1

The syntacticization of discourse1

Liliane Haegeman

Virginia Hill

Ghent University/GIST- FWO University of New Brunswick-SJ

1. Introduction

This paper examines the viability of a syntactic analysis of pragmatic markers of direct address.

They appear either on the left or the right edge of the utterance, as those illustrated in (1), from

Romanian (R) and West Flemish (WF).

(1) a. Hai, plecăm. R

hai leave-1PL

„We are leaving (injunction)‟/‟There, we are leaving now.‟

b. K‟en kennen da nie wè. WF

I en know-1SG that not wè

„I don‟t know that, you know.‟

Discourse particles have received a lot of attention in the context of discourse studies but, with

some notable exceptions (e.g. Munaro and Poletto 2004, del Gobbo and Poletto 2008), they have

so far been left aside by syntacticians. In this paper, we examine to what extent a syntactic

analysis is possible and what it would look like.

Our proposal builds on our own earlier work in this area. We refer in particular to Haegeman

(1984, 1993) for early discussion of the syntax and interpretation of some WF discourse particles, and

Hill (2007a, b,2008) for the discussion of the syntax and interpretation of Romanian particles. Inspired

2

by Speas and Tenny (2003), we will elaborate an account that postulates a speech act layer dominating

the left periphery (LP) of clauses, and which hosts the relevant pragmatic markers as well as vocative

phrases in WF and R. Differently from Speas and Tenny, we will postulate that the layers we identify are

directly related to the speech event as such, that is, the establishment of a rapport between speaker and

hearer in terms of either „attention-seeking‟ or of „bonding‟. Thus our account corroborates proposals for

the syntactic representation of the speech event made by Munaro and Poletto (2004), Sigurðsson (2004,

2011), and Giorgi (2010) a.o.

Because this is pioneering work, we will confine our analysis to a restricted set of data,

namely verb-based particles that „profile the speaker-hearer relationship‟ (Kirsner and van

Heuven 1996), and which are used as pragmatic markers, as in (1). We will show that our two

investigations along these lines, which had initially been undertaken independently of each other,

for R and WF, respectively, turned out to concur in the syntactic analysis that was elaborated,

which seems to us a promising direction2.

2. The empirical data

Both R and WF display a wide range of particles that contribute conversational pragmatic

information to the compositional reading of sentences, as discussed in Hill (2007a,b, 2008) and

Haegeman (to appear). Among these particles, we focus on those (i) that are etymologically verb-

based, and (ii) which convey the speaker‟s relation to the speech event and to the interlocutor.

The etymology and the conversational import of each particle are briefly summarized in Table 1.

3

Table 1: List of verb-based particles with pragmatic role

Lg Particle Etymology3 Conversational use Approx. gloss

R Hai Turk. hajde (=Fr. allez) injunction; evaluative „c‟mon‟, „ok‟, „really‟

Lasă lăsa „ allow/let‟ Injunction „it‟s ok‟

Uite se uita „look‟ ostensive; surprise „look here‟

WF Allé Fr. aller „go‟ Injunction „c‟mon‟

Gow gaan „go‟+weg „away‟ Injunction „c‟mon‟

nè(m) nemen „take‟ ostensivity; surprise „so there‟; „take that‟

Wè weet je „know you‟

wil je „want/will you‟

the authority of the

experience

„you know‟

Zulle zul je „shall you‟ same as wè same as wè

Zé

Zè

zien „see‟ attention drawer;

evidentiality

„look here‟

In this paper, we concentrate on nè(m), wè , zé/zè for WF, and on hai and its co-occurrence with

lasă for R. We leave the other particles (in WF and R) for future study4.

In both WF and R, some particles have two distinct conversational uses that correlate with

their distribution in the clause and/or with the intonation. For example, R hai has an injunction

reading when followed by a subjunctive clause (2a) introduced by the subjunctive marker să, but

it has an evaluative reading when followed by an indicative clause introduced by the indicative

conjunction că „that‟ (2b).

4

(2) a. Hai să citim.

hai SUBJ read-1PL

„C‟mon, let‟s read.‟

b. Hai că este nemaipomenit.

hai that is unbelievable

„It is unbelievable, really.‟

With falling intonation, WF zè occurs in final position and can be followed only by de-stressed

material. It has an evidential reading, conveying to the interlocutor that there is salient contextual

evidence for the content of the proposition expressed by the utterance. With rising intonation, on

the other hand, zé has an attention-drawing function in the speech event and can appear either

clause-initially, or clause-finally. The two particles may co-occur, with specific distribution

restrictions, summarized in (3); in our notation the accents in zé vs. zè are intended to represent

rising (zé) and falling (zè) intonation. The intonation pattern determines the position: with falling

intonation the particle must be final, with rising intonation zé can be initial (3a) or final (3d). In

the latter case the particle with rising intonation must follow that with final intonation (3e). We

return to these points in section 6.2.

(3) a. Zé, Valère is doa!

zé, Valère is there!

„Look, Valère is there!‟

b. Valère is doa zè.

Valère is there zè

„Valère is there, as you see‟

5

b‟. *Zè, Valère is doa.

c. Zé, Valère is doa zè.

d. Valère is doa zè, zé.

e. *Valère is doa zé, zè.

We interpret the various constraints on the form of the particles, their distribution and their

interpretation to indicate that the insertion of these particles in the clause is structure dependent;

hence, the necessity to formalize the behaviour of these particles in the syntax.

3. The syntactic properties of the particles

The proposal for a syntactic analysis of these particles and their relation with vocatives is in

keeping with the cartographic line of enquiry as in Cinque and Rizzi (2010), which advocates the

syntactization of the interpretive domains, extending it to the speech act layer of the clause. Early

cartographic approaches to the syntax of particles are found in Munaro and Poletto (2004), del

Gobbo and Poletto (2008), Poletto and Zanuttini (2009), a.o. For a cartographic analysis of

vocatives see Moro (2003). Though differing in implementation, our analysis is also in line with a

tradition first set out in Ross (1970) and recently taken up by Miyagawa (2012).

We believe that the justification for a syntactic analysis of discourse particles partly must

depend on how successful it is (or is not) in handling the data. If an insightful syntactic analysis

can be provided that captures the distribution and interpretation of such particles, we consider

that this in itself would offer some support for such an enterprise.

6

There are a number of empirical grounds for attempting a syntactic approach to verb-

based discourse particles. The distribution of the verb-based particles in R and in WF is

constrained by factors that also constrain the merging of elements in narrow syntax, as follows:

(i) Some of these particles display sensitivity to clause-typing: e.g., R uite and WF wè are

incompatible with interrogatives in fluent intonation:

(4) a. *Uite cine vine?

uite who comes

b. *Ee‟j gedoan wè?

have you finished wè

(ii) R verb-based particles show selectional properties; e.g., they may select a CP with a Force

head că („that‟).

(5) Hai că vine.

hai that comes

„Ok, s/he‟s coming.‟

(iii) Unlike R adverb-based particles, R verb-based particles show inflectional properties that

correlate with their distribution. For instance the particles hai and lasă show up in a number of

different forms, shown in Table 2:

7

Table 2: inflectional properties

Hai Lasă

2SG hai/haide las’/lasă

1PL Haidem

2PL hai/ haideţi lăsaţi

The inflection on these particles is also systematically constrained. In particular, the two particles

may co-occur as in (6a), but that constrains the inflectional morphology, which is then allowed

only on the second particle (6c), and not on both (6b). The word order is also fixed to hai > lasă;

the order lasă > hai as such is possible, but with a significant intonation break between the two

particles, which we interpret to mean that they belong to different utterances instead of being part

of one utterance.

(6) a. Hai-lasă nu te enerva.

hai lasa not REFL upset

„C‟mon, don‟t upset yourself.‟

b. *Haideţi-lăsaţi nu vă enervaţi.

hai-2PL-lasa-2PL not REFL upset-2PL

c. Hai-lăsaţi nu vă enervaţi.

hai lasa-2PL not REFL upset-2PL

„C‟mon, don‟t upset yourselves.‟

Similar restrictions on the appearance of inflectional morphology have been observed for verb-

8

based particles in other languages. For instance, Bazzanella (2001) provides examples for Italian.

(iv) The R clitic particle ia marks illocutionary force and can be prefixed only to verb-based

particles and to full-fledged imperative verbs; it marks injunctive mood ((7a) vs. (7b)), in the

same way mood markers signal subjunctive (i.e., să) or infinitive (i.e., a) mood, as in (7c,d).

(7) a. Ia hai, că vine. (injunctive)

ia hai that comes

„Let‟s move, s/he‟s coming.‟

b. Hai că vine. (injunctive OR informational)

hai that comes

„Let‟s move because s/he‟s coming.‟//‟Ok, s/he‟s coming.‟

c. Să vină!

SUBJ come-3SG;

„S/he better come!‟;

d. A nu deranja!

INF not come

„ „Do not disturb!‟

(v) The particles display rigid ordering restrictions. Some are initial, others are final, others may

be initial and final, but with constraints. We have already mentioned the fixed word order of R

hai > lasă above. In WF, for instance, zé / zè „see‟ can be both initial and final, with different

intonation patterns and interpretations, whereas wè is always final. Wè can be followed by zé with

rising intonation, but not by zè, with falling intonation. Neither zè nor zé can precede wè.

9

(8) a. K‟een gedoan wè zé.

I have finished wè zé

b. *K‟een gedoan wè zè.

c. *K‟een gedoan zè/zé wè .

I have finished zè/zé wè

(vi) As shown in Hill (2007b) and Haegeman (to appear), the verb-based particles interact with

the syntax of vocatives in terms of their distribution and (for R) their inflection. In both languages

examined, the vocative displays distributional constraints in relation to the particle (9), regardless

of whether it occurs in initial or final position. R displays person and number agreement between

the vocative and the verb-based particle (9e,f). This agreement relation looks similar to

subject/object-verb agreements in narrow syntax, in general. We interpret this agreement relation

as a reflex of a local Spec-head relation between the vocative nominal and the verb/particle.

(9) a. K‟een gedoan wè Valère.

I- have finished wè Valère-VOC

b. *K‟een gedoan Valère wè.

c. Zé Valère k‟een gedoan.

zé Valère-VOC I-have finished.

d. *Valère zé k‟een gedoan.

e. Lăsaţi fetelor că plecăm.

lasă -2PL girls-the-VOC that go-1PL

„There-there, girls, we‟ll be leaving.‟

10

f. *Lasă fetelor, că plecăm.

lasă -2SG girls-the- VOC that go-1PL

(vii) Further compelling support for a syntactic treatment of direct address particles is

discussed in Miyagawa (2012), who examines the distribution of allocutive agreement in

Souletin, (Basque dialect; data from Oyharçabal 1993). This agreement is determined by the

interlocutor addressed in the utterance but not otherwise expressed overtly. If agreement is

considered to be the spell-out of uninterpretable features which have to be valued by a goal with

the corresponding interpretable features, the existence of allocutive agreement presupposes the

presence of a goal controlling the agreement. This sort of data offers support for encoding

speaker and interlocutor roles in the syntax.

4. The interpretation of discourse particles

4.1. Interpretive properties

Consider the examples in (10a, b). The utterances expressing the propositional content „we

already have a medal‟ are accompanied by a discourse particle.

(10) a. M‟een al een medalie wè / zè2 // zé1 /né. WF

we have already a medal wè /zè // zé/ né

„We already have a medal.‟ (evaluation; relief, satisfaction// attention-seeking)

11

b. Hai că deja avem o medalie. R

hai that already have-1pl a medal

„We already have a medal.‟ (evaluation; relief; satisfaction)

Characterizing the interpretive value of discourse particles is not an easy task. Clearly the

particles are not truth-conditional: questioning the truth-value of the sentences in (10) („Is that

true?‟) cannot challenge the particle as such; in (10) we can only question whether we indeed

have a medal. Likewise, the particles in (10) are inaccessible to dissent or to consent, outside the

scope of negation and tense.

Although the existence of clause-typing particles has been reported in the literature

(Munaro and Poletto 2004, Poletto and Zanuttini 2009), the direct address particles are not

clause-typers. Some particles may select certain clause types, but they do not provide the typing

itself.

The particles examined here are „expressive‟ in the sense of Kratzer (1999), and have an

„illocutionary‟ or „interpersonal‟ value: they signal the speaker‟s attitude or his/her commitment

towards the content of the utterance and/or of his relation towards the interlocutor. Thus, they are

„conversational‟ in the sense that they presuppose direct speaker-interlocutor contact and would,

for instance, be inappropriate in broadcasts.

Being related to speaker and/or to hearer, and directly anchored in the speech event, the

particles can also be said to be „deictic‟. However, they are not discourse-bound: they do not need

to be used in a response to a preceding utterance.

Each particle has many slightly different overtones/values, depending on the context. For

instance, the WF particle wè can be used to accompany a statement, in which case the speaker

will use it to strongly endorse the content of his utterance by somehow appealing to his

12

experience, and (thus) to reassure the interlocutor, or even to threaten him - near-equivalent to

Engl. „you see‟, or „you know‟ (11a). However, the same particle may also be used with an

imperative (11b, c), implying that the speaker has the authority to utter the imperative – more like

the emphatic use of English do.

(11) a. Dat is nie gemakkelijk wè. WF

that is not easy wè

„It‟s not easy, you know.‟

b. Zet je mo wè.

sit you PRT wè

„Do sit down.‟

c. Komt doa nie an wè.

come there not on wè

„Don‟t you dare touch that.‟

4.2. Epistemic vigilance

Focussing on Japanese discourse connectives such as yo, kana and tte, Wilson (2010a,b) sees

them related to epistemic vigilance, rather than to the inferential process of comprehension itself.

She describes the concept of epistemic vigilance:

The hearer assesses the information re: its content and source.

13

This forces the speaker to present the information by relating it to the hearer‟s background

assumptions (e.g., his argument must show that the proposition in question follows

logically from, or is strongly supported by the hearer‟s background beliefs/assumptions);

hence, the use of discourse connectives to create the logical/evidential relations. (Wilson

2010a: handout)

The particles we are investigating have similar functions: when used to accompany statements,

WF wè, for instance, serves to display the communicator‟s competence and trustworthiness; with

imperatives, it may display either benevolence or authority, depending on the content of the

imperative. Along the lines of Wilson‟s (2010a,b) analysis, we formulate the hypothesis that the

direct address particles are geared to influencing the interlocutor‟s epistemic vigilance.

5. The syntax of the Romanian speech act particle hai

In independent work on particles from a variety of languages, Speas and Tenny (2003) have

proposed the syntacticization of speech act features: the interfacing between syntax and

conversational pragmatics is established through a functional predicative structure in the same

way that the argument structure of a lexical verb is projected. Due to their categorial feature [V],

the verb-based particles we examine here may be argued to provide empirical support for such

hypotheses. In this section, we recast Speas and Tenny‟s (2003) hypothesis in a cartographic

representation of the LP (Rizzi 1997, 2004), whose aim is to maximally syntacticize the

interpretive domains (cf. Cinque and Rizzi 2010).

14

5.1. General properties

Etymologically, R hai/haide/haideţi is a verb-based particle, which is said to be derived from the

frozen imperative form of Turkish (h)ajde („go!‟; let‟s go!) (Tschizmarova 2005).5 The particle

hajde has been adopted in the entire Balkan area as an invariable pragmatic marker, but in R it

got enriched with person and number agreement inflection.

In R, hai is used in two main environments, as shown in (2), repeated as (12) for

convenience: as an injunctive (12a), and as an evaluative/evidential (12b).

(12) a. Hai să citim.

hai SUBJ read-1PL

„C‟mon, let‟s read.‟

b. Hai că este nemaipomenit.

hai that is unbelievable

„Really, it is unbelievable (I say!).‟

Hill (2007b) argues that, in both these contexts, hai/haide does not qualify as a verb, and

so it does not project a vP or a TP structure. As an injunctive marker (12a), hai does not affect the

semantics of the host verb (e.g., unlike invariable modals or causatives), but only intensifies or

attenuates the strength of command, and qualifies the relation between speaker and hearer (e.g.,

in terms of familiarity, endearment, etc.). We conclude that hai is not an independent imperative

verb, but rather is an injunctive element with a pragmatic function.

15

In its second use in (12b), hai takes a clause introduced by că „that‟ as its complement.

Evidential and epistemic „adverbs‟ and verbs tend to select CPs introduced by că „that‟ in mono-

or bi-clausal structures (Hill 2012)6. Thus, for this particular pattern, one might be tempted to

propose that hai has become re-analyzed (through some form of lexicalization) as a non-

thematic/non-raising verb, such as (se) părea („seem‟). However, the diagnostics developed in

Hill (2007b) – which we cannot reproduce because of space limitation – show that that is not the

case. In particular, hai cannot be regularly embedded (unlike the analogues of „seem‟), and it

allows for imperative morphology, which „seem‟ type verbs rule out.

We are, thus, faced with a paradoxical situation: where hai is not a verb, it has no TP

domain, but it can display the morphology of imperative verbs. That is, hai can have the endings

for phi-features and mood (i.e., haideţi „haide+2PL.IMP‟), which are a property of T.

5.2. Matching position and interpretation

An important observation is that hai is underspecified in the lexicon. As seen for (12), its

interpretation depends on the type of complements it displays and on its position in relation to the

utterance. Accordingly, hai may function as an enhancer for illocutionary force or as an assessing

tool. These functions, however, are not interchangeable: the illocutionary reading depends on the

presence of an imperative verb (12a), whereas the evaluative reading depends on the presence of

a CP introduced by the complementizer că („that‟). These data lead to the hypothesis that the

value of hai is read off the syntactic configuration7.

Furthermore, even in relation to the same type of verb, the interpretation of hai varies

depending on its position in the clause. Compare (12a), with clause-initial hai, and (13), with

16

clause-final hai. The verb of the main utterance is always in the imperative, but in (12a) the

speaker feels entitled to give a command, whereas in (13) the speaker is negotiating the event

with the addressee, mitigating the tone by appealing to the interlocutor.

(13) Să citim, hai.

SUBJ read-1PL hai

„Let‟s read, please.‟

The same contrast emerges from (12b) and (14), where the main verb is maintained in the

indicative. First, the complementizer că „that‟ is unavailable when hai is clause-final (14a).

Without că, the clause is marginally acceptable (14b), but the interpretation is different from that

in (12b). Whereas in (12b) the speaker expresses a personal assessment with mirative force, in

(14b) the speaker is making a reassuring statement, for which hai is only a marginal candidate (a

more appropriate particle in this context would be, for instance, zău „cross my heart‟).

(14) a. (*Că) este nemaipomenit, hai.

that is unbelievable hai

b. #Este nemaipomenit, hai.

is unbelievable hai

„It is unbelievable, really.‟

The central point is that the interpretation of hai varies, rather than being fully fixed in the

lexical semantics of this particle. While the distribution of hai in this mapping is limited, the

combination between its location, complementation and the predication it belongs to allow for

17

significant proliferations in the compositional meanings, which also explains the list of numerous

pragmatic values associated with hai.8

5.3. Co-occurring elements and ordering restrictions

In order to sort out the syntactic mapping of hai, we will examine its position in relation to other

elements that occur in speech acts; notably, the presence of vocatives and of particles expressing

the speaker‟s point of view.

In (15), we show the co-occurrence of hai with the particle vai („ah‟), which is a

„lamenting‟ marker expressing the speaker‟s feelings. Vai is a speaker- oriented particle, while

hai is a hearer-oriented particle of direct address. There are ordering restrictions on the

sequencing of vai and hai: (i) vai may only precede hai (15a /15b). (ii) The vocative phrase is

always adjacent to hai, either preceding it (15a) or following it (15c). (iii) Though a vocative may

also linearly precede vai, as in (15d), the intonation pattern of such examples is markedly

different, with a clear prosodic break between the vocative and vai. We conclude from this that in

(15d) we are dealing with two utterances. Note that in (15) vai is optional, and the sequence

Vocative > hai or hai > Vocative occurs independently of vai. 9

(15) a. Vai Dane hai că nu te cred.

vai Dan- VOC hai that not you believe-1SG

„Ah, Dan, c‟mon, I don‟t believe you.‟

b. *Dane, vai hai că nu te cred.

Dan- VOC vai hai that not you believe-1SG

18

c. Vai, hai Dane că nu te cred.

vai hai Dan-VOC that not you believe-1SG

„Ah, c‟mon, Dan, I don‟t believe you.‟

d. Dane #. Vai, nu te cred.

Dan-VOC, vai not you believe-1SG

„Dan, ah, I don‟t believe you.‟

There are two relevant differences between the morpho-syntax of vai and that of hai: (i) Only hai

may display person/number agreement with the vocative (15e), and (ii) only hai can license a că

(„that‟) indicative clause ((15a, c) versus (15f)). We take the observed distribution of the vocative

in (15a-d) as well as the observed restrictions on agreement between the vocative and the

particles to indicate that: (i) the vocative is in a local relation with hai, not with vai, and (ii) the

că „that‟ CP is in a sisterhood relation with hai, but not with vai. Thus, vai helps us to determine a

hierarchy for particles; however, it is not a particle of direct address, so we do not further analyze

its properties and syntactic behavior.

(15) e. Vai(*ţi) fetelor haideţi că nu e bine.

vai (2PL) girls hai.2PL that not is good

„Ah-ah, girls, this is not good, really.‟

f. Vai (*că) nu te cred, Dane, hai.

vai that not you believe Dan- VOC hai

„Ah, c‟mon Dan, I don‟t believe you.‟

19

5.4. The cartography of discourse particles and of vocatives

Hill (2007b) argues that discourse particles are in a hierarchical relationship with each other. The

vocative is licensed in a Spec-head relation with hai, as is shown in (16). Her analysis is in line

with and indebted to the proposal in Speas and Tenny (2003) to the effect that the speech act

layer is mapped as a predicative structure encoding the conversational set-up (who is the speaker,

who is the addressee and what is the power relation in the conversation), which is conveyed

compositionally in the interpretation of the utterance. In terms of their analysis, the speech act is

computed in the same way as the functional „little v‟.

Based on Speas and Tenny‟s proposal,10

we assume that ForceP in the sense of Rizzi

(1997) is selected by an articulated Speech Act projection headed by the Speech Act (SA) head,

with a layered articulation, much as is the case with transitive verbs which project a VP shell and

a vP shell. In Hill‟s articulation the SA projection is articulated around two shells: the lower

„hearer‟ shell (SA), analogous with VP, and the higher „speaker‟ shell (sa), analogous with vP.

We tentatively assume, with Hill (2007b) that the particle hai is inserted in the lower SA head,

which is directly associated with the „hearer‟. SA takes two arguments: its „direct object‟, the

ForceP complement (which may be introduced by că „that‟)11

; and its „indirect object‟, the

vocative phrase, which is the specifier of SA. This structural articulation captures the privileged

relation between vocatives and injunctive particles, which had already been noticed in the

descriptive grammars of languages from various genetic groups (e.g., Schadeberg 1990 for

Umbundu).

(16) [saP [sa] [SAP VOCATIVE [SA hai] [FORCEP Utterance/ că …]]]

20

The unmarked order is that in which the vocative precedes hai, with hai in SA. When the

vocative is preceded by hai we assume that this is due to hai moving up to the higher head sa.

The interpretation of the utterance is determined by the relative positions of the particle and the

vocative: (16), in which the vocative precedes hai, foregrounds the vocative, for attention-

drawing, whereas the order in which hai precedes the vocative foregrounds hai, conveying

exasperation or enhanced mitigation. The pitch is higher on the foregrounded item. Hence, the

movement maps a different value for the inter-personal feature.

The interpretive differences indicate that the movement of the particle from SA to sa in R

is not optional. Of course, it would be desirable that the trigger and constraints for the head

movement of the particle be clarified. At this point we have two suggestions for looking at this,

and we intend to explore these in future work. On the one hand, as suggested in Hill (2007a,b,

2008), the movement of the head SA to sa is reminiscent of the way V moves to v. In this view

particle syntax is tied to verb syntax, which ties in with the fact that many (though not all, cf. note

3) discourse particles on the periphery of the utterance are etymologically related to verbs. On the

other hand, the derivation proposed here is also reminiscent of the derivation of PPs as that

proposed by Kayne (1999), according to which both the preposition and its complement are

merged on the clausal spine and in which the „constituency‟ of the PP is derived through

subsequent movement of the complement and the preposition.12

Utterance-final hai is derived by the movement of the complement ForceP to the specifier

of the head sa. Alternatively, the vocative itself may also be fronted (in which case the word

order is again Voc > hai, although it is obtained at a higher level). Such movement is argued to

occur when the address is a call (Hill and Stavrou 2012).

21

Data with co-occurring particles such as (6) either suggest that particles may cluster in

one head, in the same way that in many languages clitics can be seen to cluster; or, alternatively,

that more than one particle layer may be available. The latter option is the conclusion reached

independently by Haegeman (to appear) for WF and which we discuss in the next section. Having

discussed the syntax of the WF particles we briefly examine its applicability to the R particles in

section 6.5.

6. The syntax of West Flemish particles

6.1. General properties

WF verb-based particles are either clause-initial or clause-final, and cannot appear in the middle

field of the clause. This is shown for né in (17).

(17) Né, doet (*né) da (*né) mo mee!

né do né that né PRT with

„Here you are, you can have this!‟

The availability of the particles is partly dependent on clause-typing. For instance, the particle wè

is incompatible with questions or sentences with question intonation. This suggests that wè has

selectional properties w.r.t. its complement. For similar selectional restrictions concerning da see

Haegeman (1993).

(18) a. Een z‟al een medalie, wè?

22

Have they already a medal, wè?

b. Ze zoun al een medalie een wè?

Furthermore, the WF particles qualify as main clause phenomena: for example, initial né cannot

occur at the LP of the complement clause (19a) and final né in (19b) is computed as a main

clause particle related to the speaker rather than as an embedded particle related to the „embedded

speaker‟:

(19) a. *Je zei [né dat da roare was.]

he said né that that strange was

b. Je zei [dat da roare was] né.

When initial, the particles precede the first constituent of a V2 clause suggesting that they are

„outside‟ the regular CP/ForceP layer, which accounts for their restricted distribution13

.

(19) c. Né, dienen medalie een me a.

né that medal have we already

„There we are, the medal is ours.‟

These WF data offer additional support for the hypothesis elaborated in Section 5 that the particle

heads a layer of structure above CP, and selects CP. The relevant layer encodes properties of the

speech act.

WF clauses may display up to two utterance-peripheral particles at a time, with either a

combination of an initial particle and one final one, or with a combination of two final particles.

23

The combined particles are subject to specific ordering constraints. (20) illustrates some of the

combinations.

(20) a. Né, men artikel is gedoan wè (*zé).

né, my article is finishedis gedoan wè (*zé)

b. *Men artikel is gedoan wè (*zè) (né).

Based on restrictions on the number of co-occurring verb-based particles, as in (20), and, in

particular, on the observation that an initial particle (né) may co-occur with a final one (wè, zè),

we conclude that two speech act layers must be available. Given the restriction on the number of

particles in WF, we provisionally postulate two articulated speech act projections. The observed

distributional restrictions on the WF particles provide further clues concerning the internal

organization of the functional domain that encodes speech event properties.

6.2. Matching position and interpretation

Clause-initially, both zé (with rising intonation) and né draw attention to the content of the

utterance. Clause-final zè, with falling intonation, expresses epistemic vigilance, qualifying the

speaker‟s source of evidence for his utterance. The particle zè signals the speaker‟s authority with

respect to the content of the utterance and with respect to the addressee, thus inspiring more

confidence in his interlocutor, and making him more likely to pay attention to it and (where

relevant) to act upon it (21a). Wè is only clause-final and also signals the speaker‟s authority

24

w.r.t. the contents of the utterance and the addressee (21b). Final wè cannot co-occur with final zè

(21c):

(21) a. Zé men artikel is gedoan, zè.

Zé my article is finished zè

„Look, my article is finished, there you are.‟

b. (*Wè) men artikel is gedoan wè.

(*wè) my article is finished wè

„My paper is finished, you know.‟

c. Men artikel is gedoan wè (*zè)/(* zè) wè.

my article is finished wè (*zè)/ (*zè) wè

Attention-seeking né has the same distribution as initial zé, and the two cannot co-occur:

(22) Né (*zé) men article is gedoan.

6.3. Co-occurrence in final position and ordering restrictions

With rising intonation, the particles né and zé can be found in initial position, but they can also be

found in final position. In both contexts only one of the two particles can occur. When final, né

or zé can co-occur with just one particle of the „final‟ type, i.e. wè or zè. When particles co-occur

in final position, the particle with falling intonation precedes that with rising intonation, and there

may only be one particle of each type.

25

(23) a. Men artikel is gedoan wè zé/*zé wè.

b. Men artikel is gedoan wè né/(*né wè)

c. Men artikel is gedoan zè zé/*zé zè.

d. Men artikel is gedoan zè né/(*né zè)

6.4. The cartography of the WF particles

6.4.1. Two SAP

Pursuing our hypothesis that the particles of direct address head functional projections, the

distribution and the ordering restrictions on the WF particle can be derived on the basis of a

structure such as that in (24), in which there are two projections for particles, provisionally

labeled SAP1 and SAP2. The linearly final position of the particles in SAP1 follows from

movement of CP to the specifier position within the particle phrase. (24) offers a first schematic

account, to be refined presently. The final position of wè in (24b) is derived by (obligatory)

movement of the complement CP to its specifier position; the final position of né in (24c) is

derived by the (optional) movement of SAP2, the projection headed by (final) wè, to the specifier

of né in SA1. Observe that the derivations in (24) violate the anti-locality constraint which

disallows the movement of a complement to the specifier of the same projection (Abels 2003,

Grohmann 2003). The analysis developed in section 6.4.2 eliminates this problem.

(24) a. [SAP1 né [SAP2 wè [CP ….]]]

b. [SAP1 né [SAP2 [CP ….] wè [CP ….]]]

c. [SAP1 [SAP2 [CP ….]wè [CP ….]]né [SAP2 [CP ….]wè [CP ….]]]

26

6.4.2. Vocatives and SAP

A further complication needs to be introduced here. As was the case in R, WF particles interact

with vocative phrases, suggesting that here too, the two SA projections must be further

articulated in a structure with two shells, saP and SAP. In WF, vocatives always follow the

particle, whether this be initial (25a) or final (25b).

(25) a. Né Valère, men artikel is gereed (wè).

b. *Valère né, men article is gereed (wè).

c. (Né) Men artikel is gereed wè Valère.

d. *(Né) Men artikel is gereed Valère wè.

The different positions of the vocative – initial or final - coincide with a difference in

interpretation: the initial vocative has an „appeal‟ or attention-seeking function, aiming at

establishing a discourse relation; the final vocative consolidates the already established relation

of the speaker with an „addressee‟ (see also Schegloff 1968 among others).

In line with Hill (2007b), Haegeman (to appear) proposes that, in order to accommodate

the vocatives, each of the two SA projections be articulated following our proposal in section 5:

that is: each speech act projection is a shell structure, modelled on the structure of vP/VP, where,

analogously to a ditransitive V, the SA head selects a complement (ForceP) and has an „indirect

object‟ (the vocative phrase) as its specifier. The resulting SAP is dominated by a higher shell

„saP‟ (analogous to vP), with the particle moving from the lower head SA to sa, the higher head.

Thus the structure in (16) is updated to include a representation of the vocative syntax and the

sequence né –vocative – CP is schematically represented as follows:

27

(26) a. [saP [sa né [SAP VOCATIVE [SA né ….[ForceP…

Furthermore, given that two particles can co-occur, Haegeman (to appear) proposes that there are

two speech act projections, each articulated in the double shell structure, and each with a

specialized discourse function, where sa1P dominates sa2P. (26b) summarizes our conception of

the fully articulated speech event layer: it includes an attention-seeking layer (sa1/SA1P) and a

consolidating/bonding layer (sa2/SA2P):

(26) b. [sa1P [sa1 né] [SA1P VOC [SA1 né ]

[sa2P [sa2 wè] [SA2P VOC [SA2 wè ] [ForceP]]]]]]]]

(27a), with initial né and final wè is derived as in (27b-c). In (27b) wè is merged as SA2, it selects

ForceP as its complement and the vocative Valère as its specifier. Then sa2 merges with SA2P:

the particle wè moves from SA2 to sa2 and attracts ForceP to its specifier, leading to the linear

sequence in which the clause precedes the particle, which in turn, precedes the vocative. As

shown in (27c), the particle né is then merged as SA1, the lower head of the higher speech act

projection, selecting saP2 as its complement. Again, né head-moves from SA1 to sa1.

(27) a. Né, k‟een a gedoan wè Valère.

b. [saP2 ForceP [sa2 wè] [SAP2 Valère [SA2 wè] [ForceP]]

c. [saP1 [sa1 né] [SAP1 [SA1 né]

[saP2 ForceP [sa2 wè] [SAP2 Valère [SA2 wè] [ForceP]]]]

28

The utterance-final position of né in (28a) is derived on the basis of (27c), by further moving

saP2 in (27c) to the specifier of sa1, headed by né, as in (28b).

(28) a. K‟een a gedoan wè Valère né.

b. [saP1 [saP2 ForceP [sa2 wè] [SAP2 Valère [SA2 wè] [ForceP]]

[sa1 né] [SAP1 [SA1 né] [saP2]]]

6.5. Romanian particles

Having developed a more articulated system on the basis of WF, let us now apply the system to

some R data. Like WF, R has co-occurring particles that display ordering restrictions, as shown

in (6). We discuss examples (29)-(35) here: for each example we provide the derivation in terms

of the system developed in section 6.4.

In (29), the vocative is attention-drawing and by hypothesis merges in the high

SpecSAP1. The particles stay in their merge position, the lower one, lasă, selecting the utterance

as its complement. Differently from WF, the R particles need not head-move to the sa shell. At

this point we only observe this and hope to be able to account for this in the future.

(29) a. Dane, hai lasă nu te enerva.

Dan.VOC hai lasă not REFL upset

„Dan, c‟mon, dont get upset.‟

b. [saP1 [sa1 ][SAP1 Dane [SA1 hai ] [saP2 [sa2 ] [SAP2 [SA2 lasă] [ForceP]]]]]

29

In (30), the vocative has a bonding reading, and is merged in the lower vocative position,

SpecSAP2. The particles remain in the same positions as in (29).

(30) a. Hai Dane lasă [nu te enerva.]

hai Dan.VOC lasă not REFL upset

b. [saP1 [sa1 ] [SAP1 [SA1 hai] [saP2 [sa2 ] [SAP2 Dane [SA2 lasă] [ForceP]]]]]

In (31), the bonding vocative Dane is in the lower vocative position. In this example, the lower

particle lasă precedes the vocative, and we assume that it has moved to sa2. The displacement of

the particle is accompanied with emphasis on the particle (e.g. conveying exasperation). In (31)

we also place the particle hai as the head of the lower shell of saP1, though the data obviously

would also be compatible with it moving to the head of the higher shell, sa1.

(31) a. Hai lasă Dane nu te enerva.

hai lasă Dan.VOC not REFL upset

b. [saP1 [sa1 ] [SAP1 [SA1 hai] [saP2 [sa2 lasă ][SAP2 Dane [SA2 lasă ] [ForceP]]]]]

In (32), ForceP precedes lasă as well as the bonding vocative. To derive this order we propose

that ForceP moves to Spec,saP2. This movement has an effect on the information structure: the

moved ForceP is foregrounded. We tentatively assume that this displacement is thus related to a

focusing movement.

(32) a. Hai nu te enerva Dane, lasă.

hai not REFL upset Dan.VOC lasă

30

b. [saP1 [sa1 ] [SAP1 [SA1 hai ] [saP2 ForceP [sa2 ] [SAP2 Dane [SA2 lasă ] [ForceP]]]]]

In (33), the vocative is initial, and precedes lasă, and hai is in final position. Though initial, the

vocative does not carry the high pitch for attention-drawing and retains its bonding function

associated with SAP2. We assume therefore that the vocative occupies SpecSAP2. To derive the

final position of hai we assume that saP2 moves to the initial position, Spec,saP1. As a result, the

attention is focused on the entire proposition, while final hai conveys empathy.

(33) a. ? Dane, lasă, nu te enerva, hai.

Dan.VOC lasă not REFL upset hai

b. [saP1 [saP2 [sa2 ] [SAP2 Dane [SA2 lasă] [ForceP]]] [sa1 ] [SAP1 [SA1 hai] [saP2 [sa2 ]

[SAP2 Dane [SA2 lasă] [ForceP]]]]]

(34) is slightly degraded. In this example the vocative is merged in the vocative position of the

higher SAP1, in which it has an attention-drawing function. The final position of the particle lasă

is derived by moving ForceP to Spec,saP2 with a foregrounding effect. At this point it is not clear

to us what causes the degradation of this example.

(34) a. ? Dane, hai nu te enerva, lasă.

Dan.VOC hai not REFL upset lasă

b. [saP1 [sa1 ] [SAP1 Dane [SA1 hai ] [saP2 ForceP [sa2 ] [SAP2 [SA2 lasă] [ForceP]]]]]

31

In (35), both the low particle and ForceP have moved to higher positions. The double movement

is a marked option and the sentence is awkward, although acceptable. Again it is not clear to us

yet what causes this degradation.

(35) a. ??Hai nu te enerva, lasă, Dane.

hai not REFL upset lasă Dan.VOC

b. [saP1 [sa1 ] [SAP1 [SA1 hai] [saP2 ForceP [sa2 lasă]

[SAP2 Dane [SA2 lasă] [ForceP]]]]]

Though the articulated structure developed here for WF can be put to use to derive the positions

of the R particles, there are differences between WF and R particles. For instance, the order in

(36a) is ungrammatical in WF, whereas the corresponding order are grammatical R in (36b).

(36) a. *(Né) Men artikel is gereed Valère wè. WF

b. (Hai) nu te enerva Dane, lasă. R

The different word order patterns can often be related to the position of the particle. We proposed

that in WF particles are generated in the lower shell of the speech act projection (SA1, SA2) and

move to the higher head (sa1, sa2), whereas in R this movement though available, as shown

above, is not generalised. It is not clear at this point what accounts for the obligatory head

movement of the Flemish particle. One option is that the WF particles are somehow syntactically

deficient or „light‟ and have to be licensed in the higher head.14

We obviously need to examine

this variation further in the light of a wider-ranging comparative context which includes

additional languages. Observe, though, that, given the very strong interpretive similarity between

32

the particles, an account that is purely cast in semantic or pragmatic terms will not at first sight be

able to handle all the variation in distribution of particles with the same interpretive import.

Assuming, for instance, that R hai and WF né have the same semantic-pragmatic function, their

different positions could not plausibly be made to follow merely from their interpretation. In

order to fully substantiate this point we need to increase our data base of particles to be able to

compare like with like cross-linguistically. We also need to develop ways of determining the

precise shades of the interpretation of the particles.

7. General conclusions

In this paper we have examined to what extent the distribution and interpretation of so called

discourse particles can be captured in terms of a syntactic model along the lines of the

cartographic approach (Cinque and Rizzi 2010). We have concentrated on particles of direct

address in Romanian and in West Flemish, and their relation with the vocative.

Data from two unrelated languages support the view that particles of direct address are

computed syntactically at the edge of clauses, outside what is usually referred to as the CP

domain. Evidence comes from various restrictions on the particles, which are determined by

clause type, morpho-syntactic manipulations for agreement and mood marking, and from their

distribution in relation to their complement CPs.

In keeping with cartographic guidelines we have elaborated an articulated structure that

maps the speech act domain as a layered shell structure, similar to that elaborated for vP/VP. In

this respect, the analyses proposed for R and WF coincide in their general outlines.

33

At a finer-grained level, Haegeman‟s (to appear) analysis of WF suggests that the

articulation of the speech act is two-layered (vs. the one-layered proposal in Hill 2007b). This

two-layered structure is better able to represent the relevant distribution of particles, vocatives

and the complement clause. Thus, inclusion of the WF data in the analysis has led to a refinement

to Hill‟s original proposal: the higher speech act layer encodes the setting up of the discourse

layer („attention-seeking‟), while the lower layer encodes the consolidation of the discourse

relation („bonding‟).

Further research will have to shed light on the cross-linguistic differences in the

distributions of the particles an in particular what triggers head movement of the particles and/or

the propositional complements.

References

Abels, Klaus (2003). Successive cyclicity, anti locality and adposition stranding. University of

Connecticut, PhD. diss.

Bazzanella, Carla (2001). „I Segnali Discorsivi‟, in L. Renzi, G. Salvi and A. Cardinaletti (eds.),

Grande Grammatica Italiana di Consultazione. Vol. III, Bologna: il Mulino, 225-257.

Cinque, Guglielmo. (1999) Adverbs and Functional Heads. Oxford: Oxford University Press.

Cinque, Guglielmo, and Rizzi, Luigi (2010). „The Cartography of Syntactic Structures,‟ in B.

Heine and H. Narrog (eds), The Oxford Handbook of Grammatical Analysis, Oxford:

Oxford University Press, 51-65.

Craenenbroeck, Jeroen van, and Haegeman, Liliane (2007). „The Derivation of Subject Initial

V2‟, Linguistic Inquiry 38: 167-178.

Derolez, Réné and Anne-Marie Simon-Vandenbergen (1988). „Gotich saihw and sai, with some

34

Notes on Imperative Interjections in Germanic. In M.A. Yazayery and W. Winter, eds.

Languages and Cultures, Studies in Honor of Edgar C. Polomé, Berlin: Mouton-De

Gruyter, 97-110.

Giorgi, Alexandra (2010). About the Speaker. Toward a Syntax of Indexicality. Oxford/New

York: Oxford University Press.

Gobbo, del Francesca, and Poletto, Cecilia (2008). „A Typology of Sentential Particles‟, Giornata

di Dialettologia, Università di Padua, 20th June 2008. Handout.

Grohmann, Kleanthes (2003). Prolific Domains: on the Anti-locality of Movement Dependencies.

Amsterdam: John Benjamins.

Haegeman, Liliane (1984). „Interjections and Phrase Structure‟, Linguistics 22: 41-49.

Haegeman, Liliane (1993). „The Interpretation of the Particle da in West Flemish‟, in D. Wilson

and N. Smith (eds), Relevance theory (2). Lingua 90: 111-128.

Haegeman, Liliane. (1996). „Verb Second, the Split CP and Initial Null Subjects in Early Dutch

finite clauses,‟ Geneva Generative Papers 4: 133-175.

Haegeman, Liliane (to appear). „The Cartography of Discourse Markers in West Flemish,‟ Studia

Linguistica.

Hill, Virginia (2007a). „Romanian Adverbs and the Pragmatic Field‟, The Linguistic Review 24:

61-86.

Hill, Virginia (2007b). „Vocatives and the Pragmatics-syntax Interface‟, Lingua 117: 2077-2105.

Hill, Virginia (2008). „Pragmatics Markers as Syntactic Heads: a Case Study from Romanian‟, in

K. Grohmann and P. Panagiotidis (eds.), Syntaxfest Cyprus Conference., Newcastle-upon-

Tyne: Cambridge University Scholars Press, 237-265.

Kayne, Richard. S. (1999). Prepositional Complementizers as Attractors‟, Probus 11, 39-73.

35

Kirsner, Robert, and van Heuven, Vincent. (1996). „Boundary Tones and the Semantics of the

Dutch Particles hè, hoor, zeg and joh‟, in C. Cremers and M. den Dikken (eds),

Linguistics in the Netherlands 1996., Amsterdam: John Benjamins, 133-146.

Kloots, Hanne (2007). „ “Allee, wat was het weer?” Het Tussenwerpsel allee in Spontaan

Gesproken Standaardnederlands‟, Over Taal 46, 1: 17-19.

Kratzer, Angelika (1999). „Beyond „Oops‟ and „Ouch‟. How descriptive and expressive meaning

interact‟. Paper presented at the Cornell Conference on Theories of Context Dependency,

26 March 1999.

Miyagawa, Shigeru (2012). „Agreements that occur mainly in Main Clauses‟, in L. Aelbrecht; L.

Haegeman; R. Nye (eds), Main Clause Phenomena: New Horizons. Amsterdam: John

Benjamins, 79-112.

Moro, Andrea (2003). „Notes on Vocative Case: a Case Study in Clause Structure‟, in J. Quer et

al. (eds.), Romance Languages and Linguistic theory. Amsterdam: John Benjamins,. 247-

261.

Munaro, Nicola, and Poletto, Cecilia (2004). „Ways of Clause-typing‟, Rivista di Grammatica

Generativa 27: 87-105.

Oyharçabal, Bernard (1993). „Verb Agreement with Non arguments: On Allocutive Agreement‟,

in J. I. Hualde and J. Ortiz de Urbina (eds.), Generative Studies in Basque Linguistics.

Amsterdam: Benjamins, 89-114.

Poletto, Cecilia, and Zanuttini, Raffaella (2009). „Sentential Particles and Remnant Movement‟,

in P. Benincà and N. Munaro (eds.), Mapping the Left Periphery. New York: Oxford

University Press. to appear.

Rizzi, Luigi (1997). „The Fine Structure of the Left Periphery‟, in L. Haegeman, (ed.), Elements

of Grammar: Handbook of Generative Syntax. Dordrecht: Kluwer, 281–339.

36

Rizzi, Luigi (2004). „Locality and Left Periphery‟, in A. Belletti, Structures and Beyond. The

Cartography of Syntactic Structures 3, New York: OUP, 223-251.

Ross, John R. (1970). „On Declarative Sentences‟, in R. A. Jacobs and P. S. Rosenbaum (eds),

Readings in English Transformational Grammar. Waltham, Mass.: Ginn, 222-277.

Schadeberg, Thilo (1990). A Sketch of Umbundu. Köln: Rüdiger Köppe Verlag.

Schegloff, E. A. (1968). „Sequencing in Conversational Openings‟, American Anthropologist 70:

1075-1095.

Sigurðsson, Halldór Ármann (2004). „The syntax of Person, Tense, and Speech Features‟, Italian

Journal of Linguistics / Rivista di Linguistica 16: 219-251.

Sigurðsson, Halldór Ármann (2011). „Conditions on Argument Drop‟, Linguistic Inquiry 42:267–

304.

Speas, Peggy, and Tenny, Carol (2003). „Configurational Properties of Point of view roles‟, in

A.-M. Di Sciullo (ed), Asymmetry in Grammar. Amsterdam: John Benjamins, 315-344.

Tschizmarova, Ievelina (2005). „Hedging Functions of the Bulgarian Discourse Marker xajde’,

Journal of Pragmatics 37: 1143–1163.

Wilson, Deidre (2010a). „Word Meaning, Concepts and Procedures‟. Conference on Word

meaning. Universityof Oslo, 2-3.9.2010.

Wilson, Deidre (2010b). „Modality and the Conceptual-Procedural Distinction‟. Manuscript

UCL/ CSMN, University of Oslo.

Zanuttini, Raffaella (2008). „Encoding the Addressee in the Syntax: Evidence from English

Imperative Subjects‟, Natural Language and Linguistic Theory 26: 185-218.

37

1 Liliane Haegeman‟s research is funded by the FWO 2009-Odysseus-Haegeman-G091409.

We thank the audience of ONLI for their comments. Thanks to two reviewers for their

very useful comments, which have helped us in the revisions of the paper. Needless to

say, they can be held responsible for the way we have used their input.

2 Though we concentrate on verb-based particles, we assume that other direct address

particles will have a similar distribution to those discussed here and will be amenable to a

similar analysis. Cf. Haegeman (1993) for the WF particle da.

3 Observe that while the large majority of these particles are etymologically related to verbs

of movement or perception, this cannot be generalised. WF né for instance is probably

related to the verb nemen, „take‟, and the WF particle da (Haegeman 1993), which is not

discussed here but shares the distribution of utterance final particles such as wè is not

obviously related to any verb. Similarly, R. lasă is not related to a verb of movement or

perception. However, it remains true cross-linguistically that verbs of movement and

perception verbs have often grammaticalized into particles and this point certainly merits

further study.

4 As pointed out by an anonymous reviewer, many of the particles seem to display the

imperative form, which in some Romance languages is the same found in compounding;

maybe this is the „barest‟ possible form, and this is why we find an imperative.

Alternatively, the imperative form is related to the direct address to the interlocutor

(Zanuttini 2008) and are encoded in the discourse projections. For discussion of

imperatives as particles see also Derolez and Simon-Vandenbergen (1988).

5 It is similar to Flemish allé, which derives from French allez, the imperative of aller, „go‟.

38

6 Our proposal is that the features associated with evaluation and evidentiality are not

lexical properties of hai, but of the configuration in which it is used, i.e., with „that‟ CP

complement. Nuances related to the interpretation (e.g., mirativity) arise compositionally,

depending on the lexical properties of other items involved (e.g., the adjective

nemaipomenit „unbelievable‟ in (12b)).

7 The insertion of hai may yield a variety of pragmatic readings (e.g., irony, compassion,

impatience, hedging etc.; Tschizmarova 2005). Future work should aim to sort out to what

extent syntax is responsible for such variety (see also Speas and Tenny 2003 for

discussing this point).

8 We refer the reader to the discussion of the various hedging values in Tschizmarova 2005:

1148).

9 An anonymous reviewer informs us that the order corresponding to the R sequence

vocative > hai appears in Greek as well with the particle ade (see also the word order in R

(29)):

(i) Jani ade parata mas.

John.VOC ade leave.IMP us

„John do leave us alone‟ (in the sense of „stop pestering us‟)

10 Speas and Tenny equate the SA layer with the high modals in Cinque‟s (1999) proposal.

We remain non-committal concerning this particular aspect of the analysis.

11 Observe that the ForceP selected by hai itself has a regular left periphery, with Topic and

Focus positions available to the right of the complementizer că „that‟:

(i) Hai că la mare numai Maria poate merge

hai that to sea.TOP only Maria.FOC can go

39

„Ok, to the sea, only Maria can go.‟

Data such as these clearly confirm that the particles we are concerned with here are above

the left periphery in the sense of Rizzi (1997).

12 We thank an anonymous reviewer for OUP for drawing our attention to this point.

13 Obviously the precise analysis of these data must depend on the syntax of V2. We refer to

the literature. For one proposal for the analysis of V2 in cartographic terms see Haegeman

(1996) and Craenenbroeck and Haegeman (2007).

14 See Derolez and Simon-Vandenbergen (1988) for some discussion of the properties of the

particles.