Authier, Gilles, and Katharina Haude (eds.). Ergativity, Valency and Voice. Berlin/New York: Mouton...

Table of contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Gilles Authier and Katharina Haude

Ergativity and voice in Mayan languages: a functional-typological

approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Colette Grinevald and Marc Peake

Ergativity and the passive in three Mayan languages. . . . . . . . . . . . 51

Valentina Vapnarsky, Cedric Becquey, and Aurore Monod

Becquelin

The passives in Cavinena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Antoine Guillaume

The detransitive voice in Kryz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

Gilles Authier

Laz middle voice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

Rene Lacroix

Argument demotion in Japhug Rgyalrong. . . . . . . . . . . . . . . . . . . . 199

Guillaume Jacques

The Katukina-Kanamari antipassive . . . . . . . . . . . . . . . . . . . . . . . . 227

Francesc Queixalos

Undergoer orientation in Movima . . . . . . . . . . . . . . . . . . . . . . . . . 259

Katharina Haude

Case patterns and verb classes in Trumai . . . . . . . . . . . . . . . . . . . . 287

Aurore Monod Becquelin and Cedric Becquey

Ergativity in the Adyghe system of valency-changing derivations 321

Alexander Letuchiy

The evolution of transitive verbs in Basque and emergence of

dative-marked patients 353

Celine Mounole

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 13:01) WDG (155mm�230mm) TimesNRMT 1351 Authier pp. 5–6 1351 Authier_00a_Contents (p. v)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 13:02) WDG (155mm�230mm) TimesNRMT 1351 Authier pp. 5–6 1351 Authier_00a_Contents (p. vi)

Introduction

Gilles Authier and Katharina Haude

1. Preliminaries1

This volume explores voice phenomena in a substantial sample of lan-

guages with predominantly ergative features, some of them already well

studied, others less so. All the findings presented here are based on first-

hand data and personal fieldwork in speech communities. The articles

take up and elaborate on oral presentations given at the monthly seminar

Ergativite: typologie, diachronie et cognition, which took place between 2005

and 2009 in Villejuif (Paris) under the direction of Francesc Queixalos and

with the financial support of the Federation de Typologie et Universaux

Linguistiques of the CNRS, whose major purpose is to bring together lin-

guists with first-hand experience of languages with ergative features from

di¤erent parts of the world.

2. Ergativity

Ergative alignment – defined as the formal singling out of the agent (A) of

transitive verbs as opposed to the patient (P) of transitive verbs and the

single argument of intransitive verbs (S) – can be found at many levels of

grammar, both in the marking of grammatical relations and at the syntactic

level (within a single clause or in interclausal relations). The former is

usually referred to as ‘‘morphological ergativity’’, while the latter, along

with ergative alignment in terms of constituency and reflexive control, is

called ‘‘syntactic ergativity’’ (Dixon 1994).

At the clause level, ergative alignment can be recognized in the morpho-

logical encoding of cross-reference pronouns on the predicate or in the use

1. The editors’ names are in alphabetical order. Haude acknowledges the supportof the Deutsche Forschungsgemeinschaft (HA 5910/1-1) during the course ofthis project. We thank F. Queixalos and Anna Siewierska for helpful com-ments on this introduction. Needless to say, all mistakes and shortcomingsare entirely our own responsibility.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:24) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 1–14 1351 Authier_00b_Intro (p. 1)

of cases to mark the function of its dependents (‘‘core arguments’’ in Role

and Reference Grammar, Foley and Van Valin, 1984; ‘‘actants’’ in the

terminology of Lazard 1994), or in both. Alignment could in theory also

be ergative at a syntactic level within simple clauses, but if word order or

control of reflexive pronouns may not show strictly accusative behaviour

in some languages, these syntactic parameters are, to our knowledge,

never clearly ergative either.

In terms of inter-clausal syntax, ergative behaviour manifests itself in

terms of so-called ‘‘pivots’’ (Dixon 1994). Very few languages have been

claimed to have an ergative pivot, that is, a constraint in equi-deletion

that groups an omitted S with P and not A. In such languages, like Dyirbal

or its close relative Yidiny, the ergative (S ¼ P) pivot seems to coexist in

basic clause-types with the most widely attested type of pivot, S ¼ A, also

called ‘‘subject’’ by Dixon.

In some families of ‘‘ergative’’ languages, such as East Caucasian, there

is no clear category of pivot, and argument recovery between clauses is

subject to pragmatic tendencies rather than syntactic rules; but this is also

true of some accusative or active-stative languages. However, the vast

majority of languages with accusative morphology display clear accusative

pivot properties.

It should not be seen as overly cautious to refer to ‘‘ergative languages’’

in quotation marks. The expression ‘‘ergative language’’ continues to be

used here for the sake of brevity, as in many reference works on the topic,

but it should be recognized that languages are not expected to be typolog-

ically consistent across all aspects of their alignment behaviour. Some

notoriously ‘‘accusative’’ languages such as Latin, French and in fact

many Indo-European languages may have some hints of ergativity, espe-

cially in perfective nominalizations (past participles; see Lehmann 1985),

whereas many languages with ergative alignment on nouns, including lan-

guages like Dyirbal which exhibit strongly ergative interclausal syntactic

rules (Dixon 1972; Dixon 1994: 14–15, 160), show accusative alignment

on personal pronouns referring to speech act participants (see Silverstein

1976). Indeed, most if not all ergative languages also display some accusa-

tive alignment according to a set of parameters which all contribute to a

global, multifaceted, and scalar definition of transitivity (Hopper and

Thompson 1980). According to these parameters, a variety of syntactic

constructions reflect di¤erent perceptions of events involving two partici-

pants, and may deviate from the prototypical, maximally transitive con-

struction in which the Agent is also the main topic and in control of the

event (an action) while the Patient is wholly a¤ected by this action. Devia-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


2 Gilles Authier and Katharina Haude

tions from this scenario account for a number of variations in transitivity

also called actancy splits (Lazard 1997) or simply ‘‘splits’’ (Dixon 1994).

The very common ‘‘pronoun split’’ mentioned above, in which pronouns

align accusatively and nouns ergatively, was first described by Silverstein

(1976), who argued that accusative marking is more expected at the upper

end of a scale of animacy, individuation, and natural topicality or em-

pathy, because constituents referring to higher animates, especially personal

pronouns, are for semantic reasons less expected in direct object position.

This does not mean that ergative marking is unknown on personal pro-

nouns, but it is rare, and often an accusative or specific non-ergative marker

is found on personal pronouns in otherwise morphologically fully ergative

languages, as in Kryz and Cavinena (see Authier and Guillaume, this

volume). A very common intermediate situation in ergative languages is

for personal pronouns to have neutral alignment, because ergative case

on nouns and word order make case marking dispensable on pronouns

referring to the most naturally agentive participants.

Conversely, ergative alignment is more likely to be found at the lower

end of the Silverstein hierarchy, because inanimates need to be more heavily

marked when used in the less expected syntactic position of agents of tran-

sitive verbs, and indeed, some overall ‘‘accusative languages’’ such as Hittite

have ergative case marking for inanimates only (Laroche 1962).

In accusative languages with little case marking morphology, e.g. many

modern Indo-European languages (English, Spanish or French), only per-

sonal pronouns retain the old case marking system making use of supple-

tive stems (I / me). And symmetrically, even in accusative languages with

well developed core case marking, accusative case is rarely marked, or at

least less heavily expressed, on nouns referring to entities placed very low

in the hierarchy, for example on members of the neuter gender in Latin.

Aspectual splits also famously interfere with alignment, and represent

another essential parameter in the definition of transitivity as a polypara-

metric scalar phenomenon. Although such splits have been described in

languages which display very di¤erent morphological systems to mark

grammatical relations, such as Kurdish, Georgian (a relative of Laz, in

this volume) or Mayan, the rule remains the same: ‘‘if a split is condi-

tioned by tense or aspect, the ergative is almost always found either in

the past tense or in perfective aspect’’ (Dixon 1994: 99). Conversely,

many so-called ergative languages tend to exhibit exceptions to their preva-

lent ergative alignment in predicates expressing progressive-imperfective

aspect or present and future tense, because these imply a lower degree

of a¤ectedness of the object. One clear instance of this tendency is the

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Introduction 3

obligatory use of antipassive voice to express progressive aspect or present

tense. In some documented instances, e.g. in Georgian, a historically derived

antipassive has evolved into the only means of expressing non-perfective,

progressive processes.

This volume gives a representative impression of the diversity of these

often intricate situations. Before presenting the voice phenomena found

in the language at issue, contributors provide succinct but precise informa-

tion about its particular pattern of ergative behaviour. Some languages,

for example the majority of Mayan languages, only have ergative cross-

referencing of arguments on the verb. But, as shown in one of the con-

tributions to this volume, Tseltal and Yucatec in fact possess a split

intransitive system: two kinds of S can be distinguished, one of which

aligns with A while the other aligns with P. Many languages represented

in this volume (Adyghe, Basque, Cavinena, Kryz, Laz) have both ergative /

absolutive case marking on nominals and an ergative system of verb

indexes, but this situation may not be the one most commonly found:

accusative indexing on verbs often coexists with ergative case marking,

while the reverse situation seems to be unattested; and a few languages

under scrutiny in this volume, such as Movima and Trumai, have two

competing constructions – neither of which can be considered ‘‘basic’’ –

for predicates with two arguments, one showing ergative case marking on

nouns while the other shows accusative alignment.

We can confirm that, although they may be quite diverse in nature,

alignment splits are the rule for languages which have ergative alignment

somewhere in their grammar. In fact, it seems that no language has been

documented which would qualify as ergative at all levels of its grammar.

Not all aspects of those phenomena which fall under the heading of

ergativity will be touched upon in this volume; but notwithstanding the

great morphosyntactic variety of the languages in this sample, all papers

focus on the relationship of ergativity with valence change and voice phe-

nomena. The paper on Basque, which deals with the diachronic evolution

of transitive valency indexation under the influence of Spanish di¤erential

object marking, addresses a somewhat di¤erent aspect of the relationship

between ergativity valence systems.

3. Voice

Taking both formal and functional criteria into account, voice alternations,

as understood in this volume, show the following basic characteristics: in

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



terms of form, they determine the number, formal encoding, and semantic

role of verbal argument(s); in terms of function, they serve to describe an

event from di¤erent perspectives, and to retain the same participant as the

central argument through larger stretches of discourse; voice alternations

ideally form a productive system.

Many definitions of voice are more restrictive than the one given above.

They usually treat the notion of ‘‘subject’’ (for the agent in a transitive

clause and the single argument of an intransitive clause) as central to voice

alternations (cf. Siewierska 1984; Mithun 1994); some (e.g. Kulikov 2010)

accept only those alternations that either increase or decrease the number

of arguments, but not those where the semantic roles of the core argu-

ments are reassigned without a¤ecting the valency of the verb. Voice is

generally restricted to the verbal domain and taken to involve explicit

morphological marking (see e.g. Klaiman 1991: 1; Creissels 2006: 6).

However, as has frequently been noted, applying these more restrictive

definitions can complicate the study of voice phenomena in predominantly

ergative languages. In particular, it is well known that the notion of sub-

ject in ergative constructions is problematic (see e.g. Blake 1976; Mithun

1994: 247; Shibatani 1998: 120), since it encompasses a relation which is

treated di¤erently in transitive and intransitive clauses. Moreover, while

in most predominantly ergative languages the ergative-marked argument

of the transitive clause can be identified as the synactic subject (see e.g.

Anderson 1976), in some languages (e.g. Katukina, Queixalos this volume)

it is the absolutive, rather than the ergative, argument of transitive clauses

whose syntactic status is comparable to that of the subject in a nominative-

accusative construction.

The hybrid character of the relation encompassing the transitive agent

argument and the single argument of intransitive clauses is especially

interesting for a consideration of the passive and antipassive, the classic

valency-decreasing voices that serve to maintain the syntactically privileged

status of the topical participant (see Dixon 1994, Van Valin and LaPolla

1997, Primus 1999). As for the passive – an operation whereby the patient

becomes the subject of a derived intransitive clause – it has sometimes

been assumed that this operation is not usually available in predominantly

ergative languages (Shibatani 1998: 120). However, as is shown by several

articles in the present volume, many ergative languages do possess passive

operations, especially when the agent argument of the transitive clause is

syntactically privileged. At the same time, the antipassive – an operation

whereby the agent becomes the single argument of a derived intransitive

clause – remains a typical operation of ergative languages, as it allows

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Introduction 5

the role which is marked in the transitive construction to appear unmarked

in the intransitive construction (analogously to the accusative/nominative

marking alternation in the passive constructions of accusative languages;

see Silverstein 1976: 140). In nominative-accusative languages, the functions

of antipassives (including aspectual functions, see Cooreman 1994) are often

fulfilled by the intransitive use of an otherwise transitive verb, which entails

no di¤erence in the marking of the agent argument (nominative).

Languages can have two or more basic transitive constructions, includ-

ing both accusative and ergative types, which alternate in ways similar to

voice. The best-known example is found in the ‘‘symmetrical voice’’ systems

of many Austronesian languages, which provide ‘‘for any transitive event

[. . .] at least two representations, one in actor voice and one in undergoer

voice’’ (Himmelmann 2005: 135), both of which are equally morphologi-

cally marked. Some Austronesian languages clearly display ergativity in

their morphosyntax (see Mithun 1997 on Kapampangan); in others, like

Tagalog, the most frequent transitive construction takes the patient as the

privileged argument (the ang phrase), which is why some scholars consider

it to be ergative in nature (see the overview in Shibatani 1988). What is of

particular interest here is that these systems correlate with certain features

which, to a greater or lesser degree, also appear in languages with more

straightforward ergative traits, like those of the Mayan or Eskimo families:

verbs share at least some syntactic characteristics with nouns, and the

ergative argument shows similarities to an adnominal possessor. A possible

diachronic scenario for the rise of ergative structures, attested in Indo-

Aryan languages (e.g. Benveniste 1966a: 176–186, 1966b: 127–36), is that

ergative constructions arose from the frequent use of passive participles as

intransitive predicates, which had the patient as subject and the agent as

an oblique-marked (genitive or dative) argument.

Another particular case in the domain of argument adjustment that

does not change the number of verbal arguments can be found in so-called

inverse systems. These systems display two transitive constructions, direct

and inverse, which are chosen depending on the position of the event par-

ticipants of the arguments on a referential hierarchy; the choice is typically

determined by the opposition between speech-act participants and third

persons (see DeLancey 1981). However, the direct-inverse opposition often

holds in the third-person domain as well, and here the relative topicality of

the arguments plays a role (see Gildea 1994); for instance, in Algonquian

languages the choice of the direct or inverse constructions in the third-

person domain depends on whether a nominal referent is marked as prox-

imate or obviative, and this marking can, in turn, be based on the relative

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



topicality of the referents (see Zuniga 2006: 71). Therefore, when the dis-

course status of nominal referents is relevant for the selection of the

construction used, an inverse system can be analysed in terms of voice

(see Givon 1994, 2001), and perhaps (when there is no di¤erence in marked-

ness between the two constructions) even interpreted as a more strongly

grammaticalized form of symmetrical voice systems (see Farrell 2005).

A discourse-based opposition between accusative and ergative transitive

constructions is also found in languages of the Amazonian Arawa family

(Aikhenvald 2009; Dixon 2000). Thus, there are numerous examples of

alternations based on discourse properties of referents, which means that

voice is not necessarily tied to a change in the number of arguments.

Voice is often viewed as a morphological category of the verb and as

such is seen as necessarily requiring verbal marking. Such an interpre-

tation of voice excludes from its domain two alternations which do not

involve verbal marking, namely lability and lexical alternations, which

never seem to be fully productive but can assume a voice function to

some degree (see Creissels 2006: 5–6). Lability involves the use of a verb

in both transitive and intransitive constructions without any morphological

modification. There are two types of lability, as defined by Drossard (1998):

orientation-maintaining (e.g. English ‘‘I eat it’’ – ‘‘I eat’’) and orientation-

changing (English ‘‘I break it’’ – ‘‘It breaks’’). The pattern of lability found

in an ergative system is the opposite of that found in an accusative system,

since in the latter the verb is oriented towards the agent, while in the former

the verb is oriented towards the patient.

A complex issue is seen in the situation where di¤erent lexical verbs

with contrasting argument structures can be chosen depending on which

participant is to be highlighted, since potential semantic contrasts between

lexical verbs are even less easily understood than those involved in morpho-

syntactic alternations. A possible example is presented by Trumai (Monod

Becquelin and Becquey, this volume), a language without voice morphol-

ogy, where voice functions seem to be realized through lexical alterna-

tions: as argued by the authors, the ergative and accusative constructions

are used to preserve the discourse topic as the unmarked argument (the

‘‘subject’’).

Finally, while voice is almost always seen as belonging to the verbal

domain, di¤erential marking of nominal constituents may have a similar

function (see Shibatani 2006: 229) in cases where di¤erential argument

marking is not exclusively determined by person or animacy but also by

topicality, as reflected e.g. by definiteness. In this volume, Mounole’s descrip-

tion of the emergence of di¤erential object marking through language con-

tact in Basque hints at this possible connection.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Introduction 7

Aside from these issues, languages with predominantly ergative morpho-

syntax (like many others) also display other valency-decreasing devices, such

as middles, anticausatives, and noun incorporation, as well as language-

specific mechanisms or subclasses of the above; these are discussed in

many of the contributions presented here. Valency-increasing devices such

as causatives, benefactives or applicatives can be counted as belonging to

the domain of voice as well, since they can be employed for discourse-

pragmatic purposes, e.g. topic maintenance. However, the wide range of

constructions and linguistic variation which is found in this domain de-

serves discussion of its own, and therefore such phenomena are not specifi-

cally dealt with in this volume.

4. Contributions to this volume

Mayan languages have been at the core of discussions on ergativity ever

since the 1970s, and it is only natural that the volume starts out with two

articles on this language family. C. Grinevald and M. Peake o¤er an intro-

duction to transitivity and voice phenomena in Mayan languages, drawing

mainly on their expertise in two languages of the Qanjobalan branch of

the family, Jakaltek Popti’ and Tojolabal. They illustrate the mechanisms

by which voice alternations in Mayan languages serve to promote and

demote clausal arguments and thereby structure the discourse-pragmatic

organization of sentences and texts.

The article by V. Vapnarsky, A. Monod Becquelin and C. Becquey

focuses on the passive voices found in Ch’ort’i, Tseltal and Yucatec, lan-

guages from di¤erent branches of the Mayan family which display various

degrees and reflexes of morphological ergativity. Analysing the conditions

of occurrence of the passive constructions in these languages, the authors

show that (possibly in all Mayan languages) the passive has semantic and

discourse-pragmatic (agent-defocusing) rather than syntactic functions. By

also comparing the languages with respect to the other voices they display,

especially the ‘‘focus antipassive’’ typical of many Mayan languages, they

arrive at the conclusion that the array and type of voice constructions a

language displays does not, by itself, provide clues about its basic syntactic

alignment.

The subsequent four papers deal with passive and passive-like construc-

tions in unrelated languages from di¤erent geographical areas.

A. Guillaume describes two apparently related verbal morphemes (-ta

and -tana) in Cavinena (Tacanan, lowland Bolivia), both somewhat re-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



stricted in use, which yield passive and anticausative interpretations respec-

tively. In his article, the author revises earlier hypotheses on the origin of

these morphemes and their semantic and diachronic relationship.

G. Authier’s account of the detransitive voice in Kryz first outlines the

morphosyntax of transitive verbs as a subset of bi-actant constructions

(a typical East Caucasian case of semantic encoding of various types of

subjects). The semantic spectrum of the Kryz detransitive voice, which

synchronically has much in common with those of Cavinena and Laz,

is described in detail and set alongside comparable constructions and

morphemes in related languages. East Caucasian languages rarely have

valency-decreasing voices, and indeed, Kryz is the only member of this

family known to have a passive voice for most transitive verbs; this form

also has anticausative value with most verbs, and allows an antipassive

interpretation with a just a few of them. The passive interpretation is

shown to be a recent innovation linked to contact with sociolinguistically

dominant Azeri, a genetically unrelated, uniformly accusative language.

R. Lacroix describes the di¤erent functions of the voice marker i- in the

South Caucasian (Kartvelian) language Laz, which he analyses as a marker

of middle voice, that is, a category encompassing a variety of ways in

which the transitivity of the prototypical two-participant construction can

be modified. While this marker is widespread in related languages, Lacroix

o¤ers a novel approach and a fine-grained description of this element of the

Kartvelian set of valency-changing morphemes.

Japhug Rgyalrong (Qiangic, Tibeto-Burman) has several detransitivizing

voices, some of which are more frequently used than others. G. Jacques’

article provides an overview of the rich morphology associated with the

demotion of indefinite (unknown or generic) arguments in this language.

Patients can be suppressed by means of generic, antipassive, and incorpo-

rating constructions and labile verbs, while agents can be suppressed in the

generic and passive constructions.

The three subsequent articles deal with languages of the Amazon area,

all three of which display ergative syntactic patterns. F. Queixalos demon-

strates that the absolutive argument in Katukina has a privileged syntactic

status and shows that the agent argument of transitive clauses can acquire

this privileged status only through an antipassive operation. In this lan-

guage, therefore, the antipassive clearly has a syntactic function.

K. Haude discusses the correlation between ergative patterns and patient

orientation in Movima. Movima has two basic transitive constructions,

direct and inverse, with direct main clauses displaying ergative, and inverse

main clauses displaying accusative syntax. The direct and inverse markers

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Introduction 9

belong to a paradigm of verbal voice morphemes that determine the argu-

ment structure of the verb. It is argued that the verbs participating in this

scheme are basically oriented towards the non-agent, and that this may be

one of the reasons why the direct/ergative construction is less pragmatically

marked than the inverse/accusative construction.

C. Becquey and A. Monod Becquelin’s article discusses textual data

from Trumai, a linguistic isolate spoken in the Amazonian region of Brazil.

Trumai lacks derivational voice alternations, but there are two transitive

constructions – one ergative, the other accusative, depending on the lexical

verb involved – whose interplay is employed for topic-maintaining pur-

poses. While the authors make it clear that the domain of verb semantics

in Trumai requires further research, the findings suggest that we may be

dealing with a system in which voice functions are taken over by lexical

alternations.

The last two papers bring the discussion back to Europe. A. Letuchiy’s

article on Adyghe (West Caucasian, not genetically related to Laz and

South Caucasian or to Kryz and East Caucasian) explores ergative align-

ment from the point of view of voice phenomena, claiming that the behav-

iour of arguments in voice derivation (Adyghe derived voices have low

textual frequency and are not very productive) is a more decisive criterion

for the identification of syntactic ergativity than the pivot phenomena pre-

viously studied by Dixon, which generally appear not to be characteristic

of ergative languages of the Caucasus.

C. Mounole’s paper deals with the emergence of di¤erential direct object

marking in Basque, showing that it has been progressively induced by con-

tact with Spanish during the modern period. Di¤erential object marking

is rare in ergative languages, but elsewhere it is attested at least in West

Iranian ergative languages (cf. Bossong 1985) and in Udi (East Caucasian,

cf. Schulze 2008), giving rise to tripartite alignment systems. Although this

paper does not deal with voice specifically, it may be viewed as an incen-

tive to pursue the exploration of potential links between various parameters

of transitivity and the diversification of valence frames in predominantly

ergative languages.

5. Summing up: What this book contributes to typology

In sum, this volume shows once again that so-called ‘‘ergative languages’’

do not represent a homogeneous group. A language whose basic main-

clause morphosyntax is predominantly ergative does not necessarily show

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



the same voice phenomena as another language with ergative characteristics,

probably because manifestations of ergativity are far from homogeneous,

ranging from the overt case marking of an agent without any syntactic

e¤ects to the sharing of some syntactic interclausal properties of the patient

of transitive clauses with the single argument of intransitive clauses.

Many articles in this volume show that passive voices (often with asso-

ciated anticausative value) are very common in ergative languages and signal

the demotion or even the total suppression of agents, but such semantic

factors as the expression of perfective aspect or deontic modality seem to

be responsible for their extension and diversification. The antipassive

voice, which demotes or suppresses the patient, is often taken to be an

expected phenomenon in languages with ergative systems, and a substan-

tial number of the languages represented in the volume do display anti-

passive constructions; however, even in ergative languages its use seems

to be comparatively rare. Finally, some articles show that ergative struc-

tures themselves may belong to a voice system in which they alternate

with accusative constructions. In conclusion, voice phenomena, as a sub-

category of information structure modifying devices, are certainly related

to patterns of interclausal syntactic alignment, but probably not to the

coding of grammatical relations within the clause.

References

Aikhenvald, Alexandra Y.2009 ‘‘Syntactic ergativity in Paumarı.’’ In Topics in Descriptive and

African Linguistics. Essays in Honor of Distinguished ProfessorPaul Newman, Samuel Gyasi Obeng (ed.), 111–127. Munchen:Lincom Europa.

Anderson, Stephen S.1976 ‘‘On the notion of subject in ergative languages.’’ In Subject and

topic, Charles Li. (ed.), 1–24. New York: Academic Press.Andrews, Avery D.

2007 The major functions of the noun phrase. In Language Typologyand Syntactic Description. Vol. 1: Clause Structure. TimothyShopen (ed.), 132–223. Cambridge: Cambridge University Press.Second Edition.

Benveniste, Emile1966a Problemes de Linguistique Generale. Vol. 1. Paris: Gallimard.

Benveniste, Emile1966b Problemes de Linguistique Generale. Vol. 2. Paris: Gallimard.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Introduction 11

Blake, Barry1976 ‘‘On ergativity and the notion of subject.’’ Lingua 39(4): 281–300.

Bossong, Georg1985 Di¤erentielle Objektmarkierung in den neuiranischen Sprachen.

Tubingen: Narr.Cooreman, Anne

1994 ‘‘A functional typology of antipassives.’’ In Fox, Barbara A. andPaul J. Hopper (eds.), 49–87.

Creissels, Denis2006 Syntaxe generale. Une introduction typologique. Vol. 2: La phrase.

Paris: Hermes-Lavoisier.Dixon, R.M.W.

1972 The Dyirbal Language of North Queensland. Cambridge: Cam-bridge University Press.

Dixon, R.M.W.1994 Ergativity. Cambridge: Cambridge University Press.

Dixon, R.M.W.2000 ‘‘A-constructions and O-constructions in Jarawara.’’ International

Journal of American Linguistics 66: 22–56.Drossard, Werner

1998 ‘‘Labile Konstruktionen.’’ In Typology of Verbal Categories. Paperspresented to Vladimir Nedjalkov on the occasion of his 70th birth-day, Leonid Kulikov and Heinz Vater (eds.), 73–84. Tubingen:Max Niemeyer.

Farrell, Patrick2005 Grammatical Relations. Oxford: Oxford University Press.

Foley, W. and Van Valin, R.1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge

University Press.Fox, Barbara A. and Paul J. Hopper (eds.)

1994 Voice: Form and Function. Amsterdam/Philadelphia: Benjamins.Gildea, Spike

1994 ‘‘Semantic and pragmatic inverse: ‘Inverse alignment’ and ‘inversevoice’ in Carib of Surinam.’’ In Givon, T. (ed.), 187–230.

Givon, T.1994 ‘‘The pragmatics of de-transitive voice: functional and typologi-

cal aspects of inversion.’’ In Givon, T. (ed.), 3–44.Givon, T.

2001 Syntax: an Introduction. Vol. 2. Amsterdam/Philadelphia: Ben-jamins.

Givon, T. (ed.)1994 Voice and Inversion. Amsterdam/Philadelphia: John Benjamins.

Himmelmann, Nikolaus P.2005 ‘‘The Austronesian Languages of Asia and Madagascar: Typo-

logical Characteristics.’’ In The Austronesian Languages of Asia

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



and Madagascar, Alexander Adelaar and Nikolaus P. Himmel-mann (eds.), 110–181. London/New York: Routledge.

Himmelmann, Nikolaus P.2008 ‘‘Lexical categories and voice in Tagalog.’’ In Voice and Gram-

matical Functions in Austronesian Languages, Peter Austin andSimon Musgrave (eds.), 247–293. Stanford: CSLI.

Keenan, Edward L. and Matthew S. Dryer2007 ‘‘Passive in the World’s Languages.’’ In Language Typology and

Syntactic Description. Vol. 1: Clause Structure. Timothy Shopen(ed.), 325–361. Cambridge: Cambridge University Press. SecondEdition.

Klaiman, M.H.1991 Grammatical Voice. Cambridge: Cambridge University Press.

Kulikov, Leonid I.2010 ‘‘Voice typology.’’ In The Oxford Handbook of Linguistic Typology,

Jae Jung Song (ed.), 368–398. Oxford: Oxford University Press.Kulikov, Leonid and Heinz Vater (eds.).

1998 Typology of Verbal Categories. Papers presented to Vladimir Ned-jalkov on the occasion of his 70th birthday. Tubingen: Niemeyer.

Lehmann, Christian1985 ‘‘Ergative and active traits in Latin.’’ In Relational Typology,

Frans Plank (ed.), 243–255. Berlin: Mouton.Laroche, Emmanuel

1962 ‘‘Un ‘‘ergatif ’’ en indo-europeen d’Asie Mineure’’. Proceedingsof the Berkeley Linguistics Society 57: 23–43.

Lazard, Gilbert1994 L’actance. Paris : Presses Universitaires de France.

Lazard, Gilbert1997 ‘‘Ergativity (review of R.M.W. Dixon, Ergativity).’’ Linguistic

Typology 1: 243–268.Primus, Beatrice

1999 Case and Thematic Roles. Ergative, Accusative and Active. Tubin-gen: Niemeyer.

Schulze, Wolfgang2009 ‘‘Grammar.’’ In The Caucasian Albanian Palimpsests of Mount

Sinai, Jost Gippert, Wolfgang Schulze, Zaza Aleksidze, Jean-Pierre Mahe (eds.), vol. I., 21–60. Turnhout: Brepols.

Shibatani, Masayoshi1988 ‘‘Voice in Philippine Languages.’’ In Passive and Voice, Masayoshi

Shibatani (ed.), 85–142. Amsterdam/Philadelphia: Benjamins.Shibatani, Masayoshi

1998 ‘‘Voice Parameters.’’ In Kulikov, Leonid and Heinz Vater (eds.),117–138.

Shibatani, Masayoshi2006 ‘‘On the conceptual framework for voice phenomena.’’ Linguis-

tics 44(2): 217–269.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Introduction 13

Van Valin, Robert D. and Randy J. LaPolla1997 Syntax: Structure, Meaning, and Function. Cambridge: Cambridge

University Press.Zuniga, Fernando

2006 Deixis and alignment. Inverse systems in indigenous languages ofthe Americas. Amsterdam/Philadelphia: Benjamins.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Ergativity and voice in Mayan languages:a functional-typological approach

Colette Grinevald and Marc Peake

1. Introduction1

The Mayan family is a large language family that resembles in many ways

the European side of the Indo-European language family, by its time depth,

its number of branches and sub-branches, and its distinct languages, around

30 languages2. Mayan languages are spoken in a relatively contiguous

area in Guatemala and Mexico (in the southern regions of Yucatan and

Chiapas), with the exception of the geographically isolated Wastekan

branch spoken further north, in central Mexico near the Atlantic coast.

A tradition of ‘‘Mayan Linguistics’’ developed in the last quarter of the

20th century. Historical and comparative studies led first to a fairly extensive

reconstruction of Proto-Maya, and the identification of major pan-Mayan

morpho-syntactic characteristics, such as its ergativity markers (see for

example Kaufman (1974), Robertson (1977), Norman and Campbell 1978,

Campbell and Kaufman (1985)). A network of field linguists also coordi-

nated themselves to produce (morpho-)syntactic descriptions of topics of

particular interest in the contemporary languages. The Mayan family was

therefore one of the earlier and better known of the American continent,

and contributed interestingly, on several occasions, to ongoing discussions

of theoretical and typological interest to general linguistics, as will be

mentioned later.3

1. Colette Grinevald was known earlier as Colette Craig. This chapter is partlybased on materials from her courses on Mayan Grammar in the Mayan Pro-gram of the Department of Amerindian Linguistics of the ‘‘Institut Nationaldes Langues et Cultures Orientales’’ (INALCO) in Paris, France. It is meantas an introduction and orientation to Mayan studies of ergativity and voices.Our thanks to various readers for helpful comments on drafts of this paper: ananonymous reviewer, Judith Aissen and the editors of the volume.

2. See Kaufman (1974) for a view of the whole family, reproduced in England(1996).

3. The network organized topical workshops and produced a Journal of MayanLinguistics. It also contributed to the training of native Mayan speakers,

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:24) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 15–50 1351 Authier_01_grinevald (p. 15)

And ergativity and voices have certainly been some of the features of

the Mayan family of languages that have attracted most attention for their

being key to the understanding of how Mayan grammars function, and of

great interest to general linguistics. This chapter is therefore meant to give

an overview of both the facts of Mayan ergativity and voices, but also a

sense of how their study has developed and participated to certain develop-

ments of descriptive and theoretical linguistics in the past decades.

This study of Mayan ergativity and voices begins, in Section 2, with an

overview of Mayan verbal morphology, to bring to attention the striking

multiplicity of transitivity markers typical of most languages of the family.

Section 3 then o¤ers a description of the forms and functions of the verbal

person markers that justify talking of ergative alignment and of occasional

patterns of split ergativity, of these markers known as the ‘‘set A’’ and ‘‘set

B’’ in Mayan linguistics terminology. The discussion turns, in Section 4, to

the role of these set A and set B markers in the identification of a fairly

elaborate system of voices, with clear morphosyntactic markings of various

types of antipassive, passive and applicative voices and of a typologically

rarer and morphologically more ambiguous ‘‘agent focus’’ voice.

This presentation of Mayan ergativity and voices has been cast in a

‘‘functional-typological’’ approach to grammar, of the kind originally pro-

posed by Givon (1979, 1986, 2001). This means here that the description

of the ergative patternings considered will systematically appeal to the

notion of patterns of ‘‘alignments’’ between semantic and syntactic roles

(as defined in the literature on ergativity, as will be noted then). The choice

of this framework also accounts for the quick survey of studies about the

discourse use and pragmatic motivations for such a variety of voices in

Mayan languages that is presented in section 5.4

2. Mayan verbal morphosyntax

Like many languages of the American continent, Mayan languages are

head-marking, so that their verb forms carry all the information necessary

leading today to a new generation of native linguists (these noteworthy develop-ments are traced in England 1992, 1996, 2007, England and Woodbury 2004,Grinevald 2002, 2007).

4. This is not to say that all the discussions that will be cited have been expresslyformulated within such a functional-typological framework, rather that it isthe mode of organization of this presentation that is.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


16 Colette Grinevald and Marc Peake

to identify the core argument structure of the clause, in the widespread

absence of anaphoric pronouns. The ergativity phenomenon is located in

these verbal structures: it is found in the special patterning of di¤erent sets

of person markers a‰xed to the verb. Other markers contribute also to

establishing the transitivity level of the verbal construction. The argumen-

tation in Mayan languages for particular syntactic descriptions is also

greatly facilitated by a strictly agglutinating morphology, with relatively

little in the way of morphophonemics.

2.1. Pan-Mayan vs language specific verbal morphology

This sub-section surveys the di¤erent elements of the verbal complex, con-

sidering in which way they are characteristic of the family as a whole and

in which way they may vary across branches or languages of the family.

– All Mayan languages have a rich system of Mayan roots, predominantly

of the form CVC,5 and verbal constructions built on either ‘‘radical’’ or

derived verbs. This di¤erence is particularly relevant to our present

purpose to the extent that, in some languages, the status of radical or

derived verbs may determine a choice between variant forms of voice

or aspect markers.

– In the TAM (tense-aspect-modality) domain, practically all languages

make a basic distinction between ‘‘completive’’ and ‘‘incompletive’’

aspect, while some have further tense and aspect distinctions (such as

recent vs. distant past or future, and progressive). Aspect markers are

usually prefixed but sometimes su‰xed; some aspects can also be

expressed using auxiliary-type forms in periphrastic constructions, par-

ticularly the progressive. The main morphological mood markers are

for imperative and subjunctive (sometimes found under the label of

‘‘irrealis’’ or ‘‘future’’). It is worth noting that, in some languages, the

choice of variant forms of these TAM markers may be determined also

by the transitivity level of the verbal construction.

– In the domain of person marking, the indexation of the core arguments

in the verb form constitutes a clear pan-Mayan feature. The person

markers are organized into two paradigms, known traditionally as

‘‘set A’’ and ‘‘set B’’6. This complex person marking system is one of

5. Of which there are hundreds in each language participating in a large numberof highly productive morphological processes (See Lois and Vapnarsky 2003a, bfor the especially productive case of Yucatec Maya).

6. In some languages, like Tsotsil, the plurality of the person is marked by specificadditional su‰xes (see 3.3. below).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Ergativity and voice in Mayan languages 17

the best-known characteristics of the Mayan languages, and will be

amply discussed below in Section 3.

– The rich voice system is a family trait of Mayan languages too, and is

by and large marked by su‰xation. The variety of Mayan voices and

of their subtypes will be taken up in Section 4 below.

– Specific transitivity markers that participate in the redundant marking

of transitivity of verbal forms are found in some Mayan languages.

They consist of a su‰x (often known as a ‘‘theme vowel’’ or ‘‘status

vowel’’7) that specifies the syntactic transitivity of the verbal construc-

tion, marking it as either transitive or intransitive. This su‰x may

directly follow the verb or the voice marker.8

The potential complexity of the Mayan verb form will be illustrated below

with verb templates and data for two di¤erent Mayan languages, to show

their commonalities (such as distributed transitivity marking) and their

specificities (such as the presence or absence of certain verbal categories).

2.2. Verb template and data from Jakaltek Popti’9

Jakaltek Popti’ is a language of the Q’anjob’alan branch of the family,

spoken by approximately 88,000 speakers in the area around Jacaltenango

in the Cuchumatanes mountains in western Guatemala and some recent

settlements in Chiapas, Mexico.

The verbal template of Jakaltek Popti’ shown in Table 1 below is fairly

representative of the verbal morphology of Mayan languages. Its specificity

is mostly in its typologically rare and complex system of directionals,

7. Although the expression is somewhat a misnomer to the extent that it caninvolve a semi-vowel, and thus ‘thematic a‰x’ or ‘thematic extension’ wouldbe a more appropriate label.

8. It may also be su‰xed to totally grammaticalized directionals. Many Mayanlanguages have directionals, functioning like systems of verbal satellites, fromfree forms like English verb particles (in Tsotsil for instance) to a‰xed sets(su‰xed in Jakaltek Popti’, prefixed in Mam).

9. Craig’s (1977) work on the language, then called and spelled Jacaltec, was thefirst extensive syntactic study of a Mayan language, but at a time when theterminology of ergativity and voices as presented here did not exist yet. Italso used an orthography superseded now by the o‰cial orthography used inthis text.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



which is actually insensitive to the transitivity of the construction (Craig

1993). The common thematic vowel is verb final, but it disappears if the

verb itself is not clause final.

Examples (1a), (1a0) and (1b) illustrate some of the typical pan Mayan

verbal categories listed above, including the use of the final theme vowel

to signal transitivity status. (1c), on the other hand, shows one of the

language specific traits of the verbal morphology in this language, the

category of directionals.

(1) [Jakaltek Popti’, Craig (1977, 1993)]

a. xk-ach to-yi a0 . xk-ach w-il-a

cmp-b2 go-vi cmp-b2 a1-see-vt

‘you went’ ‘I saw you’

b. xk-ach il-lax-i

cmp-b2 see-pass-vi

‘you were seen’

c. x-Ø-s-muj-kan-ay-toj11 heb’ naj naj

asp-b3-a3-bury-dir1-dir2-dir3.suff pl cl/man cl/man

‘they buried him 1.once and for all-2.down-3.away’

(1c) also serves to illustrate three specificities of the language: a rigid VSO

word order; the existence of a rare type of classifiers, here noun classifiers

in their function of anaphoric pronouns; di¤erent sets of su‰xed direc-

tionals, here in a rare example of their possible maximal extension of three.

10. The set A person marker is sensitive to the nature of the initial segment of theverb to which it is prefixed. There is one form for consonant initial verbs (-C)and another for vowel initial verbs (-V).

11. The directional -toj of (1c), always verb final, is actually a fusion of -to-oj,a directional resulting from the grammaticalization of the motion verb to‘to go’, and a subjunctive mood marker -oj.

Table 1. Jalkatek Popti’ Maximal Verb Extensions

T/A- -set B set A-10 VERB -voice -mood -dir -mood -theme

_C_V

radicalderived

pass.antipass.

irrealis set 1-2-3vs. pre-verbaux

irrealis trans.intrans.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



2.4. Verb template and data from Tojol Ab’al12

Tojol Ab’al is usually classified as a member of the Chujean sub-branch of

the Q’anjob’alan branch of the family, most closely related to Chuj.13 It

is spoken by approximately 35,000 in the southern part of the state of

Chiapas in Mexico.

The indications of transitivity are more numerous yet in Tojol Ab’al

than in Jakaltek Popti’ and are more widely distributed across the verbal

construction.

As shown in the examples of (2) a thematic su‰x is always present and set

B is always su‰xed. The various examples show a su‰xed plural marking

specifically for set A arguments (otherwise marked with number neutral

prefixes) in contrast to set B su‰xes sensitive to number (2b and 2c). The

contrast (2b and 2d) confirms the role of the thematic su‰x in marking the

verb form as either transitive or intransitive:

(2) Tojol Ab’al (Peake (2007) and field notes)

a. wa la-waj-y-on

icp icp12-go-vi-b1s

‘I go/I am going’

b. wa x-aw-il-aw-on-ex

icp icp-a2-see-vt-b1s-pl

‘You (pl) see me’

12. Tojol Ab’al in two separate words, rather than the more usual Tojol’ab’al isthe form preferred today by many native speakers.

13. It is one of the rare languages of the family for which the exact classificationis still pending. The Chujean classification is the position of Kaufman andCampbell (1985). Robertson (1977) however has claimed that it is part of theTseltalan branch and is more closely related to Tseltal and Tsotsil.

Table 2. Tojol Ab’al Maximal Verb Extensions

T/A- set A- VERB -voice -mood -theme -set B -set A

(sg)_C_V

radicalderived

passiveantipassive

irrealis transitiveintransitive

(sg/pl) plural

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



c. Ø-jak-tikon och14 b’a chonab’

cmp-come-b1pl.ex dir:hither loc town

‘We came here, towards town’

d. jel ixta wa x-il-j-y-e’

int toy icp icp-see-pass-vi-b3pl

‘They were badly mistreated’ (lit. ‘They were seen as toys’)

2.5. Conclusions: distributedness of transitivity marking

One of the main points of this section was to demonstrate how transitivity

is heavily marked in the morphology of the verbal complex of Mayan

languages. As expected, the choice of person markers and the presence of

voice markers are the essential elements for the determination of the level

of transitivity of the construction. However, in this family of languages,

transitivity may be signaled by the choice of particular tense/aspect/

mood and voice markers sensitive to the nature of the verb itself (as either

transitive or intransitive, whether by root or derivation). And as if to top it

o¤, the final thematic su‰x, when it appears, adds to this rather typically

Mayan insistence on indicating the level of transitivity of the whole verb

complex.

Inversely, the clear and detailed morphosyntactic markings of the various

voices, through a number of verbal a‰xations and the choices of set A and

set B person markers, has certainly been one of the main reasons why data

from Mayan languages have found their way to a host of typological dis-

cussions dealing with ergativity and voices in recent decades.

3. Terminological and typological approach to Mayan ergativity

This section will situate the discussion of the specificity of the ergative

marking of Mayan languages in the context of two linguistic traditions.

One is the now widespread approach to typologically oriented studies of

ergativity that appeals to the distinction of three basic grammatical rela-

tions (rather than the traditional two of subject/object). The specific labels

to be used here for those three ‘‘primitives’’ are A/P/S15, and correspond

14. Tojol Ab’al has directionals too, but of the independent and uninflected typewhen compared to the Jakaltek Popti’ ones of example (1c).

15. A/P/S is found in Comrie (1978) as opposed to A/S/O of Dixon (1987, 1994).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



to subject of transitive/object of transitive/subject of intransitive, respec-

tively. The second tradition referred to here is that of the terminology

used in Mayan linguistics literature, where the person markers involved

in the marking of ergativity have been labeled ‘‘set A’’ and ‘‘set B’’.

3.1. ‘‘Set A’’ and ‘‘Set B’’, a Mayan linguistics tradition

Before modern typological studies had explored the phenomenon of erga-

tivity, Mayan linguistics had already assigned (neutral) labels of ‘‘Set A’’

and ‘‘Set B’’ to the two sets of person markers found in all Mayan lan-

guages, markers that were finally identified as functioning in a so-called

‘‘ergative alignment’’ only later, in the seventies. As recognized then, one

of the functions of the Set A markers corresponds to that of the so-called

ergative case, while the main function of the Set B corresponds to that of

an absolutive case. But the names have stuck, and most linguistic studies

of Mayan languages continue to refer to them in this way. Hence in the

Mayan examples of this text, the ergative marker is glossed ‘‘A’’ and the

absolutive marker ‘‘B’’.

The Mayan person markers are worth a few general remarks at this

point. First, they are omnipresent throughout Mayan languages because

of their multiple functions in verbal, nominal and adpositional phrases,

as will be shown below. Second, they are a solid pan-Mayan characteristic

in contemporary languages, although with detailed language specific varia-

tion, also to be considered below. Third, they have been reconstructed for

Proto-Maya, and the linguistic changes leading to the individual systems of

modern languages are known (See in particular Norman and Campbell

1978, and Robertson 1980). The following paragraphs will review the di¤er-

ent forms and functions of Set A and set B.

The Set A person markers are prefixes that attach to the verbal com-

plex in all Mayan languages, except Huastec (Zavala 1994), with some

languages also having additional plural Set A su‰xes used in conjunction

with these prefixes. In addition, again with the exception of Huastec, there

are two phonologically determined Set A paradigms: a pre-consonantal

and a pre-vocalic one. The Set A markers are further used in several types

of phrases. In transitive verbs, they cross-reference the subject (the ‘‘A’’

primitive) and, in those languages that display split-ergativity, they cross-

reference the subjects of intransitive clauses (the ‘‘S’’ primitive) following a

nominative alignment pattern (see Section 4 below, and Vapnarsky et al.

this volume). Set A markers are also found both in noun phrases as markers

of the possessor in possessive constructions, in which case the marker is

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



prefixed to the possessed noun, and again in prepositional phrases, when

prefixed to the so-called ‘‘relational nouns’’ that function as (inflected)

prepositions and far outnumber the prepositions of those languages.

The Set B person markers, on the other hand, exhibit more variety of

a‰xation. They may be prefixed and/or su‰xed, the variation, where it

exists, being determined by various factors, such as the presence or absence

of an aspect marker. There may be multiple forms of each Set B a‰x as

well, depending on where it occurs and on its function, although, if a set

B a‰x is prefixed, it does not exhibit distinct pre-consonantal and pre-

vocalic forms as Set A prefixes do. Set B markers may in fact be clitics or

free morphemes that occur (semi)independently of the verb form. In terms

of their functions, the Set B person markers are equivalent to the absolu-

tive case and cross-reference the subjects of either intransitive, verbal or

non-verbal predicates. In a typical ergative type of alignment, the same

set B is therefore associated to the functions of the ‘‘S’’ primitive (subject

of all intransitive predicates) and the ‘‘P’’ primitive (objects of transitive

predicates).

In addition, the plural marking found in some languages for set A and

set B can be of additional interest in the discussion of ergativity, particu-

larly when there are distinct plural forms for the two sets, with the result

that this di¤erence can be exploited in syntactic argumentation (see below

the case of the applicative voice in Tsotsil, section 4.4.).

To illustrate the pan-Mayan and language particulars of the basic func-

tioning of Set A and Set B in independent clauses, data from three partic-

ular languages will be given below.16

3.2. Jakaltek Popti’ – a language with both Set B prefixes and su‰xes

There are two Set B forms in the first and second person (the form being a

Ø morpheme for the third person across the family) depending on the

nature of the predicate. In verbal predicates preceded by aspect markers,

set B markers for either subjects of intransitive verbs (‘‘S’’ primitive) or

objects of transitive verbs (‘‘P’’ primitive) take the form of an enclitic

cliticized to the aspectual marker, with which it forms an independent

phonological word (as shown in 3a). However, when the P argument it

16. See England (1983) for an early study of ergativity marking in the Mameanlanguages.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



references is in the third person (i.e. is Ø) the aspectual marker is directly

prefixed to the verb stem (as shown in 3b). In non-verbal predicates (not

aspectually marked in Mayan languages), set B is a free form immediately

following the predicate (as shown in 3c):

(3) [Jakaltek Popti’, Craig (1977)]

a. ch-onh way-i

icp-b1pl sleep-vi

‘We sleep’

b. x-Ø-kam no’ cheh

cmp-b3-die cl horse

‘The horse died’

c. meb’a honh

poor b1pl

‘We are poor’

Set A markers, on the other hand, are always prefixes (in all Mayan lan-

guages). They are a‰xed directly to the verb stem when cross-referencing

the subject of transitive constructions (the A primitive, as in 3d and 3e):

d. ch-in haw-il-a

i cp-b1 a2-see-vt

‘You see me’

e. x-Ø-s-watx’e naj te’ nhah

cmp-b3-a3-make cl/he cl house

‘He made the house’

As markers of nominal possession (as shown in 3f ), or of relational nouns

(as in 3g–h) they are also directly a‰xed to those elements:

f. haw-atut

a2-house

‘Your house/home’

g. s-wi’ te’ te’

a3-top cl tree

‘at the top of the tree’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



h. ch-onh tzuj-cha haw-u17

icp-b1pl follow-pass a2-by

‘We are followed by you’ (in the passive adversative sense of

‘you are pursuing us’)

3.3. Tsotsil18 – a language with distinctive plural marking

Tsotsil is a language of the Tseltalan branch of the family, spoken by

approximately 250,000 people in the state of Chiapas.

This language will be used to show the existence of two subsets of Set A

and Set B a‰xes. For instance, as with most other Mayan languages,

Tsotsil has preconsonantal and prevocalic Set A forms, shown in Table 3:

In this language, the Set B a‰xes are usually su‰xes, but can also be pre-

fixes, as shown in Table 4:

17. It is traditional to call the Mayan adpositional elements that take ergativemarking ‘‘relational nouns’’ (also called relator nouns), although this is some-what of a misnomer in this case, since -u is actually a grammaticalization ofverbal rather than nominal origin. Relational nouns are a pan-Mayan trait.

18. Tsotsil was previously written Tzotzil, but has recently changed to conform tothe spelling systems of the other Mexican Mayan languages. All informationand examples here are from the works of Judith Aissen (1987, 1997, 1999),a specialist of Tsotsil syntax and major contributor to the study of Mayansyntax in general.

Table 3. Prevocalic and preconsonantal Set A prefixes in Tsotsil

Person Prevocalic Preconsonantal

1st k- j-

2nd av- a-

3rd s- y-

Table 4. Prefixed and su‰xed Set B in Tsotsil

Person Prefix Su‰x

1st i- -on

2nd a- -a

3rd Ø- -Ø

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Examples in (4) demonstrate this variation between Set B prefixes (4a) and

su‰xes (4b, c):

(4) [Tsotsil, Aissen (1987)]

a. l-i-bat

cmp-b1-go

‘I/we went’

b. tal-em-oncome-perf-b1

‘I have come’

c. ch-a-mil-onicp-a2-kill-b1

‘You (sg/pl) are going to kill me’

A special feature of Tsotsil is also that all of the so-called singular forms

are underspecified for number (except for the 1st and 2nd person Set B

su‰xes). Therefore, in the absence of additional explicit plural marking

(see 4a and 4c above and examples 5 below), the prefixed person marking

(and su‰xed 3d person set B, which is actually Ø) could be interpreted as

cross-referencing either singular or plural arguments, depending on the

context:


a. k-il-oj-Ø

a1-see-perf-b3

‘I/we have seen it/them’

b. s-man-oj-Ø

a3-buy-perf-b3

‘He/she/they have bought it/them’

c. av-ixlel

a2-younger.sister

‘Your (sing/pl) younger sister’

To explicitly mark plural the language has various plural person sufixes.

In the first person plural there is a double distinction, one between set A

and B and the other between inclusive and exclusive shown in Table 5

and illustrated in examples (6) below:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40




a. k-il-oj-Ø-tik

a1-see-perf-b3-a1pl.inc

‘We (inc) have seen it/him/them’

b. ch-i-tal-otik

icp-b1-come-b1pl.inc

‘We (inc) are coming’

c. ch-i-s-mil-otikotik

icp-b1-a3-kill-b1pl.excl

‘He is going to kill us (excl)’

For second and third person the explicit plural marking is the general -ik

su‰x:


a. i-s-man-Ø-ik

cmp-a3-buy-b3-pl

‘They bought it’

b. ch-a-bat-ik

icp-b2-go-pl

‘You(pl) are going’

This type of plural marking provided one of the arguments used by Aissen

(1987) to demonstrate the existence of a ‘benefactive voice’ in the lan-

guage, and the fact that the benefactive argument had taken on the role

of direct object (or P grammatical relation).

3.4. Split ergativity

Two major patterns of split ergativity are attested in the Mayan family,

one by aspect and the other by complementation. Ergative split by aspect

Table 5. Set A and B first person plural su‰xes in Tsotsil

Person Set A Plural su‰x Set B Plural su‰x

1 plural inclusive -tik -otik

1 plural exclusive -tikotik/-kotik -otikotik

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



was originally demonstrated for Yukatek by Bricker (1978), and is amply

exemplified in Vapanarsky et al (this volume). Attention will be therefore

given here to the phenomenon of split ergativity by complementation,

which has been one of the contributions of Mayan linguistics to the typo-

logical literature.

Split ergativity in complementation was first identified in the literature

in (Grinevald) Craig’s study of Jakaltek Popti’ complex sentence structure

(1977: 115–116) and is a trait of Q’anjob’alan languages.19 It consists of a

nominative-accusative alignment pattern in certain types of aspectless

complement clauses, i.e. set A on subject of intransitives (‘‘S’’ primitive),

in the context of ergative alignment everywhere elsewhere (which would

have meant set B for that argument). This situation is summarized in

Table 6 below.

The examples in (8) show how, in main (non-embedded clauses) Jakaltek

Popti’ follows the normal ergative alignment pattern of Mayan languages:


a. xk-ach to-yi

cmp-b2 go-vi

‘You went’

b. ch-in haw-il-a

icp-b1 a2-see-vtr

‘You see me’

Table 6. Split ergativity by complementation type in Q’anjob’alan languages

Main Clauseperson marking Primitives

Complement Clauseperson marking

Intransitive B V1 S . . . [ A V2itr]Transitive B A V1 P A . . . [B A V2tr]

19. The argument for considering it is a true case of split ergativity relies on therebeing no morphological evidence of a nominalisation of any type in thisparticular context, while the language can display nominalizing morphologyelsewhere. More examples of split ergativity by complementation type inAkateko, another language of the Q’anjob’alan branch, can be found inZavala (1997).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



However, as shown in the schema above and illustrated with examples in

(9) below, certain types of complement clauses exhibit a pattern of split

ergativity, in the context of a lack of aspect marking (shown in 9b and

9d), and the presence of an intransitive type of su‰x -ni20 for transitive

complement clause (in 9b). The ergative split manifests itself by the pres-

ence of the set A2/ergative marker used as indexation for the subject of an

originally intransitive embedded verb (in 9d), as opposed to the expected

set B2/absolutive:


a. xk-ach hin-kol-o

cmp-b2 a1-help-V

‘I helped you’

b. x-Ø-w-ilwe [hach hin-kol-ni ]

cmp-b3-a1-try b2 a1-help-suff.iv

‘I tried to help you’

c. xk-ach kanhalw-i

cmp-b2-dance-iv

‘You danced’

d. x-Ø-w-il [ha-kanhalw-i ]

cmp-b3-a1-see a2-dance-iv

‘I saw you dance’

3.5. Conclusions on Mayan ergativity

Mayan ergativity consists of a typologically relatively rare system of person

indexation in the verb, and the interplay of two sets of markers (each with

possible subsets). These two sets have been traditionally labeled set A and

set B in Mayan linguistics writing, and correspond to ergative and absolu-

tive markers found in the general linguistic literature. Both sets have been

reconstructed for proto-Mayan, attesting to the long history of ergativity

20. This su‰x -ni resembles the antipassive su‰x of main clauses, and is a com-bination of the reflex of proto-Mayan *-on antipassive su‰x and intransitivethematic vowel -i. For another use of this su‰x -ni see the discussion of theagent focus voice below in 3.4., and the example (12).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



in the family. In addition, several factors converge to facilitate syntactic

argumentation about the ergative nature of Mayan grammar: for one,

transitivity is marked throughout with multiple a‰xations, not just by

these person markers, and, furthermore, the segmentability of the forms

and their limited involvement in morphophonological processes make

them quite transparent and easy to use for syntactic argumentation.

4. A functional-typological approach to Mayan voices

The rich agglutinating verbal morphology of Mayan languages, with its

explicit person markers and su‰xes to mark voice and transitivity has

made it so that the languages of the family have been the source of much

descriptive work on voice systems, in a constant back and forth interac-

tion with general typological discussions ongoing at the time.21 As already

mentioned, a broad functional-typological approach will be applied here

to the presentation of the morphosyntax of voices in Mayan languages,

first in an overview showing the contrastive structural features of the

di¤erent voices, then in a rapid illustrated tour of the voices.22

4.1. Overview of the morphosyntax of the di¤erence voices

The presentation of the morphosyntactic characteristics of the Mayan voices

will be cast in a framework handling the interaction of the following types

of elements:23

21. This interaction between Mayanists and general linguists developed first in theexploratory times of the phenomenon of ergativity itself (identified but notnamed as such until the late seventies in fact), then in a period of discoveryof systems of voices such as antipassives and applicatives. More recently ithas been related to the themes of the typology of inverse voices and of thephenomenon of obviation. See Section 4 below.

22. See Dayley (1981 and 1990) for early informative overviews of ergativity andvoices in Mayan languages, and Campbell (2001) for a thorough discussionof ergativity and voices in a specific language, K’ichee’, a majority Mayanlanguage of Guatemala, cast in today’s terms of discussion of such topics.

23. The model proposed is an adapted version of Givon’s functional-typologicalframework, itself partially heir to early versions of Perlmutter and Postal’s Re-lational Grammar (RG). It is not meant to go into an elaborate formalism,and will not address the challenge that ergativity rises in terms of hierarchyof syntactic roles, but aims to o¤er a way of grasping all the di¤erent levelsof analysis that must be kept in mind to fully describe voice phenomena inthis type of languages.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



– A set of semantic roles, roles determined by the nature of the verb itself

and unchangeable. For the purpose at hand, the ones considered here

are those of agent (Agt), patient (Pat), benefactive (Ben) or instrument

(Instr).

– The three primitive syntactic relations considered in ergative studies: A

(subject of transitive), S (subject of intransitive) or P (object of transi-

tive), to which one should add the function of oblique.

– The notion of alignment between these semantic and syntactic roles,

considering as prototypical (of basic active voice) the alignments of an

agent with the role of subject of transitive verb (A) (or with that of

subject of intransitive or (S) role, but less relevant to the present dis-

cussion of voices), of a patient with the role of object (P), and of a

benefactive with that of oblique (Obl).

– The Mayan specific morpho-syntactic encoding of the syntactic roles

realized through verbal indexation, indicated with lines and arrows in

the schema below. This indexation consists of the assignment of markers

of absolutive case (ABS) or ergative case (ERG), those also labelled in

Mayan linguistics as set B (B) and set A (A), as established earlier in

section 3.

In this type of framework (partially reconstituted for the purpose at hand

of describing the variety of Mayan voices), one needs to add the notion of

hierarchies of semantic, syntactic, as well as pragmatic roles.24 Finally, the

last notion needed is that of variations on prototypical alignments, and the

appeal to the notions of ‘‘promotion’’ or ‘‘demotion’’ of certain arguments

in case of non prototypical alignment between semantic and syntactic

roles. For example, instances of ‘‘demotion’’ would be cases of an agent

or a patient argument being realized as an oblique (or nothing), while,

conversely, an instance of ‘‘promotion’’ would be a semantic benefactive

being realized as the object of a transitive verb (P), and not an oblique

(Obl).

Figure 1 below puts the five major voice types found in the Mayan

family in perspective, while focusing on their argument structure. It starts

with the schema of an active transitive construction, presented here with a

24. Here again the framework is presented in shortcut version. But it needs toinclude the notion, introduced by Givon (1983, 1986), of two topics in transi-tive clauses TOP1 (agent) and TOP2 (patient), as demonstrated by elaboratetext counts. This is discussed in more detail below, in Section 4.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



VSO pattern typical of a few Mayan languages (in particular Jakaltek

Popti’) but the verb initial status of Mayan languages at large.25 The juxta-

position of the next two de-transitivized constructions, the passive and the

antipassive, highlights the motivation and origin of the term ‘‘anti-passive’’.

As seen, both constructions are totally parallel structurally, which is why

they share the term ‘‘passive’’. On the other hand, they contrast in the

semantic role associated with the S function: patient in the passive but

still agent in the antipassive, a contrast that is underlined by the term

‘‘anti’’(-passive).

The case of the agent focus construction is more complex and varied in

its indexing mechanism and its ambiguous transitivity status in some lan-

guages, as will be seen below in 3.4. Finally, in the case of the applicative,

the structure is always transitive, and, at least in the case of the benefac-

tive voice shown here, the benefactive third argument is in the P role (the

cases of instrumental voices are more varied: see Section 3.5. below).

Figure 1. Schema of the main voice types in Mayan languages

25. Although the question of word order is a complex one in Mayan languages,they are essentially verb initial languages, so that both core arguments arecontiguous, a typological word order feature often noticed in ergative lan-guages. The variation in word order is between rigid VOS vs. VSO, and vari-able VOS/VSO in some languages (England 1991).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In all cases but the agent focus voice (and some instrumental voices)

one of the two core arguments of the corresponding active voice construc-

tion has been ‘‘demoted’’, through one of several strategies (deletion, rele-

gation to oblique, or ‘‘incorporation’’). The term ‘‘chomeur’’ used in the

applicative construction for the patient argument is meant to attract the

attention to the special status of this argument, still present in the struc-

ture but not in an active object (P) role.26

The above schema only deals with the morphological structure of these

voice constructions. A comprehensive study within a functional typologi-

cal framework also takes into account their actual use, as will be evoked

in section 4 below. The di¤erent voices will now be presented in turn, with

a discussion of their major characteristics illustrated with examples of

selected languages, and wherever relevant mention will be made of some

of the history and impact of the first descriptions of these Mayan voices.

4.2. Passive in Mayan languages

It had been initially assumed in early studies of ergativity that such lan-

guages were not expected to have a passive voice. This emerged from a

pseudo logical reasoning that the antipassive voice newly discovered in

ergative languages had to be for those languages what passives were for

nominative-accusative languages, hence that there was no use for passives

in those languages.27 Craig (1977: 77–83) presented in fact not only evi-

dence of a passive voice in Jakaltek Popti’, but of the existence of up to

four di¤erent passives. Such a multiplicity of subtypes of passives has since

been shown to be quite widespread in the family.

The morphological marking of passive voices in Mayan languages varies

from simple verbal su‰xation to phonological root processes or peri-

phrastic constructions, with possible sensitivity to the nature of the verb

itself, whether radical or derived. Much discussion has involved the treat-

26. The expression ‘‘chomeur’’, invented by David Perlmutter in the early daysof Relational Grammar, is borrowed here to attract attention to the specialstatus of the patient argument of the applicative. It is meant to underlinehow it is there in the clause but has no syntactic status, as evidenced by theabsence of indexing for it on the verb, for instance. Within the RG frame-work, demoted agents of passives (and potentially patients of antipassives)were also deemed to be ‘‘chomeurs’’, i.e. arguments that had ‘‘lost theiremployment’’, which is what the term means in French.

27. This position was reinforced by the absence of passives in Australian lan-guages that were on the other hand some of the languages fueling discussionsof ergative and antipassive voice systems at the time (such as Dixon 1972 forDyirbal).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



ment of the agent of passive clauses, and the discourse functions of the

di¤erent passives. See Vapnarsky et al (this volume) for a detailed study of

the forms and functions of passives in three languages (Yucatec, Ch’orti’

and Tseltal) of di¤erent branches of the family.

4.3. The Mayan antipassive(s)

Discussions of the existence of an antipassive voice in Mayan languages

date from the mid seventies and are encapsulated in a seminal paper by

Smith-Stark (1978). The paper is the result of much discussion among field

linguists that exchanged data and analyses over a period of time, as the

idea of ‘‘antipassive’’ was introduced to them and each one went about

checking it in the language he or she was working on at the time. The

core of the paper is based on data collected for the purpose on five Guate-

malan languages, Pokomam, Kiche, Kekchi, Ixil and Jakaltek Popti’, also

representing di¤erent branches of the family.

In his paper, Smith-Stark tried to make sense of the variety that emerged

of the so-called antipassive voices in Mayan languages from new field

studies. He proposed for instance to distinguish between a so-called ‘‘absolu-

tive voice’’, an antipassive voice in which the patient argument was either

not present at all or expressed in an oblique, and an ‘‘incorporative anti-

passive’’, in which the patient is lexically expressed but does not fulfill any

syntactic role (cf. the discussion of chomeur above), while either literally

incorporated into the verb form or simply next to the verb.

The third type of voice discussed in Smith-Stark’s paper launched the

pan-Mayan discussion of a so-called ‘‘agentive voice’’ which gets its name

from the fact that it is specifically associated with operations that a¤ect an

A argument (i.e. agentive subject of a transitive verb), such as question,

focus or relativisation. This typologically relatively rare construction is

the one labeled ‘‘agent focus’’ voice in Figure 1 above, and which is

considered further in Section 3.4. below. Although the phenomenon of

antipassive (and agentive voice) is a solid pan-Mayan characteristic,

Smith-Stark demonstrated in his paper how it could be traced back to two

reconstructed proto-Mayan verbal su‰xes, although which reconstructed

form is used for which antipassive in today’s languages is very variable.

Structurally, Mayan antipassives are easily identified by their verbal

markings. This will be exemplified with data from an absolutive anti-

passive in Tojol Ab’al. To be noted in (10b) are the various markings of

detransitivization that accompany the actual antipassive su‰x -wan, such

as the set B/absolutive agent indexation, the intransitive theme vowel

-i(y), and, in addition in this particular language, the use of a special

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



form of the incompletive marker restricted to first and second person

intransitive constructions, la-.

(10) [Tojol Ab’al, Peake field notes]

a. wa x-j-pay-aw-a

icp icp-a1-call-vt-b2s

‘I am calling you’

b. wa la-pay-wan-y-on

icp icp-call-ap-vi-b1s

‘I am calling’

The following Jakaltek Popti’ examples primarily illustrate the general pro-

cess of detransitivization of both passive and antipassive voices, evidenced

by a‰xation of voice markers and the maintenance of a set B/absolutive

indexation for the semantic patient in its S (subject of intransitive) func-

tion, in (11b, c and d). The examples can also be taken in pairs, to support

the rationale for the name of ‘‘antipassive’’ in (11b and c), and to show the

di¤erence between so-called absolutive antipassive (with oblique patient

and su‰x -wa) and incorporated antipassive (with non referential patient

and su‰x -wi),28 in (11c and d):

(11) [Jakaltek Popti’ (Craig 1977)]

a. xk-ach s-kol naj

cmp-b2 a3-help cl/he

‘He helped you’

b. xk-ach-kol-lax y-u29 naj

cmp-b2-help-pass a3-rn cl/he

‘You were helped by him’

28. One can identify di¤erent final vowels in -wa and -wi for the two subtypes ofantipassives. In an interesting way, -a# is more reminiscent of a transitivefinal vowel and -i# of an intransitive one. Both constructions show mixedsigns of intransitivity with set B for S argument but overt patient NP. Thispatient argument is oblique in one (with maybe transitive like final vowel)and ‘incorporated’ (or chomeur) in the other (with intransitive like -i). Nospecial glosses will distinguish here the two.

29. A reminder here that oblique NPs are marked by relational nouns taking anergative set A marker, as seen in FN 20 above linked to example (3h). Thisapplies to the agentive oblique -u of (11b) and the patientive oblique -inh ofthe next example (11c).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



c. xk-ach-kol-wa y-inh naj

cmp-b2-help-ap a3-rn cl/he

‘You helped him’ (‘You gave him a hand’)

d. ch-ach kol-wi anma

inc-b2 help-ap people

‘You help people’

Examples of antipassives with patients demoted to an oblique function

and introduced by a relational noun can be also found in Campbell (2001)

for K’ichee’, for example, and a prototypical case of incorporation of non

referential patient arguments in antipassive can be found in early work on

Yukatek by Bricker (1978).

4.4. The agent focus construction

The agent focus construction present in most Mayan languages was first

considered to just be another type of antipassive voice, because it shares

with antipassives the use of one of the two reconstructed antipassive verbal

a‰xes. This construction is not strictly speaking an antipassive, however,

and it has turned out to take many specific forms across the family. It

has been alternatively called the ‘‘agentive’’ voice or the ‘‘agent focus’’

voice due to the fact that, as already mentioned, it is specifically found in

cases of operations on the agent of transitive clauses (A).

Like the antipassive voice, the verb form itself is morphologically mono-

valent, with set B indexation, and an antipassive looking su‰xation. How-

ever, the construction retains two core arguments (in that the patient does

not appear in the oblique), but the two compete for one indexation slot,

filled with a set B marker. In some languages (like Jakaltek Popti’, Craig

1977, 1979), this set B marker can only refer to the patient argument, but

in other languages it may alternate between agent and patient marking (as

in K’ichee’, Mondloch 1978, Cambell 2001), hence the double lines and

arrows in Figure 1 above for this construction. In fact the complexities of

the argument marking process of this construction constituted an early

challenge for theoreticians, and have been the source of discussions on

issues of person hierarchy marking and surface constraints on audible person

marking (set B3 as a zero marker vs. audible B1/2) typical of some Mayan

languages30. In addition, while the construction is obligatorily used in

30. For a discussion summarizing these issues, see Campbell (2001). For a discus-sion of the phenomenon cast into a formal syntactic framework, see Stiebels(2006).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



some languages such as Jakaltek Popti’, Tz’utujiil, K’iche’ or Ixil), it may

be optional (i.e. sensitive to discourse conditions) in others (Akatek or

Tsotsil, for example).31

An early analysis of this construction in Jakaltek Popti’ (Craig 1977,

1979), presented it as a disambiguation mechanism in verb initial languages,

on the basis of contrasting sentences like those of (12a vs. b and c) below.

(12a) shows the two postverbal arguments in the rigid order VSO (or

VAP). (12b) shows the change in the verb form if it is the subject of the

transitive that is questioned, while (12c) shows no change in the verb if

it is the object. Hence the di¤erence in strategies signals which of the two

arguments has been questioned:

(12) [Jakaltek Popti’ Craig 1979]

a. xil naj ix32

x-Ø (y)-(’)il naj ix

cpl-b3 a3-see cl/he cl/her

‘He saw her’

b. mak x’ilni ix

mak x-Ø-’il-ni ___ ix

who cmp-b3-see-suff ___ cl/her

‘Who saw her?’

c. mak xil naj

mak x-Ø (y)-(’)il naj ___

who cmp-b3 a3-see cl/he ___

‘Whom did he see?’

More recent work relying on developments in the typology of inverse voices

and of the notion of obviation has taken a new look, two decades later, at

this construction. Casting it in a new light, it argues that this agent-focus

voice is in fact a type of ‘‘inverse’’ voice, fulfilling functions similar to

31. Craig (1979) gave an early analysis arguing for an agent extraction rule, withsimple set A/ergative deletion required to disambiguate the remaining postverbal NP, and that there was no agentive ‘antipassive’ in that language, althoughthe construction took on more antipassive-like forms in other languages.

32. Morphophonemics render opaque the person indexing on the verb in this case:there is the combination of a set B zero marker, and the deletion of the rootinitial glottal of the verb and of the A3 marker ( y), but the contrast is audiblein (12b) where the absence of an A3 marker prevents the fall of the verb initialglottal.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



those of inverse voices found in the Algonquian languages, for instance

(Aissen 1999; Zavala 1997).

4.5. Mayan applicatives

The analysis of another voice found in Mayan languages contributed in

its time to establish the inventory of a typology of voices found in the

languages of the world.33 The work was carried out on an applicative

voice in the Tseltalan languages, more specifically the description of the

‘‘benefactive voice’’ of Tsotsil originally done by Aissen (1987) within a

relational grammar syntactic framework, followed by descriptions of so-

called ‘‘instrumental voices’’ in the K’ichean languages.

The benefactive applicative of Tsotsil34 is marked by the verbal su‰xa-

tion of a specific voice marker -be, and a non prototypical alignment of

the benefactive third argument with the P syntactic function (rather than

an oblique one), while the patient argument actually remains in the clause,

in a type of ditransitive construction.35

The examples in (13) below show first an active transitive clause (13a),

followed by a corresponding applicative construction marked with the

applicative voice -be su‰x, in which the third argument is a benefactive.


a. i-Ø-j-meltsan j-p’ej na

cmp-b3-a1-make one-nc house

‘I made a house’

b. i-Ø-j-meltsan-be j-p’ej na li Xun-e

cmp-b3-a1-make-appl one-nc house the Xun-cl

‘I made a house for Xun’

33. The approach taken here is that of the path of discovery of voices in Mayanlinguistics, in the early days of the first syntactic descriptions of indigenouslanguages of America. In this case, the initial analysis of the ‘‘benefactivevoice’’ of Tsotsil was cast in the Relational Grammar framework that o¤eredat the time the possibility of a precise descriptive analysis of the construction.This was of course well before any possibility of large typological studies, suchas the very recent and major study of ditransitive constructions by Malchukov,Haspelmath and Comrie (2010). Today these constructions of Mayan languageswould be reconsidered in terms of derived ditransitive constructions.

34. Benefactive is a cover term for the construction, in which the third argumentcan be either a benefactive or a malefactive, or an addressee or target.

35. Considered an instance of ‘chomeur’ (unemployed) argument in the frame-work of Relational Grammar (Perlmutter 1983).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The following examples (13c and d) demonstrate that it is the third argment,

the benefactive, that is cross-referenced on the verb, either with a set B1

marker (su‰xed to the verb in the absence of an aspect marker) as in

(13c), or with both a set B marker and a plural su‰x as in (13d)36. (13e)

demonstrates an applicative passive construction, in which the benefactive

argument aligns with the S primitive (and is thus marked with a set B

su‰x):

c. meltsan-b(-o)-n37 lek i garafon-e

fix-appl-imp-b1 good the jug-cl

‘Fix the jugs carefully for me’

d. ch-a-j-mil-be-ik

icp-b2-a1-kill-appl-pl

‘I’ll kill it/them for you(pl)’

e. ’ak’-b-at-Ø jun.syen soltaro li j-chamu’

give-appl-pass-b3 100 soldier det e3-chamulan

preserente ’une

president cl

‘The Chamulan president was given a hundred soldiers’

The same -be applicative voice marker found in the ‘benefactive’ voice

of Tsotsil has also been identified in other languages in constructions

that exhibit more or less canonical applicative type structures, all involv-

ing, semantically, instrumental arguments. Interestingly, all the cases of

‘instrumental’ voices seem to function and be structured more like the

agent focus constructions than applicative ones. Like agent focus con-

structions, they are found in cases of operations on instruments (with

extraction of the instrument to sentence initial position). They are more

varied yet than the agent focus constructions, however, in the treatment

of the patient argument, which is sometimes clearly demoted to an oblique

in an antipassive-like structure (as in K’ichee’, see Campbell (2001)), but

36. In (13d) the plural -ik must be interpreted as that of the semantic beneficiaryof kill, and not the semantic patient, which is not directly indexed in theconstruction and underspecified for number. An interpretation in which thesecond person beneficiary is singular is not possible in this configuration.

37. A morphophonological rule erases the IMP marker.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



can still be indexed in the verb with the set B (as in Tz’utujil, Craig (1978)

and Dayley 1985).38

5. Conclusion: On the functions of the multiple Mayan voices

As the multiple voices and subtypes of voices were first being described

in the late seventies by a coordinated network of Mayan language field

workers, questions started to be raised about the use and motivation for

such a wealth of constructions.39

Craig (1977), for instance, had provided an early account of a text-

based study of the various passive forms for Jakaltek Popti’.40 For a

much more advanced study of the question, in Yucatec, Tseltal and

Ch’orti’, see Vapnarsky et al (this volume).

Otherwise, in much the same way that Mayan syntactic studies had

developed hand in hand with theoretical advances of the time, a series of

discourse studies of Mayan languages came out, in a second wave, that

were framed within developments in a more functionally oriented linguistics.

This was in particular the case of Mayan studies carried out within the

so-called ‘West Coast functionalist’ approach, using quantitative method-

ologies to study information flow and packaging of information through

discourse, as proposed by Chafe (1979) and Givon (1979).

38. This has been a very complex area of Mayan syntax, where a number ofthemes recur in di¤erent combinations: the notion of voice-like constructionslinked to operations on specific arguments (instrument vs agent focus con-struction), the many variations about the fate of the patient argument (eitherdemoted to oblique function or still P and indexed with set B/absolutive onthe verb) and the nature of the person indexing on the verb. Taken together,these variations on the theme of the possible levels of transitivity of a clauseand possible patterns of person indexation, made Mayan languages a labora-tory for the study of the nature of voice systems, and their possible forms andfunctions.

39. Early studies of the discourse use of voices in Mayan languages relied on someof the new grammatical descriptions. Such was the case for the discourse studyof Jakaltek Popti’ by Datz (1980) following the syntactic description of thelanguage by Craig (1977), as well as the discourse study of Tojol Ab’al byBrody (1982) after the grammar of the language by Furbee (1976).

40. This early analysis focused on the semantics and pragmatics of the agent. Itshowed that two passives, in -ot and -lax, occur preferentially with no agentexpressed, and if expressed, with an agent restricted to the third person. -ot is

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



One of the better known products of the famous ‘Pear Stories’ project

led by Chafe (1980) at the University of Berkeley41 is the seminal paper by

du Bois (1987)42. The study is based on a quantitative analysis of data

from Sakapultek, a language of the K’ichean branch of the family. The

original idea of the paper was to argue for a certain legitimacy of the

phenomenon of ergativity, relatively new to linguistics and still considered

as a puzzling pattern.

The demonstration consisted in showing, quantitatively, the existence

of a dominant pattern in discourse of one lexical argument per clause,

associated with new information, in a pattern that came to be known

as the ‘‘preferred argument structure’’. The fact that this one argument

turned out to be either an S or a P argument argued for what the title

claimed: a discourse basis for ergativity. The quantitative study showed

also another way in which the A argument of a Mayan language was

constrained (as in the case already seen of the agent focus construction),

this time in what du Bois labeled specifically the ‘‘given A constraint’’,

the constraint that the A argument of a clause be associated with given

information.

As it were, the demonstration of a preferred argument structure in dis-

course that came out of the study of a (Mayan) ergative language, turned

out to have much wider implications, as it was found to operate in non

ergative languages as well, as discussed in du Bois’ paper.43

actually preferred in completely impersonal constructions and in the comple-tive aspect, whereas -lax presupposes an impersonal authority or collectiveagent and is preferred with other aspects. The other two, with -lo and -cha,are much less productive and characteristically take an oblique agent whichcan be in any of the three persons. The -cha passive, of limited productivity,is a type of adversative passive, with emphasis on an agent in control of theaction and obligatorily expressed.

41. This is one of the early linguistics projects based on video stimuli, in this casea short silent film made to elicit a narrative about a boy on a bicycle, a man ina tree, and pears being picked, hence the title ‘‘Pear Stories’’.

42. This study was presented numerous times starting in 1981 to diverse groups oflinguists, as mentioned in a striking initial foot note where many illustriouslinguists of the time are listed and who all participated in the shaping of thisresearch. Many subsequent studies of other languages refer to it.

43. In the same way discussions of the existence of antipassive voices in ergativelanguages were later extended to identifying antipassive-like voices in non-ergative languages, as argued early by Postal for French (1977) for instance,the construction Givon treats as non-canonical voices, lacking all the mor-phosyntactic devices of the corresponding canonical voice but functioning indiscourse-like one. This is discussed further below.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The innovative patterns of text counts proposed by Givon for studies of

topic continuity in discourse (1983) were also the source of the ‘‘Functional

analysis of Akatek voice constructions’’ by Zavala (1997). This study con-

sisting of text counts of referential distance (how far to the left an argu-

ment was last mentioned) and topic persistence (how many times in the

next few clauses it was maintained in discourse) demonstrated the contras-

tive discourse functions of passive and antipassive voices linked to topicality.

In this model, where the A argument of a transitive verb is taken to bear

primary topicality but the P argument is also endowed of topicality,

although to a lesser degree (hence the A/TOP1 and P/TOP2 of this model)

a passive construction can be shown to signal the discursive promotion to

higher topicality (TOP1) of the argument P of a transitive verb, while an

antipassive one signals the demotion of the same argument P (to a com-

plete loss of topicality).

More recently, Mayan languages have been revisited from the point of

view of their grammar being sensitive to a number of semantic and prag-

matic hierarchies, beyond the topicality hierarchy just mentioned. New

work on Mayan languages has approached them in terms of their exhibit-

ting signs of functionally inverse and obviation systems, with or without

specific morphology for them. Working within the typological framework

presented by Gildea (1994) for inverse systems (a question of hierarchy of

persons) Zavala (1994) reveals, for instance, the presence of a morpholog-

ical inverse in Huastec (Teneek). Later, following work on the relevance

of the concept of obviation in Mayan languages by Aissen (1997), Zavala

(2007) argues that the use of the passive voice in Chol and Akatek can be

linked to the function of an obviation system, in case the P argument of a

transitive verb is higher on a hierarchy of animacy than the A argument.44

For more extensive discussions and illustrations from three Mayan lan-

guages of the workings of the ranking of third persons in terms of animacy,

definiteness as well as topicality hierarchies, see Vapnarsky et al (this

volume).

This chapter has meant to demonstrate the centrality of the notion of

ergativity in the grammar of Mayan languages. Ergativity is very clearly

44. A remarkable feature of this advanced work on voices in Mayan languages(cast into discussions of inverse and obviation, and their relation to topicaliza-tion and use of voices, including the agent focus voice) is that it has relied ingreat part on detailed studies produced by a new generation of native Mayanlinguists (such as the ones working at CIESAS under Roberto Zavala: GutierrezSanchez (2004), Osorio May (2005), Vazquez Alvarez (2002) cited in England(2007), as well as Pascual (2007) and Pascual and Curiel (2007).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



a major Pan-Mayan trait, with relatively minor language specificities,

marked in the context of a pervasive distributed marking of transitivity.

It also happens that the Mayan patterns of ergatively aligned verbal

indexation are of a nature that is typologically relatively rare.

As demonstrated, the literature on the morphosyntax of Mayan lan-

guages is rich, much of it being concerned with the issue of voice marking.

As this paper has tried to show, Mayan languages have amply contributed

to the general linguistics discussions of ergativity and voices, partly thanks

to its very rich (as well as easily identifiable) verbal morphology. The

contributions of Mayan languages to the establishment of a typology of

voices, for instance, have ranged from the demonstration that ergative

languages may have passives, to the clarification of typological variants

of antipassive voices, and the existence of an agentive voice system, voice

originating in the need for disambiguation in operations on one of two

post verbal core arguments.

More recently studies of Mayan languages have also contributed to

explore the functional load of the various voices. They are shown the func-

tion of passive and antipassive voice as (de-)topicalizing processes, and

have identified the relevance of inverse and obviation phenomena in the

grammar of languages, even those without specific morphological mark-

ings for them.

In conclusion of it all, and to never separate advancements in knowl-

edge from the people that produce that knowledge, we draw attention

to the dynamics of the network of linguists that have been working on

the description of Mayan languages for the last four decades. As already

mentioned, a network of field linguists coordinated the descriptions of

aspects of Mayan languages during the seventies and eighties, through

thematic workshops (the summer ‘‘talleres maya’’) and thematic sessions

at the annual meetings of the AAA, later SSILA. Finally, also remarkable

for the development of Mayan linguistics, several linguists, principally

Nora England, Judith Aissen and Roberto Zavala45, have been working

actively to create a new generation of native Mayan linguists, resulting

in new indepth studies that come to enrich our understanding of Mayan

languages from the inside. This is a rare enough and exemplary develop-

ment on the American continent to be worth mentioning in closing.

45. From di¤erent academic bases: firstly OKMA in Guatemala, then CIESAS-Sur Este of Mexico, and finally CCILA of the University of Texas (England2007, Grinevald 2002, 2007, Woodbury and England 2004).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Abbreviations

a(123) Set A, ergative (1st, 2nd, 3rd person)

af agent focus

ap antipassive

appl applicative

b(123) Set B, absolutive

caus causative

cl classifier

cmp completive (aspect)

det determiner

dir directional

dist distal

erg(123) ergative (1st, 2nd, 3rd person)

excl exclusive

exist existential

foc focaliser

gen generic

icp incompletive (aspect)

inc inclusive

int intensifier

itr intransitive

loc locative (preposition)

neg negative

nc numeral classifier

nf non-finite

nom nominaliser

pass passive

perf perfective

pl plural

prog progressive

rn relational noun

suff su‰x

top topic

tr transitive

vi intransitive theme vowel

vt transitive theme vowel

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



References

Aissen, Judith1979 The syntax of causative constructions. New York: Garland Pub.

Aissen, Judith1983 Indirect Object Advancement in Tzotzil. In Studies in Relational

Grammar 1, David M. Permutter (ed.), 272–302. Chicago: TheUniversity of Chicago Press.

Aissen, Judith1987 Tzotzil Clause Structure. Boston: D. Reidel.

Aissen, Judith1997 On the Syntax of Obviation. Language 73: 705–750.

Aissen, Judith1999 Agent Focus and Inverse in Tzotzil. Language 75: 451–485.

Bricker, Victoria R.1978 Antipassive constructions in Yucatec Maya. In Papers in Mayan

Linguistics, Nora C. England (ed.), 3–24. Columbia: Museum ofAnthropology, Columbia University.

Brody, Jill1982 Discourse Processes of Highlighting in Tojolabal Maya Morpho-

syntax. Ph.D. diss., Washington University.Campbell, Lyle

1978 Quichean prehistory: linguistic contributions. In Papers in MayanLinguistics, Nora C. England (ed.), 25–54. Columbia: Museum ofAnthropology, Columbia University.

Campbell, Lyle2001 Valency Changing Derivations in K’iche’, in Changing Valency:

Case Studies in Transitivity, R.M.W. Dixon and A. Aikhenvald(eds.), 236–281. Cambridge: Cambridge University Press.

Campbell, Lyle, and Terrence Kaufman1985 Mayan Linguistics: where are we now? Annual Review of Anthro-

pology 14: 187–198.Chafe, Wallace

1979 The flow of thought and the flow of language. In Discourse andSyntax (Syntax and Semantics 12.), T. Givon (ed.), 159–181.New York: Academic Press.

Chafe, Wallace (ed.)1979 The Pear Stories: Cognitive, Cultural, and Linguistic Aspects of

Narrative Production. Norwood, NJ: Ablex.Comrie, Bernard

1978 Ergativity. In Syntactic Typology: Studies in the Phenomenologyof Language, Winfred P. Lehmann (ed.), 329–294. Hassocks,Sussex: Harvester Press.

Comrie, Bernard1989 Language Universals and Linguistic Typology, 2nd Edition.

Chicago: University of Chicago Press.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Craig, Colette Grinevald1977 The structure of Jacaltec. Austin: University of Texas Press.

Craig, Colette Grinevald1978 On the Promotion of Instrumentals in Mayan Languages. In

Papers in Mayan Linguistics, Nora England (ed.), Columbia:Museum of Anthropology, Columbia University.

Craig, Colette Grinevald1979 The Antipassive and Jacaltec. In Papers in Mayan Linguistics

Vol. 2, L. Martin (ed.). Lucas Brothers Publishing.Curiel, Alejandro

2007 Estructura de la Informacion, Enclıticos y Configuracion Sintacticaen Tojol Ab’al. M.A. Thesis, Centro de Investigaciones y EstudiosSuperiores en Antropologıa Social, Mexico.

Datz, Margaret J.1980 Jacaltec syntactic structure and the demands of discourse. PhD

dissertation, University of Colorado.Dayley, Jon P.

1981 Voice and ergativity in Mayan languages. Journal of MayanLinguistics 2: 3–82.

Dayley, Jon P.1985 Tzutujil Grammar. Berkeley: University of California Press.

Dayley, Jon P.1990 Voz y ergatividad en idiomas mayas. In Lecturas sobre la linguıs-

tica maya, Nora England & S. Elliott (ed.), 335–98. Antigua,Guatemala: Centro de Investigaciones Regionales de Mesoamerica.

Dixon, Robert M. W.1972 The Dyirbal language of North Queensland. Cambridge Studies

in Linguistics, 9. Cambridge: Cambridge University Press.Dixon, Robert M. W.

1987 Studies in Ergativity. Amsterdam: New York.Dixon, Robert M. W.

1994 Ergativity. Cambridge: Cambridge University Press.du Bois, John W.

1987 The Discourse Basis of Ergativity. Language 63: 805–855.England, Nora C. (ed.)

1978 Papers in Mayan linguistics. Columbia: Museum of Anthropology,Columbia University.

England, Nora C.1983 Ergativity in Mamean (Mayan) Languages. International Journal

of American Linguistics 49: 1–19.England, Nora C.

1991 Changes in basic word order in Mayan languages. InternationalJournal of American Linguistics 57: 446–486.

England, Nora C.1996 Introduccion a la linguıstica: idiomas mayas. Antigua: Proyecto

Linguıstico Francisco Marroquın-Editorial Kamar.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



England, Nora C.2007 The influence of Mayan-speaking linguists on the state of Mayan

Linguistics. In Endangered languages, P. Austin and A. Simpson(eds.), 92–111. Hamburg: Helmut Buske Verlag.

Furbee-Losee, Louanna1976 The Correct Language, Tojolabal: a Grammar with Ethnographic

Notes. New York: Garland Publishing.Gildea, Spike

1994 Semantic and pragmatic inverse: inverse alignment and inversevoice in Carib of Surinam. In Voice and Inversion, T. Givon(ed.), 187–230. Amsterdam: John Benjamins.

Givon, T. (ed.)1983 Topic Continuity in Discourse: A quantitative cross-language study

(Typological Studies in Language 3). Amsterdam: John Benjamins.Givon, T.

1979 On Understanding Grammar. New York: Academic Press.Givon, T.

1986 Syntax: A functional-typological introduction. Volume 1. Amster-dam: John Benjamins.

Givon, T.1994 The pragmatics of de-transitive voice: functional and typological

aspects of inversion, in Voice and Inversion, T. Givon (ed.), 3–44. Amsterdam: John Benjamins.

Givon, T.2001 Syntax: An introduction. Amsterdam: John Benjamins.

Grinevald, Colette2000 A morpho-syntactic typology of classifiers, in Systems of Nominal

Classification G. Senft (ed.), 50–92. Cambridge University Press.Grinevald, Colette

2002 Linguistique et langues mayas du Guatemala. Faits de Langues,Meso-Amerique, Caraibes, Amazonie, Orphys 1: 17–27.

Grinevald, Colette2007 Endangered Languages of Mexico and Central America, in

Language Diversity Endangered, M. Brenzinger (ed.). Berlin:Mouton de Gruyter.

Instituto Indigenista Nacional1988 Lenguas Mayas de Guatemala: documento de referencia para la

pronunciacion de los nuevos alfabetos oficiales. Guatemala: Insti-tuto Indigenista Nacional, Ministerio de Cultura y Deportes.

Instituto Nacional de Lenguas Indıgenas2008 Catalogo de las Lenguas Indıgenas Nacionales. Mexico City:

Secretarıa de Educacion Publica.Kaufman, Terrence

1971 Tzeltal phonology and morphology. Berkeley: University of Cali-fornia Press.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Kaufman, Terrence1972 El proto-tzeltal-tzotzil: fonologıa comparada y diccionario recon-

struido. Mexico City: UNAM, Coordinacıon de Humanidades.Kaufman, Terrence

1974 Idiomas de Mesoamerica. Guatemala: Editorial Jose de PinedaIbarra, Ministerio de Educacion.

Kaufman, Terrence1986 Outline of Comparative Mayan Grammar I: Morphology and

Particles. Unpublished manuscript.Lois, Ximena and Valentina Vapnarsky

2003a Polyvalence of root classes in Yukatekan Mayan languages.Munich: Lincom Europa.

Lois, Ximena and Valentina Vapnarsky2003b Les raciness verbo-nominales du maya yucatec. Faits de Langues

21.Malchukov, Andrej, Martin Haspelmath and Bernard Comrie

2010 Ditransitive constructions: a typological overview. In Studies inDitransitive constructions: a Comparative Handbook, Malchukov,A. M. Haspelmath and B. Comrie (eds.) Berlin: Mouton deGruyter.

Norman, William M. and Lyle Campbell1978 Towards a Proto-Mayan syntax: a comparative perspective on

grammar. In Papers in MayanLlinguistics, Nora C. England(ed.), 136–56. Columbia: University of Missouri.

Pascual, Adan2007 Transitividad y dependencia sintactica y discursiva en Q’anjob’al.

MA thesis, CIESAS, Mexico, D.F.Peake, Marc

2007 Approche sociolinguistique et linguistique du Tojol Ab’al, languemaya du Chiapas. M.A. Thesis, University of Lyon-2.

Perlmutter, David (ed.)1983 Studies in Relational Grammar. Chicago University Press.

Postal, Paul1977 Antipassive in French. Lingvisticæ Investigationes 1:2: 333–374.

Robertson, John S.1980 The Structure of Pronoun Incorporation in the Mayan Verbal

Complex. New York: Garland.Smith-Stark, Thomas C.

1978 Mayan Antipassive: some Facts and Fictions, in Papers inMayan Linguistics, Nora England (ed.), 169–187. Columbia:Museum of Anthropology, Columbia University.

Stiebels, Barbara2006 Agent Focus in Mayan languages. Natural Language & Linguistic

Theory 24: 501–70.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Woodbury, Anthony and Nora C. England2004 Training speakers of indigenous languages of Latin America at

a US University. In Language Documentation and DescriptionVol. 2, Peter Austin (ed.), 122–139. London: Hans RausingEndangered Languages Project.

Zavala, Roberto1994 Inverse Alignment in Huastec. Funcion 15–16: 27–81.

Zavala, Roberto1997 Functional Analysis of Akatek Voice Constructions. International

Journal of American Linguistics 63: 439–74.Zavala, Roberto

2006 Inversion and obviation in Mesoamerica, in Linguistische BerichteSonderheft 14, A. Simpson and P. K. Austin (eds.). Hamburg:Helmut Buske Verlag.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Ergativity and the passive in three Mayan languages

Valentina Vapnarsky, Cedric Becquey, andAurore Monod Becquelin

1. Introduction1

This paper o¤ers a comparative analysis of the passive voice in three

languages – Yucatec, Ch’orti’ and Tseltal – belonging to di¤erent Mayan

branches. Our analysis takes a usage-based approach building on previous

studies of the passive in Mayan languages, and shows that despite impor-

tant di¤erences between the ergative patterns they display, the passive voice

fulfils fairly similar functions across the family. These are minimally syntac-

tic and mostly semantic and discourse-pragmatic.

Ergative languages express in a specific way the relationship between

the agent and the other participants in a multi-participant action, and

one might expect this to influence the configuration of the voice system. It

has been argued that in ergative languages the argument corresponding to

the agent of transitive predicates is structurally – perhaps even conceptually –

secondary with respect to the patient and to the argument of an intransi-

tive predicate. Thus, in a mirror-image reversal of what is typically found

in accusative languages, the antipassive with its object demotion is an ex-

pected phenomenon in ergative languages, whereas from a syntactic point

of view, the passive would not be, since it demotes an already secondary

argument (Jacobsen 1985, Shibatani 1988 among others). In support of

this view, it is known that passive and antipassive may serve to feed the

pivot of constructions requiring accusative or ergative alignment respec-

tively (Dixon 1994: 17, 152 ¤.). However, voices have various other func-

tions, and ergative languages show contrasting patterns as regards valency

change. While some have no passive voice, others, such as the Mayan

languages, have developed a very rich array of passive and middle con-

structions expressing di¤erent types and degrees of relationship between

1. Our thanks go to an anonymous reader and the editors of the volume for theirhelpful comments on previous drafts of this paper, as well as to the revisor ofthe English text.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 11:55) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 51–110 1351 Authier_02_vapnarsky (p. 51)

the agent and the action expressed, in connection with semantic hierar-

chies and discourse functions. Similarly the antipassive constructions of

Mayan languages are not only motivated by syntactic concerns (England

1988: 539, Zavala 1997).

As will be shown in the course of this paper, Mayan languages possess

a variety of passive forms, which may be synthetic or analytic, and may

signal valency change by means of vowel alternations, a‰xation, or peri-

phrases involving auxiliary or light verbs. The distribution of these con-

structions and markers depends on various considerations: the nature of

the stem (root vs. derived), syntactic rules of morphosyntactic alignment,

distinctions concerning the degree of demotion of the agent or promotion

of the patient, and semantic features such as the position of the partici-

pants in the animacy hierarchy. The multiple parameters configuring

passive voices in Mayan languages will be further analysed by a study of

the syntactic, semantic, and discourse-pragmatic factors which favour use

of the passive, as well as how these factors interact with each other. It will

be shown that syntactic constraints have only limited influence, while the

other parameters play a major role, in line with Shibatani’s (1985) notion of

‘‘agent defocusing’’ and Fried’s (2006) proposal on ‘‘agent backgrounding’’.

The three Mayan languages studied here are of special interest since

they show contrastive ergative patterns in the context of this linguistic

family. Ergativity is mainly a morphological phenomenon in all Mayan

languages, but cross-reference systems, split patterns and syntactic properties

linked to ergativity vary and are still much discussed. Tseltal is morpho-

logically a fully ergative language (Shklovsky 2005, Polian 2006). In con-

trast, Yucatec and Ch’orti’ show split ergativity, with intransitives mainly

triggered by aspect and mood (Bricker 1981, Quizar 1994). Ch’orti’ is

unique in the Mayan family in having developed a third paradigm of

personal markers, specialized for the single argument of intransitives in

those contexts where the split occurs. Another relevant property of these

languages is that they do not have the ‘‘focus antipassive’’ construction,

which is used by other Mayan languages such as Mam (England 1988) or

Tzotzil (Aissen 1999) when a nominal clause is questioned, focused or

relativized, and which can be taken as one piece of evidence for syntactic

ergativity.

In section 2, we present some relevant general features of the three lan-

guages, in particular their di¤erent patterns of ergativity and restrictions

on syntactic alignment. In section 3 their voice systems will be briefly

described, with special attention given to passive forms, argument encod-

ing, applicatives, and the di¤erent means available for the expression of

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


52 Valentina Vapnarsky, Cedric Becquey, and Aurore Monod Becquelin

the agent. Section 4 deals with the various constraints and other motiva-

tions leading to passive use in the languages under consideration; we will

compare syntactic factors, semantic hierarchies and discourse prominence,

genericity, and topicality. Section 5 will o¤er concluding remarks. A brief

excerpt from a Yucatec narrative is appended as an illustration of the

points made in section 4.4.1.

2. General observations on the languages under study

2.1. Speakers and Languages

The three languages analysed in the present paper, Tseltal, Ch’orti’ and

Yucatec, belong to three di¤erent branches of the Mayan family, known

as Tseltalan, Cholan and Yucatecan. The Tseltalan and Cholan branches

are grouped together as ‘‘Great Tseltalan’’.2 Tseltal, Ch’orti’ and Yucatec

are spoken in geographically separated areas, except for some sporadic

contacts in big cities in Mexico and elsewhere in North America due to

recent migrations. Tseltal is mainly spoken in the west of the Maya area,

in the Chiapas Highlands, Mexico, by about 400,000 speakers. As many

descriptions attest, Tseltal is highly multidialectal.3 It has also become a

vehicular language in recently settled regions (Selva Lacandona and the

southern part of Campeche). Ch’orti’ is spoken in the far south-east of

the Maya area in Guatemala, close to Honduras, with speaker numbers

estimated at between 11,000 and 30,000.4 Yucatec is spoken in the northern

lowlands of the Maya area, throughout the Peninsula of Yucatan, by about

2. It may be noticed, however, that some Cholan languages, especially Chol andYokot’an (also called Tabasco Chontal), have properties closer to those of theYucatecan languages.

3. See for the Bachajon variant: Slocum, Gerdel and Aguilar 1999, MonodBecquelin 1997; for Tenejapa: Kaufman 1971, Brown 1997, 1998 (amongothers); for Oxchuc: Polian 1999, 2006, to appear; for Petalcingo: Shklovsky2005; Dayley 1981, 1990, Hinman Smith 2004 and Robinson 1999 for otherlanguages of the Great Tseltalan branch. Di¤erent orthographies are used(Tzeltal/Tseltal) by di¤erent authors, as there is no specific recommendationin Mexico.

4. The SIL website mentions 30,000 speakers, Fought (1967) has 20,000, andthe last census of Guatemala gives a figure of 11,734 (XI Censo Nacional dePoblacion y VI de Habitacion, Instituto Nacional de Estadıstica de Guatemala);the last of these seems to come closest to reflecting the present linguisticsituation.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Ergativity and the passive in three Mayan languages 53

800,000 speakers. Compared to Highland Mayan languages, Yucatec shows

little dialectal variation, although insu‰cient research has been carried out

on the subject as yet (but see Pfeiler and Hofling 2006).

2.2. Ergativity

Mayan languages are head-marking: core arguments must be indexed by

personal markers on the verb, whereas lexical arguments are not obliga-

tory and are often omitted. Ergativity mainly a¤ects the encoding of argu-

ments with regard to verbal cross-reference marking. Nominals make use

of the same two sets of person markers as verbs, one to index possessors

and the other to index the argument of stative predicates. Like most parts

of speech, nominals can be used as predicates without the need for a

copula (for a general presentation of these features see Grinevald and

Peake, this volume).

While most languages of the family show an ergativity split triggered

either by aspect and/or mood, person, subordination, or negation, Tseltal

and Tzotzil (Haviland 1981) are the only languages of the family with

no split. Thus in Tseltal, single arguments of intransitive verbs or stative

predicates are consistently encoded with the same personal markers as

objects in transitive predicates, i.e. with ‘‘set B’’, the Absolutive paradigm

(ABS), see (1). Unlike in Mayan languages which possess the ‘focus anti-

passive’ construction (see below), no distinctive treatment is found for

arguments in focalizations, content questions or relative clauses. These

constructions show no syntactic or semantic pivot, although subordination

still needs to be analysed in more detail.

(1) a. Ya x way-at(tse) icp icp.intr sleep-2b

‘You sleep’

b. Ya k-il-aticp 1a-see-2b

‘I see you’

c. Winik-atman-2b

‘You are a man’

In contrast, Yucatec has an ergative split mainly based on aspect and

mood. In the completive aspect and in the subjunctive mood, single argu-

ments of intransitives are encoded with set B, the Absolutive paradigm,

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



i.e. like objects, as shown in (2)a and (2)c with the completive aspect. In

the incompletive they are encoded with ‘‘set A’’, the Ergative paradigm

(ERG), like agents of transitive predicates, as shown in (2)b and (2)d.

Unlike Tseltal and Ch’orti’ but like other Mayan languages such as

K’ichee’, Mam or Chol (Smith Stark 1978, Zavala Maldonado 2003,

2007), Yucatec has two clear classes of intransitives, agentive and non-

agentive, the di¤erences being reflected in aspectual and transitivization

morphology.5 However, in Yucatec these two intransitive classes behave

identically with respect to the ergativity split, as can be seen in (2).6

Non-agentive intransitive Agentive intransitive

(2) a. H luub-en c. H meyah-n-ah-en(yuc) cp.intr fall-1b cp.intr work-ap-intr-1b

‘I fell’ ‘I worked’

b. K-in-luub-ul d. K-in-meyah

icp-1a-fall-icp/nom icp-1a-work

‘I fall’ ‘I work’

The syntactic correlates of split ergativity in Yucatec have been analysed

more than once. Whereas Lehmann (1990) argues that Yucatec could

just as easily be considered basically accusative, Kramer and Wunderlich

(1999), on the basis of a markedness and semantic feature analysis, con-

sider it an ergative language with no ergative-accusative split. Other

authors, such as Bohnemeyer (2004, 2007) and Verhoeven (2007), have

brought into question the notions of subject, syntactic pivot and alignment

in Yucatec. Bohnemeyer considers the traditional notions of subject and

object to be inadequate for this language. He follows previous analyses

of other Mayan languages, Jakaltek-Popti’ and Tz’utujil, for which Van

Valin (1981) and Van Valin and LaPolla (1997: 282–284) argue that the

notion of ‘‘subject is not a meaningful concept’’, on the basis of the variable

5. These classes have been characterized as ‘‘active’’ vs. ‘‘inactive’’ (Lehmann1993, Bohnemeyer 2001); ‘‘agent-salient’’ vs. ‘‘patient-salient’’ (Lucy 1994);‘‘agentive’’ vs. ‘‘non-agentive’’ (Gutierrez Sanchez & Zavala 2007, ZavalaMaldonado 2008). Kramer and Wunderlich (1999) argue that inherent aspectrather than control is the relevant feature that distinguishes Yucatec intransi-tive classes.

6. This is not true of all Mayan languages; see for instance Danziger (1996) fora di¤erent pattern in Mopan. Mopan has an intransitivity split wherebythe argument of agent-salient intransitive forms is always expressed by setA/Ergative personal markers.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



syntactic pivots found in these languages. Verhoeven (ibid: 143–144) shows

that Yucatec also presents di¤erent patterns of neutralization depending

on the construction involved. But she argues for a ‘‘weakly implemented

(accusatively aligned) subject’’ due to (i) the existence of a few construc-

tions which neutralize the contrast between the single argument of intran-

sitives (including passives) and the agent of transitives and (ii) the absence

of a construction neutralizing the argument of intransitives (including

antipassives) and the object of transitives. We will see later that an analysis

of passive use and its syntactic motivation in Yucatec provides further evi-

dence in favour of the view that there is no clear dominant alignment.

In spite of its distinctiveness, there have been very few descriptions of

Ch’orti’.7 Diachronically, Ch’orti’ may have shown a type of split intran-

sitivity similar to that seen in Yucatec. However, this language innovated

by developing a new paradigm of personal markers, used for the indexa-

tion of the single argument of intransitive verbs in the incompletive. The

form and position of the new personal paradigm (prefixal like set A rather

than su‰xal like set B) clearly show that this paradigm evolved, at least

partially, from the ergative markers. The result today is a three-paradigm

system, where the single argument is encoded by Set B (ABS1) in the com-

pletive aspect (3)b, like the object of a transitive verb (3)a, but by a specific

and exclusive ‘‘Set C’’ (ABS2) in the incompletive (3)c.8

(3) a. A-kuch-iy-en(chr) 2a-carry-tr-1b

‘You carry me’

b. Tar-iy-etCome-intr-2b

‘You came’

c. I-tar-i

2c-come-intr

‘You come’

Table 1 o¤ers a synthetic comparative view of the cross-reference patterns

in the languages under study.

7. Fought (1967, 1972), Quizar (1994), Alvarez Ramırez (2004), Perez Martınez(1994), Quizar and Knowles-Berry (1988).

8. The similarity between the transitive and intransitive status su‰xes in examples(3) a and b results from diachronic phonological changes.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 1. Types and positions of personal markers on verbal predicates in Tseltal,Yucatec and Ch’orti’

Language before base base after base

TseltalYucatecCh’orti’

A (ERG)C (ABS2)

intransitiveincompletive

B (ABS)


intransitivecompletive

& subjunctive

B (ABS)B (ABS)B (ABS)


A (ERG)A (ERG)A (ERG)

transitiveB (ABS)B (ABS)B (ABS1)

Many Mayan languages present an agent focus construction whereby a

semantically transitive action is expressed with only one argument indexed

on the verb (see Grinevald & Peake, this volume). Depending on the lan-

guage, the argument indexed may be the patient or the agent, the latter

often occurring along with an antipassive marker on the verb. Ch’orti,

Tseltal and Yucatec have productive antipassive forms, but these do not

fulfil the agent focus function. Nevertheless, in Yucatec, a cleft construc-

tion used only for agent focus (Bricker 1979) shows syntactic properties

reminiscent of the focus antipassive found elsewhere. Yucatec agent focus

is illustrated in (4)b. Although the verb refers to a semantically transitive

action and takes an active inflection, only the patient is cross-referenced

on the verb, contrary to the general rule. The agent is fronted without

cross-referencing. As in the focus antipassive, the construction retains two

core arguments, only one of which is indexed. The verb appears without

preverbal TAM and, in the completive, it bears the subjunctive instead of

the completive transitive su‰x.9 Compare this to the non-focus construc-

tion in (4)a.

9. In the incompletive, the verb bears the same transitive su‰x (-ik) as in non-focus constructions. It might be noticed that both this su‰x and the transitivesubjunctive su‰x -eh are involved in dependent clauses, unlike the completivemarker -ah. It is also worth underlining that Yucatec Agent focus requires thetransitive status su‰xes, in contrast with languages with antipassive focusconstructions, but also in contrast with Akatek where the non-antipassiveAgent focus construction retains the Patient as the indexed argument as inYucatec, but requires the intransitive status su‰x (Zavala 1997: 452–453).Recent works on Yucatec Agent focus are Gutierrez-Bravo and Monforte2009 and Tonhauser to appear.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(4) a. T-uy-il-ah-Ø

(yuc) cp.tr-3a-see-tr.cp-3b

‘She saw it’

b. Le’ti’ il-eh-Ø

3pr see-sbj-3b

‘She was the one who saw it’

There is no equivalent construction for the other arguments, as shown

with wh-fronting constructions which follow the same pattern. Compare

patient fronting (5)a and agent fronting (5)b.10

(5) a. Maax t-uy-il-ah-Ø?

(yuc) Who cp.tr-3a-see-tr.cp-3b

‘Whom did she see?’

b. Maax il-eh-Ø?

(yuc) Who see-sbj-3b

‘Who saw it/him/her?’

By contrast with Yucatec and antipassive focus languages, in Tseltal

(6)a and Ch’orti’ (6)b, agent and patient content questions, relativization,

and focalization do not di¤er from other clause types in their treatment of

core arguments (but see section 4.2).

(6) a. Mach’a la y-il-Ø?

(tse) Who cp.tr 3a-see-3b

‘Whom did she/he see? or Who saw her/him?’

b. Chi uw-ir-a-Ø?

(chr) Who 3a-see-tr-3b

‘Whom did she/he see? or Who saw her/him?’

10. The Agent focus construction in Yucatec holds for all persons (Maax il-ech?‘Who saw you?’ vs. ?? Maax tuyilech?). As Verhoeven notices (2007: 142), dis-ambiguation of the Agent and Patient is carried out by passivization in somerare cases of focus where the agent focus construction does not apply (e.g.with terminative and progressive aspect). This construction retains P as acore argument (whereas in Tz’utujil agent focus is achieved by using an anti-passive, thus keeping A as the core argument, Dayley 1981, Van Valin and LaPolla 1997: 582).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



3. Passive constructions

3.1. Multiple voices

Transitivity is a key feature in all Mayan languages, with important

reflexes in phonology, morphology and syntax. Accordingly, change of

valency is a crucially rich domain of the grammar. There are many transi-

tivizing derivations (causatives and applicatives) and even more intransi-

tivization processes. Among the latter, we find di¤erent types of passives,

antipassives and middles, whose exact number and properties depend on

the language in question (see Grinevald and Peake, this volume, and refer-

ences therein). Thus, for instance, the middle voice is no longer productive

in Tseltal whereas Ch’orti’ and Yucatec have several middle forms which

di¤er semantically and functionally. Tseltal, Ch’orti’ and Yucatec also

have di¤erent antipassive forms depending on distinct criteria (type of

base, aspect, incorporation, and agency properties).

Voice is expressed in verbal, participial, and also nominal-like forms.

These constructions involve di¤erent morphophonological realizations of

the stem as well as potential changes in the encoding of the core argu-

ments on the predicate. We will focus here on productive (non-lexicalized)

passive alternations. In the following sub-sections, we present their forma-

tion in the three languages at issue.

3.2. Tseltal morphological and periphrastic passives

Tseltal valency changes are mainly realized by su‰xation and, in the case

of passives, also by complex predicates with light verbs. The morphologi-

cal passive requires the passive su‰x -ot, available in both completive and

incompletive aspects (as in (7) and (8) respectively), the patient being

cross-referenced by set B (ABS) markers.11

(7) Mam, ahtay-ot-otik-ix, na’-ot-otik-ix,

(tse) Grandfather count-pas-1b.pl-already, know-pas-1b.pl-already

huk-pik-otik

7x8000-1b.pl

‘Grandfather, we have already been counted and identified and

there are 56,000 of us’

11. In the transcription of the following examples, the non-overt third personset B marker (Ø) is omitted unless it is implicated in the constructions wherecoreference is analysed.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(8) Ya x tsak-b-ot-Ø ha’ te s-flecha

(tse) icp icp.intr take-appl-pas-3b dem deic 3a-arrow

‘Their [the enemies’] arrows are taken from them’

Periphrastic passives can be formed with two kinds of light verbs, tran-

sitive and intransitive; they all require an animate subject referring to the

patient but show important di¤erences in argument marking, as we will

see in section 3.6. The verb ich’ ‘‘take, receive’’ is used for a passive with

neutral meaning as seen in (9); the verb a’iy ‘‘feel, hear’’ creates a passive

which may be restricted to adversative contexts in some dialects, as illus-

trated in (10) (Polian 2006: 116–7), due to the experiential value of the

light verb. These two passives take the patient-oriented nominalized form

as their object argument. The intransitive verb la( j) can also be used as

a light verb in a periphrastic passive construction, but it requires the

preposition ta before the Verb-el form, as in (11): see 3.5.12

(9) Te mukenal-e la y-ich’ pas-el ta y-ahlanil

(tse) deic grave-td cp.tr 3a-receive make-nr.pas prep 3a-down

te banti la y-ich’ jijpan-el-e

deic where cp.tr 3a-receive hang-nr.pas-td

‘The grave, it was made at the foot of where he was hanged’ (txt1)

(10) Ya k-a’iy uts’in-el

(tse) icp 1a-hear bother-nr.pas

‘I am bothered’

(11) La y-il laj-em-ix ta bul-el ta chambalam

(tse) cp 3a-see end-pf-already prep destroy-nr prep animal

ha’ te s-k’al

dem deic 3a-field

‘He saw that his milpa had already been destroyed by animals’

In Tseltal, the morphological passive and the periphrastic passive con-

structions are not in strict complementary distribution, although they tend

to be associated with specific functions (see obviation in 4.2) and bear dif-

ferent semantic nuances, representing ‘‘neutral’’ and ‘‘adversative’’ passives

(ich’ and a’iy/laj respectively).

12. Notice that the verb a’iy is also used to form the antipassive voice (Lois et al.2010) and that laj is also the completive marker of transitive verbs.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



3.3. A‰xation and vowel alternation in Yucatec and Ch’orti’ passives

In Yucatec and Ch’orti’, as in some other Mayan languages, voice is ex-

pressed in di¤erent ways depending on the nature of the base, i.e. whether

or not it is a canonical CVC bare root, and of which type.13 In most

Mayan languages showing such distinctions, the passive is formed on

canonical roots by means of infixation of -h- or alternation of the nucleus

vowel, whereas other bases require su‰xation. As shown in table 2, this

is the case in Ch’orti’, where passives of canonical transitive roots are

formed by -h- infixation. These stems take the status su‰x -a used for

all derived intransitives, see (12).14 For non-canonical transitive bases,

passive voice is derived by adding the su‰x -n, which again is followed

by the status su‰x -a, as in (13).

Table 2. Ch’orti’ passive forms

active voice passive voice

Canonicaltransitive

Non-canonicaltransitive

completive aspecta-Vb-tr-b

CVhC-a-b Vb-n-a-b

incompletive aspect c-CVhC-a c-Vb-n-a

(12) A-hihx-a u-’ut e ixim

(chr) 3c-shell.pas-intr 3a-grain det maize

‘Maize grains are shelled.’

(13) E pahbur.sa’ a-che’-e-n-a

(chr) det shuco.atole 3c-do-tr-pas-intr

‘The shuco atole is done.’

13. In Ch’orti’, a canonical transitive verb is defined as an underived verb rootwith a CVC phonological template; these verbs require the verbal status su‰x-i (or -e for CeC roots) in the active voice. In this language, non-canonicaltransitive verbs include verbs with derivations (causative, factitive, etc.), non-CVC underived verbs, and also a few CVC verbs requiring a status su‰xother than -i or -e in the active voice. In Yucatec the distinction is between asignificant class of CVC roots with transitive/intransitive alternations basedon changes to the nucleus vowel, and a‰xally derived transitives or rootswith special phonological features.

14. A crasis occurs when the status su‰x is followed by a personal marker. Itresults in the reduplication of the vowel of the personal marker; the statussu‰x vowel is elided while the vowel of the personal marker is reduplicated.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Yucatec passives, as well as antipassives and middles of canonical multi-

valent roots, are formed by vowel alternation involving modification of

the length, tone and/or rearticulation of the nucleus vowel of the basic

CVC root template (Lois & Vapnarsky 2003, 2006). This system makes

use of all four possible vocalic values (short, rearticulated, high-toned

and low-toned long vowel) present in the language, as shown in table 3.

Example (14) illustrates a passive form with a rearticulated vowel, to be

contrasted with the active form with a short vowel, exemplified in (63)

below.

Table 3. Vowel alternation in Yucatec voice (for canonical non-intransitive roots;examples are in the incompletive)

Voice category Vocalic template example

active CvC k-u-huch’-ik ‘She grinds it’

canonical passive Cv’vC k-u-hu’uch-ul ‘It is (being) ground’

middle CvvC k-u-huuch’-ul ‘It gets ground’

antipassive CvvC k-u-huuch’ ‘She grinds’

(14) Mix t-u-han-t-ah-Ø bin-e’

(yuc) neg.emph cp.tr-3a-eat-trzer-tr.cp-3b rs-td

ka h ya’ach’-Ø men le waah-o’!

conj cp.intr crush.pas-3b cause det corn.bread-td2

‘He didn’t even eat them, and he was crushed by the [loaves of ]

bread [falling on him]’

In contrast, for derived transitives and non-canonical roots, the change

involves a‰xation and passive forms require a special su‰x, which varies

for aspect/mood status: icp. -a’al, cp. -a’ab, sbj. -a’ak, (15). For derived

transitives, the passive su‰x directly follows the transitivizer, as in (15)

with kıin-s formed from kıim ‘‘die’’þ causative -s, or in (16) with meen

‘‘do’’þ extraversive transitivizer -t. The various passive su‰xes can be

analysed as -a’þ non-agentive inflectional markers (Lois & Vapnarsky

2003).15

15. See also McQuown (1967), Bricker (1978), Ola and Bricker (2000) and Loisand Vapnarsky (2003) for proposals on the origin of the derivational passiveform.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(15) a. K-u-kıin-s-a’al ‘He is (being) killed’

(yuc) b. Kıin-s-a’ab-ih ‘He was killed’

c. Kıin-s-a’ak-Ø ‘(that/may) he be killed’

(16) K-u-chen-mee(n)-t-a’al u-maatan

(yuc) icp-3a-only-do-trzer-pas.icp 3a-o¤ering

‘O¤erings are made for them’ (lit. their o¤erings are made)

As an overview, table 4 below presents a selection of voice alternations

in Tseltal, Ch’orti’ and Yucatec for a cognate root, JUCH ‘‘grind’’. It shows

that Tseltal mainly resorts to a‰xal morphology (or a periphrastic passive)

where Yucatec has developed systematic vowel alternation marking.

Ch’orti’ is characterized by a mixed pattern, where the passive infix -h-

(creating a complex nucleus Vh) is still productive.16 Ch’orti’ and Yucatec

use su‰xal morphology for derived transitive or non-canonical roots.

Table 4. Voice formations for a cognate root

Tseltal Ch’orti’ Yucatec

Active ya s-juch’-Ø u-huch’-i k-u-huch’-ik She grinds it

Passive(canonical)

ya x juch’-ot-Ø a-huhch’-a k-u-hu’uch’-ul It is (being) ground

Middle ya x juhch’-Ø /ya x juuch’-Ø(unproductive)

a-huch’-p-a k-u-huuch’-ul It gets ground

Antipassive ya x juch’-awan-Ø a-huch’-m-a k-u-huuch’ She grinds

3.4. Passive participle

The passive voice is also expressed by participial forms. The passive parti-

ciple su‰x -BIL, found in many Mayan languages, has values and uses that

vary across the family. Tseltal -bil and Ch’orti’ -bir form perfect passive

16. In Bachajon Tseltal, the -h- infix only appears as a vestige in a few CVC roots,with a middle value of spontaneous event. In some phonetic contexts, -h-is realized by a lengthening of the root vowel (or not realized, in the Oxchucdialect: Polian, to appear). However, due to the lack of studies on phoneticsand prosody, the possible grammatical function of vowel lengthening in thislanguage is still a matter of investigation.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



participles and may be used with all types of transitive bases.17 Unlike

all other participial forms, -bil and -bir can have an attributive function

without further marking, as in the first clause of the Tseltal example (17),

as well as a predicative function, as illustrated in the second clause of (17)

for Tseltal and (18) for Ch’orti’.

(17) Oxom: chik’-bil pak’-bil ahch’al, banti pay-bil-Ø te ch’enek

(tse) Pot burn-ppp make-ppp clay where cook.in.water-ppp deic beans

‘A pot is a fired and fashioned piece of clay where beans are (to be)

cooked’

(18) Usta-bir-Ø e koror.te’ ka-men-er

(chr) Fix/arrange-ppp-3b det trap 1a.pl-rel-prr

‘The trap has been fixed by us’

The Yucatec participial su‰x -bil is also productive on all transitive

forms and again displays passive orientation. Its extensive range of uses,

often conveying manner or purpose senses, as in (19), involves aspectual

values more akin to the incompletive or gerundive (Lehmann 1998: 27,

Bohnemeyer 1998: 279).18

(19) Le meetik-e’ ‘eermanos-o’ob-e’ k-uy-a’a-[i ]k-o’ob-e’

(yuc) det why-td4 evangelist-3b.pl-td4 hab-3a-say-tr.icp-3b.pl-td4

hum.p’e chan kruus-e’ took-bil-Ø!

one-cn.inan small cross-td4 burn-ppas-3b

‘This is why evangelists say that little crosses should be burned

(lit. as for a little cross, it should be burned!’)

As we will see in greater detail in section 4.1, the Yucatec participle in

-bil is preferred in certain subordinate clauses to maintain coreference

between a core argument of the main clause and the patient of the depen-

dent clause.

17. The Tseltal -bil su‰x has also been called a marker of ‘‘perfect’’ (Polian 2006:85) and ‘‘patient-oriented transitive perfect’’ (Shklovsky 2005: 60). It is worthnoticing, however, that the -bil participle may have a purposive sense asin Yucatec, as shown by pay-bil in example (17). In Ch’orti’ -bir is the onlyperfect form for transitive verbs (18), which is not the case for Yucatec andTseltal.

18. Yucatec has other participles with patient orientation, including resultative(-a’an) and middle ([low tone on root vowel]þ -Vl ).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



3.5. Passive nominal forms

In Mayan languages, voice also a¤ects nominal forms. In Yucatec, a word

marked as a passive incompletive without preverbal aspect markers can be

used as a nominal, a property shared by all intransitive forms (Lois &

Vapnarsky, 2009). The nominal passive can be used as an action noun in

nominal contexts like (20), where it appears with a determiner, or as a

non-finite form in subordinate clauses (see section 4.1).

(20) Teep k-a-ts’o’on-o[l ]! Le ts’o’on-o[l ]-o’ mix

(yuc) interj hab-2a-shoot.pas-icp det shoot.pas-icp-td2 neg.emph

aw-ohel bix ken a-defender-t-[i ]k-Ø a-bah.

2a-know how pros 2a-defend-trzer-tr.icp-3b 2a-refl

‘You’re being shot at! You don’t even know how you’re going to

defend yourself from being shot!’

In Ch’orti’, the passive action noun is derived from passive stems (bear-

ing the a‰xes -h- for canonical bases and -n for non-canonical bases) and

shows the su‰x -a’(a)r.19 Ch’orti’ deverbal nouns appear in any nominal

context, as seen in (21), but very rarely with modal auxiliaries (unlike in

other Mayan languages).20

(21) Tama u-bohn-a’r e otot sat-p-a e tumin

(chr) prep 3a-paint.pas-nr det house lost-mp-intr det money

‘In painting the house, the money was wasted (lit. In its being

painted of the house, the money was wasted).’

In Tseltal, the su‰x -el is used to create non-finite forms from intran-

sitive verbs irrespective of the semantic role of the argument, see (22).

However, when it is applied to a transitive base, this results in a patient-

oriented form, see (23). The passive orientation of transitive verbs in their

deverbal form with -el has been reported in early and modern descriptions

19. The su‰x -a’(a)r results, diachronically, from the combination of the archaicform of the status su‰x -ah with the nominalizing su‰x of intransitive verbs, -er.

20. The only case is the impersonal obligative construction with uk’ani ‘‘it isnecessary (to)’’ which requires as complement a deverbal noun based on thepassive:

uk’ani pahk’-a’r e iximit.is.necessary sow.pas-nr det corn

‘it is necessary to sow the corn’.

No similar construction has been found in Ch’orti’ with other auxiliaries.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



of Tzotzil (Haviland 1981: 271) and Tseltal dialects (Ruz 1989: 110,

Hinman Smith 2004: 79, Shklovsky 2003: 92, Polian 2006: 71).21

aux-b ta [Vintransitive-el ] (agent-oriented)

(22) Hich lijk-Ø-ik ta beh-el te cheb alal-etik

(tse) Thus begin-3b-pl prep walk-nr deic two children-pl

‘Thus the two children began to walk’

prog-b ta [Vtransitive-el ] (patient-oriented)

(23) Yakal-on ta tun-el(tse) prog-1b prep follow-nr.pas

‘I am being followed’

These forms can be used as full nominals, as in Ch’orti’ and Yucatec.

(24) ya k-ak’-b-at te aw-u’-el

(tse) icp 1a-give-appl-2b deic 2a-be.able-nr

‘I give you your power/ability’

When su‰xed to a transitive verb, the -el forms are better interpreted

as ‘‘infinitives’’ with more verbal properties (see Lois et al. 2010). In this

21. The form in -el on transitive is found in the following constructions.a) In one of these constructions the non-finite transitive in -el, introduced by

ta, bears a 3A marker which cross-references the patient of the action. Thisconstruction is found in particular with movement verbs and the progressive:prog ta [A-trV-el ], see (a).Hich lok’-ø laj bah-el te mamal ta s-tah-el te karibio-etik-eso go.out-3b d.r go-nr deic old.man prep 3a-reach-nr deic Caribs-pl-td

‘So the old man went to the encounter with the Caribs’;b) Obligatory in Bachajon for first and second person agents, the second con-

struction uses the su‰x -bel and shows transitive cross-referencing of thearguments (set A/ERG for agent; set B/ABS for patient) on the -bel form:[prog-b a-trV-bel-b], see (b). The agent of the -bel form (active orientation)corefers with the set-B form of the subject of the progressive form.Teme yak-on ta j-tij-bel-ø ha’ te j-k’ayob-e. . .conj prog-1b prep1a-play-bel-3b dem deic 1a-drum-td

‘When I am playing my drum. . .’.c) The third construction involves the use of an action noun (for example

CVC root plus su‰x -aw) instead of the non-finite form with -el, and noagreement marking is found on the action noun, see (c).tal-on-ix bah-el ta tah-awarrive-1b-already go-nr prep reach-an

‘I came for the encounter’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



case, the 3A personal marker (1A and 2A are ungrammatical: see note 21,

case b) does not encode the possessor of the nominalized verb but the

object of the transitive verb, as in (25).

(25) yak-on ta a’tel, yak-on ta s-tsun-el k-ixim

(tse) prog-1b prep work, prog-1b prep 3a-plant-nr.pas 1a-corn

‘I am at work, I am planting corn’ (lit. ‘I am in its being planted

of my corn’)

Non-finite forms in -el are also required in many complex construc-

tions, involving the progressive, other modal/aspectual auxiliaries or motion

verbs. To obtain an active transitive interpretation of the non-finite form,

more complex constructions must be used, which vary depending on person

(extra- vs. intralocutive) and dialect.

In sum, passive nominal forms are common in the three languages: in

Yucatec and Ch’orti’ they are based on the verbal passive, whereas in

Tseltal a transitive form su‰xed with the general nominal su‰x -el conveys

passive value. In all three languages, transitive nominal forms are either

impossible or more complex than intransitive ones, including passives

(Polian, to appear, Lois & Vapnarsky 2009).

3.6. Passive and argument encoding

In Mayan languages, passive forms, encoded as intransitives, follow an

ergative pattern. However, di¤erences between these languages with regard

to ergativity and the ergativity split a¤ect the way argument encoding is

realized in the passive.

As Tseltal exhibits a full ergative pattern at the morphological level, in

this language the patients of derived passives, whether finite (-ot) or parti-

cipial (-bil ), are consistently cross-referenced by the absolutive marker.

(26) Manchuk y-ak’ x k’as-b-ot-Ø ta ik’ s-bakel

(tse) imp.neg 3a-give icp.intr break-appl-pas-3b prep wind 3a-bone

‘May his bones not be broken by the wind’

By contrast, Tseltal periphrastic passives show two main argument-

encoding patterns, depending on the valency of the light verb – transitive

or intransitive/stative – but not on that of the dependent verb.

In the first type of argument encoding, the light verbs ich’ ‘‘take,

receive’’ and a’iy ‘‘feel, hear’’ behave like transitive predicates (restricted

to an animate experiencer). The object of the light verb (indexed by a

zero third person morpheme from set B) refers to the action which is

expressed by a passively oriented base (transitive root/stem-el ) – in (27),

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



the fact of being killed. The patient of the passive-oriented non-finite form

is indexed by a set A personal marker on the transitive light verb and not

on the rootþ el form.

(27) Ya y-ich’-tiklan-Ø mil-el te s-winkilel te lum-e

(tse) icp 3a-receive-freq kill-nr.pas deic 3a-people deic village-td

‘The people of this village are killed one after the other’

In the second type of periphrastic passive with laj ‘‘finish’’ (28), as with

the progressive in (29), the head of the complex predicate is monovalent.

In both types, raising of the argument occurs: the patient is indexed

by set B on the first predicate and equi-deleted in the [verb-el ] form that

follows.

(28) Ya x laj-on ta tah-el

(tse) icp icp.intr finish-1b prep reach-nr.pas

‘I have just been caught!’

(29) Yakal-on ta tah-el

(tse) prog-1b prep reach-nr.pas

‘I’m being caught’

In Yucatec, due to the aspect-based ergativity split, argument encoding

in a passive construction depends, as with all intransitives, on the aspect/

mood of the verbal form. In the active/passive alternation, a change of

encoding of the patient occurs in the incompletive (30), which follows the

accusative pattern, but not in the completive and subjunctive, where the

ergative pattern prevails (31).

(30) a. K-uy-il-ik-en(yuc) hab-3a-see-tr.icp-1b

‘He sees me’

b. K-inw-il-a’al

hab-1a-see-pas.icp

‘I am seen’

(31) a. T-uy-il-ah-en(yuc) cp.tr-3a-see-tr.cp-1b

‘He saw me’

b. Il-a’ab-ensee-pas.cp-1b

‘I was seen’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In Ch’orti’, due to the aspect-based ergativity split and the second

intransitive set of personal markers, patients in the passive clause are

never cross-referenced by set A ergative markers but instead by one of

the absolutive sets, either C in the incompletive (32)b, or B in the comple-

tive (32)c.

(32) a. uwire’n

(chr) uw-ir-a-en3a-see-tr-1b

‘He sees me / He saw me’

b. i’inrna

Vn-ir-n-a

1c-see-pas-intr

‘I am seen’

c. irne’n

ir-n-a-ensee-pas-intr-1b

‘I was seen’

To summarize, whereas in Tseltal patients are always cross-referenced

by the same personal marker in active and morphological passive forms,

in Ch’orti’ and Yucatec cross-reference marking di¤ers in the incompletive

aspect, following the general ergative split. In this aspect, patients are

expressed by set A (ERG) in Yucatec and by a special and innovative set

of markers in Ch’orti’. Notably, Tseltal also expresses the patient by a set

A (ERG) marker in two periphrastic constructions involving a transitive

light verb.

3.7. Expression of the agent in passive constructions

Passives di¤er in the degree to which they suppress, defocus or demote the

agent (Thompson 1994: 47). In this section, we will present how these

e¤ects are manifested in the languages at issue.

The demoted agent is never cross-referenced on the verbal core. When

expressed, it is mentioned in an oblique phrase introduced by a ‘relator’.

This element may be a relational noun, tum(e)en in Yucatec (often con-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



tracted to men or ten), (33), Set A-men in Ch’orti’, (40), and 3a-u’un in

Tseltal; or a preposition, ta in Tseltal (36).22

(33) Bey bin t-u-beet-ah-Ø t-uy-oox-pak-o’,

(yuc) How rs cp.tr-3a-do-tr.cp-3b prep-3a-three-cn.times-td4

tumeen bin taan u-chi’ib-il tumeen u-ch’ik-el

relat rs prog 3a-bite.pas-icp relat 3a-flea-pos

‘This is how he made it the third time, because he said he was

being bitten by the fleas.’

The relators have diverse functions besides introducing the passive agent.

In Tseltal ta corresponds to the general non-oriented preposition of Mayan

languages (expressing source, goal, cause, purpose, location, instrument);

-u’un is also non-oriented and can introduce causes (34), including the

agent, as well as result, purpose (35), or any kind of generic relation.

(34) X mahl k’ahk’al y-u’un te mak-bil k’inal

(tse) icp.intr fade.mp day 3a-relat deic cover-ppp weather

‘The day faded because of the overcast sky’

(35) Bayel ya x a’teh-otik

(tse) Much icp icp.intr work-1pl.incl

y-u’un ya j-mak’lin-otik te k-al.nich’n-ab

3a-relat inac 1a-sustain-pl deic 1a-child-pl

‘We work a lot in order to sustain our children’

In Ch’orti’ and Yucatec (tu)m(e)en/-men can introduce any kind of causal

argument or proposition, as seen in example (33) above from Yucatec

where tumeen appears twice, introducing first a causal clause and then a

causal agent.

22. With the exception of ta, these relators are descended from a relational noun(marked as possessed with set A indexing the possessor), a common formationin Mayan languages (Kaufman 1990: 75–78). They correspond to the ‘agentedpassive’ clause in Akatek marked by u’u ‘by’, ‘because of ’, as described byZavala (1997: 453). In Ch’orti’ the relator -men still functions as a full rela-tional noun, including pronominal uses under the form pos-men-(er) withall persons (ni-men ‘‘by/because of/in connection with me’’). In contrast, theTseltal use of the relational noun with -u’un is only available for the thirdperson, whereas in Yucatec the relator tumeen (often contracted to men orten), which clearly has the same origin as Ch’orti’ -men, has been grammati-calized and can no longer be used in pronominal function.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



We find some di¤erences between the languages under study, and among

their dialects, regarding the existence of restrictions on the expression of

the oblique agent. These restrictions depend on animacy and person. In

the Tseltal of Bachajon, which represents one of the most conservative

dialects, one observes a clear tendency for humans and assimilated entities

to be introduced by the relator -u’un (36), whereas non-human agents,

such as animals (37) or inanimates (38), are introduced by the preposition

ta. In Oxchuc (and less commonly in Bachajon), a human agent may be

introduced by ta, but only if indefinite.

(36) K’alal laj te il-ot-Ø y-u’un te kabinal-etik-e,

(tse) When finish deic see-pas-3b 3a-relat deic Carib-pl-td,

ya la x laj. . . alak’in-ot-Ø y-u’unicp rs icp.intr finish. . . pet-pas-3b 3a-relat

‘When they [women] were noticed by the Caribs, (it is reported)

they were . . . petted by them’

(37) La y-ich’ nuts-el lok’-el ta ts’i’ te mis-e

(tse) cp.tr 3a-receive pursue-nr.pas go.out-nr prep dog deic cat-td

‘The cat was chased out by a dog’

(38) Te (a) j-kanan xot’-bil ta machit-e

(tse) deic agp-protect cut-ppp prep machete-td

‘The angel/guardian was cut in half by the machete’

In other Tseltal dialects, such as Oxchuc, the demoted agent is rarely

introduced by a relator or a preposition (Polian 2006: 218), as in (39).

(39) Nahil nop-tes-ot-Ø ha’ te pukuj

first learn-caus-pas-3b dem deic devil

‘First the devil was taught’

‘First he was taught by the devil’

This results in ambiguity in the attribution of syntactic functions and

semantic roles to the nominal phrases of the sentence, especially when

they have equivalent animacy status. In sentences with two lexical argu-

ments, word order may help to disambiguate (Polian 2005).

With respect to person, it is worth noticing that Ch’orti’ allows first

and second person to be introduced as agent/cause by the relator -men,

whereas this is impossible in Tseltal and Yucatec.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(40) k’ahs-a ni-k’ab a-men-er(chr) break.pas-intr 1pos-hand 2a-relat-prr

‘My hand has been broken by you.’

A complex scale operates in the marking of the relationship between a

verb and its demoted agent, involving intertwined parameters, from syntax

to pure stylistic free choice. We will deal in sections 4.3 and 4.4 with the

frequent cases where the passive is used without an overt agent, either

because the agent is topical and easely identifiable from the context or

because it is intentionally veiled, hidden or removed from discourse. Other

intransitive voices or derivations, such as middle forms in Ch’orti’ and

Yucatec and the so-called ‘‘agentless passive’’ of Yucatec (Bricker et al.

1998: 346), allow for the depiction of an action without reference to an

agent; it is less clear that demotion is involved in these cases.23

3.8. Applicative and ditransitive passivizations

Tseltal, like most Mayan languages, but unlike Yucatec and Ch’orti’, has

an applicative voice, called ‘‘referential voice’’ by Dayley (1981) and taken

by Dryer (1986) to identify Tseltal as a primary object language. A -b(ey)

su‰x allows a peripheral argument (addressee, beneficiary, maleficiary,

recipient, target) to be promoted and encoded as a direct object, i.e.

indexed by set B. Morphologically the resulting form is still a two-place

predicate, but one from which the patient has been erased (two set B

markers cannot co-occur). This construction may be analysed as a ditran-

sitive construction with a chomeur argument (see Grinevald & Peake, this

volume). The three arguments may be expressed lexically and no relator is

needed for either of them, see (41).

(41) Ya j-k’an chikan-tes-bey-ex s-tojol

(tse) icp 1a-want appear-caus-2b.pl 3a-price

‘I want to show you its price’

The applicative form can be passivized, with indexation of the pro-

moted argument as absolutive on the verb. In (42) the absolutive plural

marker makes it clear that the indexed argument refers to the recipient.

23. The so-called ‘‘agentless passive’’ -p-ah derivation of Yucatec is lexicalized,only appearing with a few roots. Semantically it seems more appropriate toconsider it as a middle form. As with the other middles, a cause may be intro-duced with tumeen, because of the generic causal value of this relator.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(42) Hich bin.u’til y-uch’-ik-ix ul a’-b-ot-Ø-ik(tse) So how 3a-drink-pl-already gruel give-appl-pas-3b-pl

laj ul

rs gruel

‘That is how they drank gruel, they were given gruel they say’

Yucatecan languages and Ch’orti’ (unlike other Cholan languages) lack

the -bey applicative construction. In most constructions the beneficiary

must be expressed as an oblique argument introduced by a preposition.

However, some semantically ditransitive verbs (in Yucatec a subset of verbs

of transfer, communication, secretion, etc.; in Ch’orti’, to our knowledge,

only toyi ‘‘pay’’) accept two alternative constructions, where the syntactic

object can be either the patient (43)a or the beneficiary/recipient (43)b.24

The same slot and the same personal marker set are used for cross-

reference indexation on the verb. In Yucatec, when the patient is indexed

on the predicate, the beneficiary/recipient appears as an oblique argument

introduced by the preposition t(i), or as an independent pronoun (43)a. In

contrast, when the beneficiary/recipient is coindexed as an object on the

predicate (43)b, the theme is excluded as a core argument (43)c.25 In the

corresponding passive forms of these Yucatec and Ch’orti’ constructions,

either the patient or the beneficiary/recipient can be promoted to the func-

tion of passive argument, as exemplified in (43)d and (43)e for Yucatec.

(43) a. K-u-bo’o[l ]-t-ik-Ø ten in-meyah

(yuc) hab-3a-pay-trzer-tr.icp-3b 1pr 1a-work

‘He pays me for my work’ (lit. ‘he pays my work to me’)

b. K-u-bo’o[l ]-t-ik-enhab-3a-pay-trzer-tr.icp-1b

‘He pays me’

24. In Yucatec, most of the verbs concerned correspond to extraversive transitiv-ization, as defined by Lehmann and Verhoeven (2006). As these authors show,extraversion in Yucatec introduces a direct object which may encode a varietyof participants (patient, goal, stimulus, addressee, etc.) (ibid: 471–473). Suchsemantically diverse direct objects can also appear as single arguments ofextraversive verbs in passive form.

25. For verbs which do not allow a transitive construction with the beneficiary in-dexed as absolutive, the indirect object cannot be passivized (see also Verhoevenibid: 144): Sıi-a’ab-en (o¤er-pas.cp-1b) means only ‘I was o¤ered (to/by some-one)’ and not ‘it was o¤ered to me’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



c. ?? K-u-bo’ot-ik-en in-meyah

hab-3a-pay-trzer-tr.icp-1b 1a-work

d. Bo’o[l ]-t-a’ab-Ø ten in-meyah

Pay-trzer-pas.cp-3b 1pr 1a-work

‘I have been paid for my work’ (lit. ‘my work has been paid

to me’)

e. Bo’o[l ]-t-a’ab-enPay-trzer-pas.cp-1b

‘I have been paid’

Ch’orti’ possesses a couple of trivalent verbs, which exhibit the same

properties as active transitive verbs marked by -bey in Tseltal but without

morphological marking (ahk’u ‘‘give’’, a’re ‘‘say’’), see (44)a. The corre-

sponding passive form of these verbs allows promotion of either the

patient or the beneficiary/recipient. This passive form can still involve

two lexically expressed participants, but only the promoted argument is

indexed on the verb, either the patient or the beneficiary/recipient (as in

the Tseltal passivized applicative), see (44)b.

(44) a. Kaw-ahk’-u-Ø e wya’r e ah-ch’uyma’r

(chr) 1a.pl-give-tr-3b det food det agp-baptism

‘We give the godfather the food’

b. Ni-tatanoy che ahk’-u-n-a-Ø in-te’ tah.rum

1pos-grandfather rs give-tr-pas-intr-3b one-nc wooded.land

‘My grandfather, he said, was given woodland [for cutting]’

Thus, like other Mayan languages, Tseltal has a very productive ap-

plicative voice which allows passive formation with recipients or similar

participants as the core argument. Ch’orti’ and Yucatec lack such a form;

however, they have sets of verbs which, without special marking, also

allow the recipient to be treated as a direct object in the active form and

as a single argument in the passive.

3.9. Synopsis of voice constructions

Ch’orti’, Tseltal and Yucatec, like Mayan languages in general, show

many voice alternations which grammaticalize the di¤erent degrees and

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



ways the agent may be implicated in the action. From the agent focus

construction to the agentless passive or the middle voice, several devices

emphasize the role of the agent or, on the contrary, demote it from the

core to oblique argument status, from a central to a peripheral participant

or to one omitted altogether.

These e¤ects are achieved by means of changes on the morphophono-

logical and syntactic levels, with some di¤erences evident between the

languages under study, such as the Tseltal deployment of periphrastic

constructions contrasting with the Yucatec exploitation of root vowel

alternations. These two languages exhibit opposing tendencies, the former

centrifugal, where changes of valency are exteriorized from the root and

distributed over several clausal constituents, and the latter centripetal,

where changes of valency a¤ect the very core of the root.

Another important di¤erence concerns the way arguments are indexed

on the verb. Whereas in the Ch’orti’ and Tseltal morphological passive,

patients are never cross-referenced by the Set A (ERG) personal markers

(as agents are), this is very common in Yucatec because of its aspectual

ergative split, which a¤ects all intransitives, including passive forms.

Notice nevertheless that in Tseltal two periphrastic constructions with

transitive light verbs involve a patient encoded with set A (ERG). Table

4 presents an overview of passive formations in Tseltal, Ch’orti’ and

Yucatec.

The agent of the passive forms under analysis can be expressed as

an oblique argument, introduced by a generic causal relator or, in Tseltal,

by a non-oriented preposition. Some Tseltal dialects show contrastive means

of introducing the oblique agent, triggered by the animacy hierarchy – a

semantic parameter especially salient in the grammar of this language, as

we will see in the following sections.

In all three languages a recipient can be promoted to the function of

unique argument of a passive clause. Whereas in Tseltal promotion of the

recipient is licensed by a very productive applicative voice, in Ch’orti’ and

Yucatec it results from the lexical properties of certain roots which accept

either a patient or a recipient as their object.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 5. Comparison of passive formations in Tseltal, Ch’orti’ and Yucatec

Tseltal Ch’orti’ Yucatec

Active a-Vt-b a-Vt-tr-b asp-a-Vt-tr-b

Passive on canonicalroots

icp

asp Vt-ot-bcp

Vt-ot-b

icp

c-CVhC-acp

CVhC-a-b

icp

asp-a-Cv’vC-vlcp

Cv’vC-b

Passive on non-canonical &derived transitives

icp

c-Vt-n-acp

Vt-n-a-b

icp

asp-a-V(-trzer)-a’alcp

V(-trzer)-a’ab-b

Periphrasticpassive

asp a-ich’-b [Vt-el ]asp a-a’iy-b [Vt-el ]asp laj-b ta [Vt-el ]

Passiveparticiple

Vt-bil-b Vt-bir-b Vt-bil-b

Nominal passive:canonical roots

Vt-el

CVhC-a’ar

Same as incompletiveintransitive formNominal passive:

non-canonical &derived tr.

Vt-n-a’ar

Expression of theagent

non-orientedrelator:(-u’un)þ animateAgenttaþ inanimateAgent

causalrelator:a-menþAgent

causal relator:(tu)m(e)enþAgent

Applicative passive (asp) Vt-b-ot-b

4. Use of the passive: constraints and motivations

We will now turn to an analysis of the motivations for the use of passive

forms in the three Mayan languages under study. It will be shown that,

although some syntactic constraints do operate, the functions of the

passive are principally semantic and discourse-pragmatic in nature. We

consider coreference constraints, obviation, agent backgrounding and

topicality. The importance and precise contribution of these factors depend

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



on the language involved. In line with the terminology traditionally em-

ployed for Mayan languages, we will use the label ‘‘subject’’ to cover

both the single argument of monovalent predicates, including passives,

and the non-object core argument of transitives.

4.1. Coreference constraints

The use of the passive may be triggered syntactically by coreference con-

straints in subordination. Subordination in the Mayan family is a complex

matter and has been little studied: the details of subordination constructions

(including the need for a relator element, the type of base required – verbal,

non-finite, nominal or participial – and the choice of aspect/mood and

indexation) may vary depending on the type of predicate found in the

matrix clause and the nature of the matrix and subordinate verbs, as well

as the patterns of coreference involved.

In Yucatec, coreference and syntactic control between the core argu-

ments of the main and the subordinate clauses is an influencing factor in

voice selection. As a general rule, the transitive object of an embedded

clause cannot be coreferential with any core arguments of the matrix

clause: in other words, it cannot be the target of control. Specific construc-

tions are required, involving a conjunction (kaaþ transitive in subjunctive),

passivization, or both. Depending on the type of matrix verb involved,

passivization is realized either by the participial passive (-bil ) or by the

canonical passive in the incompletive, the latter sometimes introduced by

the relator (kaa) (see Verhoeven 2007, Bohnemeyer 2009). Whereas no

argument control appears to be at work in the case of embedded clauses

with the canonical passive, the embedded participle in -bil arguably entails

coreference between the subject of an intransitive or the object of a transi-

tive matrix clause and the subject of the embedded passive participle.

Coreference with a core argument other than the object of the dependent

clause does not trigger these constructions, and involves equi-deletion

when the subordinate argument is the single argument of an intransitive.

Observe (45) (where a. and c. are adapted from Verhoeven 2007: 139–

140), in which there is referential identity between the patient of an under-

lying bivalent subordinate predicate and the object of the transitive main

predicate. The patient of the dependent clause cannot appear as the object

of a transitive predicate but has to be expressed by means of one of the

constructions given in (45), which involve a passive form. For some sub-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



ordinate clauses, such as those of perception, the canonical passive with-

out conjunction is also possible, as shown in (46).26

(45) a. Pedro-e’ t-u-tuuxt-ah-Ø Maria isıint-bil

(yuc) Pedro-td4 cp.tr-3a-send-tr.cp-3b Maria bathe-ppas

meen u-maamah

cause 3a-mother

‘Pedro sent Maria to be bathed by her mother’

b. Pedro-e’ t-u-tuuxt(-ah)-en isıint-bil

Pedro-td4 cp.tr-3a-send-tr.cp-1b bathe-ppas

‘Pedro sent me to be bathed’

c. Pedro-e’ t-u-tuuxt-ah-Ø Maria

Pedro-td4 cp.tr-3a-send-tr.cp-3b Maria

kaa isıin-t-a’ak-Ø meen u-maamah

conj bathe-trzer-pas.sbj-3b cause 3a-mother

‘Pedro sent Maria so that she would be bathed by her mother’

(46) Chen bin ka’ t-uy-u’uy-o’ob u-t’a’an-[a]l-oo’

(yuc) only rs conj cp.tr-3a-hear-3pl 3a-speak.pas-icp-3pl

men hun-p’e nohoch wıinik.

cause one-nc big man

‘Then they heard that they were being called by an old man’

26. The canonical passive without a relator is barely acceptable for the example in(45)c: */? Pedro-e’ t-u-tuucht-ah Maria uy-isıint-a’l meen u-maamah[Pedro-td4 cp.tr-send-tr.cp-3b Maria 3a-bathe-pas.icp by 3a-mother]‘Pedro sends Maria to be bathed by her mother’ (Verhoeven 2007: 139).Verhoeven (ibid ) also gives the alternative in (i) where a transitive dependentclause is introduced by a relator. However, when we checked this with Mayaspeakers, this sentence (along with other similar examples) was understood asreferring to Maria performing the action of bathing rather than being bathed,in line with the constraint on coreference with the object of the dependentclause, despite the presence of the relator.

(i) Pedro-e’ t-u-tuucht-ah-Ø Maria(yuc) Pedro-td cp.tr-3a-send-tr.cp-3b Maria

kaa uy-isıint u-maamahconj 3a-bathe 3a-mother

‘Pedro sends Maria so that her mother bathes her’ (Verhoeven)‘Pedro sends Maria so that she bathes her/his mother’(our informants)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In the construction with the passive participle -bil (45)a, equi-deletion

of the patient takes place and no personal marker appears on the sub-

ordinate predicate. (45)b shows evidence that the argument of the -bil

form is controlled: set B on the matrix predicate is always interpreted

as coreferent with the sole argument of the -bil form, and the sentence

is unambiguous.27 In constructions involving the canonical passive, the

patient is expressed by set B (ABS) if the passive is in the subjunctive after

the conjunction kaa as in (45)c, or by set A (ERG) when the passive is

in the incompletive/nominal form without a conjunction, as in (46). This

follows the general pattern of the Yucatec aspect/mood ergative split.

However, it can be argued that none of the constructions involving the

canonical passive imply control. No equi-deletion takes place, unlike in

all other cases of intransitive embedded clauses. Moreover, in construc-

tions of the type seen in (45), where the embedded clause is not introduced

by the conjunction, there is no coreferentiality between the object (set B)

of the matrix clause and the passive subject; rather it is the embedded

clause itself which is indexed by means of set B on the transitive predicate.

Although this cannot be illustrated here for reasons of space, the same

constraints on coreference of the transitive patient of a subordinate

clause hold with O-, A- or S-controlled subordinate clauses (see ex. 138

in Verhoeven 2007), which indicates no clear accusative or ergative align-

ment pattern. Other properties of subordinate constructions in Yucatec

reveal di¤erent types of pivot, but notably none where S and O behave

alike in contrast to A (see Verhoeven ibid: 144). One important exception

might be the agent focus construction already discussed in section 2.2.

Tseltal shows some coreference constraints similar to those of Yucatec.

However, these only apply to third persons, are overridden by the animacy

hierarchy, and vary across dialects. For Tenejapa Tseltal (Robinson 1999:

148) and Oxchuc Tseltal (Polian 2008, to appear, part 9.3.6), passivization

has been reported to be obligatory in the complement clause if the matrix

27. Our analysis di¤ers from that of Bohnemeyer (2009: 29–30) who argues thatthe ‘‘sole argument of the gerundive [the form with -bil ] is not controlled atall’’. The evidence provided is twofold: 1) the ungrammaticality of complexsentences where a coreferential Undergoer argument of -bil would be identifi-able by the set-B marker on the matrix verb; however, such ungrammaticalityseems to depend on the type of the matrix predicate, in that it is expected inpredicates of desire but is not found in others, as illustrated by (45)a; 2) thepossible inclusion of a new participant as the subject of the embedded -bilform; but according to our data, this inclusion results in sentences judged byspeakers as awkward at best.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



subject is coreferential with the dependent patient of a bivalent predicate

(whereas an active form is used in the other cases). Thus, when no semantic

hierarchies or topicality constraints are in play, any core argument which

is coreferent with the subject of the matrix clause is taken as the syntacti-

cally highest argument.

This is illustrated in the version of the construction seen in Oxchuc

(where the passive agent does not have to be introduced by a preposition).

The subject of the matrix clause can only corefer with the agent of the

transitive dependent clause in (47), and with the patient of the dependent

passive in (48); both sentences are unambiguous (examples adapted from

Polian, to appear).

(47) Y-u’un ja’ yi-al [te ya y i-uts’in-Ø j te ch’in kerem]

(tse) 3a-relat foc[3b] 3a-say conj icp 3a-bother det dim boy

‘he thinks that he will bother the young one’

(48) Y-u’un ja’ yi-al

(tse) 3a-relat foc[3b] 3a-say

‘he thinks

[te x-’uts’in-ot-Ø i te ch’in kerem]

conj icp.intr-bother-pas-3b det dim boy

that he will be bothered by the young one’

Further investigation is required as regards dialectal variation in this

domain. In Bachajon Tseltal, coreferentiality between the subject of the

main clause and the object of the dependent clause does not obligatorily

entail use of the passive voice, as illustrated in (49) where an active form

is used.28

(49) Ya (xi)-xi’-Ø ha’ te ini j ya y j-il-Ø i

icp icp.intr-fear foc[3b] det dem icp 3a-see-3b

‘Hei fears this [guy] sees himi’

In Tseltal, use of the passive may also be entailed by syntactic constraints

on the expression of possession. When there is coreference between the

28. The same is true for asyndetic coordination. In the following example, agentor patient can be the subject of the coordinate clause: te winik laj y-il te mut,patil lok’-Ø bah-el [deic man cp 3a-see deic bird afterwards go.out-3b go-nr]‘the man saw the bird, then he/it went’. In the second interpretation, the sub-ject of the first clause is not coreferential with the subject of the second clause.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



possessor of the matrix subject and the patient of a dependent bivalent

predicate, the dependent patient cannot be expressed as an object and

hence a passive form is used, see (50), adapted from Polian (to appear).

This can again be understood according to the hypothesis that the out-

ranking core argument must be indexed as the higher argument in the

embedded clause. As with other complement clauses, this constraint may

be overridden by semantic hierarchies (in particular, when the possessed

participant is higher in animacy than the possessor a dependent active

form may be used, see (51)).

(50) ya x tal-Ø1 il-ot-uk-Ø2 y2-ihts’in1(tse) icp icp.intr come-3b see-pas-sbj-3b 3a-younger.brother

‘her younger brother comes to see her’ (lit. for her to be seen)

(51) ha’-nix te s-winkilel te ya s-lajin-ik i lum.k’inal-to

(tse) foc-adv deic 3a-people conj icp 3a-end-pl prox territory-td

‘It is its inhabitants who end up with this land’

Thus, Yucatec shows syntactic restrictions on coreference with the object

in subordinate clauses, which favour the use of the passive. Similar con-

straints are in play in some Tseltal dialects, albeit in interaction with

semantic hierarchies. At least in Yucatec, the construction is independent

of any specific alignment (either accusative or ergative), since the use of

the passive depends on restrictions of coreference between the object of

the subordinate clause and any of the matrix arguments.

Ch’orti’ does not seem to show the constraints on coreferentiality observed

in Yucatec and Tseltal. Dependent clauses are expressed either in a serial

construction with equi-deletion phenomena or with a conjunction (mainly

used when there is no coreference between the matrix and the dependent

arguments).29 In all cases, any type of coreference may be attested, and

voice alternations are not required.

(52) b’a’kta-Ø a’i uw-ira-Ø

(chr) fear-3b they.say 3a-see-3b

‘They say she fears he can see her’ or ‘They say she fears she

can see him’

29. Ch’orti’, in contrast with the other Cholan languages, does not require anyspecific verbal inflectional morphology to mark syntactic dependency. Equi-deletion optionally occurs when markers of set C or of set B are repeated: uy-akta kukreme’n [3a-leave fall-1b] ‘he let me fall’ which can be realized withoutequi-deletion: uy-akte’n kukreme’n [3a-leave-1b fall-1b] ‘idem’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



4.2. Semantic hierarchies and discourse prominence

The influence of animacy on syntactic structures is a phenomenon which

has received increasing attention in cross-linguistic studies in recent decades

(Aissen 2003, Branigan et al. 2008, among others). Among the Amerindian

languages, it is attested for instance in the Algonquian, Tupi-Guarani and

Carib language families. The passive is one of the most common construc-

tions linked with animacy constraints. The influence of animacy on the use

of the passive is also observed in Tseltal and Ch’orti’, where animacy hier-

archies have strong e¤ects on the grammar. In Yucatec, animacy a¤ects

alignment patterns to a lesser extent.

Tseltal is particularly interesting in this respect because of the obviative

function which is currently being acquired in some dialects by the mor-

phological passive in -ot (Polian 2005, to appear).30 The same phenome-

non has been observed in closely related languages such as Tzotzil, Cholan

and K’ichean (Aissen 1992, 1997, Robinson 1999, Zavala 2007).31

Thus, in her analysis of the Agent Focus construction in Tzotzil –

basically a passive construction with intransitive morphology and transi-

tive semantics – Aissen shows that this type of passive clause ‘‘express[es]

the inverse function only along one dimension, that of obviation’’. This

occurs only in clauses where both arguments are third person and both

are specified (Aissen 1999: 479). A similar analysis is proposed by Polian

for Oxchuc Tseltal, although agent extraction seems less constrained in

Bachajon Tseltal, as we will now see.

30. We follow the contrastive definitions o¤ered by Guillaume (2006, 2009): ‘‘themechanism of hierarchical agreement refers to the fact that transitive verbsagree with the core participant of the clause which is higher in terms of theperson hierarchy, regardless of its grammatical function [. . .] The mechanismof obviation is manifested in combinations of third person participants. Theparticipants are ranked on a hierarchy according to the semantics of theirreferent along an animacy scale or their respective topicality. [. . .]. The phe-nomenon of inversion refers to the fact that, in addition to agreement marking(on the verb) and proximate/obviative marking (on the NPs), Algonquianlanguages have special verbal markers to indicate whether the higher rankedparticipant is the subject (direct marking) or the object (inverse marking)’’.As we do not find special marking of this kind in Bachajon Tseltal, we preferto avoid the term ‘inverse’ despite its use by Polian (to appear) for Oxchuc.

31. Robinson (1999: 163) considers that Tseltal is better characterized as having a‘‘true passive’’, ‘‘since person and agreement converge on the promotion of thepatient and the demotion of the agent’’. This is in contrast with Tzotzil, whichpresents an agent focus inverse (Aissen 1999), and Akatek, which has two‘inverse’ constructions (Zavala 1997: 457–60).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In its obviative function in Tseltal, the morphological passive, rather

than demoting the agent argument, signals the higher saliency of the

patient in terms of animacy (53), or other hierarchies such as definiteness

(54), specificity, possession (53) and topicality (for a detailed analysis, see

Polian 2005 and 2006).

In (53), the men’s heads (standing metonymically for the men them-

selves) are given a higher degree of animacy than the magic powder and

constitute the possessed NP – both features leading to use of the passive –

while the powder is focused by means of a cleft construction.

(53) Pero ha’ te sibak te la s-tek’-e

(tse) emph dem deic powder rel cp.tr 3a-trample-td

jim-b-ot-Ø s-jol

spoil-appl-pas-3b 3a-head

‘But it is this powder they trampled on, (the reason why) their

minds got confused’

In (54), the patient (the women) constitutes the topic, and is encoded by

a definite NP, which is also more highly specified than the agent thanks to

the accompanying relative adjunct (‘‘who were bathing’’).

(54) tawaltay-ot-Ø y-u’un j-yakubel ha’ te ants

(tse) threaten-pas-3b 3a-relat agp-drunkard dem deic woman

te ay ta atimal

rel exist prep bath

‘Those women who were bathing were threatened by a drunkard’

Obviation with passive -ot only concerns the third person (Ø). It helps

to disambiguate the role of arguments in a system without case and with a

relatively flexible constituent order. In dialects (Oxchuc, Tenejapa) where

the relator -u’un, used to introduce passive agents, is frequently omitted

(see section 3.7), this disambiguating e¤ect seems all the more important.

First and second person are not a¤ected by this pattern. There is apparently

no restriction on their use as the object of a transitive predicate with an

inanimate or a less animate agent, as in (55), where the stone appears as

subject.

(55) La y-ehchen-tes-on te ton-e

(tse) cp.tr 3a-hurt-caus-1b deic stone-td

‘The stone hurt me’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



However, with certain verbs, especially those of interlocution, a passive

applicative construction is often preferred when the ranking [1/2> 3] is

violated, as in (56) where in natural discourse one would say (a) rather

than (b).

(56) a. Wojey, al-b-ot-on-ix (y-u’un)

(tse) Yesterday say-appl-pas-1b-already 3a-relat

‘Yesterday, I had already been told (by him)’

b. Wojey, laj y-al-b-on-ix

Yesterday cp 3a-say-appl-1b-already

‘Yesterday, he had already told me’

To summarize, obviation in Tseltal, as in Tzotzil, depends on several

hierarchies which are often interrelated in complex ways: animacy (animate/

inanimate, human or personified non-human/non-human), definiteness and

specificity (including proper noun/common noun), topicality, and the distinc-

tion between intralocutive and extralocutive persons are all relevant factors.

In cases of conflict between animacy and definiteness or animacy and posses-

sion, active and passive voices are both possible. Topicality, to be discussed

in 4.4, interplays with semantic features but overall it represents a weaker

factor. In (57), the topical and definiteness properties of the stones outweigh

their lack of animacy, with the result that no passive is used despite the fact

that an inanimate participant (stones) is ‘acting’ on human patients:

(57) Ch’ay tal koh-el ton-etik, lom muk’-ix nax

(tse) Get.lost come fall-nr stone-pl very big-already only

i te muk’ul ton-etik-e la s-mil-tiklan

and deic big stone-pl-td cp 3a-kill-freq

j-me’-tik.j-tat-ik-etik-e

1a-mother-rev.1a-father-rev-pl-td

‘Stones fell down, they were really big, and the big stones killed

many of our people (lit. our respected mothers-fathers)’

It is the contextual relation between the two entities that matters and

not their putative absolute values. The entity which is considered more

important at a certain point in a given speech event, either by nature or

by temporary attribution, or because of its role in the narrative or conver-

sation (see 4.4), tends to be presented as the subject. This is a subtle issue

which requires further research, particularly taking into account di¤erent

discourse genres.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In Tseltal the passive can be seen as conditioned by a saliency hierarchy,

which is also of importance for other constructions such as the use of the

relational marker u’un versus the preposition ta to introduce a passive

agent, or the constraint on an animate agent required for the periphrastic

passives with ich’, a’iy and laj (see 3.2). But it is also a strategy used to

disambiguate agent and patient roles when subject and object are third

persons, equivalent in animacy and sometimes without a di¤erentiating

agent relator. Passive voice is mostly linked to hierarchy and word order

to topicality, but the two criteria often overlap: ‘‘information structure is

responsible for word order – in passive and active sentences’’ (Polian

2005: 43, our translation).

In Ch’orti’, the animacy hierarchy [human > animal > inanimate (vege-

table, object, abstract concept)] is crucial in diathetic relations. The study

of a corpus of narratives shows that the argument which is higher in the

hierarchy always appears as the subject in two-participant events. This

rule also governs uses of the passive. Thus, in (58), where both arguments

are topicalized (as shown by their fronted position), the fact that the

patient, a human, is higher in animacy than the agent, an animal, triggers

the use of a passive form.

(58) Inw-ir-a ke e winik u-men e ts’i’ kuhx-a

(chr) 1a-see-tr conj det man 3a-relat det dog bite.pas-intr

‘I saw the man being bitten by the dog’

Again, the ways in which entities are located on scales of saliency must

be read with an awareness of the context. In Mayan tales animals are

often given human attributes and appear in an equipollent relationship

with them, as in (59).

(59) War u-takr-iy-et yer e mama’ ch’o’k

(chr) prog 3a-help-tr-2b dim det uncle mouse

‘The uncle mouse is helping you [Sun and Moon’s son].’

As will be further analysed in the next section, in Ch’orti’ topicality

also plays a central role in the use of passive constructions. However,

when animacy and topicality are in conflict, animacy seems to prevail.

This is seen in example (60), where the patient (the ears of corn) is the

topic, and as such might be expected to be expressed as the subject of a

passive clause. But due to the lower status of the patient on the animacy

scale, the passive is not used.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(60) Ya’ a-tur-an yer e nar twa’ yer e nar

(chr) There 3c-sit-posit dim det ear.of.corn for dim det ear.of.corn

ira ma’chi u-k’a’-p-es wa’kchetaka u-winkir e

dem neg 3a-end-mid-caus quickly 3a-master det

wi’nar, e xiximay

hunger det Xiximay

Lit. ‘Here are the ears of corn, so that these ears of corn the

master of hunger Xiximay doesn’t destroy quickly’

Ch’orti’ does not seem to be sensitive to the person hierarchy [1st >2nd > 3rd], as seen in (59) where the object is a second person and the

subject a third person.

In Yucatec, the animacy hierarchy may influence the use of passive

forms, but it does not appear to be a decisive factor. Indeed, Bohnemeyer

(2009) argues that it represents the weakest factor in the prominence hier-

archy [topicality > definiteness > humanness > animacy] which constrains

alignment and word order in Yucatec (see also Skopeteas & Verhoeven

2005). If the patient outranks the agent (U > A) in terms of one of these

parameters – bearing in mind the hierarchy in which these parameters are

ranked – a transitive form is excluded and the speaker may resort to

di¤erent strategies: passivization, left dislocation (for topics), or clefting

(agent focus). Since topicality is a major factor in alignment constraints

(see section 4.4) and topics are more often humans than animals or (a

fortiori) inanimates, instances where the animacy hierarchy acts as the

sole factor of alignment are at best rare in natural conversation. However,

the weakness of animacy as a relevant factor can be exemplified with cases

where animals or even inanimate beings are encoded as subject agents and

humans as object patients, as seen in example (61) below. In (61) the puma

is topical and appears as a transitive subject (set A) with the human

patient indexed as object (set B). Of the elicited examples in (62), although

version (a) sounds more natural because the human participant Juan is

understood as topical, version (b), where the loaves of bread are expressed

as subject agents, is also readily accepted. As in Tseltal and Ch’orti’, first

and second person are not a¤ected by semantic hierarchies, and no restric-

tion has been observed on their use as objects of transitive predicates; in

(63), from a tale, the second person is indexed as the object of a transitive

predicate whose subject agents are the loaves of bread.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(61) T-u-lah-Ø in-taat e koh-o’

(yuc) cp.tr-3a-hit.with.hand-3b 1a-father det puma-td2

‘The puma struck my father’

(62) a. Niik-Ø ya’abach x-tuti.waah yook’ol le Juan-o’,

(yuc) Fall-3b lots gn-corn.bread on det Juan-td2

ka kıin-s-a’ab-Ø men e waah-o’

conj die-caus-pas.cp-3b cause det bread-td2

‘Many loaves of (ritual) corn bread fell on top of Juan and he

got killed by the corn breads’

b. Niik ya’abach x-tuti.waah yook’ol le Juan-o’!

Fall-3b lots gn-corn.bread on det Juan-td2

Ka t-u-kıin-s-ah-Ø!

conj cp.tr-3a-die-caus-cp.tr-3b

‘Many loaves of corn bread fell on top of Juan! And they

killed him!’

(63) Ka t-aw-a’al-e’ ka laah eem-ek-Ø

(yuc) conj cp.tr-2a-say-td4 conj all go.down-sbj.intr-3b

tulaaka le waah-a’, myeentras t-u-yach’-ech !

all det bread-td1 conj cp.tr-3a-crush-2b

‘And you said that all the loaves of bread should fall, but they

crushed you!’ [txt2]

In sum, the animacy hierarchy is a factor motivating use of the passive

in the three languages, though its importance and the details of its opera-

tion vary. In Tseltal its influence is embodied in an incipient obviative sys-

tem, and in Ch’orti’ animacy imposes strong constraints on the use of the

passive for marking topicality (see below), whereas in Yucatec it interplays

more freely with topicality and other discursive preferences. As far as we

know, in none of the languages under study does the person hierarchy

seem to strictly determine changes in diathesis. This weak asymmetry

between the first/second and third person contrasts with what is seen in

other Mayan languages, in particular Mocho, where split ergativity is trig-

gered by person (Larsen and Norman 1979: 353). But it argues for the

idea that the use of passive voice can serve as part of a disambiguating

strategy when two third person participants are involved.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



4.3. Generic, backgrounded or avoided agent

Related to saliency is the fact that, in the three languages analysed here,

passive constructions are very frequently used when the agent is generic,

collective, unspecified, unknown, irrelevant or obvious. In such cases –

which often correspond to contexts where Mayan speakers would use a

transitive with third person plural subject in their Spanish translation –

the agent is not mentioned. This is very common in descriptions of the tra-

ditional way of doing things, or prescriptions for proper ways of behaving

or executing an action, as in (64) and (65). In fact, this use of the passive

might be considered as one characteristic feature of discourse genres

related to the description of traditional practices.

(64) Ma’ hach u-han-t-a’al kan way-e’

(yuc) neg very 3a-eat-trzer-pas.icp snake here-td4

‘Snake is not eaten very much here (¼we don’t eat snake here)’

In Ch’orti’, the generic passive construction is almost systematically

used in narratives presenting recipes, or the sequence of certain routine

events.

(65) Tya’ k’ani a-che’-e-n-a e pahbur.sa’, a-hihx-a

(chr) when vol 3c-do-tr-pas-intr det shuco.atole 3c-shell.pas-intr

u-’ut e ixim bahxan i a-ts’ahy-a tama e ha’.

3a-grain det maize first and 3c-soak.pas-intr prep det water

‘When one wants to make shuco atole, first maize grains are

shelled and soaked in water’

Alternatively, transitive forms with a second person (in Yucatec) or a

first/third person plural (Tseltal) set A marker can also be used to express

a generic agent. Example (66) shows a generic agent expressed in two

ways: first with third person plural (s- . . . -ik), and second by a passive

form (-ot).

(66) La s-na’-ik te bin.u’ til ya x [s-]chik’-ik te

(tse) cp.tr 3a-know-pl deic how icp icp.intr 3a-burn-pl

tan-e, te bin.u’til ya x chik’-ot-Ø te

deic limestone-td deic how icp icp.intr burn-pas-3b deic

tan-e sok ts’u-bil taj-etik

limestone-td with cut-ppp pine-pl

‘They knew how to burn limestone, how limestone was burnt with

pine chips’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In other cases the use of a passive is motivated by the intentional deletion

of the agent for pragmatic and cultural reasons (see also Ochs Keenan 1976,

Shibatani 1985, among others). This backgrounding or deletion of the agent

figures among Tseltal politeness devices (Brown 1987), and is common in

Yucatec in conversations concerning supernatural agents in particular (see

also Vapnarsky, to appear). For instance in (67), an extract from a Yucatec

personal narrative, the speaker describes the disturbances and misdemean-

ours of the souls of the dead, who drove his ancestors to abandon their

village. Throughout the narrative, the speaker systematically uses passive

forms to refer to the actions of the malign souls, and by doing so he avoids

mentioning these agents explicitly.

(67) Mun-cha’-[a’]al a-maan-i’!

(yuc) neg.3a.icp-let-pas.icp 2a-pass-td3

Taan-chi’in-i [l ] yeete tuunich

prog.2a-throw.pas-icp with stone

taan-chi’in-i [l ] yeete che’,

prog.2a-throw.pas-icp with stick,

taan-muk-lu’um-t-a’a [l ]

prog.2a-bury-earth-trzer-pas.icp

‘They don’t let you pass (lit. your passing there is not being

permitted), you’re pelted with stones, you’re pelted with sticks,

you’re covered with earth’

4.4. Topicality

Topicality is another important parameter of passive use in the three lan-

guages under discussion here. The topic is that element of a sentence

which is presented as already existing in the discourse, and/or which the

subsequent portion of speech is expected to be ‘‘about’’. It has been argued

that, in Maya Tz’utujil, the primary topical participant in a discourse

tends to function ‘‘as the syntactic pivot in each of the clauses in which it

occurs’’ (Van Valin & LaPolla 1997: 283¤ ).

The topic appears in ‘‘topic chains’’ (Dixon 1972) either as the agentive

subject of a transitive or as the unique argument of an intransitive predi-

cate. When the topic is the patient of a transitive predicate, it tends to be

expressed not as an object, but as the subject of a passive. This corre-

sponds to a discourse use of the passive found cross-linguistically, both

within the Mayan language family (see Zavala 1997 for a thorough study

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



on Akatek) and also commonly outside it (Cooreman 1987, Givon 1994

among others).

4.4.1. Topical patients as ‘‘subjects’’

Analysis of coordinate constructions in Ch’orti’ narratives reveals a strong

tendency for the subject (whether marked on the verb as ergative or absolu-

tive) to coincide with the topic. This constraint determines constituent

order, but it is overridden by the animacy hierarchy.

According to Quizar (1994), the unmarked word order in Ch’orti’ is SVO,

largely determined by pragmatic conditions: the topical subject appears in

preverbal position, and the subject must be the topic in transitive clauses.32

As for intransitive clauses, both VS and SV seem equally acceptable; the

condition that the topical subject should appear in preverbal position

explains the SV pattern.33

In her analysis, Quizar only considers the contrast between transitive

and intransitive verbs, without taking into account that the latter are often

passive, middle, or antipassive derived intransitives. This is illustrated in

table 6 with a narrative collected in Oquen, Jocotan.

Table 6. Types of intransitive forms in a Ch’orti’ narrative (wi’nar)

Number of occurrences(% on monovalent verbs)

underived intransitives 27 (41.54%)

passive forms 21 (32.31%)

middle forms 12 (18.46%)

antipassive forms 5 (7.69%)

32. Whereas in Yucatec transitive clauses the preferred order is VS when theobject is not lexically expressed and the agent is new information, in Ch’orti’the subject appears predominantly in preverbal position (SV) when the agentis new information, even if the lexical object is omitted.

33. For example, VS represents 47.62% of monovalent predicates in the narrative‘‘Harvest’’ and 54.84% in the narrative Wi’nar ‘Hunger’ (see table 6). All theexamples from Ch’orti’ come from a corpus of texts recorded in Oquen,Jocotan, Guatemala in 2007. The text ‘‘Harvest’’ (10 mins) describes certainrituals and traditions that precede the corn harvest. The text Wi’nar (8 mins)describes the great famine of 1914 and the di¤erent means used by the Ch’orti’people to survive.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Quizar mentions some rare cases where the patient encoded as an

object is topical. A closer look at these examples reveals that if a transitive

construction is maintained in such cases, this is because the non-topical

agent is higher in animacy, a configuration which excludes passivization,

as seen in (60) above. However, when both arguments rank equally in

terms of animacy, or when the patient is higher in rank, the passive con-

struction tends to be systematically used if the patient of a transitive

action is the topic. In (68) below, the ‘‘angry people’’ are topical, and are

expressed as subjects regardless of their semantic role. This is made possi-

ble by the fact that the people and Jesus are treated as equal in rank with

respect to animacy.

(68) E ah.k’ihna’r-ob [. . .] kora.kora a-sut-p-o’b war

(chr) det angry.people-pl [. . .] anywhere 3c-return-mid-intr.pl prog

u-sahk-o’b i wix che a-mahres-n-o’b

3a-look.for-tr.pl and prog.already rs 3c-deceive-pas-intr.pl

u-men e Jesus ma’chi u-tahw-iy-ob

3a-relat det Jesus neg 3a-find-tr-pl

‘The angry people [. . .] are again looking everywhere for Jesus

and they’re being deceived by him, they haven’t found him’

Tseltal and Yucatec show a similar pattern, although only in the former

does animacy tend to prevail over topicality (see section 4.2). In Tseltal

any passive, morphological (-ot) or periphrastic passive can be used for

the purpose of topicalization.

In (69), both arguments are equal in animacy, and the topic is the man.

He is encoded first as the subject of a modal phrase, and then as the

subject (patient) of the passive verb mil-ot-Ø.

(69) K’an to baht s-t’un te y-ihnam i mil-ot-Ø y-u’un

(tse) want adv go.3b 3a-follow deic 3a-wife and kill-pas-3b 3a-relat

te ants komo ya (x) suht-em-ix ta chambalam

deic woman adv icp icp.intr turn-cp-already loc animal

‘He wanted to follow his wife, but he was killed by the woman

[his wife] as he had already turned into an animal’

In example (70), the referents of the patient argument in the subordinate

clause, (the ‘‘jaguar-puma’’ and the puma) are lower in humanness and

definiteness than the human agent; however, as primary topic in this

section of the narrative, the animals function as the passive subject in the

subordinate clause (indexed by set A y- on the light verb ich’ followed by

the non-finite passive oriented form in -el ).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(70) Te burum.choj ya (x) s-jo’in-tay te choj-e

(tse) deic jaguar.puma icp icp.intr 3a-go.with-trzer deic puma-td

te x-xi’-oj-ik te ya y-ich’-ik nuts-elconj 3a-fear-pf-pl deic icp 3a-receive-pl pursue-nr.pas

y-u’un te kristano-etik

3a-relat deic people-pl

The jaguar-puma [focus] joined the puma because they were afraid

of being pursued by the people’ (Txt2)

Passivization to maintain topic continuity despite the presence of an

agent higher in animacy is also observed in (71). Here the patient argu-

ment corresponds to the noun phrase with s-k’uk’um-al ‘‘its feathers’’.34

(71) Ma’ba ts’ak-al s-k’uk’um-al s-ne te hukpik-e,

(tse) neg complete-adj 3a-feather-pos 3a-tail deic momoto-td

ha’ laj ta s-kaj te tek’-b-ot-Ødem rs prep 3a-cause deic stand-appl-pas-3b

y-u’un te kabinal-etik-e

3a-relat deic caribs-pl-td

‘The feathers of the momoto’s tail are not complete, because they

have been trampled by the Caribs’

The passive can also be used for changes of topic for discourse purposes.

In (72), taken from the description of a ritual performance, the belly and

its possessor become the topic expressed as the subject of a passive form

due to the importance of the as yet unborn child inside it:

(72) ya s-kuch-ik ta ahk’ot te kapitan,

(tse) icp 3a-carry-pl loc dance deic captain

Ya x baht ta [s-]xat-el-ix x-ch’uht,

icp icp.intr go.3b prep 3a-flatten-nr.pas-already 3a-belly

ya [x] xat-b-ot-Ø x-ch’uht,

icp icp.intr flatten-appl-pas-3b 3a-belly

y-u’un te ay laj te t’ut’ alal

3a-relat rel exist rs deic little child

34. K’uk’um ‘feathers’ functions as a metonymic representation of the momotobird; in this section of the narrative, allusion is being made to the origin ofthe shape of the tail of the real bird, and as such, feathers are the topic ofthis episode, although they are clearly inanimate in contrast to the fierceCaribs, enemies of the Bachajontecos.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



‘They (the Caribs) carry the captain for dancing, he goes to the

‘being flattened’ of his belly [ritual], his belly is flattened, because,

they say, there is the little child’

The topically motivated use of the passive in Yucatec is illustrated well

by a section of a narrative whose speaker, an old woman, recalls how she

once killed a puma (see extract in the Appendix). Three participants are

present in the fight being described: the puma, the speaker, and her dogs,

which helped her to fight the feline. Of 51 verbal forms, 35% correspond

to intransitive movement predicates, 33% are transitive active, and 18%

are passive. Leaving aside movement predicates, the narrator is mainly ex-

pressed as a transitive subject but also as an object; the dogs are expressed

as the agents of transitives or the demoted agents of passive forms; while

the puma is expressed as the subject or the object of a transitive verb in

almost equal proportions, and unlike the other participants it also often

appears as the patient argument of a passive form (see the Appendix for

precise figures). This use is arguably connected with the predominant

topical status of the puma in this section of the story. However, changes

in the way the puma is presented also depend on who is dominating

whom during the fight.

In the central section (lines 16 to 26) the puma is consistently the topic,

as is reflected in the discourse by its maintenance as the subject of transi-

tive and passive intransitive predicates. At this point, the puma is mainly

fighting against the dogs, which may be explicitly mentioned but always as

an oblique argument (introduced by the relator ten). However, the main

duel confronts the woman with the puma, at the beginning and the end

of the extract. When these protagonists are involved, the fight is always

presented with transitive predicates clearly indicating the dominating/

dominated roles, and topicality becomes a weaker factor in voice selection.35

The section begins with transitive predicates (1–4, 11–16) whose subjects

refer to the woman and her dogs, while the puma is expressed as the

object. Later on, the puma repeatedly strikes the woman; this action is

also expressed by transitive forms, with the woman (first person pronoun)

now encoded as the object (line 24). Eventually the dogs manage to bring

down the puma (22–33) and the woman kills it by cutting o¤ its head (34–

36). In this final climax, the woman, who is less topical but dominates the

action, recovers her subject position as agent of the verbs of injuring. No

35. This is also seen in column 2 of the table presented in the Appendix, whichshows that the narrator and the puma may be indexed as the object of transi-tives even when they are topical.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



further passive expressions are used with the puma as subject. The defeated

animal is now only referred to as the object of a transitive predicate, a

form presumably emphasizing its patienthood (32–35). This extract shows

the intimate interaction between control of the action, animacy, and

topicality which operates in the choice of passive versus active voice in

Yucatec, and whose analysis crucially requires us to take into account the

roles and types of event depicted, as well as the progression of the narra-

tive, besides/beyond information structure.

4.4.2. Maintaining topic continuity with oblique arguments

Some other interesting cases concern the use of the passive to maintain

topic continuity with oblique arguments. In Yucatec, one construction in-

volves the expression of the topical argument as possessor of the patient

(encoded as set A), which appears as a passive subject. This is exemplified

in (73), where the topic is the subject of intransitive and transitive forms,

as well as the possessor of the passive subject (the demoted agent of this

last action is clearly the protagonist’s wife, a specific human mentioned

just beforehand in the narrative).

(73) Pos le’ti’-e’ mix ween-ih, t-uy-aah-s

(yuc) thus 3pr-td4 neg.emp sleep-intr.icp.3b cp.tr-3a-awake-caus

u-faamilya, ka ts’-u-mee-t-a’al u-yuk’-ul,

3a-wife conj term-3a-do-trzer-pas.icp 3a-drink-nom

u-haan-a[l ] tulaaka bin bey-o’

3a-eat-nom all rs like-td2

‘So, as for him, he didn’t sleep, he woke up his wife, and his drink

and his food were prepared’

Another construction involves verbs of saying in narratives. As a general

rule, these verbs are found in the active form when the topic is the agent,

the speaker of the reported speech event, but in the passive when the topic

is the addressee of the reported speech. Syntactically, the reported speech

is cross-referenced on the verb as the argument of the passive ‘‘being said’’

and the topical addressee is expressed by the indirect pronoun ti’. This use

is fairly systematic and it is an important way for the audience to keep

track of who is speaking to whom in Yucatec narratives, which are tradi-

tionally very rich in dialogue and direct reported speech.36

36. Direct speech is also signalled by the verbatim quotative marker ki (k-þergative su‰x, also translated by ‘‘say’’), to which the rule explained heredoes not apply.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(74) Tunhaanal bine’ kuya’alik bin:

(yuc) He is eating, so-they-say, he says [to his wife], so-they-say’’:

– ‘‘Maadre, senyoora wa ka akre’erten,

– ‘‘Damn, woman, if you would believe me,

hum p’e ba’a ken intsikbatik tech.’’ Ki bin.

I would tell you something’’, said he, so-they-say.

(. . .)

– ‘‘Maadre, pos wa ka xıikech achen ilik ka’achih ki bin,

– ‘‘Damn, and if only you had gone to look, said she so-they-say,

ts’awilik ba’axi’’ kuya’alah bin ti’

‘‘you would have seen what it was!’’ it was said [by his wife] to

him so-they-say

An equivalent distinction is made in Tseltal by means of a passive

applicative. In narratives, the active form x-chi ‘‘he said’’ is postposed to

the quoted speech, but if the addressee of the reported speech is the topic,

then the passive applicative of another verb ‘‘say’’ (al ) is used. This is illus-

trated in (75), where chi is used when the grasshoppers are the speakers

and al-b-ot when they are the addressees.35

(75) – Yak-uk aw-ak’-b-onkotik tehbuk j-we’el-kotik?

(tse) icp-def 2a-give-appl-1b.pl.excl some 1a-food-pl.excl

x-chi-Ø-ik; al-b-ot-Ø y-u’un xanich’-etik:

icp.intr-say-3b-pl say-appl-pas-3b 3a-relat ants-pl

– ay a-wokol-ik y-u’un ma’ba a’tej-ex!

exist 2a-trouble-pl 3a-relat neg work-2b.pl

– Would you give us some (of ) our food?’’

they (grasshoppers) say. They were answered by the ants:

– you got into trouble because you didn’t work!

In summary, motivations for the use of the passive cover the same set

of parameters in Ch’orti’, Tseltal and Yucatec, but these parameters are of

di¤erent importance in each language. The passive is triggered by syntactic

constraints only in restricted cases, which involve cross-reference in certain

subordination constructions. Semantic hierarchies, especially animacy, play

a significant role. The passive is also a privileged and common way to

express a generic, implicit or hidden agent. Finally, topicality seems to be

one of the most important factors for the use of the passive in all three

languages. Genericity and topicality are of equal importance in the three

languages. Overall, Yucatec seems to be more sensitive to syntactic con-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



straints and Tseltal and Ch’orti’ to be more significantly a¤ected by

animacy and other hierarchies.

5. Concluding remarks

On the basis of a study of three morphologically ergative Mayan lan-

guages, Ch’orti’, Tseltal and Yucatec, we have shown the predominance

of discursive, semantic and discourse-pragmatic factors, rather than syn-

tactic ones, as motivations for passive use. This is true independently of

the di¤erent configurations of ergativity that characterize each of these

languages, and is coherent with observations made for other Mayan lan-

guages such as Mam or Akatek. In Yucatec, which presents a stronger

mixed accusative pattern than the other two languages as regards cross-

reference marking, some syntactic constraints on the use of the passive

do operate. However, these constraints do not reveal any clear or pre-

dominant accusative alignment. Furthermore, Yucatec has an agent focus

construction showing a specific syntactic treatment of the agent which is

comparable to an ergative pattern, and which makes Yucatec closer to

those Mayan languages which possess the focus antipassive. These properties

confirm that there is no unified syntactic alignment in this language. Tseltal,

which is fully ergative in terms of cross-reference marking, shows syntactic

constraints entailing the use of the passive similar to those seen in Yucatec.

But in Tseltal these constraints are subordinated to semantic hierarchies,

in particular animacy, due to the prominence of the obviative system in

this language. While Tseltal has no focus antipassive, Tzotzil, a closely

related language, does possess such a construction; it is again governed

by obviation and not by syntactic alignment. In Ch’orti’, voice does not

serve a syntactic realignment function, but instead has semantic, discourse

and information structure motivations.

These facts show that the existence of the passive in Mayan languages

has little to do with their ergative or accusative properties, and that its be-

haviour cannot be taken to provide much information about the putative

‘basic’ status of any particular alignment pattern. Nevertheless, the passive

is an essential part of the grammar and discourse strategies of Mayan lan-

guages, for several reasons: it represents a crucial means of maintaining

topic continuity or signalling a change in the topic in both narrative and

conversation, and hence of disambiguating roles and participants; it repre-

sents one way in which ontological contrasts between di¤erent types of

entities and events have been grammaticalized and are subject to mani-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



pulations by the speakers for discourse-pragmatic reasons; it operates in

strategies of backgrounding or omission of the agent, which depend on

cultural norms of speaking and referring to others. Thus, rather than

being systematically triggered by object promotion and/or agent demo-

tion, the passives analysed in this paper, together with other voices found

in these languages, reveal the grammatical sophistication available for the

expression of the relationships between participants and their particular

levels of involvement in an action, as well as for the subtle control of dis-

course interaction by speakers.

Abbreviations

Ø Third person singular set b (abs)

1 first person

2 second person

3 third person

a set A personal marker (ergative, subject of transitive verbs,

possessor of nouns)

abs1 Ch’orti’ set B su‰xal absolutive personal marker

abs2 Ch’orti’ set C prefixal absolutive personal marker

ag agent

agp agent prefix j-/h-

adj adjectival a‰x

ap antipassive

appl applicative

asp aspect

b set B personal marker (absolutive, subject of intransitive

verbs, non-verbal predicates, object of transitive verbs)

caus causative

chr Ch’orti’

conj conjunction

cp completive

cp.intr completive intransitive initial

tam, cp.tr completive transitive initial tam

def deferential

deic deictic

dem demonstrative

det determiner

dim diminutive

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



emph emphatic

excl exclamative

exist existential

foc focus

freq frequentative

gn gender

hab habitual, generic, incompletive

icp incompletive

icp.intr incompletive intransitive

inan inanimate

imp imperative

interj interjection

intr intransitive

intrzer intransitivizer

mid middle voice

irr irrealis

nc numeral classifier

neg negation

nom nominal

nr nominalizer

nr.pas passive nominalizer for transitive verbs

o syntactic object of transitive verbs

p patient

part particle

pas passive

pf perfect

pl plural

pos possessive su‰x

posit positional

ppas passive participle

ppp perfect passive participle

pr independent pronoun

prep preposition

prog progressive

prr pronominalizer

pros prospective

refl reciprocal/reflexive

rel relative (te)

relat relational marker

rep repetitive

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



retr retrospective

rev reverential

rs reported speech

s syntactic subject of transitive verbs

sbj subjunctive

tam tense, aspect, mood

td terminal deictic

td1 terminal deictic (proximal/new information)

td2 terminal deictic (distal/shared knowledge)

td3 terminal deictic (locative/negation)

td4 terminal deictic (topic)

temp temporal particle

term terminative

tr transitive

trzer transitivizer

tse Tseltal

yuc Yucatec, ?? not attested.

Texts references

CM1 Cuentos Mayas

Txt1 Kuxulon, Estoy vivo (see Entzin Hernandez)

Txt2 Te ame, chanul xux sok te xanich’e; the spider, the wasp and

the ant (see Gomez Ramırez). Examples without reference

are taken from the field materials of the authors – narratives

or elicited glosses – recorded in the years 1986–1996 for

Tseltal, 1994–2007 for Yucatec and 2007–2008 for Ch’orti.

Appendix

Excerpt from a personal narrative told by an old Yucatec woman in Kopchen-

Quintana Roo (‘‘How I killed a puma!’’, 1996, VVA0155-A, 18’25-19’50).

(1) Ka h k’uch-en waalka’ bey-o’,

conj cp arrive-1b run like-td2

And I arrived running like this,

(2) ka t-in-hıi-paay-t in-maachete bey-o’,

conj cp.tr-1a-drag-extract-trzer 1a-machete like-td2

and I took out my machete

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(3) ka tun ok-en waalka ka tun chok’ e peek’ chi’-o’,

conj thus enter-1b run conj part cram det dog bite-td2

I started running and then the dogs surged forward to bite it,

(4) ka ok-en oombre ‘‘cheeepem!’’ in-ch’ak u-pool!

conj enter-1b man interj 1a-cut 3a-head

And I . . . man! I joined in, ‘‘Cheeepem!’’ I cut its head.

(5) Lıik’ u-sıit’ e kooh u-xum-p’aah-t e peek’-o’!

rise 3a-jump det puma 3a-end-leave-trzer det dog td2

The puma jumped to get away from the dogs!

(6) Haa ka t-u-wa’a-kat u-bah.

interj conj cp.tr-3a-stand.up-trzer 3a-refl

Oh! And it stood up.

(7) Laam-Ø e peek’ tuka’aten

Penetrate-3b det dog again

The dogs go into it again,

(8) le k-u-ka’-taal chuk-bi tuka’aten ten e peek’-ob-o’

det icp-3a-rep-come catch-ppas again by det dog-3pl-td2

and it comes and gets caught again by the dogs,

[lit. it comes to be caught again by the dogs]

(9) ka chuuch-mach-t-a’ab-Ø ten e peek’-o’ tuka’aten-e’,

conj suck-take-trzer-pas.cp-3b by det dog-td2 again-td4

it’s grabbed by the dogs again

(10) p’aata ka chu’uk-Ø ten e peek’-ob-e’ ko’lel

part conj catch.pas-3b by det dog-3pl-td4 woman

And it’s caught by the dogs, woman!

(11) ka t-u-ka’-chok’-int-ah-Ø ich yook;

conj cp.tr-3a-cram-trzer-cp.tr-3b in foot

and they run again in between its paws

(12) ka u-chuk-Ø e peek’-o’ u-chok’e ti’

conj 3a-catch-3b det dog-td2 3a-cram prep.3pr

and the dogs catch it, and they get in,

(13) k-u-chi’-chi’-ik-Ø e peek’ ich yook-o’,

icp-3a-bite-bite-tr.icp-3b det dog in foot-td2

the dogs keep biting at its paws,

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(14) ka t-in-kaalama-ch’ak-t ti’ te’ bey!

conj cp.tr-1a-adv-cut-trzer prep.3pr loc like

and I cut it strongly like this! (I gave it a great blow with my

machete)

(15) T-in-ch’ak-ah–Ø ti’ bey-o’.

cp.tr-1a-cut-tr.cp.3b prep.3pr like-td2

I cut it there like this!

(16) Ka ti’ale’ ka t-uy-oot-ah-Ø tun wa’a-kat

conj temp conj cp.tr-3a-want-tr.cp-3b thus stand.up-trzer

u-bah-e’,

3a-refl-td4

And then it tried to stand up,

(17) tun-taal u-lah-en ka’achih-a’ le [le le] kooh-o’!

prog.3a-come 3a-slap-1b temp-td1 det [det det] puma-td2

the puma was coming to claw me!

(18) Tun-taal u-lah-en! Men-m u-k’a ten bey k-u-taal-a’!

Prog.3a-come 3a-slap-1a! do-pf 3a-hand 1pr like icp-3a-come-td1

It was coming to claw me! It lifted its paw towards me and it’s

coming like this!

(19) Ka ti’a lel-a’ ten in-mach/ preparaado’-en yeet in-maaskab-o,

conj temp det-td1 1pr 1a-take/ ready-1b with 1a-machete-td2

And then, me, I took/ I was ready with my machete,

(20) Ka ti’ale’ Chaa!’’ ka la’ap’-Ø paach-i ten le peek’-o’,conj temp interj conj catch.pas-3b behind-nom by det dog-td2

And then tcha! It was caught from behind by the dogs

(21) saam koo’Lpay-t-a(’a)k-Ø!retr pull-extract-trzer-pas.sbj-3b

It was dragged away!

(22) Koo’Lpay-t-a’a(b)-Ø ten e peek’-ob-o’,Pull-extract-trzer-pas.cp-3b by det dog-3pl-td2

Dragged away by the dogs,

(23) ma’ ch’a’-ab-Ø u-taal u-lah-en-e’!

neg let-ps.cp-3b 3a-come 3a-slap-1b-td4

they didn’t let it come and claw me!

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(24) Ka ti’al u-taal u-lah-en t-u-ka’a-ten

conj temp 3a-come 3a-slap-1a prep-3a-two-time

And it’s coming to claw me again

(25) bey ix uuch u-koo’-paay-t-a(’a) ,

like conj happen 3a-pull-extract-trzer-pas.icp,

and the same way it was dragged away,

(26) haan-pay-t-a’(ab)-Ø,

roughly-extract-trzer-pas.cp-3b

it is pulled away,

(27) ka ti’a’lale’ chok’ e peek’-o’, ka tun chook’ e

conj temp cram det dog-td2 conj part cram det

peek’ ich yook-o’,

dog in foot-td2

and then the dogs surged, the dogs surged into its paws,

(28) ka’ ti’al ook-en waalkab ‘‘Cheepen cheepen cheepen, kıilin!’’

And temp enter-1b run interj. . . .

and I started running ‘‘Cheepen cheepen cheepen, kıilin!’’,

(29) ka luub-ih!

conj fall-cp.intr.3b

and it fell!

(30) Aaa! tun-luub-l tun o’tsil kooh tun-e’!

interj prog.3a-fall-icp thus poor puma thus-td4

Aaah! And so the poor puma is falling!

(31) Ay ko’lel he’ tun k-u-luub-u tun-a’,

interj woman ost thus icp-3a-fall-icp thus-td1

Oh woman! There it is that it falls!

(32) he’ tun k-u-cho’ le peek’-o’ tun-chi’-ik-a’!

ost thus icp-3a-cram det dog-td2 prog.3a-bite-tr.icp-td1

And there are the dogs surging forward, they’re biting it

(33) Taan a tun u-ch’a’-ab-l u-lıik’-i?!

prog quest part 3a-let-pas-icp 3a-rise-icp

Are they letting it stand up?! [lit. is its standing up being allowed?]

(34) Ka’ ti’a’l-e’ saam in-he-hen-ch’ak-t u-pool!

conj temp-td4 retr 1a-red-adv-cut-trzer 3a-head

And I already hacked its head!

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(35) Saamih in-kin-s-m-ah-Ø kooh bey-o’!

retr 1a-die-caus-pf-cp puma like-td2

I’ve already killed the puma!!

(36) Ma’ sa’ahk-en-i’! Ma’ sa’ahk-en-i’, chen ch’ak-bi

neg afraid-1b-td3 neg afraid-1b-td3 only cut-ppas

t-in-meet-ah-Ø!

cp.tr-1a-do-tr.cp-3b

I’m not afraid! I’m not afraid! I just did it with my machete!

(37) Ch’ak-bi t-in-meet-ah-Ø!

cut-ppas cp.tr-1a-do-tr.cp-3b

I did it with my machete!

Topicality measurement

We use the quantitative text-based method initially developed by Givon

(1983) and presented in Zavala (1997). Only the referential distance mea-

sure (anaphoric accesibility) is given, since for this short excerpt no signi-

ficant di¤erence is observed by taking topic persistence into account. The

first figure corresponds to the total number of occurrences. In brackets are

given the measures of topicality, from 1, highly topical, to >3, less topi-

cal. The figure before the colon or the > sign corresponds to the number

of occurrences, e.g. 6:1, 2> 3 is to be read ‘6 occurrences are highly topical,

2 are much less topical’. Two cases of the reflexive have been set aside.

Agent ofactive

transitive[A]

Object ofactive

transitive[P]

Passivepatient

[U]

Passiveagent

(Oblique)

Arg. ofmovementpredicate

[U]

Narrator(1st person)

8(6:1, 2>3)

4(3:1, 1:2)

Puma7

(7:1)5

(4:1, 1>3)8

(7:1, 1:2)11

(9:1, 2:2)

Dogs4

(4:1)1

(1:3)1

(1)4

(2:1, 1:2, 1>3)

5(1:1, 2:2–3,

2>3)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



References

Aissen, Judith1992 Topic and focus in Mayan. Language 68: 43–80.

Aissen, Judith1997 On the syntax of obviation. Language 73: 705–50.

Aissen, Judith1999 Agent focus and inverse in Tzotzil. Language 75: 451–485.

Aissen, Judith2003 Di¤erential Object Marking: Iconicity vs. Economy, Natural

Language and Linguistic Theory 21: 435–483.Alvarez Ramırez, Juan et al.

2004 U’tirach e Ojroner Maya Ch’orti’. Gramatica descriptiva Ch’orti’.Ediciones de la Academia de Lenguas Mayas de Guatemala.

Blaha Pfeiler Barbara, and Andrew C. Hofling2006 Apuntes sobre la variacion dialectal en el maya yucateco. Penın-

sula. volumen I, numero 1, primavera de 2006: 27–45.Bohnemeyer, Jurgen

2001 Argument and Event Structure in Yukatek Verb Classes. Univer-sity of Massachusetts Occasional Papers in Linguistics: the seman-tics of underrepresented languages in the Americas 25.8–19.

Bohnemeyer, Jurgen2004 Split intransitivity, linking, and lexical representation: the case of

Yukatek Maya. Linguistics 42(1): 67–107.Bohnemeyer, Jurgen

2009 Linking without grammatical relations in Yucatec: Alignment,extraction, and control. In Y. Nishina, Y. M. Shin, S. Skopeteas,E. Verhoeven, & J. Helmbrecht (eds.), Issues in functional-typological linguistics and language theory: A Festschrift forChristian Lehmann on the occasion of his 60th birthday. Berlin:Mouton de Gruyter. 185–214.

Branigan, Holly P., Martin J. Pickering, and Mikihiro Tanaka2008 Contribution of animacy to grammatical function assignment

and word order during production, Lingua, 118: 117–189, Elsevier.Bricker, Victoria R.

1978 Antipassive constructions in Yucatec Maya. Papers in MayanLinguistics, (University of Missouri Miscellaneous Publicationsin Anthropology N�6, Studies in Mayan Linguistics N�2 N.England (ed), Columbia, Missouri, 3–24.

Bricker, Victoria R.1979 WH-questions, relativisation and clefting in Yucatec Maya.

Papers in Mayan linguistics, L. Martin (ed.), Lucas Bas Brothers,Columbia, Missouri, 107–36.

Bricker, Victoria R.1981 The Source of the ergative split in Yucatec Maya. Journal of

Mayan Linguistics, vol. 2, n�2. Spring 1981.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Bricker, Victoria, Eleuterio Po’ot Yah and Ofelia Dzul de Po’ot1998 A Dictionary of the Maya language as spoken in Hocaba, Yucatan.

University of UTAH Press, Salt Lake City.Brown, Penelope

1997 Isolating the CVC root in Tzeltal Mayan: a study of children’sfirst verbs. In E.V. Clark, ed., Proceedings of the 28th AnnualChild Language Research Forum, 41–52. Stanford, CA: CSLI/University of Chicago Press.

Brown, Penelope1998 Early Tzeltal verbs: argument structure and argument repre-

sentation. In E.V. Clark, ed. Proceedings of the 29th AnnualChild Language Research Forum, 129–140. Stanford, CA: CSLI/University of Chicago Press.

Brown, Penelope and Stephen C. Levinson1987 Politeness. Some universals in language usage. Cambridge Uni-

versity Press, Cambridge.Comrie, Bernard

1981 Language universals and linguistic typology: Syntax and mor-phology. Chicago: University of Chicago Press.

Creissels, Denis2006 Syntaxe generale une introduction typologique 2 : la phrase, 2/2,

Paris, Hermes – Lavoisier Cuentos y relatos indıgenas.Creissels, Denis

1994 Ya’yejal an Sebastian sok ch’ul me’tik Veronika, 199–218. Mexico:Universidad Nacional Autonoma de Mexico.

Danziger, Eve1996 Split Intransitivity and Active-Inactive Patterning in Mopan.

International Journal of American Linguistics, 62(4): 379–414.Dayley, Jon

1981 Voice and ergativity in Mayan Languages. Journal of MayanLinguistics Linguistics 2: 6–82.

Dayley, Jon1990 Voz y ergatividad en idiomas mayas Lecturas sobre la linguıstica

maya, Nora England y Stephen Elliott eds, CIRMA: 335–399.Dryer, M.S.

1986 Primary Objects, Secondary Objects, and Antidative. Language62(4): 808–845.

England, Nora C.1983 Ergativity in Mamean (Mayan) Languages. International Journal

of American Linguistics 49: 1–19.England, Nora C.

1988 Mam Voice. In Passive and Voice. Ed. M. Shibatani. Amsterdam/Philadelphia: John Benjamins Publishing Company. 523–545.

England, Nora C.1991 Changes in Basic Word Order in Mayan Languages. Interna-

tional Journal of American Linguistics 57(4): 446–86.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Entzin Hernandez, C.2007 Kuxulon; Estoy vivo. In El bolom dice . . . venimos a este lugar.

Antologıa de cuentos, Centro Estatal de Lenguas, Arte y Litera-tura Indıgenas, Tuxtla Gutierrez, 13–20.

Fought, John G.1967 Chorti (Mayan): Phonology, Morphonemics, and Morphology.

Yale University, Department of Anthropology New Haven.Doctoral dissertation.

Fought, John G.1972 Chorti (Mayan) texts. Edited by Sara S. Fought, Philadelphia:

University of Pennsylvania Press.Givon, Talmy (ed.)

1983 Topic Continuity in Discourse: a Quantitative Cross-LanguageStudy. Amsterdam: John Benjamins.

Givon, Talmy (ed.)1994 Voice and Inversion. Amsterdam: John Benjamins.

Gomez Ramırez, Martın2001 Te amej, chanul xux sok te xanich’e; La arana, la avispa y la

hormiga. In Jich’ sk’op balam. Seleccion de cuentos tzeltales.Biblioteca Popular de Chiapas coneculta, Tuxtla Gutierrez,104–120.

Gomez Ramırez, Martın2007 Te jilinem a’teleletike; Los trabajadores incoformes. In El bolom

dice . . . venimos a este lugar. Antologıa de cuentos, Centro Esta-tal de Lenguas, Arte y Literatura Indıgenas, Tuxtla Gutierrez,68–85.

Guillaume, Antoine2006 The hierarchical agreement system with inverse e¤ect of Reyesano.

Laboratoire Dynamique du Langage, CNRS / University Lyon2. Ms. 31 pp.

Guillaume, Antoine2009 Hierarchical agreement and split intransitivity in Reyesano,

International Journal of American Linguistics, 75: 1, pp. 29–48.Gutierrez Bravo, Rodrigo y Jorge Monforte

2008 La alternancia sujeto inicial/verbo inicial y la teorıa de la optimi-dad. In Teorıa de la optimidad: estudios de sintaxis y fonologıa.Eds. Rodrigo Gutierrez Bravo y Esther Herrera Zendejas. ElColegio de Mexico-Centro de Estudios Linguısticos y Literarios-Laboratorio de Estudios Fonicos, Mexico, 2008, pp. 61–90.

Gutierrez Bravo, Rodrigo y Jorge Monforte2009 Focus, agent focus and relative clauses in Yucatec Maya. in New

Perspectives on Mayan Linguistics, H. Avelino, J. Coon, & E.Norcli¤e (eds.), MIT Working Papers in Linguistics.

Gutierrez Sanchez, Pedro and Roberto Zavala2005 Chol and Chontal: Two Mayan languages of the agentive type.

Paper presented at the conference The Typology of stative-activelanguages, Leipzig May 2005.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Haviland, John B.1981 Sk’op sotz’leb. El tzotzil de San Lorenzo Zinacantan. Universi-

dad Nacional Autonoma de Mexico, Mexico.Hinman Smith, Joshua

2004 Manual of Spoken Tzeltal, translated into English by Stuart P.Robinson, Madelief Bertens ed.

Kaufman, Terence1971 Tzeltal Phonology and Morphology. University of California

Publications in Linguistics, Berkeley, CA; 61. Berkeley and LosAngeles.

Kaufman, Terence1990 Algunos rasgos estructurales de los idiomas Mayances con refer-

encia especial al K’iche. In Nora England & Stephen R. Elliott(eds). Lecturas sobre la linguıstica maya, 59–114. La AntiguaGuatemala: CIRMA.

Kramer, Martin and Dierter Wunderlich1999 Transitivity alternations in Yucatec and the correlation between

aspect and argument roles. Linguistics 37–3, 431–479.Larsen, Tomas W., and William M. Norman

1979 Correlates of ergativity in Mayan grammar. In Frans Plank (ed.)Ergativity: Towards a theory of grammatical relations, 347–370.London/New York: Academic Press.

Lehmann, Christian1990 Yukatekisch. Zeitschrift fur Sprachwissenschaft 9(1/2): 28–51.

Lehmann, Christian1993 Predicate classes in Yucatec Maya. Funcion 13–14: 195–272.

Lehmann, Christian and Elisabeth Verhoeven2005 Extraversive transitivization in passive Maya and the nature of

the applicative. In L. Kulikov, A. Malchukov and P. de Swart(eds) Case, Valency and transitivity, 465–493. Benjamins, Am-sterdam.

Lois, Ximena and Valentina Vapnarsky2003 Polyvalence of lexical roots in Yukatekan Mayan Languages.

Lincom Europa, Munchen.Lois, Ximena and Valentina Vapnarsky

2006 Root indeterminacy and polyvalence in Mayan languages. Lexi-cal Categories and Root Classes in Amerindian languages. X.Lois and V. Vapnarsky (eds), Peter Lang, Bern, 69–115.

Lois, Ximena and Valentina Vapnarsky2009 Morphosyntactic and semantic issues on polycategoriality in

Yucatecan Mayan languages. Paper presented at the ALT8meeting, Berkeley, University of California. July 23–27 2009.

Lois, Ximena, Valentina Vapnarsky, Cedric Becquey et Aurore Monod Becquelin2010 Polycategoriality across Mayan languages: Action nouns and

ergative splits. Paper presented at the International Conference

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



on Polycategoriality. Cross-linguistic, cross-theoretical and lan-guage acquisition approaches. Ecole Normale Superieure, Paris,Oct. 2010.

Ola, Olanike, and Victoria Bricker2000 Placeless and historical laryngeals in Yucatec Maya. In Interna-

tional journal of American linguistics vol. 66, no. 3: 283–317.Perez Martınez, Vitalino

1994 Gramatica del idioma ch’ortı’. Antigua, Guatemala: ProyectoLinguıstico Francisco Marroquın.

Polian, Gilles1999 Morpho-syntaxe du verbe tzeltal. Memoire de DEA, Universite

Paris III.Polian, Gilles

2003 Sustantivos verbales e infinitivos en tzeltal. Quinto Congreso deEstudios Mayas. Guatemala Ciudad.

Polian, Gilles2005 Dinamica de la oracion tzeltal. La topicalidad como un factor

determinante del orden lineal. Trace 47: 30–45.Polian, Gilles

2006 Elements de Grammaire du tseltal, L’Harmattan to appear Gra-matica del tseltal. CIESAS, Mexico.

Quizar, Robin1994 Split Ergativity and Word Order in Ch’orti’, International Journal

of American Linguistics, Vol. 60, No. 2 (Apr., 1994): 120–138.Quizar, Robin and Susan M. Knowles-Berry

1988 Ergativity in the Cholan Langages. International Journal ofAmerican Linguistics 54, 1: 73–95.

Robinson, Stuart1999 Voice and obviation in Greater Tzeltalan. Australian National

University.Ruz, Mario Humberto (ed.)

1989 Las Lenguas del Chiapas Colonial, Universidad Autonoma deMexico, Mexico.

Sanchez Gomez F.J. et al.2003 Sk’op bats’il winiketik; Gramatica basica de la lengua Tzeltal.

Unidad de escritores Mayas-Zoques. San Cristobal Las Casas.Shibatani, Masayochi

1985 Passives and related constructions: a prototype analysis. Language,61, 4: 821–848.

Shklovsky, Kirill2005 Person marking in Petalcingo Tzeltal. Thesis, Reed College.

Skopeteas, Stavros, and Elisabeth Verhoeven2005 Postverbal argument order in Yucatec Maya. Sprachtypologie

und Universalienforschung 53(1), 71–79.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Slocum, Marianna C., Florence L. Gerdel and Manuel Cruz Aguilar1999 Diccionario tzeltal de Bachajon, Chiapas, Instituto Linguistico de

Verano, Mexico.Smith-Stark, Thomas

1978 The Mayan antipassive: some facts and fictions. Papers in MayanLinguistics, N. England ed. 169–187. University of Missouri Mis-cellaneous Publications in Anthropology and Studies in MayanLinguistics, Columbia, University of Missouri.

Stross, Brian1978 Tzeltal tales of Demons and Monsters. Museum of Anthropology,

University of Missouri-Columbia (Columbia), Museum brief;no. 24.

Thompson, Chad1994 Passive and inverse constructions. Voice and inversion, Talmy

Givon (ed.) Amsterdam: John Benjamins, 47–63.Tonhauser, Judith

In Press Agent Focus and Voice in Yucatec Maya. Proceedings of the39th Meeting of the Chiacgo Linguistic Society.

Van Valin, Robert D.1981 Grammatical Relations in Ergative Languages. Studies in Lan-

guage, 5:3.Van Valin, Robert D., and Randy J. La Polla

1997 Syntax: structure, meaning, and function. Cambridge: CambridgeUniversity Press.

Vapnarsky, Valentinato appear Le passif peut-il eclairer les esprits? Actance, interactions et

esprits-maıtres en maya yucateque, Ateliers d’Anthropologie(http://ateliers.revues.org).

Verhoeven, Elisabeth2007 Experiential constructions in Yucatec Maya. John Benjamins

Pub. Company, Studies in Language Companion series 87.Zavala Maldonado, Roberto

1997 Functional analysis of Akatek voice constructions. InternationalJournal of American Linguistics, 63, 4: 439–457.

Zavala Maldonado, Roberto2008 The agentive alignment pattern in Cholan and K’ichean (Maya).

Paper presented at the AALLED conference, Lyon, February2008.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The passives in Cavinena

Antoine Guillaume

1. Introduction1

This paper consists of a detailed analysis of two verbal su‰xes with passive

meanings in Cavinena, an ergative language from the Tacanan family spoken

in Amazonian Lowland Bolivia.

These markers, -tana and -ta, are interesting for a number of reasons.

First, it has been argued that passive constructions are rare – or even

absent – in Amazonia (Derbyshire 1987: 321, Payne 1990: 3, 2001: 596).

However, recent surveys like that by Siewierska (2005) in the World Atlas

of Language Structure do mention quite a few in the area. It is also some-

times thought that passives are normally features of accusative languages,

not of ergatives ones. Yet, as pointed out by authors like Lazard (1986: 9),

Dixon (1994: 149¤.), Keenan and Dryer (2007: 359), a fair number of

ergative languages are known to display passive constructions. This paper

contributes to the idea that passives might not be that exceptional in

Amazonia and in ergative languages in general.

Second, because of the lack of detailed studies on Tacanan languages,

the existence of passives in this family has so far not been very clear. For

example, in the WALS article mentioned above, Cavinena is said to lack

this derivation while Araona, a related Tacanan language, is said to have

1. The research presented in this paper is based, for the most part, on first handdata that I collected from Cavinena native speakers in traditional communitiesthrough 15 months of fieldwork between 1996 and 2003 (complemented bymaterial published by two SIL missionaries, E. Camp and M. Liccardi). Iwould like to thank the Cavinena people for their generous hospitality andinterest in documenting and describing their language. The ideas developedin this paper benefited from comments made by the participants of the work-shop on ‘‘Passif en contexte ergatif ’’ organized by F. Queixalos in Vuillejuifon May 25th, 2007, within the project ‘‘Ergativite ’’ of the Federation de Typo-logie et Universaux Linguistiques. The preparation of this article was also im-proved by comments from Denis Creissels, Rene Lacroix, Francesc Queixalosand Francoise Rose, as well as from an external reviewer (Anna Siewierska)and from the editors Gilles Authier and Katharina Haude.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:41) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 111–132 1351 Authier_03_guillaume (p. 111)

it which, to my knowledge, is not the case. This paper will serve to clarify

the situation.

Third, the distinction between the passive markers -ta and -tana is a

very slim one. In earlier work (Guillaume 2004, 2008), I treated -tana and

-ta as two free variants of the same morpheme. This paper is a major revi-

sion of this initial analysis. By looking more closely at their respective

semantics, distribution and use, I show that they di¤er in terms of produc-

tivity and ability to express or not anticausative semantics, and I claim

that -ta and -tana are distinct morphemes. The observations I make lead

me to propose several historical scenarios for the evolution of the two

markers.

The paper is organized as follows. In §2, I provide an overview of the

main characteristics of the Cavinena ergative argument-coding system.

Particular attention is paid to the sensitivity and rigidity of the system to

the category of transitivity, and to the fact that the ergative patterning is

restricted to the morphological coding domain, not extending to the level

of behavior-and-control properties, in other words that there is no ergative

‘‘pivot’’ in the language. In §3, I introduce and discuss the two passive

derivations, realized by the verbal su‰xes -tana and -ta, in terms of their

syntactic, semantic, and distributional properties. Section 4 takes a dia-

chronic perspective. Building on another study that traced back the origin

of the marker -ta as a third person plural marker (Guillaume forthcom-

ing), I here argue for the development of a productive marker -tana either

because of the loss of productivity of -ta or because of the semantic overlap

between the two.

2. The Cavinena argument-coding system

Cavinena is an Amazonian language from the Tacanan family spoken in

Northern Bolivia by approximately 1200 speakers who are also all bilin-

gual in Spanish. Cavinena encodes the grammatical functions of its core

arguments by way of case-marking and cross-referencing. Both systems

manifest ergative alignment.2 The transitive agent/subject (henceforth A)

NP is marked by the enclitic ¼ra, as with iba ‘jaguar’ in (1a). By contrast,

the transitive patient/object (henceforth O) NP is unmarked, as with takure

2. This section only provides a brief summary of the argument-coding system.For more details, see Guillaume (2006, 2008, Forthcoming-a).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


112 Antoine Guillaume

‘chicken’ in (1a), similar to the intransitive unique argument (henceforth S),

as with iba ‘jaguar’ in (1b).3

(1) a. Transitive clause

Iba¼ra ¼tu4 iye-chine takure.

jaguar¼erg ¼3sg.abs kill-rec.past chicken(abs)

‘The jaguar killed the chicken.’ (elicited)

b. Intransitive clause

[Tuke tupuju] ¼tu iba tsajaja-chine.

3sg following ¼3sg.abs jaguar(abs) run-rec.past

‘The jaguar ran behind him.’ (from text)

The ordering of the core arguments is free and does not play any role in

the disambiguation of grammatical functions. There is an argument cross-

referencing system realized by bound pronominal clitics in second position

in the clause (not on the verb). These clitics specify the person, the number

and the grammatical function of the core arguments in the same ergative

fashion as NPs and independent pronouns. Neither core argument NPs

and independent pronouns, nor bound pronouns are obligatory. Free and

bound forms can co-occur (representing the same argument), as in (2), or

only one of them can be present, i.e., the free form, as in (3), or the bound

form, as in (4). When referring to a third person singular, neither a free

nor bound form is obligatory, as shown in (5); with other person and

number combinations, however, there needs to be some expression of the

relevant argument.

(2) co-occurrence of free and bound forms

a. Ejetupu ¼tuke5 ¼mikwana e-a-u pista?

when ¼3sg.abs ¼2pl.erg pot-a¤ect-pot airstrip(abs)

‘When could you make (lit. a¤ect) an airstrip?’ (from text)

3. Although not illustrated here, the expression of core arguments by way ofindependent pronouns follows the same ergative patterning (see Guillaume2006, 2008, Forthcoming-a).

4. Enclitics preceded by a space are second position enclitics, as discussed furtherbelow.

5. Second position bound pronouns can be subject to a morpho-phonologicalrule that modifies their segmental make-up. For example, this rule is responsi-ble for the fact that the third person singular absolutive marker is ¼tu in(1a,b) while it is ¼tuke in (2a); see full details in Guillaume (2006, 2008,Forthcoming-a).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The passives in Cavinena 113

b. Jadya ¼tura ¼Ø a-kware bari¼ra.

thus ¼3sg.erg ¼1sg.abs a¤ect-rem.past giant_anteater¼erg

‘This is what the giant anteater did to (lit. a¤ected) me.’

(from text)

(3) only free forms

E-wane¼ke¼ra amena ba-ti-kware tuke.

3-wife¼3¼erg bm see-go-rem.past 3sg.abs

‘His wife went to see him.’ (from text)

(4) only bound forms

Enapa-wa ¼taa ¼tunara ¼ike.

cry_for-perf ¼emph ¼3pl.erg ¼1sg.abs

‘They (my dogs) cried for me.’ (from text)

(5) neither free nor bound form (only for third person singular)

Ka-bajeje-ti-tsu shana-nuka-kware.

ref/recip-prepare-ref/recip-ss leave-reitr-rem.past

‘He prepared himself and left it (a viper) again.’ (from text)

In Cavinena, the category of transitivity is entrenched at various levels

of structure, from lexicon to morphosyntax, at least at the intra-clausal

level. Starting with morphosyntax, the ergative marking on the A argu-

ment (and the absence of marking of the S and O arguments) is obligatory

in all (syntactic, semantic or pragmatic) contexts. Within the predicate,

several su‰xes have ‘‘transitivity harmony’’ restrictions, having two allo-

morphs, the selection of which depends on the transitivity of the verb

stem they attach to. This is the case, for example, with the completive

su‰x, which shows up as -tere on intransitive verb stems, as in (6a), and

-tirya on transitive verb stems, as in (6b) (Guillaume 2008: 191 ¤.).

(6) a. Tiru-tere-wa [ekwe budari ].

burn-comp.itr-perf 1sg.gen banana

‘My banana burnt completely.’ (elicited)

b. [Iyuka biti ] ¼tu yupu-tirya-kware iba¼ra.

head skin ¼3sg.erg tear_o¤-comp.tr-rem.past jaguar¼erg

‘The jaguar tore o¤ his whole scalp.’ (from text)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



At the lexicon level, the rule is for all verbal lexemes to be either intran-

sitive or transitive; I could only find two ambitransitive verbs (kike-/keke-6

‘shout, shout at O’ and kwina- ‘give birth, give birth to O’). One conse-

quence of this is that there is a functional need for this language to have

overt valency-changing mechanisms for playing with transitivity, which it

indeed has. There are two passive markers (to be discussed in detail further

below), a reflexive/reciprocal marker, two antipassive mechanisms (involv-

ing reduplication or the exchange of auxiliaries) and three causative markers

(see Guillaume 2008: ch. 8).

At the inter-clausal level, the transitivity distinction is somewhat more

relaxed. In particular, the co-reference restrictions that exist between certain

dependent clauses and their controlling matrix clause operate accusatively,

i.e., treating the S and the A identically and in opposition to the O. This is

exemplified in (7) with the temporal dependent clause, the verb of which

is marked by the su‰x -tsu and which requires its subject (S or A) to be

co-referential with the subject (S or A) of the matrix clause.

(7) a. Dependent S ¼Matrix S

Tudya ¼tatse amena [kwaba¼ju ani-bute-tsu]

then ¼3dl.abs bm canoe¼loc sit-go.down-ss

tsura-kware.

go_up-rem.past

‘Then they (dl) sat down in their canoe and went up(river).’

(from text)

b. Dependent A ¼Matrix A

Tudya ¼tuke ¼Ø [imeta-tsu] mare-kware.

then ¼3sg.abs ¼1sg.erg point_at-ss shoot_at-rem.past

‘Then I pointed (my rifle) at it (a peccary) and I shot at it.’

(from text)

c. Dependent S ¼Matrix A

[Babi¼ra kwa-atsu] ¼tuja ¼tu

hunt¼purp.mot go-ss ¼3sg.dat ¼3sg.erg

tsuru-kware [ peadya matuja ].

encounter-rem.past one cayman

‘When hei went hunting, hei met a caiman.’ (from text)

6. The alternate forms kike and keke are in free variation and have nothing to dowith transitivity.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



d. Dependent A ¼Matrix S

Bajida¼jipenee ¼Ø ju-kware [tuke peta-tsu].

scared¼almost ¼1sg.abs be-rem.past 3sg.abs look_at-ss

‘I was a bit scared, looking at it (something moving among the

manioc leaves).’ (from text)

Having laid down the main characteristics of the Cavinena argument-

coding system, we will now turn to the passive derivations in this language.

3. The passives -tana and -ta

Cavinena has two verbal markers with core passive meanings, -tana and

-ta. Apart from the fact that they are segmentally very close – they both

contain the syllable ta – their syntactic and semantic e¤ects are very simi-

lar. A semantic distinction between these two markers was postulated

by Camp and Liccardi in their grammatical sketch (1989: 315). However,

their lack of argumentation7 and my inability at the time to determine

any obvious di¤erences made me treat them as free variants of the same

morpheme in Guillaume (2004, 2008).

The goal of this paper is to revisit my preliminary analysis. By looking

more closely at the semantics, use and distribution of these su‰xes in

texts, I argue that -tana and -ta are distinct morphemes and that their

respective characteristics are as follows:8

Marker -tana:

– distribution: transitive verbs only

– syntactic e¤ect: reduction of the verb valency by one

– semantic e¤ect: (1) passive and (2) anticausative9

– discourse use: major use in depiction of generic customary practices

and procedures

– productivity: full

7. The authors say that with -tana the agent is ‘‘optional’’ whereas with -ta theagent is ‘‘unknown’’. However, since they do not provide any clear illustrativeexample or further discuss their claim, it is completely unclear what exactlythey had in mind.

8. With these properties, both markers fulfill the criteria for a ‘‘basic passive’’,according to Keenan (1985) and Keenan and Dryer (2007).

9. In this study, I use ‘‘passive’’ and ‘‘anticausative’’ in the following senses,following Creissels (2006: 9–10, 31, 35–37). Both are mechanisms that applyto a transitive verb and produce a derived intransitive form. In the case of a

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Marker -ta:

– distribution: identical to -tana

– syntactic e¤ect: identical to -tana

– semantic e¤ect: passive only

– discourse use: major use for specific events in narratives; not used for

depicting generic customary practices and procedures

– productivity: restricted

These characteristics are discussed, refined and exemplified in §3.1 (with

-tana) and §3.2 (with -ta), and summarized in §3.3.

3.1. The passive and anticausative -tana

Formally, -tana attaches to transitive verbs and reduces their valence by

one. The semantic correlates of this process are either passive or anticausa-

tive. When they are passive, the event involves and agent (notional A) act-

ing on a patient (notional O expressed as S), but the identity of the agent is

backgrounded, and the agent must be left unexpressed. This is illustrated

with a pair of elicited examples in (8) and a pair of text examples in (9).

(8) a. Active

Roberto¼ra akwi abu-ya.

Roberto¼erg tree(abs) carry-impfv

‘Roberto carries the tree.’ (elicited)

b. Passive

Akwi abu-tana-ya.

tree(abs) carry-pass1-impfv

‘The tree is carried.’ (elicited)

passive, the subject receives exactly the same semantic role as the object of thetransitive construction, the participation of the agent of the transitive con-struction in the event being always implied, whether it is overtly expressed(by an oblique) or not; e.g., ‘the eggs were broken (by the child)’. In the caseof an anticausative (decausatif in Creissels’ terms), the subject undergoes thesame process as the object of the transitive construction but unlike in a pas-sive, this process is interpreted as more or less spontaneous; e.g., ‘the eggsbroke’. In this sense, anticausative is one of the various possible meanings ofthe more general category ‘‘middle’’, which is characterized by the fact thatthe subject does not receive exactly the same role as the object of the transitiveconstruction – the other possible meanings of the category of ‘‘middle’’ arereflexive, reciprocal, autocausative and autobenefactive.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(9) a. Active

Irare¼kwita¼dya¼jutidya ¼tuna tsume-kware.

plaited_sieve¼restr¼foc¼disemph ¼3pl.erg use-rem.past

‘(In old times, our Cavinena ancestors) would only use plaited

sieves (for sieving corn).’ (from text)

b. Passive

Jadya¼tibu ¼tu [tumeke emajaka¼ju]

thus¼reason ¼3sg.abs that place¼loc

[ jeeke karetu] tsume-tana-ya. . .

this cart(abs) use-pass1-impfv

‘(In Cavinena communities there aren’t any motorized vehicles.)

For that reason, this cart is used in these places.’ (from text)

In anticipation to the discussion of the di¤erences between -tana and

-ta, it is worth mentioning that -tana in its passive sense can be used for

depicting both actual and generic events. By actual (or referential), I mean

actions with a specific agent acting on a specific patient at a particular

place and time. Textual examples of this use can be found in (10).

(10) a. Dutya ekana tibarirya-tana-wa.

all 3pl encircle-pass1-perf

‘They were (already) all encircled.’ (from text)

b. Ne-diru-ra¼dya! Tirya-tana-ya¼dya ¼ekwana.

hort.pl-go-hort.pl¼foc finish-pass1-impfv¼foc ¼1pl.abs

‘Let’s go! (Otherwise) we will all be killed (lit. finished).’

(from text)

c. A-tana-wa ¼ekwe [ekwe e-bui¼ekatse]!

a¤ect-pass1-perf ¼1sg.dat 1sg.gen 1-nephew¼dl(abs)

‘My two nephews got killed (lit. a¤ected)!’ (from text)

By generic events, I mean actions with non-specific/impersonal agents and

patients as used for the expression of customary practices and procedures.

A good example of this use can be found in (9) above, where the speaker

explains that traditional carts (rather that motorized vehicles) are normally

used to travel in the places where he lives. In (11) below, the speaker

explains the traditional way of making a canoe.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(11) a. Ikwene¼dya ¼tu e-duku¼ju rure-tana-ya.

first¼foc ¼3sg.abs npf-inside¼loc carve-pass1-impfv

‘First, (the canoe) is carved on the inside.’ (from text)

b. Amena tuekedya ¼tu chamakama e-pere

bm then ¼3sg.abs finally npf-side(abs)

a-tana-ya amena.

a¤ect-pass1-impfv bm

‘And then finally the (canoe’s) side is made (lit. a¤ected).’

(from text)

When the semantic interpretation of the application of -tana to a tran-

sitive verb is anticausative, the event is not carried out by any agent

(notional A) but instead by the patient (notional O expressed as S) itself

spontaneously. This is illustrated in (12).10

(12) a. [Ekwe karusune iyakwake utsa-wa¼ju]

1sg.gen pants new(abs) wash-perf¼ds

dyuru-tana-chine.

shorten-pass1-rec.past

‘(The women) washed my new pants and they shrunk.’ (elicited)

b. Amena ¼tuke ¼ekwe arusu

bm ¼3sg.abs ¼1sg.dat rice(abs)

puku-tana-bare-ya. . .

crack-pass1-distr-impfv

‘My rice is cracking open.’ (Camp and Liccardi 1989: 20)

It is unclear to what extent the distinction between the passive or anti-

causative interpretations is linked to the semantic type of verbs. In the

available corpus, I have not found any clear example of a verb showing

both readings. However, I would assume that -tana can potentially have

a passive reading with any transitive verb and that with some of these

verbs it can also have an anticausative interpretation, as long as they

express events that can occur spontaneously. More work is needed on

this topic to confirm this hypothesis.

10. Note that -tana does not yield autocausative interpretations. These areexpressed by the reflexive/recriprocal marker k(a)- . . . -ti (ex. utsa- ‘wash’,k-utsa-ti- ‘wash oneself ’); see Guillaume (2008: 270).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The su‰x -tana is a clear detransitivizer. There are several pieces of

evidence for this claim. First, as I said, the agent (notional A) cannot be

overtly expressed: there is no single passive example in the whole corpus

where this happens and in elicitation, my attempts at expressing the

notional A (by way of di¤erent kinds of oblique phrases) were systemati-

cally rejected. Second, passivized verbs can only bear the intransitive

allomorph of those markers which harmonize with the transitivity of the

verb stem they attach to; see the discussion of this phenomenon in §2.

Accordingly, a passivized verb can only select the intransitive completive

allomorph -tere, as in (13); the transitive completive allomorph -tirya

would be ungrammatical.

(13) Dutya ekana iye-tana-tere-wa.

all 3pl kill-pass1-comp.itr-perf

‘They were all killed, to the last one.’ (from text)

Third, at least when the interpretation of -tana is anticausative, there is

evidence that the notional O (expressed as S) has behaviour-and-control

subject properties. In particular, in sentences involving clauses with subject-

to-subject restrictions, the notional O of the verb marked by -tana counts

as co-referential with the subject of the other clause, as shown in (14).11

(14) a. [Ebakwapiji riwi-tsu] tuku rake-tana-wa.

child(abs) fall-ss onom break-pass1-perf

‘The child fell, ‘‘tuku!’’, and (some part of his body) broke.’

(elicited)

b. . . . kunu¼ju [ekwe e-wachi¼ekatse] karya-tana-tsuliana¼loc 1sg.gen npf-foot¼dl(abs) hook-pass1-ss

ike riwi-kware.

1sg.abs fall-rem.past

‘(As I was running after the coati,) my feet got caught

(lit. hooked) into lianas and I fell down.’ (from text)

11. In some cases, co-reference is between a part and its whole, in which case onemight say that co-reference is not strict. However, this type of co-referenceis possible between non-passivized clauses, so this is not a problem for thepresent claim.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



c. [Estakilla wika-tana-tsu] ¼tu pakaka-ya

wooden_peg(abs) extract-pass1-ss ¼3sg.abs fall-impfv

karetu.

cart(abs)

‘When the peg (of the cart) went out (lit. extracted), the cart fell

down.’ (Camp and Liccardi 1989: 56)

Before turning to -ta, let us add that -tana is probably fully productive.

This is suggested by the fact that in texts it is used fairly frequently and

with a wide range of transitive verbs, and in elicitation, its use was never

considered ungrammatical.

3.2. The passive -ta

Similar to -tana, -ta applies to transitive verbs12 and reduces the valency

by one. The semantic correlates are exclusively passive. As will be dis-

cussed further below, unlike -tana, -ta is not a productive marker.

(15) a. A-ta-nuka¼dya ekwana.

a¤ect-pass2-reitr¼foc 1pl.abs

‘We will be attacked (lit. a¤ected).’ (from text)

b. Jadya tirya-ta-wa¼ju. . .

thus finish-pass2-perf-ds

‘Having been exterminated (lit. finished) that way (they decided

to go live somewhere else).’ (from text)

There are tests that indicate that the construction is intransitive. First,

it is impossible to express the notional A, whether by an ergative NP, as in

the b-example in (16), or by an oblique NP.

(16) a. [Peadya senora] ¼tu kweja-ta-ya.

one woman ¼3sg.abs inform-pass2-impfv

‘A woman is being informed.’ (elicited)

b. *[Peadya senora¼ra] ¼tu kweja-ta-ya.

one woman¼erg ¼3sg.abs inform-pass2-impfv

(he was informed by a woman) (elicited)

12. The su‰x -ta is found on one intransitive verb, maju- ‘die’, in which it has anindefinite meaning ‘someone’; see Guillaume (Forthcoming-b).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Second, if the completive su‰x is used on a verb marked by -ta, it is the

intransitive allomorph, -tere, that shows up, not the transitive one, -tiriya

(see §2), as illustrated in (17).

(17) Dutya ekana iye-ta-tere-wa.

all(abs) 3pl(abs) kill-pass2-comp.itr-parf

‘They were all killed’ (elicited)

Regarding subject properties, I do not have enough material to be able

to claim with confidence whether the notional O has behavior-and-control

subject properties or not. In the only example available of a verb marked by

-ta in a sentence with obligatory subject-to-subject co-reference between its

clauses, reproduced in (18), the co-referential argument is the notional A

of the passive, not the notional O.

(18) [[Ekwanaja e-bakani¼kwana] waraji¼kwana¼keja duju-tsu]

1pl.gen npf-name¼pl(abs) authority¼pl¼loc.gnl take-ss

ekwana tsume-ta-ya.

1pl.abs use-pass2-impfv

‘(The land owners) give (lit. take) our names to the authorities and

use us.’ (spontaneous, given during elicitation session)

This example suggests that the notional O does not have subject behavior-

and-control properties. More examples of this type are needed, in particular

examples from texts, in order to know with certainty that this is indeed the

case.

Semantically, -ta seems to be always used in a passive sense; that is,

unlike -tana, it is never found with anticausative meanings.13 In addition,

in all the textual examples available, verbs marked by -ta only refer to

actual (referential) events, that is, actions with a specific agent acting

on a specific patient at a particular place and time. Generic events, with

non-specific/impersonal agents and patients cannot normally be expressed

by a verb marked by -ta,14 as shown by the ungrammaticality of the

b-example in the following pair:15

13. Note that no elicitation was conducted on this issue.14. Unless it is adjectivized with the resultative prefix e-; see further below.15. It is unclear whether this example would be acceptable with a di¤erent context,

like ‘a fan is made/will be made of chonta palm (by someone in particular)’. Myimpression is that it would not, since the collocation of -ta on a- ‘a¤ect’ onlyresults in the meaning ‘kill, destroy’ in the data.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(19) a. Epiki ¼tu a-tana-ya abari16.

fan(abs) ¼3sg.abs a¤ect-pass1-impfv chonta_palm

‘Fans are (traditionally) made (a¤ected) of chonta palm.’

(elicited)

b. *Epiki ¼tu a-ta-ya abari.

fan(abs) ¼3sg.abs a¤ect-pass2-impfv chonta_palm

(elicited)

In other words, semantically, the two markers only overlap in their

capacity to express actual passive events. The su‰x -tana has a much

broader semantic scope than -ta in its possibility to express situations with

no agent (anticausative) and generic events.

Another major di¤erence between -tana and -ta concerns their respec-

tive productivity. Unlike -tana, -ta manifests many symptoms of an non-

productive marker. In the available texts, it is only found very infrequently

and only with a very limited number of roots, actually limited to a- ‘a¤ect’

and tirya- ‘finish, terminate’, as in (15) above. While testing -tana and -ta

on other verbs, speakers are sometimes reluctant to use -ta with some

items while this does not happen with -tana.17 For example, with iye-

‘kill’, one speaker said that he had heard some people use iye-ta, as in

(17), but he himself preferred the use of -tana here. Another speaker also

had a few problems with iye-ta, until he said that it could be correct in a

figurative expression where the ‘killer’ was for example a disease (while with

-tana it could be a person or an animal). Similarly, with the verb ara ‘eat’

(tr.), the first speaker only accepted it with -tana at first, as in (20a) and

(20b). When asked to reflect on the ungrammaticality of -ta with this verb,

he came up with a context where this could nevertheless work, namely a

figurative expression where ara-ta ‘eat-pass’ refers to a moon eclipse, a

phenomenon that can be explained by reference to their traditional belief

that the moon is eaten by piranhas. The example is provided in (20c).

(20) a. Kashi ¼tuke ¼mikwe18 ara-tana-ya.

banana(abs) ¼3sg.abs ¼2sg.dat eat-pass1-impfv

‘Your bananas are eaten (by some pest, in your plantation).’

(elicited)

16. abari is an exceptionally unmarked optional (oblique) argument specifying thematerial used during the process of making something (a fan in this case).

17. Note that I did not conduct any systematic study.18. The dative enclitic ¼mikwe has the meaning of an ethical dative here, i.e., the

second person is the participant that is the most a¤ected by the event.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



b. *Kashi ¼tuke ¼mikwe ara-ta-ya.

banana(abs) ¼3sg.abs ¼2sg.dat eat-pass2-impfv

c. Badi ¼tu ara-ta-ya.

moon(abs) ¼3sg.abs eat-pass2-impfv

‘There is going to be a moon eclipse (lit. the moon is going to

be eaten).’ (elicited)19

What seems to happen, then, is that -ta, unlike -tana, cannot be applied

in an ad hoc way for depicting any transitive situation that can be described

with a passive in a language like Spanish (as in elicitation) or English. The

su‰x -tana, on the other hand, does not appear to have such restrictions.

For example, the expression referring to a moon eclipse can also be

expressed by ara-tana, as in (21), an example from Camp and Liccardi’s

(1989) dictionary.

(21) Badi ¼tu ara-tana-wa.

moon(abs) ¼3sg.abs eat-pass1-perf

‘The moon desappears in the eclipse (lit. the moon is eaten).’

(Camp and Liccardi 1989: 8)20

Another di¤erence in productivity between -tana and -ta can be found

in their distribution and/or semantics in certain derivational processes. In

particular, Cavinena has a productive resultative prefix e- that turns a

verb into a predicative adjective (see full details in Guillaume 2008: 397 ¤.).

Such adjectives normally express a state that results from the occurrence of

the verb event over the participant expressed as the S argument (if intransi-

tive) or the O argument (if transitive), as in (22a) and (22b), respectively.

(22) a. . . . e-rara e-jaki¼kwana ju-kware.

res-dry npf-leaf¼pl(abs) be-rem.past

‘. . . the leaves were dry/had dried/were in a dried state.’

(from text)

b. E-muya ¼tu ju-wa upati¼kwana.

res-scare_away ¼3sg.abs be-perf animal¼pl(abs)

‘The animals are scared away/have been scared away/are in the

state of being scared away (so they won’t show up for quite

some time now).’ (Camp and Liccardi 1989: 41)

19. The translation given in Spanish is ‘La luna se va eclipsar’.20. The translation given in Spanish is ‘La luna desaparece en la eclipse’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



When a verb derived by -tana is adjectivized by e-, the meaning is

resultative, as expected (i.e., as if applied to any intransitive verb, like

rara- ‘dry’ above).21

(23) E-penune-tana ¼tu akwi.

res-twist-pass1 ¼3sg.abs tree(abs)

‘The tree is twisted/has been twisted/is in a twisted state.’

(Camp and Liccardi 1989: 44)

However, when the prefix e- is applied to a verb passivized by -ta, the

resulting meaning is not resultative/stative but generic, as in (24).

(24) a. [Kwanubi jasa] ¼tu e-ara-ta.

animal lung(abs) ¼3sg.abs res-eat-pass2

‘(In our culture) the lung of animals is eaten/edible/we eat the

lung of animals/one eats the lung of animals (*is eaten/has been

eaten/is in the state of being eaten).’ (elicited)

b. Epiki ¼tu e-a-ta abari.

fan(abs) ¼3sg.abs res-a¤ect-pass2 chonta_palm

‘Fans are (traditionally) made (lit. a¤ected) of chonta palm.’

(*in the state of being made of chonta palm).’ (elicited)

This observation can probably be linked to the lack of productivity of

-ta and the concurrent tendency for it to manifest idiosyncratic aspectual

and modal meanings. In co-locations with e-, the resulting meaning is not

the one expected but one that expresses genericity rather than resultativity/

stativity.

3.3. Summary

Before turning to section §4, where I put the two detransitivizing markers

in a diachronic perspective, let us sum up the observations made above:

– -tana and -ta both apply to transitive verbs and detransitivize them,

with the syntactic e¤ect that the notional A cannot be overtly expressed;

– -tana is fully productive; -ta has limited productivity;

– -tana and -ta both have passive meanings. They are equivalent in that

function;

– -tana has functions that -ta lacks; it is used to express generic passive

and anticausative;

– -tana and -ta are segmentally very similar.

21. Note that the copula verb of copula clauses is not obligatory in Cavinena.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In the present study, I hypothesize that, even though there are very

strong formal and semantic similarities between the two markers, -tana

and -ta are di¤erent morphemes. The question I now attempt to address

is: How did those two markers arise diachronically?

4. A historical perspective

The low productivity of -ta suggests that this morpheme is an old mor-

pheme that is falling out of use in the language. This idea is reinforced by

comparative data. In Guillaume (forthcoming) I argue that cognates of -ta

are present in all the remaining Tacanan languages. In those languages,

-ta is not a passive marker but a cross-reference marker for a third person

plural S in an intransitive clause and a third person number-neutral A.

This is illustrated with data from Reyesano, a sister language of Cavinena,

in (25) and (26).22

(25) Reyesano intransitive clauses

a. 3SG b. 3PL

a-puti-a a-puti-ta(-a)23

past-go-past past-go-3s.pl-past

‘he went’ ‘they went’

(26) Reyesano transitive clauses

a. 3! 3 b. 3! 1

a-ba-ta(-a) m-a-ba-ta(-a)

past-see-3a-past

‘he/they saw him/them’

1sg-past-see-3a-past

‘he/they saw me’

These observations led me to propose that the source of all -ta su‰xes

in proto- (or pre-proto-) Tacanan is a third person plural marker in both

intransitive and transitive clauses. In Cavinena, with transitive verbs, this

marker would have yielded the present-day passive marker -ta through an

intermediate stage where it would have referred to an indefinite A argument

(like the indefinite use of the pronoun they in English); this grammaticaliza-

tion path is well attested (Haspelmath 1990, Heine and Kuteva 2002).

22. See Guillaume (2009) for a detailed description of the Reyesano person-marking system.

23. In the Reyesano examples, the morphemes between brackets do not show upon the surface.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Viewed from the perspective of the findings, the postulated third person

plural origin of the su‰x -ta is in line with the fact that -ta is restricted to

passive meanings, and not to anticausative ones.

The origin of -tana is more problematic. Its high productivity suggests

that it is more recent. This idea is corroborated by the fact that of all

Tacanan languages, it appears to be found only in Cavinena.24 Its formal

similarity with -ta suggests that they are historically related in a way that

-tana could be made up of -ta plus a su‰x -na. Reasoning in functional

terms, one could imagine that the loss of productivity of -ta created

a functional need for a productive passive and that -tana appeared as a

consequence via a process of reinforcement,25 that is by way of adding

morphological material to -ta – in that case a syllable na – in order to

reactivate the original passive meaning. This would explain why both

markers are so similar formally and semantically. The question remains

however of the origin of na. A possible candidate is the motion su‰x -na

‘come’26 but I have no evidence to prove it. Another potential problem

with this view concerns the presence of anticausative meanings for -tana.

If -tana is a reinforced passive marker, where do its anticausative mean-

ings come from? We know from the predictions of grammaticalization

theory that passive morphemes can arise from anticausative markers, not

the other way around (cf. Haspelmath 1990; Creissels 2006: 39–40). In

other words, it is unlikely that the anticausative meaning appeared as

a natural semantic extension of -tana. A possibility is language contact,

in particular with Spanish which does display a passive-anticausative

polysemy. Indeed, Cavinena speakers are bilingual in Spanish and have

been so for a fairly long time, having been living in a (Franciscan) mission

for more than two centuries (Guillaume 2008: 5–7).

Alternatively, -tana could have had a di¤erent origin from -ta. In that

case, one could think of two distinct scenarios. In one scenario, -tana

could have started as an anticausative and later developed the passive

meaning, either in order to fill the gap left by the disappearing passive

-ta, or under the influence from Spanish. In the other scenario, -tana could

24. Emkow (2006: 558) reports a middle su‰x -tana in Araona. However, there istoo little exemplification and discussion in her study to ascertain the existenceof such a marker in that language.

25. I thank Gilles Authier for this suggestion.26. This su‰x is used to associate a motion component to a verb stem event. The

motion is directed towards the speaker (ex. ara- ‘eat’, ara-na- ‘come and eat’);see full details in Guillaume (2008: 212¤ ).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



have been a passive and anticausative marker right from the beginning. In

that situation, one could imagine that -tana would be in the process of

replacing -ta because of the semantic overlap between the two markers

and because -tana has much broader meanings that -ta, which make it

more broadly applicable. An argument in favour of this hypothesis is the

fact that it is not clear why a passive marker would lose its productivity if

nothing is there from the start to replace it.27

With all the above scenarios we are still faced with the problem of the

origin of -tana, whether it is related to -ta or it has a completely distinct

origin, in which case the resemblance between the two markers is accidental.

At the present time this issue cannot be answered.

5. Conclusions

In this paper I have argued that the two detransitivizing markers -tana and

-ta in Cavinena are distinct morphemes, even though they are very similar

in their segmental make-up, their syntactic properties and their semantics.

I have brought to light a number of di¤erences, in particular in terms of

productivity and semantic extensions. The second part of the paper was

an attempt at proposing a historical scenario for their respective develop-

ment. I argued that -ta is an old morpheme (originally a third person plural

marker) that is losing ground, that -tana is younger and is replacing -ta,

and that the replacement of -ta by -tana is probably linked in one way or

another to the fact that there is a semantic overlap between the two.

Abbreviations

1 first person

2 second person

3 third person

a transitive subject

asf adjective su‰x

assoc associative

bm (syntactic) boundary marker

comp completive

dat dative

27. I thank Francesc Queixalos for this observation and the suggestion that thedisappearance of -ta could be due to a competition with -tana.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



disemph disemphatic

distr distributive

dl/dl dual

ds di¤erent subject

emph emphatic

erg ergative

fm formative

foc focus

gen genitive

hort hortative

imp imperative

impfv imperfective

int interrogative

itr./ itr intransitive

juss jussive

lig ligature

loc locative

loc.approx locative approximative

loc.gnl general locative

neg negative

npf nounprefix

nsg non-singular

o object

onom onomatopoeia

pass passive

perf perfect

perl perlative

pl/pl plural

pot potential

purp.mot purpose of motion

quest question (marker)

rec.past recent past

redup reduplication

ref reflexive

reitr reiterative

rem.past remote past

rep reportative

res resultative

restr restrictive

s intransitive subject

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



sg singular

simlr similarity

ss same subject

strg.emph strong emphasis

tr./tr transitive

uncert uncertain

References

Camp, Elizabeth, L. and Millicent R. Liccardi1989 Diccionario Cavinena-Castellano Castellano-Cavinena con bosquejo

de la gramatica Cavinena, Dallas: Summer Institute of Linguistics.Creissels, Denis

2006 Syntaxe generale: un introduction typologique. Vol. 2. Paris:Lavoisier.

Derbyshire, Desmond C.1987 Morphosyntactic areal characteristics of Amazonian languages.

International Journal of American Linguistics, 53 (3): 311–326.Dixon, R.M.W.

1994 Ergativity. Cambridge: Cambridge University Press.Emkow, Carola

2006 A Grammar of Araona. Ph.D. diss., Research Centre for Linguis-tic Typology, La Trobe University.

Guillaume, Antoine2004 A Grammar of Cavinena, an Amazonian Language of Northern

Bolivia. PhD. diss., Research Centre for Linguistic Typology, LaTrobe University.

Guillaume, Antoine2006 Revisiting ‘split ergativity’ in Cavinena. International Journal of

American Linguistics. 72 (2): 159–192.Guillaume, Antoine

2008 A Grammar of Cavinena. (Mouton Grammar Library no. 44.)Berlin and New York: Mouton de Gruyter.

Guillaume, Antoine2009 ‘‘Hierarchical agreement and split intransitivity in Reyesano.’’

International Journal of American Linguistics. 75 (1): 29–48.Guillaume, Antoine

2010 ‘‘How ergative is Cavinena?’’ In Ergativity in Amazonia, SpikeGildea and Francesc Queixalos (eds.), 97–120 [TypologicalStudies in Language, 98]. Amsterdam/Philadelphia: Benjamins.

Guillaume, AntoineForthc. ‘‘Third person agreement and passive marking in Tacanan lan-

guages: a historical perspective.’’ International Journal of AmericanLinguistics.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Haspelmath, Martin1990 The grammaticalization of passive morphology. Studies in Lan-

guage 14 (1): 25–72.Heine, Bernd and Tania Kuteva

2002 World Lexicon of Grammaticalization. Cambridge: CambridgeUniversity Press.

Keenan, Edward L.1985 Passive in the World’s languages. In Language typology and

syntactic description, Vol. 1, Tim Shopen (ed.), 243–281. Cam-bridge: Cambridge University Press.

Keenan, Edward L. and Matthew S. Dryer2007 Passive in the World’s Languages. In Language Typology and

Syntactic Description, Vol. 1, Timothy Shopen (ed.). Second Edi-tion. 325–361. Cambridge University Press.

Lazard, Gilbert1986 Formes et fonctions du passif et de l’antipassif. Actances 2: 3–57.

Payne, Doris L.1990 Introduction. In Amazonian Linguistics. Studies in Lowland South

American Languages. Doris L. Payne (ed.), 1–10. Austin: Univer-sity of Texas Press.

Payne, Doris L.2001 Review of The Amazonian languages, R.M.W. Dixon and

Alexandra Aikhenvald (eds.), 1999, Cambridge: Cambridge Uni-versity Press. Language 77 (3): 594–598.

Pitman, Donald1980 Bosquejo de la gramatica Araona. (Notas Linguısticas No. 9.)

Riberalta: Instituto Linguıstico de Verano.Siewierska, Anna

2005 Passive constructions, In World Atlas of Language Structure.Martin Haspelmath, Matthew S. Dryer, David Gil and BernardComrie (eds.), chapter 107: 434–437. Oxford: Oxford UniversityPress.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The detransitive voice in Kryz

Gilles Authier

1. Introduction

This paper examines the detransitive voice in Kryz, an unwritten language

belonging to the Lezgic branch of the North-East Caucasian family1.

Nowadays three dialects of Kryz (Kryz proper, Jek, and Alik-Khaput)

are spoken as a first language by at most 2000 speakers, in fewer than ten

localities of north-eastern Azerbaijan. Despite generalized bilingualism in

Azeri, Kryz preserves typical Proto-Lezgic features. In particular, gender-

number agreement with S/P (Single argument or Patient) nouns is pre- or

infixed to the lexical stems of synthetic verbs2, which form a closed class.

Person is expressed by free pronouns. Word order is strictly possessor-

possessed, adjective-noun, and basically Agent-Patient-Verb; case marking

on nouns and cross-referencing on the verb is ergative. Valency increase

is expressed periphrastically by means of auxiliaries (‘do’ or ‘give’). The

Kryz synthetic detransitive voice, to be described in this paper, is unique

within Lezgic3 and the East Caucasian language family as a whole.

While East Caucasian languages have long been believed to lack a

passive construction altogether, it is beyond doubt that in Kryz there

exists a detransitive voice with a prominent passive reading, alongside

1. This article is based on first-hand data collected by the author in Azerbaijanover the last ten years. Field trips were funded by the Centre National de laRecherche Scientifique (UMR 7192) and the Institut National des Langues etCivilisations Orientales. My host in the village of Alik and main informant forthis dialect is Madjlis Shamseddinov, to whom I am indebted more thanwords can say.

2. Verbal predicates are also represented by compound verbs, which are notaddressed here. They switch valency by changing their auxiliary (‘do’ /‘become’): this valency-changing device is equipollent and belongs to syntax,not derivation.

3. This branch comprises the core languages Lezgian, Tabassaran, Aghul, Tsakhur,Rutul, Kryz, and Budugh, all located close to the watershed line of the eastern-most part of the Great Caucasus range, and the outliers Udi (to the south)and Archi (to the north).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:41) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 133–164 1351 Authier_04_authier (p. 133)

various other potential values (anticausative and antipassive, as well as

aspectual and modal nuances). The use of this voice is restricted by the

semantic parameters and lexical properties of individual verbs, which

accounts for the fact that the few previous studies on Kryz (principally

Saadiev 1994) make no mention of any detransitivizing derivation. Indeed,

the Kryz detransitive voice is rarely heard in natural speech: it is mostly

employed in prescriptive discourse and technical instructions. But the deri-

vational morphology is old and can be compared with similar detransitive

forms found in some other East Caucasian languages only distantly related

to Kryz.

Section 2 of this articler gives an overview of grammatical relations and

the pragmatically marked order of constituents in non-derived predication.

Section 3 presents the morphology of detransitive forms. Section 4 presents

the lexico-semantic restrictions which operate on the passive voice. Section

5 describes the particulars of the aspectual and modal values associated

with the anticausative and antipassive forms. Section 6 proposes compara-

tive material and a hypothesis regarding the origin of this rare and residual

voice phenomenon in East Caucasian.4

2. Coding of syntactic and pragmatic functions

In Kryz, as in most East Caucasian languages, basic grammatical relations

in declarative clauses are organized according to an ergative/absolutive

case marking system with gender-number cross-referencing of the Single

or Patientive core argument on the verb. The details are described in the

following subsections.

2.1. Case marking

Kryz has ergative case marking on nouns. Nouns expressing S and P are

in the unmarked ‘absolutive’ case, while the Agent noun5 of a transitive

4. The new Azeri alphabet is used. The diphthongs /a-u/ /e-u/ and /i-u/ are real-ized as [o], [ø] and [y] in closed syllables; the sounds [q:], [g] and [g] (uvularfricative) are in complementary distribution and written <g>. The digraphs<xh> and <gh> represent voiceless and voiced velar fricatives respectively.

5. The pronominal system operates on a di¤erent basis from the nominal system.Speech Act Participants are expressed by free or cliticized pronouns with noergative marking. There is also no specialized morpheme on verbs headingreflexive and reciprocal constructions, which make use of special reflexiveand reciprocal pronouns. Peculiarities of case marking on Kryz pronouns aredetailed in Authier (2008).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


134 Gilles Authier

predicate is marked with ergative case (-r or one of its allomorphs), as

illustrated in ex. (1):

(1) a. q’al-ar-ir yuva-yar guzay.c-a gvat’ats’-re

mouse-pl-erg nest-pl north-in dig-prs(npl)

‘Mice dig their nests to the north.’

b. q’al-ar yuva-yar-a girxhar-e

mouse-pl(abs) nest-pl-in dwell-prs(npl)

‘Mice dwell in nests.’

Note that in contrast with many other East Caucasian languages, in

Kryz the ergative case is never used to mark semantic roles other than

Agent and two ergative-marked nouns with di¤erent semantic roles cannot

coexist in the same clause. But ergative-marked Agents can be inanimate,

as in examples (2) and following, and while the basic order of constituents

is S/A-(P)-V, non-prototypical Agents are often postposed to a more topical

Patient, as in (3) and (4):

(2) naq’ �a-xhr-i duli.c-ir xayla ziyan vu-yic

yesterday pv-come.pf-part storm-erg much damage give-aor.n

‘Yesterday’s hailstorm has caused a lot of damage.’

(3) yagin jin gara.c-ir basmis b-ar-ca-b

sure 1plexcl illusion-erg pressing hpl-do-prf-hpl

‘We are certainly victims of an illusion.’

(4) kaha-c siy babrux-ci xal.id-ir yi-qir-ca

cave-gen mouth spider-gen roof-erg pv-catch-prf

‘A cobweb covers the entrance to the cave.’

2.2. Indexation of S/P

Kryz verbs have the gender and number6 of the Patient or Single argument

cross-referenced in a pre-root (post-preverbal) slot, as shown in ex. (5):

(5) a. buba-r ris-imbi yi-b-qir-cib

father-erg daughter-pl pv-hpl-catch-aor.hpl

‘The father has caught the daughters.’

6. The gender-number system comprises five ‘gender-number agreement classes’:human masculine (M), human feminine and other animates (F, which alsoincludes many inanimates and certain abstract concepts), other inanimate (Nfor ‘neuter’), non-human plural (NPL), and human plural (HPL).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The detransitive voice in Kryz 135

b. ris-imbi ga-b-qil-cib

daughter-pl pv-hpl-lie-aor.hpl

‘The daughters went to sleep.’

Nominatively aligned indexation of S/A persons is also present in the

imperative paradigm only. The overall morphological structure of finite

verb forms is summed up in table 1.

Table 1. Structure of synthetic verb forms in Kryz

-3 preverb (usually with transparent spatial meaning)

-2 prefixes cross-referencing the gender-number of S/P

-1 aspect-marking sonorant (-r, -l, -n)

0 root consonant

1 perfective aspect or detransitivity markers

2 Tense & Mood su‰xes

3 su‰xes cross-referencing the gender-number of S/P (S/A in imperatives)

2.3. Verb classes according to valency

Verbs in Kryz fall into clear semantic classes, which are defined by their

specific morphological and syntactic properties, especially valency changing

properties: strict intransitive, extended intransitive (a¤ective and sensorial

predicates), strict transitive, and ditransitive. Lability (ambitransitivity) is

a very marginal phenomenon.

Transitive verbs have in their valency an Agentive argument which, if

expressed, is in the ergative case; these all permit valency increase of the

causative type by means of the auxiliary ‘give’, including those which are

ditransitive like ‘give’ itself. Most but not all transitive verbs allow syn-

thetic (morphological) detransitive derivation, a restriction which we find

to have a clear semantic motivation, to be described below.

Subclasses are also observable within the class of intransitive verbs –

for instance, most but not all of these have synthetic imperative forms,

nor are they all able to undergo periphrastic causative derivation with the

auxiliary ‘do’. But all are characterized by a stem ending in -aR.

2.4. Order of arguments and Patient topicalization

Agents are usually topics, and their unmarked position is clause-initial;

non-topicalized Patients come after the Agent and before the verb. But, as

is common for languages which express grammatical relations by means

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


136 Gilles Authier

of case-marking, the order of arguments can be modified for pragmatic

purposes. Focused constituents tend to occur in preverbal position, with

heavy stress, and topicalization of a Patient in the absolutive does not

require the use of a special form of the verb. Kryz allows the Patient to

be topicalized simply through alteration of the word order. In this ‘‘func-

tional passive’’ (Givon 2004) the ergative-marked Agent can remain, back-

grounded, in the construction:

(6) u-c seksenbes san a-n-ir vul cixe� ya-u-qur-cu

3-gen.nh eighty_five year 3-h-erg ewe behind pv-f-spend-prf.f

‘[He lived for a hundred years,] eighty-five of which he spent tending

the sheep.’

Non-prototypical Agents are often moved into non-initial position. The

most natural translation in English is a passive construction:

(7) a-c hicvaxt sel.c-ir tu-gats’-da-d

3-n never flood-erg pv-carry-neg.evt-n

‘This (bridge) will never be carried away by the flood.’

Note that with appropriate intonation on the preverbal position, the

Agent can be focused instead of backgrounded. It then translates not as a

passive, but as a cleft sentence:

(8) xinib fura-r v-a-u-q-ryu, nisi maltal.ci-r

woman man-erg f-pv-f-keep-prs.f cheese cheese.bag-erg

‘The woman is kept by the MAN, as cheese (is kept) by the

CHEESE-BAG.’

(9) ceuhur-ci ghala-d mesa.c-a sar.id-ir ulats’-ru

pear-gen good-notn forest-in bear-erg eat-evt.f

‘It is the BEAR who will get to eat the best pears in the forest.’

(10) ts’e�.il-kar azar-i yux yiggacig ebil-ir girats’-re-ni

goat-subel milk.mp-part milk every.day wolf-erg drink-prs-past

‘The milk drawn from the goat every day, the WOLF would drink it.’

2.5. Omission of Agents

In Kryz predicative constructions, the expression of arguments as overt

NPs is not grammatically obligatory, but at least one is encoded on the

verb: the minimal predicate consists of a single verb form on which a

Single argument or a Patient is cross-referenced:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(11) a. ( furar) (xhin) ge-t’-ic

man.erg grass.n pv-strike-aor.n

‘(The man) cut it (the grass).’

b. (xinib) yi-p-du

woman(abs) pv-f.go-aor.f

‘(The woman /) she left.’

Easily recoverable or non-referential Agents are very often left out. In

the original of the following English translation the word ‘you’ is used

only once, with obviously generic, non-referential value, and Kryz uses

only active forms (see text in Authier 2009) whereas English has to employ

the passive:

‘The cheese-bag is made to keep cheese in. You have to flay the skin to keepthe cheese in without making any cuts. Then, as soon as it is flayed, salt ispoured into it to dry it out. Then it has to be worked on, and tied firmly atthe fore and hind legs. If there is a great deal of wool, it is shorn, then it iswashed with water while being beaten with a comb (?), then left to dry.Then it is turned inside out (the cheese is on the woolly side) and stored forwinter.’

Non-explicitation of a generic Agent is frequent in proverbs such as the

following (on ingratitude):

(12) tur.ud-zina ula-ci ciyar-zina �ul gva-s-ryu

spoon-instr eat-seq tail-instr eye.f pv.f-pull.out-prs.f

‘[They] eat with the spoon and then pull (your) eye out with the

handle!’

A preliminary conclusion, then, is that in Kryz as in many other (genet-

ically related or unrelated) ergative languages, backgrounding through

omission of the Agent NP and concomitant foregrounding of the semantic

Patient is possible without any change in the verb morphology. But Kryz

also has a detransitive voice which shows properties commonly associated

with the label ‘passive’: it is a morphological derivation which transforms

transitive predicates into intransitive ones by removing the agentive NP

at a syntactic level. But let us first examine the formal manifestation of

(in)transitivity on Kryz verbs.

3. Marking of (in)transitivity and verb classes

Kryz synthetic (non-compound) verbs form a closed class comprising some

200 items. They are either transitive or intransitive, a distinction which is

marked morphologically on imperfective stems only.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


138 Gilles Authier

3.1. Morphology of intransitive vs transitive verbs

All of the 70 synchronically underived intransitive verbs display on their

imperfective verb stem one of three allomorphs of the su‰x -aR (-ar/-al/-an):

Table 2. Structure of intransitive imperfective verb stems

�a-r-c’-ar- ‘pv-ipf-enter-intr’

ke-r-xh-ar- ‘pv-ipf-move-intr’

la-l-s-al- ‘pv-ipf-turn-intr’

ha-n-gv-an- ‘pv-ipf-run-intr’

Conversely, the 130 transitive synthetic verbs do not show these su‰xes.

They are not overtly marked as transitive, but their imperfective forms

should be considered zero-marked for transitivity, given the absence of

the characteristic intransitive su‰xation:

Table 3. Structure of transitive imperfective verb stems

ya-r-t’- ‘pv-ipf-cut’

yi-r-q- ‘pv-ipf-catch’

ki-l-t’- ‘pv-ipf-yoke’

yi-n-gh- ‘pv-ipf-pull’

Overt marking of intransitivity is not to be found on the perfective

forms of underived intransitive verbs, which are thus indistinguishable

from transitive perfectives:

Table 4. Structure of perfective stems of non-derived verbs

intransitive verbs, pf cf. transitive verbs, pf

�a-c’- ‘pv-enter.pf’ ya-t’- ‘pv-cut.pf’

ke-xh-r- ‘pv-move-pf’ yi-q-r- ‘pv-catch-pf’

la-s-l- ‘pv-turn-pf’ ki-t’-l- ‘pv-yoke-pf’

ha<r>gu-n- ‘pv<M>run.pf’ yi-gh-n- ‘pv-pull-pf’

Because of the general omissibility of overt argument NPs, transitive

and intransitive predicates are hard to distinguish in the perfective tenses.

Allowance of explicit ergative arguments is the only criterion available,

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



and no distinction can be drawn by means of syntactic tests between a

Single argument construction and a transitive construction in which the

Agent is left out.

3.2. Detransitive morphology on transitive roots

Transitive verbs may be classed into two groups according to whether they

allow morphological detransitivization or not. When available, detransi-

tive voice is formed synthetically in the imperfective through the addition

of the morpheme -aR- to the right of the verbal root; on most but not all

verbs, a sonorant imperfective aspect marker occurs before the root, and

R is a duplicate of this pre-root aspect marker:

Table 5. Derivation of detransitive imperfective stems

transitive ipf ¼> detr. ipf

‘catch’ yi-r-q- yi-r-q-ar- ‘be caught’

‘pull’ yi-n-gh- yi-n-gh-an- ‘be pulled’

‘yoke’ ki-l-t’- ki-l-t’-al- ‘be yoked’

Intransitive verbs in Kryz thus look like detransitive verb forms with no

corresponding non-derived form, i.e. ‘deponents’ or ‘media tantum’ in the

grammar of classical languages.

The perfective is formed analytically with a verbal adjective derived

from the imperfective detransitive stem, to which the auxiliary xhiyi ‘having

become’ (perfective participle) is added:

Table 6. Analytic perfective of detransitive verbs

detr. ipf detr. pf

‘catch’ > yirq-ar-i ‘being caught’ yirq-ar-a xhiyi ‘caught’

‘pull’ > yingh-an-i ‘being pulled’ yingh-an-a xhiyi ‘pulled’

‘yoke’ > kilt’-al-i ‘being yoked’ kilt’-al-a xhiyi ‘yoked’

3.3. The detransitive derivation (prototypical transitive verbs)

Most transitive verbs allow a detransitive derivation with Patient-maintaining

semantics, either an Agent-backgrounding ‘passive’ or an Agent-suppressing

‘anticausative’. These forms are found in appropriate contexts which usually

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


140 Gilles Authier

imply modal or aspectual nuances, and in fact they are rare in narrative

and everyday speech.

As is clear from this sample, transitive verbal lexical items in the passive

voice retain prototypical transitive (voluntary Agent, Patient-transforming)

semantics:

Table 7. Some typical transitive verbs and passive forms

pf ipf detransitive pf

gaxur-i garxv-i ‘knead’ garx-ar-i / garxvara xhiyi

na<d>q’un-i nanq’v-i ‘churn’ nanq’v-an-i / nanq’vana xhiyi

cakva-y cak(vats’)-i ‘mix’ cakv-ar-i / caukvara xhiyi

raxva-y raxv(ats’)-i ‘shave’ raxv-ar-i / raxvara xhiyi

reuha-y reuh(ats’)-i ‘grind’ reuh-ar-i / reuhara xhiyi

yixha-y yixh(ats’)-i ‘harvest’ yixh-ar-i / yixhara xhiyi

yiza-y yiz(ats’)-i ‘plough’ yiz-ar-i / yizara xhiyi

The following example shows a typical passive transformation (note

that here the passive imperfective stem is not derived from the current

transitive one):

(13) a. har cu�ma-ca Hazratbaba.ci-r lu kel kura-ts’-ryu-ni

every Friday-in Saint.Baba-erg this lamb slay-ipf-prs.f-past

‘Every Friday Saint Baba would sacrifice this lamb.’

b. har cum�a-ca lu kel kur-ar-yu-ni

every Friday-in this lamb slay.detr-prs.f-past

‘This lamb would be sacrificed every Friday.’

Unlike in English, in Kryz the Agent cannot be expressed as an oblique

NP, a frequent feature of passives cross-linguistically. The detransitive con-

struction in Kryz thus complies with Kazenin’s (2001) and Comrie’s (2008)

criteria for passive voice7:

7. We do not retain the definition of passive as a means of promoting the Patientof a transitive verb to grammatical subject, because this would presupposethat Kryz has a clear-cut definition of syntactic ‘subject’, which is not the case:as in most East Caucasian languages, behaviour properties tend to group Swith A, while we have seen that coding properties clearly align S with P, thatis, ergatively. Note also that Kryz detransitive forms never receive reciprocal

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



– it is morphologically heavier, usually derived from the active;

– the construction is ‘‘less frequent, functionally specialized, not fully

productive’’ (Haspelmath, 1990);

– the new subject is not a semantic Agent;

– the semantic role of the maintained argument does not change. In the

case of a language with ergative case marking like Kryz, the unmarked

absolutive case of the maintained Patient is also retained.

3.4. Transitive verbs lacking a detransitive form

Three semantically coherent classes of verbs systematically lack a detran-

sitive form: 1) verbs of in- or ex-corporation; 2) verbs of exchange and

social interaction; 3) transitive verbs of motion (for which see below).

The following two lists are exhaustive:

Table 8. Verbs of social interaction and in- or ex-corporation

gaynic ‘take, buy’ �adgulic ‘swallow’

vaxhayc ‘borrow’ c’udxunic ‘suck’

aqayc ‘keep’ gitnic ‘drink in one draught’

lipic ‘say’ c’agvayc ‘gnaw’

k’isic ‘bite’

gadgunic ‘put on (trousers)’

gacayc ‘vomit’

gahayc ‘extract; hatch’

xuyic ‘give birth’

These verbs are not prototypical transitive verbs. They deviate from the

transitive prototype in that social interactions imply two agentive partici-

pants, and present semantic similarities with reciprocal actions, which

cannot be deactivized. In the second class the Agent is strongly a¤ected

by the event. Such non-prototypical features probably account for the

restriction on passivization observed here, as the main functions of the

or autocausative (reflexive) interpretations (there are no ‘‘grooming middles’’,cf. Kemmer 1993). The fact that processes implying subject-object coreferenceare not expressed by this form of the transitive verb in Kryz has led us to dis-card the term ‘middle’ in favour of detransitive voice. We also discard themore all-embracing appellation ‘Medio-Passive’, because a couple of detransi-tivized verbs have an antipassive interpretation, for which see below.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


142 Gilles Authier

passive (backgrounding of the Agent and foregrounding of the Patient)

make no sense with these verbs.

Among those Kryz transitive roots which do not allow valency decrease

we also find the whole semantic class of verbs denoting transitive motion

events. One motivation for their failure to allow detransitivization may

be the atypically agentive role implied by ‘transportation’ situations: the

Patient is not transformed or internally a¤ected, and retains a great deal

of agentivity if it is animate.

In accordance with this principle, agent backgrounding in a process of

bringing is expressed by using an intransitive counterpart of ‘bring’ with a

meaning such as ‘go’, as in ex. (14):

(14) mahraka.c-a va-xhr-i sabas

contest-in pv-come.together-part presents.for.musicians

‘Gifts gathered at the wedding party.’

Another – lexical – motivation is the fact that these roots neatly match

a class of intransitive motion verbs which do not form causative construc-

tions with ‘do’, unlike the majority of Kryz intransitives: they pair up with

the corresponding transitive roots. Any valency-changing construction is

thus superfluous for both classes, a form of symbiosis which makes this

suppletive subsystem particularly economical:

Table 9. Other verbs lacking detransitive forms: suppletive motion verbs

�agayc ‘bring’

�aqric ‘take down’

�a�ayc ‘push into’

�ac’ic, �a<r>fic ‘enter, go down’

�axhric ‘come (down)’

aska-yc ‘put (down)’ guq’ric ‘touch, intr.’

gagayc ‘create’

gaqric, ga�ayc ‘take, push out’

�ac’ic, gadfic ‘go out, escape’

gaxhric ‘appear’

ge�ayc ‘build’ gi<d>fic ‘rise’

gigayc ‘turn, tr.’ gixhric ‘turn, intr., live’

ke�ayc ‘push under’ kec’ic ‘go under’, kedfic ‘pass under’

va�ayc ‘push aside’ vac’ric, vadfic ‘turn from’

vagayc ‘gather, tr.’ vaxhric ‘gather, intr.’

ya�ayc ‘push aside’ ya<d>fic ‘stray’

yaqric ‘take across’ yac’ic ‘pass, cross’

yatric ‘leave, tr.’ i<d>knic ‘remain’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



4. Passive voice and its additional values

A clause involving a transitive verb in its passive form is usually not to be

understood as a mere paraphrase of the non-derived, transitive clause. Not

only does the passive background the Agent and topicalize the Patient,

but it also has an impact on modal and aspectual parameters. This section

examines these associated e¤ects. We have found passives of transitive

verbs either with habitual and deontic value in the imperfective, or with

state-focusing (resultative) value in the perfective. The referential proper-

ties of the remaining argument are also a¤ected by a change of aspect:

(15) a. gul ambar.c-a va-nsan-e

corn barn-in pv-weigh.detr-prs

‘Corn is usually weighed in the barn.’ (non-referential)

b. gul ambar.c-a va-nsan-a xhi-yic

corn barn-in pv-weigh.detr-a be-aor.n

‘The corn has been weighed in the barn.’ (referential)

4.1. Deontic and Habitual meaning of the imperfective passive

The association of deontic meaning with the passive is well attested cross-

linguistically. Collective or typically feminine chores culturally entail the

de-individualization of the agent, and the generic, non-referential (or

taboo?) status of the unmentioned Agent entails the use of a detransitivized

verb form:

(16) cindir-a q’vahac-ci q’an ke-rt’ar-e

worn-a sock-gen bottom pv-sew.detr-prs

‘When the bottom of a sock is worn out, one patches / must patch it.’

(17) riki yiggacig va-rcar-e

yard every.day pv-sweep.detr-prs

‘The yard is / must be swept every day.’

There are many examples in proverbs in the ‘eventual’ mood (covering

both habitual events and gnomic sentences with deontic modality):

(18) irac irca-zina zimar-da-d

blood blood-instr wash.detr-negevt-n

‘One does not wash away blood with blood.’ (Prov.)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


144 Gilles Authier

(19) �u-du-fa riki ge-t’ar-da-d

pv-neg-close.pf(part) door pv-strike.detr-negevt-n

‘An open door is not knocked at.’ (Prov.)

(20) d-isar-i �ayal.ci-z maxar vuts’ar-da-b

neg-cry.ipf-part child-dat breast.f give.detr-negevt-f

‘A child who does not cry is not given the breast.’ (Prov.)

(21) q’usi.c-a �ayal �a-lt’al-e / gi-rqar-e

cradle-in child pv-bind.detr-prs pv-hold.on.detr-prs

‘The child is (¼should be) firmly bound / held in the cradle.’

(22) heyvanat yif.un-a q’acil-a �a-rgvar-e

cattle night-in everyday pv-push.in.detr-prs

‘Cattle are / have to be penned during the night.’

(23) gurux.c-a�ar barkan-bi hu-rt’ar-e

garden-inel horse-pl pv-expel.detr-prs

‘Horses should be banned from gardens.’

These habitual and deontic values are present in the overwhelming

majority of examples of detransitive forms in the corpus; but passive voice

can also be found when the precise Agent of an action is unknown in the

context. A generic, non-referential interpretation is then preferred, espe-

cially in future time reference:

(24) a-n-iz �ark’a caza vuts’ar-iya

3-h-dat heavy punishment give.detr-fut

‘He will be given a severe punishment.’

4.2. Resultative meaning of the perfective passive

While imperfective passive forms have habitual and deontic readings, the

corresponding perfective forms have a resultative reading. Emphasis is

laid not on the event itself but on the resulting state. This is sometimes a

way to convey the fact that the event was not witnessed by the speaker:

(25) riki cigac-a �a-rt’ar-a xhi-yic

door place-in pv-fit.in.detr-a be-aor.n

‘The door has been put back in place.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(26) dauga q’um-ug gva-xvar-a xhi-d-u

soup ground-super pv.f-pour.detr-a be-aor-f

‘The soup has been spilt on the ground.’

It seems that this resultative interpretation does not necessarily apply

with negative forms:

(27) zin lu yas.in-iz karta hicvaxt atar-a xhi-de-r

1 this year-dat until never beat.detr-a be-negperf-m

‘Right up to my current age, I have never been beaten up.’

4.3. Stylistic variation?

The passive marking of verbs is often optional, and passive forms can

coexist in the same text as unmarked active forms. In the following text,

the first few forms are passive, and set the general ‘deontic’ tone of the

whole passage. Most forms later in the passage are left unmarked, proba-

bly because active forms are ‘lighter’ and systematic passive marking

would be redundant (recall that the Agent can also be omitted in transitive

constructions):

(28) vul sina q’va-d safar vu-dar-yu garayaz-ca

ewe year.in two-n time pv-shear.detr-prs.f springtime-in

vudar-i yis.a-gar yapagu li-re u-c-kar

pv.shear.detr-part wool-supel spring.wool say-prs 3-nh-subel

gi-uxvar-a xhi-yi xiy-ar-kar q’vahac xirats’-re

pv-spin.detr-a be.pf-part thread-pl-subel sock knit-prs

ic-kar an xi-xar-a xhi-yi say-ri

selff-subel an red-plait.detr-a be.pf-part thing-pl

mahkam sare; cixe�an yis yanxuc-a vudats’-ryu

solid be-prs then wool.f autumn-in shear-prs.f

lu safar vu-da yis git sa-re

this time pv-shear.pf(part) wool.f few be-prs

u-c-kar gi-ux-i xiy davam-suz si-u-ryu

3-nh-subel pv-spin-part thread.f continuity-without be-f-prs.f

The wool of the sheep is shorn (PASS) twice a year. The wool shorn

(PASS) in spring is called ‘yapaghu’. With the spun (PASS) threads, one

knits (ACT) socks, and objects knitted (PASS) from them are robust. The

second wool is collected (ACT) in the autumn; this time the wool shorn

(ACT) is not abundant; the thread spun (ACT) from it does not hold.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


146 Gilles Authier

4.4. Orientation of relative participles

The participles found in East Caucasian languages are ‘contextually oriented’

(Haspelmath, 1994), due to the fact that their arguments retain the same

case marking as in independent clauses. In ex. (29), the same participle is

used for both orientations: in a) Patient orientation is indicated by erga-

tive marking on an NP and the absence of an absolutive NP in the relative

clause, whereas in b) Agent orientation is inferred from the absence of an

ergative-marked NP:

(29) a. xinib.ci-r hala da-b-cir-i xasul

woman-erg yet neg-f-bake.pf-part stew.f

‘the stew that the woman has not yet cooked’

b. hala xasul da-b-cir-i xinib

yet stew.f neg-f-bake.pf-part woman

‘the woman who has not yet cooked the stew’

But Kryz also uses derived, specifically negative passive participles, made

up of a transparent analytic combination of the detransitive verbal adjective

and the auxiliary ‘be, become’:

(30) a. reudeuha q’el sifra.c-a a-ma-sku!

neg.grind.ppf salt table-in pv-proh-put

b. reuhar-a dauxha

grind.detr-a neg.be.ppf

‘Do not put unground salt on the table!’

These analytic forms seem to add a slight nuance of possibility:

(31) a. ga-dauxvi xamir.ci-kar fu sa-dad

pv-neg.knead.ppf dough-subel bread be-negdeb

b. ga-rxvar-a dauxha

pv-f.knead.detr-a neg.f.be.ppf

‘Dough does not make bread unless it is kneaded.’

(32) a. ga-da-xi xamir-bekar fu cirar-de-d

pv-neg-knead.ppf dough-pl.subel bread bake.detr-negprs-n

b. ga-rxar-a da-xha

pv-knead.detr-a neg-be.ppf

‘Bread made of unleavened dough does not bake.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Some degree of Agent orientation inherent in imperfective active par-

ticiples is probably the reason why, in the following example, a specifically

passive participle has to be used, because the active participle would be

Agent-oriented and prevent recovery of the Recipient (rather than Agent)

role for the head:

(33) a. caza *vu-ts’-i / vuts’ar-i adami

punishment give-ipf-prt give.detr-prt person

‘The man who is given a punishment.’

b. caza *vu-yi / vuts’ar-a xhi-yi adami

punishment give.pf-prt give.detr-adj be.pf-prt person

‘The man who has been given a punishment.’

4.4. Topic continuity and syntactic accusativity

Morphologically active sequential converbs in Kryz usually imply a corefer-

ent subject (S/A) in the linked matrix clause (see examples in Authier 2009),

that is, an apparent accusative pivot in terms of Dixon 1994. This might

lead us to assume that passive predicates promote Patients to the status

of syntactic subjects. In the following example, coreference of subjects

obtains as the imperfective coordinating converb heading the first clause

is in the passive voice:

(34) lu guyi yiggacig gvat’-ar-a, halu �aranxhin-a xhi-cu

this well every_day pv.dig-detr-manner this depth-in be-prf.f

‘This well being dug every day, it became this deep.’

This may well be true for this imperfective converb: more investigation

is needed. On the other hand, perfective converbs (sequential, ending in

-ci) can be followed by a passive form whose Single argument co-refers

with the Patient of the preceding subordinate clause, as in ex. (35):

(35) gugarti yi-t’a-ci har adami-c busq’ab.ci-g gi-yts’ar-e

greens pv-cut-seq every person-gen plate-super pv-pour.detr-prs

‘The salad is cut and served on everyone’s plate.’

This property of sequential converbs shows that Kryz generally lacks

any clear ‘syntactic pivot’, and this is by no means the only instance of

ergative inter-clausal syntax. In ex. (36), the Single argument in the sequen-

tial subordinate clause co-refers with the Patient of the matrix clause, in

the absence of an overt Agent:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


148 Gilles Authier

(36) sad tak yif xinib-cizina irkin-ci

one.n only night woman-instr m.remain.pf-seq

laqatasi u-d vul-bevas hay yi-re

the_day_after prox-m.abs sheep-pl.cum sent do.ipf-prs.m

‘After remaining only one night with his wife, on the following day

he is sent to (watch over) the sheep.’

4.5. Preliminary conclusions on the use of the Kryz passive voice

The main Kryz passive-forming device seems to be linked to imperfective

aspect, with either habitual nuance or deontic modality. Perfective forms

are derived analytically from the imperfective, and synchronically these per-

fective periphrastic passive forms occur just as often in texts as imperfective

ones, because they o¤er valuable new aspectual nuances (inceptive or resul-

tative). These modal or aspectual nuances are much more crucial to the

actual use of passive forms than any syntactic rule or topic-maintaining

function.

5. Anticausative and antipassive interpretations

More archaic than the passive appear to be those uses of the detransitive

forms to which we now turn, in which the detransitive voice is anticausa-

tive (associated with Patient-oriented meaning) or antipassive (with a few

Agent-oriented verbs such as ‘eat’).

5.1. Anticausative

The same detransitive morphology is used with a number of verbs in con-

texts where the Agent is not only backgrounded, but semantically suppressed:

that is, it encodes anticausative value. For instance:

Table 10. Detransitive forms with anticausative meaning

pf ipf detr. ipf & pf participles

ugv-a- ugv-(a-ts’)- ‘burn’ ugv-ar- / ugvara xhiyi

cir-a- cira-ts’- ‘cook’ cirar-i / cirara xhiyi

�uf-a- �uf-(a-ts’) ‘close’ �ufar-i / �ufara xhiyi

saku-r- sarkv- ‘trim’ sarkvar-i / sarkvara xhiyi

get-a- get- ‘beat’ getar-i / getara xhiyi

surh-a- su<d>ha-ts’- ‘drag’ surhar-i / surhara xhiyi

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



When the detransitive derivation is applied to these verbs, only one

participant remains in the situation. The semantic role of the absolutive

argument is also retained in the anticausative construction, but unlike in

the passive, where an unexpressed agent is implicit, here the agent is con-

ceptually removed. In the anticausative interpretation the event is sponta-

neous, often with significant semantic modification of the verb.

Note that two distinct interpretations may coexist, with or without an

implicit Agent:

(37) zang ge-rt’ar-a xhi-yic-zina k’ul.c-a ha-r-gun-d

bell pv-strike.detr-adj be-msd-instr house-in pv-m-run-aor.m

‘The bell a) was rung (passive) or b) rang (anticausative), and he

ran home.’

While the passive interpretation of detransitive forms is slightly more

prevalent in narrative texts, the anticausative or ‘spontaneous’ meaning is

found in idiomatic expressions:

(38) c’ebic gi-yts’ar-e

rain pv-pour.detr-prs

‘The rain is pouring.’

(39) palas ge-tar-e

carpet pv-beat.detr-prs

‘The carpet slaps (in the wind) / *starts to slap (cf. ex. 46).’

In contrast to the e¤ect noted above for the passive use, the imperfec-

tive gives rise to no aspectual (habitual) or modal (deontic) nuances:

(40) za kil.a-k rib cu-nq’van-yu

1.gen arm-sub needle(f) pv-stick.detr-prs.f

‘A needle sticks itself in my arm.’

5.2. Technical spontaneous meaning

The semantic role of the absolutive argument is often modified along with

the meaning of the verb put into the anticausative. For instance, the derived

detransitive form of ‘catch’ means either ‘be caught (by the police)’ (passive

interpretation) or ‘coagulate’ or ‘be eclipsed’ (anticausative):

(41) yux yi-rqar-e

milk pv-catch.detr-prs

‘The milk ‘‘catches’’ ¼ turns to cheese.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


150 Gilles Authier

(42) varag yurqar-yu

sun.f pv.f.catch.detr-prs.f

‘The sun is eclipsed.’

5.3. Perfective anticausatives: additional inceptive value

The non-derived, labile, perfective form has anticausative meaning:

(43) za kil.a-k rib cu-b-q’vun-du

1.gen arm-sub needle(f) pv-f-stick-aor.f

‘A needle has stuck itself in my arm.’

To give a passive reading, the derived, explicitly detransitive form is

required:

(44) rib za halav.ci-k cu-nq’van-a xhi-du

needle 1.gen dress-sub pv-stick.detr-a be-aor.f

‘The needle has been stuck in her dress (by someone).’

Derived perfective anticausatives are also found, usually with an addi-

tional aspectual value of inceptivity (to be contrasted with the resultative

meaning of passive perfectives; note also that Kryz lacks a general verb

‘start, begin’):

(45) c’ebic gi-yts’ar-a xhi-yic

rain pv-put.on.detr-a be-aor.n

‘The rain started to pour.’

(46) palas ge-tar-a xhi-yic sa-re

carpet pv-beat.detr-a be-aor.n be-prs

‘The carpet started / starts to slap (in the wind).’

(47) yux yi-rqar-a xhi-yic

milk pv-catch.detr-a be-aor.n

‘The milk turned to cheese.’

Much rarer is the resultative reading, with the auxiliary in the perfect

tense:

(48) za galu yi-rqar-a xhi-ca

1.gen throat pv-catch.detr-a be-prf

‘My throat is blocked.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



5.4. Magic autocausative

Detransitive forms can be used in magical contexts to emphasize the un-

natural absence of the expected, supposedly necessary Agent:

(49) a-c-iz amr v-ar-izma�an k’ul va-rcaryu-ni

3-nh-dat order.f f-do-before room.f pv-sweep.detr-prs.f-past

‘(She had a magic ring:) as soon as she gave it the order, the room

would be swept.’

(50) div-ci k’ul.c-a yis icic-igan ga-har-e-ni

demon-gen house-in wool selff(e)selff-equ pv-card.detr-prs-past

‘In the house of the demon, the wool would card itself.’

The analytic perfective detransitive form also serves to express impera-

tive modality with spontaneous and inceptive meaning. For instance, in

addressing his own instrument, a bard says:

(51) ca-rt’ar-a sak, ay saz!

pv-strike.detr-a be.imp Oh lute

‘Start playing, my lute!’

5.5. Instantiation of the semantic role of ‘force’

While the passive interpretation (which preserves semantic roles) does not

allow any oblique expression of the backgrounded agent, the anticausative,

semantically intransitive interpretation of the detransitive voice is com-

patible with the expression of an argument in the semantic role of ‘force’.

If this force is external, it is found in the subelative case, and normally

appears in initial position:

(52) (kulak.ci-kar) riki �u-f-ar-a xhi-yic

wind-subel door pv-close-detr-a be-aor.n

‘The door closed because of the wind.’

(53) (varag.ci-kar) za �ic ugv-ar-a xhi-yic

sun-subel 1.gen skin burn-detr-a be-aor.n

‘My skin was burned by the sun.’

These detransitive forms are not Patient-promoting and even less Agent-

demoting: the corresponding transitive clauses are not ungrammatical, but

seem extremely awkward to speakers:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


152 Gilles Authier

(54) ?kulak.ci-r riki �u-fa-c

wind-erg door pv-close-aor.n

‘The wind closed the door.’

(55) ?varag.ci-r za �ic ugva-c

sun-erg 1.gen skin burn-aor.n

‘The sun burned my skin.’

A force a¤ecting the subject internally may be expressed in the subloca-

tive case in the following example (technical):

(56) ge� yig-in.a�ar sehirci azar.ci-k yi-rqar-iya vun

today day-inel magical illness-sub pv-catch.detr-fut 2

‘From today on, you will be a¿icted with a magical disease.’

5.6. Antipassive with ‘eat’, ‘drink’, and other verbs

The detransitive forms of the two verbs ‘eat’ and ‘drink’ are ambiguous

and can be interpreted as either Patient- or Agent-preserving, i.e. as anti-

causative or (more usually) as antipassive. The Agent-backgrounding, pas-

sive use is found in gnomic contexts:

(57) lem-ird yak ugval-de-d, yux gi-gar-de-d

donkey-gen meat eat.detr-negprs-n milk red-drink.detr-negprs-n

‘The meat of the donkey is not eaten, nor is its milk drunk.’

but also in the perfective, if the Agent is unspecified:

(58) ugval-a daxha yak xvar-imez sa-�a-c

eat.detr-a neg-be.pf(part) meat dog-pl.dat pv-throw-aor.n

‘The meat which was not eaten has been thrown to the dogs.’

(59) dauga gi-gar-a xhi-d-u

doogh red-drink.detr-a be-aor-f

‘The doogh has been drunk.’

The antipassive use also has a technical meaning:

(60) u-be-k ibe bala-yar gi-gar-e

3-npl-sub selfnplg child-pl red-drink.detr-prs

‘Their foals drink under them.’

(61) vul-bi ugval-ciz. . .

ewe-pl eat.detr-simul

‘While the sheep were grazing. . .’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The perfective antipassive is inceptive:

(62) vul-bi ma�an halu huq’-a ugval-a xhi-yic

ewe-pl again this meadow-in eat.detr-a be-aor.n

‘The sheep started to graze in the meadow again.’

(63) leh-ar gi-gar-a xhi-yic

calf-pl red-drink.detr-a be-aor.n

‘The calves have started to drink.’

Other spontaneous or quasi-reflexive (non-passive) readings of the de-

transitive voice are also found, which preserve non-controlling Agents

such as animals or the radio, in processes involving the body or voice:

(64) hava-yar ghira xhi-yic-zina kis-ar gi-nghan-e

air-pl warm be-msd-instr hen-pl pv-lay.detr-prs

‘As soon as the weather becomes warm, the hens start laying eggs.’

(65) pirennik uxvar-e / uxvar-a xhi-yic

radio recite.detr-prs recite.detr-a be-aor.n

‘The radio is on / started to play.’

(66) vul-bi ca-rhar-e

ewe-abs.pl pv-soil.detr-prs

‘Sheep get dirty.’

Table 11. Detransitive verb forms with antipassive reading

pf ipf (tr.) detr. ipf & pf participles8

gira- ‘drink’ gig- gigar- 1) ‘be drinkable;2) ‘start / be made to drink’

u(gu)la- ‘eat’ ul(ats’)- ugval- / 1) ‘be edible’;2) ‘pasture’

gighn- ‘lay (eggs)’ gingh- ginghan-

caha- ‘soil’ cac- carhar-

uxva- ‘read, sing’ uxv-(ats’)- uxvar-i

8. Recall that the analytic forms (gigara xhiyi; ugvala xhiyi; carhara xhiyi;uxvara xhiyi) do exist, but with a new, passive interpretation. The form ging-hana xhiyi means ‘having been knitted’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


154 Gilles Authier

This tripartite morphology, with a ‘labile’ perfective form opposed to

the imperfective transitive/detransitive pair, certainly reflects an archaic

situation, as is shown by the existence of parallels in non-Lezgic branches

of East Caucasian.

6. East Caucasian origins of the detransitive voice

The passive voice is restricted to Kryz among the East Caucasian lan-

guages, which more commonly have derivations with anticausative or

antipassive readings. The passive reading probably became prevalent only

recently, under the influence of Azeri: it first appeared with verbs for which

the event encoded cannot come about without an external causer (i.e. which

are semantically Agent-oriented, cf. Haspelmath 1993) and as a result

cannot be cast in the anticausative (with a Patient-oriented meaning com-

ponent), and then it became an option with most other transitive verbs in

Kryz.

We may account for the fact that the main – passive – use of the su‰x

-aR- in Kryz cannot be found in any other East Caucasian language if we

bear in mind that ellipsis of the Agent is always available for the purpose

of Agent backgrounding, and that other semantics linked to Kryz passive

forms (modal or aspectual) are expressed independently of diathesis in

other languages.

However, a comparable detransitive marker seems to be attested as

such in nearly all branches of the East Caucasian family. Although these

languages have no passive constructions, they do have other detransitive

voices with comparable morphology.

6.1. Tsezic, Avar and Dargi detransitives

In other branches of the East Caucasian family, we find a significant sample

of (non-productive) detransitive voices in Tsezic languages, apparently in

at least one Dargic language, and in Avar.

Avar is well known for possessing derived intransitive verbs with ‘itera-

tive’ or ‘durative’ meaning, which show the morpheme -aR; they are

clearly of the antipassive type:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 12. Transitive and iterative detransitive stems in Avar (Charachidze 1982)

b-ets-ize ‘mow’ w-ec-ar-ize ‘be busy mowing’

b-uq’-ize ‘sew’ w-uq’-ar-ize ‘be busy sewing’

qver-ize ‘slay’ qveqar-ize ‘slay cattle for the winter’

as-ine ‘sting’ w-as-and-ize ‘dance, frolic’

heq’e-ze ‘drink’ heq’old-ize ‘be busy drinking’

b-its-ize ‘tell’ bits-ard-ize ‘reproach’

b-etsts-ize ‘praise’ v-etsts-ar-ize ‘praise oneself ’

(67) a. he-ł he-sda b-its-anila 50 tumen b-it’-i

this-f.erg this-m.loc n-tell-pstevid 50 toman n-send-msd

‘She told him about the (sending of the) 50 tomans.’

b. ros he-lde kves bitsard-anila

husband this-f/n.lat bad reproach-pstevid

‘The husband was reproaching her bitterly.’

Tsez (see Comrie 2000) has an antipassive making use of a morpheme

-na-:

(68) a. Pat’aa uji esa-yxo

Fatima.erg boy.abs(m) (m)wash-prs

‘Fatima washes the boy.’

b. Pat’i ker-aa y-esa-na-yxo

Fatima.abs(f ) river-in f-wash-detr-prs

‘Fatima does the laundry in the river.’

Bezhta (van den Berg 2005, quoting Dr. Majid Khalilov, a native

speaker) has a morpheme -lA:, which derives the antipassive form of

certain verbs (note that the original absolutive-marked object can be ex-

pressed in the instrumental case in some of the new derived constructions):

(69) a. ojdi qarandi y-o:t’o-yo

boy.erg hole.abs(n) n-dig-pst

‘The boy dug the hole.’

b. ojo qarandi-yad o:t’o-la:-yo

boy.abs(m) hole-instr (m)-dig-detr-pst

‘The boy was digging at the hole.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


156 Gilles Authier

Other verbs lose any ability to express a Patientive argument. Note the

aspectual shift in the English translation:

(70) a. kid-ba hak’a kl’eq’e-yo

girl-erg boots sew-pst

‘The girl sewed the boots.’

b. kid kl’eq’e-la:-yo

girl(abs) sew-detr-pst

‘The girl was sewing.’

Hunzib (closely related to Bezhta, see van den Berg 1994) shows two

possible interpretations of the same morpheme -la:-, antipassive and anti-

causative:

(71) a. ołul bex kose

he. erg grass mow

‘He mows the grass.’

b. eg kose-la:

he.abs mow-detr

‘He mows (often, usually).’

(72) a. iyul q’utila: zinkl’ay m-utsu-r

mother.erg trunk.dat ring(a) a-hide-pst

‘Mother hid the ring in the trunk.’

b. kid q’utila: y-utsu-la:-r

girl trunk-dat f-hide-detr-pst

‘The girl hid in the trunk.’

Table 13. Detransitive derivations attested in Tsezic

transitive Patient-preserving:anticausative

Agent-preserving:antipassive

Tsez ‘wash’ -esa- -esa-na-

Bezhta ‘sew’ -o:t’o- -o:t’o-la:-

Bezhta ‘dig’ tl’eq’e- tl’eq’e-la:-

Hunzib ‘hide’ -utsu- -utsu-laa-

Hunzib ‘mow’ kose- kose-la:-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Icari is a language of the Dargic branch, and can now be considered

well-described thanks to Sumbatova and Mutalov (2006). Like Kryz,

it systematically marks intransitivity on stems with the allomorphs -ar-,

-al-, -an- for third person subjects. These stems are attested in all dialects,

and more commonly used as infinitives. I suspect that they are former

impersonal stems, genetically linked with the Kryz detransitive voice.

Table 14. Detransitive derivation in Icari Dargi

tr. 3d person intr.

‘rot’ pf ø er’-ar-

ipf ‘break’ ø u’-ar-

pf elq’w- elqw-an-

ipf luq’w- luqw-an-

Unfortunately, the grammar provides no examples of sentences show-

ing the possible contexts in which this detransitivization may take place.

The great productivity of this derivational marker in Icari and Kryz may

or may not represent a recent, parallel development.

6.2. Lezgic languages and the nominal hypothesis

The detransitive derived stems are obviously old in Kryz itself: often they

are not derived from the transitive imperfective stem currently in use.

Transitive forms seem younger than their detransitive counterparts, as

they contain a recent imperfective morpheme -ts’-:

Table 15. Irregular detransitive derivation in Kryz

thematic pf ipf DETR. ipf

‘cook’ cir-a- cira-ts’- cir-ar-

‘slay’ kur-a- kura-ts’- kur-ar-

‘shear’ vud-a- vuda-ts’- vud-ar-

‘weave’ xir-a- xix- / xira-ts’- xir-ar-

‘wash’ zim-a- zima-ts’- zim-ar-

compare, athematic:

‘tie’ vat’l- valt’- valt’-al-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


158 Gilles Authier

Furthermore, in some instances the detransitive verbal adjective retains

strong gender agreement on the stem, whereas strong forms of agreement

are normally found only on perfective stems:

PRS.F AOR.F DETR PRS.F DETR AOR.F

‘yoke’ kilt’ilyu ki<b>t’ildu kilt’al-yu ki<b>t’ala xhidu

As for the analytic perfective detransitive, it is certainly a recent creation

in Kryz, prompted by the influence of the highly productive passive found

in Azeri. The anticausative voice with labile perfective is older, as witnessed

by the closely related Budugh language, which shows no trace of a passive

construction but has a couple of anticausative imperfective forms directly

comparable with those of the corresponding verbs in Kryz:

Table 16. Budugh detransitive imperfective stems

pf (labile) ipf DETR. ipf

‘cut’ ya-t’- ya-r-t’- yart’-ar

‘catch’ sı-q-r sı-r-q- sırq-ar

The su‰x -aR- is related to imperfective aspect markers, which in turn

are akin to the nominal (collective) plural markers -r-, -l-, -n- found in

all branches of the East Caucasian family. The following table shows that

the three sonorants must be considered polycategorial plurality/iterative

morphemes in Kryz.

Table 17. Intransitivity, imperfectivity and plurality markers in Kryz

nominal plurals prefixes > ipf aspect su‰xes > intransitivity

-r- gub-ri ‘frog-PL’ yi-r-q- ke-rxhar- ‘move.ipf’

-l- Ø ki-l-t’- la-lsal- ‘turn.ipf’

-n- xiy-ni ‘thread-PL’ yi-n-gh- ha-ngvan- ‘run.ipf’

The detransitivizing su‰x -aR- is not restricted to inflected verbs. Lezgian

has a few verbal nouns (e.g. zw-er ‘running’, zw-al ‘boiling’) which display

the same ‘intransitive’ su‰x, Agent- or Patient-oriented. We propose to

recognize it also on a substantial set of Kryz nouns (note that some of

these are also found in Rutul, from another branch of Lezgic):

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 18. ‘Detransitive nouns’ in Kryz

derived noun cf. verbal root

-ar xh-ar: ‘wind’ ke-xh-r-ic ‘move’

gh-ar: ‘snake’ yi-gh-nic ‘draw, pull’

c’-ar: ‘overhang’ ga-c’-ic ‘exit, protrude’

z-ar: ‘cow’ ¼ ‘milch cow, Ger. melkende Kuh’ a-z-ayc ‘milk’

-al x-al: ‘roof; cobweb’ x-irayc ‘weave’

We may now propose the following diachronic scenario: some of these

intransitive deverbals came to be used as substantives, while others never

became lexicalized. But the latter, used as predicates with semantic Patient-

or subject-orientation, evolved into anticausative or antipassive forms,

which are not only preserved in the Southern sub-branch (Kryz and

Budugh) of the Lezgic languages but also quite well distributed in a range

of languages belonging to other branches. The last stage of this evolution

in Lezgic is the Kryz passive use of this common East Caucasian detransi-

tive voice (for the evolution of anticausative into passive markers, see for

instance Haspelmath 1990).

Conclusion

The Kryz passive is unique in East Caucasian, and its existence and pro-

ductivity are apparently linked to associated modal and aspectual values,

rather than to such syntactic features as an accusative pivot (although

reference tracking predominantly operates on an accusative basis) or topic

continuity. Its habitual interpretation in the older, imperfective forms is

also a typical semantic property of antipassives, and indeed, other East

Caucasian detransitive voices are not passive but antipassive in function.

Other East Caucasian detransitive, Agent-preserving voices are also

highly restricted in their use, more so than the passive in Kryz, and they

have only been brought to light by recent descriptions based on original

texts. But it is significant that, although rare, the instances of voice forma-

tion discussed here are distributed widely across the di¤erent branches of

the East Caucasian family, meaning that a contact explanation is not likely.

We therefore believe that this is a recessive, and in the case of Kryz, quite

resilient, feature inherited from Proto-East-Caucasian morphology.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


160 Gilles Authier

Abbreviations

1plecl 1st person plural exclusive

-a attribute

ad adlocative

adel adelative

adr addressative

all allative

aor aorist, on a pf stem

apud apudlocative

apudel apudelative

deb debitive

dir directive

el elative case

equ equative

evt ‘eventual mood’, on an ipf stem

f singular human female, animals, plants and some animates

h human

if conditional converb

int interrogative

itr intransitive

ipf imperfective

msd masdar, on a pf stem

neg negation

nh non-human

notn (pronominal su‰x) non-neuter

pf perfective

ppf perfective participle

self reflexive pronoun

sub sublocative

subel subelative

supel superelative

super superlocative

References

Authier, Gilles2005 ‘‘Split Ergativity on Kryz Pronouns’’ in Cahiers du CELIA: Erga-

tivity in Amazonia, III: Coreference. ed. Francesc Queixalos.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Authier, Gilles2009 Grammaire kryz. Paris/Leuwen: Peeters.

van den Berg, Helma1994 Hunzib. Munich: Lincom.

van den Berg, Helma2001 Dargi Folktales. Leiden: Universiteit Leiden, Research School of

Asian, African and Amerindian Studies.van den Berg, Helma

2005 ‘‘The North-East Caucasian Languages’’. Lingua 115(1).Charachidze, Georges

1982 Grammaire de la langue avar Paris: Jean Favard.Comrie, Bernard

2000 ‘Valency-changing derivations in Tsez’. In Robert M.W. Dixonand Alexandra Y. Aikhenvald, eds.: Changing Valency: CaseStudies in Transitivity, 360–374. Cambridge: Cambridge Univer-sity Press.

Dixon, Robert M.W.1994 Ergativity, Cambridge: Cambridge University Press.

Dixon, Robert M.W. & Aikhenvald, Alexandra Y.2000 Changing Valency. Cases studies in transitivity Cambridge: Cam-

bridge University Press.Haspelmath, Martin

1990 ‘‘The grammaticization of passive morphology’’. In Studies inLanguage 14.1: 25–71.

Haspelmath, Martin1993 A grammar of Lezgian. Berlin/New York: Mouton DeGruyter.

Haspelmath, Martin1993 ‘‘More on the typology of inchoative/causative verb alterna-

tions’’. In: Bernard Comrie, & Maria Polinsky, (eds.) Causativesand transitivity. (Studies in Language Companion Series, 23.)Amsterdam/Philadelphia: Benjamins, 87–120.

Haspelmath, Martin1994 ‘‘Passive participles across languages’’. In: Barbara Fox, & Paul

J. Hopper, (eds.) Voice: Form and Function. (Typological Studiesin Language, 27.) Amsterdam/Philadelphia: Benjamins, 151–177.

Givon, Tom1994 ‘‘The pragmatics of de-transitive voice: Functional and typol-

ogical aspects of inversion’’. In: Talmy Givon, (ed.). Voice andinversion. Amsterdam/Philadelphia: Benjamins, 3–46.

Kazenin, Konstantin2001 ‘‘Passive’’. In: Martin Haspelmath, Ekkehardt Konig, Werner

Osterreicher & Wilhelm Raible. Language Typology and Lan-guage Universals. An International Handbook. Berlin/New York:DeGruyter, 970–978.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


162 Gilles Authier

Kemmer, Susan1993 The middle voice. Amsterdam/ Philadelphia: Benjamins.

Kibrik, Aleksandr E.1997 ‘‘Beyond subject and object: Toward a comprehensive relational

typology.’’ Linguistic Typology 1. 1997: 279–346.Malchukov, Andrej

2006 ‘‘Transitivity parameters and transitivity alternations: constrain-ing co-variation’’. In Leonid Kulikov, Andrej Malchukov &Peter de Swart (eds.). Studies on case, valency and transitivity,329–359. Amsterdam/ Philadelphia: Benjamins.

Malchukov, Andrej2005 ‘‘Case Pattern Splits, Verb Types and Construction Competition.’’

In: Mengistu Amberber & Helen de Hoop (eds.) Competition andVariation in Natural Languages: The Case for Case. Amsterdam,Boston: Elsevier, 73–118.

Nichols, Johanna1992 Linguistic Diversity in Space and Time. Chicago: University of

Chicago Press.Saadiev, Shamseddin

1994 ‘‘Kryts’’. In: Rieks Smeets (ed.) The Indigenous Languages of theCaucasus, vol. IV. New York: Delmar, Caravan.

Sumbatova, Nina & Mutalov, Rasul2004 A Grammar of Icari Dargi, Munich: Lincom.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Laz middle voice

Rene Lacroix

1. Introduction

The South Caucasian languages (Laz, Mingrelian, Georgian and Svan, also

called the Kartvelian languages) share a common morpheme, i-, which

appears in finite verb forms immediately before the root. This paper aims

to present the functions of i- in Laz. As an introduction to the problem, let

us consider the ways i- is commonly described for Georgian, the best-

known language in the family.

In Georgian grammars (Sani‰e 1953, Tschenkeli 1958, Hewitt 2005,

Aronson 1990), no unified account is given of the morpheme i-. Consider

for instance example (1). In (1b), i- indicates that the action takes place for

the benefit of the subject himself: in this use it is often labelled ‘‘subjective

version’’ (subj.vers). In (1a), by contrast, the verb does not specify any

beneficiary; the morpheme a-, which appears in the same slot as i-, is

labelled ‘‘neutral version’’ (neutr.vers). The glossing of the pronominal

a‰xes in the verb is explained in section 2; thematic su‰xes (ths) such as

-eb appear only in certain tenses, including the present and the imperfect.

They do not have any clear semantic value.1

(1) a. Me a-v-a-sen-eb saxl-s.

1s pv-i1-neutr.vers-build-ths house-dat

‘I will build a house.’ (Tschenkeli 1958: 247)

b. Me a-v-i-sen-eb saxl-s.

1s pv-i1-subj.vers-build-ths house-dat

‘I will build a house for myself.’ (Tschenkeli 1958: 247)

1. I would like to thank Winfried Boeder, Denis Creissels, Antoine Guillaumeand Katharina Haude for their valuable comments on earlier drafts of thisarticle. Thanks are due to my informants as well. In the transcription of Laz,the apostrophe marks glottalized consonants; when several glottalized con-sonants follow each other, I indicate glottalization on the last one only. ‰Łcorresponds to [d‰], ‰ to [dz] and c to [ts]. Since /r/ often drops in Laz, certainmorphemes may appear with /r/ in some examples but not in others.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:42) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 165–198 1351 Authier_05_lacroix (p. 165)

There is a class of verbs, often labelled ‘‘passives’’, which are also

marked by i-. Compare (2b) and (3b) with their transitive counterpart in

(2a) and (3a).

(2) a. St’udent’-ma da-c’er-a c’erili.

student-erg pv-write-aor.i3s letter

‘The student has written the letter.’ (Tschenkeli 1958: 254)

b. C’erili da-i-c’er-a st’udent’-is mier.

letter pv-pass-write-aor.i3s student-gen by

‘The letter has been written by the student.’

(Tschenkeli 1958: 254)

(3) a. Is c’am-s xil-s.

dem eat-i3s fruits-dat

‘He eats fruits.’ (Tschenkeli 1958: 255)

b. Es xili ar i-c’m-eb-a.

dem fruits neg pass-eat-ths-i3s

‘These fruits cannot be eaten / are not edible.’

(Tschenkeli 1958: 255)

Finally, with certain verbs, the morpheme i- is found in the formation

of some tenses such as the future and the aorist. I gloss it with tns for

‘‘tense’’.

(4) a. v-t’rial-eb b. v-i-t’rial-eb

i1-spin-ths i1-tns-spin-ths

‘I spin’ ‘I will spin’ (Tschenkeli 1958: 295)

The grammar of Georgian has long served as a model for the analysis

of the other languages of the family. According to Holisky (1991), for

instance, in Laz the morpheme i- marks ‘‘subjective version’’ in diviboni

‘I washed myself ’ (p. 438); on the other hand, this morpheme serves to

derive intransitive forms which ‘‘have passive meaning or express possibility’’

(p. 422), one of the examples given being igiben ‘it boils’.

This paper aims at presenting the functions of Laz i- in a unified manner.

It will be argued that the uses of i- correspond to what has been labelled

middle voice in other languages.

Laz is spoken in north-eastern Turkey. Estimates of the number of

speakers range from 45,000 (Andrews 1989) to 500,000 (Holisky 1991).

Laz is an unwritten and endangered language; although young people still

understand it, they speak only Turkish.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


166 Rene Lacroix

According to some scholars (Marr 1910; Cikobava 1936), Laz is divided

into three main dialects. Kutscher (2001) distinguishes four dialects. The

data presented here is from the dialect of the town of Arhavi and neigh-

bouring villages. My study is based on a corpus of published Laz texts

(Dumezil 1937, 1967, and 1972, Z'ent’i 1938, Q’ipsi‰e 1939, K’art’ozia

1972 and 1993) as well as data from my work with native speakers. Unless

otherwise stated, the examples are from my informants.

In section 2, I present the coding properties of A, O and S arguments;

in section 3, I briefly summarize the functions of the verbal valency opera-

tors; I then propose a classification of Laz verbs (section 4), which will

serve as a basis for the presentation of the di¤erent uses of the middle

marker i- in section 5; in section 6, I discuss other accounts of the func-

tions of i- in the Kartvelian languages, before the conclusion in section 7.

2. Coding of syntactic functions

In Arhavi Laz, syntactic functions are indicated by cases and cross-

referencing a‰xes. There are two sets of cross-referencing a‰xes, glossed

with Roman numerals. Set I cross-references the A and S arguments,

while Set II cross-references the O argument.2 Sentence (5) illustrates the

basic transitive construction. The A argument (bere) is in the ergative and

is cross-referenced on the verb by a Set I a‰x (-u); the O argument (ocxo‰Ł)is in the absolutive and is cross-referenced by a set II a‰x:

(5) Bere-k ocxo‰Ł me-Ø-tk’oc-u.

boy-erg comb pv-ii3-throw-aor.i3s

‘The boy threw the comb.’ (Dum37.10.3)

First and second person objects, unlike third person objects, are overtly

cross-referenced on the verb (see ex. 9).

In the terminology used here, ‘‘object’’ is taken in the restrictive sense

of ‘‘direct object’’; it is synonymous with ‘‘O argument’’. The term ‘‘subject’’

will be used to cover both A and S arguments, which exhibit the same

behavioural properties.

2. Kartvelologists often use the terms ‘‘subject’’ and ‘‘object’’ a‰xes. The problemwith this terminology is that there are constructions where the argument cross-referenced by ‘‘object’’ a‰xes exhibits subject properties, such as number agree-ment in the verb. For this reason, I prefer a more neutral terminology.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Laz middle voice 167

The S argument is cross-referenced by Set I a‰xes. Certain intransi-

tive verbs take an absolutive subject (ex. 6) while others take an ergative

subject (ex. 7).

(6) Mk’yapu xrock-u-n.

jackal die-ths-i3s

‘The jackal is dying.’ (Z'.96.26)

(7) K’oci-k cind-um-s.

man-erg sneeze-ths-i3s

‘The man sneezes.’

Comparison of (6) with (5) shows that the alignment of absolutive-S

verbs is of the mixed type: S behaves like O with respect to case marking

(both are in the absolutive), but it behaves like A with respect to cross-

referencing (both are cross-referenced by Set I a‰xes). On the other

hand, comparison of (7) with (5) shows that the alignment of ergative-S

verbs is accusative: S behaves like A in terms of case marking (both are

in the ergative) and cross-referencing (both are cross-referenced by Set I

a‰xes).

Strictly speaking, then, the major alignment types of Arhavi Laz are

not ergative, but mixed for absolutive-S verbs and accusative (with

‘‘marked nominative’’) for ergative-S verbs.

There is an alignment split between nouns and third person pronouns

on the one hand and first/second person pronouns on the other, as these

have the same form in the ergative, absolutive and dative cases:

Table 1. Declension of first and second person pronouns

ergative, absolutive and dative

1sg ma

2sg si

1pl cku

2pl tkva

The alignment of first and second person pronouns is thus neutral with

respect to case marking. First and second person cross-referencing, how-

ever, remains accusative. The examples below illustrate the use of the

second singular pronoun si in A, O and S functions respectively. In the

paradigm of Set I cross-referencing a‰xes, the second person singular is

the only cell which bears no overt marking (except in the future).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


168 Rene Lacroix

(8) Si mu cu-me?

2s what wait-ths

‘What are you waiting for?’ (Dum67.XX.42)

(9) Ma si e-k-c’op-are.

1s 2s pv-ii2-marry-fut.i1/2s

‘I will marry you.’ (Dum37.121.6)

(10) Si guruni ye-i?

2s donkey be-int

‘Are you a donkey?’ (K’art’72.193.26)

Basic word order is SOV. Word order does not indicate grammatical

functions, but instead reflects information structure.

The dialect of Ardesen has lost the ergative and dative cases (Dumezil

1972; Kutscher 2001). Compare in this respect (11a), taken from Ardesen

Laz, with (11b), from Arhavi. In (11a), the subject mtuti ‘bear’ and the

applicative argument arkadasi-musi ‘his friend’ are not case-marked, while

in (4b), the subject is in the ergative and the applicative argument in the

dative (for more details on the applicative construction, see below).

(11) a. Mtuti arkadasi-musi u‰Ł i k-el-u-d-u.

bear friend-poss3s ear pv-pv-ii3.appl-put-aor.i3s

‘The bear put his ear on his friend’s side.’ (Dum72.4.5)

b. Mtuti-k arkadasi-musi-s u‰Ł i el-u-d-u.

bear-erg friend-poss3s-dat ear pv-ii3.appl-put-aor.i3s

‘The bear put his ear on his friend’s side.’ (Dum72.4.5)

Ardesen Laz thus di¤ers from the other dialects in that no cases are

used to encode core arguments; cross-referencing, however, functions as

in the other dialects.

The alignment system of Laz di¤ers from that of the other Kartvelian

languages. Throughout the family, cross-referencing functions in roughly

the same way, but case marking di¤ers (Harris 1991b: 3.2.2 and 4.2.1).

Alignment in these languages cannot be presented here in detail. Su‰ce

it to say that ergative, mixed and accusative alignments are found with

nouns and third person pronouns; first and second person pronouns have

the same form in the absolutive, ergative and dative cases, as in Laz.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



3. Valency operators

The morphological structure of finite verb forms in Laz may be summarized

as follows:

-3 preverbs

-2 cross-referencing prefixes

-1 valency operators

0 root

1 causative

2 TAM and cross-referencing su‰xes

The immediately pre-root slot (–1) may contain one of five valency

operators, each of them consisting of one vowel: o- indicates that the

verb is transitive, and is also used to derive applicative forms; i- marks

middle voice; i/u- mark mainly applicative derivation; a- marks, among

other things, applicative derivation and middle voice simultaneously. No

two valency operators may occur together.

Such markers are often called ‘‘preroot vowels’’ or ‘‘preradical vowels’’

in Kartvelian linguistics. As has been pointed out to me by Denis Creissels

and Antoine Guillaume, such terms refer to the phonological rather than

the morphosyntactic level. Since the general function of these markers is to

indicate the valency of the verb they occur in, the term ‘‘valency operator’’

will be preferred here.

As was mentioned above, the functions of the valency operator i- are

generally not given a unified account in the literature on South Caucasian

languages. I will return to this point in section 6. In this paper, I show that

the valency operator i- in Laz functions as a middle voice marker. Con-

sider for instance example (12). In (12b), the addition of i- to the verb

yields an intransitive verb with an anticausative reading – one subtype of

the middle.

(12) a. Bozo-k nek’na ge-nk’ol-um-s.

girl-erg door pv-close-ths-i3s

‘The girl closes the door.’

b. Nek’na ge-i-nk’ol-e-n.

door pv-mid-close-ths-i3s

‘The door closes.’

The term ‘‘middle’’ is taken here in a broad sense, corresponding to that

given in Kemmer (1993). It covers intransitive constructions such as the

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


170 Rene Lacroix

anticausative and the autocausative, but also includes some transitive con-

structions, as in ‘I build a house for myself ’.3 The uses of the valency

operator i- are the focus of this article and are examined in more detail in

section 5.

Transitive verbs may be classified into two groups according to whether

they involve the valency operator o- or not. This operator (glossed tr for

‘‘transitive’’) does not correlate with any semantic distinction; its presence

is lexically conditioned. It corresponds etymologically and functionally to

the ‘‘neutral version’’ of Georgian grammar (see example [1a]). Some

examples are given below.

Some transitive verbs which do not employ the valency operator o-

p-sinax-am i1-hide-ths ‘I hide sth’

p-kos-um i1-sweep-ths ‘I sweep sth’

me-m-‰Ł'on-am pv-i1-send-ths ‘I send sth’

p-xazi-um i1-prepare-ths ‘I prepare sth’

p-c’ar-um i1-write-ths ‘I write sth’

Some transitive verbs which employ the valency operator o-

b-o-rd-am i1-tr-raise-ths ‘I raise sb’

b-o-rg-am i1-tr-plant-ths ‘I plant sth’

b-o-gzal-em i1-tr-send-ths ‘I send sb’

b-o-cil-am i1-tr-marry-ths ‘I marry him (to a girl)’

go-b-o-kt-am pv-i1-tr-turn-ths ‘I turn sth’

Causative verb forms contain the vowel o- by default:

(13) o-mt’-in-am-s / *mt’-in-am-s

tr-flee-caus-ths-i3s

‘he makes it flee, he chases it away’ (Dum67.XXIII.8)

In examples such as the following, the fact that the middle marker is

added to a form devoid of any explicit valency marking can be viewed as

evidence that the direction of the derivation is from the non-middle to the

middle:

3. Thus, the term ‘‘middle’’ is not taken here in the sense it bears in traditionalGeorgian grammar, for instance in Tschenkeli (1958, Lesson 28: ‘‘Mittelverben’’).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



non-middle > middle

p-xazi-um b-i-xazi-am

i1-prepare-ths

‘I prepare something’

i1-mid-prepare-ths

‘I prepare myself ’

In contrast, when the transitive verb shows the vowel o-, the fact that

both forms are of equal morphological complexity means that there is no

way to posit a direction of derivation on purely morphological grounds:

non-middle < > middle

b-o-cil-am b-i-cil-e

i1-tr-marry-ths

‘I marry him (to a girl)’

i1-mid-marry-ths

‘I am getting married (to a girl)’

Throughout the present article, I will use the terms ‘‘base verb’’ to refer

to a non-middle verb possessing a middle counterpart, and ‘‘base subject’’

to refer to the subject of a base verb. These terms have no theoretical

significance, since a ‘‘base’’ verb like bocilam is equal in morphological

complexity to its middle counterpart bicile, and therefore is not mor-

phologically more basic. The terms ‘‘base verb’’ and ‘‘base subject’’ are

used merely for simplicity of description.

The valency operators i/u- are used, among other things, to derive ap-

plicative forms. The applicative construction is illustrated in (14b) below,

and can be compared with (14a), the corresponding non-derived con-

struction. The applicative argument Xasani is in the dative and is cross-

referenced by a Set II a‰x (u-).4 It may be considered as a core argument

on the basis that it is cross-referenced on the verb.

(14) a. Hemu-k oxoi k’od-um-s.

demp-erg house build-ths-i3s

‘He builds a house.’

b. Hemu-k Xasani-s oxoi u-k’od-um-s.

demp-erg Hasan-dat house ii3.appl-build-ths-i3s

‘He builds a house for Hasan.’

4. The applicative construction in Laz (except in the Ardesen variety) and in theother South Caucasian languages di¤ers from prototypical applicatives in thatthe applicative argument is not case-marked in the same way as O (Lacroix2007; Peterson 2007: 1). The operator u- as illustrated in example (14b) isknown in the literature on Kartvelian languages as ‘‘objective version’’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


172 Rene Lacroix

The applicative marker is realized as u- when the applicative argument

is third person and i- when it is first or second person:

m-i-k’od-um-s

g-i-k’od-um-s

u-k’od-um-s

ii1-appl-build-ths-i3s


ii3.appl-build-ths-i3s

‘he builds it for me’

‘he builds it for you’

‘he builds it for him’

The applicative argument may express semantic roles such as benefi-

ciary, maleficiary, recipient, allative and possessor.

The valency operator o- also is used to derive applicatives. Although it

is homonymous with the o- indicating transitive verb forms (see above),

the two must be distinguished on formal grounds. Compare sentence (15a),

which involves a non-derived intransitive verb (‰Łant’u), and (15b), where

the same root appears in an applicative verb form, marked by the operator

o-. The applicative argument is nek’na-s ‘door-dat’.

(15) a. Xasani dido ora-s ‰abuni ‰Łan-t’u.

Hasan much time-dat sick lie-impft.i3s

‘Hasan was lying sick for a long time.’ (Z'.5.8)

b. Didi kva n-o-‰Łan-t’u nek’na-s.

big stone pv-appl-lie-impft.i3s door-dat

‘There was a big stone against the door.’ (K’art’72.166.6)

Two di¤erences can be pointed out between o- indicating transitive verb

forms and o- indicating the applicative. First, verb forms with transitive

o- (other than causatives) cannot be shown to derive from simpler forms.

By contrast, verb forms like no‰Łant’u in (15b) result from the addition of

o- to a more basic verb. Secondly, transitive o- indicates the presence of an

absolutive argument (the object), while applicative o- indicates the presence

of a dative argument (the applicative argument).

The valency operator a- is used, among other things, to derive applica-

tives from middle verbs. Compare (12b) above with (16), which involves

the addition of a dative argument, bic’is ‘boy’, bearing the semantic role

of maleficiary:

(16) Bic’i-s ixi-te nek’na gy-a-nk’ol-u.

boy-dat wind-instr door pv-appl.mid-close-aor.i3s

‘Because of the wind, the door closed on the boy.’

4. Verb classes

Finite verb forms in Laz may be classified into two broad groups, accord-

ing to the shape of the Set I third person singular su‰x: Class 1 verbs take

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



the su‰x -s, while Class 2 verbs take -n. This classification is also reflected

in the shape of the thematic su‰xes: Class 1 verbs either take one of the

thematic su‰xes -am, -em, -im, -om, -um or -mer or take no thematic su‰x

at all, while Class 2 verbs take the thematic su‰xes -er, -ir, -ur (/r/ drops

before a su‰x beginning with /n/).

This purely morphological classification has syntactic correlates: while

Class 1 contains both transitive and intransitive verbs, Class 2 verbs are

exclusively intransitive. Most intransitive Class 1 verbs take an ergative

subject, while most Class 2 verbs take an absolutive subject. This is

summarized in the chart below.

Table 2. Characteristics of Class 1 and 2 verbs

Class 1 Class 2

Set I third singular su‰x -s -n

thematic su‰xes -Vm, -mer, -Ø -Vr

transitivity tr. and intr. intr.

case of the subject mostly ergative mostly absolutive

Class 2 is made up primarily of middle verbs. Some examples of non-

middle Class 2 verbs are: -ul- ‘go’, illustrated in example (46a), -x- ‘be

sitting’, -dg- ‘be standing’ and -'ur- ‘die’. Such verbs do not take any

valency operator.

5. Classification of middle verb forms

Morphologically, middle verb forms may belong to Class 1 or Class 2.

I illustrate this with the root -kun- (þpreverb dolo-) ‘put on (clothes)’.

From this root, a ditransitive verb may be formed:

(17) Bozo-k bee-s porca ko-dol-o-kun-am-s.

girl-erg child-dat shirt pv-pv-tr-put.on-ths-i3s

‘The girl dresses the child in the shirt.’

The presence of the Set I third person singular su‰x -s and the thematic

su‰x -am indicates that this verb belongs to Class 1. We also see that this

verb exhibits the valency operator o- characteristic of transitive verbs (see

section 3).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


174 Rene Lacroix

To mark coreferentiality between the agent and the recipient, the verb

takes the valency operator i-, which replaces the operator o-. The resulting

middle verb remains a Class 1 transitive verb:

(18) Bee-k porca ko-dol-i-kun-am-s

child-erg shirt pv-pv-mid-put.on-ths-i3s

‘The child puts the shirt on.’

An intransitive verb with anticausative reading may be formed from

the same root. The former object porca ‘shirt’ appears in subject position.

No agent is involved:

(19) Ha porca va dol-i-kun-e-n.

demd shirt neg pv-mid-put.on-ths-i3s

‘This shirt cannot be put on [because it is dirty, too small, etc].’

The presence of the Set I third person singular su‰x -n and the thematic

su‰x -e indicates that this verb belongs to Class 2.

In the remainder of this section, I present the di¤erent uses of the middle

marker i-, examining first Class 1, then Class 2 verbs.

5.1. Class 1 middle verb forms

5.1.1. Subject-Object coreference

In the subject-object coreference construction, the subject of the middle

verb undergoes the same process as the object of the base verb, but is at

the same time the initiator of the process denoted by the verb.

Firstly, we find autocausative verbs (Creissels 2006: 29). One example

is given in (20b) below. As in the remainder of this article, the correspond-

ing non-middle construction is illustrated in the a. sentence. The subject of

(20b), nana ‘mother’, cumulates the roles represented by the subject and

the object of (20a).

(20) a. Nana-k bee-musi o-n‰Łi-am-s.

mother-erg child-poss3s tr-put.to.bed-ths-i3s

‘The mother puts her child to bed.’5

5. In such examples, it is often di‰cult to find a satisfactory English gloss for theverbal root: a choice must be made between ‘‘simple’’ and ‘‘causative’’ glosses(‘go to bed’ and ‘put to bed’) which is not reflected in the Laz root itself. Thegloss used in this example has been chosen arbitrarily.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



b. Nana i-n‰Ł i-s.

mother mid-put.to.bed-i3s

‘The mother goes to bed.’

Both verbs belong to Class 1, as indicated by the pronominal su‰x -s

and the thematic su‰xes. The verb in (20a) is transitive: this is indicated

by the operator o-. The verb in (20b) is intransitive.

Other examples are: eysels ‘he gets up’ (compare the corresponding

non-middle yoselams ‘he makes him get up’); oxink’ans ‘he moves about’

(cp. oxonk’anams ‘he moves sth’).

Although the autocausative verbs illustrated above generally appear

with an absolutive subject, there are some rare occurrences of ergative

subjects in the corpus of published Laz texts. One of these is given in (21)

below. The use of the ergative in this example, as well as in (20b) above, is

not accepted by my informant.

(21) Nana-musi-k e-y-sel-u.

mother-poss3s-erg pv-mid-stand.up-aor.i3s

‘His mother stood up.’ (Z'.102.6)

As further instances of the subject-object coreference construction, we

find body care verbs:

(22) a. Bozo-k bee bon-um-s do cxon-um-s.

girl-erg child wash-ths-i3s and comb-ths-i3s

‘The girl washes and combs the child.’

b. Bee-k i-bon-s do i-cxon-s.

child-erg mid-wash-i3s and mid-comb-i3s

‘The child washes [himself ] and combs his hair.’

As we will see in section 5.1.3, body care verbs may also be used with

an object NP designating the body part being acted upon. Autocausative

and body care verbs belong to ‘‘body actions’’, a typical sub-category of

the middle (Kemmer 1993: 53).

Coreferentiality between the subject and the object of a transitive verb

may also be expressed through the reflexive NP ti-ckimi ‘my head’ (ti-skani

‘your head’, etc.):6

6. In Kartvelology, this kind of construction is known as ‘‘tavization’’ (fromGeorgian tavi ‘head’). It has been discussed on the basis of Georgian material(see, among others, Braithwaite 1973). The details of this construction in Laz,however, require further research.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


176 Rene Lacroix

(23) K’oci-k ti-musi il-om-s.

man-erg head-poss3s kill-ths-i3s

‘The man kills himself (lit.: his head).’

Thus, Laz appears to be a two-form language in Kemmer’s sense in

that it has a reflexive NP in addition to the middle marker i-.7 The fact

that the use of the reflexive NP in object position appears almost exclu-

sively with expressions meaning ‘kill oneself ’ is consistent with Kemmer’s

observation that in two-form languages, the heavier marker (in this case

the reflexive NP) is used with reflexive verbs which involve high distin-

guishability of participants, while the lighter marker (operator i-) appears

with verbs with a lower degree of distinguishability of participants.

Both the reflexive NP and the operator i- may appear in the same con-

struction, as shown in (24) below. This construction can be analysed as an

example of ‘‘object possession’’ (see 5.1.3).

(24) Ti-ckimi do-b-i-'ur-in-am.

head-poss1s pv-i1-mid-die-caus-ths

‘I will kill myself.’ (Dum67.XXIX.4)

5.1.2. Subject-Dative coreference

As we have seen in section 3, the beneficiary may be expressed as an appli-

cative argument (Xasanis in ex. 25).

(25) Hemu-k Xasani-s oxoi u-k’od-um-s.

demp-erg Hasan-dat house ii3.appl-build-ths-i3s

‘He builds a house for Hasan.’

When the subject is coreferential with the applicative argument, the middle

marker i- appears on the verb and replaces any other valency operator:

(26) Hemu-k oxoi i-k’od-um-s.

demp-erg house mid-build-ths-i3s

‘He builds a house for himself.’

(27) Ma oxoi b-i-k’od-um.

1s house i1-mid-build-ths

‘I build a house for myself.’

7. Two-form languages are opposed to one-form languages, where the reflexiveand reciprocal markers are morphologically identical to the middle marker.An example of the reciprocal pronoun artikati is given in (57).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In this construction, the subject cumulates the semantic roles of agent

and beneficiary, expressed in the corresponding applicative construction

(25) by the subject and the applicative argument.

The constructions in (25) and (26) respectively can be schematized as

follows (‘‘Appl’’ stands for ‘‘applicative argument’’):

Table 3. Applicative construction

Syntactic level Aerg Oabs Appldat

Semantic level agent theme beneficiary

Table 4. Middle construction

Syntactic level Aerg Oabs

Semantic level agentþ beneficiary theme

Dative core arguments may also appear in non-applicative construc-

tions, as for instance with the ditransitive verb dolo-o-kun- ‘put on (clothes)’

illustrated in (17) above. Another example is mo-o-k’id- ‘hang sth on sth’:8

(28) a. Xasani-k cxomi cxeni-s m-o-k’id-am-s.

Hasan-erg fish horse-dat pv-tr-hang-ths-i3s

‘Hasan hangs the fish on the [back of the] horse.’

In this example, the dative argument cxeni ‘horse’ fulfils the role of goal.

Correspondingly, the subject of the related middle construction cumulates

the semantic roles of agent and goal:

b. Xasani-k cxomi mo-i-k’id-am-s.

Hasan-erg fish pv-mid-hang-ths-i3s

‘Hasan hangs the fish on his own back.’

Another example is Ar c’uvali dik’a kec’ib'it ‘spill under yourself a sack

of wheat’ (Dum67.II.15).

8. The fact that the dative argument of these verbs belongs to the core becomesclear in the first or second person, where it is overtly cross-referenced on theverb. When citing verbs, I give the preverb (if any), the valency operator (ifany) and the root. In mo-o-k’id, for instance, mo- is the preverb, o- the operatorand -k’id- the root.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


178 Rene Lacroix

The body part or item of clothing towards which the action is directed

may be added in the construction as a dative oblique:9

(29) Xasani-k cxomi k’ap’ula-s mo-i-k’id-am-s.

Hasan-erg fish back-dat pv-mid-hang-ths-i3s

‘Hasan hangs the fish on his own back.’

(30) T’abak’a do mendil

snu¤box and handkerchief

yele'i-s ‰Łeb-epe-s dol-i-du-mer-nan.

vest-gen pocket-pl-dat pv-mid-put-ths-i3p

‘They put into the pockets of their vest their snu¤box and their

handkerchief.’ (Dum37.128.6)

Subject-dative coreference also includes reciprocal events. Sentence (31a)

illustrates the verb ela-purcin- ‘whisper’ in its basic form; (31b) gives the

same verb in the applicative derivation; (31c) illustrates the corresponding

middle form.

(31) a. Bee-pe-k muntxa ela-purcin-am-an.

child-pl-erg something pv-whisper-ths-i3p

‘The children whisper something.’

b. Bee-pe-k bozo-pe-s muntxa el-u-purcin-am-an.

child-pl-erg girl-pl-dat something pv-ii3.appl-whisper-ths-i3p

‘The children whisper something to the girls.’

c. Bee-pe-k muntxa el-i-purcin-am-an.

child-pl-erg something pv-mid-whisper-ths-i3p

‘The children whisper something to each other.’

Coreference between the subject and a dative complement may also be

expressed by the reflexive NP in the applicative argument position:

(32) K’oci-k ti-musi-s u-c’u-me-s-ki. . .

man-erg head-poss3s-dat ii3.appl-tell-ths-i3s-comp

‘The man says to himself: . . .’

9. Such examples could alternatively be analysed as instances of oblique posses-sion constructions, whereby the valency operator i- indicates that a relation-ship of possession exists between the subject and the dative oblique (comparewith the object possession construction in section 5.1.3).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The translation of some of the examples above would involve a middle

verb form in French: Je me construis une maison (27), Les enfants se

chuchotent quelque chose (31c).

5.1.3. Object possession

The object possession construction is illustrated in (33b) below. The subject

hemuk is the possessor of the object pantoloni.

(33) a. Hemu-k bere zin‰Łir-epe-te k’o-um-s.

demp-erg child chain-pl-instr bind-ths-i3s

‘He ties the child up with chains.’

b. Hemu-k pantoloni gel-i-k’o-am-s.

demp-erg trousers pv-mid-bind-ths-i3s

‘He ties his [own] trousers up.’

This is a type of ‘‘external possession’’ (or ‘‘possessor raising’’) con-

struction: the possessor is not expressed as a possessive determiner in the

possessed NP (‘his trousers’), but appears as a distinct NP, and the rela-

tion of possession follows from the presence of the operator i-.

As a subtype of the object possession construction, we find body care

verbs in constructions where the body part is overtly expressed as an

object NP.

(34) a. see (22a) above

b. Bozo-k xua i-bon-s.

girl-erg body mid-wash-i3s

‘The girl washes.’

Such uses of a middle marker are known in other languages as well (cf.

French il se lave les mains ‘he is washing his hands’).

Although subjects of body care verbs normally appear in the ergative,

we find the following two exceptional sentences:

(35) Bozo nuk’u d-i-bon-u.

girl face pv-mid-wash-aor.i3s

‘The girl washed her mouth.’ (Z'.63.6)

(36) K’ulan-epe nuk’u i-bond-es.

girl-pl face mid-wash-aor.i3p

‘The girls washed their faces.’ (Z'.151.17)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


180 Rene Lacroix

These are the only examples of the ‘‘double-absolutive’’ construction I

have found in my corpus. They are not accepted by my informant, who

uses an ergative subject instead. I propose a possible explanation for this

exceptional pattern below (5.3).

Other examples of the object possession construction are: K’iti mck’oni

nik’vatu ‘He cut his little finger’ (Z'.5.30); Peskirite nuk’u bikosare ‘I will

wipe my face with a towel’ (Z'.60.20); Divik k’ibirep dilasirudort’un ‘The

giant had sharpened his teeth’ (Dum37.85.12).

The object possession construction appears primarily with body parts,

clothes and weapons, but it seems also to be possible with other kinds

of objects, for instance in k’ocik araba dicxu ‘the man washed his car’

(elicited example).

5.1.4. Antipassive

Example (37) illustrates the construction of the ditransitive verb do-o-gur-

‘teach sth to sb’.

(37) Padisahi-k zur c’ut’al-epe-s zanaxat’ d-o-gur-am-t’u.

sultan-erg two little-pl-dat trade pv-tr-teach-ths-impft.i3s

‘The sultan taught a trade to his two youngest sons.’ (Dum37.1.1)

The corresponding middle verb do-i-gur- ‘learn sth’ (ex. 38) illustrates

the subject-dative coreference construction (see section 5.1.2). This verb is

monotransitive.

(38) Bere-k ir sey ko-d-i-gur-u-dort’un.

child-erg all thing pv-pv-mid-learn-i3s-pperf

‘The boy had learnt everything.’ (Dum37.56.12)

Example (39) shows another use of the middle verb do-i-gur-. The sub-

ject (hentebe) is in the absolutive and the verb takes no object.

(39) Hentebe i-gur-am-t’es Amerik’a-s.

demp.pl mid-learn-ths-impft.i3p America-dat

‘They studied in America.’ (Z'.103.24)

Comparison of (38) and (39) reveals an antipassive pattern: in (39), the

object of the transitive construction in (38) disappears and the subject

stands in the absolutive. The construction thus becomes intransitive.

The construction in (39) cannot, however, be analysed as a prototypical

antipassive (Dixon and Aikhenvald 2000: 9), as no overt marking distin-

guishes the antipassive from the corresponding transitive construction:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



both are marked by the middle. I do not have enough information to state

whether the base object can appear in the antipassive construction or not.

The transitive construction in (38) and the antipassive construction in

(39) can be schematized as follows:

Subject-dative coreference

Syntax Aerg Oabs

Semantics agentþ beneficiary theme

Antipassive construction

Syntax Sabs

Semantics agentþ beneficiary

Middle voice markers are used in antipassive constructions in other lan-

guages as well, for instance in Spanish and, more marginally, in French

(Creissels 2006: 34).

5.1.5. Aspectual distinctions

The Class 1 intransitive verbs -k’iy- ‘crow’, -‘ur- ‘scream’ and -m'or-

‘scream’, which take an ergative subject,10 appear with the operator i- when

used perfectively (ex. 40), and without the operator when used imperfec-

tively (ex. 41). Ex. (42) shows that a form with the operator i- cannot be

used when a durative adverbial occurs in the sentence. Note that in other

languages as well, the presence of a middle marker has aspectual implica-

tions (Creissels 2006: 31).

(40) Divi-k ar zor ko-d-i-ur-u.

giant-erg one strongly pv-pv-mid-scream-aor.i3s

‘The giant uttered a loud shout.’ (Dum37.81.7)

10. The only exception I have come across is C’ic’ila ur-am-t’u ‘the snake wasscreaming’ (Z'.125.29). In the very next sentence in the same text, however,we find this verb used with an ergative subject (c’ic’ilak-na uramt’u yeri ‘theplace where the snake was screaming’).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


182 Rene Lacroix

(41) ‰Ł inaze n-o-yon-am-t’a-n-si

corpse pv-tr-take-ths-subj-i3p-when

oxor‰Łal-epe-k dido ur-am-an.

woman-pl-erg much scream-ths-i3p

‘When people take the corpse [to the cemetery], the women scream

a lot.’ (Z'.10.26)

(42) Sum saat’i-s k’iy-asen. *d-i-k’iy-asen

three hour-dat crow-fut.i3s pv-mid-crow-fut.i3s

‘He will crow for three hours.’

5.1.6. Lexicalized items

Two types of middle verb form may be considered to be lexicalized. First,

we find middle verbs which do not have any corresponding non-derived

form. Such ‘‘frozen’’ middle verbs are similar to ‘‘deponents’’ or ‘‘media

tantum’’ in the grammar of classical languages. They include transitive

and intransitive verbs. The list below is not exhaustive.

b-i-cin-om i1-mid-know-ths ‘I know sb, I recognize sb’b-i-pxor i1-mid-eat ‘I eat sth’b-i-Ø-mer i1-mid-take-ths ‘I take sth away’b-i-p’aramit-am i1-mid-talk-ths ‘I talk, I say sth’b-i-calis-am i1-mid-work-ths ‘I work’b-i-xoron-am i1-mid-dance-ths ‘I dance’b-i-bgar i1-mid-cry ‘I cry’me-b-i-xi pv-i1-mid-steal ‘I steal sth’

Intransitive deponents generally take an ergative subject (ex. 43). Here

again, however, we find rare examples involving absolutive subjects (ex. 44).

(43) Bere-pe-musi-k i-bgar-nan.

child-pl-poss3s-erg mid-cry-i3p

‘Her children cry.’ (Z'.81.7)

(44) Padime i-bgar-s.

Padime mid-cry-i3s

‘Padime cries.’ (Z'.63.35)

My informant uses an ergative subject in example (44).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Secondly, we find middle verbs whose meaning cannot be straight-

forwardly derived from the related non-middle form: -k’itx- ‘ask’ vs. i-

k’itx- ‘study; read’; -‰Łoxon- ‘be called’ (e.g. Ma C’ip’ut’ina m‰Łoxons ‘my

name is C’ip’ut’ina’ [Dum67.III.10]) vs. i-‰Łox- ‘shout’.

The presence of the middle marker i- on ‘eat’, ‘take away’, ‘steal’ and

‘read’ is not surprising, as these verbs necessarily have an autobenefactive

reading (Creissels 2006: 31). Neither is it surprising on ‘know, recognize’

(‘cognition middle’, Kemmer 1993: 127); ‘shout’, ‘cry’ (‘emotion middle’,

Kemmer 1993: 130); and ‘dance’, which bears some resemblance to ‘‘non-

translational motion’’ verbs (Kemmer 1993: 56).

5.2. Class 2 middle verb forms

The subject of the middle verbs examined so far corresponds to the base

subject, if any, and these verbs are mostly agentive (an example of a non-

agentive Class 1 verb is i-bgar- ‘cry’). We are now going to consider Class

2 middle verbs, which include primarily non-agentive verbs (anticausa-

tives, among others).

Morphologically, Class 2 middle verbs are more homogeneous than

Class 1 middle verbs, as they all take the thematic su‰x -er.

5.2.1. Passive

The passive construction, illustrated in (45b) below, is rare in my corpus

compared to other uses of i-. The subject of the passive verb form corre-

sponds to the base object.

(45) a. Bic’i-k sum k’oc do-yl-u.

boy-erg three man pv-kill-aor.i3s

‘The boy killed three men.’ (Z'.144.2)

b. Tabi baba-musi d-i-yl-u.

of.course father-poss3s pv-mid-kill-aor.i3s

‘[The boy fired when his father stood up.] Of course, his father

got killed.’ (Dum67.XXXIV.8)

The corresponding present form is d-i-yl-e-n pv-med-kill-ths-i3s.

The base subject (i.e. the agent) cannot be expressed in the passive con-

struction. Its presence, however, is implicit. This point distinguishes the

passive from the anticausative, which does not make reference to any

agent (Creissels 2006: 31, where the equivalent term ‘‘decausatif ’’ is used).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


184 Rene Lacroix

Formally, the passive in Laz is not distinct from other Class 2 middle

verb forms. The functions of i- could thus have been labelled ‘‘medio-

passive’’, but since the passive is quite restricted in Laz, I have prefered

to use simply ‘‘middle’’.

The history of Romance and Slavic languages attests a scenario

whereby a middle marker develops passive uses, as in se invitaron muchos

amigos ‘many friends were invited’, which involves the middle marker se

(Creissels 2006: 32). Although the history of the operator i- in Laz is

not known, it may be suggested that this language underwent a similar

development, whereby passive uses of i- developed secondarily out of

middle uses.

5.2.2. Impersonal middle

By impersonal middle, I refer to a middle construction with generic seman-

tics derived from an intransitive verb:

(46) a. K’oci mezare-sa mend-ul-u-n.

man tomb-all pv-go-ths-i3s

‘The man goes to the tomb.’

b. Hac’ineri mc’ima do ixi-s

contemporary rain and wind-dat

mezare-sa mend-i-l-in-e-n-i?

tomb-all pv-mid-go-caus-ths-i3s-int

‘Do people go to the tomb when it is raining and windy, as it is

now?’ (K’art’93.88.17)

The causative su‰x -in appears in some impersonal middle verbs. In

these verbs, however, it does not retain its causative function (see Lacroix

2009: 481–482).

5.2.3. Anticausative

The anticausative, illustrated in (47b), is one of the most frequent uses of

the operator i-. It corresponds to a situation which does not involve an

agent (Creissels 2006: 31) and di¤ers in this respect from the passive, which

semantically does involve an agent (albeit one which is not expressed).

(47) a. Bozo-k nek’na ge-nk’ol-um-s.

girl-erg door pv-close-ths-i3s

‘The girl closes the door.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



b. Nek’na ge-i-nk’ol-e-n.

door pv-mid-close-ths-i3s

‘The door closes.’

The cause may be expressed as an oblique in the instrumental case:

(48) Hea mzoa-te i-c’v-e-n.

demp sun-instr mid-burn-ths-i3s

‘It burns because of the sun.’

(49) Mera'i-te d-i-zabun-u-doren.

anxiety-instr pv-mid-get.sick-aor.i3s-evd

‘He became sick with anxiety.’ (Dum67.I.133)

The anticausative may also have a deontic reading, whereby it denotes

a general event reflecting the social norm:

(50) On‰Ł'ore ren-ya, var i-tkv-e-n-ya.

shame be.i3s-quot neg mid-say-ths-i3s-quot

‘It is a shame, it cannot be said.’ (Z'.66.22)

An anticausative form used with negation implies a nuance of deontic

(51) or physical (52) impossibility:

(51) He p’et’emezi ckva va i-ck’om-e-t’u.

demd treacle more neg mid-eat-ths-impft.i3s

‘[A mouse drowned in the treacle.] This treacle could not be eaten

any more.’ (Z'.56.4)

(52) Badi var es-i-'-e-t’u.

old.man neg pv-mid-take.out-ths-impft.i3s

‘The old man could not be taken out [of the well he had fallen into].’

(Dum67.LIII.25)

Under the term facilitative, Kemmer (1993) includes among other things

‘‘expressions of intrinsic ability of an object to undergo a particular process’’

(p. 147). Example (53) could be subsumed under the facilitative reading:

(53) K’ule-sen mteli Mp’oli i-‰ ir-e-t’u-doren.

tower-abl all Istanbul mid-see-ths-impft.i3s-evd

‘From the tower you could see all of Istanbul.’ (Dum67.XLVII.4)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


186 Rene Lacroix

Transitive Class 1 verbs may be derived from adjectives; the meaning

of such verbs can be paraphrased by ‘make something A’ (where A ¼adjective; ex. 54a); these verbs correspond to Class 2 anticausatives with

inchoative semantics (‘become A’; ex. 54b).

(54) a. Mzoa-k camasur-epe o-kcan-am-s.

sun-erg linen-pl tr-whiten-ths-i3s

‘The sun fades the linen.’

b. K’at’a yei d-i-kcan-e-n.

every place pv-mid-whiten-ths-i3s

‘[When it snows,] all becomes white.’

Deadjectival Class 1 verbs always take the operator o-.

Other examples of anticausative verbs are: Mumiten iguben ‘It cooks

over a candle’ (Dum67.VI.29); Orsi n‰Łumu stey dipuncxoludoren ‘The

anvil crumbled away like salt’ (Dum67.XLIII.21); Namazi dicoden ‘The

prayer finishes’.

Subjects of passive and anticausative verbs are non-agentive. They are

generally in the absolutive. One exception must be mentioned: the verb

i-xel- ‘to be glad’ takes an ergative subject (oxor‰Ła-k ixeludoren ‘the

woman was glad’ [Dum67.VI.51]; the present tense is i-xel-e-n and the

corresponding transitive form is o-xel-am-s ‘he makes him happy’). I turn

now to Class 2 middle verbs with agentive subjects.

5.2.4. Subject-Object coreference

We have seen above examples of Class 1 autocausative verbs. Class 2 also

includes some autocausatives:

(55) a. Badi-k nk’ola g-o-kt-am-s.

old.man-erg key pv-tr-turn-ths-i3s

‘The old man turns the key.’

b. Badi go-y-kt-e-n.

old.man pv-mid-turn-ths-i3s

‘The old man turns back.’

Other examples are: K’oci itk’ocen ‘The man leaps’; Hea mt’ut’as dixven

‘He buries himself in ash’; K’ules kenayis kamik’idudoren ‘He hung from

the edge of the tower’ (Dum67.XLVII.14).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Among the Class 2 agentive verbs, we find a verb with reciprocal

semantics: ‘hit each other, fight’.

(56) a. Xasani-k Ali il-om-s.

Hasan-erg Ali hit-ths-i3s

‘Hasan hits Ali.’

b. Hemtepe i-il-e-nan.

demp.pl mid-hit-ths-i3p

‘They are fighting.’

The same verb may be used transitively with the reciprocal pronoun

artikati in object position. In this case, the verb form belongs to Class 1:

(57) Sum bere-k biga-pe-te artikati il-om-t’es.

three child-erg stick-pl-instr recp hit-ths-impft.i3p

‘Three guys were hitting each other with sticks.’ (Dum67.I.415)

Lastly, the verbs ‘prepare oneself ’ and ‘marry (intr.)’ admit two forms:

one in Class 1, bixaziam ‘I prepare myself ’, bicil ‘I marry (a girl)’

(Dum67.VIII note 14), and one in Class 2, bixazie ‘I prepare myself ’,

bicile ‘I marry (a girl)’.

Although in general Class 2 agentive verbs take an absolutive subject,

we do find some rare occurrences of an ergative subject (ex. 58). I will

return to this point below (5.3).

(58) Orc’-ayi-k go-y-kt-u.

Orc’i-sfx-erg pv-mid-turn-aor.i3s

‘The man from Orc’i turned back.’ (Dum67.LI.44)

5.2.5. ‘look round’

Another example of a Class 2 agentive verb is i-ck’- ‘look round’. The

corresponding non-middle Class 1 verb ‘look at’ takes an ergative subject

and a dative complement:

(59) a. Bere-k bozo-s mend-o-ck’e-s.

child-erg girl-dat pv-tr-look-i3s

‘The child looks at the girl.’

The middle form is used when the act of looking is not directed towards a

specific object. The subject is still in the ergative. No complement appears

in the construction.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


188 Rene Lacroix

b. Xo‰Ła-k hekolhakole i-ck’-e-t’u.

hodja-erg here.and.there mid-look-ths-impft.i3s

‘The hodja was looking here and there / was looking round.’

(Dum67.XXX.7)

The present tense form is i-ck’-e-n mid-look-ths-i3s. This is one of the

few examples of Class 2 verbs which always appear with an ergative sub-

ject. Unlike with the verbs illustrated under ‘‘subject-object coreference’’

and ‘‘subject-dative coreference’’, here the subject does not cumulate two

semantic roles.

The e¤ect of the middle marker on this verb is reminiscent of the anti-

passive. Syntactically, the second core argument of the verb in (59a) is

deleted in (59b). However, the antipassive demotes the object, which is

not the case here, since the demoted argument is marked by the dative.

Furthermore, the subject remains in the ergative. Semantically, in the

situation described in (59b), the thing being perceived is unimportant – a

feature usually associated with the antipassive (Givon 2001: 168).

Some languages attest a scenario whereby middle markers develop

antipassive uses (Creissels 2006: 40).

5.2.6. Lexicalized item

One final example of a Class 2 verb with an agentive subject is the lexical-

ized verb i-mt’- ‘flee’ (60a). This verb lacks a corresponding non-derived

form; the corresponding transitive verb is morphologically a causative

(60b):

(60) a. K’oci i-mt’-e-n.

man mid-flee-ths-i3s

‘The man runs away.’ (Z'.91.7)

b. Xasani-k mc’a‰Ł i o-mt’-in-am-s.

Hasan-erg fly tr-flee-caus-ths-i3s

‘Hasan chases away the flies.’

The verb ‘flee’ is attested with middle markers in other languages as

well (see Kemmer 1993 on ‘‘translational motion’’ verbs [p. 56]).

5.3. Summary

We have seen that middle verb forms may belong to Class 1 or Class 2.

The subject of a Class 1 middle verb corresponds to the base subject. Class

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1 middle verbs are in general agentive. Subject-dative coreference (5.1.2)

and object possession (5.1.3) are transitive constructions, whereas the

subject-object coreference construction (5.1.1) and the antipassive (5.1.4)

are intransitive. A few intransitive verbs include the operator i- when

used in a perfective context (5.1.5).

The subject of passive (5.2.1) and anticausative verbs (5.2.3), which

belong to Class 2, corresponds to the base object, the base subject being

backgrounded. Such verbs are non-agentive. Class 2 also contains some

agentive verbs (5.2.2, 5.2.4 and 5.2.5). All Class 2 middle verbs are

intransitive.

Both Class 1 and Class 2 middle verbs include lexicalized items.

As we have seen, body action verbs are found in Class 1, which is

mostly restricted to agentive verbs, and Class 2, which comprises for the

most part non-agentive verbs; secondly, there is some variation between

the use of the ergative and the absolutive with body care (5.1.3), auto-

causative (5.1.1, 5.2.4) and media tantum verbs (5.1.6); finally, the verbs

‘prepare oneself ’ and ‘marry (intr.)’ admit both Class 1 and Class 2 forms.

All these facts correlate with Kemmer’s observation that, from a semantic

point of view, body action verbs are intermediate between prototypical

two-participant (agentive) and prototypical one-participant (non-agentive)

events (Kemmer 1993: 73).

The functions of the middle vowel i- in the two verb classes are

summarized below.

Table 5. Summary of the functions of the middle marker i-

Class 1 middle verbs Class 2 middle verbs

subject-object coreferencesubject-dative coreferenceobject possessionantipassivesensitivity to aspectlexicalized items

passiveimpersonal middleanticausativesubject-object coreference‘look round’lexicalized item

These uses correspond to events typically expressed by middle markers in

languages possessing a middle system.

The wide range of uses covered by the middle vowel i- and the number

of lexicalized items (especially in Class 1) indicate that this operator is of

ancient origin. Furthermore, we have seen that Laz middle verbs may be

used transitively. In this respect, Kemmer (1993: 34) notes: ‘‘Transitive

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


190 Rene Lacroix

middle marker verbs arise diachronically when the middle marker has

been grammaticalized to the point where it no longer denotes a referential

entity, but only marks middle semantics’’.11

Indeed, comparative data show that the operator i- is ancient. It is

found in the other languages of the family – Mingrelian, Georgian and

even Svan, the most remote member, which is said to have branched o¤

from Proto-Kartvelian no later than the last centuries of the third millen-

nium B.C. (Klimov 1998: IX; this approximation is based on Swadesh’s

method of glottochronology). The operator i- is reconstructed for Proto-

Kartvelian (Fahnrich 2007: 209). Its position immediately before the

verbal root is an indication that it became attached to the verbal root

even before the cross-referencing prefixes, which are also reconstructed

for Proto-Kartvelian.

6. Other accounts of the valency operator i-

As mentioned in the introduction, Georgian has an operator i- with

roughly the same uses as the Laz element. In descriptions of Georgian,

the operator i- used with Class 1 verbs in examples such as ‘build a house

for oneself ’ or ‘comb one’s hair’ is referred to as ‘‘subjective version’’

(Sani‰e 1953, Tschenkeli 1958, Boeder 1969, 2005, Hewitt 2005). The

‘‘subjective version’’ indicates that the subject acts ‘‘for his own benefit’’,

that the action ‘‘goes back to the subject’’ (Tschenkeli 1958: 245).

Georgian Class 2 verbs (which correspond roughly to Laz Class 2

verbs) are sometimes referred to as ‘‘passives’’ (Tschenkeli 1958, Fahnrich

1993). They are of di¤erent formal types. ‘‘Passive’’ verbs marked by the

operator i- include true passives, but also anticausative (ic’veba ‘it burns

[intr.]’, irk’veva ‘it becomes clear’), facilitative (itargmneba ‘it can be trans-

lated’) and even autocausative verbs (imaleba ‘he hides [intr.]’, in‰reva

‘he moves [intr.]’). The term passive is thus misleading: strictly speaking,

passive is only one of the values of i- in this verb class.

This presentation of Georgian has served as a model for the analysis of

the other languages of the family: Laz (Holisky 1991: 422, 438), Mingrelian

(Harris 1991c: 354, 360) and Svan (Tuite 1997: 26, 35). However, it does

11. This explanation, however, would hold only if the valency operator i- could beshown to derive from a marker denoting a referential entity, such as a reflexivemarker. Although this is the most frequent origin for middle markers, thereare other possible sources (Kemmer 1993: 197–200).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



not capture the functional unity of the operator i- in these languages. In

fact, as we have seen, the ‘‘subjective version’’ and the ‘‘passive’’ marker

i- represent one and the same morpheme, which indicates middle voice, in

the sense of Kemmer (1993). Passive uses indicated by i- may occasionally

be found in Laz, but such examples are not frequent in the corpus.

Some authors (Deeters 1930, Sani‰e 1953, Schmidt 1965, Tuite 2007,

Gurevich 2006) have pointed out the similarities between the operator i-

and the middle, especially as seen in Indo-European. However, this has

not led them to take the view that, beyond mere resemblance, i- can

indeed be considered as a marker of middle voice.

Gurevich (2006: 176) argues that ‘‘there are . . . significant di¤erences

between the Greek and Georgian situations. The Georgian parallels to

middle voice are much more heterogeneous than the Greek ones appear

to be, and more lexicalized. . . . Moreover, the Georgian middle-voice

formations cover only a portion of the semantic classes described for

Greek middle voice. . .’’. These di¤erences, however, naturally follow

from the fact that the diachronic development of middle markers is not

necessarily identical across languages (although, as shown by Kemmer

[1993], some generalizations may be drawn). Lexicalization is not a sys-

tematic process, but rather the sum of independent instances of evolution.

As a result, the exact semantic domains covered by middle markers in dif-

ferent languages need not coincide. In addition, divergences in the number

of lexicalized items may simply reflect di¤erent stages in the chronology of

the grammaticalization process.

Boeder (1969), Forest (1999) and Mac’avariani (1987), working primarily

on Georgian data, treat the middle operator i- and the operator i- used in

applicative derivation (see section 3) as the same marker, and try to give a

unified account of them. Thus, in Boeder’s view, the i- markers in (61a)

and (61b) are functionally identical.

(61) a. subject-dative coreference

i-k’od-um-s

mid-build-ths-i3s

‘he builds sth for himself ’

b. applicative

g-i-k’od-um-s


‘he builds sth for you’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


192 Rene Lacroix

Indeed, in both examples, the marker i- indicates that the action is

directed towards a beneficiary: the second person addressee in (61b) and

the agent himself in (61a). Thus, the comparison of such examples might

suggest an analysis of both instances of i- as the same marker. However,

such an analysis becomes much more di‰cult when one takes into account

the other uses of i-, in particular the passive and the anticausative: a

sentence such as ‘The door opens (goinck’en)’ has no beneficiary. Further-

more, the middle i- and the applicative i- have distinct syntactic character-

istics: the applicative adds an argument to the base construction, which is

never the case with the middle.

Thus, although there may be a historical relationship between the

middle and the applicative i-, it seems preferable to keep these two markers

distinct in a synchronic description.

7. Conclusion

In this paper, I have shown that the multiple uses of the valency operator

i- in Laz may be accounted for in a unified manner if we take the view that

this morpheme indicates middle voice.

Thus, the operator i- does not constitute a grammatical category peculiar

to Kartvelian. Its originality lies rather in its formal manifestations. The

divergences that have been noted between i- in the Kartvelian languages

and middle markers in Indo-European languages follow from more general

principles of grammaticalization.

The operator i- in Laz and, I believe, in the other Kartvelian languages

covers a functional domain which is not essentially di¤erent from that of

middle markers in accusative languages. But the fact that in Laz subjects

may be in both the ergative and in the absolutive raises questions on

the relation between ergative alignment type and middle voice which

have no equivalents in languages with case marking displaying accusative

alignment.

Givon (2001: chap. 13) distinguishes between semantic and pragmatic

voice mechanisms. Pragmatic voices ‘‘render the same semantically-transitive

event from di¤erent pragmatic perspectives. These perspectives turn out to

involve, primarily although not exclusively, the relative topicality of the

agent and patient’’ (p. 93). ‘‘Primarily pragmatic voice constructions are

those whose functional definition depends on some facets of the wider,

extra-clausal, discourse context’’ (p. 92). Such mechanisms do not a¤ect

the semantics of the predicate. Primarily pragmatic voices include the

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



passive, the antipassive and the inverse. These voices interfere with other

mechanisms which may be involved in the expression of pragmatic per-

spective (for example, constituent order). In particular it is commonly

assumed that they interfere with alignment, and that (for example) the

antipassive may have functional motivations in morphosyntactically erga-

tive languages which find no equivalent in accusative languages.

On the other hand, primarily semantic voice mechanisms operate on

argument structure at a semantic level, modifying the semantic roles a

predicate assigns to its arguments. Unlike pragmatic voices, their ‘‘func-

tional definition does not depend on entities outside the boundaries of the

event-clause’’ (p. 92). Primarily semantic voices include reflexive, reciprocal

and middle mechanisms.

Middle voice, which belongs to the latter group, has its own motiva-

tion; that is, it does not interfere with pragmatic mechanisms. Hence, there

need not be any correlation between middle voice and the alignment type

of a language. This is corroborated by the existence of a middle voice in

partially ergative languages such as Laz.

Abbreviations

abl ablative mid middle

all allative neg negation

aor aorist pl plural

appl applicative poss possessive

caus causative pperf pluperfect

comp complementizer pv preverb

dat dative quot quotative

dem demonstrative pl plural

demd demonstrative determiner pv preverb

demp demonstrative pronoun recp reciprocal

erg ergative sfx su‰x

evd evidential subj subjunctive

fut future ths thematic su‰x

gen genitive tr, tr. transitive

impft imperfective i Set I

instr instrumental ii Set II

int interrogative 1, 2, 3 first, second, third person

intr. intransitive S, P singular, plural

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


194 Rene Lacroix

For Dumezil 1937, Z'ent’i 1938 and K’art’ozia 1993, the references

accompanying the examples give the page and the line of the Laz text, so

that Dum37.10.3 means Dumezil 1937, page 10, line 3. For Dumezil 1967

and 1972, the references give the text number and the line number pro-

vided by Dumezil, so that Dum67.XXIII.8 means Dumezil 1967, text

XXIII, line 8.

References

Andrews, Peter (ed.)1989 Ethnic groups in the Republic of Turkey. Compiled and edited

with the assistance of Rudiger Benninghaus. Wiesbaden: Reichert.Aronson, Howard

1990 Georgian: A Reading Grammar. Slavica Publishers.Boeder, Winfried

2005 The South Caucasian Languages. Lingua 115(1): 5–89.Boeder, Winfried

1969 Uber die Versionen des georgischen Verbs. Folia Linguistica 2:82–152.

Boeder, Winfried2002 Syntax and morphology of polysynthesis in the Georgian verb.

In Evans, Nicholas and Sasse, Hans-Jurgen (eds.) Problems ofPolysynthesis: 87–111. Berlin: Akademie Verlag.

Boeder, Winfried2005 The South Caucasian languages. Lingua 115: 5–89.

Braithwaite, Kim1973 Case shift and verb concord in Georgian. PhD thesis. University

of Texas at Austin.Creissels, Denis

2006 Syntaxe generale, une introduction typologique. Tome 2, la phrase.Paris: Hermes Science.

Cikobava, Arnold1936 C’anuris gramat’ik’uli analizi (t’ekst’ebiturt) [A grammatical

analysis of Laz (with texts)]. Tbilisi: Sakartvelos SSR Mecniere-bata Ak’ademiis Gamomcemloba.

Deeters, Gerhard1930 Das kharthwelische Verbum: vergleichende Darstellung des Verbal-

baus der sudkaukasischen Sprachen. Leipzig: Markert und Petters.Dixon, R.M.W and Aikhenvald, Alexandra Y.

2000 Introduction. In Dixon, R.M.W and Aikhenvald, Alexandra Y.(eds.) Changing Valency. Case studies in transitivity. Cambridge:Cambridge University Press.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Dumezil, Georges1937 Contes lazes. Paris: Travaux et memoires de l’Institut d’Ethnolo-

gie, XXVII.Dumezil, Georges

1967 Documents anatoliens sur les langues et les traditions du Caucase,IV. Recits lazes (dialecte d’Arhavi). Paris: Presses Universitairesde France.

Dumezil, Georges1972 Textes en laze d’Ardesen. Bedi Kartlisa, vol. XXIX–XXX.

Fahnrich, Heinz1993 Kurze Grammatik der georgischen Sprache. Leipzig: Langenscheidt

Verlag Enzyklopadie.Fahnrich, Heinz

2007 Kartwelisches Etymologisches Worterbuch. Brill.Forest, Robert

1999 Empathie et linguistique. Presses Universitaires de France.Givon, Talmy

2001 Syntax. Volume II. John Benjamins.Gurevich, Olga

2006 Constructional Morphology: the Georgian Version. PhD thesis.University of California, Berkeley.

Harris, Alice C. (ed.)1991a The Indigenous Languages of the Caucasus. Volume I: The Kart-

velian Languages. Delmar, New York: Caravan.Harris, Alice C.

1991b Overview on the History of the Kartvelian Languages. In Harris,Alice C. (ed.) The Indigenous Languages of the Caucasus.

Harris, Alice C.1991c Mingrelian. In Harris, Alice C. (ed.) The Indigenous Languages

of the Caucasus.Hewitt, George

2005 Georgian: A Learner’s Grammar. London: Routledge.Holisky, Dee Ann.

1991 Laz. In Harris, Alice C. (ed.) The indigenous languages of theCaucasus.

Kemmer, Suzanne1993 The middle voice. John Benjamins.

Klimov, Georgij A.1998 Etymological Dictionary of the Kartvelian Languages. Mouton de

Gruyter.Kutscher, Silvia

2001 Nomen und nominales Syntagma im Lasischen. Eine deskriptiveAnalyse des Dialekts von Ardesen. Munchen: Lincom Europa(Lincom Studies in Caucasian Linguistics 17).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


196 Rene Lacroix

K’art’ozia, Guram1972 Lazuri t’ekst’ebi [Laz texts]. Tbilisi: Mecnierebata

K’art’ozia, Guram1993 Lazuri t’ekst’ebi [Laz texts]. Tbilisi: Mecniereba.

Lacroix, Rene2010 Benefactives in Laz. Submitted to Kittila, Seppo and Zuniga,

Fernando (eds), to appear, Benefactives and malefactives. Casestudies and typological perspectives, John Benjamins.

Lacroix, Rene2009 Description du dialecte laze d’Arhavi (caucasique du sud, Tur-

quie). Grammaire et textes. PhD thesis. University Lyon 2.Available at http://theses.univ-lyon2.fr/documents/lyon2/2009/lacroix_r.

Mac’avariani, Maia1987 Kcevis gramat’ik’uli k’at’egoriis semant’ik’a [Semantics of the

grammatical category of version]. Tbilisi: Mecniereba.Marr, Nicolas

1910 Grammatika canskago (lazskago) jazyka [Grammar of Can (Laz)].Materialy po jafeticeskomu jazykoznaniju 2. St. Petersburg:Akademija.

Peterson, David A.2007 Applicative Constructions. Oxford: Oxford University Press.

Q’ipsi‰e, Ioseb1939 C’anuri t’ekst’ebi [Laz texts]. Tbilisi: SSRK’ Mecnierebata

Ak’ademiis Pilialis Gamomcemloba.Schmidt, Karl Horst

1965 Indogermanisches Medium und Sataviso im Georgischen. BediKartlisa 19–20: 129–135.

Sani‰e, Ak’ak’i1953 Kartuli gramat’ik’is sapu‰vlebi, I: morpologia [The fundamentals

of Georgian grammar I: morphology]. Tbilisi: TSUG.Tschenkeli, Kita

1958 Einfuhrung in die georgische Sprache. Band I: theoretischer Teil.Zurich, Amirani Verlag.

Tuite, Kevin1997 Svan. Munchen: Lincom Europa.

Tuite, Kevin2007 Liminal morphosyntax: Georgian deponents and their kin.

Chicago Linguistics Society 39(1): 774–788.Z'ent’i, Sergi

1938 C’anuri t’ekst’ebi (arkabuli k’ilok’avi) [Laz texts (dialect of Arhavi)].Tbilisi: SSRK’ Mecnierebata Ak’ademiis Pilialis Gamomcemloba.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Argument demotion in Japhug Rgyalrong

Guillaume Jacques

1. Introduction1

Japhug (Chabao in Chinese), a Sino-Tibetan language spoken in China

(Mbarkhams county, Rngaba autonomous region, Sichuan Province), is

unusual among the languages of this family in displaying complex verbal

morphology. Alongside Tshobdun (Caodeng in Chinese), Zbu (also known

as Showu, Ribu or Rdzongmbur) and Situ (Eastern Rgyalrong), it belongs

to the so-called Rgyalrong languages.2

Like some other Sino-Tibetan languages of Sichuan, Japhug has an

ergative case marking system and a verb agreement system which indexes

two arguments (for transitive verbs) following a hierarchical pattern. No

ergative syntactic pivots are found in nominalization, complementation or

equi-NP deletion constructions.

This paper is divided into seven parts. First, we will discuss the morpho-

logical and syntactic marking of transitivity in Japhug, the basis for any

study of transitivity-changing devices in this language. Second, we will

1. I collected the data presented in this article in a series of field trips to Chinaundertaken from 2002 onwards. My main informant for this language wasChenzhen. I wish to thank Peter Austin, Gilles Authier, William Croft, HenrietteDaudey, Katharina Haude, Steven Kaye and an anonymous reviewer for theirinsightful comments, corrections and discussions. This article was completedduring my stay as a visiting scholar at the Research Centre for LinguisticTypology, LaTrobe University: I am grateful to Randy LaPolla for makingthis visit possible. The glosses generally follow the Leipzig Glossing Rules,except for the following: adv adverbializer, anticaus anticausative, apass

antipassive, cis cislocative, conj conjuction, const constative, downstr

downstream, evd evidential, genr generic, hum human, intsv intensive, inv

inverse, neu neuter (indefinite possession), nonhum non-human, n.pst non-past, stat stative, trans translocative, upstr upstream.

2. The Rgyalrong languages are themselves a sub-branch of the Qiangic branchof Sino-Tibetan, which comprises the extinct Tangut language, as well asQiang, Pumi, Muya, Queyu, Zhaba, Lavrung and Rtau. Guiqiong, Shixing,Namuyi and Ersu, generally thought to belong to this branch, should probablybe classified di¤erently (Jacques and Michaud 2011).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 12:04) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 199–226 1351 Authier_06_jacques (p. 199)

describe the generic argument marking system, which, like nominal case

marking, shows ergative alignment. Third, passive and anticausative forms

will be considered. These forms have a fairly restricted range of uses, since

‘‘generic’’ forms are employed in most cases where the passive would be

found in European languages. Fourth, we will analyse the two antipassive

prefixes found in Japhug, which are not productive, but have the interest-

ing property of distinguishing between human and non-human suppressed

patients. Fifth, we will briefly describe the de-experiencer prefix, which can

derive an intransitive verb from a transitive verb of perception. Sixth, we

will discuss the labile verbs, a small class of verbs which can be either tran-

sitive or intransitive without any derivational marking; these verbs are

uniformly agent-preserving and never patient-preserving. Finally, we will

present incorporation, the last morphological means of suppressing the

patient in Japhug.

2. Transitivity marking in Japhug

Transitivity is an essential feature of the Japhug verbal system. There is

never any ambiguity about whether a given verb is transitive or not. A

complete account of person agreement and TAM markers in Japhug is

beyond the scope of the present article, but in this section all a‰xes rele-

vant to the marking of transitivity will be discussed.

2.1. Morphological transitivity

In Japhug, transitive verbs regularly agree with two arguments, so that

many transitive verbal forms (such as 1>2, i.e. first person agent and

second person patient) have no equivalent in intransitive verbs.

However, third person patient forms (1>3, 2>3 and 3>3) are in some

cases almost indistinguishable from intransitive verbal forms. Two of the

markers which are found exclusively with transitive verbs are restricted

to a phonologically conditioned subset of these verbs: the regular stem 3

formation,3 which only occurs with verbs whose basic stem ends in -a, -u,

-o and -�, and the -t 1sg>3 / 2sg>3 past tense su‰x, which only applies

to open syllable roots.

3. In the Rgyalrong languages, verbs can have up to three or four distinct stems(Sun 2000). Stem 3 is used in Non-past, Imperfective, Irrealis and Imperativeforms with a singular agent and a third person patient (Jacques 2008: 246–7).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


200 Guillaume Jacques

Fortunately, two morphological tests can be applied to any verb to

determine whether it is transitive or intransitive, even if this verb only

allows third person patients and has a closed syllable stem. Verbs which

possess a stem 3 and make use of the -t past tense su‰x always confirm

their transitive nature in these two tests as well.

Firstly, transitive verbs take the prefix a- in direct aorist 3>3 forms.

Compare the following examples:

(1) p�-a-mto

aor-3>3-see

‘He saw him.’

(2) p�-fflqhl�t

aor-fall

‘He fell down (into a trap).’

In the first example, the verb is transitive, and the a- prefix appears

after the directional prefix p�- which marks the aorist, whereas in example

(2) the verb is intransitive and no such prefix appears.

Secondly, the nominalized forms of transitive and intransitive verbs

are distinct: intransitives build their S-nominalization by means of a prefix

k�-, while for transitive verbs the A-nominalized form makes use of the

same k�- prefix (see example 3) preceded by a possessive prefix coreferent

with the O. For instance:

(3) �-k�-mto

3sg.poss-nmlz:S/A-see

‘The one who sees him’

In example (3), the nominalized form is marked with a third person

singular possessive prefix �- indicating the O. Compare this form with (4):

(4) k�-si

nmlz:S/A-die

‘The dead one’

In this example, no such possessive prefix appears. Thus, on the basis

of the direct aorist 3>3 form and the nominalization in k�-, it is always

possible to determine whether a given verb is transitive or not.

The strict morphological transitivity marking found in Japhug and

other Rgyalrong languages is relatively rare among Sino-Tibetan lan-

guages, where transitivity is often di‰cult to define (as for example in

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Argument demotion in Japhug Rgyalrong 201

Chinese or Tibetan). However, Rgyalrong languages are not the only ones

in this family to show transitivity marking: Dulong/Rawang languages

(LaPolla 2001: 284) and Kiranti languages also have a fully-fledged set of

transitivity-marking a‰xes on the verb.

2.2. Case marking

The formal distinction between transitive and intransitive verbs is not

limited to their morphology. Japhug has a simple case marking system,

which presents ergative alignment: A arguments are marked with the

enclitic k�,4 while S and O are left unmarked, as can be seen in sentences

(5) and (6):

(5) ‚d�rffii k� ´amu p�-a-mto

Rdorje erg Lhamo aor-3>3-see

‘Rdorje saw Lhamo.’

(6) ´amu ci ��-n�re

Lhamo a_little evd-laugh

‘Lhamo laughed a little.’

When both participants are overt, it is therefore possible to determine

whether the verb is transitive or intransitive on the basis of case marking.

Ergative marking is obligatory with third person participants in Japhug

for all TAM categories, but only rarely appears with SAP pronouns such

as affio ‘I’ and n�ffio ‘you’.

With intransitive verbs, the ergative case can sometimes be used, but it

conveys a specific comparative meaning:

(7) n�ffio k� ��-t�-cha

you erg const-2-be_capable

‘You are more capable.’

Word order is verb-final, the agent usually preceding the patient. Sen-

tences with an overt agent and patient are fairly rare in actual texts. As

in many polysynthetic languages with indexation of two arguments, it is

quite common to omit both agent and patient NPs (Mithun 1999: 190–

193): the only necessary element of a sentence is the finite verbal form.

Covert arguments are normally definite. A minimal sentence such as:

4. This clitic is probably borrowed from the Tibetan ergative marker gyis.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(8) to-ndza

evd-eat

‘He ate it.’

can only appear if both the eater and the thing eaten have been mentioned

before or are implicit. Therefore, unlike in European languages such as

French or English, omission of the O argument is not available as a poten-

tial means of expressing an indefinite patient. Other morphological devices,

such as antipassive, generic or incorporation constructions, are required to

serve that purpose.

2.3. Ditransitive verbs

In ditransitive verbs involving a recipient, such as ‘give’ or ‘tell’, we observe

in Japhug both indirective and secundative alignment (Haspelmath 2005).

No more than two arguments may be indexed on the verb, and the nature

of the second argument indexed (recipient or theme) is lexically determined

for each verb. This can be illustrated with two verbs meaning ‘give’, mbi

and kho.

The first of these, mbi, encodes the recipient as its second argument

(Haspelmath’s secundative alignment):

(9) ki ��-ta-mbi

dem ipfv-1>2-give

‘I give this to you.’ (Gesar, 197)

Meanwhile, the verb kho allots the theme to the second argument posi-

tion, and the recipient can only be marked as an external argument taking

the dative su‰x -fflki or -phe (indirective alignment):5

(10) a-me ��-kham-a Ðu

1sg.poss-daughter ipfv-give[3]-1sg n.pst:be

‘I will give (you) my daughter.’ (The Frog, 78)

The alignment of ditransitive verbs in Japhug is quite strict, and no verb

can be both indirective and secundative.

2.4. Semi-transitive verbs

Some two-place verbs in Japhug are distinctive in that they neither present

transitivity markers nor require ergative marking on the agent. This class

5. The external argument is not overt in example (10); if present, it would takethe form n�-fflki or n�-phe.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



of ‘semi-transitive’ verbs includes verbs of motion and some verbs of

perception. We provide some examples to illustrate this intriguing phe-

nomenon:

(11) n� ra ffl-k�-ru j�'dem pl trans-imp:east-look n.pst:could

‘Just go to see them.’ (Smanmi, 110)

In this example, the verb ru ‘look at’ clearly has two distinct arguments,

but it fails to show any transitive marking. If it were a normal transitive

verb, the verb form seen in (11) – along with all other non-past, imper-

fective and imperative forms – would make use of the special ‘stem 3’

mentioned above, involving the vowel alternation u! e. Meanwhile, other

perception verbs such as mto ‘see’ are fully transitive: the person seeing

receives ergative marking and both arguments are indexed on the verb.

Like the verb ‘look at’, motion verbs are intransitive, although they

might be considered to have two arguments: the person/animal moving

and the place travelled towards (the latter is not usually marked with

an oblique case in Japhug). This class includes the verb ffle (Aorist stem

ari ) ‘go’.

(12) t�-pi ni t»u k�-wxti j�-arı-ndffii

neu-elder_sibling du way nmlz:stat-be_big aor-go[2]-du

‘The two elder brothers followed the big road.’ (The Fox, 19)

This phenomenon is illustrated here in example (12), where the persons

going (the elder brothers) do not receive ergative case. Other tests for

transitivity (nominalization, transitivity a‰xes on the aorist, etc.) would

also fail.

Interestingly, Dulong/Rawang, the other sub-branch of Sino-Tibetan to

share the strict transitivity marking seen in the Rgyalrong languages, also

possesses a class of ‘semi-transitive’ verbs, whose members overlap with

those in Rgyalrong: in particular, perception and motion verbs are again

found to display this behaviour (LaPolla 2008).

3. Generic marking

As mentioned in the introduction, Japhug and other Rgyalrong languages

have a direct/inverse marking system (DeLancey 1981, Sun and Shi 2002).

The inverse marker w'- appears on verb forms where the agent is third

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



person and the patient is SAP, or when the agent is non-human and the

patient human, for instance:

(13) tha a-m�-p�-t�� -w'-sat

otherwise irr-neg-pfv-2-inv-kill

‘Otherwise she would kill you.’ (Nyima vodzer, 36)

The use of the inverse in Japhug has been described in Jacques (2010a),

so this topic will not be discussed in detail here. It should be noted, how-

ever, that unlike passive and antipassive prefixes this a‰x does not cause a

change in valency: the verb remains transitive (the inverse marker cannot

appear on an intransitive verb) and both arguments can be marked in the

verb morphology.

A special use of the inverse prefix w'- is nevertheless highly relevant to

the topic covered in this paper: namely what I will call its ‘‘generic’’ use, a

function cross-linguistically associated with passive forms. No other lan-

guage with direction marking seems to share this feature, as can be seen

from Zuniga’s (2006) recent survey of direct/inverse systems in languages

of the New World. Japhug generic forms often appear in sentences ex-

pressing general truths, and generic marking is extremely common in pro-

cedural texts:

(14) s� jno sm�n tu-w'-¯t� tffl�n, n� lu-w'-l�t

grass medicine ipfv-inv-buy conj dem ipfv-inv-throw

‘After one buys fertilizer, one spreads it (in the field).’

(Rtsampa, 49)

The generic form can be strictly distinguished from the normal inverse.

Inverse verbal forms imply that the agent is lower or equal to the patient

on the Empathy Hierarchy. Inverse marking is obligatory if an inanimate

agent acts upon a human, and conversely, it can never appear when a

human acts upon an inanimate. This e¤ect of the Empathy Hierarchy

does not apply to generic forms. As example (14) shows, the generic inverse

can be used with a (generic) human agent and an inanimate patient. It never

takes dual or plural su‰xes. This implies that generic arguments, even

humans, are lower on the empathy hierarchy than inanimates. The follow-

ing hierarchy should thus be postulated:

(15) SAP > human > animal > inanimate > generic argument

Non-generic inverse forms di¤er from generic ones in two ways. Firstly,

they can take number su‰xes, and secondly, they cannot appear if the

patient is inanimate. Consider the following example:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(16) s�p'o �-ta‚ p�� -w'-ta-ndffii nd�re,

stake 3sg.poss-on aor-inv-put-du conj

ndffii-pa smi t�-a-”l�� -n� nd�re,

2du.poss-under fire aor-3>3-burn-pl conj

t�-pi ‚na‚na ffio p�� -w'-sat-ndffi i �� -�u.

neu-elder_brother both adv aor-inv-kill-du ipfv-be

‘People put the two of them on the stake, lit the fire under them,

and killed both of them, the two brothers.’ (The Fox 180–181)

All three verbs in (16) share the same agents (unspecified in the story,

but probably the king’s servants or the villagers). The first and the third

have inverse marking (and agree in number – dual – with their patient),

while the second is direct, with plural agent marking because its patient is

inanimate (the fire, smi). The absence of an inverse form on this verb con-

firms the fact that the inverses in the first and the second are not generic,

for otherwise we would expect the second verb to be in the inverse form

too. In (16), the inverse can be used because both the agents and the

patients are human, and thus equal in terms of the Empathy Hierarchy.6

The inverse prefix is only used when the generic argument is the agent

of a transitive verb. For intransitive verbs or patients, a di¤erent a‰x is

used: the k�- prefix. This prefix is homophonous with the nominalizer de-

scribed in section 2, but is unrelated to it at least synchronically. The uses

of this prefix are exemplified in sentences (17), from a story about the yeti,

and (18), from a procedural text:

(17) �-‚�ri n� �-‚�ri ffio

3sg.poss-front conj 3sg.poss-front adv

ju-k�-ph'o a-p�-Ðu tffle,

ipfv-genr:S/O-flee irr-ipfv-be conj

maka ffio m�-pj�-k�-mto khi

at_all adv neg-evd-genr:S/O-see hearsay

‘If one runs in the direction in front of (the yeti), one will not be

seen by (him).’ (The Yeti, 17)

In this example, both the first verb, ‘flee’ (intransitive), and the second,

‘see’ (transitive), have the generic prefix k�-. In the second verb, the

6. In example (16), direct forms would also be grammatical. The inverse appearshere for pragmatic reasons, because the patient (the two brothers) is moretopical than the agent (the unnamed people who perform their execution).See Jacques (2010a) for a more detailed discussion of this phenomenon.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



generic argument is not the A (the one who sees) but rather the O (the one

seen), as the A is definite (the yeti).

The generic forms are compatible with both imperfective and aorist

forms, as shown in the following example. Imperfective forms are the

most common in procedural texts, but the aorist forms are used to express

a ‘twist’ in the action, which can usually be translated into western lan-

guages with the conjunction ‘when’.

(18) n�n� mb'�jro‚ n� ci �� -k�-ffle

dem furrow dem once ipfv:west-genr:S/O-go

ki tu-k�-fse tffle

dem ipfv-genr:S/O-be this way conj

k�-k�-'e tffle li ki

aor:east-genr:S/O-come[2] conj again dem

tu-k�-fse tffle, n� k��-w'-ffl lu

ipfv-genr:S/O -be this way conj dem aor-inv-plough

n� k� li ch�-fka”dem erg again ipfv: downstr-cover

‘One goes along the furrow this way, but when one comes back

again this way, as one ploughs (the furrow), (the earth of the new

furrow) covers (the groove of the older furrow).’ (Rtsampa, 30–31)

Generic forms, however, are not compulsory in procedural texts. Verbs

in the imperfective, often (but not always) with plural marking, can express

the same meanings. See the following passage, which contains no generic

markers:

(19) ju-ph'o tffle k�spo‚-Ðg� lu-n�ffle

ipfv-flee conj hole-inside ipfv:upstr-go_back

tffle k�-sat m�-kh� tffle

conj aor-kill neg-n.pst:be_possible conj

�-k�m smi pj�-”l�-n� tffle

3sg.poss-door fire ipfv-burn-pl conj

m�-t�-tfflha‚ tffle ch�-n�´o‚ tffle

neg-aor-bear conj ipfv:downstr-go_out conj

pj�-sat Ðgr�l

ipfv-kill n.pst:be_usually_the_case

‘When (the badger) goes back into his hole, one cannot kill him, so

they light a fire in front of the hole, he cannot bear it and goes out,

and then one can kill him.’ (Dictionary entry)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



It is clear from these data that the generic markers in Japhug show

ergative alignment (S¼OAA): w'- is an A-generic, and k�- an S/O-

generic. This is one of the few contexts in Japhug where ergativity is

present, other than in case marking. In other syntactic structures, no erga-

tive alignment is found: nominalization patterns in Rgyalrong languages

follow accusative alignment (Sun 2003: 497), and equi-NP deletion shows

no syntactic pivot (Jacques 2010a).

Japhug generic forms di¤er markedly from those described in Tshobdun

(Sun 2005), where the inverse is not used in this function and a strict distinc-

tion is observed between human and non-human generic forms.

Unlike in Kryz, where the passive form conveys deontic/habitual mean-

ing (Authier, this volume §4.1), in Japhug such contexts instead make use

of the generic form, and the passive, as we will show, never bears this kind

of meaning.

4. Passive

Japhug has two agentless passive forms, which present complex morpho-

phonological alternations. These alternations have been described in Jacques

and Chen (2007), and we will concentrate our discussion here on the func-

tions of these forms. Unlike the generic described above, the passive belongs

to the domain of derivational morphology, and changes the valency of the

target verb, which becomes intransitive.

4.1. a- passive prefix

The first passive form, marked by the prefix a- (see Table 1), is fully pro-

ductive. It has three allomorphs, a-, �- or k�-, depending on the preceding

prefix (Jacques and Chen 2007). Most transitive verbs can be prefixed with

this element, and the resulting intransitive verb cannot have an overt

agent.

Table 1. Examples of passive forms

basic verb meaning derived verb meaning

mto see a-mto be seen

pr�t cut a-pr�t be cut

rku put in a-rku be put in

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In the imperfective, past imperfective and past evidential forms, passives

have a stative meaning, though a passive interpretation is also possible:

(20) s�tfflha �-Ðg� n� tfflu rÐ�l t�-tango‚ ��-�-rku

place 3sg.poss-in dem loc silver one-basket ipfv-pass-put_in

‘In that place, there is/someone has put a basket of silver.’

(The Divination, 68)

(21) tfflheme n� ci th�-sta ri,

girl det a_little aor-wake_up conj

�-mph�s th�cu n� p�-a-ta

3sg.poss-buttocks downstream dem pst.ipfv-pass-put

fflti

N.PST:be.a‰rmative

‘The girl woke up, and it (the horse embryo) was there / someone

had put it under her bottom.’ (The Three Sisters, 106)

(22) �-ph�Ðg� n� tfflu qap�t�m ci

3sg.poss-bosom dem loc pebble one

na-rku ��-Ðu ‘‘tffletha qhuj tffle

aor:3>3-put_in ipfv-be a_while this_evening conj

ki a-ph�Ðg� a-rku tffle

this 1sg.poss-bosom n.pst:pass-put_in conj

tffletha p�-mto-t-a ffio tffle, r˜ �lpu

a_while aor-see-pst-1sg adv conj king

�-phe tu-ti-a ��-ra’’

3sg.poss-dat ipfv-say-1sg ipfv-have_to

He put a pebble in his bosom, thinking: ‘‘This evening, it will be in

my bosom, and as I see it, I will certainly tell (the story) to the

king.’’ (Kunbzang, 279)

No overt agent can appear in the same clause, but this is not to say that

the passive is forbidden when the agent is known. For instance, in one text

we find the following sentence:

(23) t�-rdo‚ p�-a-qr� tffle, �-Ðg� n� tfflu

one-piece aor-3>3-tear conj 3sg.poss-in dem loc

rÐ� l qho‚qho‚ t�-rdo‚ pj�-k�-mph�� r-ch�,

silver ingot one-piece ipfv.evd-pass-wrap-evd

‘He opened one piece (of bread), and there was a silver ingot wrapped

inside it.’ (The Raven, 112)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Although not formally expressed in sentence (23), the agent who put

the silver ingots in the bread is known. A few sentences earlier in the

same story, we read:

(24) rÐ� l qho‚qho‚ t�-rdo‚ nts� ko-mph� r

silver ingot one-piece always evd-wrap

‘She (a character named Lhamo) put a silver ingot in each (piece of

bread).’ (The Raven, 109)

In the aorist, passive forms do not always have resultative meaning

(unlike in Kryz: Authier, this volume 3.2):

(25) tffle ‚ffi�n� ci rcan� k�-wxt�Pwxti

conj boy one unexpectedly nmlz:stat-intsvPbig

ffio n�-a-”zu ��-Ðu.

adv aor-pass-make ipfv-be

‘Then the fox changed into a big boy.’ (The Fox, 193)

The a- passive is rare in narratives and also in procedural texts, where

generic verb forms are generally used in clauses with a generic agent. It

is not commonly used to express situations where the agent is unknown:

more usual is inverse (or plural direct) marking on the verb, as was seen

in the previous section.

Additionally, on occasion a passive form can display an idiosyncratic

meaning which has evolved independently from that of the base verb.

This phenomenon is exemplified by a pair of very common verbs in

Japhug. The verb pa originally meant ‘do’ in proto-Rgyalrong, but in this

meaning it was supplanted by the Tibetan loanword ”zu and was pre-

served only in its secondary meaning, ‘close’. However, the corresponding

passive form a-pa, originally meaning ‘be made’, did not come to mean

‘be closed’; instead, it independently developed the meaning ‘become,

change’, and thus pa and a-pa are no longer related synchronically.

A prefix related to the a- passive is the s�- prefix, which represents the

fusion of the causative s�- and the passive a-. Verbs with this prefix are

few in number, and the situation is quite confused due to the fact that

two other homophonous s�- prefixes exist in Japhug, one of which is an

antipassive marker (cf. the following section). The combination of passive

and causative generates the special meaning ‘ask sb to’. For instance, from

the verb mbi ‘give’7 one derives s�-mbi ‘ask sb for sth’, in other words

‘cause sb to give sth to oneself ’.

7. This verb has primative/secundative alignment (Haspelmath 2005). The recipientis treated as the O, and the theme is not encoded on the verb. Other ditransitiveJaphug verbs, however, have indirective alignment, as mentioned above.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(26) ¯�jtsu t�-kh�tsa a-n�-t�-s�-mbi

chilli one-bowl irr-pfv-2-caus:pass-give

‘You will ask him for a bowl of chilli.’ (The Smart Rabbit, 40)

Up to this point, we have taken for granted that a- is to be considered a

real passive formant. We wish to dispel any suspicion that this might not

be the case. Keenan and Dryer (2007) warn against confusing the passive

with other constructions such as middle (p. 352), unspecified subject

(p. 354), inverse (p. 356) and antipassive (p. 359). However, we know

that Japhug has separate forms for generic reference (section 3), inverse

(Jacques 2010a) and antipassive (section 5), all of which are entirely dis-

tinct from the a- prefix. Therefore, the main issue to be discussed is

whether the a- form is a genuine passive or should be seen as expressing

middle voice.

Cross-linguistically, the term ‘middle’ is used for various verbal forms

which generally cover a wide functional range, such as passives, reciprocals,

reflexives, autobenefactives, and forms to denote actions a¤ecting an object

possessed by the agent. The presence of middle marking has been reported

for several Sino-Tibetan languages, such as Rawang (LaPolla 2001). There-

fore, the presence of a typologically similar grammatical category in Japhug

would not be surprising.

However, it seems clear that the a- prefix has a much more restricted

range of functions than we would expect if it were a middle marker. In

particular, the a- passive never has reflexive meaning: reflexive forms are

regularly made from transitive verbs by adding the prefix ffi'�- (e.g. sat

‘kill’, ffi'�-sat ‘commit suicide’).8 However, it is possible that in an earlier

stage of the language, the a- prefix had a broader range of uses than

merely the agentless passive.

One productive verbal form which is historically related to the passive

is the reciprocal (see Table 2). It is produced by adding the a- prefix and

reduplicating the last syllable of the verb stem. A minority of reciprocal

verbs show no stem reduplication, but instead add the double prefix

a-m�-. All reciprocal verbs are morphologically intransitive.

8. On the origin of this prefix, see Jacques (2010b).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 2. Examples of reciprocal forms


ndza eat a-ndz�-ndza eat each other

n�r�tsa envy a-n�r�ts�-tsa envy each other

l�t throw a-l�-l�t fight each other

mto see a-m�-mto see each other

ti say a-m�-ti tell each other

The reciprocal form in a- is obviously related to the a- passive, and this

could challenge our claim that the prefix a- is a real passive formant.

However, the relationship between the passive and the reciprocal is purely

historical, and is not synchronically valid in modern Japhug.

The common ancestor of the passive formant a- and the a- element seen

in the reciprocal must have taken the shape *Ða- in proto-Japhug (Jacques

and Chen 2007: 889). This *Ða- prefix must have functioned as a much

broader intransitive marker and not strictly a marker of the passive.

From a typological perspective, it may be instructive here to compare

intransitive a‰xes in other Sino-Tibetan languages. In Rawang, the intran-

sitive v- [�] prefix can derive passive-like verb forms, whose S corresponds

to the O of the transitive verb (LaPolla 2001: 288):

(27) tvl-o-e # ! v-tvl-e #roll-3.tr-n.pst

‘roll (tr)’

intr-roll-n.pst

‘roll (intr)’

Moreover, according to LaPolla, ‘if the single direct argument of the

derived intransitive is a plural animate argument, then the meaning is

reciprocal’. Here is one of his examples:

(28) angmaq v-shvt-e #they intr-hit/kill-n.pst

‘They are fighting.’

The functions of the v- prefix in Rawang are reminiscent of those of the

proto-Japhug *Ða- prefix.9 The main di¤erence is that in Japhug (as in all

four of the Rgyalrong languages) the reciprocal meaning only appears if

the verb stem is reduplicated.

9. This was suggested to me by an anonymous reviewer of the journal ‘Languageand Linguistics’, whom I wish to thank again for insightful comments.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Interestingly, in Rtau, a language related very closely to the Rgyalrong

languages10, verb stem reduplication expresses plurality in the subject (S/A)

(Huang 1991: 29–30), as shown in Table 3.

Table 3. Paradigm of the verb ‘go’ in Rtau.

person singular plural

1 fflo-Ð ffl�-fflo-Ð2 ffli-n ffl�-ffl i-n

3 ffl� ffl�-ffl�

If we assume that proto-Japhug *Ða- had a function similar to the

Rawang v-, and that the reduplication present in Rtau as a marker of plu-

rality is a survival from proto-Rgyalrongic, the formation of the reciprocal

in the modern Rgyalrong languages can be explained: when a *Ða- pre-

fixed verb had a plural animate argument, the verb stem was reduplicated,

and it developed a reciprocal meaning. After verb stem reduplication ceased

to function as a marker of plurality, its presence on *Ða- prefixed verbs

was reanalysed as a reciprocal marker.

Therefore, although the a- passive prefix and the a- element in recipro-

cal forms are historically related, they no longer represent the same mor-

pheme in modern Japhug. This is why we argue that the form in a- is a

genuine passive and not a vague intransitive or middle.

Finally, it should be noted that many intransitive verbs, such as armbat

‘be near’ or afflqhe ‘cough’, have a prefixal a- element although no corre-

sponding transitive verb exists. We have no reason to identify such verbs

as passives, and can consider the a- element as part of the verb root. How-

ever, it should be stressed that no transitive verb has this a- element.

4.2. Prenasalized anticausative

The prenasalization alternation in Japhug derives an anticausative verb

from a transitive one. Only sixteen pairs of verbs present this alternation

(Jacques 2008: 84–5), and at least one of them (¯t�r, Table 4) is a loan-

10. Sun (2000) argues convincingly that Rtau and Lavrung are the closest relativesof Rgyalrong within Qiangic. He calls the sub-branch including the Rgyalronglanguages, the Lavrung languages and Rtau ‘Rgyalrongic’, a term which wealso adopt here.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



word from Tibetan (gtor), which shows that this morphological process

retained its productivity until recently.11

Table 4. Examples of the prenasalized anticausative in Japhug

transitive meaning intransitive meaning

ftsi melt (tr) ndzi melt (itr)

pr�t break (tr), cut mbr�t break (itr), be cut

q�t separate ng�t be separated

¯t�r scatter ‚nd�r be scattered

The major di¤erence between the a- passive and the prenasalized anti-

causative is that the agent is omitted when the a- passive form is used,

but semantically, the existence of an indefinite external agent is not ex-

cluded, whereas in the case of the prenasalized anticausative, no agent is

present, and the action is viewed as taking place spontaneously.

(29) ‘‘wo a-ffi i ra n�-mk�'� r

Oh 1sg.poss-lady pl 3pl.poss-necklace

p�-mbr�t ’’ tı-n�aor-anticaus:break n.pst:say-pl

‘Oh, my lady, your necklace has broken!’ they said. (Kunbzang 214)

In this example, the characters uttering this sentence believe that the

necklace broke by itself, without an external agent: he therefore uses an

anticausative form. If the a- passive a-pr�t had been used instead, it could

have implied that someone had broken the necklace on purpose.

However, since only a few verbs have a special anticausative form, for

the majority of verbs only the a- passive is available to express anticausa-

tive meaning. There are phonological constraints on prenasalization: all

16 known examples of transitive verbs to which this process applies have

an unvoiced unaspirated stop or a¤ricate in initial position.12 Verbs with

nasal or fricative initial consonants have no distinct anticausative form.

11. However, the phonological correspondence -or : -�r shows that it belongs tothe earliest layer of Tibetan loanwords (Jacques 2008: 136–146).

12. We use ‘initial’ here in the sense defined in Jacques (2004: 12–73).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



5. Antipassive

Sun (2006: 8) was the first scholar to notice the existence of antipassive

forms in a Rgyalrong language, namely Tshobdun (Caodeng). He pointed

out that two antipassive prefixes exist in this language: r½-, which is used

when the patient is non-human, and s½-, used with human patients (see

Table 5).

Japhug is very closely related to Tshobdun, and the same distinction

applies for the cognate prefixes r�- and s�-, as can be seen from the follow-

ing examples: r�- is added to verbs with a prototypical non-human patient,

and s�- when the patient is necessarily human.

Table 5. Examples of antipassive derivation in Japhug


nd�n read r�-nd�n read sth.

nts'e sell r�-nts'e do business

fflar search r�-fflar look for sth.

fst�n serve s�-fst�n serve sb.

‚nd� hit sa-‚nd� hit sb.

fflar search s�-fflar look for sb.

The resulting verbs are intransitive: they never display any of the signs

of verbal transitivity described in part 2 of this paper. The antipassive is

not simply a means of changing the case marking of the Agent from Erga-

tive to Absolutive: it involves a change in valency, with the result that the

original patient can no longer be expressed. Usually, as in examples (30)

and (31), the semantic patient becomes indefinite:

(30) t�-rffia” n� pj�-r�-fflph�t

neu-wife det ipfv.evd-apass:nonhum-mend

‘The wife was mending (clothes).’ (The Raven, 19)

(31) tfflhi tu-t�-ste Ðu k�-s�-fst�n

what ipfv-2-do_this_way[3] n.pst-be inf-apass:hum-serve

‘How do you serve (your husband and the people from his family)?’

(The Frog, 128)

In a few cases in our texts, the (human patient) antipassive seems to be

used as a way to avoid using the first person when asking for something,

thereby making the request more indirect:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(32) t�cirq� k�-s�-jtshi �� -tucold_water nmlz:O-apass:hum-make_drink Q-n.pst:have

‘Is there any cold water to give (us) to drink?’ (Nyima Vodzer, 35)

The human/non-human distinction is not always strictly observed. In

some examples, a s�- prefixed verb may have a non-human animate (animal)

as its intended patient:

(33) th�-wxti �-j�ja tffle,

aor-big 3sg.poss-following conj

k�-s�-ndza k�-Ðu n� pj�-s�¯s�l

nmzl:S/A-apass:hum-eat nmzl:S/A-be det ipfv-realize

‘As (the buzzardi) grows bigger, (the bunting) realizes that iti eats

(other birds).’ (The Buzzard, 26)

In this example, since semantically both the patient and the agent of the

verb ‘eat’ are birds and therefore equals on the Empathy Hierarchy, the

antipassive with suppressed human patient prefix s�- is found instead of

the expected r�-.

The Japhug antipassive does not seem to be very productive, and few

verb roots are able to take both r�- and s�- prefixes (fflar ‘‘to search’’ in

Table 5 is one example). In both narratives and procedural texts, these

forms are quite rare (cf. the text counts in the conclusion).

6. De-experiencer

The de-experiencer prefix s�-, homophonous with the antipassive prefix

for human patients discussed in the previous section, di¤ers from the other

morphological devices presented in this paper. Its basic function is not to

decrease valency per se: rather, it derives an intransitive stative verb from

an intransitive verb or a transitive verb of perception.

The S of the derived verb denotes the stimulus of the state or action,

and it has the meaning ‘‘be liable to cause sb/sth to X’’, where X is the

meaning of the basic verb.

In the case of intransitive verbs, the original S argument is suppressed

and replaced by the stimulus. For transitive perception verbs, the A (corre-

sponding to the experiencer) is suppressed and the original O (the stimulus)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



becomes the S of the derived verb. The examples in Table 6 illustrate this

derivation.

Table 6. The de-experiencer prefix with intransitive and transitive verbs

Basic verb meaning Derived verb meaning

Ðgio slip (itr.) s�-Ðgio be slippery (of the ground)

scit be happy s�-scitbe nice (of a situation), be funny(of a person)

fflke be burned s�-fflkebe burning (of a boiling kettle, whichburns the hand of anyone touching it)

rga like (itr.) s�-rga be nice

mto see s�-mto be easy to see

mtsh�m hear s�-mtsh�m be easy to hear

Since the deleted argument is always the experiencer (whether the verb is

transitive or intransitive), we label this prefix ‘‘de-experiencer’’.13 Although

in the case of intransitive verbs there is no decrease in valency, the addition

of this prefix is nonetheless a demotion in the sense that a stimulus is lower

than an experiencer in terms of agentivity (for instance, humans are less

likely to be stimuli).

7. Lability

While transitive and intransitive verbs can be easily distinguished on formal

grounds (see section 2), a small class of verbs can be either transitive or

intransitive. Table 7 presents a list of the labile verbs identified in Japhug

up to this point.

13. This terminology was suggested by Peter Austin. The counterpart of the de-experiencer prefix among transitivizing a‰xes is the estimative prefix n�-,which derives a transitive verb from an intransitive one whose A correspondsto the experiencer and whose O corresponds to the stimulus (the S of the basicverb), for instance mpffl�r ‘‘be beautiful’’ > n�mpffl�r ‘‘consider to be beautiful’’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 7. Labile verbs in Japhug

Japhug meaning Japhug meaning

s�Ðo listen r��˜o‚�˜ i crush

n�m�o look 'ndffi� r grind

rÐu fry n�khaja resist

m� rk� steal s� la‚rd�” kick (of an animal)

rpu bump s�qarts� kick

fffl i forge n�ffi�mÐ�n envy

ffl lu plough n�mbr�p� ride (on horseback)

ta‚ weave

For instance, the verb rpu ‘bump into’ can appear both with and without

the 1sg/2sg Aorist -t su‰x, cf. the following examples:

(34) �-ta‚ k�-rpu-a

3sg.poss-up aor:east-bump_into-1sg

‘I bumped into him.’

(35) a-rpa‚ �-ta‚ k�-rpu-t-a

1sg.poss-elbow 3sg.poss-up aor:east-bump_into-pst-1sg

‘I bumped into him with my elbow.’

In sentence (35), the verb bears a transitive -t su‰x, and the patient of

the clause, ‘my elbow’, is overt (the person bumped into is indicated with

an oblique case). In sentence (35), in contrast, the body part involved is

not expressed, and the verb lacks the transitive marker; here �-ta‚ ‘on

him’ is syntactically not the object of the verb but an adjunct. The lability

observed in these two examples is agent-preserving: the agent of the tran-

sitive verb, and not the patient, remains when the verb is used intransi-

tively.14 We will label the S/A argument of labile verbs as the ‘subject’.

Since nouns in Japhug take ergative marking, this implies that the sub-

ject is marked with the ergative case when the verb is transitive, and has

absolutive marking when the verb is intransitive. Consider the following

examples, involving the labile verb n�mbr�p� ‘ride’.

(36) t�-rdo‚ n� k�� t�-a-n�mbr�p�one-piece det erg aor-3sg>3-ride

‘One of them rode it.’ (The Tiger, 11)

14. This phenomenon is described by Næss (2007: 125) as ‘Indefinite Object deletion’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(37) �-ta‚ n� tfflu t�tffl� n� to-n�mbr�p�3sg.poss-up det loc boy det evd-ride

‘The boy rode on her.’ (The Flood, 83)

In example (36), the agent is marked with the ergative. The patient is

not overt, but the transitive a- prefix on the verb and the presence of erga-

tive case indicate that the verb is to be interpreted as transitive: the patient

is definite (it refers to a tiger mistakenly stolen by three thieves).

In example (37), the subject ‘the boy’ does not bear ergative case, and

the verb must be interpreted as intransitive (though in this example the

evidential verbal form itself is ambiguous as regards transitivity). As in

examples (34) and (35), �-ta‚ ‘on her’ is an adjunct, and does not partic-

ipate in the verb’s argument structure. Similar examples could be found

for all verbs in Table 7.

These verbs belong to two distinct semantic categories: perception/

feeling (‘listen to’, ‘look at’, ‘envy’) and concrete action (‘grind’, ‘bump

into’, ‘forge’ etc). Interestingly, no ‘verbs of ingestion’ such as ‘eat’ are

found in this list. Næss (2007: 126) claims that ‘‘For most languages for

which I have been able to find data, it is the case that if they allow indefi-

nite object deletion with any verb, they will allow it with ‘eat’’’, but the

verbs ndza ‘eat’ and tshi ‘drink’ are strictly transitive in Japhug; intransi-

tive verbs of ingestion do exist, such as r�ndz�tshi ‘have lunch’, but they

are not related to their transitive equivalents by means of lability.

Interestingly, none of the labile verb pairs discovered so far seems to

show an unpredictable change in meaning, although these are very com-

mon cross-linguistically, as for instance with English ‘drink’, which as an

intransitive principally bears the meaning ‘drink alcohol’.

No examples of patient-preserving lability have so far been found in

Japhug.

8. Incorporation

The last argument-demotion device in Japhug is verbal incorporation. It

is not widespread but restricted to a few verbs, though the fact that some

examples involve recent Chinese and Tibetan loanwords shows that the

process is still potentially productive.

Incorporated nouns involve objects (see ‘timber’ and ‘money’ in the

following table), subjects of intransitive verbs (‘horse’) and a few adjuncts

(‘back’), but never the agent of a transitive verb. The incorporated noun

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



root appears before the verb root, following the normal strict verb-final

order of the language. Some examples are given in Table 8.

Table 8. Examples of incorporated nouns in Japhug

noun verb incorporation meaning

mbro ‘horse’ r˜�' ‘run’ it. n�-mbr�-r˜�' gallop it.

qhu ‘back’ ru ‘look’ it. n�-qha-ru turn around, look back it.

si ‘timber’ ph�t ‘chop’ tr. '�-s�-ph�t fell trees it.

pfflawts� ‘money’15 fso‚ ‘earn’ tr. '�-pfflawts�-fso‚ earn money it.

qhu ‘back’ Ðga ‘wear’ tr. n�-qh�-Ðga wear on the back tr.

Morphological marking on the incorporated noun is extensive. Firstly, the

incorporated noun appears between a derivational prefix (n�-, n�- or '�-)

and the verb root. Secondly, if the noun ends in an open syllable, the

vowel usually undergoes one of a set of regular changes also found in

compound nouns, giving a form we call status constructus:

-o ! -�/-a

-u ! -�/-a

-i ! -�

When the verb incorporates the patient of the original verb (see ‘fell

trees’ and ‘earn money’ above), the incorporated verb becomes intransi-

tive: this is a case of saturating incorporation. When, however, the incor-

porated noun is an adjunct (‘wear on the back’), it does not a¤ect the

valency of the verb.

Most incorporated objects can appear with the same verb as free objects,

but there are a few frozen incorporated objects which are no longer felt as

such. Sometimes both forms appear next to one another in the same story,

as can be seen in examples (38) and (39).

(38) ´asa ju-k�-ffle tffle, n� tfflu

Lhasa ipfv-genr:S/O-go conj dem conj

pfflawts� k�-fso‚ ��-mbat

money inf-earn const-easy

‘If one goes to Lhasa, money is easy to earn there.’ (Lobzang, 22)

15. The word pfflawts� ‘paper money’ comes from the colloquial Chinese piaozi.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(39) n�-mbro n�-rÐ� l tu-rke-a tffle

2sg.poss-horse 2sg.poss-silver ipfv-put_in[3]-1sg conj

k�-'�-pfflawts�-fso‚ j�-ffle tffle

nmlz:S/A-derivation-money-earn imp-go conj

‘I will prepare a horse and some silver for you: go to earn money.’

(Lobzang, 17)

In (39), pfflawts� ‘money’ is incorporated, whereas in (38) it appears

as a free object. As these examples show, the semantic di¤erence between

free and incorporated object is minimal here; in both cases the object is

indefinite.

The decrease in valency caused by incorporation shows some similarity

with the e¤ects of antipassivization. The incorporated object can never be

definite, and there are strong restrictions on which objects can be incorpo-

rated. The activity expressed by the verb must be, as Mithun (1984: 848)

puts it, ‘‘recognized su‰ciently often to be considered name-worthy in its

own right’’. For instance, the verb mtshi ‘lead’ can only appear with two

incorporated objects, jla ‘yak/cow hybrid’ and mbro ‘horse’, never with

any other animal.

9. Conclusion

In Japhug, covert arguments of transitive verbs are generally interpreted as

definite, though a few exceptions are treated in this article. Therefore, indef-

inite arguments, in particular unknown or generic ones, must be demoted.

Patients can be suppressed by means of four distinct constructions –

generic, antipassive, lability and incorporation; while only two possibilities

exist for the suppression of agents, namely the generic and passive con-

structions. Japhug has a wide range of very specialized argument demo-

tion devices, including passive, reciprocal, de-experiencer and reflexive

(not mentioned in the present article; see Jacques 2010b).

We have already outlined some important di¤erences in the use of these

forms, but additional insight may be gained from text counts. The distri-

bution we found in a corpus of eight narrative texts, and another compris-

ing two procedural texts, is presented in Table 9.16

16. No examples of the de-experiencer s�- with a transitive verb were found in thispart of the corpus.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 9. Text counts of argument demotion

narrative procedural text

inverse (excluding generic) 35 3.1% 0 0%

A-generic 2 0.2% 215 49.2%

O-generic 0 0% 0 0%

S-generic 4 0.4% 2 0.5%

labile verb as transitive 2 0.2% 0 0%

other transitive 564 51.1% 46 10.5%

antipassive 3 0.3% 0 0%

passive 8 0.7% 6 1.4%

labile verb as intransitive 27 2.4% 0 0%

incorporation 1 0.1% 0 0%

other intransitive 459 41.4% 168 38.4%

In both narrative and procedural texts, S/O generic forms, antipassive

and incorporation are all barely attested. Passive verb forms are less rare,

but the majority of examples involve one of just three common verbs:

a-rku ‘be placed in’, a-ta ‘be placed’ and a-”zu ‘become’. In comparison

with such languages as Kutenai (Dryer 1994: 75), where passive forms

constitute up to 27% of all semantically transitive clauses, it is clear that

the Japhug passive is very restricted in usage.

The major di¤erence between the two categories of texts lies in the

generic use of the inverse. Almost unattested in narratives, it is by far the

most common form in procedural texts. Of all five argument-demoting

mechanisms, the generic is the only one which is fully productive.

It is instructive to evaluate to what extent the syntactic properties of the

four patient-demoting constructions in Japhug are common for such con-

structions cross-linguistically. Bickel et al. (2007: 18) propose a set of ten

properties commonly associated with antipassive (A), incorporation (I)

and optional agreement (OA). The properties of the Japhug constructions

are given in Table 10.17

17. The abbreviations G (J), A (J), L (J) and I (J) correspond to the Generic, Anti-passive, Lability and Incorporation constructions of Japhug. The symbols y,n, n/a and ? stand for ‘yes’, ‘no’, ‘not applicable’ and ‘true in some languages’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 10. Typological comparison of syntactic properties of Patient-Demotingconstructions

A I OA G (J) A (J) L (J) I (J)

O is adjacent to verb n y n n n n y

O and verb in one word n y n n n n y

O is obligatory n y n n n n y

O is deleted ? n n y y y n

O is a fully-fledged NP y ? y n/a n/a n/a n/a

Relativization on O y ? y n/a n n n

Regular O-case on O n n y n/a n/a n/a n/a

Transitive A-case on A n n y y n n n

Suspended O-agreement y y y y y y y

Generic O or O with unknowncardinality

n y ? y either y y

Incorporation aside, the three other constructions di¤er very little as far

as these ten properties are concerned.

As is the case for many antipassive constructions cross-linguistically,

in the Japhug Antipassive the demoted patient cannot be retained in the

clause concerned, even in an oblique case. The Japhug Passive behaves

analogously, as its use rules out the specification of an Agent in the clause.

As we have seen in the present article, the Japhug Passive and Anti-

passive have a variety of functions, but they do not serve to topicalize the

Patient/Agent (word order, elision of the argument and inverse marking

are used instead) or make the Patient/Agent argument accessible to rela-

tivization (as S, A and O can all be relativized).

They do not share the whole set of parameters which are commonly

associated with the terms ‘passive’ and ‘antipassive’, but nevertheless these

terms seem to be appropriate labels to describe the behaviour of the a‰xes

discussed here. Whether these properties are cross-linguistically common

or rare, and whether some of them are somehow correlated with ergative

alignment, are questions which lie beyond the scope of this article.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



References

Bickel, Balthasar, Martin Gaenzle, Arjun Rai, Prem Dhoj Rai, Shree Kumar Rai,Vishnu S. Rai and Narayan P. Sharma

2007 ‘‘Two ways of suspending object agreement in Puma: betweenincorporation, antipassivization and optional agreement.’’ Hima-layan Linguistics 7: 1–19.

DeLancey, Scott1981 ‘‘The category of direction in Tibeto-Burman.’’ Linguistics of the

Tibeto-Burman Area 6(1): 83–101.Dryer, Matthew S.

1994 ‘‘The discourse function of the Kutenai inverse.’’ In Voice andInversion, T. Givon (ed.), 65–99. Amsterdam/Philadelphia: Ben-jamins.

Haspelmath, Martin2005 ‘‘Argument marking in ditransitive alignment types.’’ Linguistic

Discovery 3(1): 1–21.Hopper, Paul J. and Sandra S. Thompson

1980 ‘‘Transitivity in grammar and discourse.’’ Language, 56(2): 251–299.

Jacques, Guillaume2004 Phonologie et morphologie du Japhug (rgyalrong), these de

doctorat, universite Paris VII-Denis Diderot.Jacques, Guillaume

2008 Jiarongyu yanjiu [Study on the Rgyalrong lan-guage]. Beijing: Minzu chubanshe.

Jacques, Guillaume2010a ‘‘The inverse in Japhug Rgyalrong.’’ Language and Linguistics,

11(1): 127–157.Jacques, Guillaume

2010b ‘‘The origin of the reflexive prefix in Rgyalrong languages.’’Bulletin of the School of Oriental and African Studies, 73(2):261–268.

Jacques, Guillaume and Chenzhen‘‘ Chabaohua de bujiwuqianzhui ji xiangguan wenti [The intransitive prefix in Japhugand related problems].’’ Language and Linguistics, 8(4): 883–912.

Jacques, Guillaume and Alexis Michaud2011 The reconstruction of Proto-Naic: a preliminary study of the

historical phonology of highly eroded Sino-Tibetan languages,Diachronica, 28.4.

Keenan, Edward L. and Dryer, Matthew S.2007 ‘‘Passive in the World’s Languages.’’ In Clause Structure, Language

Typology and Syntactic Description, Vol. 1, Timothy Shopen (ed.),325–361. Second Edition. Cambridge University.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



LaPolla, Randy2001 ‘‘Valency-changing derivations in Dulong/Rawang.’’ In Changing

valency: Case studies in transitivity, R.M.W. Dixon and A.Y.Aikhenvald (eds.), 282–311. Cambridge: Cambridge UniversityPress.

LaPolla, Randy2008 ‘‘Transitivity and transitivity alternations in Rawang and

Qiang.’’ In Proceedings of the Workshop on Tibeto-Burman Lan-guages of Sichuan, 239–248. Academia Sinica, Institute of Lin-guistics, November 21–22.

Mithun, Marianne1984 ‘‘The evolution of noun incorporation.’’ Language 60: 847–94.

Mithun, Marianne1999 The Languages of Native North America. Cambridge Language

Surveys. Cambridge: Cambridge University Press.Næss, Ashild

2007 Prototypical Transitivity. Amsterdam/Philadelphia: John Benjamins.Sun, Jackson T.-S.

2000 ‘‘Parallelisms in the verb morphology of Sidaba rGyalrong andLavrung in rGyalrongic.’’ Language and Linguistics 1(1). 161–190.

Sun, Jackson T.-S.2003 ‘‘Caodeng rGyalrong.’’ In Sino-Tibetan Languages, Graham

Thurgood and Randy J. LaPolla (eds.), 490–502. London:Routledge.

Sun, Jackson T.-S.2005 ‘‘Linguistic Coding of Generic Human Arguments in rGyalrongic

Languages,’’ 11th Himalayan Languages Symposium, 6–9December, 2005, Chulalongkorn University, Bangkok, Thailand.

Sun, Jackson T.-S.2006 ‘‘ Jiarongyu dongci de paisheng xing-

tai [Derivational morphology in the Rgyalrong verb].’’ Minzuyuwen 4: 3–14.

Sun, Jackson T.-S. and Shidanluo2002 Caodenghua

yu « rentong dengdi » xiangguan de yufa xianxiang [CaodengRgyalrong and grammatical phenomena related to the EmpathyHierarchy], Language and Linguistics, 3.1: 79–99.

Zuniga, Fernando2006 Deixis and Alignment: Inverse Systems in Indigenous Languages

of the Americas. Amsterdam/Philadelphia: Benjamins.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The Katukina-Kanamari antipassive

Francesc Queixalos

1. Introduction1

If we take linguistic form seriously – a stand which does not preclude the

necessity of taking seriously other levels of linguistic patterning such as

semantics and pragmatics – then some level ought to be defined on purely

formal grounds. I take the grammatical relations ‘subject’ and ‘object’ as

being among such formally defined linguistic notions. In dealing with a

linguistic entity, and with grammatical relations in particular, we are

likely to have much to gain by clearly distinguishing between the defini-

tion of that entity, i.e. the set of properties which identify it, and its moti-

vation, i.e. the factors which lead a speaker to put it to use (see section

2.5). The main aim of this paper is to present data from a very sparsely

documented, strongly ergative language of Amazonia as an illustration of

these considerations. The particular grammatical point chosen for this

purpose concerns the basic bivalent transitive clause and its associated

valency-reducing voice, the antipassive. I will attempt to show that in a

clearly hierarchized system of grammatical relations, voice alternation

has a primary target, namely the accessibility restrictions bearing on the

lower-ranked argument of a two-place clause. Of course, the idea that

such a generalization could encompass antipassives in all the di¤erent

kinds of so-called ergative languages, and even passives in languages of

all kinds, is on the face of it a blatant impossibility: functional (semantic

and pragmatic) motivations for the existence of voice are too pervasive

everywhere for a purely syntactic explanation to be convincing. Never-

theless, the interplay between these two types of motivation for voice

alternations is somewhat obscured by the quest for all-or-nothing (formal

vs. functional) typological definitions (for the passive, see Comrie 2008;

Givon 2008). In the spirit of Cooreman’s (1994) work on antipassives

or Givon’s (2009) on passives, diachrony – a repository for the e¤ects of

1. Many thanks to Katharina Haude, Tomas Givon and Gilbert Lazard forcomments on a previous version of this paper.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:43) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 227–258 1351 Authier_07_queixalos (p. 227)

formal and functional pressures – still has much to say on voice mecha-

nisms as partial subproducts of the interaction between 1) the basic align-

ment type of a language, and 2) the communicative needs of speakers.

Katukina-Kanamari, apparently the only surviving language of the

small Katukina family, is spoken by at most two thousand people in

a large region of Amazonia located between the Purus and the Javari

rivers and between the Japura river and the southern edge of the state of

Amazonas, Brazil. Adelaar (2000) suggests a genetic link with the Peruvian

isolate Harakmbut/Amarakaeri.

The language can be characterized typologically by the following fea-

tures: simple phonology, sparse morphology, clear-cut lexical classes with

no primitive adjectives, no trivalent verbs other than ‘say’, head marking,

predicate-initial2 constituent order, strong configurationality in terms of

constituency, a neat grammatical relations hierarchy, predominant erga-

tive patterning in almost all aspects of its grammar, and split transitivity.

The paper is organized in the following manner. I first present the basic

patterns in terms of formal features such as coding, constituency, behaviour,

and coreference control, followed by an explicitation of the consequences

of these phenomena for the question of grammatical relations. Then I turn

to describe a voice alternation which operates upon the basic bivalent

transitive clause: the participant mapped to the highest grammatical rela-

tion is deprived of its core argument status, thereby allowing the extant

participant to fill the unique argument slot of a one-place clause.

2. Basic patterns

The most unrestricted way of rendering an event involving ‘‘Mayon’’

(personal name), ‘‘cut’’ and ‘‘wild meat’’ is shown in

(1)ITQ Mayon-na¼ tukman barahai

Mayon-MkCase¼ cut wild_meat

‘Mayon cut the wild meat’3

2. But with some amount of flexibility, as we will see. ‘‘Predicate’’ is taken in thesense ‘‘predicate phrase’’.

3. ITQ refers to the Itaquai river, where the data from the Kanamari dialectwere collected. BIA will stand for the river Bia, a Jutai tributary and theorigin of the data from the Katukina dialect.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


228 Francesc Queixalos

where the verb is in medial position, the patient phrase (unmarked for

case) follows the verb, and the agent phrase (case-marked by na) precedes

the verb.4

For an event involving ‘‘man’’ and ‘‘go away’’, we have

(2)ITQ daan piya

go;walk man

‘the man went away’

Lexically monovalent verbs such as daan do not appear in other types

of clause structure.

2.1. Coding

Comparison of (1) and (2) demonstrates that Katukina-Kanamari has

clear ergative alignment in terms both of case marking – the agent in (1)

is explicitly marked for case, while the patient in (1) is unmarked, like the

unique argument in (2) – and word order, agents being located preverbally

but patients and unique arguments postverbally. Moreover, pronominal

forms are bound for agents but free for patients and unique arguments,

as shown in (3) and (4) respectively:

(3)BIA singular plural

1 yo-5 tyo-

2 no- na-

3 a- ma-

4. Specific semantic roles are not particularly at issue in this paper, so for two-place verbs I will simply rely on the prototype semantics-based terms agentand patient. MkCase stands for ‘marked case’, for which a justification willbe proposed below. Phonologically, the case su‰x cliticizes to the verb, yield-ing the sequence of phonological words #mayon# #natukman# #barahai#.This is assumed to be the result of a diachronic process of procliticization ofthe case su‰x to the phrase head (other examples of head attraction includeauxiliarization, see 2.4). In spite of the well-established tradition of organizinggrammatical examples on the basis of the phonological properties of cliticsrather than on their grammatical properties, I adopt a di¤erent convention insuch a case of strict contiguity between the grammatical host and the phono-logical host of an intermediate form: the clitic na is restored to the positionbeside its grammatical host, the notation {A� x ¼ B} reflecting the restitu-tion of the element x – phonologically bound to an adjacent following elementB – to its adjacent preceding grammatical host A.

5. The tables follow the dialect showing the simplest allomorph inventories.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The Katukina-Kanamari antipassive 229

(4)ITQ singular plural

1 adu adik

2 idi:k idi:ki

3 anyan6 anyan hinuk

No coding of gender surfaces in pronominal forms.

2.2. Constituency

The predicate and its external (sister) argument7 appear in that order. The

reverse order is also allowed, as in (5), though this probably brings with it

a slight pragmatic e¤ect.

(5)ITQ piya daan

man go;walk


A subclass of lexical predicate heads generates syntactically complex

predicate phrases. These are: bivalent verbs (6), bivalent nouns (7), and

postpositions (8). All take an obligatory internal, pre-head argument. All

other lexical predicate heads, i.e. monovalent verbs (9), monovalent nouns

(10), and adverbs (11)–(12), only take an external, post-predicate argu-

ment. Examples follow.

Bivalent heads:

verb

(6)BIA [Ayobi-na¼ bo:dak] tawami

Ayobi-MkCase¼ roast manioc_dough

‘Ayobi is roasting the manioc dough’

noun

(7)ITQ [opatyin-na¼ wadik] Warohan

child-MkCase¼ name Warohan

‘Warohan is the child’s name’

6. Third person pronouns seem to be demonstrative in origin.7. I use ‘‘argument’’ in its formal sense of ‘‘linguistic expression of a participant

required by the semantic structure of the verb’’, as I do for ‘‘internal’’ and‘‘external’’, that is, instantiated respectively inside or outside the predicatephrase.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



postposition

(8)BIA [Raidi-na¼ katu] Apikaru

Raidi-MkCase¼ com.instr Apikaru

‘Apikaru is with Raidi’

Monovalent heads:

verb

(9)BIA datikan pi:na

sink hook

‘the hook sank’

noun

(10)BIA totyawa idi:k wa

shaman you prospect

‘you will be a shaman’

adverb

(11)ITQ kodo kamodya

in_the_higher_part monkey_sp.

‘the monkey sp. is up there’

(12)BIA kiman idi:k

quickly you

‘quick!’

Clause-initial, final and intermediate positions can be occupied by dis-

course particles and adverbs in adverbial function. However, none of

these, nor any other kind of word, can intervene between a phrase head

and its left-adjacent case-marked dependent. For example, the particle

niama ‘then’ can be added to the sentence in (6) above as follows:

(13)BIA niama [Ayobi-na¼ bo:dak] tawami

then Ayobi-MkCase¼ roast dough

[Ayobi-na¼ bo:dak] niama tawami

Ayobi-MkCase¼ roast then dough

[Ayobi-na¼ bo:dak] tawami niama

again-MkCase¼ roast dough then

*[Ayobi-na¼ niama bo:dak] tawami

Ayobi-MkCase¼ then roast dough

*[Ayobi niama -na¼bo:dak] tawami

Ayobi then -Mkcase¼roast dough

‘then Ayobi roasted the manioc dough’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



As will be amply illustrated below, all the external arguments listed so

far occupy one and the same syntactic position. They also share the same

coding features in terms of case (zero marking) and position (typically

postpredicative). Since what we might wish to call the ergative case of (1),

marked by -na¼ , is in all coding respects identical to that of the genitive,

example (7), and of the ‘‘object’’ of a postposition, example (8), I employ

for these three semantically di¤erent but structurally identical instances

of -na¼ the generic label marked case. In fact, -na as an allative marker

is the only case a‰x allowed to occur on a noun phrase in an adjunct rela-

tion to the predicate, as in (14). All other meanings are rendered by post-

positions.8

(14)BIA Koni-na¼ dahu wankurun hak-na

Koni-MkCase¼ take_away pot house-all

‘Koni took the pot to the house’

2.3. Behaviour

This section is devoted to the syntactic asymmetries between external and

internal arguments, and will show the extent to which the unique argu-

ment of monovalent verbs does indeed pattern together with the patient

of transitive verbs.

2.3.1. Movement

Only external arguments can be moved from their postpredicative position

to a prepredicative, pragmatically marked one. This was seen in example

(5) above, resumed here.

(15)ITQ piya daan

man go;walk


This capacity for movement is shared by the external arguments of

predicates of all kinds, e.g. bivalent verbs and nouns (illustrated in (16)

and (17) respectively):

8. Of course, the diachronic hypothesis of a grammaticalization path allative >genitive > agentive, together with an ancillary hypothesis on the origin ofpostpositions as bivalent nouns, is appealing mostly because of its far-reachingconsequences for the understanding of the genesis of Katukina ergativity (seeQueixalos 2010). Postpositions display di¤erential object marking (see footnote14).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(16)ITQ ma-obatyawa kotyia-na¼ dyoro

3pl-wife otter-MkCase¼ copulate_with

‘the otter copulated with their wives’

(17)ITQ Inu Aro-na¼ tyo

Inu Aro-MkCase¼ daughter

‘Inu is Aro’s daughter’

One pragmatic e¤ect available for external argument fronting is that of

attenuated contrastive focus (see below). No movement of the internal

argument is allowed unless other formal changes take place – such as the

loss of case marking on the noun and the appearance of a personal prefix

on the verb, (19)–(20). Compare, starting from the basic constituent order

in (18):

(18)ITQ nyama-na¼ kionyuk a-okpu

mother-MkCase¼ comb 3sg-son

(19)ITQ a-kionyuk nyama a-okpu

3sg-comb mother 3sg-son

(20)ITQ nyama a-kionyuk a-okpu

mother 3sg-comb 3sg-son

‘the mother combs her son’

Regarding the pragmatic e¤ects of these movements, no information is

available for (19), while (20) seems to bear a nuance absent from (16),

along the lines of left-dislocated ‘as for the mother, she combs her son’.

As both (19) and (20) show, extraction of the agent noun phrase has to

leave behind the pronominal prefix on the verb. The constraint underly-

ing this alternation is simply that the internal argument must be realized

phonologically. In addition to any potential prosodic subtleties, the mere

presence of the pronominal prefix instead of the proclitic case marker is

evidence that the initial noun phrase nyama in (20), unlike in (18), is not

a component of the verb phrase. Now, section 2.3.5 will show that while

displaced external arguments retain their grammatical relationship with

the predicate, displaced internal arguments fall outside the clause core.

No attestation, whether spontaneous or elicited, shows the two move-

ments simultaneously – both left-dislocation for the internal argument

and fronting for the external argument. Moreover, my lack of control as

regards prosodic cues for constituency has prevented me from submitting

to my informants a plausible tentative example of a clause showing this

behaviour. The default hypothesis, then, would be that (20) shows a

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



moved agent phrase and a patient phrase in situ, rather than something

like (19) plus a fronted agent phrase.

2.3.2. Elision

Only external arguments can be elided.

(21)ITQ kitan-nin

sleep-dur

‘(he) is sleeping’

(22)BIA Koni-na¼ dahu

Koni-MkCase¼ take_away

‘Koni took (it) away’

The ‘‘zero pronoun’’ allows for a third person indefinite reading (see

section 3.2.10).

Elision of an internal argument has the same consequence as extraction,

namely the need for a pronominal verb prefix referring to this argument.

(23)ITQ a-hudyi homo

3sg-bring hammock

‘he brought the hammock’

An indefinite reading of the plural third person prefix is the closest

equivalent to a functional passive in this language.

(24)ITQ ma-dahu tyowipikon tyo

3pl-carry glass_beads excl

‘someone took away the glass beads!’

2.3.3. Ostension

Only external arguments can be modified or replaced by a demonstrative,

(25)–(26) and (27)–(28), respectively.

(25)ITQ kitan-nin itiyan wa:pa

sleep-dur this dog

‘this dog was sleeping’

(26)BIA yo-hoki ityian9 oman

1sg-put this log

‘I put this log (over there)’

9. The phonological form of the demonstrative is slightly di¤erent in the twodialects.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(27)ITQ kitan-nin itiyan

sleep-dur this_one

‘this one was sleeping’

(28)ITQ wa:pa-na¼ ti itiyan

dog-MkCase¼ kill this_one

‘the dog killed this one’

2.3.4. Coordination

No explicit element other than concatenation is involved in coordinating

noun phrases. External arguments can be coordinated, (29)–(30), but not

internal ones, (31).

(29)ITQ opatyin-na¼ wu awa nyama a-ponhanya

child-MkCase¼ want his_one mother 3sg-sister

‘the child loves his mother and his sister’

(30)ITQ tyuku wa:pa takara

die dog hen

‘the dog and the hen died’

(31)ITQ *Nodia Hanani-na¼ hoho-nin Owi10

Nodia Hanani-MkCase¼ call-dur Owi

‘Nodia and Hanani are calling Owi’

2.3.5. Focalization

Contrastive focus is achieved on noun phrases by the device of moving

them to clause-initial position and postposing the particle (ka)na. This

process is available for external arguments but not for internal ones.

(32)ITQ waro kana kitan-nin

parrot foc sleep-dur

‘it is the parrot that is sleeping’

(33)ITQ wiri na tyo-ikihak

wild_pig foc 1pl-spear

‘it is a wild pig that we speared’

10. One single instance of this construction was accepted by a speaker fromanother geographical area (Jurua).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



On the basis of (34), related sequences that were rejected by consultants

in elicitation include (35)–(36), where the scope of the focus particle would

be the internal argument.

(34)ITQ mapiri-na¼ duni takara

anaconda-MkCase¼ catch hen

‘the anaconda caught the hen’

(35)ITQ *mapiri-na (ka)na duni takara

(36)ITQ *mapiri (ka)na na¼duni takara

Focus displays an asymmetry which can be viewed as evidence that

the pre-predicate position – that is, the position preceding the predicate

phrase – is di¤erent for displaced external vs. internal arguments. The

former undergoes a dislocation which keeps it within the clause core,

and in that position it can be focused, (32)–(33). The latter is truly left-

dislocated (see section 2.3.1) and no longer available for focus, (37).

(37)ITQ *waro kana a-boni wa:pa

parrot foc 3sg-peck dog

‘it is the parrot that pecked the dog’

2.3.6. Constituent questions

External argument positions are eligible to be questioned.

(38)ITQ hanian tu11 tatan koniok-nin

who(m)/what int here talk-dur

‘who is talking here?’

(39)ITQ hanian tu no-toman?

who(m)/what int 2sg-shoot

‘what did you shoot?’

Internal arguments cannot be questioned as such.

2.3.7. Relativization

Data on relativization are too fragmentary to allow for reliable conclu-

sions regarding the structure of relative clauses and the nature of the rela-

11. In this dialect, the negation particle tu – or a homophonous element – occursin constituent and yes/no questions. In the Bia dialect the form of the interro-gation particle is yu, whereas the negation particle remains tu. See below fortwo comparable examples.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



tivizing element. The following is o¤ered as a first insight into one more

extraction process that seems to show the same asymmetries as have

already been seen in focus and interrogative constructions. In the Itaquai

dialect the presumably deictic element nyan – perhaps related to the free

third person pronoun, see (4) – opens the relative clause. The relativized

noun generally follows; this fronting is strongly preferred, though the

noun is sometimes found in situ. This means that relatives are basically

‘‘head-internal’’ in this language. The verb is su‰xed by -nin, which in

independent clauses denotes durative aspect, see (38), and on subordinate

predicates marks dependence. Only external arguments can be relativized.

(40)ITQ i-hik nyan anyan piya waokdyi-nin

1sg-know deic this_one man arrive-dep

‘I know the man who arrived’

(41)ITQ i-hi:k nyan tukuna Kontan-na¼ dahudyi-nin

1sg-know deic person Kontan-MkCase¼ bring-dep

‘I know the person that Kontan brought’

2.3.8. Nominalization

Morphology in bivalent verb nominalizations shows that these are inher-

ently patient-oriented, (42), as monovalent verb nominalizations are oriented

toward the unique participant whatever its specific semantic role, as in

(43)–(44). Nominalization is achieved by postposing to the lexical verb

stem the deictic element nyan (seen just above for relativization), while

retaining, in bivalent verbs, the original bound expression for the agent.12

Noteworthy di¤erences between nominalized verb and relativized clause

include not only the position of the deictic element nyan, but also the

absence vs. presence respectively of both a relativized noun and the sub-

ordinating verbal su‰x -nin. As for aspectuality, nominalization seems to

be equally suitable for denoting entities involved in events, as in (42) and

(43), and those characterized by properties / habitual activities, as in (44).

(42)aITQ yo-wahak barahai

1sg-cook wild_meat

‘I cooked the wild meat’

12. As far as the Itaquai dialect is concerned. In the Bia dialect nominalizationsare quite di¤erent in form, which suggests a diachronically recent innovationin at least one dialect, presumably the former.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



bITQ bak tu yo-wahak nyan

be_good neg 1sg-cook deic

‘the thing I cooked (lit. my cooked thing) is not good’

(43)aITQ dapoki opatyin

fall child

‘the child fell’

bITQ dapoki nyan

fall deic

‘the one that fell’

(44)aITQ donman piya

go_fishing man

‘the man went fishing’

(44)bITQ donman nyan adu

go_fishing deic 1sg

‘I am a fisherman’

The bivalent verb agent is not directly accessible to nominalization, see

3.2.8.

We have so far demonstrated that in basic bivalent clauses the argument

representing the patient ranks formally above the argument representing the

agent as far as constituency and behaviour properties are concerned (loca-

tion with regard to verb phrase, elision, movement, extraction), and aligns

with the unique argument of monovalent clauses in terms of these same

properties as well as coding properties (case marking, pronominal paradigms).

Let us now turn to a less neatly hierarchized domain, that of coreference.

2.4. Control

The simplest way of characterizing how arguments establish coreference

pivots between lexical noun phrases and zero or bound pronominal forms

is to say that in some sub-domains the hierarchy is straightforward

whereas in others the situation we face is somewhat fuzzy. (I will not supply

here a fully-fledged exposition of the topic, which can be found in Queixalos

2004, 2010.) Let us begin with the latter.

At the intraclausal level, functional (semantic, pragmatic) conditions

prevail over syntactic constraints such as linear order and rank in the con-

stituency hierarchy (‘‘c-command’’) for core arguments, (45)–(47) (square

brackets delimit the verb phrase). In (45) the external argument controls

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



the possessive marker prefixed to the internal argument, as expected from

the constituency hierarchy but not from linear order. In (46) the internal

argument controls the possessive on the external argument, counter the

constituency hierarchy but in agreement with linear order. Disjunct refer-

ence appears in (47).

(45)ITQ [a1-obatyawa-na¼ todiuk] Mayon13sg-wife-MkCase¼ hate Mayon

‘Mayon1’s wife hates him1 (lit.: His1 wife hates Mayon1’)

(46)ITQ [Dawi1-na¼ bobo] a1-obatyawa

Dawi-MkCase¼ beat 3sg-wife

‘Dawi1 beat his1 wife’

(47)BIA [ pi:da1-na¼ buro:] a2-mimi

jaguar-MkCase¼ lick 3sg-blood

‘Jaguar1 licked his2 blood’

Other domains show a slight preference for pivots where the patient

argument is involved. One example is intraclausal coreference between

core arguments and adjuncts; (48) and (49) were originally uttered with

the meanings given here, but in the absence of contextual clues informants

tend to interpret the latter with the patient as the antecedent (i.e. ‘. . . in her

house’). For disjunct reference a free pronominal form (anyan) is appealed

to, (50).13

(48)ITQ Dawi1-na¼ toman Poroya2 [a2-wa mokawa katu]14

Dawi-MkCase¼ shoot Poroya 3sg-grn gun com.instr

‘Dawi1 shot Poroya2 with his2 gun’

13. The gender of pronouns in the English translations reflects the circumstancesin which the examples occurred. Square brackets delimit the postpositionalphrase. I have no explanation for the need to code disjunct reference on thepostpositional phrase in (50) but not on the external argument of (47). Inter-sentential coreference is another domain where pivots involving the patientare preferred.

14. Mokawa, ‘gun’, and hak, ‘house’, are monovalent nouns – i.e. nouns unable tohead a phrase containing a genitive. To do so, they must let the generic rela-tional noun (NGR) -wa mediate between them and the genitive expression,whether this is a person prefix, as in the example, or a case-marked lexicalnoun. Usually, non-human nouns as internal arguments of postpositions donot take the case marker -na¼.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(49)ITQ Dawi1-na¼ tohik ityaro2 [a1-wa hak to]

Dawi-MkCase¼ see woman 3sg-grn house loc

‘Dawi saw the woman in his house’

(50)ITQ Dawi1-na¼ tohik ityaro2 [anyan3-na¼ wa hak to]

Dawi-MkCase¼ see woman 3sg-MkCase¼ grn house loc

‘Dawi1 saw the woman2 in his3/her3 house’

Involvement of the patient in the coreference pivot is the default with

intraclausal adverbs of manner and location, and in subordination. I give

an example of the latter only, in order to keep this section within reason-

able limits (square brackets delimit the dependent clause). As mentioned

above, the su‰x -nin operates both as a durative aspect marker on matrix

predicates, (21), and as a subordinator.15

(51)BIA a1-makaudyaran Ø2 [dyahian-nin Ø2 ama]

3sg-stand_astride stand_up-dep goal

‘He stood over her to have her stand up’

An obligatory patient pivot shows up in control constructions.16 Com-

pare (52)–(53) for alignment between patient and unique argument (square

brackets delimitate the predicate phrases).

(52)ITQ [[i-ti-nin¼ ] wu] idi:k

1sg-kill-dep¼ want 2sg

‘I want to kill you’

(53)ITQ [[donman-nin¼ ] wu] adu

go_fishing-dep¼ want 1sg

‘I want to go fishing’

A few comments are in order. In this kind of construction we have a

main finite verb, here wu, ‘want’, heading a clause whose internal argu-

ment is a clause complement containing the subordinate non-finite lexical

verb, here bivalent ti, ‘kill’ and monovalent donman, ‘go fishing’, marked

15. I assume a diachronic link between the two functions. Plausibly this was orig-inally a nominalizer, a function which survives in subordination but was rean-alysed as aspect on main verbs.

16. A possible alternative interpretation (raising) is suggested in Queixalos (2010).Since in terms of hierarchy the conclusion remains untouched – patientprivileged – I will not go into details here.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



for dependency by -nin, which phonologically procliticizes to the syntacti-

cally main verb wu – the ‘‘auxiliary’’.17

The crucial point is that the external argument of the bivalent non-

finite complement clause is either the patient, ‘you’ in (52), or the unique

in its monovalent counterpart, ‘I’ in (53). (It should also be noted that the

external argument of the main verb is coreferential with the external argu-

ment of the complement clause.18)

I label as ergative all the two-place clauses seen so far. They feature the

formal properties induced by basic active transitive predicates. However,

an alternative two-place clause type exists which is accusatively aligned,

as far as I can tell at present, in all the aspects enumerated in the previous

sections (see 3.2.10 for details).

2.5. Subject and object

What we have, then, is a morphosyntactic configuration where, in bivalent

clauses, all the formal properties currently attributed to the notion of

grammatical pivot converge almost perfectly on the patient argument of

the bivalent clause, as they do on the unique argument of the monovalent

clause.

I wish, however, to make two caveats on aspects of the evidence

adduced here that could weaken my interpretation.

The first of these concerns the value of nominalization in identifying

the alignment patterns of a language. A particular set of so-called ergative

alignments attested cross-linguistically should deserve a special status

in typology, since they appear 1) in peripheral regions of grammatical

systems – ‘peripheral’ meaning non-basic alignments and non-basic clause

types; and 2) in languages of all kinds, be they ergative, accusative, active

or other. Argument alignments in nominalizations are among the most

17. Phonologically, these utterances are organized as /iti##ninwu##idi:k/ and/donman##ninwu##adu/. Like the case marker -na¼, the dependence-marking su‰x -nin has undergone a diachronic process of head attraction,leaving its grammatical host to bind phonologically to the phrase syntactichead that immediately follows it. Consistently with the stand taken in footnote4, and even at the cost of introducing some diachrony in a synchronic account,I let considerations of grammatical structure override those of phonologicalstructure in presenting the examples.

18. Which means that, for the construction with a bivalent complement clause atleast, this argument is not a semantic participant of the main verb at all. Thisis the point in favour of a raising reading of this structure, in spite of theatypicality of the putative raising verb.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



notorious exponents of what I call ubiquitous ergativity, along with

number distinctions on the arguments of simple / reduplicated verbs,

‘‘possessor’’ raising through nominal incorporation, and more (Keenan

1984). Indeed, if the facts of nominalization displayed in 2.3.8, and in

3.2.8 below, are merely instances of ubiquitous ergativity, they should

not count as criterial for the establishment of the Katukina language type

in terms of alignment. But what is at stake here, as will shortly become

evident, is the morphology of voice as captured for nominalization purposes,

which shows that bivalent verbs are clearly patient-oriented.

Secondly, there is the possibility that my claim is underdetermined by

the data on coreference. The language can be grossly characterized as

pivotless for coreference or, more accurately, as slightly biased towards

an ergative-type pivot. For this reason, coreference can only be counted

as a weak indicator, agreeing with much more robust ones in pointing to

ergative syntax for Katukina. However – and this is crucial to our under-

standing of ergativity – the fact that the only real weak zone of erga-

tive syntax is precisely coreference shows, in my view (Queixalos 2010),

that this kind of grammatical organization is of recent appearance in the

language.

We return now to the issue of grammatical relations. The original moti-

vation for the notion of pivot in Dixon (1994) was to subsume the formal

properties of subjects in a single cluster so as to let semantically-based

notions like agent permeate the notion of subject. It should be clear from

what precedes that this mixing of levels is in my view the key factor under-

lying a significant portion of the chronic misunderstandings that have

weighed upon the di¤erent approaches to ergativity. If the mapping of

semantic roles on to the expression of arguments were not taken into con-

sideration, the facts adduced above would lead anyone to clearly posit the

existence of a grammatical subject and a grammatical object in Katukina

ergative clauses. This is exactly the stand I take,19 since the ergative clause

presented so far cannot be seen as an inverse: there is no direct clause to

be held as its more basic counterpart. In Katukina, then, the linguistic

expression of the semantic role of patient in a basic bivalent clause dis-

plays the formal properties of subjects; if, as I believe, the grammatical

relation subject is formal in nature, then the patient argument is the sub-

ject, along with the unique argument of the monovalent clause; the other,

lower-ranked argument in the basic active bivalent clause, can only be an

19. At first sight this is in line with Givon’s (1997: 34) statement that formal prop-erties reflect grammatical relations ‘‘more faithfully’’; my view is in fact qual-itatively di¤erent in that, for me, this is not a question of more or less.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



object, despite its correlate in terms of semantic roles, namely the agent.

This, of course, runs counter to the general assumption of a radical incom-

patibility between the semantic role of agent and the grammatical relation

of direct object (e.g. Givon 2001: 200). Now, the fact that grammatical rela-

tions are, synchronically, formal entities does not rule out the possibility of

diachronic grammaticalization processes whereby functional motivations –

mainly pragmatic, e.g. topic maintenance – can be the source of converging

forces that lead a given argument to capture the set of characteristics which

will make it the syntactically privileged argument of a given construction,

i.e. the subject (I take up the issue of diachrony in Queixalos 2010).

3. Antipassive

As we have seen, in the ergative clause the expression of the agent is barred

from a number of properties attached to subjects, such as sisterhood with

the predicate phrase (i.e. sharing its level of constituency), zero case mark-

ing, pronominalization by free forms, extraction, and so on. However, the

expression of the agent can access all of these properties by means of a

voice process.

3.1. Form

The antipassive is built on the basis of the ergative clause by suppressing

the internal argument position. Its formal properties are:

e the agent prefix paradigm slot is made inaccessible to any referent by

means of an invariable morpheme wa- that blocks the agent’s morpho-

logical slot;e no noun phrase can appear within the verb phrase;e the agent surfaces as an external argument;e the patient, demoted from its external argument position, is either

omitted (54), instantiated as the object of a postposition (55), or instan-

tiated with no relational marking (whether case or postposition) (56);

the postposition for the demoted noun phrase in the antipassive clause

is the comitative instrumental marker katu, already seen in examples

(8) and (48).20

20. Only the Bia dialect allows for the oblique patient antipassive. Several indica-tions converge on the idea that, with regard to the evolution of ergativity, thisdialect may be more conservative than Kanamari.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



See examples:

(54)ITQ wa-pu adik tyo

antip-eat 1pl excl

‘we eat!’ (Context: ‘We are happy in our land’.)

(55)BIA wa-wu dyara tukuna anya-na¼ katu

antip-want white_people Indian woman-MkCase¼ com.instr

‘Whites like Indian women’

(56)ITQ piya wa-pu-nin barahai

man antip-eat-dur wild_meat

‘men are eating wild meat’

As far as the verb-argument core is concerned, constituency factors do

not constrain word order any longer, since the predicate phrase is devoid

of any internal, dependent, noun phrase. As an external argument, the

agent phrase is basically post-verbal, as in (54)–(55). But as such it can

also be fronted, as in (56). If both participants are overtly present, which

is rare in spontanous discourse but accepted unproblematically in elicita-

tion, a few restrictions obtain.

When instantiated as the object of a postposition, the patient phrase

is post-core, as in (55). One clue to its adjunct status is the behaviour of

the future particle wa. Its position in the clause is necessarily post-core

and predominantly – though not obligatorily – final. Both (57) and (58)

are grammatical, whereas (59) is not.

(57)BIA wa-toman adu wiri katu wa

antip-shoot 1sg wild_pig com.instr fut

‘I will shoot wild pigs’

(58)BIA wa-toman adu wa wiri katu

antip-shoot 1sg fut wild_pig com.instr

‘I will shoot wild pigs’

(59)BIA *wa-toman wa adu wiri katu

antip-shoot fut 1sg wild_pig com.instr

Despite its adjunct status, the patient expression has a privileged status

among non-core constituents, since all other adjuncts appear after it, even

when headed by the same postposition -katu, (60).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(60)BIA wa-toman adu wa wiri katu mokawa katu

antip-shoot 1sg fut wild_pig com.instr gun com.instr

‘I will shoot wild pigs with a gun’

It is not unambiguously clear whether the instantiation of a patient by

means of a bare noun phrase still bears any grammatical relation to the

predicate, but the answer is presumably negative, see section 3.1. As

regards word order, noun phrases are either both located preverbally, in

which case the agent always precedes the patient, (61), or distributed one

on either side of the verb, the agent almost always appearing before the

patient, (62), with very few cases of the reverse order, (63).

(61)ITQ kaina moroho idi:k wa-binik tyo

toad_sp. 2sg antip-swallow excl

‘the toad sp. swallowed you!’

(62)ITQ piya wa-pu barahai

man antip-eat wild_meat

‘the man eats wild meat’

(63)BIA tawa wa-pukni Kirak

manioc antip-pull_out Kirak

‘Kirak harvested manioc’

Note that the preverbal position of the patient is compatible with an

adjunct status, since postpositional phrases are accessible to fronting, as

in the following primitive one-place clause:

(64)ITQ ityowa ityonin naki adik tyuru

poss.1pl territory loc 1pl grow

‘in our territory we grew up’

No saliency hierarchy seems to be at work here, in the sense that some-

thing of a core argument status is being conferred to the patient phrase

in a more or less inverse fashion. See the following examples, where the

patient is either preverbal but low on the animacy scale, (65), or first

person but in postverbal position, (66).

(65)ITQ Wura poako wa-buhuk a-ama21

Wura paddle antip-make 3sg-goal

‘Wura made a paddle for him’

21. See this postposition as purpose subordinator in (51).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(66)ITQ Owi wa-hoho adu

Owi antip-call 1sg

‘Owi called me’

When only one participant is granted overt expression in the antipassive, it

is practically always the agent. This state of a¤airs is massively repre-

sented, whereas instances of patient noun phrases are extremely rare in

texts and straightforwardly rejected in elicitation: our informant, when

presented with (67), rejected it on the grounds that ‘‘Pacu fishes don’t

spear’’.

(67)BIA *wa-hak bamak

antip-spear pacu_fish

Antipassive clauses consisting of the verb alone are more frequent in texts

than all antipassives with one or two lexically instantiated participants put

together. Most of the time we observe that what is being reported is an

activity, not an event,22 and that the agent is either a stable topic in the

fragment of discourse under scrutiny, or first person,23 or both, as in this

excerpt from an ode to the happiness of life in ancient times (when there

was plenty of food, etc.).

(68)ITQ wa-pu niama kotuda

antip-eat then again

‘then we ate again’

To sum up, the Katukina antipassive displays a formal device that

consists in preventing the verb from taking an internal argument. The

morphological means of e¤ecting this is to block the agent prefix para-

digm with a marker wa-. The agent migrates to the external argument

position, which is also that of the single argument of monovalent verbs

and the patient of bivalent verbs. Since the patient either migrates to an

adjunct (sometimes obliquely marked) position or, more frequently, is

deprived of linguistic expression altogether, we can safely say that we are

22. Here I wish to posit a distinction between event: conditions of existenceendowed with 1) dynamicity, 2) spatio-temporal coordinates, and 3) one ormore participant(s); and activity: conditions of existence serving to characterizean entity because they involve that entity in a repetitive or (more or less) exclu-sive manner.

23. First person as external argument frequently remains unrealized in spontane-ous speech.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



left with a typical one-place clause. However, the patient can also be instan-

tiated through a bare noun phrase. One could think of this as occupying a

non-syntactic position, since no clause type with two external arguments is

otherwise attested in the language.24 It might be analysed as something of

an afterthought element, were it not for examples like (61), recalled here:

(69)ITQ kaina moroho idi:k wa-binik tyo

toad_sp. 2sg antip-swallow excl

‘the toad sp. swallowed you up!’

We could perhaps be led to admit the possibility that in some specific

cases, the Katukina antipassive retains the non-promoted argument. After

all, if the antipassive in a syntactically ergative language is overwhelmingly

devoted to the function of lending pivothood to the agent, the natural out-

put of the voice process ought to be one that retrieves both arguments, since

the action depicted by the verb has undergone no change as to the number

and identity of its central participants. I return to this issue below.

3.2. Motivations

Obviously, such a large number of restrictions imposed upon the expres-

sion of the ergative clause agent – as detailed in sections 2.3 and, to a

lesser extent, 2.4 – necessarily entails, as a direct e¤ect, the existence of

a voice device whereby these restrictions are circumvented. Because this

language is almost homogeneously ergative, the antipassive has mainly

formal motivations, and its functional ones are somewhat di‰cult to detect.

Here I will take one by one the processes enumerated in sections 2.3

and 2.4 as being barred for access to the ergative internal argument, in

order to show how they apply to the expression of an antipassive agent.

3.2.1. Movement

An antipassive agent can be moved to clause-initial position, as in

(70)ITQ ikik wa-pu-nin barahai

one antip-eat-dur wild_meat

‘only one is eating wild meat’

24. With the exception of ‘say’ clauses: see the Introduction.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



3.2.2. Elision

An antipassive agent can be elided, if recoverability of reference is granted.

(71)ITQ wa-o-nin

antip-drink-dur

‘(she) was drinking’

3.2.3. Ostension

An antipassive agent can be modified, as well as pronominalized, by a

demonstrative. Respectively:

(72)ITQ itiyan kawahiri wa-duni tyon

dem.prox cat antip-catch rat

‘this cat caught the rat’

(73)ITQ itiyan wa-duni tyon

dem.prox antip-catch rat

‘this one caught the rat’

3.2.4. Coordination

Two nominals referring to two participants in the same agent role can

be coordinated if expressed as the external argument of an antipassive

predicate.

(74)ITQ Nodia Hanani wa-hoho-nin Owi

Nodia Hanani antip-call-dur Owi

‘Nodia and Hanani were calling Owi’

3.2.5. Focalization

An antipassive agent can be focused.

(75)ITQ Aro kana wa-nuhuk a-batyawa kariwa-na¼ ton

Aro foc antip-give 3sg-wife non_Indian-MkCase rec

‘It was Aro who gave his wife to the white man’

3.2.6. Constituent questions

An antipassive agent can be questioned.25

25. Recall the slightly di¤erent form taken by the interrogative pronoun in thetwo dialects (hanian / hanin), as well as the interrogative particle (tu / yu).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(76)BIA hanin tan wa-dyuman tahi yu?

who(m) here antip-spill water int

‘Who spilled the water here?’

(77)ITQ hanian tu kana Pawi wa-toman tyo?

who(m) int foc Pawi antip-shoot excl

‘who killed Pawi?’

3.2.7. Relativization

An antipassive agent can be relativized. Compare (78) to the relativization

of the patient in an active clause, (41), renumbered here as (79).

(78)ITQ i-hi:k nyan piya wa-dahudyi-nin Hiowai

1sg-know deic man antip-bring-dep Hiowai

‘I know the man who brought Hiowai’

(79)ITQ i-hi:k nyan tukuna Kontan-na¼ dahudyi-nin

1sg-know deic person Kontan-MkCase¼ bring-dep

‘I know the person that Kontan brought’

3.2.8. Nominalization

A bivalent verb (a-examples below) cannot undergo agent nominalization

unless it is recast in the antipassive (b-examples):

(80)aITQ a-bi:wik-nin obakon

3sg-smoke-dur cigar

‘he is smoking the cigar’

bITQ i-toman wa-bi:wik nyan

1sg-shoot antip-smoke deic

‘I shot the smoker’

(81)aBIA a-hak bamak

3sg-spear pacu_fish

‘he speared a pacu fish’

bITQ ki:tan wa-hak nyan

sleep antip-spear deic

‘the spearer slept’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



3.2.9. Control

Briefly, in active ergative clauses no straightforward coreference pivot is

observed except for proper control structures, where patient (and unique)

pivots are mandatory (see 2.4 above in fine). Otherwise, either patient or

agent can establish distant coreference links. But a bias towards the

patient can be observed in several areas, such as adverbial phrases (posses-

sion on object of postposition, semantic scope of manner and location

adverbs) and interclausal relations (subordination, coordination).

(82) is a sequence of clauses where the antipassive is required to license

the inclusion of a bivalent verb agent and a monovalent verb participant

in a coreference pivot. In elicitation, when faced with strictly symmetrical

extralinguistic situations leading to potential ambiguity such as (83), the

speaker spontaneously antipassivizes the verb to make clear the involve-

ment of its agent in a coreference pivot.

(82)ITQ padyi, wa-pu niama, koniohin niama

arrive antip-eat then dance then

‘they arrived, then they ate, and then they danced’

(83)ITQ Nodia-na¼ toman Yowai a-tohi:k-nin annin

Nodia-MkCase¼ shoot Yowai antip-stare_at-dep emph26

‘Nodia shot Yowai while the latter was staring at something’

We have listed no fewer than eight purely syntactic situations where

the antipassive is required to allow accessibility to the agent of a bivalent

verb. None of these applies to the genitive noun phrase, despite the fact

that its coding and constituency properties are identical to those of the

agent of an ergative clause. For instance, modifying or pronominalizing a

noun by means of a demonstrative, (84)–(85), or questioning a referent,

(86), are plainly admitted for a genitive noun. This proves that the internal

arguments of verb and noun phrases are distinct syntactic elements.27

(84)ITQ daan niama itiyan ityaro-na¼ tyo

go;walk then dem woman-MkCase¼ daughter

‘then the daughter of this woman went away’

26. The basic meaning of tohi:k is simply ‘look at’. It is still unclear what particularkind of pragmatic emphasis this form – an-nin, copula-dependence – conveys.

27. Or, to put it di¤erently, that noun and verb phrases are not alike in syntacticstatus.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(85)ITQ itiyan-na¼ tyo kana tona-nin tyo

dem-MkCase¼ daughter foc leave-dur excl

‘it is this one’s daughter that is leaving’

(86)ITQ hanian-na¼ okpu tu annin?

who(m)/what-MkCase¼ son int emph

‘whose son is this?’

Concerning the focus example (75), notice that although the motivation

for focusing is in itself of a pragmatic nature, its syntactic consequences –

the fronting of the noun phrase and the postposing of the focus particle

(ka)na – are the direct formal reason for resorting to the antipassive when

the process applies to the internal argument of the active verb phrase.

We turn now to the functional motivations for the antipassive.

3.2.10. Semantics & pragmatics

A few common semantic or pragmatic constraints inducing use of the

antipassive cross-linguistically are taken over in Katukina-Kanamari by

other, and diverse, formal devices.

One possible function for the antipassive is the pragmatic promotion of

the agent. Specifically in Katukina, this promotion is tantamount to con-

trastive focus. We know that this kind of pragmatic process is achieved by

means of the particle (ka)na on external arguments. Section 2.3.5 shows

examples (32)–(33) for unique and patient arguments, and section 3.2.5

shows example (75) for the antipassive agent. However, the simple front-

ing of an external argument (remember that the canonical position for an

external argument is post-verbal, cf. section 2.2), hence of the antipassive

agent, has something of an attenuated contrastive focus e¤ect. Informants

insist that the best equivalent for (87) is not the plain ‘‘my wife cooked

wild meat’’ but something like ‘‘it’s my wife that cooked wild meat’’. A

fine spontaneous example of this e¤ect is seen in (61), despite the more

neutral translation given above. Plausibly more than a single degree of

focus may be available to speakers.

(87)ITQ yo-obtayawa wa-wahak bara

1sg-wife antip-cook wild_meat

Among the morphosyntactic devices triggered by the properties of the

patient we find the accusative pattern referred to in section 2.4. The lan-

guage features a transitive split whereby an accusatively aligned construc-

tion is employed when the patient is semantically generic, (88)–(89). Iterative

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



or habitual aspect is not necessarily involved, as the first clause of (90)

shows. The accusative clause is a perfect inversion of the ergative clause

in terms of constituency: it has the patient as its internal preverbal argu-

ment, and the agent as its external and typically postverbal argument.

The patient is obligatorily instantiated through a noun phrase since no

person prefix attaches to the verb. The formal properties of both argu-

ments are more or less identical for ergative and accusative patterns, as

long as we express these properties in terms of internal / external noun

phrases. They are, of course, inverted in terms of semantic roles.

(88)ITQ [wiri hak] adu

wild_pig spear 1sg

‘I speared wild pigs’

(89)ITQ [takara duni ] mapiri dawa

hen catch anaconda today

‘the anaconda is catching hens today’

(90)ITQ [mokawa wu] adu [wiri toman] niama

gun want 1sg wild_pig shoot purp28

‘I want a gun to shoot wild pigs’

While the language allows for noun incorporation, the accusative pattern

cannot be analysed in those terms since monovalent (i.e. ‘‘alienable’’) nouns

may only be incorporated if the verb also undergoes an applicative process

(see in Queixalos 2008 the notion of redistributive incorporation). Compare

(91), with a bivalent noun incorporated, to (92)–(93) with a monovalent

noun.

(91)ITQ nyama-na¼ ki-onyuk a-okpu

mother-MkCase¼ head-scratch 3sg-son

‘the mother combed her son’

(92)ITQ *yo-obtayawa-na¼ bara-wahak

1sg-wife-MkCase¼ wild_meat-cook

‘my wife cooked wild meat’

28. The grammatical status of niama is unclear here. This particle serves as thediscourse connector ‘then’, and also as a purpose subordinator, as it seems todo here in spite of the lack of the verbal dependence su‰x -nin in the example.A more accurate translation might perhaps be ‘I want a gun. Then I’ll shootwild pigs’, but the problem in that case would be the absence of the futureparticle wa.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(93)ITQ yo-obtayawa-na¼ ma-bara-wahak

1sg-wife-MkCase¼ appl-wild_meat-cook

‘my wife cooked wild meat for him’

However, there seems to be some overlap in the semantics of the

patient between antipassive and accusative clauses: no di¤erential quanti-

fication feature has so far been identified between the patients in (88) and

(95). Nor can the referential status of the patient be held as distinctive

between the two patterns. Compare, for the accusative clause, (88) with

a referential patient to (90) with two non-referential patients, and for

the antipassive clause, (94) with a referential patient to (95) with a non-

referential patient.

(94)BIA hanin koya wa-buhuk?

who(m) pap antip-make

‘who made the pap?’

(95)ITQ adu don wa-buhuk-nin¼ bak

1sg fish antip-make-dep¼ be_good

‘I am a good fisherman (lit. I am good at making fish)’

Another common functional feature associated with antipassive patients

is indefiniteness. This language shows no incompatibility between an in-

definite patient and the ergative clause, and the simplest means of packag-

ing an indefinite patient is to let a zero pronoun (‘‘pro’’, see 2.3.1) fill the

external argument slot of the ergative clause provided that no plausible

referent is available in the situational or discourse environment. This will

automatically bring about an indefinite reading. The following examples

show a piece of information containing an indefinite patient with no overt

expression, (96), followed by a quite natural question, (97). A single noun

phrase preceded by the indefinite prefix a- is no less natural an answer.

This prefix allows the indefinite patient to be represented as an explicit

noun phrase, (98).

(96)ITQ oman-na¼ ti na tyo

tree-MkCase¼ kill foc excl

‘the tree killed someone!’

(97)ITQ hanian ti tu na oman-na¼ ti tyo?

who(m)/what restr int foc tree-MkCase¼ kill excl

‘whom did the tree kill, precisely?’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(98)ITQ oman-na¼ ti a-tukuna

tree-MkCase¼ kill indef-Indian

‘the tree killed someone [who is an Indian]’

We have come across several instances of ergative clauses in which the

agent outranks the patient in the saliency hierarchy, e.g. (1), (6), (14), (47),

to mention but a few. Now, not only can the same pattern obtain in anti-

passive clauses, as in (56), but the reverse pattern (in which the patient is

more salient than the agent) does not motivate the use of an antipassive,

as (99) shows. Nor do semantic hierarchies in the accusative clause favour

an inverse interpretation, although generic patients tend per se to be less

salient. (100) shows a human patient alongside a non-human agent, to be

compared with (101) where, in the same accusative pattern, a non-human

‘‘patient’’ faces a human ‘‘agent’’.29 Hence, neither antipassive nor accusa-

tive clauses can be viewed as forming part of a direct / inverse system.

(99)ITQ dyoko-na¼ hak-dyi adu tyo

dart-MkCase¼ perforate-centrip 1sg excl

‘the dart perforated me’

(100)BIA pi:da ityaro botyana

jaguar woman follow

‘<the> jaguar>s< follow<s> women30’

(101)BIA pi:da ohiya Ayobi

jaguar fear Ayobi

‘Ayobi fears jaguar(s)’

Now, if we assume that strictly formal motivations should lead anti-

passive clauses to retain overt expression of the patient, whether bare or

oblique, the very existence of antipassive clauses with covert patients is a

clear clue to the possibility of concomitant functional motivations for

voice alternation in this language. We have come across a typical instance

of such a functional antipassive above: description of an activity (vs. an

event), in (68). An additional example is (102). Close to this function is

29. Quotation marks are for the non-prototypical agent and patient required bythis particular verb.

30. The ‘inverted angle brackets’ notation is used for obligatory disjunctive occur-rence: <x> y <z> stands for xy, yz, *y, *xyz, that is: either x or z must occur,but not both.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



the description of an ability (or inability), as in (103).31 What both types

have in common is that the patient is not interesting in its own right: its

identity is insignificant, irrelevant, obvious.

(102)ITQ opatyin hinuk niama wa-o ti

children group then antip-drink restr

‘then the children just drank’

(103)BIA wa-hak tu (adu)

antip-spear neg 1sg

‘I’m a bad fisherman (lit.: I don’t spear)’

4. Conclusion

To sum up, the formal motivations for the antipassive in Katukina are

clear. This does not rule out the possibility that functional motivations

exist alongside them. The latter, however, partially overlap and compete

with those of other morphosyntactic devices catering to the speaker’s

intention of semantically / pragmatically promoting an agent or demoting

a patient. The residue of antipassives whose motivations are not formal

or whose functions are not also taken over by other grammatical devices

consists, at the present state of our knowledge, of the description of two

semantically contiguous configurations: activities and agentive abilities.

The overwhelming importance of formal motivations in voice selection

settles, in my view, the issue of the grammatical relation hierarchy in

this language, and confirms the direct objecthood of an argument – the

agent – which is in the marked case, internal to the verb phrase, and pro-

motable to syntactic subject status by means of recessive voice change. In

sum, the formal side of morphosyntax in a syntactically ergative language

is much the same as that observed in many accusative languages. At first

sight, a substantial qualification on this similarity is the ever-present fea-

ture of split transitivity which is so characteristic of ergativity. But note

that accusative languages have their own areas of split transitivity: besides

the existence of ‘ubiquitous ergativity’ as mentioned above, we find di¤er-

ential marking of objects and even di¤erential marking of subjects. The

crucial specificity of syntactically ergative languages is not, then, their

31. These aspects are akin to imperfectivity, a common feature of antipassives.Imperfectivity plausibly accounts for the observed slightly higher frequencyof the durative (-nin) in antipassive clauses. I was led to verify this point aftera remark by K. Haude.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



formal mechanism per se, which is in itself rather common whatever the

basic alignment type, but merely the mapping of semantic roles on to

grammatical relations (Mel’cuk 1979, Marantz 1984). Our current ideas

on this topic are certainly valid generalizations for 99.9 per cent of the

documented languages of the world, but the tiny 0.1 per cent remainder

means that they cannot be treated as strict defining features of human

language.

Abbreviations

1 first person

2 second person

3 third person

all allative

antip antipassive

appl applicative

centrip centripetal

com comitative

deic deictic

dem demonstrative

dep dependent marker

dur durative

excl exclamatory marker

foc focus

fut future

goal goal

grn generic relational noun

indef indefinite

instr instrumental

int interrogation

loc locative

MkCase marked case

neg negation

pl plural

poss possessive

prospect prospective

prox proximal

purp purposive

rec recipient

restr restrictive

sg singular

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



References

Adelaar, Willem F. H.2000 ‘‘Propuesta de un nuevo vınculo genetico entre dos grupos lin-

guısticos indıgenas de la Amazonıa occidental: Harakmbut yKatukina.’’ In Actas del I Congreso de Lenguas Indıgenas deSuramerica 219–236. Universidad Ricardo Palma.

Comrie, Bernard2008 ‘‘What is a passive?’’ In Studies in Voice and Transitivity, Zarina

Estrada, Søren Wichmann, Claudine Chamoreau and AlbertAlvarez (eds.), 1–18. Munchen: Lincom Europa.

Cooreman, Ann1994 ‘‘A functional typology of antipassives.’’ In Voice: Form and Func-

tion, Barbara Fox and Paul J. Hopper (eds.), 49–88. Amsterdam/Philadelphia: John Benjamins.

Givon, T.1997 ‘‘Introduction.’’ In Grammatical Relations: a Functionalist Per-

spective, T. Givon (ed.), 1–84. Amsterdam / Philadelphia: JohnBenjamins.

Givon, T.2001 Syntax: an Introduction. Vol. 1. Amsterdam / Philadelphia: John

Benjamins.Givon, T.

2008 ‘‘On the relational properties of passive clauses: A diachronicperspective.’’ In Studies in Voice and Transitivity, Zarina Estrada,Søren Wichmann, Claudine Chamoreau and Albert Alvarez(eds.), 19–32. Munchen: Lincom Europa.

Givon, T.2009 The Genesis of Syntactic Complexity. Amsterdam / Philadelphia:

John Benjamins.Keenan, E.

1984 ‘‘Semantic correlates of the ergative/absolutive distinction.’’ Lin-guistics 22: 197–223.

Marantz, A.1984 On the Nature of Grammatical Relations. Cambridge (Mass.): MIT

Press.Mel’cuk, I.

1979 Studies in dependency syntax. Ann Arbor: Karoma.Queixalos, F.

2004 ‘‘Split transitivity and coreference in Katukina.’’ In Ergatividadena Amazonia III, Atas do terceiro encontro do projeto Manifesta-coes da ergatividade na Amazonia, Queixalos, F. (ed.), 175–188.Paris: CELIA/CNRS.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Queixalos, F.2008 ‘‘Incorporation nominale en sikuani et en katukina-kanamari.’’

Amerindia 31: 61–86.Queixalos, F.

2010 ‘‘Grammatical relations in Katukina-Kanamari.’’ In Ergativityin Amazonia, Spike Gildea and Francesc Queixalos (eds.), 237–285. Amsterdam / Philadelphia: John Benjamins.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Undergoer orientation in Movima

Katharina Haude

1. Introduction1

This article explores the system of verbal voice morphemes in Movima

(unclassified, Amazonian Bolivia) and seeks to explain why most transi-

tive main clauses in Movima pattern ergatively. Movima has two basic

transitive constructions, direct and inverse, overtly distinguished by verbal

morphemes. In main clauses, to which the discussion in this paper is

restricted, the direct construction patterns ergatively and the inverse con-

struction patterns accusatively. In terms of statistical frequency, the large

majority of transitive main clauses in texts is direct, i.e. ergative.

1. The data on which the study is based were collected in Santa Ana del Yacumabetween 2001 and 2009, financed by the Spinoza programme Lexicon andSyntax (Radboud University Nijmegen) and by the Movima project of theDobeS programme of the Volkswagen Foundation. I am deeply greatful tothe Movima speakers who shared their knowledge with me. Spike Gildea andFrancesc Queixalos are thanked for their critical remarks on an earlier versionof this paper, without being responsible for any shortcomings. Elicited exam-ples are marked with [el], all other examples stem from recorded spontaneousdiscourse. Symbols (partly adapted from the Leipzig Glossing Rules) are:¼internal cliticization; – external cliticization; < > infixation; P reduplica-

tion. Abbreviations in glosses are: 1 ¼ first person; 2 ¼ second person;3 ¼ third person; ab ¼ absential; abs ¼ absolute state; agt ¼ agentive; appl ¼applicative; art ¼ article; be ¼ bound nominal element; ben ¼ benefactive;caus ¼ causative; co ¼ co-participant; d ¼ dummy; dem ¼ demonstrative;det ¼ determiner; detr ¼ detransitivizer; dr ¼ direct; dsc ¼ discontinuous;dur ¼ durative; ep ¼ epenthetic vowel; ev ¼ evidential; f ¼ feminine;frust ¼ frustrative; inal ¼ inalienable; inv ¼ inverse; hyp ¼ hypothetical;imm ¼ immediate past; instr ¼ instrumental; irr ¼ irrealis; itn ¼ inten-tional; intr ¼ intransitive; loc ¼ location; m ¼ masculine; md ¼ middle;mod ¼ modal; mov ¼ moving; n ¼ neuter; neg ¼ negation; nmz ¼ action/state nominalization; nstd ¼ nonstanding; ntr ¼ neutral; obl ¼ oblique;obv ¼ obviative marking; pst ¼ past; pl ¼ plural; prc ¼ process; pro ¼ freepronoun; reas ¼ reason; rel ¼ relativizer; res ¼ resultative; r/r ¼ reflexive/reciprocal; sg ¼ singular; trc ¼ truncated element; vbz ¼ verbalizer.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:43) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 259–286 1351 Authier_08_haude (p. 259)

The direct and inverse markers, which are employed according to the

relative position of the participants on a referential hierarchy (Haude 2009,

2010), belong to a paradigm of verbal morphemes (reflexive/reciprocal,

resultative, agentive, middle) that indicate the transitivity of the verb and

the participant (macro)role – actor or undergoer – of its subject. They can

only be applied productively to one class of verbal bases, which in their

majority denote two-participant events. When a verb of this class occurs

without an overt voice marker, it is syntactically intransitive and denotes a

state, which means that its subject has the undergoer (theme) role. Further-

more, resultative verbs, which also take an undergoer (patient) as subject,

can be identified as being morphologically the least marked of all voice-

marked verbs. Unmarked verbs of this class can therefore be considered

undergoer oriented. The proposal of this paper is to consider the direct

voice marker as a morpheme that derives a transitive verb by simply adding

a syntactic position for an actor argument, leaving the undergoer orienta-

tion of the verb untouched and thus creating an ergative structure. The

inverse marker, under this view, is a secondary derivation, reversing the

participant roles of the arguments of a transitive clause according to the

referential hierarchy and discourse status.

The paper is structured as follows. Section 2 presents the syntactic

properties of transitive and intransitive predicates and their arguments:

Section 2.1 illustrates the distinction between transitive and intransitive

predicates, the encoding of the arguments of transitive clauses and the

function of direct and inverse marking; Section 2.2 demonstrates that

the argument that represents the undergoer in a direct transitive clause

has the same formal and behavioural properties as the S argument of the

intransitive clause, leading to an ergative pattern, whose status is further

discussed in 2.3. Section 3 describes the verbal voice morphemes. Section

4 gives an overview of the verbal bases that can (4.1) and cannot (4.2)

participate in the voice system, showing that the members of the first class

typically denote two-participant events and the members of the second

class one-participant events. Section 5 argues that semantically bivalent

verbs are undergoer-oriented and that this may contribute to the default

ergativity of transitive clauses. The conclusions are drawn in Section 6.

2. Clause structure and ergativity

2.1. Transitive and intransitive clauses

There are two classes of predicates in Movima: transitive and intransitive.

They can be identified by the ability to be combined with two overt argu-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


260 Katharina Haude

ment expressions. A transitive predicate can take two overt argument expres-

sions, as in (1).

(1) tikoy-na¼us os rulrul

kill-dr¼3m.ab art.n.pst jaguar

‘He killed the jaguar.’

An intransitive clause may only contain one argument expression; any

other event participant can only be expressed as an adjunct, marked by

the oblique prefix. Example (2) shows that the verb kaykay ‘eat’, despite

its meaning, is intransitive: the eaten object (‘those nuts’) can only be

expressed as an adjunct.

(2) kayPkay is karak ni-kis ney choPchoK-kwa

mdPeat art.pl macaw obl-art.pl.ab def redPnut-abs

‘The macaws eat those nuts.’

The two core arguments in a transitive clause, identified by the fact that

they do not carry the oblique marker, are distinguished from each other

by their constituency properties: one is expressed by a constituent internal

to the predicate phrase, the other by a constituent external to it, as illus-

trated in (3). These properties are reflected by the following features: the

internal nominal constituent (¼us in (1)) is inseparably attached to the

predicate by ‘‘internal cliticization’’ (causing stress shift; marked as ¼)

and is obligatorily expressed; the external nominal constituent (os rulrul

in (1)) is attached through ‘‘external cliticization’’ (when bound pronoun;

leading to resyllabification but no stress shift; marked as –) or not phono-

logically attached at all (when free pronoun or NP), other elements can

occur between it and the predicate phrase, it is not obligatorily expressed,

and it is easily replaced by a free pronoun in clause-initial position (for

more details see Haude 2006, 2010).

(3) [PRED¼ARG] [ARG]

Whether an argument is represented by the internal or the external con-

stituent depends primarily on its referential properties (see Haude 2009,

2010). The expression of speech-acts participants (except second person

plural; see Haude forthcoming) is restricted to the internal position.

When two third persons interact, the more topical one is represented by

the internal constituent; typically, the internal constituent is a pronoun

and the external one a noun phrase, as in the above examples.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Undergoer orientation in Movima 261

Constituency alone does thus not indicate the participant roles (actor or

undergoer) of the arguments.2 These are indicated by verbal morphemes.

When the internal constituent is the actor and the external constituent the

undergoer, the verb is marked as direct; when the situation is reversed, the

verb is marked as inverse, as shown in (4).

(4) tinok-poj-kay-a¼us os merek rulrul

scare-caus-inv-ep¼3m.ab art.n.pst big jaguar

‘The big jaguar scared him.’

The fact that the arguments are primarily encoded according to their

referential properties and not according to participant roles, makes it di‰-

cult to provide them with uncontroversial labels. Following Bickel (2010),

I use terms that are based on the referential properties of the arguments,

labelling the argument internal to the predicate phrase (high-ranking)

as ‘‘proximate’’ and to the argument external to the predicate phrase

as ‘‘obviative’’ (short PROX and OBV, respectively). As we will see in

the following section, the OBV argument has the syntactically privileged

status.

2.2. The syntactic subject

The single argument of an intransitive clause (S) has the same morphosyn-

tactic properties as OBV of a transitive clause: it is phonologically inde-

pendent, cliticized only when bound pronoun, not obligatorily expressed,

and can easily replaced by a free pronoun. Moreover, on the syntactic

level, OBV and S have access to syntactic operations to which PROX

does not have access.

The clearest case is relativization. A relative clause is introduced by the

particle di’ following the relativized noun phrase, which is not expressed

again in the relative clause. Example (5) shows an intransitive relative

clause.

2. The terms ‘‘actor’’ and ‘‘undergoer’’ (the ‘‘macroroles’’ of Role and ReferenceGrammer; Foley and Van Valin 1984; Van Valin and LaPolla 1997; VanValin 2005) instead of ‘‘agent’’ and ‘‘patient’’ are chosen for two reasons:firstly because a Movima clause can maximally contain two core arguments,which can represent a large range of semantic roles, including recipients; andsecondly because the undergoer role encompasses the semantic roles patientand theme (Van Valin and LaPolla 1997: 140–141; Van Valin 2005: 52),which is relevant for the point made in the present paper, i.e., the similarorientation of stative and resultative verbs.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


262 Katharina Haude

(5) kinos ney ay’ku [di’ jayna kayni ]

art.f.ab here aunt rel dsc die

‘That aunt of mine who has died already.’

Examples (6) and (7) illustrate the relativization of OBV. In (6), the

relativized argument is the undergoer, therefore the predicate of the rela-

tive clause is marked as direct; in (7), the relativized argument is the actor,

therefore the predicate is marked as inverse.

(6) kinos alwaj-a¼us [di’ naye-Ke-na¼us]

art.f.ab spouse-ep¼3m.ab rel marry-co-dr¼3m.ab

‘his wife, whom he had married’

(7) is pa:ko [di’ lap-kay-a¼as]

art.pl dog rel bite-inv-ep¼3n.ab

‘(the) dogs who bit it [the jaguar]’

In order to relativize the participant represented as PROX, a detransi-

tivizing voice operation has to be applied. Here, the predicate is preceded

by a particle kwey (or kaw), the clause is intransitive with the original

PROX as its single argument; the original OBV is (optionally) expressed

as an adjunct, marked by the oblique prefix. This process is generally

found with direct predicates, where it has an antipassive e¤ect. It is illus-

trated in (8)b, which contrasts with (8)a, the transitive construction.

(8) a. jayna rey way-na¼ is kinos rey sonsa

dsc mod lift-dr¼3pl.ab art.f.ab mod silly

‘Then they had already taken up that silly (woman).’

b. is juyeni [di’ jayna kwey way-na n-i’ne]

art.pl person rel dsc detr lift-dr obl-pro.3f

‘the people who had taken her up’

While the syntactically privileged status of OBV as opposed to PROX

in relative clauses is clear, there is no evidence of any construction to

which PROX, but not OBV, has access (Haude 2009). Therefore, in the

remainder of this paper and in line with other contributions in this volume

(Monod-Becquelin and Becquey; Queixalos), I use the term ‘‘subject’’ to

refer to the grammatical relation encompassing {OBV,S}. While based

here entirely on syntactic grounds, the term seems suited best for making

the patterns in Movima comparable to voice patterns in other languages,

since discussions of voice phenomena generally make use of the subject

notion (see e.g. Mithun 1994; Kulikov 2010).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



2.3. Ergativity

Given that OBV aligns with S, it is obvious that when the participant roles

of the arguments are considered, the direct/inverse alternation results in

two basic transitive clause types, one with an ergative and the other with

an accusative pattern. In the direct construction, OBV represents the

undergoer, and in the inverse construction, OBV represents the actor.

With respect to formal marking, the split is entirely parallel: whenever

PROX is the actor and OBV the undergoer, the pattern is ergative; when-

ever OBV is the actor and PROX the undergoer, the pattern is accusative.

In both cases, the predicate is overtly morphologically marked. However,

with respect to discourse frequency and pragmatics, there is evidence that

direct/ergative construction is the default for transitive clauses (see also

Haude 2010).

First of all, the direct construction is statistically more frequent, occurring

in about 80% of the transitive clauses with two third-person arguments.

This is not surprising, since a high-ranking actor and a low-ranking under-

goer represent the typical participant constellation in a two-participant

event (cf. DeLancey 1981; Givon 1994; Croft 2003).

Evidence for the pragmatically unmarked character of the direct con-

struction is also apparent from the fact that this construction is occasion-

ally found in opposition to the referential hierarchy, while this is never the

case with the inverse construction. When two third-person arguments are

represented by pronouns or both by full noun phrases, reflecting an equal

discourse status, then automatically the direct construction is used. This is

the case in elicitation, and it is illustrated by the text example in (9), where

both arguments are encoded as noun phrases: even though the actor is

an animal and the undergoer a human, the construction is direct and the

actor is represented as PROX. The text corpus contains no example of the

reversed case, i.e. the inverse construction with the referentially lower-

ranking participant represented as PROX. In elicitation, the inverse con-

struction in such a context is accepted, but never proposed spontaneously.

(9) jayna lap-na¼as mimi:di us majni!

dsc bite-dr¼art.n snake art.m my_o¤spring

‘Now the/a snake bit my son!’

The inverse construction, furthermore, has only limited access to the

detransitivising voice operation with kwey. A kwey-construction with an

inverse predicate occurs nowhere in the text corpus. In elicitation, certain

examples with a kwey-construction and an inverse predicate, like the one

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


264 Katharina Haude

in (10)b, tend to be rejected; the exact conditions of this still need to be

investigated, but in any case, the limited acceptability of such examples

means that the use of an inverse predicate in the kwey-constructon is

problematic.

(10) a. inKa kwey dul-na n-isne

pro.1sg detr visit-dr obl-pro.3f.ab

‘It was me who visited her.’

b.? inKa kwey dul-kay n-isne

pro.1sg detr visit-dr obl-pro.3f.ab

(‘It was me who was visited by her.’)

I conclude that while there is no di¤erence in morphological marked-

ness between the inverse and the direct construction, the direct construc-

tion is the default transitive construction, used when two third-person par-

ticipants are ranked equally in terms of discourse status. The inverse

construction, in contrast, is restricted to the situation in which the under-

goer outranks the actor with regard to person, animacy and discourse pro-

minence. Furthermore, the detransitivising operation with kwey is largely

restricted to the direct construction. The default transitive a‰rmative

main clause in Movima, therefore, has an ergative pattern.

3. Voice markers

The direct and inverse markers belong to a paradigm of verbal morphemes

that indicate the transitivity of a verb and the participant role(s) of the

core argument(s). They are listed in Table 1.

Table 1. Movima voice markers

transitivity marker meaning subject role

transitive -na/-a- direct Ug

-kay inverse Ac

intransitive -cheK reflexive/reciprocal ActþUg

-’i resultative Ug

-eKe agentive Act

<REDP> middle ActþUg

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



A first illustration of the e¤ect of the voice markers is provided in (11)

with the transitive root jat- ‘hit’. (The reduplicative middle marker is not

illustrated here because it is not fully productive and does not occur with

this root.)

(11) direct: jat-na¼Ø ‘I hit X’

inverse: jat-kay¼Ø ‘X hits me’

reflexive: jat-cheK ‘X hit(s) themselves/each other’

resultative: jat-’i ‘X has been hit.’

agentive: jat-eKe ‘X hits (continuously/habitually)’

These morphemes can be adequately analysed as voice markers because

of their property to indicate the transitivity of a verb and the participant

role of the verb’s subject, which are characteristics of voice marking as

defined cross-linguistically. They also serve to present an event from vary-

ing perspectives (see Shibatani 2006). They do not seem, however, to be

employed for syntactic purposes (this is the role of the particle kwey, see

2.2 above), and none of them derives an intransitive verb from a transitive

one, a feature present in many traditional definitions of voice (see Kulikov

2010). They have aspectual (Aktionsart) and sometimes modal connota-

tions. In the following discussion, however, I will focus on their voice-

marking property, i.e. the way in which they assign arguments to the verb.

3.1. Transitive voice markers: direct and inverse

As was shown in 2.1 above, transitive predicates contain either the direct

or the inverse morpheme. The direct morpheme has two allomorphs: base-

final -na and base-internal -a-, which are both illustrated in (12). The base-

internal allomorph -a- is applied like an infix: it is inserted in morphologi-

cally complex verbal bases immediately after the root, as in (12)b, pro-

vided that the root be a single closed syllable not followed by an aspectual

su‰x (e.g. -ka ‘mlt’). Being based on prosodic properties of the base, the

insertion of the direct marker -a- can create non-linear morphological

patterns. For example, in (12)b the direct marker precedes the causative

su‰x (-poj ) in linear order, but it is applied at a later stage of the verbal

derivational process, after the causative su‰x has derived the bivalent base.

(12) a. lat-na¼ is kis ko’o,

chop-dr¼3pl.ab art.pl.ab tree

b. ´ok-a-poj-a¼ is ba:ra kis ko’o

fall-dr-caus-ep¼3pl.ab all art.pl.ab tree

‘They chop down the trees, they fell all the trees.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


266 Katharina Haude

In all other environments, the su‰x -na is applied, as illustrated in (12)a

with a simple root and in (13) with a complex base with an open-syllable

root.

(13) jayna chi-poj-na¼ is us majni

dsc go_out-caus-dr¼3pl.ab art.m my_o¤spring

‘They took my child out (of my body).’

Some bivalent bases seem to be historically complex, since they take

the direct marker -a- instead of -na, but synchronically, their components

cannot be properly identified. An example is given in (14), with the verb

base kayKe ‘give’, whose final syllable is most probably the applicative

su‰x -Ke, but whose first syllable kay (homophonous with the root kay-

‘eat’) cannot be identified as a root on the synchronic level.

(14) kay<a:>Ke¼Ø n-i’ko jayna

give<dr>¼1sg obl-pro.3pl dsc

‘I’ll give it (the money) to you.’

The inverse marker invariably consists of the su‰x -kay (presumably

unrelated to the verb root ‘eat’ or ‘give’) independently of whether the

verb takes the su‰x -na ((15), cf. (13)) or its allomorph -a- ((16), cf. (14))

as direct marker. Since its final phoneme /j/ (y) is a consonant, the epen-

thetic vowel -a is added before an internal enclitic (as in (4) and (7) above;

see Haude 2006: 98–99). As was shown in 2.1, the inverse su‰x derives a

transitive predicate whose subject is the actor.

(15) chi-poj-kay¼Ø isnos kayni di’ nonok

go_out-caus-inv¼1sg art.f.pst die rel my_grandmother

‘My late grandmother took me out (of school).’

(16) kayKe:-kay¼Ø–isne n-is narasa-mes-a¼sne

give-inv¼1sg–3f.ab obl-art.pl orange-cl.fat-ep¼3f.ab

‘She gave me her orange oil.’

In the third-person domain, the application of the direct and inverse

markers depends very much on the relative topicality of the arguments,

since the topical argument is generally represented as PROX. Here, the

inverse assumes the function of a voice marker (cf. Gildea 1994), as is

reflected by the fact that it can generally be translated by a passive, as

in (17).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(17) us bi:jaw, jiw-a-Ke-kay-a¼us

art.m old come-dr-co-inv-ep¼3m.ab

us kayni di’ Koyimbra

art.m die rel proper_name

‘The old (man), he was brought by the late Coimbra.’

3.2. Intransitive voice markers

The intransitive voice markers create predicates that can only take one

overt core argument expression.

3.2.1. -cheK reflexive/reciprocal

The reflexive/reciprocal su‰x -cheK is the most productive intransitive

voice marker, combinable with any verb that can be marked as direct or

inverse. It creates a verb whose subject represents one or more partici-

pant(s) that simultaneously carry out and undergo an action. The interpre-

tation as either reflexive (18) or reciprocal (19) depends on the context.

(18) didi’ tikoy-cheK n-os kachi:ra

frust killed-r/r obl-art.n.pst knife

‘(He) wanted to kill himself with a knife.’

(19) ban jayna don-cheK–is

but dsc dislike-r/r–3pl.ab

‘. . . but they already disliked each other.’ [HRR tx 191]

In 4.2 below it will be shown that when attached to so-called monova-

lent bases, this su‰x is deprived of its reflexive/reciprocal meaning and

only marks an activity.

3.2.2. -’i resultative

The su‰x -’i marks an intransitive verb denoting a resultative state, i.e.

a verb expressing ‘‘both a state and the preceding action it has resulted

from’’ (Nedjalkov and Jaxontov 1988: 6). The subject of the resultative

verb represents the undergoer.

(20) ben-’i is chorimpa¼sne

paint-res art.pl fingernail¼3f.ab

‘Her fingernails were painted.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


268 Katharina Haude

(21) bo rey os bet’i vus-’i–as

reas mod art.n.pst grassland carbonize-res–3n.ab

che kaw-ra is ve’e:-vus

and much-be.ntr art.pl fire-cl.dust

‘. . . because the grassland, it had been burnt [by the farmers] and

there were lots of ashes.’

Depending on the context, the su‰x -’i can also express a deontic

modality (‘X has to be done’); it is used in instructions and procedural

texts:

(22) che jayna n-as ba:-paj-uk-wa¼n jayna

and dsc obl-art.n finish-split-nmz¼2 dsc

il-’i–is, tamol-’i–is daya’a

spread-res–3pl.ab expand-res–3pl.ab dur.nstd

‘And when you finish splitting (the straws), then they are/have to be

spread out, they are/have to be extended.’

(23) raK-’i is pe:ra, il-’i, jayaw-’i,pull_out-res art.pl reed spread-res nice-res

werel-ni che pil-’i no-kos do:nojbet ney

dry-prc and rolled_up-res obl-art.n.ab cloth here

‘The reed is pulled out, (it) is spread, (it) is made nice, (it) dries and

is rolled up with a cloth like this. . .’

In traditional voice terminology, the resultative comes closest to a

passive (cf. Comrie 1981), since it creates an intransitive predicate whose

subject is the undergoer. However, in di¤erence to a canonical passive

(cf. e.g. Siewierska 1984; Shibatani 1985; Dixon and Aikhenvald 2000;

Kazenin 2001; Keenan and Dryer 2007), there is no sign of derivation

from an unmarked transitive, active verb: the base to which the resultative

su‰x is attached is stative rather than active (see Section 5 below), and

transitive predicates are overtly morphologically derived. In fact, instead

of being a derived form, the resultative even seems to be the morphologi-

cally least-marked form of a bivalent verb root (see Section 5 below).

The agent is not expressed in a clause with a resultative verb; the few

cases in the corpus where an oblique phrase in a resultative clause might

be interpretable as an agent are cases like (24), where an oblique phrase

(nis alamre) denotes the entity by which the state was caused. In general,

also in resultative clauses obliques encode peripheral roles such as loca-

tions, times or instruments, as in (23) above (nokos do:nojbet) or (25).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(24) biK-’i is dinoj-a¼ ’ne n-is alamre

scratch-res art.pl thigh-ep¼3f obl-art.pl wire

‘Her legs were scratched from the wire.’

(25) pay’-’i n-is bubutkwa os ro:ya

smear-res obl-art.pl mud art.n.pst house

‘The house was plastered with mud.’

3.2.3. -eKe agentive

The agentive marker -eKe derives an intransitive verb whose subject is the

actor, as illustrated in (26). In this way, the agentive su‰x resembles an

antipassive, and like many antipassives (see Cooreman 1994), it indicates

a durative or habitual action. However, in the same way as the resultative

marker, the agentive marker does not derive an intransitive from a transi-

tive verb, which belongs to the definition of a typical antipassive (see Dixon

1994: 146).

(26) jayna jo’yaj, sal-e:Ke–y’Ki

dsc arrive search_for-agt–1pl

‘Then (we) arrived, we searched.’

Verbs containing the su‰x -eKe are often combined with an oblique-

marked noun phrase denoting the patient, as in (27) and (28). However,

oblique phrases can encode many relations, which can only be inferred

from the context. For example, in (27) and (28), the oblique NP is identi-

fied by the context as a patient, while in (29), the context identifies it as a

location.

(27) jayna jot-e:Ke n-is chekwesKa

dsc gather-agt obl-art.pl taruma

‘Then (I) gathered taruma fruits.’ [JAO tx 026]

(28) raK-pit-e:Ke n-is kade:na

tear-be.half-agt obl-art.pl chain

‘[The wild cat] tore the chain [and escaped].’

(29) sal-e:Ke–us n-is wolsiko¼us

search_for-agt–3m.ab obl-art.pl pocket¼3m.ab

‘He searched in his pockets [for bullets].’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


270 Katharina Haude

3.2.4. Middle reduplication

The middle marker consists of a full reduplication of monosyllabic and a

partial reduplication of disyllabic roots (see Haude 2006: 345¤.). I call it

‘‘middle’’ because reduplicated verbs generally denote events that a¤ect

the subject participant in some way. Consider the di¤erence between the

direct-marked verb in (30) with the middle verb in (31):

(30) iK Kap-na¼Ø is ona:cho

1sg bathe-dr¼1sg art.pl my_grandchild

‘I bathe my grandchildren.’

(31) jayna KapPKap–i’ne

dsc mdPbathe–3f

‘Then she bathed.’

The same di¤erence is illustrated with a longer verb root, chumay-,

whose middle form is created by reduplication of the penultimate syllable.

Example (32) illustrates the direct, (33) the middle form of the verb.

(32) chumay-na¼n kos boPboj-a¼kos ko’

smoke-dr¼2 art.n.ab redPbase-ep¼art.n.ab tree

‘You smoke out the trunk of the tree [for collecting honey].’

(33) kos da’ ve’ chu<ma:P>may

art.n.ab dur.nstd fire smoke<mdP>

‘The fire is smoking.’

The middle marker is not very productive on the so-called ‘‘bivalent’’

bases (4.1); in contrast, it frequently occurs with ‘‘monovalent’’ bases,

from which it derives activity verbs (see 4.2).

4. Types of verb bases

Two types of verbal bases can be distinguished in Movima: bases that

participate fully in the voice system and bases that do not. The defining

criterion for their distinction is a formal one: on verbs of the first class,

the su‰xation of -na derives a transitive predicate in the way described in

2.1, whereas on verbs of the second class, the su‰xation of -na derives a

locational noun. The two classes are furthermore characterized semanti-

cally: bases of the first class denote two-participant events, while most

bases of the second class denote one-participant events; bases of the first

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



class, when unmarked for voice, are stative predicates, while unmarked

bases of the second class typically denote activities. The two classes are

described in the following sections.

4.1. Bivalent bases

The bases that participate fully in the voice system typically denote events

that imply at least two participants, including prototypical transitive events

(Hopper and Thompson 1980) like ‘hit’ or ‘kill’. Therefore, I refer to the

bases of this class as ‘‘bivalent’’.3 Bivalent bases can be either simple (i.e.

roots) or complex; most complex bases can occur independently, while

roots cannot.

Examples of bivalent verb roots, of which there are about 150 (see

Haude 2006: 555¤.), are listed in (34).

(34) jat- ‘hit’

Kek- ‘kick’

lap- ‘bite’

sal- ‘search for’

yey- ‘want’

ju:- ‘scold’

tikoy- ‘kill’

ela- ‘leave behind’

ji:sa- ‘make’

Verb roots cannot occur independently; they must be combined with

another morpheme, e.g. a voice marker, as illustrated in (34) above. In

Section 5 below I will argue that the resultative voice, marked by -’i, can

be considered the least marked form of bivalent roots.

Bivalent bases can also be morphologically complex, i.e. can consist of

a root plus a derivational morpheme (like causative or applicative) or a

modifying incorporated element (see Haude 2006: Ch. 9). Consider the

base jam-Ke in (35), which combines with most voice markers:

3. Adjectives (words like ra:pal ‘red’, merek ‘big’ and ja:yaw ‘nice’) are excludedfrom this discussion. They can participate in the voice system (e.g. jayaw-na ¼Ø ‘I make X nice’ or jayaw-’i ‘be made nice’ in (23)), but can be distinguishedfrom the verbal bases in that they can be combined with the verbalizing su‰xes-ni ‘to be/become X’ and -tik ‘to make/do X’, and that they can be reduplicatedin subordination (see Haude 2006: 119). There are borderline cases where adjec-tives and stative verbs cannot be easily distinguished, and more fine-grainedinvestigation still needs to be undertaken. However, this is not directly relevantfor the present study.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


272 Katharina Haude

(35) a. jam-a:-Ke¼Ø

tie-dr-co¼1sg

‘I tie X onto (sth.).’

b. jam-Ke:-kay¼Ø

tie-co-inv¼1sg

‘X tie(s) me (onto sth.).’

c. jam-Ke:-cheK

tie-co-r/r

‘X tie(s) itself/each other onto (sth.).’

d. jam-Ke-’i

tie-co-res

‘X has been tied onto (sth.)’

e. jam-<Ke:P>Ke

tie-<mdP>co

‘X gets tied onto (sth.).’

Unlike roots, most complex bivalent bases can occur independently, in

which case they denote a state. Example (36), illustrating the base jamKe

from (35) above, is from elicitation; (37) illustrates the occurrence of an

unmarked bivalent bases in texts (see Section 5 below for more examples):

(36) jam-Ke–i n-is waPwa-n-kwa

tie-co–3n obl-art.pl redPliana-ln-abs

‘They are tied onto (sth.) with a liana.’

(37) am-poj is Santo rey jayna

enter-caus art.pl Saint mod dsc

‘The Saints had already been put in again.’

4.2. Monovalent bases

The so-called monovalent bases, in contrast, cannot be combined with the

full range of voice markers. This group, though semantically less con-

sistent than the one described above, includes elements referring to one-

participant events like ‘sleep’ or ‘stand’, which is why I refer to them as

monovalent. Examples of monovalent roots are listed in (38) (see Haude

2006: 340¤.).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(38) joy- ‘go’

en- ‘stand’

as- ‘sit’

de:- ‘lie’

josi- ‘laugh’

chi- ‘go out’

jiwa- ‘come’

kay- ‘eat’

dejal- ‘cook’

Kokot- ‘boil’

The formal diagnostic for identifying a monovalent base is the e¤ect of the

su‰x -na. With bivalent bases, the addition of this su‰x (or its allomorph

-a- on certain complex bases) creates a direct transitive predicate, while

on monovalent bases, the su‰xation of -na creates a locational noun; the

di¤erence is illustrated in (39) (bivalent) and (40) (monovalent).

(39) jat-na¼Ø

hit-dr¼1sg

‘I hit (you/him/her/it/them)

(40) kay-na¼Ø

eat-loc¼1sg

‘the place where I eat’

The derived locational nouns can be identified as nouns because, when

functioning as predicates, their subject can only be expressed by a clause-

initial free pronoun and not by an externally cliticized pronoun, as shown

in (41)a and (41)b, respectively. They share this syntactic limitation with

possessed nominal predicates (see Haude 2010).

(41) a. bo a’ko as-na¼Ø

reas pro sit-loc¼1sg

‘because it is my house’

b. *bo as-na¼Ø–areas sit-loc¼1sg–3n

(‘because it is my house’) [el]

Just like bivalent roots, monovalent roots cannot occur independently;

to form a main-clause predicate, they must be combined with another

morpheme, the choice of which depends on the root. Most monovalent

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


274 Katharina Haude

roots are combined either with the reflexive/reciprocal su‰x -cheK or with

the middle reduplication, as illustrated in (42) and (43), respectively, for

the roots listed above. (Others take distributionally more restricted, unpro-

ductive su‰xes with possibly specific meanings; see Haude 2006: 342–

344.) Many verbs formed in this way express activities, by which the actor

may or may not be seen as a¤ected. Some middle verbs denote transitive

events whose patient can optionally be expressed by an oblique noun

phrase, as was shown in example (2) for the verb kaykay ‘eat’.

(42) joy-cheK ‘go, leave’

en-cheK ‘stand (up)’

josi:-cheK ‘laugh’

pen-cheK ‘land (plane, bird)’

tes-cheK ‘limp’

de:-cheK ‘lie (down)’

(43) kayPkay ‘eat’

chi:Pchi ‘go out’

KapPKap ‘bathe’

ji<wa:P>wa ‘come’

Ko<ko:P>kot ‘boil’

de<ja:P>jal ‘cook’

A text example of middle verbs based on monovalent roots is given in

(44). While identical with respect to morphological marking, the verb

deja:jal ‘cook’ in (44)a denotes an activity and takes the agent (the cook)

as subject, while Koko:kot ‘boil’ in (44)b denotes a process and takes the

a¤ected entity (the cooked food) as subject.

(44) a. ena’ de< jaL>jal–isne che jayna

dur.std cook<mdP>–3f.ab and dsc

b. da’ Ko<ko:L>kot is dej-na¼sne4

dur.nstd boil<mdP> art.pl cook-dr¼3f.ab

‘She was cooking, and what she cooked was already boiling.’

Complex monovalent bases can be derived through argument incorpo-

ration (see Haude 2006: 283–286). Although a verb with an incorporated

argument (usually in the form of a bound lexical element or classifier)

obligatorily contains the direct marker, it is syntactically intransitive; this

4. While probably historically related, the monovalent root dejal- has to be consid-ered as distinct from the bivalent root dej-, since the ending al is unanalyzable.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



can be seen from the oblique marking of the NP in (45), which is co-

referential with the incorporated argument. Furthermore, the addition of

the su‰x -na creates a locational noun, as shown in (46).

(45) dan-a:-so–is n-is pokso

chew-dr-cl.chicha–pl.ab obl-art.pl chicha

‘They chewed (the) chicha.’

(46) asko yok-a-mo-na¼ is n-eys

pro.n.ab catch-dr-cl.bird-loc¼3pl.ab obl-dem.spk.pl

karak di’ sere:re

macaw rel wild

‘That (was) where they caught those wild macaws.’

In addition, there are monovalent verbs that do not show any sign

of (synchronic) complexity and that cannot be combined with a voice

marker; examples are given in (47).

(47) ja:yi ‘run’

ja:raK ‘fight’

jo’yaj ‘arrive’

te:lo ‘dance’

salmo ‘return’

yolmoK ‘go for a walk’

ya:lo:we ‘drink’

Also on these verbs, the addition of -na creates a locational noun, as illus-

trated in (48):

(48) che asko jayi-na¼us

and pro.n.ab run-loc¼3m.ab

‘And that (was) where he ran to.’

While the aspectual properties of monovalent bases require further

study, the data so far indicate that these bases denote activities (cf. Van

Valin 2005: 55) rather than states, and in this respect di¤er significantly

from unmarked bivalent bases.

5. The undergoer orientation of bivalent bases

5.1. Orientation

The term ‘‘orientation’’ is sometimes used to refer to the participant role a

verb assigns to its central argument. In the transitive domain, the accusa-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


276 Katharina Haude

tive pattern implies actor orientation and the ergative pattern undergoer

orientation. In the intransitive domain, so-called unaccusative verbs are

typically undergoer-oriented and unergative verbs actor-oriented.

The preceding sections have shown that in Movima, there is a split

in the domain of morphologically unmarked verbal bases: the so-called

bivalent bases, forming stative predicates when unmarked, are undergoer

oriented; the majority of monovalent bases, in contrast, denote activities

and hence can be considered actor oriented.5 A major function of the

voice markers is to overtly indicate and/or modify the orientation of

the voice-marked verb (see Serzisko 1991): in Movima, a verb marked as

resultative (intransitive) or direct (transitive) is oriented towards the

undergoer, a verb marked as agentive (intransitive) or inverse (transitive)

is oriented towards the actor. Reflexive/reciprocal and middle-marked

verbs, whose actor is a¤ected by the action it instigates, have a tendency to

be actor oriented; this is apparent from the e¤ect of middle and reflexive/

reciprocal marking of monovalent bases, which denote activities (see (42),

(43) above). Thus, since the bivalent bases, which participate fully in the

voice system, are undergoer oriented and the default transitive clause

patterns ergatively, the hypothesis is that the function of the ergative

marker is to permit the additional expression of the actor argument.

This hypothesis is strengthened by the fact that the resultative marker

-’i, which overtly marks the verb as undergoer oriented, shows signs of

being closer to a zero marker than the other voice morphemes (and was

analysed as such in Haude 2006). The correlation between resultative

marking and absence of marking is both semantic and phonological.

5.2. Semantic parallels between unmarked and resultative verbs

When in elicitation, speakers are faced with a minimal pair of an unmarked

bivalent base and a resultative form marked with -’i, their explanations

make it clear that the unmarked form is interpreted as denoting a state

‘‘without any implication of its origin’’ (Nedjalkov and Jaxontov 1988:

6), while in the resultative form, a (possibly unknown) actor is implied.

5. It may be tempting to use the terms ‘‘unaccusative’’ and ‘‘unergative’’ for the‘‘bivalent’’ and ‘‘monovalent’’ verbal bases of Movima, since these terms aresyntactically based and include lexical idiosynchrasies (e.g. Movima Kokot-‘boil’, which falls in one class of predominantly activity verbs like dejal-‘cook’). However, they are not very felicitous here because they rely heavilyon the notions of subject (agent) and object (patient) (see Perlmutter 1978)and on syntactic tests.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Some verbs that were tested this way are listed in (49) (the parentheses in

the glosses provide the interpretation of the verb with -’i; bases whose

stress or lengthening pattern changes when occurring with the su‰x are

presented independently).

(49) katpit(-’i) ‘be broken in halves (by someone)’

Kok-poj(-’i) ‘be felled (by someone)’ (-poj caus)

way’choK(-’i) ‘be mended (by someone)’

ja:rat ( jarat-’i) ‘be thrown away (by someone)’

do’waj (dowaj-’i) ‘be moved to another place (by someone)’

ji:sa ( ji:sa:-’i) ‘to be made (by someone)’

However, the distinction between simple and resultative states is not so

clear-cut in natural discourse, where the forms seem to be used inter-

changeably.

The following examples may illustrate this. As was shown in 3.2.2, the

resultative occurs in descriptions and instructions, where sequences of

events are described. However, here also the unmarked forms are found,

like the verb nanra in (50) (a description of raising cattle) and Kokpoj and

dakato:lej in (51) (a description of how a slash-and-burn field is made):

(50) n-as jayna tawakeni-wa¼ i jayna

obl-art.n dsc wake_up-nmz¼3pl dsc

nanra–i reyka, nanraset_free–3pl.ab mod set_free

bo as joy-wa¼ i di:ra

reas art.n go-nmz¼3pl at_least

n-as kay-wa¼ i n-as chapmo

obl-art.n eat-nmz¼3pl obl-art.n bush

‘When they [the cows] wake up, they are set free again, (they) are set

free so that they go at least to feed in the bush.’

(51) jayna Kok-poj kis ko’

dsc fall-caus art.pl.ab trees

che das-ka-to:lej jayna

and cut-mlt-branch dsc

bo as de:-wa¼kis tolej-a¼kis ko’o

reas art.n lie-nmz¼art.pl.ab branch-ep¼3pl.ab tree

‘Then the trees are felled and their branches cut o¤ (lit. ‘. . . [they]

are repeatedly branch-cut), so that the branches of the trees lie flat.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


278 Katharina Haude

(52) kiro’ kis lotoba¼ is di’ pokso,

dem.pl.ab art.pl.ab jug¼3pl.ab rel chicha

joy-Ke no-kos kavildo

go-co obl-art.n.ab Cabildo

‘There were their jugs of chicha, (they had been) taken to the

Cabildo.’

On the other hand, verb roots, which cannot occur without additional

phonological material, take the resultative marker -’i when denoting states

that do not imply an actor, like yey’i in (53) or Kek’i in (54):

(53) rim<a:>Ke¼Ø n-is wa:ka-wandi buka’

sell<dr>¼1sg obl-art.pl cow-instr:be.house dur.mov

bo rey yey-’i–is

reas mod want-res–3pl.ab

‘I sold (the hats) on the ranches because they were appreciated.’

(54) yey-na¼ ’ne os ma:kina di’ Kek-’iwant-dr¼3f art.n.pst machine rel kick-res

‘She wanted a (sewing) machine that is foot-driven.’

The equivalence of unmarked and resultative forms is also apparent

when unmarked stative verbs verbs and verbs ending in -’i cooccur in the

same clause, as in (55) (a joke), where they denote a sequence of events:

(55) dak-Ke che jarat-’i, jayaw-Ke¼a

cut-co and dump-res good-neg¼3n

‘[Your head] needs to be cut o¤ and thrown away, it’s useless!’

Furthermore, there are verbs where, without any obvious morpho-

phonological reason, the attachment of -’i is not possible, while others

cannot occur in the unmarked form. In both cases, the resultative and the

stative reading are indistinguishable. For instance, the verb rimKe ‘sell’

in (56) cannot occur with the resultative marker, whereas the verb base

rimeK- ‘buy’ in (57) cannot occur without it when denoting a state:

(56) a. ba:ra rimKe

all sell

‘It is all sold.’/ ‘It has all been sold.’

b. *ba:ra rimKe-’iall sell-res

(‘It has all been sold.’) [el]

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(57) a. ba:ra rimeK-’iall buy-res

[ba:.ra riomeK?i]

‘It is all bought.’/ ‘It has all been bought.’ [el]

b. *ba:ra ri:meK

all buy

(‘It is all bought.’) [el]

Another verb that, like rimKe, cannot occur with the ending -’i to

denote a resultative state, is jommi ‘eat up’ in (58):

(58) ban kiwa il-na¼Ø bo nokowa jommibut dem spread-dr¼1sg reas right_now eat_up

‘But I have spread it there because now it will be eaten.’

Resultative verbs that, like rimeK, cannot occur without the ending -’i

(besides the monosyllabic roots) are listed in (59). For the first three verbs

in this list, the most probable explanation is that they constitute verb roots

and therefore cannot occur on their own; for the last two, no such expla-

nation can be given, since they are clearly composed of a root and the

causative su‰x.

(59) tikoy-’i ‘be killed’

e:la:-’i ‘be left behind’

dewaj-’i ‘to be seen’

chi-poj-’i ‘to have been taken out’ (‘go out’þcausþres)

ju:-poj-’i ‘to have been punished’ (‘punish’þcausþres)

5.3. Phonological parallels between -’i and zero

The hypothesis that of all voice markers, the resultative marker -’i comes

close to zero marking receives support from the fact that the element [?i] is

homophonous with a dummy element that occurs on some prosodically

deficient nouns and pronouns. In Movima, a content word (noun, verb,

adjective) must be minimally disyllabic. Apart from very few exceptions

(Haude 2006: 196), monosyllabic noun roots are either augmented by

reduplication or by attachment of the ending [?i], the choice being lexi-

cally determined. Example (60) illustrates the augmentation of the mono-

syllabic noun root nun- with the dummy -’i. The resulting form behaves

like any other noun; for example, it can be marked for alienable and

inalienable possession, as shown in (60)b and (60)c, respectively. Examples

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


280 Katharina Haude

(61) and (62), however, where the noun forms part of a compound and

an incorporating verb, respectively, show that the root of the word is

nun- alone (see also Haude 2006: 207).

(60) a. nun-’i b. nun-’i¼a c. nun-<’iP>’i¼a

bone-d bone-d¼3n bone-<inalP>d¼3n

‘bone’ ‘its bone (al.)’ ‘its bone (inal.)’ [el]

(61) punta:-nuntip-bone

‘the bone from the tip (of the rib cage)’

(62) it dan-a:-nun1intr chew-dr-bone

‘I chew on bones.’ [el]

The element [?i] also occurs in initial position on pronouns that do not

display the required syllable structure. It augments the first-person prono-

minal elements (i)K and (i)t, which consist of a simple consonant and

therefore need to be attached to a preceding vowel. In (63), this vowel

is provided by the preceding word; in (64), the preceding word ends in a

consonant, therefore the element [?i] is inserted as a dummy host.

(63) jayna t joy-cheK

dsc 1intr go-r/r

‘Then I went.’

(64) ban it joy-cheK

but 1intr go-r/r

‘But I went.’

The personal pronouns (i)’ne ‘3f’, (i)sne ‘3f.ab’, (i)y’Ki ‘1pl’, and

(i)y’bi ‘2pl’, which would have a complex onset if occurring independ-

ently, are preceded by an element /i/ when occurring as free forms or

when externally cliticized to a consonant-final host. Compare the variants

of the third-person feminine pronoun (i)sne ‘3f.ab’ when cliticized to a

vowel-final host, (65)a, when occurring independently, as in (65)b, and

when cliticized to a consonant-final host (resyllabifying with that con-

sonant), as in (65)c. The stress and syllable pattern is given in the phonemic

representation.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(65) a. salmo–snereturn–3f.ab

/osal.mos.ne/

‘She returned.’

b. isne salmo

3f.ab return

/o?is.ne osal.mo/

‘She returned.’

c. a:mon–isneenter–3f.ab

/o?a:.mo.%nis.ne/

‘She came in.’

Note that the bound pronoun in (65)c is not preceded by a glottal stop,

the pronoun being resyllabified with the preceding consonant. It may be

argued that here, the element /i/ cannot be compared to the sequence [?i]

in the above examples. However, as was shown in Haude (2006: 101), it is

a property of external cliticization that referential elements are resyllabi-

fied with the preceding consonant, and that the glottal stop, which other-

wise precedes all vowel-initial morphemes, is dropped.

If -’i is a prosodically triggered element in other environments, then it is

very well conceivable that also on verbal bases, it is synchronically per-

ceived as a phonological dummy. This may be the reason why in actual

speech the resultative verbs are used in the same way as unmarked bi-

valent verbs, given that both denote states.

6. Conclusion

Movima has a set of verbal voice markers that indicate the transitivity of a

verb and the participant role(s) (actor, undergoer, or both) of its subject.

Six voice markers can be identified, four of which mark intransitive verbs

(reflexive/reciprocal, resultative, agentive, and middle) and two that mark

transitive verbs (direct and inverse). Verbs unmarked for voice are in-

transitive, while all transitive verbs are overtly morphologically derived

through direct or inverse marking. Movima can therefore be characterized

as a ‘‘fundamentally intransitive language’’ (Nichols 1982; see also Nichols,

Peterson, and Barnes 2004).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


282 Katharina Haude

Only one class of verbal bases, termed ‘‘bivalent’’ because they denote

two-participant events, can be productively combined with the voice

markers. The other class, termed ‘‘monovalent’’ because most of its mem-

bers denote one-participant events, only has limited access to the voice

markers, and their combination with the reflexive/reciprocal or middle

marker is lexically determined. In particular, monovalent bases cannot be

marked as direct; on them, the su‰xation of the element -na (the direct

marker on bivalent bases) derives a locational noun.

Bivalent verbs unmarked for voice denote states. They are in this

respect similar to the resultative forms, marked by the su‰x -’i. The simi-

larity is confirmed by the distributional equivalence of unmarked and

resultative forms, the fact that some can only be occur unmarked and

others only marked by -’i when denoting a state, and by the fact that the

su‰x -’i is homophonous with a dummy element [?i] found in other envi-

ronments. Both unmarked and resultative verbs are undergoer oriented,

i.e. have a non-actor as their subject.

Direct-marked predicates are also characterized by undergoer orienta-

tion. Therefore, although both the direct and the inverse derivation are

overtly morphologically marked, the direct derivation can be seen as the

simpler operation, since it introduces an actor without changing the orien-

tation of the verb. Maybe this fact has contributed historically to the basic

status of the direct (ergative) construction. The inverse marker, in con-

trast, introduces an undergoer and at the same time changes the orienta-

tion of the verb. The inverse voice can, from this perspective, be seen

as operating on the transitive level, enabling a referentially high-ranking

participant to take the undergoer role.

The cross-linguistically unusual ergative bias found in Movima discourse

can thus be explained by the underlying undergoer orientation of bivalent

verbal bases.

References

Bickel, Balthasar2010 ‘‘Grammatical relations typology.’’ In The Oxford Handbook of

Linguistic Typology, Jae-Jung Song (ed.). Oxford: Oxford Uni-versity Press.

Comrie, Bernard1981 ‘‘Aspect and voice: some reflections on perfect and passive.’’ In

Syntax and Semantics 14: Tense and Aspect, Philip Tedeschi andAnnie Zaenen (eds.), 65–78. New York et al.: Academic Press.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Cooreman, Ann1994 ‘‘A functional typology of antipassives.’’ In Voice: Form and

Function, Barbara A. Fox and Paul J. Hopper (eds.), 49–87.Amsterdam/Philadelphia: Benjamins.

Croft, William2003 Typology and Universals. Cambridge: Cambridge University

Press. Second edition.DeLancey, Scott

1981 ‘‘An interpretation of split ergativity and related patterns.’’ Lan-guage 57(3): 626–657.

Dixon, R.M.W.1994 Ergativity. Cambridge: Cambridge University Press.

Dixon, R.M.W. and Alexandra Y. Aikhenvald2000 ‘‘Introduction.’’ In Changing Valency. Case Studies in Transitivity.

R.M.W. Dixon and Alexandra Y. Aikhenvald (eds.), 1–29.Cambridge: Cambridge University Press.

Foley, William A. and Robert D. Van Valin, Jr.1984 Functional Syntax and Universal Grammar. Cambridge: Cam-

bridge University Press.Gildea, Spike

1994 Semantic and pragmatic inverse: ‘‘inverse alignment’’ and ‘‘inversevoice’’ in Carib of Surinam. In Voice and inversion, T. Givon(ed.), 187–230. Amsterdam/Philadelphia: Benjamins.

Givon, T.1994 ‘‘The pragmatics of de-transitive voice: functional and typologi-

cal aspects of inversion.’’ In Voice and inversion, T. Givon (ed.),3–44. Amsterdam/Philadelphia: Benjamins.

Haspelmath, Martin1994 ‘‘Passive participles across languages.’’ In Voice: Form and Func-

tion, Barbara Fox and Paul J. Hopper (eds.), 151–177. Amsterdam/Philadelphia: Benjamins.

Haude, Katharina2006 A grammar of Movima. Ph.D. diss., Radboud Universiteit

Nijmegen.Haude, Katharina

2009a ‘‘Reference and predication in Movima.’’ In New Challenges inTypology 2: Transcending the Borders and Refining the Distinc-tions, Alexandre Arkhipov and Patience Epps (eds.), 323–342.Berlin: Mouton de Gruyter.

Haude, Katharina2009 Hierarchical alignment in Movima. International Journal of

American Linguistics: 513–532.Haude, Katharina

2010 ‘‘The intransitive basis of Movima clause structure.’’ In Ergativityin Amazonia, Spike Gildea and Francesc Queixalos (eds.), 285–315. Amsterdam/Philadelphia: Benjamins.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


284 Katharina Haude

Haude, KatharinaForthc. ‘‘Argument encoding in Movima: the local domain.’’ In Argu-

ment Coding Systems of Lowland Bolivian Languages (SpecialIssue of International Journal of American Linguistics), AntoineGuillaume and Francoise Rose (eds.)

Hopper, Paul and Sandra A. Thompson1980 ‘‘Transitivity in grammar and discourse.’’ Language 56(2): 251–

299.Kazenin, Konstantin I.

2001 ‘‘The passive voice.’’ In Language Typology and Language Uni-versals: an International Handbook. Vol. 2., Martin Haspelmath,Ekkehard Konig, Wulf Oesterreicher and Wolfgang Raible (eds.),899–916. [Handbucher zur Sprach- und Kommunikationswissen-schaft, 20.] Berlin/New York: Walter de Gruyter.

Keenan, Edward L. and Matthew S. Dryer.2007 ‘‘Passive in the World’s Languages.’’ In Language Typology and

Syntactic Description. Vol. 1. Clause Structure, Timothy Shopen(ed.), 325–361. Cambridge: Cambridge UP. Second Edition.

Kulikov, Leonid2010 ‘‘Voice typology’’. In The Oxford Handbook of Language

Typology, Jae-Jung Song (ed.), 368–398. Oxford: Oxford Uni-versity Press.

Mithun, Marianne1994 ‘‘The implications of ergativity for a Philippine voice system.’’ In

Voice: Form and Function, Barbara Fox and Paul J. Hopper(eds.), 247–277. Amsterdam/Philadelphia: Benjamins.

Nedjalkov, Vladimir P. and Sergej Je. Jaxontov1988 ‘‘The typology of resultative constructions.’’ In Typology of

resultative constructions, Vladimir P. Nedjalkov (ed.): 3–62.Amsterdam/Philadelphia: Benjamins.

Nichols, Johanna1982 ‘‘Ingush Transitivization and Detransitivization.’’ In Proceedings

of the Eighth Annual Meeting of the Berkeley Linguistics Society,445–462.

Nichols, Johanna, David A. Peterson, and Jonathan Barnes2004 ‘‘Transitivizing and detransitivizing languages.’’ Linguistic Typol-

ogy 8(2): 149–211.Perlmutter, David M.

1978 ‘‘Impersonal passives and the Unaccusative Hypothesis.’’ In Pro-ceedings of the Fourth Annual Meeting of the Berkeley LinguisticSociety: 157–189. Berkeley Linguistic Society, University ofCalifornia, Berkeley.

Serzisko, Fritz1991 ‘‘Orientierung.’’ In Partizipation, Hansjakob Seiler and Waldfried

Premper (eds.), 273–308. Tubingen: Gunter Narr.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Shibatani, Masayoshi1985 ‘‘Passive and related constructions: a prototype approach.’’ Lan-

guage 61: 821–848.Shibatani, Masayoshi

2006 ‘‘On the conceptual framework for voice phenomena.’’ Linguistics44(2): 217–269.

Siewierska, Anna1984 The Passive. A Comparative Linguistic Analysis. London et al.:

Croom Helm.Van Valin, Robert D., Jr.

2005 Exploring the Syntax-Semantics Interface. Cambridge: CambridgeUniversity Press.

Van Valin, Robert D., Jr. and Randy R. LaPolla1997 Syntax. Structure, Meaning and Function. Cambridge: Cambridge

University Press.Zuniga, Fernando

2006 Deixis and Alignment. Inverse Systems in Indigenous Languagesof the Americas. Amsterdam/Philadelphia: Benjamins.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


286 Katharina Haude

Case patterns and verb classes in Trumai

Aurore Monod Becquelin and Cedric Becquey

1. Introduction

1.1. Ethnographic and sociolinguistic context

The Trumai are one of the groups of the Upper Xingu (Mato Grosso,

Brazil), first contacted at the end of the 1880s by Carl von den Steinen

(1942). A series of brief expeditions in the early twentieth century was

followed by a period of relative isolation that lasted until the 1960s. Since

then, the growing number of ethnographic studies has produced a series of

monographs on nearly every group in the region, highlighting the multi-

ethnic and multilingual character of this territory (Basso 1973) – fourteen

groups have been identified, belonging to five families: Tupi (Kamaiura,

Kayabi, Aweti, Yudja), Caribe (Kalapalo, Kuikuro, Matipu, Nahukua,

Ikpeng), Arawak (Waura, Yawalapiti), Ge (Kayapo, Kisedje) and Trumai.

The region can be divided into two cultural areas, northern and southern;

the latter (the Uluri area, Galvao 1953), which has formed over the course

of several decades, comprises various southern groups with a certain

number of features in common and a strong multilingual practice (for a

history of the region, see Franchetto and Heckenberger 2001).

Quain was the first anthropologist to work with the Trumai (1938): he

described a very tense situation in which a drastically reduced population

existed in a state of conflict with neighbouring groups (Murphy & Quain

1955).

The creation of the Xingu National Park in 1961 resulted in demo-

graphic growth among the groups of the Upper Xingu and for the Trumai

in particular, but the increasing pressures of modernization, in the face of

which they are still weak numerically, linguistically and culturally, places

them in an unstable situation with regard to the other groups (Monod

Becquelin et al. 2008).

Regional multilingualism is significant – because of the multiple matri-

monial, ritual and economic relations between groups – but also specifically

significant for the speakers represented in the two main corpora: the corpus

of Aurore Monod Becquelin (AMB) shows the influence of Kamayura

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:53) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 287–320 1351 Authier_09_becquelin (p. 287)

(Tupi), while there is Portuguese influence, particularly through schooling,

on the corpus of Raquel Guirardello Damian (RGD).1

1.2. The crux of the matter

This article examines the two bivalent verb classes of Trumai (for descrip-

tion of Trumai verb classes see AMB 1975, 1976 and RGD 1999, 2003).2

Alongside class 1, composed of monovalent verbs taking an unmarked core

argument, class 2 is bivalent and has been labelled ‘‘ergative’’, while the last

major class, class 3, has been analysed as ‘‘extended intransive’’ following

Dixon’s terminology, and employs the unmarked argument together with

the ‘‘extended to core’’ argument E (Dixon 1994: 122–4, Dixon and

Aikhenvald 2000: 3, Guirardello 1999, 2002). In class 2, semantic roles

are canonically encoded by an ergative marker interpreted as an agent

and an absolutive form (identical to the unmarked subject of an intransi-

tive verb) interpreted as the patient. In class 3, the agent takes the absolu-

tive case (unmarked) and the patient is marked by one of three markers

(-s, -ki, -tl ) otherwise used for encoding oblique cases (space and time

locatives and recipient dative).

The di¤erent morphosyntactic alignments and the di¤erent possible

forms taken by argument structure in Trumai have given rise to two

opposing interpretations. The first of these (Trumai as an ergative lan-

guage) tends to favour the ‘‘ergative’’ argument as syntactic subject and

agent (1a), and the unmarked argument as the object of the transitive

(1b) and the subject of an intransitive, whether extended (1c) or not (1a).

The alternative point of view (Trumai as an accusative language) con-

siders the unmarked argument (which would then be labelled ‘‘nomina-

tive’’) as the central argument in verbal constructions (2a,c), necessitating

1. The first survey of Trumai was initiated by A. Monod Becquelin (AMB) in1966, and she continued to work on the language into the 1990s. The researchconducted by the linguist Raquel Guirardello Damian (RGD) began in 1989,and formed the subject of her thesis and subsequent works (see the bibliogra-phy). Texts from the ethnological research of Emmanuel de Vienne (first field-work in 2002) have also been used, as well as a Master’s thesis on Trumai byCedric Becquey based on the documentation gathered by previous researchers.

2. Up to this point Trumai has usually been considered a linguistic isolate. Itsphonological system includes a phoneme unique on the whole American con-tinent, /T/ (this apico-alveolar occlusive, apparently made with the tongueflattened against the palate, has yet to be defined acoustically), and anotherwhich is very rare in South America, /tł/ [ł], [tł], here tl (see Monod Becquelin1975, Guirardello 1999). The main syllabic structures are CVCV(C) andVCVC.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


288 Aurore Monod Becquelin and Cedric Becquey

an analysis of the agent in class 2 (2b) as an agentive complement to a

verb without passive verbal morphology (see Mel’cuk 1983: 248); in class

3, the patient would be interpreted as bearing a set of accusative/dative

su‰xes (2c).

Ergative representation:

(1) a. hine yotl

3abs sleep case labels

subject syntactic functions

‘He sleeps’

b. hine-k k’ate iki

3p-erg fish.abs arrow case labels

agent patient semantic roles

subject object syntactic functions

‘He arrows (the) fish’

c. hine ma k’ate-s/-ki/-tl3abs eat fish-e1/e2/e3 case labels


subject oblique syntactic functions

‘He eats (the) fish’

Accusative representation

(2) a. hine yotl

3nom sleep case labels

subject syntactic functions

‘He sleeps’

b. hine-k k’ate iki

3p-obl fish-nom arrow case labels


oblique subject syntactic functions

‘The fish is arrowed by him’

c. hine ma k’ate-s/-ki/-tl3nom eat fish-acc1/acc2/acc3 case labels


subject object syntactic functions

‘He eats (the) fish’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


Case patterns and verb classes in Trumai 289

These divergences in the analysis of case assignment patterns have

often made it impossible to define syntactic functions in Trumai (RGD

1999: ch.7.3.1) because of the complexity and diversity of the morpho-

syntactic alignments of verbal arguments. Both the descriptive terminology

of the Trumai verbal system and the typological position of Trumai need

to be reconsidered. In particular, this reinterpretation should take into

account the following specific features: in Trumai, no operations on

argument structure are found which involve a change in morphological

diathesis3; changes to argument structure are, however, possible either

by means of a lexical alternation between two parallel forms we call

‘‘doublets’’, or in connection with a syntactic property belonging to certain

verbs known as ‘‘fluid verbs’’ (see 2.1.4); moreover, Trumai lacks a clear

syntactic pivot (RGD 2004). We must a) take into account the very frequent

deletability/optionality of ergative arguments, and b) specify the relationship

between valency and semantic transitivity for each verb class.

The next section briefly presents the verb classes and associated case

patterns using, for the present, the ‘‘ergative’’ terminology established by

earlier publications (RGD 1999). Section 3 takes into account the relation-

ship between semantic roles and morphosyntactic coding. Section 4 shows

that a relevant analysis of Trumai verb classes requires a thorough textual

inquiry in order to provide a sounder understanding of the relationship

between verbal valency and argument structure in classes 2 and 3. The

morphosyntactic alignment split governed by the choice between lexical

‘‘doublets’’ has been scrutinized for numerous texts and with various types

of elicitation; we have found no semantic di¤erences leading to the use of

one or the other class of verbs, and our principal claim is that the use

of the verb classes is discourse-determined.

3. With the exception of the causative construction, which, without changing thetwo-argument structure of the verb, introduces a causer in the ergative caseand a causative morpheme on the verb:

axos ma k’ate-schild eat fish-E1

‘The child eats some fish’

hine-k axos ma-ka k’ate-s3p-erg child eat-caus fish-E1

‘He gets the child to eat some fish’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



2. A brief outline of the verb classes and case patterns

The verbal morphology of Trumai is not extensive. Essentially, it consists

of three sets of markers:

– The first is the absolutive 3rd person (or non-person) su‰x required in

the absence of any lexical expression of this argument. This has two

allomorphs, depending on whether it is preceded by a vowel, V-n, or a

consonant, C-e. The personal pronouns, meanwhile, are independent

forms and in addition to person they indicate number (singular, dual

and plural), gender (for 3rd person), clusivity (inclusive/exclusive) and

case (AMB 1975, RGD 1999).

– Modal, aspectual, directional and circumstantial postverbal grammatical

markers.

– Transcategorial su‰xes (nominalizer, adjectivizer, etc.).

2.1. Verb classes

Case assignment is the most visible criterion for defining a verb class.

Verbs assign cases to all their dependents, and these cases are encoded by

morphological markers governed by schemes which di¤er from one verb

class to another.

Valency refers to the capacity of a verb to take a specific number of

arguments. Class 1 verbs are monovalent; class 2 verbs are bivalent; in

the ergative interpretation of Trumai, class 3 verbs are classifed as mono-

valent and analysed as extended intransitives on morphological grounds.

But if we take into account pragmatics (topic analysis), frequency (distri-

bution over the lexicon), and semantic roles (agent/patient), verbs of class

3 should be considered bivalent (see below). In addition to this, the three

di¤erent forms of E (-s, -ki, -tl ) exhibit domains of application – spatial,

temporal and notional – and semantic values which must be scrutinized

for a better understanding of their syntactic function.

AMB’s first analysis (1975) attributed as much importance to semantic

roles as to morphosyntactic patterns, mixing levels in a confusing way.

This interpretation considered as transitive what RGD calls extended

intransitive; the unmarked argument, which encodes the agent, was con-

sidered to be in the nominative case, while the marked argument, which

encodes the patient, was taken as accusative. Both verb classes were con-

sidered to display equally strong semantic transitivity: the definition of

transitives 2 was based on ergative morphology whereas that of transitives

3 was based on semantics. We will now see that the typological identifica-

tion of Trumai as an ‘ergative language’ is not satisfactory.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



2.1.1. Class 1

This class is composed of monovalent intransitive verbs. The unique argu-

ment is morphologically unmarked.

(3) ha tsula.tsula, ha demTıtapat

1abs be.lying.down 1abs rest.a.little

‘I am lying down, I am resting a little.’ (Plant.)

(4) kachı-n ale hen

go-3abs one.says then

‘Then he went on, one says.’

Other verbs of this class are: ora ‘‘cry (in pain / of animals)’’, wanle

‘‘be over’’, chı ‘‘go’’, fakdits ‘‘die’’, demle ‘‘be tired’’, lafku ‘‘swim’’, otl

‘‘sleep’’ (yotl when immediately following the personal pronouns ha, hi,

hine), pata ‘‘arrive’’, watkan ‘‘weep’’, tsula ‘‘lie down’’.

2.1.2. Class 2

This class is made up of bivalent verbs with ergative alignment. If there

is any prototypical agent it will correspond to the marked argument

(ergative), and the unmarked argument (absolutive) will correspond to

the prototypical patient (but see (6)).

(5) adis.pa-ek ha-wan disi

Indians.of.Xingu-erg 1-pl.abs kill

‘The Indians of Xingu killed us.’ (Murder)

(6) kasoro-k ha tako

dog-erg 1abs bite

‘The dog bites me.’ (Murder)

(7) yaw-ak padi-n de

people-erg wait-3abs already

‘People are already waiting for him.’ (AMB notes)

The imperative form di¤ers from that of the other verb classes: here,

the imperative particle takes the form waki.

(8) Waki kıtı hai-tl

Imp.erg give 1abs-e3 (beneficiary)

‘Give it to me.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Other verbs of this class include: etsi ‘‘bring’’, disi ‘‘kill’’, tako ‘‘bite’’,

padi ‘‘be waited for’’, miro.miro ‘‘rasp’’, maxkewa ‘‘paint’’, k’etan ‘‘move’’,

kapan ‘‘make’’, ts’ake ‘‘be fond of ’’, ni’tsapa ‘‘squeeze’’.

2.1.3. Class 3

The first argument of so-called ‘‘extended intransitive’’ verbs is the un-

marked argument, i.e. takes absolutive case, while the second argument is

labelled E: in the definition of Dixon and Aikhenvald, this is an argument

with a special status (E stands for ‘extension to core’). ‘‘In [some lan-

guages], E and peripheral are treated in the same way’’ (2000: 3): this is

the case in Trumai.

(9) ha ma t’aak-es, k’ate-s

1abs eat cake-E1 fish- E1

‘I eat cakes and fish.’ (Murder)

(10) wana xu’tsa de kayapi wan ami-ki

imp see already Kayabi pl word-E2

‘Remember the words of the Kayabi!’ (Murder)

(11) Uksitukuk-etl otl.taxeL-e

capybara-E3 dream-3abs

‘He dreams of capybara.’ (Karijawar)

The imperative form for this class, as for class 1, is wa/wana.

(12) wana xu’tsa, wana ma, wana yotl

imp see imp eat imp sleep

‘Look, eat and sleep!’

Verbs of this class include: lax ‘‘hunt, fish’’, api ‘‘catch’’, laT ‘‘lie’’,

da’tsi ‘‘carry’’, elka ‘‘exchange’’, fatlkamu ‘‘believe’’, fatne ‘‘clear (the plan-

tation)’’, xu’tsa ‘‘see’’, ma ‘‘eat’’, poyo ‘‘take revenge’’, waymi ‘‘tell’’.

Note that class 2 and 3 both include verbs usually considered highly

transitive, as well as some generally regarded as experiencer verbs (see 3.2).

2.1.4. ‘‘Fluid verbs’’ and ‘‘doublets’’

There is a small set of ‘‘fluid’’ verbs which accept both of the preceding

case assignment patterns while exhibiting the same semantic structure

(without any aspectual/telic motivation for choosing one or the other).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Verbs belonging to this class include: chapta ‘‘clean (the plantation)’’,

t’aka ‘‘repair’’, ami’in ‘‘answer’’, homne ‘‘meet, find’’, pechma ‘‘thread,

insert’’, pırew ‘‘destroy, finish’’.

According to RGD (2003), the di¤erent alignment patterns are some-

times correlated with a slight semantic nuance (routinized actions are

associated with ‘‘extended intransitive’’ alignment?); see Table 1.

Table 1. List of fluid verbs with semantic nuances according to RGD

Verb ergative alignment extended intransitive alignment

naha cut cut maniocochen grind pound, make flourpıTke peel peel maniockıtıw grate grate manioctıami crush, squeeze squeeze manioc pulpwen tear out pluck a bird

However, this semantic nuance is not always in evidence in the texts:

(13) a. hai-ts ıwır naha fakao-letsi

1p-erg wood cut machete-instr

b. ıwır-as ha naha fakao-letsi

wood-e1 1abs cut machete-instr

‘I cut wood with a machete’

Trumai displays another device allowing a shift between the two oppos-

ing argument structures while preserving the semantic relationship between

the two participants. The two constructions involve paired lexemes which

we call ‘‘doublets’’: see table 2.

Table 2. List of doublets

ergative translation extended intr.

chınaha cut nahachoku hit ikidisi kill, beat fafada put to flight chayokapan do chudamapa break ku’kupanu exchange elkatako bite maketı distribute detnetuxa’tsi push dama

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Below we find sentences where both verbs are used successively.

(14) – ‘‘aka! naide! hi-wan fa hai-tl, ha dua’’

Ah it’s.over 2p-pl.abs kill 1p-e3, 1abs bleed’’

‘‘–Ah! It’s over! You’ve killed me, I’m bleeding.’’

disi pchetsi-n ale.

get.killed fall-3abs one.says

‘He got himself killed and fell, one says.’ (Murder)

(15) Atlat paT-es ku’ku-n ale [.]

cooking.pot little-e1 break-3abs one.says [.]

‘He has broken the little cooking pot, they say [.]

iyi mapa.mapaIYI get.broken

‘It got broken.’ (Anu Bird, RGD 2004)

2.1.5. Three-participant constructions

There exist subsets of class 2 (‘‘ergatives’’) and class 3 (‘‘extended intransi-

tives’’) which allow a third argument. The recipient or addressee role is

marked by the same cases as the ‘extended to core’ argument -s/-ki/-tl.

(16) hine-k hit kıtı hai-tl3p-erg arrow.abs give 1p-e3

‘He gives me an arrow.’ (AMB notes)

(17) Ha deTne kain k’ate-s kiki wan-ki.1abs distribute foc fish-e1 man pl-e2

‘I am distributing fish to the men.’ (RGD lexicon)

Verbs displaying this behaviour include: kıT ı ‘‘give’’, pap ‘‘retrieve’’,

tı ‘‘distribute’’, xoma ‘‘teach’’, pechma ‘‘insert’’, waymi ‘‘tell’’.

2.1.6. Minor classes and patterns

There is also a very small class of verbs (sa ‘‘dance’’, hain ‘‘celebrate’’)

both of whose arguments are unmarked:

(18) wan sa-n tawarawana

pl dance-3abs tawarawana

‘They are dancing the tawarawana.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Finally, there is a construction often described as ‘‘object incorpora-

tion’’, which occurs when the possessor of the absolutive argument of an

ergative verb is coreferent with the agent of the process. This device is re-

stricted to a few verbs, which exhibit a type of lexicalization whereby the

absolutive argument of the ergative construction combines with the verb

to produce a monovalent compound verb; the agent takes the unmarked

absolutive form.

(19) mut.pupe-n

garment.take.o¤-3abs

‘He is undressing.’

2.2. Cases

In the previous examples we considered the di¤erent verb classes; we now

examine the associated case markers.

2.2.1. Unmarked case

This case encodes various semantic roles (agent, experiencer, patient. . .):

ha fa ‘‘I kill’’, ha ma ‘‘I eat’’, ha xu’tsa ‘‘I see’’, ha yotl ‘‘I sleep’’, ha padi

‘‘I am waited for’’.

It is used for the unique argument of a monovalent verb (intransitive),

or for the unmarked argument of a bivalent verb, whether this is the

absolutive in an ergative frame (patient-oriented: ha padi hinek ‘‘I get

waited for by him’’) or the absolutive in an extended intransitive frame

(agent-oriented: ha fa hinetl ‘‘I kill him’’). In morphosyntactic terms the

unmarked argument, whether labelled absolutive or nominative according

to the theoretical interpretation, encodes a proto-agent in a ‘‘pseudo’’

extended intransitive construction, or a proto-patient in the ergative con-

struction, and any role in the intransitive construction.

2.2.2. -(V)k: ergative case

This case encodes not only the agent of ‘‘ergative’’ verbs, but also the

causer in causative constructions involving verbs of classes 1, 2, and 3.

(20) Yakır etsi ka-wan-eksalt carry 1excl-pl-erg

‘We are carrying the salt.’ (Tapir)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(21) talel ıch wapta-ka totsit-ikdoor catch fall-caus totsit-erg

‘The totsit bird causes the door catch to fall.’ (myth AMB msc.)

(22) hai-ts chıin Atawaka-k mapa-ka-n.

1p-erg foc/tens Atawaka-erg break-caus-3abs

‘I made Atawaka break it.’ (RDG 2003: 211)

2.2.3. Extension to core

Extensions to core, recipients and locatives, which share the same set of

markers, can all be subsumed under the label ‘goal’ in the notional ((23),

(25), (28), (29)) spatial (26) and temporal (27) domains. These cases have

been given di¤erent interpretations in previous descriptions of Trumai, see

table 3.

Table 3. Specification of cases according to AMB (1975) and RGD (2004)

Marker AMB RGD

-tl DativeIndividualized, identifiable andsalient dative

-ki Targeted accusativeIndividualized, non-identifiable andnon-salient dative; Non-individualized,identifiable and non-salient dative

-s Indefinite accusativeNon-individualized, non-identifiable andnon-salient dative

As in many languages, the object of an extended intransitive verb bears

the same case marker as certain locatives – answering the ubi/quo ques-

tions, -s (‘at’) and ki (‘to’) – meaning that it is sometimes di‰cult to decide

whether one is dealing with an intransitive verb with locative (23a) or an

‘‘extended intransitive’’ (23b).4

4. Here is an interesting example where the informant obviously plays on thisambiguity:

hai.hen, sone-n ale hen misu-s misu-ki;then, drink-3abs one.says then water/river-E1 water/river-loc2

‘Then, one says that they drink water at the river’

Ina.ik otl-e. kodaka-ki ukanright.there sleep-3abs. dawn-loc2 again‘And they slept immediately. The next day’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



-(V)s: Extension to core (E1); but also spatio-temporal locative (LOC1)

(23) a. – ‘‘ha kaxmi tsimo-s’’

1abs go.to.forest cipo-loc1

‘I am going to the forest for cipo’

b. – ‘‘ma ka-wan fa kawa tsimo-s’’

Let’s.go! 1excl-pl.abs beat go cipo-e1

‘Let’s go and beat the cipo!’

(24) kawixu-ami-aduru-s hen yaw si de

rain-voice-noise-loc1 then people burn already

‘When the thunder rumbles, people have already burned

[the plantation]’ (Plant.)

-ki: Accusative, dative (E2); spatio-temporal locative (loc2)

(25) ure.ure silo-ki api kakda-n

parrot crown.of.feathers-e2 take early.in.the.morning-3abs

‘At dawn he took the parrot’s crown of feathers.’ (Orig. day)

(26) yaw si de kut’a-ki.

people burn already plantation-e2

‘People have already burned (in) the plantation.’ (Plant.)

(27) tach kodaka-ki ha yatxa.

more tomorrow-loc2 1abs work.on.the.plantation

‘Again the following day, I work on the plantation.’ (Plant.)

-(V)tl: Dative, Accusative (E3)

(28) tach ats’aek tı hen tach kiki-tl.

more manioc.gruel distribute then more man-e3

‘The manioc gruel is distributed to the men once more.’ (Plant.)

sone-n misu-ki. hai.de otl-e. kodaka-kidrink-3abs water/river-E/loc2 already sleep-3abs dawn-loc2

‘they drink (at the river? water?). They are already sleeping. The next day,’

sone-n misu-ki, sone-n misu-s.drink-3abs water/river-E2, drink-3abs water/river-E1

‘they drink ‘‘this water’’, they drink ‘‘some water’’.’ (Payetan)

We could assign the generic term ‘‘goal’’ to these two labels (E/LOC), wereit not for the fact that the valency of the verb entails a required core argumentin one case and a circumstantial one in the other.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(29) kayapi-tl wan yenye-n

Kayabi-E3 pl imitate-3abs

‘They imitate the Kayabi.’ (AMB notes)

(30) xu’tsa tak Trumai-s pıtık-etl Trumai ami ‘‘kuT’’

see neg Trumai(abs)-loc1 monkey-E3 Trumai say kuT

‘When the Trumai can’t see the monkey, the Trumai say kuT

(‘‘water spirit’’).’ (AMB notes)

2.2.4. Oblique cases

-tam: Comitative (COM)

(31) ha pine-tam ha lax kawa

1 friend-com 1abs hunt go

‘With my friend, I go hunting’

(32) hai-tl-tam werew waki kıtı

1p-e3-com a.little imp give

‘Give me a little bit!’ (Sun and Moon)

The comitative marker is su‰xed to nouns but also to personal pro-

nouns marked by -tl; case stacking occurs only on personal pronouns, as

in (32), and is impossible with the case markers -s and -ki.

-letsi: instrumental (INSTR)

(33) intsatske yaw cho xu-letsi de.

next people perforate needle-instr already

‘And then one perforates with a needle.’ (Harpoon)

-ita: Allative (ALL)

(34) ni-s sela-ita hen pech-lapchı-n ale ıd ıcha-tam

here-loc1 stern-all then run-cont-3abs one.says bow-com

‘Here, he runs towards the stern with his bow, one says.’ (Murder)

-lots: Ablative (ABL)

(35) – ‘‘hamosin hi wax, Katsini?’’ – ‘‘ha petl-lots’’ kale

– where 2abs fart Katsini? – 1abs behind-abl cit

‘– ‘‘Where do you fart, Katsini ?’’ – ‘‘From my behind.’’ ’

(Fish festival)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



-n: illative/allative temporal and spatial (LOC)

(36) tamula-n misu-n xuma-ktsu-n

zenith-loc water-loc bathe-towards.the.river-3abs

‘At midday, he bathes in the river.’ (AMB notes)

3. Limitations of the ergative analysis

The analysis of Trumai as displaying an ergative system combined with

the existence of an extended intransitive verb class was motivated mainly

by the desire to take into account the morphosyntactic alignment of verb

class 2, intuitively considered to be canonical and predominant among the

two-place verbs. This analysis, following Dixon (1994: 123), asserts that

the ergative verbs are those with the highest transitivity, possessing two

nuclear arguments (ergative and absolutive). According to this account

the extended intransitive verbs have an absolutive subject (identical in

form to the subject of a monovalent intransitive verb) and an additional

argument marked as the recipient/addressee in a three-participant construc-

tion (see 2.1.5).

But this interpretation does not entirely take the empirical data into

account. An examination of Trumai vocabulary and texts reveals a quite

di¤erent situation from that described by Dixon: the extended intransitive

construction is greatly underemphasized by Dixon (who talks of ‘‘minor’’

transitivity and an ‘‘additional’’ argument), while textual analysis reveals

its statistical weight and its often high semantic transitivity.

3.1. Is the ergative alignment dominant?

If the purpose of examining the morphosyntax of Trumai is to describe the

verbal system as a whole, or at least its dominant alignment, then an

observation of the distribution of two-place verbs in the lexicons (fig. 1)

and texts (table 4) collected for Trumai shows that ergative verbs are not

predominant.

Setting aside the fluid verbs, we can see from fig. 1 that in the vocabu-

lary extended intransitive verbs predominate slightly over verbs of the

ergative class. This superiority can also be observed in texts, where verbs

expressing one or more arguments (see below on the optionality of the dif-

ferent cases) and presenting extended intransitive alignment are distinctly

more numerous than those with ergative alignment.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 4. Distribution of two-place verbs in texts

Text5 Extended intransitive verbs Ergative verbs

Plantation 75.68%28/37

24.32%9/37

Murder 75.76%75/99

24.24%24/99

Tapir 58.82%20/34

41.18%14/34

Orig. day (N.) 58.06%36/62

41.94%26/62

Orig. day (W.) 60.38%32/53

39.62%21/53

Figure 1. Distribution of two-place verbs in the lexicons of AMB (190 verbs) andRGD (274 verbs)

5. We have chosen texts which are as diverse as possible in order to reduce dis-tortion in the counting, which is still at a preliminary stage. ‘‘Plantation’’(Plant.) is an ethnotext recounting the preparation of a field of manioc; ‘‘Murder’’is the account by a Trumai of the murder of one of his relatives (though we shallsee later that the title of this narrative, given by the author who collected it,should be changed: it is not a murder narrative but a political text, as shown bythe syntactic behaviour of its verbs); the story of the ‘‘tapir’’ recounts theorigin of ritual songs; the two narratives concerning ‘‘the origin of the day’’are two myths recounted forty years apart, one collected by AMB (orig. dayN.) in 1967 and the other by E. de Vienne (orig. day W.) in 2005.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



If these facts call into question the typological classification of this

language as ergative, they also run counter to a description which attrib-

utes extended intransitivity to non-ergative verbs. Dixon points out: ‘‘An

important point to note is that these extended subclasses are always rela-

tively minor. Most transitive verbs are canonically transitive with two core

roles; only a small number will be extended transitive (or ditransitive) with

an additional role’’ (1994: 123). He repeats this claim with Aikhenvald:

‘‘In a few languages (e.g. Tonga, Trumai), there is also an extended in-

transitive clause type, with S and E [. . .] In every language in which they

occur, extended intransitives and extended transitive clause types are

greatly outnumbered – in dictionary and texts – by plain intransitive and

plain transitive’’ (Dixon & Aikhenvald 2000: 3).

It appears that, grounded though it is on data from a wide range of

languages, a confrontation of this cross-linguistic generalization with the

Trumai facts obliges us to rethink these supposedly ‘‘additional roles’’. The

second argument of an extended intransitive verb is not a mere adjunct,

which may or may not be lexically expressed; it is required by the meaning

of the verb, and there are therefore di¤erences between this argument and

optional elements such as ‘‘comitative’’, ‘‘locative’’, ‘‘instrumental’’. In any

case, with verbs of this class a lexical object can always be expressed.

3.2. Transitivity and verb classes

Dixon’s reasoning here is based on a scalar view of the transitivity of the

di¤erent verb classes; on this scale the class of ergative verbs is held to be

systematically higher-placed than the class of extended intransitives.

Bearing in mind a definition of prototypical semantic transitivity such as

that proposed by Givon (1990: 565–566) (see also Hopper and Thompson’s

(1980) scalar concept of transitivity and Dowty’s (1991) conception of

semantic transitivity), it is easy to see that verbs such as ‘‘kill’’, ‘‘cut’’,

‘‘break’’, ‘‘destroy’’ can be found in both classes of Trumai verbs, as well

as in the class of fluid verbs (see Table 5).

Table 5. Classes of prototypically transitive verbs

Trumai translation class

chı cut extended intransitive

naha cut fluid

(o) fa kill, beat extended intransitive

disi kill, beat ergative

pırew destroy, finish fluid

fatla pierce ergative

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In fact, as Dixon himself observes, semantic transitivity and bivalence

are not coextensive. Semantic transitivity and morphosyntactic class mem-

bership, which are often observed to display a correlation among the

languages of the world (Tsunoda 1981, 1985), appear to be independent

in Trumai. In a quick survey of the lexicon, classifying verbs according

to the semantic fields they encode, no such correlation appears; the two

classes coexist in each conceptual domain, with the notable exception of

verbs of perception and cognition.6

The semantic fields presented here are based on the works of Tsunoda

(1981, 1985) and the Wordnet lexical database developed by the Princeton

University Cognitive Science Laboratory as a tool applicable to all of the

world’s languages. Here, they are intended solely as very general indica-

tive labels, delimiting broad conceptual domains which can serve as input

to a preliminary classification; they are not intended to represent the

semantic domains of Trumai.

Semantically low-transitive verbs, such as cognizer or experiencer verbs,

select particular patterns in Trumai (as in most languages) and are, there-

fore, not relevant for typological classification. On the other hand, the list

of extended intransitive verbs shows that the semantic categories recorded

for them by Guirardello do not cover all available conceptual domains:

they do indeed include verbs of perception, mental activity and verbs of

‘‘contact’’ or ‘‘movement’’ (RGD 2003: 202), but all the others are con-

sidered ‘‘habitual activities’’, such as ‘‘eating’’, ‘‘drinking’’ or ‘‘cooking’’

(RGD 2002: 6). Furthermore, what are we to make of verbs which show

the same alignment but do not refer to perception or cognition, or to

contact or routine actions, such as look for, count, arrange, trust, explode,

skip along, move away, fight, kill, tear, bury, avenge, steal, etc.?

At the same time, a large number of verbs for ‘‘peaceful’’ (i.e. not pro-

totypically transitive) actions are associated with the ergative construction,

such as advise, tickle, touch lightly, anoint, paint, wait for. . .

This apparent dissociation between the referential process, the proto-

roles and the morphosyntax of the verb is not without consequences

for the determination of the valency of the di¤erent Trumai verb classes.

6. As has been noticed by many linguists, semantically low-transitive verbs –such as experiencer or cognizer verbs – select di¤erent patterns in mostlanguages and are therefore not relevant for typological classification. Thesemantic domains associated with the extended intransitive construction are‘‘seeing, hearing, liking and wanting’’, and for the extended transitives,‘‘giving, showing, telling’’ (Dixon & Aikhenvald 2000: 3).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



General definitions, such as that provided by Creissels (2004: 3), state that

‘‘a construction will . . . be denoted as transitive if and only if it contains a

verb accompanied by two nominal terms with which that verb is con-

structed in the same way as a prototypical action verb, with the two nom-

inal constituents representing the agent and the patient, in a construction

where agent and patient are both treated as nuclear syntactic terms’’ (our

translation). But the determination of such prototypical action verbs is

extremely problematic in Trumai. The only means by which we can allot

valency to verb classes (2 and 3) is by determining the degrees of nuclear-

ity of the arguments involved in the di¤erent classes, using purely morpho-

syntactic criteria.

The criteria most frequently used to di¤erentiate between core and

oblique arguments are the following (see, for example, Lazard 1994: 68–

Figure 2. Distribution of verb classes in all the lexicons of AMB and RGDaccording to a number of conceptual domains

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



77, Creissels 2006: 273–275): verb agreement, obligatory character, mor-

phological marking, and word order constraints. We can observe in

Trumai a strong asymmetry between the absolutive argument and the

others – the absolutive argument governs verb agreement, and it is the

only obligatory argument, the only argument that is not marked mor-

phologically and the only argument subject to a strict word order con-

straint. At this stage, nothing allows us to distinguish morphosyntactically

between the ergative, accusative and dative cases and oblique arguments

such as temporal and spatial locatives.

3.3. Argument omission

In table 6, we show the asymmetry between the optionality of the extended

argument in class 3 and of the ergative argument in class 2.

Table 6. Omission of the main case-marked argument

Tokens in: % of ‘‘extended intransitive’’verbs (class 3) without ext.to core argument

% of ‘‘ergative’’ verbs (class 2)without ergative argument

Murder 29.33%22/75

54.17%13/24

Plantation 46.43%13/28

100.00%7

9/9

Orig. day (N.) 50.00%18/36

96.15%25/26

Orig. day (W.) 25.00%8/32

42.62%10/21

We might be tempted to translate ergative verbs by passives. However,

in the absence of a morphological passive marker in Trumai, in transla-

tions we prefer the expression ‘‘get (oneself )þ past participle’’ to the

canonical English passive.

In light of the fact that all arguments apart from the unmarked one are

only optionally expressed, it might be concluded that the language con-

7. We cannot explain the fact that in this ethnotext, where the speaker is describ-ing the successive activities he performs to make a plantation, there is not asingle example of the agent in the ergative case, whereas fewer than half ofthe ‘‘extended to core’’ arguments go unexpressed.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



tains only intransitive verbs. It is worth noting that the frequent absence of

the ergative argument in texts is strong evidence for the preeminence of

the absolutive argument (the unique argument of intransitives, the core

argument of ‘‘ergative’’ verbs without the ergative argument and the un-

marked argument of ‘‘extended intransitive’’ verbs). We consider the rela-

tionship between the optionality of the ergative argument in class 2 and

the optionality of the ‘‘extension to core’’ argument in class 3 to require a

more thorough examination, but Table 6 shows a much higher percentage

of omission for ergative arguments than for extension to core arguments.

4. Revisiting the Trumai data

Given the above observations based on the textual data, the analysis of

Trumai as an ergative language on the basis of the ergative morphosyntactic

alignment of some of its verbs is unsatisfactory. We are dealing with a

language which does not in fact display any predominant alignment in

the lexicon, whether ‘‘ergative’’ or ‘‘extended intransitive’’ (see Fig. 1).

Throughout the texts (see Table 4), extended intransitive is far more fre-

quent than ergative alignment. We have also noticed that class 2 and class

3 show a similar relationship between their core arguments. Finally, the

distribution of case assignment appears to be independent of the semantic

transitivity of the verb: semantically high-transitive verbs are distributed

almost equally between an ergative construction and what we now identify,

by contrast, as an accusative construction.

Liberated from the constraints imposed by its previous typological assign-

ment, we are now in a position to reconsider the Trumai verbal system as

a whole.

4.1. Determination of the syntactic subject

Based on the works of Lazard (1994), Keenan (1976) and Mel’cuk (1983),

a preliminary approach shows that the unmarked argument in the three

constructions (intransitive, ergative and extended intransitive) fulfils many

of the prerequisites for defining the syntactic subject. The unmarked argu-

ment is obligatory in every construction8; it governs verbal inflection (-n/-e

in the 3rd person); it is not marked morphologically; it represents the

8. In our texts, the total absence of the absolutive argument occurs only inimpersonal constructions such as kawixukla ‘‘it is raining’’. A few other casesrequire clarification. In a vocabulary survey, some class 2 verbs expressed

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



prototypical agent in the ‘‘extended intransitive’’ construction; it is the

addressee of the imperative9 in the ‘‘extended intransitive’’ construction;

it commands reflexive, reciprocal and middle use (in the rare cases where

the deletion of the ergative argument gives rise to a middle interpreta-

tion, see (37)); and it is placed according to syntactic constraints of linear

precedence.

(37) a. Kumaru-k ha tıchı.

Kumaru-erg 1abs scarify

‘Kumaru scarifies me.’ (RGD 2003: 22)

b. ha tıchı.

1abs scarify

‘I am scarified (by someone), I scarify myself.’

(RGD 2003: 22)

On the other hand, the ergative-marked argument fulfils only a few

functions: it represents the prototypical agent, and is the addressee of the

imperative, but it is not obligatory, does not govern verbal inflection,

is morphologically marked, does not command reflexive, reciprocal and

middle use, and is less constrained by word order.

This inventory of properties is not exhaustive, but the asymmetry

between the two sets of subject properties in favour of the unmarked argu-

ment raises the question of the latter’s actual syntactic functions and those

of the ergative argument.

The causative construction in Trumai, especially when applied to class

2, see (22), is problematic because it shows two instances of ergative mark-

ing, which on the ergative interpretation would be interpreted as two

in the 1st person ergative do not take the extralocutive su‰x that normallyappears. In her study on valency changes, Guirardello Damian mentions thatthe absolutive can be omitted when U of the intransitive and S of the extendedintransitive (in her terms) bear a semantic e¤ect of genericity: an event occurs,but one does not know who or what produced it. Tsixu’tsa kain iyi-n ‘‘it iscold’’ATsixu’tsa kain iyi ‘‘it (the weather) is cold’’. This analysis is not con-clusive, insofar as these phrases include an element – iyi – whose syntactic andsemantic value is not known for certain.

9. The allocutor of the imperative has always been analysed as the ergative argu-ment. However, it is surprising that the imperative particle changes accordingto the semantic nature of the patient (human wa or non-human waki). More-over, the ‘‘absolutive’’ argument is never mentioned in this construction.Lastly, the wana form, which is characteristic of ‘‘extended intransitive’’ (andintransitive) verbs, can be employed with ‘‘ergative’’ verbs in a medio-passiveuse where there is identity between the agent and the patient.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



separate ergative subjects, but ‘‘doubling on subjects is unknown in causa-

tive constructions’’ (Comrie, 1989: 178). In any case, this construction is

uncommon in Trumai and does not seem to be of central importance in

the identification of the subject.

One way to pin down the identity of the subject is to scrutinize the

behaviour seen in anaphora. Anaphors are generally controlled by the

subject: ‘‘the subject case tends to code the most important, recurrent, con-

tinuous topic. We may call it the primary clausal topic’’ (Givon 1984: 138,

cited by RGD 2004). However, in Trumai, the control of anaphoric ele-

ments is pragmatic, and the antecedent is neither the A nor the P of the

previous clause but the topical noun phrase which for pragmatic reasons

is the most ‘‘plausible’’ in the logic of the discourse (see for demonstration

RGD 2004).

Given that anaphora is not controlled syntactically, it constitutes a

useful device for determining the intentions of the speaker by considering

the switch of topics within narrative texts.

4.2. Marking the topic

A study making use of this feature was outlined by RDG (2004): ‘‘it seems

that in a Trumai transitive clause [i.e. verb class 2], it is the Ergative that

codifies the topical participant. If this is indeed the case, it would mean

that the Ergative is the Subject of transitive clauses in Trumai’’. In a sense

her presentation slants the results somewhat, as it counts all other verbs

(class 1, 3 and fluid) as one- or two-place intransitives. But in her sample

this topical participant occurs as ergative 6 times, and 5 times as absolu-

tive with a second argument; in none of the clauses is it presented as a full

noun phrase.

Our preliminary study was conducted on a diverse set of texts (an

account of a murder, an ethnotext on a plantation, and a collection of

myths). This study brings to light a strong correlation between the dis-

course topic and the unmarked argument, as the following examples illus-

trate (by ‘topicalization’ we mean, for the time being, the arrangement of

sentence structure according to the discourse roles chosen by the speaker

in context).

In a text narrating the murder of a Trumai (Javaritı), which the narra-

tor attributes to a neighbouring group, the Kayabi, the following excerpt

relates the events leading up to his death from the victim’s point of view.

The ‘‘victim’’ participant, whatever his semantic role, systematically ap-

pears unmarked, as the primary clausal topic and the syntactic subject

(appendix, lines 67 to 75).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Table 7. List of the 13 verbs in ‘‘Murder’’, line 67–75

Verb verb class participant abs translation participant erg/e

waymi extended intr. Javaritı speaks Kayabi


lax extended intr. Javaritı fishes/hunts Kayabi


chı intr. Javaritı rows

iki ergative Javaritı is aimed at Kayabi

iki ergative Javaritı is aimed at Kayabi

ora intr. Javaritı shouts

(i)ki ergative Javaritı is aimed at Kayabi

pech intr. Javaritı runs



ku intr. Javaritı shouts

In the subsequent passage (lines 75 to 82), the Kayabi constitute the

discourse topic: accused of the murder, they are in turn promoted to the

syntactic function of subject, as the main theme of this speech argues

for revenge against them. This promotion entails the selection of verbs

belonging to a di¤erent class from those used in the passage above for

the same semantic content (e.g. iki/chomta ‘‘aim at’’).

Table 8. List of the 11 verbs in ‘‘Murder’’, line 75–82

Verb verb class participant abs translation participant erg/e

fa extended intr. Kayabi kill

fa extended intr. Kayabi kill Javaritı

fa extended intr. Kayabi kill Javaritı

(o) fa extended intr. Kayabi kill

damtsi ergative Javaritı gets chased Kayabi

ka’chı pita intr. Kayabi go out

chomta extended intr. Kayabi aim at

chomta extended intr. Kayabi aim at Javaritı

choku extended intr. Kayabi hit tree



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Javaritı only appears once in this sequence as the syntactic subject (i.e.,

the unmarked argument) of the verb damtsi, an ergative verb. This topical

marking is made possible by resorting to two di¤erent phenomena in the

language, phenomena constrained by the syntactic and lexical resources of

Trumai.

Our analysis shows that the text we had previously entitled ‘‘the murder of

Javaritı’’ is in reality told from the perspective of a political conflict between

two groups – both antagonists being treated in turn as the topic – and would

therefore a more appropriate title would be ‘‘Trumai political evaluation of

an interethnic murder’’. The fine-grained analysis of a discourse device,

topicalization, thus reveals the true meaning (purpose) of the text.

4.3. The di¤erent interpretations of the Trumai system

Despite its preliminary state, the above study demonstrates the need to

rethink the Trumai verbal system. Drawing notably on the universal

syntactic-semantic primitives of Dixon (1994: 6–9), relabelled by Creissels

(2006: 300) as A(gentive), P(atientive) and U(nique argument), we now pro-

pose to summarize the di¤erent interpretations of this system put forward

so far and to suggest a new one.

4.3.1. Once again: the ergative interpretation?

The ergative interpretation, depicted in table 9, now appears to be un-

sustainable. The consequences imposed by an interpretation of this kind,

such as the assignment of the syntactic function of subject to the ergative

argument and the di¤erential treatment of the morphosyntactic properties

of ergative (nuclear) and extended to core arguments (see (1)), are not

warranted on the basis of the data observed in Trumai texts.

Table 9. Ergative interpretation

Coreargument 1

Casemarker

Coreargument 2

Casemarker

Extendedto core

argument

Casemarker

Class 1U

Absolutivenone

Class 2A

Ergative-(V )k

PAbsolutive

none

Class 3A

Absolutivenone E

-(V )s-ki

-(V )tl

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



4.3.2. The accusative interpretation?

The accusative interpretation (see (2)) appears to solve the problem of the

assignment of the function of subject to the unmarked argument. How-

ever, it requires us to posit the existence of verbs with a ‘‘passive mean-

ing’’, (i.e., patient-oriented with only optional mention of the agent),

reversing the preceding situation with regard to the nuclearity of the

marked arguments in -(V)k without providing a satisfactory solution.

Table 10. Accusative interpretation

Coreargument

1

Casemarker

Coreargument

2

Casemarker

Obliqueargument

Casemarker

Class 1U

Nominativenone

Class 2P

Nominativenone

AOblique

-(V )k

Class 3A

Nominativenone

PAccusative

-(V )s-ki

-(V )tl

Second PDative

-(V )tl

4.3.3. The ‘all-intransitive’ interpretation?

As we saw earlier (section 3.2.), no overt distinction can be observed at

the morphosyntactic level which could be used to di¤erentiate between

ergative / accusative / dative arguments and oblique arguments. It would

therefore be possible to postulate the non-existence of transitive verbs in

Trumai – a proposal already made for languages such as Lezgian, for

example, by Mel’cuk (1983).

According to this interpretation, there is no di¤erence between the

marked cases in terms of their relationship with ‘macroclass I’. However,

there are di¤erences on the semantic and syntactic levels, since some case

markers (namely -k, -s, -ki, -tl ) show a preferential relation with semanti-

cally transitive verbs and combine symmetrically with agent and patient

roles.

Our claim that there are two core arguments in both classes is supported

by valency changes, such as the middle construction, where the ergative

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



argument is deleted (see (37b)), and object incorporation (see (19)). One of

the two participants, either agent or patient, can be selected in the syntax

as the subject according to its topical status (see 4.2 and annexes).

The exclusive accessibility of agents and patients to syntactic opera-

tions such as those mentioned seriously undermines the ‘‘all-intransitive’’

interpretation.

4.3.4. Closing remarks

The ergative perspective removes the E(xtension to core) from core argu-

ments to the periphery; the accusative perspective disregards the ergative

argument; the all-intransitive point of view downplays both. The three

analyses thus all seem inadequate and inconclusive. We therefore posit

the existence of a division between intransitive and transitive verbs. Tran-

sitive verbs can be agent-oriented (previously ‘‘extended intransitive’’) or

patient-oriented (previously ‘‘ergative’’).

The coexistence in a single language of two sets of transitive verbs, one

of which shows ergative features and the other accusative features, is well

known among Austronesian languages. However, the remarkable charac-

teristic of Trumai is that the ergative/accusative split is lexically governed

Table 11. The all-intransitive interpretation

Coreargument

Casemarker

Oblique argumentCase

marker

Class I(1/2/3)

Ua/p

uniqueargument

none

Agent -(V )k

Definite patient, definite targetaddressee, terminative (locative)

-ki

Individualized patient, addressee/recipient (indefinite or definite)

-(V )tl

Indefinite patient, inessive,temporal locative

-(V )s

Allative -ita

Ablative -lots

Comitative -tam

Instrumental -letsi

Locative -n

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



for the class of verbs semantically high in transitivity: some semantically

high-transitive verbs select the ergative construction, others the accusative

construction. Furthermore, unlike in Austronesian languages, this split

never involves morphological marking on the verb even if both alignments

are available for the same semantic structure (fluid verbs and doublets).

Table 12. Alternative interpretation

Subjectargument

Casemarker

Complementargument

Casemarker

Class 1U

unmarked casenone

Class 2patient-oriented

Punmarked case

none Ergative -(V )k

Class 3agent-oriented

Aunmarked case

none

Accusative; Dative -(V )tl

Accusative; Dative -(V )s

Accusative; Dative -ki

It appears that one means of interpreting the di¤erence between these

two major constructions (patient-oriented and agent-oriented) might con-

sist not in attempting to define the subject with reference to either mor-

phosyntax or semantic roles, but in gauging as precisely as possible the

relationship between the two arguments of each construction as seen in

texts: is there one in which the patient is closer or more distant (ontologi-

cally, contextually or notionally) from the agent (ha ma k’ates/haits atlat

mapa ‘‘I eat fish’’/ ‘‘by me, the pot gets broken’’), or more or less a¤ected

by the action (haits k’ate naha/ha chı k’ates ‘‘by me, the fish gets cut’’ / ‘‘I

cut the fish’’)? What type of change a¤ects the relation when one of the

two arguments is not mentioned (ha ma ø / ha padi hine-k ‘‘I eat’’ / ‘‘I get

waited for by him’’)? Could the ha ‘‘I’’ of ha padi in fact be semantically

‘‘more’’ agentive than the ergative-marked hine-k ‘‘he’’ (the one waiting),

as is potentially reflected in French ‘‘je me fais attendre par lui’’? A syntac-

tic complement representing the patient is marked by di¤erent accusative

cases according to semantic properties combining semantic role and char-

acteristics such as definiteness, individuation and number, which could

explain the variety of forms available: these relations almost certainly

involve factors that we have not yet imagined, and which will only be

verifiable after considerable work on texts.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



5. Appendix: Political evaluation of an interethnic murder

This narrative was performed for AMB in 1967 by the son of the Trumai

chief, and refers to the murder of one of his uncles a few weeks before; the

speaker contrasts his own opinion about the murderers with his father’s,

each of them having a di¤erent judgement on interethnic conflicts in the

Upper Xingu region. We give detailed interlinear glosses in sections exem-

plifying points made in the article, especially with regard to the identity of

topics and subjects.

067 Kawan amihak kawanki. . .

kawa-n ami-hak ka-wan-ki

go-3abs say-for 1excl-pl-e2

Kawan hi waymi aehak kale hi waymi aehak kawanki.

kawan hi waymi ae-hak kale hi waymi ae-hak ka-wan-ki

let’s 2abs tell well-for one.says 2abs tell well-for 1excl-pl-e2

‘For you to tell us. . .

For you to tell us well, for you to tell us well.’

(waymi: class 3, J.: abs)

068 – ‘‘Huk’anik ha alax kawain ukan

huk’anik ha lax kawa-in ukan

wait 1abs fish go-foc again

heletsis kain ha waymi hi wanki’’ kale hen amin le.

heletsis kain ha waymi hi wan-ki kale hen ami-n le

and.then foc 1p tell 2abs pl-e2 one.says then tell-3abs one.says

‘–‘‘Wait! I’ll go fishing again

and then I shall tell you’’, he said, as reported;’

(alax: class 3, J.: abs / waymi: class 3, J.: abs)

069 chı lapchınes hen iki pchıkidan ale natues,

chı-lapchı-n-es hen iki-pchı.kida -n ale natu -es

paddle-cont-3abs-loc1 then get.hit-begin-3abs one.says back-loc1

‘at that moment he went paddling away upstream, then he started

to get shot in the back, one says’

(chı: class 1, J.: abs / iki: class 2, J.: patient, abs; no agent)

070 natues.

natu-es

back-loc1

‘in the back.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



071 Iki pchıkidan ale natues.‘‘Aka!’’ Kale hen

iki-pchı.kida-n ale natu -es aka! kale hen

get.hit-begin-3abs one.says back-loc1 Oh! one.says then

oran ale ‘‘Aka!’’

Ora-n ale aka!

shout-3abs one.says Oh!

‘He was shot in the back: ‘‘Ow!’’ He said then

he shouted in pain; ‘‘Ow!’’ ’

(iki: class 2, J.: patient, abs; no agent / ora: class 1, J.: abs)

072 hanhak de ha iki hi wanek.’’

han.hak de ha iki hi-wan-ek

what.for already 1abs get.hit 2p-pl-erg

‘Why am I getting shot by you?’’ ’

(ki: class 2, J.: patient, abs; K.: agent, erg)

073 Ni selaita hen pech lapchın ale ıd ıchatam

ni sela-ita hen pech-lapchı-n ale ıd ıch-tam

here prow-all then run-cont-3abs one.says bow-com

t’ı t’ı t’ı pech laxmin ale hen wıTıki.

t’ı t’ı t’ı pech-laxmi-n ale hen wıTı-ki

onom run-towards.forest-3abs one.says then dry-loc2

‘He runs to the prow with his bow, as reported, t’ı t’ı t’ı he runs

towards the forest, to a dry place.’

( pech: class 1, J.: abs / pech: class 1, J.: abs)

074 Pech laxmin;

pech-laxmi-n

run-towards.forest-3abs

‘He runs towards the forest;’

( pech: class 1, J.: abs)

075 haihen waTıma waTıma ku pchıkidan ale.

hai.hen waTıma.waTıma ku-pchı.kida-n ale

then noise.noise shout-begin-3abs one.says

‘‘Helein hi fa, Tapiukat?

hele -in hi fa Tapiukat?

what-foc 2abs kill Tapiukat?

‘and then, he begins to shout, as reported: ‘‘why are you killing

me, T.?’

( fa: class 3, K.: agent, abs; no patient)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



076 Han ma’tsik daetl hi wan fa?

han.ma’tsi.k dae-tl hi-wan fa?

for.what.reason innocent-e3 2p-pl.abs kill?

‘For what reason are you killing an innocent?’


077 Helein de tsiets haitl hi wan fa?

hele.in de tsiets hai-tl hi-wan fa?

what already why 1p-e3 2p-pl.abs kill?

‘Why are you killing me?’

( fa: class 3, K.: agent, abs; J.: patient, acc(usative))

078 Tsifan chıin de hi wanek ma’tsike

tsifan chı-in de hi-wan-ek ma’tsi.ke

thing cop-foc already 2p-pl-erg reason.for

kaaletl hi wan ofahak’’ kaale amin ale.

kaale-tl hi-wan ofa-hak kaale ami-n ale

so-? 2p-pl.abs kill-for so tell-3abs one.says

‘What makes you so violent that you kill?’’ So he said, as reported.’


079 Ina hen damtsi laxmin ale hinak wanek,

ina.hen damtsi laxmi-n ale hinak-wan-ek

and.then follow towards.forest-3abs one.says 3p-pl-erg

‘Then they followed him into the forest, as reported,’

(damtsi: class 2, J. patient, abs; K.: agent, erg)

080 wan ka’chı pita hen chomtan le.

wan ka’chı pita hen chomta-n le

pl walk go.out then aim.at-3abs one.says

‘they left the canoe and walked and they aimed at him, as reported.’

(ka’chı: class 1, K.: abs / chomta: class 3, K.: agent, abs; no patient)

081 Chomtan ale la-t’aes hen chokun ale hen

chomta-n ale la-t’a-es hen choku-n ale hen

aim.at-3abs one.says stay-perf-loc1 then reach-3abs one.says then

ıwırki tlan

ıwır-ki tlan

wood-E2 only

‘They aimed at him where he was, as reported, but then they hit

only tree branches;’

(chomta: class 3, K.: agent, abs; no patient / choku: class 3,

K.: agent, abs; tree: patient, acc)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



082 tach chomtat’ kawan ale tach chomtat’

tach chomta-t’ kawan ale tach chomta-t’

more aim.at-perf let’s one.says more aim.at-perf

They aimed more and more, as reported. . .

(chomta: class 3, K.: agent)

Abbreviations

P Alternative interpretation

1 first person pronoun

2 second person pronoun

3 third person pronoun

3abs Extralocutive marker of the 3rd person absolutive (-e/-n)

abl Ablative

abs AbsolutiveP nominative

all Allative

amb Aurore Monod Becquelin

caus Causative

com Comitative

cont Continuative

cop Copula

E1 ‘‘Extension to core’’ (-(V )s)

E2 ‘‘Extension to core’’ (-ki)

E3 ‘‘Extension to core’’ (-(V )tl )

erg Ergative

excl Exclusive

foc Focalizer

foc/tens cf. RGD (2003)

imp Imperative

instr Instrumental

i.v. Intransitive verb

iyi Particle iyi

loc Spatial/temporal locative (-n)

loc1 Spatial/temporal locative (-(V )s)

loc2 Spatial/temporal locative (-ki)

loc3 Spatial/temporal locative (-(V )tl )

n Myth told by N., Trumai chief, shaman, polyglot and

famous singer

neg Negation

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



nom Nominative

obl Oblique

onom Onomatopoeia

perf Perfect participle

pl Plural

t.v. Transitive verb

Ua/p Unique argument of intransitive verb (either agent or patient

participant)

-(V) Epenthetic vowel

vol Volitional

w Myth told by W., great Wauja curer’s wife

References

AMB (List of texts collected by A. Monod Becquelin):‘‘Anu Bird’’, ‘‘Ear-piercing, ‘‘Fish festival’’, ‘‘Harpoon’’, ‘‘Karijawar’’, ‘‘Murder’’,

‘‘Myth AMB msc.’’, ‘‘Origin of the day’’, ‘‘Payetan’’, ‘‘Planta-tion’’, ‘‘RGD lexicon’’, ‘‘Sun and Moon’’, ‘‘Tapir’’

Basso, E.B.1973 The Kalapalo Indians of Central Brazil. New York: Holt, Rine-

hart and Winston, Inc.Becquey, Cedric

2007 Le systeme verbal du trumai (langue isolee du Xingu, Bresil):Description, reanalyse et formalisation dans le cadre HPSG.Master’s thesis, Universite Sorbonne nouvelle, Paris 3.

Comrie, Bernard1989 Language Universals and Linguistic Typology. Chicago: The

University of Chicago Press.Creissels, Denis

2004 ‘‘Ergativite / accusativite et l’heterogeneite des constructions in-transitives’’, Cercle linguistique de l’INALCO, Cours de syntaxegenerale 2004.

Creissels, Denis2006 ‘‘Syntaxe generale, une introduction typologique’’. Paris: Hermes

Lavoisier.Dixon, Robert M.W.

1994 Ergativity. Cambridge: Cambridge University Press.Dixon, Robert M.W. and Alexandra Y. Aikhenvald

2000 Changing Valency: Case Studies in Transitivity. Cambridge:Cambridge University Press.

Franchetto, Bruna and Michael J. Heckenberger2001 Os Povos do Alto Xingu: Historia e Cultura. Rio de Janeiro:

Editora UFRJ.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Friedmann, Naama, Gina Taranto, Lewis P. Shapiro and David Swinney2008 ‘‘The leaf fell (the leaf ): the online processing of unaccusatives.’’

Linguistic Inquiry 39(3): 355–377.Galvao, Eduardo

1953 ‘‘Cultura e sistema de parentesco das tribos do alto rio Xingu’’.Boletim do Museu Nacional, Nova Serie, Antropologia 14.

Givon, Talmy1984 Syntax: A Functional-Typological Introduction. Vol. 1. Amsterdam/

Philadelphia: Benjamins.Givon, Talmy

1990 Syntax: A Functional-Typological Introduction. Vol. 2. Amsterdam/Philadelphia: Benjamins.

Guirardello(-Damian), Raquel1999 A Reference Grammar of Trumai. Ph. D. diss., Rice University.

[20/10/2008] http://scholarship.rice.edu/handle/1911/19387>Guirardello(-Damian), Raquel

2002 ‘‘Caso e relacoes gramaticais em Trumai.’’ in Ergatividade naAmazonia I, F. Queixalos (ed.), Brasılia: Centre d’etudes des lan-gues indigenes d’Amerique, Laboratorio de Lınguas Indıgenas.[25/01/2010]<http://celia.cnrs.fr/FichExt/Documents%20de%20travail/Ergativite/1lGuirardello.pdf>

Guirardello(-Damian), Raquel2003 ‘‘Classes verbais e mudancas de valencia em Trumai.’’ In

Proceedings of the Workshop Ergatividade na Amazonia II, F.Queixalos (ed.), Centre d’etudes des langues indigenes d’Ameri-que (CNRS/IRD), Laboratorio de Lınguas Indıgenas (UnB).[25/01/2010] http://celia.cnrs.fr/FichExt/Documents%20de%20travail/Ergativite/2mGuirardello.pdf

Guirardello(-Damian), Raquel2004 ‘‘Coreference in Trumai.’’ In Proceedings of the Workshop Erga-

tividade na Amazonia III, F. Queixalos (ed.), Centre d’etudes deslangues indigenes d’Amerique, Laboratorio de Lınguas Indıgenas,Paris. [25/01/2010] http://celia.cnrs.fr/FichExt/Documents%20de%20travail/Ergativite/3mGuirardello.pdf

Guirardello(-Damian), Raquel2005 DoBeS, Max Planck Institut; [25/01/2010]

<http://www.mpi.nl/DOBES/projects/trumai/>Keenan, Edward L.

1976 ‘‘Towards a Universal Definition of ‘Subject’.’’ In Subject andTopic, Charles Li (ed.), 303–333. New York: Academic Press.

Lazard, Gilbert1994 L’actance. Paris: PUF.

Mel’cuk, Igor1983 ‘‘Grammatical Subject and the Problem of the Ergative Con-

struction in Lezgian.’’ In Proceedings of the Second Conference

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



on the Non-Slavic Languages of the USSR, Slavica, 1983. Vol. 5.246–293.

Malchukov, Andrej L.2005 ‘‘Case Pattern Splits, Verb types and Construction Competi-

tion.’’ In Competition and Variation in Natural Languages: TheCase of Case, Mengistu Amberger and Helen de Hoop (eds.),73–117. Oxford: Elsevier.

Monod Becquelin, Aurore1975 La pratique linguistique des Indiens Trumai (Haut Xingu, Mato

Grosso, Bresil) [Langues et civilisations a tradition orale, 9, 1].Paris: SELAF.

Monod Becquelin, Aurore1976 ‘‘Classes verbales et constructions ergatives en trumai.’’ Amerindia

1: 117–143.Monod Becquelin, Aurore

1977 ‘‘Les amants punis: conte trumai (Haut-Xingu, Bresil).’’ Amerindia2: 163–173.

Monod Becquelin, Aurore1987 ‘‘ ‘Les femmes sont un bien excellent’ ’’: vision des hommes, etre

des femmes dans le Haut Xingu.’’ Anthropologie et Societes(Quebec, Universite de Laval), XI(1) [numero special: Enjeux etcontraintes, discours et pratique des femmes]: 121–136.

Monod Becquelin, Aurore and Raquel Guirardello2001 ‘‘Historias trumai.’’ In Os Povos do Alto Xingu. Historia e Cultura,

Bruna Franchetto and Michael Heckenberger (eds.), 401–443.Monod Becquelin, Aurore, Raquel Guirardello Damian and Emmanuel de Vienne

2008 ‘‘Working together. The interface between researchers and nativepeople.’’ In Lessons from Documented Endangered Languages,K. David Harrison, David S. Rood and Arienne Dwyer (eds.),43–66. Amsterdam/Philadelphia: John Benjamins.

Murphy, Robert F. and Buell Quain1955 The Trumai Indians of Central Brazil. Locust Valley, N.Y., J.J.

Augustin.Steinen, Karl von den

1942 O Brasil Central: expedicao em 1884 para a exploracao dorio Xingu. Traducao. Sao Paulo: Companhia Editora Nacional(Brasiliana, serie extra [grande formato], 3).

Tsunoda, Tasaku1981 ‘‘Split case-marking patterns in verb-types and tense/aspect/mood.’’

Linguistics 19: 389–438.Tsunoda, Tasaku

1985 ‘‘Remarks on transitivity.’’ Journal of Linguistics 21: 385–396.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Ergativity in the Adyghe system of valency-changing derivations

Alexander Letuchiy, Higher School of Economics,Moscow

1. Introduction

In this paper I will analyse the syntactic properties of valency-changing

derivations and other syntactic processes in Adyghe (a language of the West

Caucasian family spoken in the Republic of Adygheya and the Krasnodar

region of Russia, and also in some countries of western Asia such as

Turkey). My aim is to determine whether these processes testify to syntactic

ergativity or accusativity in Adyghe, or whether they in fact shed no light

at all on the question of Adyghe alignment behaviour.

In traditional descriptions, such as Rogava & Keraseva (1966), Kumakhov

(1984), and Zekox (2002), it is taken for granted that Adyghe has ergative

alignment. This is due to the fact that Adyghe is a morphologically ergative

language (see below). As I will show, the case marking of verb arguments

and the system of cross-reference markers are indeed organized ergatively.

However, with the exceptions of Serdobol’skaya (2007) and Lander (2009),

scholars have not considered the syntactic aspects of ergativity in Adyghe.

In the present paper, I base my analysis of syntactic ergativity on the

evidence of valency-changing derivation only. I choose not to consider

other pivot properties related to ergativity / accusativity (coordination

reduction, relativization, subordinate clauses etc.; see Dixon 1994; Van

Valin and LaPolla 1997). It seems to me more justifiable to restrict myself

to the data presented by derivational behaviour alone, since in a single

article it is impossible to analyse the whole range of data related to erga-

tivity in a polysynthetic language like Adyghe; moreover, the valency-

changing derivational system may be organized ergatively, for example,

while other syntactic processes are organized accusatively, or vice versa.

We assume that voice systems and syntactic alignment are closely

related. On one hand, many linguists (e.g., Shibatani 1985, Dixon 1994)

claim that syntactic alignment is crucial for the the voice system of a

language. For instance, it has long been thought that the passive is not a

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 00:44) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 321–352 1351 Authier_10_letuchiy (p. 321)

characteristic feature of ergative languages. Though this formulation may

be too strong, it is at least true that the passive is more characteristic of

accusative than of ergative languages.

Moreover, the precise nature of ergativity in a given language can be

relevant for the description and classification of voice-like phenomena.

For instance, if we are able to prove that the agent argument of transitive

verbs has syntactic pivot properties (in other words, that the language is

syntactically accusative), and there is a syntactic alternation which de-

motes the agent to indirect object while leaving the absolutive argument

una¤ected, we can describe this alternation as a prototypical passive.

However, if the language is syntactically ergative (the absolutive argument

has subject properties with both transitive and intransitive verbs), the

same alternation should be described di¤erently: in this case it lacks the

main property of the passive, in that it does not decrease the syntactic

status of the subject argument. We will see that there is a problematic

case of this type in Adyghe: two of the derivations found in the language

(the potential, marked with fe-, and the inadvertitive (involitional),

marked with �ec˙

’e-) seem similar to the passive voice in syntactic terms.

However, this similarity disappears if we assume that Adyghe is a syntac-

tically ergative language.

On the other hand, the accurate description of valency-changing mech-

anisms is a prerequisite for the analysis of syntactic alignment. Some

voices and valency-changing derivations systematically show dependence

on the subject properties of arguments: for instance, in the majority of

languages, reflexives are controlled by the subject argument. Therefore, a

description of the voice system allows us to define whether the language is

syntactically accusative or ergative.

In Section 2, I will sketch the most important features of Adyghe

grammar, such as polysynthesis, pro-drop and morphological ergativity.

In what follows, each valency-changing derivation is considered. What is

crucial for the present analysis is that each derivation will be considered in

terms of its relation to syntactic alignment: whether the particular type

of syntactic alternation is more characteristic of ergative or accusative

languages, and whether or not it shows pivot properties of one argument

or another.

Note that a similar type of analysis has been carried out for Adyghe in

a previous paper (Paris 1987). The important di¤erence between this

earlier paper and the present article is that Paris adopts a semantic point

of view: she examines which mechanisms in Adyghe play the same func-

tional role as passives in European languages. My aim is to analyse the

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


322 Alexander Letuchiy

valency-changing devices of Adyghe not only from the semantic, but also,

and perhaps more importantly, from the syntactic point of view in order

to see which derivational mechanisms are syntactically similar to voice al-

ternations.

Matasovic (2008) proposes an analysis of transitivity in Kabardian, a

language of the West Caucasian family closely related to Adyghe. How-

ever, though his analysis is similar in many respects to the present work,

Matasovic takes into account only causative and antipassive derivations.

In Section 3, Adyghe valency-changing derivations and their ergative /

accusative behaviour are considered. Finally, in Section 4 I draw some

conclusions concerning the nature of ergativity in Adyghe.

Taking into account the fact that Adyghe is a polysynthetic language,

we use a non-canonical notion of valency-changing derivation throughout

this article. Inasmuch as NPs are not obligatory in Adyghe, whereas all

core arguments have to be cross-referenced in the verb, we will mostly

describe the impact of each derivation upon the expression of arguments

inside the verb form, i.e. upon the system of cross-reference prefixes,

although traditionally, as in (Comrie 1976), (Dixon 1994), (Plungian 2000),

(Testelets 2001), valency-changing derivations are often described in terms

of their impact upon the expression of free NP arguments.

Let us now review the notions of morphological and syntactic ergativity /

accusativity. A language is morphologically ergative (or accusative) if the

morphological coding of verbal arguments, i.e. case-marking and the

system of verbal personal prefixes, follows the ergative (or accusative)

strategy. For instance, if a language is morphologically ergative, the Actor /

Agent (A) of transitive verbs must be case-marked and cross-referenced on

the verb form in a di¤erent way from the Patient (P) of transitive verbs

and Single argument (S) of monovalent intransitive verbs. The Patient

and the Single argument must, in turn, receive the same case marker and

be cross-referenced in the same way.

The notion of syntactic ergativity / accusativity refers to the organization

of syntactic processes. If the language is syntactically ergative, this means

that S and P show the same set of syntactic properties, distinct from those

displayed by A in the same syntactic processes (for instance, only S and P,

but not A, can be relativized). It is usually required that in a syntactically

ergative language, S and P must show more syntactic subject properties

than A. In contrast, in a syntactically accusative language, S and A show

the same set of syntactic properties, distinct from the properties of P. S

and A, but not P, must show subject properties in syntactically accusative

languages.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The term ‘morphological ergativity’ is used in the standard way

throughout this paper: Adyghe is a morphologically ergative language

because its argument coding system follows the ergative strategy (a spe-

cific case marks the agent of transitive verbs,1 while another case form is

used both for direct objects and subjects of intransitive verbs). However,

I choose to understand the term ‘syntactic ergativity / accusativity’ in a

specialized sense: I will say that the given phenomenon displays syntactic

ergativity / accusativity if the absolutive argument / the agent of transitive

verbs is the pivot for the morphological process involved (and thus, for

instance, that morphological reciprocalization provides one piece of

evidence for syntactic ergativity because the absolutive argument is the

antecedent of morphological reciprocals). Although the valency-changing

derivations under analysis here occur inside the word, the tendencies

observed in the present paper are parallel to syntactic processes and

tendencies in many other, non-polysynthetic languages.

2. Essentials of Adyghe grammar

2.1. Polysynthesis and pro-drop

The most salient feature of Adyghe morphosyntax is polysynthesis. Adyghe

is a canonical polysynthetic language. The verb form can encompass a large

number of locative, temporal, and modal markers. All verb arguments are

cross-referenced on the verb form, which can also take a large number of

derivational a‰xes. The language therefore apparently represents the head-

marking type, in the terminology of (Nichols 1986). The following example

illustrates the richness of Adyghe verb forms:

(1) S�-q�-t-de-p-f�-r-a-‚a-‰Łe-s’t�-‚.

1sg.abs-dir-1pl.io-com-2sg.io-ben-3sg.io-3pl.a-caus-read:ap-aux-pst

‘They were making me read it to you together with us.’2

In (1), the verb contains two applicative markers ( f�- ‘benefactive’ and

de- ‘comitative’), the causative marker ‚e-, the complex imperfect marker

-s’t�-‚(e) (which includes auxiliary stem -s’t� and past marker -‚(e)),

1. Here and below I use the term ‘agent’ to denote the ergative (oblique) argu-ment of transitive verbs, including verbs like le‚W �n ‘see’ for which thisargument does not really have the semantic role of agent.

2. I am grateful to Yuri Lander for this example.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



and four personal cross-reference prefixes (s�- ‘1SG.ABS’, t- ‘1PL.IO’, p-

‘2SG.IO’, a- ‘3PL.A’).

The personal (and all other) markers are organized in slots: each

marker has its own position (slot) in the verb form. These positions are

not interchangeable, except in some very special cases. The general scheme

is very complicated, which is why we restrict ourselves to the prefixal zone

for present purposes. Following (Smeets 1984) we can represent it as

follows:

Table 1. Prefixal slots of the Adyghe verb form

-9 -8 -7 -6

absolutive argumentcross-reference marker

directiveprefix

temporalprefix

applicative prefixestogether with markersof oblique indirectapplicative object

-5 -4 -3 -2 -1 0

oblique indirectobject cross-reference marker

oblique agentcross-referencemarker

optative negation causative root

Henceforth we will refer to slot -9 as the ‘absolutive slot’, and slots -6,

-5 and -4 as ‘oblique slots’, sometimes specifying whether the given argu-

ment is an agent or an indirect object.

The personal markers introduced by applicative derivations are paired

with derivational markers: each personal prefix must occur immediately

before the marker by which it is introduced. All arguments introduced by

applicatives are indirect objects. For instance, in (1) the 2SG indirect

object prefix p- is before the benefactive marker fe- (‘for you’), whereas

the 1PL IO prefix t- is before the comitative marker de- (‘with us’).

It is also important to note that Adyghe is a pro-drop language. The

participants are not obligatorily expressed by separate referential phrases –

what is genuinely obligatory is their cross-referencing on the verb form

by means of personal prefixes. Any pronoun in an argument position

(absolutive argument, agent, indirect object) can be dropped.

2.2. Morphological ergativity

Adyghe is a morphologically ergative language – in other words, case-

marking and the system of verbal cross-reference markers are organized

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



ergatively. In the domain of case-marking, intransitive subjects and direct

objects can both be marked with the absolutive marker -r, whereas transi-

tive agents take the marker -m; the latter is traditionally called ergative

(see Kumakhov 1989; Kumakhov, Vamling 2006), but in what follows

we will label it ‘oblique’, with the gloss OBL, because the range of its func-

tions is not limited to canonical functions of the ergative.

Cross-reference prefixes also group S and P together, as distinct from

A: the former roles are indexed by a series of prefixes occupying the

leftmost position in the verb form. Third person singular absolutive argu-

ments are cross-referenced with a zero prefix, as psase-r ‘girl (absolutive)’

in (2) and (3). The agent argument is cross-referenced in the -4 position of

the verb form. Third person singular agents are cross-referenced with the

prefix �-.

(2) C˙

’ale-m psase-r Ø-�-le‚W�-‚.

boy-obl girl-abs 3sg.abs-3sg.a-see-pst

‘The boy saw the girl.’ (transitive verb)

(3) Psase-r Ø-ma-k˙

W e.

girl-abs 3sg.abs-dyn-go

‘The girl goes.’ (intransitive verb)

Almost all verbs have an absolutive argument. Two minor verb classes

constitute an exception: the so-called impersonal verbs, which have an

oblique argument only, and the facilitives / di‰cilitives, which are analysed

in detail in 4.5.

Notably, Adyghe possesses a class of bivalent intransitive verbs – a

rather atypical feature for morphologically ergative languages. These verbs

have a subject and an indirect object. The subject of intransitive verbs

takes absolutive marking, as is typical for morphologically ergative lan-

guages. The indirect object, just like the agent of transitive verbs, is

marked with oblique case and takes the su‰x -m:

(4) C˙

’ale-r psase-m Ø-Ø-je-bew�-‚.

boy-abs girl-obl 3sg.abs-3sg.io-oblique-kiss-pst

‘The boy (intransitive subject, absolutive) kissed the girl

(IO, oblique).’

In a sense, the syntactic type of the verbal construction represented in

(4) is the reverse of the canonically transitive one (cf. (2)) as regards case-

marking.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The verb form in (6) is intransitive because it does not bear the agentive

3SG prefix �- (cf. (2) with a transitive verb). I will show below that reflex-

ivization and reciprocalization also distinguish between bivalent intransi-

tive and bivalent transitive verbs.

The verbal marker je- glossed in (4) and below as ‘OBLIQUE’ is not

a co-reference marker. It is a special type of applicative prefix: for many

bivalent intransitive verbs, the indirect object is introduced by means of

this prefix, though the precise semantics of this marker is very vague – in

other words, its function seems to be just to add an indirect object.

The oblique case has a wide range of semantic functions. It marks not

only agents of transitive verbs and indirect objects, but also complements

of postpositions, possessors, and temporal and locative adjuncts:

(5) C˙

’ale-m paje

boy-obl for

‘for the boy (complement of postposition, oblique)’

(6) c˙�f�-m j�-cet�w

man-obl 3sg.prþposs-cat

‘The man’s (possessor, oblique) cat’

Finally, both the oblique and the absolutive case markers can be dropped

when an argument is non-specific or indefinite, as in cet�w qe-KWa-R

‘a cat came’ (cat dir-go-pst) where the absolutive ending is dropped, or

c˙

’ale cet�w �-le‚W�‚ ‘a boy saw a cat’ (boy cat 3sg.a-see-pst) where

the oblique ending on c˙

’ale ‘boy’ and the absolutive ending on cet�w ‘cat’

are both omitted. They thus encode not only the syntactic position of

the noun, but also its value in terms of definiteness. This is not typical of

morphologically ergative languages.3

As we have seen, all core arguments, including oblique-marked agents

and indirect objects as well as absolutive-marked direct objects, are cross-

referenced on the verb form. The system of personal prefixes also follows

ergative alignment: intransitive S and the DO of transitive verbs are cross-

referenced with the same set of prefixes occupying the first position in the

verb form. The 3SG absolutive argument is cross-referenced with a zero

prefix (or, alternatively, is simply not cross-referenced on the verb form

3. Of course, the core cases (chiefly the absolutive) can be expressed with zeroendings (for instance, the absolutive in Tsakhur). However, if a case formbears non-zero marking, this marker usually cannot be omitted (see Dixon1994 and Kibrik 2002 for examples).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



at all). The 3PL absolutive argument does not control an agreement prefix

but is cross-referenced with the (optional) absolutive plural su‰x -x.

Transitive agents are cross-referenced by means of a di¤erent set of pre-

fixes from indirect objects and absolutive direct objects: agent markers are

situated in the -4 slot in the terms of Smeets 1984. Only agent prefixes

have the 3SG form �- / j�-.

We are now in a position to analyse the system of valency-changing

derivations in Adyghe. We will not analyse each derivation in detail. Our

purpose will be to say a few words about each derivation and to determine

whether a given derivation can be regarded as a manifestation of semantic /

syntactic ergativity, or whether, by contrast, it represents an argument

against the analysis of Adyghe as an ergative language.

3. Adyghe derivations

The system of valency-changing derivations in Adyghe includes both means

of valency and transitivity increase (causative, benefactive, malefactive,

locative) and means of valency and transitivity decrease (potential, anti-

passive, facilitive and di‰cilitive). As I have mentioned, although not

only derivations, but also some other syntactic processes can be useful as

diagnostics of ergativity, in this paper I consider valency-changing deriva-

tions only.

Let me begin by noting that the causative formation will be excluded

from the following analysis. The causative in Adyghe corresponds to

what is often called the ‘paradigm case’ of the formation, already detailed

in Comrie (1976):

The causative derivation introduces a new argument (causer) which

becomes a syntactic subject. The subject of the base non-causative verb

(causee) occupies the highest vacant position in the hierarchy: S > DO >IO > Oblique objects.

This rule is not related to ergativity and cannot be used as a diagnostic

for syntactic alignment.

3.1. Antipassive

Like many ergative languages (Alutor, Mayan languages, Dyirbal etc.),

Adyghe has an antipassive construction (see Arkadiev & Letuchiy 2008

for details). Before describing it, we need to discuss the definition of the

antipassive.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



For Cooreman (1994), the antipassive is a voice alternation which

operates on transitive verbs and decreases the syntactic status of the initial

direct object. The object is either eliminated and cannot be expressed or is

demoted to an indirect / oblique object, the verb thus becoming intransi-

tive. The subject (agent) of the initial transitive verb becomes the subject

of a derived intransitive verb: in ergative languages this leads to a change

in case-marking, as the agent which was marked with the ergative in the

base construction receives accusative marking in the new antipassive con-

struction. Cooreman (1994), Testelets (2001), Say (2007) and others claim

that one possible motivation for the use of the antipassive is to demote a

non-salient or indefinite patient which occupies the privileged direct object

position of the base verb.

Plungian (2000) distinguishes between two very similar operations which

are not always easy to tell apart in a given language. The first is the anti-

passive proper – a voice which decreases the syntactic status of the direct

object, making the verb a bivalent intransitive. The second is the object

impersonal – this transformation eliminates the object, making the verb

monovalent. In Cooreman’s (1994) sample, there are many examples where

the agent can be either expressed or left unexpressed: thus, Plungian’s dis-

tinction is not irreproachable for all languages. However, it is useful for

Adyghe: some morphological antipassives admit expression of the initial

direct object as an oblique indirect object triggering indirect object agree-

ment on the verb, but others do not admit the initial DO expression, or

permit it to be expressed only as a non-argument NP.

In Plungian’s terms, most Adyghe verbs form the object impersonal,

but not the antipassive proper. This means that most verbs are morpho-

logically monovalent in their antipassive form: thus, the initial direct

object does not control any cross-referencing prefixes. The verb sxen ‘eat

(antipassive)’ is morphologically monovalent: the initial direct object

can only be marked with the instrumental, as in (7b), and instrumental-

marked NPs do not trigger verbal cross-referencing prefixes. The variant

sx�n in (7a) is a bivalent transitive.

(7) a. Se l� Ø-s-e-sx�.

i meat 3sg.abs-1sg.a-dyn-eat.tr

‘I eat meat.’

b. Se l�-c˙

’e Ø-s-e-sxe.

i meat-ins 3sg.abs-1sg.a-dyn-eat.ap

‘I eat meat.’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



With some verbs, the object impersonal is marked by means of a stem

vowel change from -� to -e:

(8) a. Se pj�s’me Ø-s-e-tx�.

i letter 3sg.abs-1sg.a-dyn-write.tr

‘I write a letter.’

b. Se s-e-txe.

I 1sg.s-dyn-write.ap

‘I write.’ (Arkadiev & Letuchiy 2008)

Although pairs like (8a) vs. (8b) can be analysed as two (equally basic)

morphological variants, I find it more plausible to regard (8b) as derived

from (8a): there are transitive verbs which do not form variants of the (8b)

type, but there are no intransitive verbs with a generic or indefinite object

which do not have a transitive variant of the type seen in (8a).

This group includes the following lexemes (throughout this article, all

verbs will be cited in the masdar (verbal noun) form with the su‰x -n):

(9) tx�n ‘write’ w�qebz�n ‘clean’

sx�n ‘eat’ pw�n ‘sow’

thac˙

’�n ‘wash’ gW�n ‘pound’

ha‰Ł�n

etc.

‘mill’ t�n ‘sew’

Some verbs use Agent-preserving lability to express the same pair of

meanings:

(10) a. Xate-r Ø-�-pc˙

’a-‚.

garden-abs 3sg.abs-3sg.a-weed-pst

‘He weeded the garden.’

b. C˙

�f�-r mefe.rjen�-m Ø-pc˙

’a-‚e.

man-abs day.whole-obl 3sg.abs-weed-pst

‘The man weeded all day.’ (Arkadiev & Letuchiy 2008)

In general, lability is very frequent in Adyghe4 (see Gisev 1968, Hewitt

1982, Letuchiy 2009b). In (11), the two verb forms di¤er only as regards

the presence / absence of the agentive 3SG prefix �-.

4. I do not take into account P-lability, in the terminology of Dixon (1994),which is taken to be a voice-like mechanism by Paris (1987). In Letuchiy(2006), I show that the existence of P-lability does not characterize a languageas ergative or accusative.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The list of Agent-preserving labile verbs includes, for instance:

(11) pc˙

’en ‘weed’ zW en ‘plough’

w�pc˙

’en ‘mow’ p�w�pc˙

’�n ‘chop’

s’en ‘sell’ c˙

’ec˙

�n ‘lay (eggs)’

As we can see, in (7b), (8b) and (10b) the verb is morphologically mono-

valent, although in (7b) there is a peripheral object which is not cross-

referenced in the verb form. Thus, the second argument cannot be

expressed in the verb form; for the verbs listed so far, Adyghe does not

have constructions equivalent to ‘I plough at the field’ or ‘I write at a

letter’ in which the second argument controls an indirect object slot.

However, two verbs ( je‰Ł ’en ‘read (intransitive)’ / ‰Ł�n ‘read (transitive)’

and jesW en ‘drink (intransitive)’ / j�sW �n ‘drink (transitive)’) retain a

bivalent structure in the antipassive form:

(12) a. C˙

�f�-m tx�l�-xe-r Ø-�-‰Ł�-‚.

man-obl book-pl-abs 3sg.abs-3sg.a-read-pst

‘A man read the book through’.

b. Se tx�l�-m s-Ø-je-‰Ła-‚.

i(abs) book-obl 1sg.abs-3sg.io-obl-read.ap-pst

‘I read a book (for some time)’. (Arkadiev & Letuchiy 2008)

The verb has two agreement slots in both (12a) and (12b). In (12a) we

observe transitive morphology: one of the a‰xes occupies the agent slot,

and the other is found in the direct object slot. In (12b), one of the a‰xes

(s-) is in the absolutive slot, while the zero a‰x marks the indirect object;

thus, the construction is literally ‘I read at / on a book’ (the verb has the

same structure as seen in (4), where the strictly intransitive verb bew�n

‘kiss’ has a subject and an indirect object).

It is not clear why the prototypical bivalent antipassive is compatible

with these two verbs only. For instance, the verbs ‘eat’ and ‘write’, which

are semantically close to ‘drink’ and ‘read’ respectively, form object

impersonals and not antipassives. Furthermore, it must be emphasized

that for ‘read’ and ‘drink’, the antipassive variant is more frequent than

its transitive counterpart: the transitive variants ‰Ł�n ‘read through’ and

j�sW �n ‘drink up’ are used relatively rarely. Moreover, the transitive

variants of ‘drink’ and ‘read’ have a narrow meaning: they express that

the situation is either finished or is to finish soon after the moment of

speech. This is not obligatorily the case with all other transitive variants

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



of antipassive verbs: for instance, sx�n ‘eat’ and tx�n ‘eat’ do not neces-

sarily imply that the situation is or will soon be finished (see Arkadiev &

Letuchiy 2008 for details). In order to state simply that someone is drink-

ing tea or reading a book, the intransitive variant will be used (this fact is

in accordance with Hopper and Thompson (1980), where the authors

show that perfective situations are semantically more transitive than those

which are ongoing).

In principle, the existence of the antipassive could be regarded as an

argument for syntactic ergativity (see, for instance, Cooreman 1994,

Shibatani 1985). This valency-changing derivation is more productive in

syntactically ergative languages like Dyirbal than in semantically and

morphologically ergative languages such as Nakh-Daghestanian. How-

ever, this argument is weakened by the fact that the antipassive is not at

all productive in Adyghe. A canonical ergative language should have a

productive antipassive which is able to detransitivize most transitive verbs

when the need is felt to decrease the status of the absolutive direct object

or to eliminate it altogether.

Moreover, the same stem alternation -� / -e can also mark other valency-

changing derivations which are only indirectly related to antipassive func-

tion, as in bew�n ‘kiss sb.’ vs. bewen ‘kiss (in general)’:

(13) a. C˙

’ale-r psase-m Ø-Ø-je-bew�-‚.

boy-abs girl-obl 3sg.abs-3sg.io-oblique-kiss-pst

‘The boy kissed the girl.’

b. Bewe-n-�r j�-c˙

’as.

kiss-msd-abs 3sg.prþposs-love

‘He likes kissing.’ (lit. ‘To kiss is his love’). (Txarkaxo 1991: 41)

This class also includes w�nc˙

’�n ‘push sb.’ vs. w�nc˙

’en ‘push’, and pl�n

‘look at sth.’ vs. plen ‘look (in a particular direction)’. In this case, both

forms are intransitive. The �-variant is a bivalent intransitive verb. For

instance, in (13a) the fact that the verb has two arguments is evident

from the presence of the oblique indirect object psase-m and the oblique

argument prefix je-. The e-variant is also intransitive, but monovalent.

For instance, in (13b), the verb contains no oblique argument marker. In

examples like (13), -e marks not a canonical antipassive, but the elimina-

tion of the indirect object. Therefore, the motivation proposed by Cooreman

(1994), Testelets (2001) and Say (2007) for the antipassive among other

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



functions (demotion of the direct object, the most syntactically privileged

argument) is not applicable to (13), where the indirect object is eliminated.5

Thus, in most cases the stem alternation -� / -e eliminates an argument,

but this is not obligatorily a direct object. This feature shows that the anti-

passive construction in Adyghe cannot serve as evidence for syntactic

ergativity. While in many languages the antipassive formation shows the

privileged status of the absolutive argument, the antipassive in Adyghe

can eliminate di¤erent types of objects, and is not exclusively related to

the status of the absolutive argument.6

3.2. Passive-like valency-changing operations

Adyghe does not have a passive in the strict sense of the word. Following

Shibatani (1985), and the definition proposed by Aikhenvald and Dixon

(2000), I take prototypical passivization to be a voice alternation which

demotes the initial subject and often, though not necessarily, raises the ini-

tial direct object to the subject position. Passive is considered to be a voice

and not a valency-changing derivation since it does not change either the

number of arguments or their semantic properties. Passivization is sup-

posed to be a ‘syntactico-pragmatic’ change: simplifying somewhat, it can

be said that this syntactic process reflects the fact that the patient is more

pragmatically salient than the agent. This is the rationale for the use of the

passive construction, and, thus, the patient occupies the subject position.

According to Dixon (1994), syntactically ergative languages are often

characterized by the lack of a passive. Passive constructions seem to be

more widespread in accusative languages, where they serve to decrease

the status of the most salient participant (¼the agent argument of transi-

tive verbs). See, however, several papers in this volume where it is demon-

strated that passivization is not as uncommon in ergative languages as it is

often considered to be.

The Adyghe system includes some derivations which can be considered

passive-like in nature. They are semantically distinct from the passive:

unlike the passive, they add new components to the semantics of the

situation. However, syntactically they are very similar to passives in the

5. Of course, as we have already said, the morphological nature of the anti-passive marker makes it possible to consider examples like (12) to involvepairs of verbs neither of which is genuinely derived from the other.

6. Matasovic (2008: 62), arguing that the antipassive in Kabardian is not atypical antipassive, notes also that many e-stems, as well as �-stems, do nothave an �- or e-variant respectively.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



‘Standard Average European’ accusative languages, in that they demote

the initial subject to a non-subject argument (in Adyghe, it becomes an

oblique indirect object).

3.2.1. Potential and inadvertitive

Two valency-changing operations – the potential and the inadvertitive

(referring to involuntary action) – are syntactically similar to a canonical

passive: they decrease the status of the agent argument, making it an

indirect object (Figure 1).7

I call ‘inadvertitive’ or ‘involitional’ the operation marked with the

prefix �ec˙

’e-, which transforms the oblique-marked agent of the basic tran-

sitive verb into an indirect object, while the initial direct object retains its

absolutive marking but presumably changes its syntactic status to that of

subject. The semantic e¤ect of this derivation is the following: the derived

form means that the initial agent carries out the action involuntarily.

The potential derivation is marked with the prefix fe-.8 Its syntactic

e¤ect is the same as that of the inadvertitive. The meaning of the derived

form is that the initial agent is capable of carrying out the action denoted

by the base verb.

As we can see, the syntactic e¤ect of these derivations is equivalent to

that of passivization in SAE languages, except that the absolutive argu-

ment, which is usually patientive, and the oblique argument (the agent)

retain their case-marking (because both the direct object and the intransi-

tive subject are marked with absolutive case in Adyghe, while both the

agent and the indirect object receive oblique marking). What changes is

only the structure of the verb form and the syntactic properties of the

arguments (see below on the reciprocalization of potential derivatives).

Figure 1. Syntactic e¤ect of potential and inadvertitive

7. Here I illustrate with examples only the potential derivation; the inadvertitive(involitional) derivation, expressed with the prefix �ec

˙e-, seems to be rather

rare and to have additional syntactic properties which lie outside the scope ofthe present article.

8. The same prefix also expresses benefactive meaning.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(14) a. c˙

’ale-xe-m bukva-xe-r Ø-a-le‚W �-xe-re-p.

boy-pl-obl letter-pl-abs 3sg.abs-3pl.a-see-pl(abs)-dyn-neg

‘The boys do not see the letters.’

b. c˙

’ale-xe-m bukva-xe-r Ø-a-fe-le‚W �-xe-re-p.

boy-pl-obl letter-pl-abs 3sg.abs-3pl.io-ben-see-pl(abs)-dyn-neg

‘The boys cannot see the letters.’

The transitive verb le‚W �n ‘see’ (14a) has an oblique agent argument

(subject), cross-referenced with the agent prefix a-, and an absolutive

object, cross-referenced with the zero third person prefix and controlling

the absolutive plural su‰x -xe. In (14b), the marking of the patient does

not change; however, the oblique prefix a- now marks an indirect object,

an argument of the potential / benefactive prefix fe-. This is evident from

the fact that a- occupies the position immediately before fe-: as was

mentioned above, the position immediately before the applicative prefix

is reserved for an IO introduced by the applicative derivation. Interest-

ingly, the agent NP does not change its case-marking, since both agents

and indirect objects take oblique marking in Adyghe.

That the verb in (14b) is indeed intransitive follows from two con-

siderations: first of all, the oblique prefix a- cannot be an agent marker,

since it depends on the potential prefix. As I have shown, all arguments

introduced by the applicative derivations are indirect objects – thus, the

verb in (14b) does not have an agent prefix and cannot be transitive.

Second, potential derivatives cannot form reciprocals with zere-, while

this is always possible for transitive verbs independently of their semantics

(see Section 3.4).

Crucially for our analysis, the potential and involitional derivation are

compatible only with transitive verbs. Neither monovalent nor bivalent

intransitive verbs can form derivatives as in (14b) (see (14c) for a monova-

lent verb and (14d) for a bivalent intransitive):

(14) c. *c˙

’ale-xe-m a-fe-k�We-re-p.

boy-pl-obl 3pl.io-ben-go-dyn-neg

‘The boys cannot go.’

d. *c˙

’ale-xe-m psase-xe-m a-fe-bew�-xe-re-p.

boy-pl-obl letter-pl-abs 3sg.io-ben-see-pl(abs)-dyn-neg

‘The boys cannot kiss the girls.’

This restriction does not seem to be related to the nature of potential or

involitional meaning. Sentences like ‘I cannot go’ (monovalent intransitive

verb) or ‘I cannot kiss a girl’ (bivalent intransitive verb), which are incom-

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



patible with the potential construction under analysis, are semantically un-

objectionable, their English translations being fully grammatical.

We could in principle say that this is a manifestation of syntactic accu-

sativity. If the oblique agent is the most syntactically privileged argument,

the function of passive-like derivations can be to decrease its status: thus,

they are compatible only with verbs which have an agent argument,

i.e. transitive verbs. However, this explanation is valid only for passives

proper. The potential in Adyghe has a semantic function, distinct from

demotion of the agent argument: it expresses the modal meaning of possi-

bility. Thus, there is no natural reason for the valency-changing mechanism

to be incompatible with bivalent intransitive verbs possessing an agentive

subject: why should it be impossible to form a sentence like ‘The boys

cannot kiss the girls’, as in (14d)9?

In our opinion, the situation is just the reverse: the incompatibility of

the potential with intransitive verbs is an argument for syntactic ergativity –

in other words, it confirms that the absolutive argument of transitive verbs

is the most syntactically privileged argument (the subject, the pivot of the

sentence). Note that a construction of the type (14d) ‘The boys cannot kiss

the girls’ would require a change of case-marking. The absolutive subject

‘the boys’ would become the indirect object of the potential prefix,

whereas the patient which occupies the position of an oblique-marked

indirect object of the base verb (‘girls’) would take on absolutive case

(because almost10 no verb can exist without an absolutive argument):

Figure 2. Syntactic e¤ect of potential and inadvertitive with intransitive verb(hypothesis)

9. The question why structures like (14c) are impossible is simpler to answer: ifa monovalent verb were to undergo a potential derivation, it would lose itsabsolutive argument. However, any Adyghe verb, except those belonging toa very small class (see note 10), must have an absolutive argument.

10. I do not mention here the small class of so-called bezlicnyje (subjectless) verbsdescribed by Kumakhov (1984) et al., which have only an oblique and noabsolutive argument.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



The Adyghe derivational system in fact obeys a general restriction:

no derivation can a¤ect the absolutive argument. For instance, all valency-

increasing derivations add oblique arguments only (Adyghe does not have

a canonical applicative which introduces a new direct object). Moreover,

the antipassive which eliminates the absolutive argument is not highly

productive and is marked with a stem alternation rather than simply a

grammatical marker.

3.2.2. Quasi-passive (resultative)

Adyghe has two productive resultatives, one of them unmarked, and the

other marked with the prefix zere-. We do not consider the zere-resultative

here.

The unmarked resultative in Adyghe functions syntactically as a quasi-

passive construction: in the resultative construction based on transitive

verbs, as in (15b), the agent cannot be expressed. The structure of exam-

ples of the same construction based on intransitive verbs, as in (16),

remains less clear.

The unmarked resultative eliminates the oblique agent argument and

the agent cross-reference prefix, as in (15b):

(15) a. Se qebaske Ø-z-‚e-?W a-‚.

i.obl cabbage 3sg.abs-1sg.a-caus-cook-pst

‘I cooked cabbage.’

b. Qebaske Ø-‚e-?W a-‚e.

cabbage 3sg.abs-caus-cook-pst

‘The cabbage is cooked.’ (Nikolaeva 2003)

(16) a. Ha-r l˙�-me Ø-ja-ceqa-‚.

dog-obl man-obl.pl 3sg.abs-3pl.ioþoblique-bite-pst

‘The dog bit the men.’

b. l˙�-me Ø-q-ja-ceqa-‚.

man-obl.pl 3sg.abs-dir-3pl.ioþoblique-bite-pst

‘The men are bitten / have been bitten.’ (Nikolaeva 2003)

In (15a), the causative verb ‚e-?W en ‘cook’ is transitive: it bears the

agentive cross-reference marker z-. In the resultative construction in

(15b), the agent prefix and the agent NP are eliminated. The verb has

only a patient absolutive argument, which is cross-referenced with a zero

prefix.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In example (16), the verb ceqen ‘bite’ is intransitive: this is evident from

the fact that the third person singular agent cross-reference marker �- is

absent. The agent (dog) takes the absolutive case, whereas the patient

(men) is in the oblique form.

Despite the existence of examples like (16b), where the resultative con-

struction is seemingly built on the basis of intransitive verbs, we cannot

claim that this construction has the same morphological structure as that

in (15b). It would be temping to say that (16b) is an example of resultative

formation from an intransitive verb: the absolutive subject of the bivalent

intransitive verb ceqen ‘bite’ is eliminated. However, let us recall that the

3SG absolutive argument is cross-referenced with the zero prefix. Thus, it

is impossible to prove that the absolutive argument is eliminated in (16b)

in the same sense as the agent in (15b): we cannot say whether the zero

prefix is present or absent. Therefore, up to this point we do not have

any evidence that resultatives are formed in the same way from both

intransitive and transitive verbs.

The crucial question is whether the presence of a resultative construc-

tion represents an argument for or against syntactic ergativity in Adyghe.

Although passivization seems to be characteristic primarily of accusative

languages, it should be borne in mind that resultative formation is di¤er-

ent from passivization. It is generally accepted that the resultative has a

primarily aspectual function. Thus, we should not take (15) and (16) as

evidence that Adyghe has a passive proper (a feature which is not charac-

teristic of syntactically ergative languages).

The same line of argumentation is plausible for the passive-like potential

derivation in (14): the functional motivation for the potential is di¤erent

from that for a canonical passive. In (14), we are not dealing with a prag-

matically motivated agent demotion (as in the case of passivization), but

with a semantically motivated demotion. The agent is demoted because it

is no longer a prototypical agent: (14b) does not refer to the fact of bring-

ing about a transitive situation, but only to the potential to do so. Thus, in

the terms of Hopper and Thompson (1980), (14b) is less semantically tran-

sitive than (14a), because the situation is unreal, and the agent in (14b) is

less agentive.

Thus, the potential and inadvertitive derivations are syntactically simi-

lar to passives. They demote the agent, which is often the syntactic subject

in morphologically ergative languages11.

11. However, below I will show that some criteria allow us to propose that thesubject of Adyghe transitive verbs is the absolutive argument rather than theagent.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



However, all passive-like derivations in Adyghe have a semantic moti-

vation distinct from the motivation for canonical passives in accusative

languages. In contrast, the use of the antipassive in Adyghe is motivated

by principles relevant for canonical antipassives in ergative languages: for

instance, the antipassive pattern is used when the direct object is indefinite

or is not important for the speaker. Therefore, antipassives and ‘passives’

do not occupy the same place in the Adyghe derivational system: only

antipassive is a canonical voice. Thus, the evidence for syntactic ergativity

(the antipassive) is more important than the arguments for syntactic accu-

sativity (the passive-like derivations), since there is no canonical passive in

Adyghe.

3.3. Reflexive and reciprocal: semantic or syntactic motivation?

Reflexive and reciprocal are traditionally viewed as valency-rearranging

derivations. They do not change the number of arguments, but introduce

the requirement that the arguments display co-reference to each other.

In fact, Adyghe reflexives and reciprocals are not derivations in the

proper sense of the term. That is, markers of reflexivization in Adyghe

should be described as personal markers, rather than derivational markers.

In accusative languages which seem to represent the Standard Average

European type (SAE), there are cases when morphological reflexivization

is genuinely derivational: for instance, it often changes the inflectional

type of the verb:

French:

(17) a. J’ai lave la vaisselle.

b. Je me suis lave.

The verb laver ‘wash’ in (17a) is transitive. The reflexive marker se

in (17b) changes its syntactic characteristics: now the verb takes the auxil-

iary etre ‘be’ in the past, which signals that it has become syntactically

intransitive – all French transitive verbs use another auxiliary, avoir, in

the formation of the complex past (17a). We find no e¤ects of this type in

Adyghe.

In Adyghe, the reflexive and reciprocal morphemes represent personal

cross-reference markers rather than derivational markers (see Rogava &

Keraseva 1966: 265–269, 271–276, Smeets 1992: 115–117, Letuchiy 2007

for details), but with the additional requirement that the argument cross-

referenced with one of these markers must be co-referent with another

argument. The reflexive / reciprocal marker always occupies the slot

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



corresponding to one of the co-referent arguments (the -9 absolutive slot,

the -6 or -5 oblique IO slot or the -4 agent slot):

(18) a. W�-s�-w�ps�-‚.

2sg.abs-1sg.a-shave-pst

‘I shave you.’

b. Z�-s�-w�ps�-‚.

refl.abs-1sg.a-shave-pst

‘I shave (myself ).’

In (18a), the verb w�ps�n ‘shave’ bears two personal markers: the

agentive prefix s�- (1sg) and the absolutive prefix w�- (2sg). In (18b), the

reflexive marker occupies the absolutive slot. The verb form in (18b)

di¤ers from the form in (18a) only in that the absolutive slot is occupied

by a special ‘reflexive marker’ – the derivation does not change the transi-

tivity or the number of arguments of the base verb. In this sense the

Adyghe reflexives and reciprocals are analogous to reflexive pronouns,

such as sich in German or pozyn ‘oneself ’ in Khakas (Turkic, author’s

field data), which also ‘substitute’ for one of the co-referent arguments.

In French, the situation is roughly the same, except that the addition of

se changes an important morphosyntactic property of the verb, namely

its complex past formation.

In what follows, the glosses for reflexive and reciprocal markers in-

clude the designation of the slot occupied by the marker. For instance,

REFL.ABS in (18) means that the reflexive marker z�- is in the slot of

the absolutive argument (-9).

It may seem that the position of the markers under analysis can help us

to ascertain which argument of the base verb is the subject and which

is the object. Indeed, in many languages (such as the East Caucasian

languages and most languages of Europe) reflexives are subject-oriented:

they are bound by the syntactic subject and are never found in subject

position (for instance, the Russian sebja ‘oneself ’ does not have a nomina-

tive form even theoretically).

However, this prediction is not entirely borne out. First of all, with

transitive verbs it is notable that reflexives and reciprocals behave di¤er-

ently from each other. The complex reciprocal marker zere- occupies the

agentive oblique slot -4 (see also Rogava & Keraseva 1966, Letuchiy

2007), whereas the reflexive marker z�- is in the absolutive (direct object)

slot -9:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Reciprocal:

(19) Zec˙

’e c˙

�f-xe-r Ø-zere-le‚W �-z’�-x.

all man-pl-abs 3sg.abs-rec.a-see-re-pl.abs

‘All the people see each other (i.e. meet up).’

Reflexive:

(20) Zec˙

’e c˙

�f-xe-r z-a-le‚W �-z’�.

all man-pl-abs refl.abs-3pl.a-see-re

‘All the people see themselves.’

The picture which emerges in the domain of reflexivity and reciprocity

is shown in Table 2. It reflects the position and the form of both markers

for the transitive and the intransitive class of verbs. For each marker, the

slot it occupies is specified.

Table 2. Positions of reflexive and reciprocal markers in transitive andintransitive verbs

Transitive Intransitive

Reflexive z�-, absolutive slot12 ze- / z�-þ preverb of oblique argument,oblique (IO) slot

Reciprocal zere-, oblique slotze- / z�-þ preverb of oblique argument,oblique (IO) slot

It may seem that in Adyghe both reflexive and reciprocal marking (the

marker zere- aside) are organized as in semantically and morphologically

ergative languages such as the East Caucasian languages. In these, reflexives

are always bound by the agent of the transitive verb and the absolutive

argument of the intransitive verb. If we adopt the hypothesis that the

reflexive marker is always controlled by the syntactic pivot / subject, as is

the case in most of the world’s languages, this means that some syntactic

subject properties characterize the oblique agent of transitive verbs and

the absolutive argument of intransitive verbs. This is the case in many

12. I do not consider here the distribution of ze- vs. z�-, which seems to be purelymorphophonemic, and not directly related to the syntactic status of the verbarguments, according to (Arkadiev and Testelets 2009). However, it may benoted that z�- mostly cross-references absolutive arguments, and ze- is usedin most examples for oblique arguments (see also (21) vs. (22) for the di¤erencebetween reciprocal and reflexive forms).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



other Caucasian languages where the ergative argument (agent) of transi-

tive verbs and the absolutive argument of intransitive verbs possess most

subject properties (see Testelets and Toldova 1998).

However, this explanation is problematic for some minor verb classes,

such as inverse verbs13, the verb zexex�n ‘hear’ and potential derivatives

in fe-. For all of these the position of ze- / z�- is defined in a more complex

way than in transitive and intransitive verbs.

First of all, in inverse verbs the reflexive and reciprocal markers can

occupy either the absolutive slot or one of the oblique slots. In other

words, both reflexive and reciprocal forms can have ze- / z�- either in the

absolutive slot or in the oblique.

(21) T�-z�-s’�-gW �psa-‚ / z�-t-s’�-gW �psa-‚.

1pl.abs-refl.io-loc-forget-pst refl.abs-1pl.io-loc-forget-pst

‘We forgot ourselves.’

(22) A-xe-m ze-s’�-gW �pse-z’�-‚e-x /

that-pl-ob rec.io-loc-forget-re-past-pl.abs

z-a-s’�-gW �pse-z’�-‚.

rec.abs-3pl.io-loc-forget-re-past

‘They forgot each other.’

The distribution of ze- vs. z�- is sensitive to morphophonological con-

ditioning (see note 12). At the same time, as the comparison of the first

variants in (21) vs. (22) shows, reflexives tend to choose z�- even in some

cases where reciprocals choose ze- (for instance, when cross-referencing an

indirect object).

Second, in derivatives with potential meaning ze- / z�- can only be found

in the oblique slot when used reciprocally, but only in the absolutive slot

when used reflexively:

(23) S-�e me-w�z�-s z�-s-fe-w�ps�-s’t-ep.

1sg-hand dyn-hurt-conv refl.abs-1sg.io-ben-shave-fut-neg

‘My hand hurts, so I cannot shave (myself ).’

13. I use this traditional term of Caucasian linguistics (see, for instance, Rogava& Keraseva 1966 and Kumakhov 1984) for bivalent emotional and mentalverbs such as s’�gW �psen ‘forget’, whose absolutive argument is the stimulusand the oblique argument is the experiencer (the sentence ‘I forgot you’ isliterally translated into Adyghe as ‘You were forgotten to me’).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



This variation in behaviour prevents us from considering reflexivization

to be a syntactic test. What is relevant for these derivations is the agentivity

of the arguments involved. The rule for reflexivization can be stated as

follows: the reflexive marker always occupies the slot of the least agentive

of the co-referent arguments (this explanation was first proposed in Smeets

1992 for the Shapsug dialect of Adyghe). The most agentive argument

is the oblique argument of transitive and potential derivatives, and the

absolutive argument of intransitive verbs.

Note that the behaviour of s’�gW �psen ‘forget’ in example (21) con-

firms the semantic explanation. Situations like ‘forget’ involve neither

an agent nor a patient; both arguments (stimulus and experiencer) have

properties intermediate between those of agent and patient. For instance,

neither acts volitionally (like an agent), and neither is a¤ected by the situa-

tion (like a patient). This is why the position of the reflexive marker is

variable in inverse verbs.

Therefore, reflexives cannot give us any information on ergativity in

Adyghe. The choice of the argument which controls the reflexive marker

is motivated semantically and not syntactically: it is the most agentive

argument, whether this is marked with absolutive or oblique case14.

By contrast, the reciprocal marker is much more informative from our

point of view. The Adyghe reciprocal is clearly absolutive-oriented, as is

evident from four facts:

e with inverse verbs, the reciprocal marker ze- most commonly occupies

the oblique indirect object slot -5, as in the first variant of (22), and can

only rarely occupy the absolutive slot, as in the second variant of (22).

Conversely, the reflexive marker usually occupies the absolutive slot

in these derivatives, as in the second variant of (21). This di¤erence

between reflexivization and reciprocalization is not taken into account

in Smeets (1984) and Smeets (1992);e with derivatives with potential meaning, the reciprocal marker ze-

occupies the oblique slot -5 (24), whereas the reflexive marker occupies

the absolutive slot, as in (23):

(24) A-xe-r Ø-ze-fe-le‚W �-xe-r-ep.

(s)he-abs 3sg.abs-rec.io-ben-see-pl-dyn-neg

‘They hate each other’ (lit. ‘They cannot see each other’).

14. Semantically motivated agent orientation of reflexives is not unique toAdyghe: a similar situation is found in the ergative languages Tsakhur (Kibrik(ed.) 1999) and Warlpiri (Legate 2006) and many other languages (includingsome which, like Warlpiri, are syntactically ergative according to some criteria).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



e with intransitive verbs, the reciprocal marker occupies the oblique IO

slot -5;e with transitive verbs, the reciprocal marker (though in another variant)

also occupies an oblique slot, namely the agent slot -4, as in (18),

though the oblique argument is the most agentive for transitive verbs.

Thus, the reciprocal marker is generally found in one of the oblique

slots and it is controlled by an absolutive personal marker. The only verb

which admits reciprocalization exclusively in the absolutive slot is zexex�n

‘hear’.

The absolutive orientation of reciprocals in most verb classes (except

for the inverse verbs, where variation is observed as in (22)) cannot be

accounted for in semantic terms. The reciprocal prefixes occupy the obli-

que slot of most bivalent verbs, irrespective of whether the oblique argu-

ment is more agentive than the absolutive one (as with transitive and

potential verbs) or less agentive (as with intransitive verbs where the

oblique argument is an indirect object). In other words, the motivation

for the choice of the controller is not the degree of agentivity of the argu-

ments, but rather the location of absolutive marking.

Thus, we take the absolutive orientation of reciprocals as evidence

for syntactic ergativity in Adyghe: reciprocal markers are bound with the

syntactically privileged absolutive argument, whereas the behaviour of the

reflexive marker is semantically motivated.

3.4. Facilitive and di‰cilitive: semantic motivation

Two more valency-decreasing derivations in Adyghe, namely the facilitive

(-‚W es˙

W�) and di‰cilitive (-‚Waje), are treated in syntactic terms in

Rogava & Keraseva 1966 and Kumakhov 1984. These authors claim that

in the derivations in question, the personal marker in subject position (i.e.

the absolutive argument of intransitive verbs, and the ergative (‘oblique’ in

our terminology) of transitive verbs) is eliminated, and the non-subject

marker is retained. This is indeed the case with transitive and intransitive

verbs: in (25) the verb does not bear an agentive marker, which would

normally be j�-. In (26) the situation is less straightforward since the

absolutive subject prefix is zero in 3SG (this is why Smeets (1992) does

not analyse the status of absolutive subjects in facilitive constructions at

all). However, the indirect object prefix jE- is obviously present in (26),

and we can judge that the absolutive prefix is absent from the fact that it

is impossible for most native speakers to express an absolutive argument,

such as ‘It is easy for me to push this boy’, without using an applicative

prefix.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Transitive verb:

(25) M� c˙

’ale-r ‚e-s’�ne-‚W es˙

W �.

this boy-obl caus-fear-fcl

‘It is easy to frighten this boy.’ (lit. ‘This boy is easy to frighten’).

Intransitive verb:

(26) M� c˙

’ale-m Ø-je-w�nc˙

’�-‚W es˙

W �.

this boy-obl 3sg.abs-3sg.ioþoblique-push-fcl

‘It is easy to push this boy.’

However, this derivation can also be analysed in semantic terms: it is

always the most agentive argument which is eliminated. This explanation

seems more plausible, because sensation verbs, such as zexex�n ‘hear’

and le‚W �n ‘see’, often admit expression of an agent. While the case of

zexex�n ‘hear’ is complicated, because this verb is neither a canonical

transitive nor a canonical intransitive verb, le‚W �n ‘see’ is a canonically

transitive verb: the reciprocal formation indisputably demonstrates its

transitivity. The experiencer of verbs like these is in a sense more agentive

than the stimulus (see Dowty 1991 for details): only the experiencer, but

not the stimulus, is animate.

Sensation verb:

(27) Te t�-zere-le‚W �-‚W es˙

W �.

we 1pl.abs-rec.a-see-fcl

‘It is easy for us to see (to meet) each other.’

In (27), the verb includes both an agent and an absolutive personal pre-

fix (note that the verb le‚W �n ‘see’ is syntactically transitive, and there-

fore the agent prefix should be eliminated, according to Smeets 1992).

Hence, it is not only the most agentive argument, but usually a proto-

typical agent which is obligatorily eliminated in the facilitive / di‰cilitive

form – while non-agentive animate arguments can be expressed. In this

sense the facilitive derivation provides evidence against syntactic accusa-

tivity in Adyghe.

4. Conclusions

The processes analysed in this paper can be divided into two groups, based

on the kind of information they provide about ergativity in Adyghe.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



First of all, there are derivations which can be regarded as semantically

motivated (though syntactic motivation can also be proposed for these

processes). This is the case with reflexive marking, since the reflexive

marker always seems to be bound by the most agentive argument; the

facilitive and di‰cilitive always require the omission of the most agentive

argument (though this is generally the agent of transitive verbs or the

absolutive argument of intransitive verbs); the resultative is also semanti-

cally motivated. These derivations could in principle be taken as evidence

that Adyghe is syntactically accusative, but in fact this is not the case.

Secondly, there are derivations which are only compatible with transi-

tive verbs, namely the inadvertitive and potential. These transformations

are more significant for our analysis, since they show that Adyghe is

syntactically ergative. This fact manifests itself in a simple rule which

regulates all ‘passive-like’ derivations: they can only change properties

of the oblique (Agent), but not the absolutive argument. The reciprocal

formation, which is absolutive-oriented, also demonstrates that the absolu-

tive has subject properties (at least the orientation of reciprocals towards

the absolutive cannot be explained in semantic terms).

The conclusion that Adyghe is syntactically ergative seems to be con-

tradicted by the fact that the antipassive is not productive in Adyghe and

does not fully correspond to the prototypical antipassive construction.

However, there is in fact no contradiction. What is essential is that the

Adyghe system forces us to modify our assumptions about the function

served by verbal derivation in an ergative language.

Usually authors implicitly or explicitly adopt the following view on the

function of voices: they change the status of the most syntactically privi-

leged participant. For instance, in Dixon (1994), the observation that the

passive is characteristic of accusative languages, and the antipassive of

ergative languages, is explained by the fact that passives decrease the

syntactic status of the agent (which is privileged in accusative languages),

whereas antipassives decrease the status of the absolutive object (which is

privileged in ergative languages).

However, the Adyghe data lead us to adopt the opposite hypothesis for

this language: derivations can change the status of any participant, except

the most privileged one. As we have shown, no derivation or voice in

Adyghe can change the status of the direct object or intransitive subject –

with the exception of the antipassive. But the antipassive seems to be

unproductive and expressed by means of stem alternation rather than a

grammatical marker sensu stricto, and it violates the general restrictions

which apply to the system of valency-changing derivations in Adyghe.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Thus, Adyghe seems to be syntactically ergative and not syntactically

accusative. This syntactic ergativity manifests itself in the derivational

domain. We think that the view adopted by Matasovic (2008: 64), that

the absolutive case marks ‘‘the lowest ranking macrorole argument’’, may

be true from the semantic point of view. However, the morphosyntactic

processes we have examined do not show that the ergative argument is

syntactically privileged over the absolutive.

However, some other syntactic processes, such as co-referent deletion,

are not at all restrictive in Adyghe, which makes the question of syntactic

alignment and the subject/object status of participants impossible to

answer. The question why the derivational system in Adyghe is much

stricter than many other syntactic processes requires further research,

which is undertaken, for instance, in Letuchiy (2009a).

Note that the hypothesis that Adyghe is syntactically ergative makes

the case of the potential and inadvertitive (see 3.3.1) more problematic.

Since the agent is not the syntactic subject of the underlying transitive

verb, these two passive-like derivations are not entirely similar to the

passive voice: they do not demote the initial subject. In any case, we

should suppose that the agent is a syntactically privileged argument, even

if it is not a subject. The passive-like derivations demote a privileged argu-

ment to the status of indirect object. In other words, provided that Adyghe

is a syntactically ergative language, the two passive-like derivations are

similar to voice alternations in that they rearrange the syntactic status

and change the pragmatic values of the arguments, but this voice alterna-

tion di¤ers from the passive formation in that the demoted argument is

not a syntactic subject.

These data from Adyghe are also useful because they show that syntactic

accusativity in some morphologically ergative languages can in fact be

determined by semantic factors. Many processes which seem to be moti-

vated accusatively, such as the formation of the reflexive, facilitive or

di‰cilitive, may in fact be motivated in semantic terms rather than being

directly related to syntactic alignment.

Abbreviations

1,2,3 1st, 2nd, 3rd person

a agent cross-reference prefix

abs absolutive case; absolutive argument cross-reference prefix

ap antipassive variant of the verb stem

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



ben benefactive

caus causative

com comitative

fcl facilitive

ins instrumental case

io indirect object cross-reference prefix

obl oblique case

oblique oblique indirect object prefix

re refactive

rec reciprocal

refl reflexive

sg singular

pl plural

pst past tense

tr transitive variant of the verb stem

References

Aikhenvald, Alexandra and Robert M.W. Dixon2000 Changing valency: Case studies in transitivity. Cambridge: Cam-

bridge University Press.Arkadiev [Arkad’jev], Petr M., and Alexander B. Letuchiy [Letucij]

2008 Derivacii antipassivnoj zony v adygejskom jazyke [Antipassive-like derivations in Adyghe]. In Tipologija glagol’noj derivacii[The typology of verbal derivation], Vladimir A. Plungian andSergei G. Tatevosov (eds), Moscow: Jazyki slav’anskoj kul’tury.

Arkadiev [Arkad’jev], Petr M., and Yakov G. Testelets2009 O trex ceredovanijax v adygejskom jazyke [On three alternations

in Adyghe]. In Aspekty polisintetizma: Ocherki po grammatikeadygejskogo jazyka [Aspects of polysynthesis: Essays on Adyghegrammar], Yakov G. Testelets, Peter M. Arkadiev, Alexander B.Letuchiy and Nina R. Sumbatova (eds), 121–145. Moscow:RGGU.

Baker, Mark1996 The polysynthesis parameter. New York: Oxford University

Press.Belyaeva, Anna V., and Sergei V. Minor

2005 ‘‘Iskl’ucitel’naja koreferentnost’’’ v konstrukcijax s sentencial’nymiaktantami v adygejskom i kabardino-cerkesskom jazykax[‘‘Extraordinary co-reference’’ in constructions with sententialarguments in Adyghe and Kabardian]. In Vtoraja konferencijapo tipologii i grammatike dl’a molodyx issledovatelej. Tezisy

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



dokladov [Abstracts of the second conference in grammar andtypology for young researchers], Sergei S. Say et al. (eds). SaintPetersburg: ILI RAN.

Comrie, Bernard1976 The syntax of causative constructions: Cross-language similarities

and divergences. In Syntax and Semantics 6. The grammar ofcausative constructions, Masayoshi Shibatani (ed.), 261–312. NewYork: Academic Press.

Cooreman, Ann1994 A functional typology of antipassives. In Voice: Form and Func-

tion [Typological Studies in Language 27], Barbara A. Fox andPaul J. Hopper (eds), 49–88. Amsterdam and Philadelphia:Benjamins.

Dixon, Robert M.W.1994 Ergativity. Cambridge: CUP.

Gisev, Nukh T.1968 Glagoly labil’noj konstrukcii v adygejskom jazyke [Labile verbs in

Adyghe]. Majkop: Adygejskoje kniznoje izdatel’stvo.Jelinek, Eloise

1984 Empty categories, case, and configurationality. Natural languageand linguistic theory 2.1.

Hewitt, Bernard George1982 ‘Antipassive’ and ‘labile’ constructions in North Caucasian lan-

guages. General Linguistics 22.3. 158–171.Hopper, Paul and Sandra Thompson

1980 Transitivity in grammar and discourse. Language, v. 56.2. 251–299.

Kibrik, Alexander E.2002 Ocerki po obscim i prikladnym voprosam jazykoznanija. Moscow:

URSS.Kibrik, Alexander E. (ed.)

1999 Elementy tsakhurskogo jazyka v tipologiceskom osvescenii [Ele-ments of Tsakhur grammar in typological perspective]. Moscow:Nasledie.

Kumakhov, MuhaddinMorfologija adygskix jazykov [Morphology of the Adyghe lan-guages]. Moscow: Nauka.

Lander, Yuri A.2009 Mnozestvennaja reljativizacija: podlinnaja i mnimaja. In Aspekty

polisintetizma: ocerki po grammatike adygejskogo jazyka [As-pects of polysynthesis: aspects of Adyghe grammar], Yakov G.Testelets, Peter M. Arkadiev, Alexander B. Letuchiy and NinaR. Sumbatova (eds), 612–653. Moscow: RGGU.

Legate, Julie Anne2006 Split absolutive. In Ergativity. Emerging Issues, Alana Johns,

Diane Massam and Juvenal Ndayiragjje (eds). Berlin: Springer.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Letuchiy, Alexander B.2007 Reciprocals, reflexives, sociatives and comitatives in Adyghe.

In Reciprocal constructions [TSL 71], Vladimir P. Nedjalkov,Emma S. Geniusiene and Zlatka Guentcheva (eds), 773–811.Amsterdam: John Benjamins.

Letuchiy, Alexander B.2009a Verb classes in Adyghe: derivational and non-derivational criteria.

Abstract of the talk at the conference ‘‘Verb typologies revisited’’.Letuchiy, Alexander B.

2009b Kauzativ, dekauzativ i labil’nost’ [Causative, anticausative andlability]. In Aspekty polisintetizma: ocerki po grammatike ady-gejskogo jazyka [Aspects of polysynthesis: aspects of Adyghegrammar], Yakov G. Testelets, Peter M. Arkadiev, AlexanderB. Letuchiy and Nina R. Sumbatova (eds), 372–428. Moscow:RGGU.

Matasovic, Ranko2008 Transitivity in Kabardian. In Investigations of the Syntax–

Semantics–Pragmatics Interface, Robert D. Van Valin (ed.),59–74. Amsterdam, Philadelphia: John Benjamins.

Nichols, Johanna1986 Head-marking and dependent-marking grammar. Language, v.

62.1.Nikolaeva, Liudmila L.

2003 Rezul’tativ v adygejskom jazyke [The resultative in Adyghe].Expedition report. Unpublished manuscript.

Paris, Catherine1987 Comment sont remplies en tcherkesse les functions devolues

dans d’autres langues aux variations de diathese. Actances 3.Plungian, Vladimir P.

2000 Obscaja morfologija [General morphology]. Moscow: URSS.Rogava, Grigorij V. and Zejnab I. Keraseva

1966 Grammatika adygejskogo jazyka [A grammar of Adyghe].Krasnodar: Krasnodarskoje kniznoje izdatel’stvo.

Say, Sergei S.2007 K tipologii antipassivnyx konstrukcij: semantika, pragmatika,

sintaksis [Towards a typology of antipassive constructions:semantics, pragmatics, syntax]. PhD. diss. Saint Petersburg:Institute for linguistic studies.

Serdobol’skaja, Natalia V.2007 Ergativity in Adyghe. Talk in the Caucasian seminar at INALCO,

Paris.Shibatani, Masayoshi

1985 Passives and related constructions: a prototype analysis. Language,v. 61: 821–848.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Smeets, Rieks1992 On valencies, actants and actant coding in Circassian. In Cauca-

sian Perspectives, George B. Hewitt (ed.), 98–144. Munchen:Lincom Europa.

Smeets, Rieks1984 Studies in West Caucasian phonology and morphology. Leiden:

Hakuchi Press.Testelets, Yakov G.

2001 Vvedenije v obscij sintaksis [Introduction to general syntax].Moscow: RGGU.

Testelets, Yakov G., and Svetlana Yu. Toldova1998 Refleksivnyje mestoimenija v dagestanskix jazykax i tipologija

refleksiva. Voprosy jazykoznanija 4.Txarkaxo, Junus A.

1991 Adygejsko-russkij slovar’ [Adyghe-Russian dictionary]. Majkop:Adygejskoje kniznoje izdatel’stvo.

Van Valin, Robert D. and Randy La Polla1997 Syntax. Structure, meaning and function. Cambridge: Cambridge

University Press.Zekox, Ucuzuk S.

2002 Adygejskaja grammatika [A grammar of Adyghe]. Majkop:GURIPP ‘‘Adygeja’’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The evolution of transitive verbs in Basque andemergence of dative-marked patients

Celine Mounole

1. Introduction1

Basque is the only surviving pre-Indo-European language of Western

Europe. It is spoken in the Basque Country, a region situated between

France and Spain. It is a SOV language, ergative, overwhelmingly su‰x-

ing and highly agglutinating. As for the verbal morphology, it is mainly

analytic. In historical Basque, the middle voice values of autocausative

and decausative are obtained by detransitivization of the verb: the patient

of the transitive construction becomes the single argument of an intransi-

tive one. At the same time, the transitive auxiliary is replaced with an

intransitive one. As concerns the passive interpretation, it is expressed by

participial predicate structures, namely resultative constructions (Rebuschi

1983, Trask 1985, Ortiz de Urbina & Uribe-Etxebarria 1991). That is

why there are often identified as ‘stative or adjectival passives’ (Ortiz de

Urbina 2003: 299).

In this paper, we will argue that di¤erential object marking is emerging

in modern Basque. This kind of development being unusual in an ergative

language, we will try to describe the mechanisms involved. As we will

see, animacy and referentiality seem to be the main factors governing this

phenomenon. It still remains to be investigated to what degree the DOM

system is employed for discourse purposes.

After a brief presentation of the structure of actancy in Basque (§2), we

will provide the data of some varieties of this language which encode in

the dative the patients of transitive verbs (§3.1 and §3.2). Afterwards, we

1. I want to express my gratitude to Gilles Authier and Katharina Haude whogive me the opportunity of publishing this paper, and to Ricardo Etxepare,Beatriz Fernandez and the two anonymous reviewers for their comments andencouragement. I am also grateful to Luigi Anselmi, Kepa Erdozia, AitorIglesias, Alazne Landa, Arantzazu Elordieta, Julen Manterola, Jean Harit-schelhar, Maitena Laxague and the informants of Arratia, Lekeitio, Tolosa,Baigorri and Azkarate. All errors are mine.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

(V9 13/1/12 13:42) WDG-LCB (155mm�230mm) TimesNRMT 1351 Authier pp. 353–376 1351 Authier_11_mounole (p. 353)

will compare those data with the structure of transitives in Spanish, the

language in contact with the varieties of Basque displaying this pattern

(§3.3 and §3.4).

2. Actancy in Basque

Before starting the study of the evolution of transitive verbs we shall briefly

describe the structure of actancy in Basque.

Throughout the paper, I will use the following abbreviations: A for the

Agent-like term of transitive verbs, P for the Patient-like term of the same

transitive verbs, S for the Single term of intransitive verbs and D for the

Dative term of both groups of verbs. Furthermore, by transitive verb I

mean a verb governing an agent (A) in the ergative case and a patient (P)

in the absolutive case (i.e. verbs of actions, and verbs structured and

behaving like verbs of action) (§2.2). If it does not govern any P term in

the absolutive, it is intransitive. So, we will consider as intransitive the

monovalent verbs with a single term in the absolutive (§2.1) or in the erga-

tive (§2.3.1), and as bivalent intransitives the verbs bearing a term in the

absolutive and another in the dative (§2.1), and those with a term in the

ergative and another in the dative (§2.3.2).

2.1. Intransitive verbs

Intransitive constructions generally display a term S in the absolutive

indexed on the verb by means of prefixes, except for the 3rd person singular

and plural, since their absolutive NP is never overtly reflected on the verb

((1) vs (2)):

(1) Zu erori zara

You.abs fall aux.pres.s2sg

‘You have fallen down’

(2) Jon etorri da

Jon.abs come aux.pres.s3sg

‘Jon has come’

Moreover, intransitives can include a second term in the dative case,

which also triggers verb agreement by the addition of su‰xes (3–5). These

bivalent intransitive constructions are mainly found with three types of

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


354 Celine Mounole

verbs.2 First, a handful of psychological verbs require the experiencer in

the dative and the stimulus in the absolutive. This is the case for interesatu

‘to interest’, gustatu ‘to like’, damutu ‘to repent’, gaitzitu ‘to be o¤ensive

to’, dolutu ‘to repent’ (3) among others.3 Besides, the dative agreement is

also used with motion verbs to express the end-point of the motion (4),

or as a dative of interest, with possessed entities (5) (for more detail, see

Etxepare 2006): 4

(3) Zu-ri dolutzen zaizu han izana

You-dat regret aux.pres.s3sg.d2sg there being.abs

‘You regret to have been there’

(4) Jon-i Andoni joan zaio

Jon-dat Andoni.abs go aux.pres.s3sg.d3sg

‘Andoni has gone from / to Jon’

2. We also find this construction with some verbs that encode the ‘aimer’ in theabsolutive and the ‘aimed object’ in the dative (Lazard 1998). This class ofverbs that was more extended four centuries ago, is nowadays restricted to ahandful of verbs ( jarraiki ‘to follow’, jarin ‘to spill’, atxiki ‘to attach to’) thatalternate the absolutive-dative construction with the ergative-dative construc-tion or with a transitive one (ergative-absolutive) (see §2.3.2).

3. Anyway, some of them can also be constructed monovalently with the experi-encer receiving the absolutive and the stimulus the instrumental or locativemarking:

(54) Literatura-n interesatzen daLiterature-loc interest aux.pres.s1sg

‘He is interested in literature’

(55) Damutzen da bere jokamolde-azRegret aux.pres.s1sg his behaviour-inst

‘He regrets his behaviour’

4. However, this construction is obligatory neither with the possessed entities norwith the motion verbs. We can perfectly translate the possession by a genitivalphrase (56) and the end-point of the motion by an allative (57):

(56) Jon-en liburua erori daJon-gen book.abs fall aux.pres.3sg

‘Jon’s book has fallen’

(57) Jon Andoni-rengana joan daJon.abs Andoni-all go aux.pres.s3sg

‘Jon has gone to Andoni’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


The evolution of transitive verbs in Basque 355

(5) Jon-i liburua erori zaio

Jon-dat book.abs fall aux.pres.s3sg.d3sg

‘Jon’s book has fallen’ (lit. ‘The book has fallen to Jon’)

2.2. Transitive verbs

Basque transitive constructions require a term A in the ergative case, and

a term P in the absolutive case. Both are indexed on the verb: A by means

of su‰xes and P by prefixes, the latter being the same as the S prefixes of

the intransitive construction. Again, the third person is an exception to

this rule since its absolutive and ergative NPs are never overtly encoded

on the verb ((6) vs (7)):

(6) Ni-k zu ikusi zaitut telebista-n

i-erg you.abs broke aux.pres.a1sg.p2sg television-loc

‘I have seen you on TV’

(7) Jon-ek mahaina hautsi du

Jon-erg table.abs broke aux.pres.a3sg.p3sg

‘Jon has broken the table’

A third term D encoded in the dative can appear in transitive construc-

tions. It also triggers agreement on the verb by the addition of su‰xes (8),

just the same as those employed to index the dative NP in intransitive

constructions (cf. (3–5) above).5 The absolutive term of the trivalent verbs,

which is restricted to be a third person, is not reflected on the verb; neither

are the 3rd person ergative NPs:

(8) Ni-k zu-ri ogia eman dizut

i-erg you-dat bread.abs give aux.pres.a1sg.p3sg.d3sg

‘I have given you some bread’

Finally, the 3rd person is overtly marked on the verb only when it is

encoded in the dative case. This is done by the addition of the su‰x -o

(singular) or -e (plural) (see (4) and (9)):

(9) Jon-ek Mikel-i ogia eman dio

Jon-erg Mikel-dat bread.abs give aux.pres.a3sg.p3sg.d3sg

‘Jon has given Mikel some bread’

5. As nowadays, in the 16th century, the agreement with the dative object issystematic in Western (Bizkaian, Araban) and Central (Gipuzkoan) dialects,while it is variable in the Eastern ones (Low Navarrese, Zuberoan) (Ortiz deUrbina 1991; Etxepare and Oyharcabal 2008).

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


356 Celine Mounole

2.3. Split intransitivity

Apart from those typical transitive and intransitive constructions which

are the most common and widespread ones in the system, the Basque lan-

guage has some verbs sharing an unusual pattern. Those verbs possess a

term in the ergative but none in the absolutive. Even so, they trigger the

morphology usually used with prototypical transitive verbs. This phenom-

enon has been referred to as split-intransitivity (Dixon 1994), and those

verbs as unergatives (Perlmutter 1978, Levin 1983), anti-impersonals (Lazard

1985, 1995), or accusatively-aligned intransitive verbs (Creissels 2006).

2.3.1. Intransitive monovalent verbs with an NP in the ergative

The Basque language has a handful of intransitive verbs with an NP in

the ergative and none in the absolutive. As Trask (2002) reminded us the

majority of them are quite old, and ‘‘mostly appear to be historical acci-

dents’’.6 Some of them may be ancient causative verbs (for example, iraun

‘to last’ in (10), irakin ‘to boil’, see Rebuschi 1984) since they bear the

causative prefix -ra which can still be recognized in some pairs of verbs

like ikusi ‘to see’ / erakutsi ‘to make see, to show’.

Besides, Basque has many light verb constructions composed of a bare

noun (sometimes an adjective, or an adverb) plus the verb egin ‘to do’. In

many cases, both elements have joined together, leaving the construction

without any P element (see Ortiz de Urbina 1989; Trask 2002; Etxepare

2003):7

6. In this section we only consider the verbs of Basque origin. However, inSouthern dialects, many recently borrowed verbs follow this same pattern:desaparezitu ‘to disappear’, dudatu ‘to doubt’, funtzionatu ‘to function’, eskiatu‘to ski’ (Sarasola 1977; Alberdi 2003).

7. We must add that the Nþ egin ‘to do’ complex predicates are not homoge-nous since the N of some of them seems to behave as the true Patient-liketerm of egin ‘to do’. Indeed, it may be separated from the verb egin in wh-questions or focalized constructions, and partitive-marked in negative con-structions (see Ortiz de Urbina 1989, Laka 1993):

(58) ez du salto-rik eginneg. aux.pres.a3sg.p3sg jump-PART do

‘He didn’t jump’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(10) Kontzertu-ak ez zuen luzaz iraun

concert-erg no aux.past.a3sg.p3sg long last

‘The concert did not last long’

(11) Irakasle-ak barre egin zuen

teacher-erg laugh do aux.past.a3sg.p3s

‘The teacher laughed’

2.3.2. Intransitive bivalent verbs with an NP in the ergative and an NP

in the dative

The second group of verbs with an unusual pattern, like the former, has

one term in the ergative and no term in the absolutive. It di¤ers however

considerably from the first group since it involves a second term marked in

the dative. Furthermore, while showing only two terms, it requires the

morphology usually used with prototypical ditransitive verbs (see (§2.2)

above).8

The verbs sharing this unusual pattern di¤er in their origin and condi-

tions of occurrence and, in consequence it would be convenient to distin-

guish di¤erent subgroups among them. Indeed, some of them are aligned

in this way already in our oldest texts while others have undergone a

change in their actancy in historical times.

To begin with, since their first records, the verbs deitu ‘to call’, eraso ‘to

attack’, itxadon / itxaron ‘to wait’, begiratu ‘to look at’, oratu ‘to grasp,

seize’, iguriki ‘to wait’ have been constructed with an NP in the ergative,

an NP in the dative, and have displayed a trivalent morphology. In some

cases, this unusual alignment seems to be easily explicable. Itxadon /

itxaron and iguriki ‘to wait’ may be ancient compounds which got fused,

letting the construction without any apparent patient (hitz ‘word’þ *edun

‘to have’ > itxadon and egun ‘day’þ eduki ‘to hold’ > iguriki; cf. Trask

2002); eraso ‘to attack’ might be an ancient causative (Etxepare p.c.). As

for deitu ‘to call’, which is a simple transitive in the Northern dialects, it

may have su¤ered a change in actancy in the Southern dialects, becoming

an intransitive with an NP in the ergative and an NP in the dative, by

analogy with its equivalent composed form dei egin ‘‘to call’’ (dei ‘calling’

8. It must be understood that even if the verbal morphology required is origi-nally reserved to the ditransitive constructions, in those constructions none(absolutive) NP has been omissed. Furthermore, remember that in the ditran-sitive constructions the absolutive NP, that can only be a 3rd person, is notovertly indexed on the verb.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


358 Celine Mounole

plus egin ‘to do’) (Creissels 2008, Etxepare p.c.). Finally, begiratu ‘to see’

used with an NP in the ergative and an NP in the dative in some dialects,

hardly shows us a clear explanation of its actancy. In (12) and (13), we see

that the construction displays an NP in the dative (ainbesteri ‘so much’

and sazerdoteari ‘priest’, respectively), and no NP in the absolutive. The

term in the ergative is dropped but it is indexed on the verb (-k in itxadok,

and -gu in diogu):

(12) Gatx egiten boc ainbeste-ri itxadok

bad do aux.a2sg.p3sg so much-dat wait-a2sg.p3sg.d3sg

‘If you do it badly, wait for so much’ (Refranes y Sentencias

1596: 296)

(13) Begiratu bear diogu sazerdote-ari

see need aux.pres.a1pl.p3sg.d3sg priest-dat

Iesus Christo-ri berari bezala

Jesus Christ-dat himself like

‘We have to look at the priest as we look at Jesus Christ himself ’

(Beriain 1621: 12)

We must add that, depending on the dialects,9 some of these verbs can

also be used transitively without a¤ecting either the lexical meaning of the

verb or the interpretation of the process (for a detailed description, see

Fernandez 2008). However, some other verbs provide a di¤erent meaning

depending on the pattern adopted (Fernandez 2008). For example, the

verbs begiratu and oratu mean ‘to look at’ and ‘to grasp, to seize’ respec-

tively when constructed with an NP in the ergative and an NP in the

dative (13) and (15b), but ‘to protect, to save’ and ‘to capture, to catch’

when constructed transitively (14) and (15a):

(14) Nola bere promesa begiratzen du?

how his promess.abs save aux.pres.a3sg.p3sg

‘How does he save its promise?’ (Leizarraga 1571: Oth. D 6r)

(15) a. Gena-k sagua askatzen ta eullia oratzen

spider’s_web-erg mouse.abs liberate and fly.abs capture

‘The spider’s web liberates the mouse and captures the fly’

(Refranes y Sentencias 1596: 298)

9. Both constructions are possible with deitu ‘to call’, itxadon ‘to wait’ since ouroldest texts. On the contrary, iguriki ‘to wait’, begiratu ‘to see’, eraso ‘toattack’ are transitivized in classical and modern Basque, certainly by hyper-correction or analogy with the verbs of action.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



b. Oratu egijozu aga on-i

Grasp aux.imp.a2sg.p3sg.d3sg railing this-dat

‘Grasp yourself to this railing’ (Mogel 1881: 118)

Moreover, depending on the conjugation path adopted, other verbs

slightly vary in the interpretation of the process described. The clauses

with the verb deitu ‘to call’, when employed transitively have two possible

interpretations (i) ‘they called / phoned Xabier to his o‰ce’, or (ii) ‘they

called him to go to their o‰ce’ (16a). In contrast, when employed with

an NP in the ergative and an NP in in the dative (16b), the clauses have

only one possible interpretation: ‘they called him to his o‰ce’ (Etxepare

2006: 412–413):

(16) a. Xabier bulegora deitu dute

Xabier.abs o‰ce.all call aux.pres.a3sg.p3sg

‘They called Xabier to the o‰ce’

b. Xabierr-i bulegora deitu diote

Xabier-dat o‰ce.all call aux.pres.a3sg.p3sg.d3sg

‘They called Xabier to the o‰ce’

Apart from all these verbs showing this pattern since our oldest texts,

the other verbs which nowadays bear an NP in the ergative and an NP in

the dative have undergone a change in their actancy in historical Basque.

They have two sources: (i) the evolution of some intransitive bivalent

verbs and (ii) the valency-shift of some transitive bivalent verbs. Indeed,

some intransitive bivalent verbs constructed with an NP – normally the

subject – in the absolutive and an NP in the dative in the oldest texts

(17), later on shifted towards another intransitive pattern with an NP

–the subject– in the ergative, the other NP remaining in the dative (18).

It is the case of the verbs lagundu ‘to help’, eskatu ‘to ask for’, jarraiki /

jarraitu ‘to follow’, eutsi ‘to retain, to hold on’, etxeki ‘to retain’. It is

interesting to notice that among them, the last verb is the only one to

have undergone this change path in the Northern dialects of Basque. In

fact, the first three have undergone a distinct evolution in these dialects

becoming purely transitives with one term in the ergative and another in

the absolutive (19):10

10. In historical Basque eutsi ‘to retain, to hold on’ is not employed in the Easterndialects. They employ the verb etxeki ‘to retain’ instead of it. It is just thecontrary in the Western dialects.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


360 Celine Mounole

(17) Zein gauza lagun-duko iaku

which thing.abs help-fut aux.pres.s3sg.d1pl

kastidadea goardeetako?

chastity.abs save

‘What can help us to save our chastity?’ (Kapanaga 1656: 50)

(18) Zer-k lagun-duko digu kasto izateko?

what-erg help-fut aux.pres.a3sg.p3sg.d1pl chaste be

‘What will help us to be chaste?’ (Elizalde 1735: 49)

(19) Lagun gaitzatzue

help aux.imp.a2pl.p1pl

‘Help us’ (Etxeberri 1627: XLIII)

The evolution of transitive verbs is the second source for the occurrence

of this pattern. Whereas in some dialects they remain always transitive

(20), in others, their human patient shifts from the absolutive case towards

the dative (21)11 and the bivalent morphology of the verb is replaced by

the trivalent one:

(20) (Ni-k) (zu) ikusi zaitut

i-erg you.abs see aux.pres.a1sg.p3sg

‘I have seen you’

(21) (Ni-k) (zu-ri) ikusi dizut

i-erg you-dat see aux.pres.a1sg.p3sg.d3sg

‘I have seen you’

In this paper, we are going to deal with this evolution of transitive

verbs and to try to give an answer to the mechanisms causing the occur-

rence of such an exceptional pattern.

3. Emergence of dative-marked patients among the transitive verbs

In this section we shall focus on the evolution of transitive verbs causing

the occurrence of dative-marked patients in Basque. We will see some

data of archaic and old Basque, before describing in more detail the situa-

tion of modern Basque. We will finish by comparing the Basque and the

Spanish data.

11. Data collected from young speakers of the Gipuzkoan variety of Tolosa (see3.2.3)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



3.1. Archaic and Old Basque12

In sixteenth century’s texts, transitive verbs almost always governed a term

A in the ergative and a term P in the absolutive, both indexed or cross-

referenced on the verb.

At that time, we find only two examples of transitive verbs bearing

an unusual pattern. The verbs ikusi ‘to see’ and ulertu ‘to understand’,13

commonly conjugated transitively in all the dialects, carry the agent in

the ergative and the patient in the dative in two Western texts – Refranes

y Sentencias 1596 (Bizkaia) and Lazarraga c.1564 (Araba). Besides, the

trivalent auxiliary is used (see footnote 8):

(22) Joan gura dot ekustera ni-k

Go desire aux.pres.1sg.p3sg see i-erg

on deretxada-n-ari

love.s3sg.p3sg.d1sg-rel-dat

‘I want to go to see the person who loves me’

(Lazarraga c. 1564: 1164v)

(23) Trakart egiten deustak ta ulerretan

deceit do aux.pres.a2sg.p3sg.d1sg and understand

deustat

aux.pres.a1sg.p3sg.d2sg

‘You deceive me and I understand you’ (Refranes y Sentencias

1596: 49)

From the next century onwards, it has been possible to observe the

spreading of this kind of structure among the transitive verbs. We find,

for example, danatu ‘to damage’ and persegitu ‘to pursue’ with an agent

12. According to Lakarra’s periodization (1997), Archaic Basque embraces theperiod until 1600 and Old Basque the period between 1600 and 1745.

13. Until the 19th century, the verb ulertu ‘to understand’ is only used in occiden-tal dialects. Anyway, in the text where it appears with a P in the dative case(23), it is also used with a P in the absolutive in which case this latter is[-human]:

(59) uler ezak lenago taunderstand aux.imp.a2sg.p3sg first and

itz egik geroengoword do.imp.a2sg.p3sg later

‘First understand [it] and speak latter’ (Refranes y Sentencias 1596: 194)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


362 Celine Mounole

in the ergative, a patient in the dative and a trivalent auxiliary in Bizkaian

and High Navarrese texts (Southern dialects):

(24) On-ek persegietan deusku-elako

those-erg pursue aux.pres.a3sg.p3sg.d1pl-caus

‘Because they [our worst temptations] pursue us all’

(VJ 17th century, Mitxelena 1954)

(25) Nor-k danatzen digu?

who-erg injure aux.pres.a3sg.p3sg.d1pl

‘Who is injuring us?’ (Elizalde 1735: 49)

All those earliest occurrences of the dative-marked patients are to be

found in Southern texts (Spanish Basque Country) and restricted to

[þhuman] patients (see 22–25). Among them, we find Spanish loans as

well as verbs of Basque origin. We will return to this topic (§3.2.4).

3.2. Modern Basque

In Modern Basque, dative-marked patients are more widespread. Di¤erent

dialects have developed them among the transitive verbs. We find them in

some varieties of High Navarrese, Bizkaian, and Gipuzkoan, all varieties

of Southern Basque (Spanish Basque Country). Thus, first of all, we are

going to present the data of each dialect one by one, and next, we will try

to give an explanation of the mechanisms involved.

3.2.1. High Navarrese

In his study of the Basque verb, Bonaparte ([1869] 1991)14 pointed out the

absence in most varieties of meridional High Navarrese15 of many forms

of the bivalent auxiliary usually used with prototypical transitive verbs.

Indeed, the bivalent forms indexing the 1st and 2nd person absolutive

patients had disappeared, and instead of them, the speakers employed the

trivalent form, thus encoding those patients in the dative. He found only

the bivalent auxiliary indexing the 3rd person absolutive patients.

A century later, Yrizar (1981) was able to confirm that the description

of the meridional High Navarrese given by Bonaparte was in general still

14. Bonaparte’s Le verbe Basque en tableaux ([1869] 1991) was the first monographyof the Basque verb that provided the data from all dialects.

15. In all the area apart from the valleys of Erro, Artze and Burgete.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



valid.16 Furthermore, he found the same phenomenon in some Western

varieties of High Navarrese (Sakana and Burunda varieties) which in

Bonaparte’s times still used the bivalent auxiliary without any restriction.

Nowadays, in Etxarri-Aranatz (variety of Sakana), the 3rd person human

patients can also be encoded in the dative (Erdozia 2001):17

(26) Alsasu-e yaman doogu Fermin-i

Altsasu-adl bring aux.pres.a1sg.p3sg.d3sg Fermin-dat

‘We have brought Fermin to Altsasu’

However, the dialectological descriptions do not say precisely whether

the use of the dative is systematic with all the 3rd person animate patients

or whether it has any restriction (see below §3.2.4).

3.2.2. Bizkaian

Bonaparte (1991) did not gather any case of dative propagation among

the patients of transitive constructions in Bizkaian. However, Yrizar (1981)

points out its use in the varieties of Markina, Gernika, Bermeo, Arratia

and Arrigorriaga. Here, we are going to study the data of the seaside village

Lekeitio and of the valley of Arratia.

As shown by Hualde, Elordieta and Elordieta (1994), in Lekeitio, they

tend to encode the patient of transitive constructions in the dative. Indeed,

the dative case can appear with the 1st, 2nd and 3rd person singular and

plural patients even if it is less frequent with the latter type. The 3rd

person patient can be encoded in the dative if and only if it is animate

(27a, 27b):

(27) a. Pedrori ikusi dotzat

Pedro-dat see aux.pres.a1sg.p3sg.d3sg

‘I have seen Pedro’

b. Kotxia ikusi dot

car.abs see aux.pres.a1sg.p3sg

‘I have seen the car’

16. Yrizar did not give more data as this dialect was moribund. However, he wasable to add that the phenomenon occurred in the variety of Erro.

17. Unfortunately, we are not able to give modern data of Altsasu (Burunda) asthe variety spoken there is almost extinct today.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


364 Celine Mounole

Anyway, the restrictions concerning the use of the dative are even more

complex than they seem to be at first sight. Indeed, it is restricted to the

human patients (28) (as in Fernandez 2008) and moreover, the definiteness

of those human patients too seems to be an important condition for the

occurrence of the dative case (29), (30):18

(28) a. **Txakurra-ri ikusi dotzat

dog-dat see aux.pres.a1sg.p3sg.d3sg

‘I have seen the dog’

b. Txakurra ikusi dot

dog.abs see aux.pres.a1sg.p3sg

‘I have seen the dog’

(29) a. **Eztotzat ezaututen inor-i

neg.aux.pres.a1sg.p3sg.d3sg know nobody-dat

‘I don’t know anybody’

b. Eztot ezaututen inor

neg.aux.pres.a1sg.p3sg know nobody.abs

‘I don’t know anybody’

(30) a. **Morroi bat-i ikusi dotzat

guy one-dat see aux.pres.a1sg.p3sg.d3sg

‘I have seen a guy’

b. Morroi bat ikusi dot

guy one.abs see aux.pres.a1sg.p3sg

‘I have seen a guy’

In the valley of Arratia, 1st and 2nd person patients are all encoded in

the dative (31), but they would never encode the 3rd person patient in this

way –even if it is human and definite–. Indeed, it is always in the absolu-

tive case (32) (Iglesias in prep).19

18. These data were provided by a 50 year-old man, native of Lekeitio. It isremarkable that another speaker of this variety only accepted to encode the3rd person patient in the dative with the frequent verb ikusi ‘to see’. So,it seems that frequency of the verb also plays a role in the distribution ofthis pattern. Anyway, we must add that in Lekeitio the new pattern has notreplaced the old one. Thus, the marking of the patient still alternates betweenthe dative and the absolutive.

19. Arretxe (1994) described the same situation for the Basque spoken in Basauri.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(31) Artu oskun kotxean eroan oskun

take aux.past.a3pl.d1pl car carry aux.past.a3pl.p3sg.d1pl

‘They took us in their car, and carried us’

(32) A ikusı dot

him.abs see aux.pres.a1sg.p3sg

‘I have seen him’

As for the younger generation, my informant accepts to encode the 3rd

person patient in the dative only with the verb ikusi ‘to see’ (and not for

example ezagutu ‘to know (somebody)’) which is also a very frequent

verb (cf. footnote 18):

(33) Ari ikusı dotzet

him-dat see aux.pres.a1sg.p3sg.d3sg

‘I have seen him’

3.2.3. Gipuzkoan

Bonaparte (1991) and Yrizar (1981) do not find this pattern in Gipuzkoan.

Nowadays, it is, however, very productive in some of its varieties. The

data presented above have been collected in the variety of Tolosa (Central

Gipuzkoa) from speakers aged from 21 to 33.

Generally speaking, the speakers encode the 1st and 2nd person patients

in the dative. As for the 3rd person patient’s case-marking, it is restricted by

human animacy and referentiality constraints: all the inanimate patients are

in the absolutive but not all the human ones are encoded in the dative.20 As

the examples show, the definite NPs (proper names, pronouns, and definite

nouns) are in the dative, but the indefinites (inor ‘nobody’) (35), quantifiers

(asko ‘many’, guzti ‘all’) (36–37) and reciprocals (elkar ‘each other, one

another’) (38) can only be in the absolutive:

(34) (Ni-k) irakasle-ari ikusi diot

i-erg teacher-dat see aux.pres.a1sg.p3sg.d3sg

‘I have seen the teacher’

20. Interestingly, in some cases, the majority of them also accepted the transitivepattern based on a term A in the ergative and a term P in the absolutive as anequivalent of the marked pattern they usually employ (ezagutu ‘to know’, jo‘to hit’), while in some others almost all of them totally rejected the formerpattern (utzi ‘to let, to leave’, ikusi ‘to see’, aukeratu ‘to choose’), qualifying itas ‘‘wrong, ungrammatical’’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


366 Celine Mounole

(35) **(Ni-k) ez diot inorr-ei ikusi

i-erg no aux.pres.a1sg.p3sg.d3sg anybody-dat see

‘I have not seen anybody’

(36) **Jon-ek neska asko-ri ikusi dio

Jon-erg girl many-dat see aux.pres.a3sg.p3sg.d3sg

‘Jon has seen many girls’

(37) **Jon-ek neska guzti-ei ikusi die

Jon-erg girl all-dat see aux.pres.a3sg.p3sg.d3sg

‘Jon has seen all the girls’

(38) **Elkarr-i ikusi diote

each other-dat see aux.pres.a3pl.p3sg.d3sg

‘They have seen each other’

Concerning the indefinite article bat it seems that the dative-marked

ones receive referential interpretation (39b), and the absolutive-marked

ones, a non-referential interpretation (39a):

(39) a. Idazkari bat bilatzen det

secretary.abs one.abs look_for aux.pres.a1sg.p3sg

‘I am looking for a secretary’ (in general)

b. Idazkari bati bilatzen diot

secretary one-dat look_for aux.pres.a1sg.p3sg.d3sg

‘I am looking for a secretary’ (¼one of my o‰ce)

A further point is that there is a generational split in the use of this

pattern. The speakers of this variety of Basque who are over 40 encode

all their animate patients – so, the 1st, 2nd and 3rd person patients – in

the absolutive and employ the transitive bivalent morphology. However,

even if they would not employ it, they accept the dative-marking of the

human patient as a potential variant.

3.2.4. Summary

In short, the data presented above show that dative marking of the patient

in particular varieties of Basque is not done at random. Animacy and

referentiality degrees of the NP involved seem to play a role in the use

and the non-use of the dative-marking.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



In all the varieties encoding the patient in the dative case, humanhood

is a central constraint.21 However, this phenomenon has di¤erent realiza-

tions from one dialect to another. In Southern and Western High Navarrese

(§3.2.1) and in the Bizkaian variety of Arratia (§3.2.2) only the 1st and 2nd

person patients can be encoded in the dative, namely the persons ranking

highest in Silverstein’s (1976) hierarchy of animacy. The younger speakers

of those varieties seem to employ this pattern very rarely with the 3rd

person patients. In Tolosa (Gipuzkoa) and Lekeitio (Bizkaia), lower rank-

ing elements of this hierarchy can be encoded in the dative, namely the 3rd

person human NPs.

However, in this case, case-marking varies with the referentiality degree

of the NPs. As we have seen above, the definite NPs (proper names, pro-

nouns, and definite nouns) are marked in the dative case. As for the indef-

inite NPs, those receiving a referential interpretation tend to be encoded in

the dative, whereas the non-referential ones are encoded in the absolutive.

Other elements such as the quantifiers asko ‘many’, guzti ‘all’, and the

reciprocal elkar ‘each other, one another’ which may also receive a non-

referential interpretation can only be encoded in the absolutive.

Some bascologists (Rezac 2006, Fernandez 2008) have proposed that

this tendency of Basque to encode the animate patients in the dative would

have appeared by influence of the Spanish pattern. Therefore, before dis-

cussing this position, we will very briefly summarize the structure of tran-

sitive verbs in this latter language, focusing more especially on the variety

of Spanish spoken in the Basque Country.

3.3. Spanish

In Modern Spanish, the agent of transitive constructions is not morpho-

logically marked whereas the patient can be, depending on its level of

animacy and definiteness (Pottier 1968; Bleam 1999). When the patient

of a transitive construction is animate – human but also anthropomorph-

ized objects – and definite (40) it must be preceded by the preposition a,

which is homophonous with the dative case marker. As for the indefinite

NPs, those contributing a referential interpretation must be accompanied

by the preposition a (42), whereas non-referential NPs can optionally bear

it (43).

21. Inanimate patients are usually not encoded in the dative. We find only oneexception to this rule in a writing of Beriain (1621) with the verb apatu ‘tokiss’.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


368 Celine Mounole

Finally, inanimate patients are not marked at all (41):

(40) He visto a la mujer

aux.a1sg ver acc the woman

‘I have seen the woman’

(41) He visto un coche rojo

aux.a1sg ver a car red

‘I have seen a red car’

(42) Juan busca a una chica que sabe espanol

Juan look_for acc a girl that know.ind Spanish

‘Juan is looking for a girl who speaks Spanish’

(43) Juan busca (a / Ø) una chica que sepa espanol

Juan look_for acc a girl rel know.subj Spanish

‘Juan is looking for a girl who speaks Spanish’

As for the clitic system, it di¤ers from one dialect to another. In stan-

dard Spanish, the 3rd person clitics split into two groups: accusative and

dative. The accusative clitics are lo (masculine) and la (feminine), while in

the dative masculine and feminine share the same and unique clitic le. In

contrast, the variety of Spanish spoken in the Basque Country is ‘‘leısta’’,

as it employs the clitic le – etymologically dative clitic – as the animate

masculine and feminine patient clitic. So, in this variety, the animate

patient (44) and the recipient (45) – animate or inanimate – are both

represented by the clitic le, the clitics lo and la being relegated to inani-

mate patients, masculine and feminine respectively (46) (Landa 1995,

Fernandez-Ordonez 1999):

(44) le vı ayer

cl3acc see.past.a1sg yesterday

‘I saw him / her yesterday’

(45) le dı un pastel a Juan / Elena

cl3dat give.past.a1sg a cake dat Juan / Elena

‘I gave a cake to Juan / Elena’’

(46) lo comprare manana

cl3acc buy.fut.a1sg yesterday

‘I will buy it tomorrow’

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Furthermore, in leısta dialects definite (44) and indefinite animate

patients with a referential interpretation (47) can be doubled by le. How-

ever, clitic-doubling is not possible with patients which only have a non-

referential interpretation22 (48) (Landa 1995; Bleam 1999):

(47) La profesora le castigo a un nino

the teacher cl3acc punish-past.a3sg acc a child

‘The teacher punished the child’

(48) No (*le) vi a nadie

Neg cl3 acc see.past.a1sg acc nobody

‘I didn’t see anyone’ (Bleam 1999: 118)

3.4. Discussion

Without any doubt, the restrictions concerning the use of the preposition a

and the clitic doubling in Spanish are very similar to those which appear

in Basque when encoding the patients in the dative.

However, the main di¤erence between both languages lies in the fact

that until nowadays the varieties of Basque encoding the patient in the

dative only allowed it with the highest ranking elements of Silverstein’s

hierarchy – namely, the 1st and 2nd persons –, while in Spanish all the

animate person patients – 1st, 2nd and 3rd ones – can bear the preposition

a. Nevertheless, when encoding the 3rd person patients in the dative, the

same restrictions as those appearing in the case of the clitic doubling in

Spanish emerge.

Furthermore, we must remember that the dative patients appear only

in the varieties of Basque spoken in the Spanish Basque Country.23 This

suggests that if the animacy and definiteness conditioned dative patients

22. As an anonymous reviewer points out to me, not all indefinite pronounsbehave in the same way. For example, todo ‘all’ must be doubled by a clitic –the patient being animate or inanimate –: lo he hecho todo ‘I have done all’,les vı a todos ‘I saw all of them’ (vs **he hecho todo and **vı a todos).

23. Yrizar (1981, 1998a, 1998b) claimed that this pattern also occurred in thevariety of Low Navarrese, precisely in Baigorri and Azkarate, two little vil-lages of the French Basque Country. He based this idea on the first manu-script of Bonaparte on the verb of Baigorri, where under the title ‘‘Il m’a’’the trivalent auxiliary appears instead of the bivalent one. However, in thesecond manuscript on the same topic, the apparent mistake is corrected. Fur-thermore, Yrizar added that this pattern was still in use in both villages in thetwenty first century, as one informant of each village was able confirm it to

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


370 Celine Mounole

did not appear by interference with the Spanish structure, their use must at

least have been reinforced by contact with Spanish.

Nowadays, the diglossic situation is stronger than it used to be, and as

a consequence, the interferences with Spanish increase in the di¤erent

areas of the grammar. Data of Unified Basque24 seem to be a good example

of the clear influence of the Spanish structure in the transitive constructions

in synchrony. Unified Basque taught in all the schools of the Southern

Basque Country does not allow dative-marked patients. However, this

tendency is more and more perspicuous among the pupils who learn

Basque at school. In spite of all the recommendations of the prescriptive

grammars which automatically reject it, this structure is invading more

and more transitive verbs with an animate patient – Spanish loanwords

(50) as well as verbs of Basque origin (51–53) (Zubiri & Zubiri 1995: 507)

– (kritikatu ‘to criticise’, animatu ‘to encourage’, entzun ‘to hear’, gonbitatu

‘to invite’, abisatu ‘to inform’, ikusi ‘to see’, ezagutu ‘to know’, jo ‘to hit’. . .).

Moreover, just as in Spanish, it appears with 1st, 2nd and 3rd person

patients:

(49) El profesor le ha animado

The teacher him-dat aux.pres.a3sg hearten

‘The teacher has encouraged him’

(50) Irakasle-ak animatu dio

Teacher-erg heart aux.pres.a3sg.p3sg.d3sg

‘The teacher has encouraged him’

(51) Gaur kale-an ikusi dizut

today street-loc see aux.pres.a1sg.p3sg.d2sg

‘Today I have seen you on the street’

(52) Irakasle-ak jo egin dit

Teacher-erg hit foc aux.pres.a1sg.p3sg.d1sg

‘The teacher has hitten me’

him. We have spoken with his informant from Baigorri who made it clear thathe had never heard (and used) this pattern, which seemed to him ‘‘abarbarism’’. Our informants from Azkarate have also maintained that theydon’t know it. So, unfortunately based on a philological mistake and later re-inforced by misunderstanding, the description given by Yrizar seems incorrect.

24. Unified Basque is a standardized version of the Basque language. It wascreated in the 1970s by Euskaltzaindia (The Royal Academy of the BasqueLanguage). This is the version of the language used in the o‰cial texts,schools, newspapers, TV.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



(53) Neska horr-i ez diot ezagutzen

Girl this-dat not aux.pres.a3sg.p3sg.d3sg know

‘I do not know this girl’

We must remark that even if this pattern is very widespread in the spoken

language we find few examples of it in the written corpus (newspapers,

literature). This may be because of the pressure of the Standard Language

which reject it, and because the Academy of the language keeps a tight

control over the production of the language.

Anyway, these synchronic variations need to be carefully examined.

Certainly the best way to obtain more details on this phenomenon would

be the realization of oral investigations directed at di¤erent generations of

speakers in the di¤erent varieties of the language.

4. Conclusion

As has been argued at the beginning of the paper, the Basque language

has some verbs with an NP in the ergative, an NP in the dative and none

NP in the absolutive. Apart from some verbs where this pattern seems to

be lexicalized (§2.3.2), in most cases, it appears among the transitive verbs,

and it is conditioned by the syntax of the construction. Animacy and

referentiality seem to be the factors regulating the encoding in the dative

of the patients (§3.2.4).

In Basque dative-marked patients appear as soon as the sixteenth cen-

tury’s texts. In spite of that, their use seems to spread out in the nineteenth

century and even so, only in some varieties of Basque. Nowadays, this

pattern is spreading more and more at the expense of the canonical transi-

tive construction, certainly by interference of the Spanish structure.

Abbreviations

a agent

abs absolutive

acc accusative

aux auxiliary

caus causal marker

cl clitic

d dative patient

dat dative

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


372 Celine Mounole

erg ergative

foc focus marker

fut future

gen genitive

imp imperative

ind indicative

loc locative

rel relative marker

pres present

p patient

pl plural

s single argument in the absolutive

sg singular

subj subjunctive

* reconstructed

** agrammatical

References

Alberdi, Xabier2003 Euskal aditz mailegatuen erregimena : hurbilketa. [Borrowed

verbs in Basque: an approach]. In Hommage a Pierre Lafitte(1901–1985), IKER XIV. Bilbao: Euskaltzaindia, 37–60.

Arretxe, JonBasauriko euskara [The Basque of Basauri]. Basauri: BasaurikoUdala.

Bleam, Tania1999 Leısta Spanish and the Syntax of Clitic doubling. Ph. D. diss.,

Department of Linguistics, University of Delaware.Bonaparte, Louis Lucien

1991 Reprint. Le verbe basque en tableaux. In Opera Omnia vasconice.Vol. 1, 175–442. Bilbao: Euskaltzaindia. Original edition, London:Strangeways and Walden, 1869.

Creissels, Denis2006 Syntaxe generale: une introduction typologique. Paris: Lavoisier.

Creissels, Denis2008 Direct and indirect explanations of typological regularities: The

case of alignment variations. Folia Linguistica 42 (1), 1–38.Dixon, Robert M. W.

1994 Ergativity. Cambridge: Cambridge University Press.Elizalde Francisco

1735 Apezendaco dotrina kistiana uskaraz [Christian doctrine for priestsin Basque]. Pamplona.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Erdozia, Jose Luis2001 Sakana erdialdeko euskara [The Basque of Meridional Sakana].

Irunea-Pamplona: Government of Navarre.Etxeberri, Joanes

1627 Manual devotionezcoa [Devotional book]. Bordeaux: G. Millanges.Etxepare, Ricardo

2003 Valency and argument structure in the Basque verb. In A Gram-mar of Basque, Jose Ignacio Hualde and Jon Ortiz de Urbina(eds), 63–426. Berlin/New York: Mouton de Gruyter.

Etxepare, Ricardo and Oyharcabal, Benat2008 Hautazko datibo komunztadura ifar-ekialdeko euskalkietan [Dative

agreement alternations in northeastern dialects of Basque]. Ms.Fernandez, Beatriz

2008 Quirky dative objects in Basque. Power Point presented at theEuropean Dialect Syntax III, Venice (Italy).

Fernandez-Ordonez, Ines1999 Leısmo, laısmo y loısmo. In Gramatica descriptiva de la lengua

espanola, Vol. 1, Ignacio Bosque and Demonte Violeta (eds.),1317–1394. Madrid: Espasa Calpe.

Harris, Alice2002 Endoclitics and the Origin of Udi Morphosyntax. Oxford: Oxford

University Press.Hualde, Jose Ignacio, Elordieta, Gorka, Elordieta, Arantzazu

1994 The Basque dialect of Lekeitio. Supplements of Anuario del Semi-nario de Filologıa Vasca ‘‘Julio de Urquijo’’, XXXIV.

Iglesias, AitorIn progr. Igorreko hizkeraren azterketa dialektologikoa [Dialectological

study of the Basque variety of Igorre]. Ph.D. diss., Departmentof Linguistics and Basque Studies, University of the BasqueCountry.

Kapanaga, Otxoa de, Martin1656 Exposicion breve de la doctrina christiana compuesta por el P. M.

Geronimo de Ripalda de la compania de Iesus. Bilbao.Laka, Itziar

1993 Unergatives that assign ergative, unaccusatives that assign accu-sative. MIT Working Papers in Linguistics 18: Papers on caseand agreement I, 149–72. Department of Linguistics and Philos-ophy, MIT, Cambridge, MA.

Lakarra, Joseba Andoni1996 Refranes y Sentencias (1596). Ikerketak eta edizioa. (Euskararen

Lekukoak 19). Bilbao: Euskaltzaindia.Lakarra, Joseba Andoni

1997 Euskararen historia eta filologia: arazo zahar, bide berri [Historyof Basque and philology: old problems, new perspectives].Anuario del Seminario de Filologıa Vasca ‘‘Julio de Urquijo’’ 31(2), 447–535.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


374 Celine Mounole

Landa, Alazne1999 Conditions on null objects in Basque Spanish and their relation to

leısmo and clitic doubling. Ph. D. diss., Departement of Linguis-tics, University of Southern California.

Lazard, Gilbert1984 Actance variations and categories of the object. In Objects:

towards a theory of grammatical relations, Frans Plank (ed),269–292. Londres/New York: Academic Press.

Lazard, Gilbert1985 Anti-impersonal verbs, transitivity continuum and the notion of

transitivity. In Language invariants and mental operations, SeilerHansjabob and Brettschneider Gunter (eds.), 115–123. Tubin-gen: Gunter Narr Verlag.

Lazard, Gilbert1995 Le georgien: actance duale (‘active’) ou ergative? Typologie

des verbes anti-impersonnels. Sprachtypologie und Universalien-forschung 48, 275–293.

Lazard, Gilbert1998 Actancy. Berlin/New York: Mouton de Gruyter.

Lazarraga, Joanc.1564 Manuscript. Electronic Edition of Inigo Landa.

Leizarraga, Joanes1571 Iesus Krist Gure Jaunaren Testamentu Berria [New Testament of

Jesus Christ]. La Rochelle.Levin, Beth

1983 On the nature of ergativity. Ph.D. diss., MIT, Cambridge, MA.Mitxelena, Koldo

1954 Textos vascos antiguos. Un catecismo vizcaıno del siglo XVII,Boletın de la Real Sociedad Vascongada de los Amigos del Paıs10, 85–95.

Mogel, Jose Antonio1881 Peru Abarka. Durango: J. de Elizalde.

Ortiz de Urbina, Jon1989 Parameters in the Grammar of Basque. Dordrecht: Foris.

Ortiz de Urbina, Jon1991 Datibo komunztaduraren gainean [Remarks on the dative agree-

ment]. In Euskal dialektologiako kongresua, Ricardo Gomez andJoseba Andoni Lakarra (ed.), 579–588. Supplements of Anuariodel Seminario de Filologıa Vasca ‘‘Julio de Urquijo’’ XXVIII.

Ortiz de Urbina, Jon2003 Periphrastic constructions. In A Grammar of Basque, Jose Ignacio

Hualde and Jon Ortiz de Urbina (eds), 284–300. Berlin/NewYork: Mouton de Gruyter.

Ortiz de Urbina Jon & Uribe-Etxebarria, Myriam1999 Participial predication in Basque. In Memoriae L. Mitxelena

magistri sacrum, Joseba A. Lakarra & Inigo Ruiz Arzallus(eds.), 993–1012. San Sebastian: Diputacion Foral de Guipuzcoa.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40



Pottier, Bernard1968 L’emploi de la preposition a devant l’objet en espagnol. Bulletin

de la Societe Linguistique de Paris 66 (1), 345–368.Rebuschi, Georges

1983 Autour du parfait et du passif basques. Iker 2, 545–558.Rebuschi, Georges

1984 Structure de l’enonce en basque. Paris: Societe d’Etudes Linguis-tiques et Anthropologiques de France. (L’Europe de la traditionorale 3).

Rezac, Milan2006 Agreement displacement in Basque: Derivational principles and

lexical parameters. Gasteiz, University of the Basque Country,manuscript.

Sarasola, Ibon1977 Sobre la biparticion inicial en el analisis en constituyentes.

Anuario del Seminario de Filologıa Vasca ‘‘Julio de Urquijo’’ XI,51–90.

Silverstein, Michael1976 Hierarchy of features and ergativity. In Grammatical categories

in Australian Languages, Robert M. W. Dixon (ed.), 112–171.Canberra: AIAS.

Trask, Robert Larry1985 The Basque passive: a correct description. Linguistics 23, 985–

991.Trask, Robert Larry

2002 Ergativity and Accusativity in Basque. In The Nominative &Accusative and Their Counterparts, Kristin Davidse and BeatriceLamiroy (eds), 265–285. Amsterdam: John Benjamins.

Yrizar, Pedro1991 Contribucion a la Dialectologıa de la lengua vasca. 2 Vol. Donostia-

San Sebastian: Gipuzkoako Aurrezki Kutxa Probintziala.Yrizar, Pedro

1999a Morfologıa del verbo auxiliar bajo navarro occidental (estudiodialectologico). Irunea-Pamplona: University of the Basque Coun-try and Euskaltzaindia.

Yrizar, Pedro1999b Morfologıa del verbo auxiliar bajo navarro oriental (estudio dia-

lectologico). Irunea-Pamplona: University of the Basque Countryand Euskaltzaindia.

Zubiri, Ilari and Zubiri, Entzi1995 Euskal gramatika osoa. Bilbao: Didaktiker.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40


376 Celine Mounole

Authier, Gilles, and Katharina Haude (eds.). Ergativity, Valency and Voice. Berlin/New York: Mouton...

Documents

Transcript of Authier, Gilles, and Katharina Haude (eds.). Ergativity, Valency and Voice. Berlin/New York: Mouton...