The case for the partitive case: the contribution of Ancient Greek

26
THE CASE FOR THE PARTITIVE CASE: THE CONTRIBUTION OF ANCIENT GREEK* By MARIA NAPOLI University of Pisa and University of Rome ‘G. Marconi’ ABSTRACT As is well known, Indo-European languages like Vedic, Gothic and Ancient Greek allow the use of the genitive case with a partitive value. This use is traditionally explained by invoking the notion of partial affectedness of the object argument. In the present paper, a case study from Ancient Greek is analysed: accusative genitive alternation with consumption verbs. It appears that the partitive genitive has the functional property of denoting an indefinite and non-specified quantity. I will try to show that in both Homeric and Classical Greek the accusative partitive genitive alternation does not depend on the definite indefinite or referential non-referential status of the object argument, but on the more general parameter of verbal boundedness. Under the theoretical framework provided by recent typological and semantic studies, the hypothesis is investigated according to which the partitive genitive always gives rise to an unbounded reading of the verbal action, as opposed to the accusative, which can be ambiguous between a bounded and an unbounded reading. 1. INTRODUCTION The aim of this paper is to re-examine the partitive function of the Ancient Greek 1 genitive case. In the literature on partitives, what is usually assumed is that this special case has to do with the notion of partial affectedness of the object argument. As regards Indo-European languages, it is traditionally claimed that they lack a specific partitive case; however, they show a genitive case which has a broad range of uses and, in particular, may function as a partitive, alternating with the accusative case. On the basis of this interpretation, an object *This paper has benefited from remarks and suggestions from Pierluigi Cuzzolin, Romano Lazzeroni and Carlotta Viti: I want to thank them. I am deeply indebted to Philip Baldi, who revised my English and provided helpful comments. Special thanks are due to Eystein Dahl: I have greatly profited from our stimulating discussions on this topic, and from comparing our points of view. I am also very grateful to three anonymous reviewers for their insightful observations, and to the editor, Paul Rowlett. 1 In this paper, I shall use the label ‘Ancient Greek’ in opposition to the labels ‘Medieval Greek’ (from 565 AD) and ‘Modern Greek’. Texts and translations of the Greek examples are taken from the editions mentioned in the references. In general, Greek texts are not translated literally, but the meanings of single words are given in interlinear glosses. However, in some cases (23, 24a, 27a, 27b, 30a, 32a, 32c, 35b, 35c, 38a), I have slightly modified the translation in order to make the interpretation of the relevant verbal phrase clearer. List of standard abbreviations (based on Leipzig Glossing Rules: http://eva.mpg.de/lingua/files/morpheme.html): 1 = first person, 2 = second person, 3 = third person, ACC = accusative, ADV = adverb(ial), ALL = allative, AOR = aorist, DAT = dative, ELAT = elative, FUT = future, GEN = genitive, IMP = imperative, IMPF = imperfect, IMPS = impersonal form, IND = indicative, INF = infinitive, IPFV = imperfective, M = masculine, MOD = modal particle, NEG = negation, NOM = nominative, NUM =numeral, OPT = optative, PART = partitive, PFV = perfective, PL = plural, PLE = particle, POSS = possessive, PREP = preposition, PRF = perfect, PRS = present, PST =past, PTCP = participle, SBJV = subjunctive, SG = singular, TRAN = translative case. Transactions of the Philological Society Volume 108:1 (2010) 15–40 Ó The author 2010. Journal Compilation Ó The Philological Society 2010. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ and 350 Main Street, Malden, MA 02148, USA.

Transcript of The case for the partitive case: the contribution of Ancient Greek

THE CASE FOR THE PARTITIVE CASE: THE CONTRIBUTION

OF ANCIENT GREEK*

By MARIA NAPOLI

University of Pisa and University of Rome ‘G. Marconi’

ABSTRACT

As is well known, Indo-European languages like Vedic, Gothic and Ancient Greekallow the use of the genitive case with a partitive value. This use is traditionallyexplained by invoking the notion of partial affectedness of the object argument. In thepresent paper, a case study from Ancient Greek is analysed: accusative ⁄genitivealternation with consumption verbs. It appears that the partitive genitive has thefunctional property of denoting an indefinite and non-specified quantity. I will try toshow that in both Homeric and Classical Greek the accusative ⁄partitive genitivealternation does not depend on the definite ⁄ indefinite or referential ⁄non-referentialstatus of the object argument, but on the more general parameter of verbalboundedness. Under the theoretical framework provided by recent typological andsemantic studies, the hypothesis is investigated according to which the partitive genitivealways gives rise to an unbounded reading of the verbal action, as opposed to theaccusative, which can be ambiguous between a bounded and an unbounded reading.

1. INTRODUCTION

The aim of this paper is to re-examine the partitive function of the Ancient Greek1 genitivecase. In the literature on partitives, what is usually assumed is that this special case has to dowith the notion of partial affectedness of the object argument. As regards Indo-Europeanlanguages, it is traditionally claimed that they lack a specific partitive case; however, theyshow a genitive case which has a broad range of uses and, in particular, may function as apartitive, alternating with the accusative case. On the basis of this interpretation, an object

*This paper has benefited from remarks and suggestions from Pierluigi Cuzzolin, Romano Lazzeroni and CarlottaViti: I want to thank them. I am deeply indebted to Philip Baldi, who revised my English and provided helpfulcomments. Special thanks are due to Eystein Dahl: I have greatly profited from our stimulating discussions on thistopic, and from comparing our points of view. I am also very grateful to three anonymous reviewers for theirinsightful observations, and to the editor, Paul Rowlett.

1In this paper, I shall use the label ‘Ancient Greek’ in opposition to the labels ‘Medieval Greek’ (from 565 AD) and‘Modern Greek’. Texts and translations of the Greek examples are taken from the editions mentioned in thereferences. In general, Greek texts are not translated literally, but the meanings of single words are given in interlinearglosses. However, in some cases (23, 24a, 27a, 27b, 30a, 32a, 32c, 35b, 35c, 38a), I have slightly modified thetranslation in order to make the interpretation of the relevant verbal phrase clearer.List of standard abbreviations (based on Leipzig Glossing Rules: http://eva.mpg.de/lingua/files/morpheme.html):1 = first person, 2 = second person, 3 = third person, ACC = accusative, ADV = adverb(ial), ALL = allative,AOR = aorist, DAT = dative, ELAT = elative, FUT = future, GEN = genitive, IMP = imperative, IMPF = imperfect,IMPS = impersonal form, IND = indicative, INF = infinitive, IPFV = imperfective, M = masculine, MOD = modalparticle, NEG = negation, NOM = nominative, NUM =numeral, OPT = optative, PART = partitive, PFV = perfective,PL = plural, PLE = particle, POSS = possessive, PREP = preposition, PRF = perfect, PRS = present, PST =past,PTCP = participle, SBJV = subjunctive, SG = singular, TRAN = translative case.

Transactions of the Philological Society Volume 108:1 (2010) 15–40

� The author 2010. Journal Compilation � The Philological Society 2010. Published by Blackwell Publishing, 9600 Garsington Road,Oxford OX4 2DQ and 350 Main Street, Malden, MA 02148, USA.

argument marked by the genitive – instead of the accusative – has to be considered as partiallyinvolved in the verbal action.2

In the present paper, the focus is on verbs of consumption, like ‘to eat’ and ‘to drink’, theobject argument of which may be marked by the accusative or the genitive. Theaccusative ⁄genitive alternation with consumption verbs in Ancient Greek is analysed hereby taking into account two main issues: the relationship with the parameters of referentialityand definiteness, and the relationship with the parameter of actionality (the type of verbalaction expressed by the predicate plus its object argument).

The discussion is organised as follows. Section 2 introduces the theoretical frameworkwithin which the Ancient Greek alternation between accusative and genitive will be analysed.Section 3 discusses some cross-linguistic data on the partitive. Section 4 illustrates thedistinction between partitive and pseudo-partitive nominal constructions, based on recenttypological studies. Section 5 deals with the Indo-European partitive genitive, and summarisesthe literature on this topic. Section 6 focuses on data from Ancient Greek and examines theaccusative ⁄partitive genitive alternation with verbs of consumption. In section 7, the AncientGreek data are discussed in the light of the theoretical notions introduced previously and theconclusions are drawn.

2. ALTERNATING CASE MORPHOLOGY

From a cross-linguistic perspective, the object argument of a verbal action tends to be codedby different morphological markings depending on semantic and pragmatic features. Thisphenomenon, known as Differential Object Marking (DOM), is mainly determined by thefollowing parameters:3

(1) the inherent semantic properties of the object argument, i.e. ‘its position along theIndividuation hierarchy’ (Cennamo 2003: 71): this includes, in particular, its positionalong animacy, person, referentiality and definiteness hierarchies;

(2) the degree of affectedness of the object with respect to the verbal action;(3) the nature of the verbal action, as determined by categories such as aspect, actionality

and modality (more specifically, irrealis mood);(4) the pragmatic properties of the object argument, in particular, its function as theme or

rheme.

These nominal and verbal parameters mirror low transitivity status of the whole sentence,according to Hopper & Thompson’s (1980) definition (see, in particular, Haspelmath 2001:57f.; Onishi 2001: 5f.). The main difference among languages concerns which of theseparameters are more relevant and how they interact. In Icelandic, for example, some verbsrequire the dative case with animate objects, but the accusative case with inanimate objects.However, the dative is also the case taken by the object that undergoes a change of location,whereas the object undergoing a change of state takes the accusative: clearly enough, thisalternation reflects the lower or higher degree of affectedness of the object (cf. Cennamo 2003:72 and references therein). In Hindi, on the other hand, the parameters of definiteness andanimacy determine the distribution of the suffix -ko, which marks definite and animate objectsin imperfective tenses (cf. Lazard 1994; Dixon 1994: 190).

2On the problem of ‘case polysemy’ and on the notion of ‘case’ as a language-specific category, see the recentcontribution by Haspelmath (2008). For a discussion of the partitive case from different theoretical perspectives, thereader is referred to Bach, Jelinek, Kratzer & Partee (1995); Hoeksema (1996); Blake (2001); Anderson (2006).

3On this topic see e.g. Lazard (1984; 1994; 2003), Bossong (1985; 1998; 2003), Haspelmath (2001), Onishi (2001),Aissen (2003), Cennamo (2003), Levin & Rappaport (2005), Malchukov (2008). In particular, cf. Bossong (2003: 207)on DOM in Indo-European languages.

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201016

In conclusion, by investigating the phenomenon of alternating morphology, it is possible toshow how nominal and verbal parameters which do not necessarily depend on each other inan individual language may interact. Obviously, the way in which they interact is related tothe morphosyntactic structure of such a language, and to the specific device, if any, forexpressing the distinctions to which these parameters may give rise.

2.1. At the interface of nominal and verbal properties

For the purpose of the present investigation, I will introduce a sketch of the theoreticalframework that I adopt as the background of my analysis. I begin by defining some notionswhich are particularly relevant to the phenomenon of alternating case morphology in general,and to the phenomenon of the accusative ⁄partitive alternation in particular (cf. section 3).

My working definition of referentiality is based on Givon (1978: 293–4):

referentiality is a semantic property of nominals. It involves, roughly, the speaker’s intentto ‘refer to’ or ‘mean’ a nominal expression to have non-empty references – i.e. to ‘exist’ –within a particular universe of discourse. Conversely, if a nominal is ‘non-referential’ or‘generic’, the speaker does not have a commitment to its existence within the relevantuniverse of discourse. Rather, in the latter case, the speaker is engaged in discussing thegenus or its properties, but does not commit him ⁄herself to the existence of any specificindividual member of that genus.

I will adopt here a characterisation of ‘definiteness’ based on the notion of identifiability: anentity is ‘definite’ if the speaker expects the listener to be able to identify it.4 Obviously,‘referential’ is not equivalent to ‘definite’: if a specific, i.e. non-generic, entity is referred to,which is not identifiable by the hearer, it can be defined in terms of ‘referential indefinite’(Givon 1978: 296). As is well known, in languages with an article system, the definite articlecan have both a referential and a non-referential value: it may be used to denote a definitereferent, but it may also be used to denote a generic entity, without any referential properties(cf. Napoli 2009). Moreover, the category of (in)definiteness must be distinguished from thenotion of ‘quantification’: a quantifiable entity is ‘anything concerning which quantity may bepredicated’ (Sapir 1930: 23). It is perfectly conceivable to speak of an indefinite quantity takenfrom a definite entity, the referent of which is contextually determined (cf. e.g. Lazard 1994;Bossong 1998; Kiparsky 1998).

The nominal properties described above are particularly relevant to determine the actionalclass of a verb. Before explaining why, it must be stressed that the notion of actionality hasto be kept distinct from the notion of aspect. Following Bertinetto & Delfitto (2000: 190),aspect concerns ‘the specific perspective adopted by the speaker ⁄writer’, while actionalityconcerns ‘the type of event, specified according to a limited number of relevant properties’.The opposition between imperfective aspect and perfective aspect is the most widespread typeof aspectual distinction: in general terms, imperfective aspect focuses on the internalstructure of the situation, whereas the perfective aspect represents a situation as a singlewhole, without considering the individual phases of which it is constituted (Comrie 1976:16f.).

This brief discussion on actionality is based on the well-known Vendlerian classification(Vendler 1957), which distinguishes four classes (states, activities, accomplishments, achieve-ments) on the basis of properties as stativity, durativity, punctuality, telicity. Here, the focus is

4It is not my task here to discuss theoretical issues concerning the category of definiteness, on which a vast literatureexists. For a detailed discussion of the debate on definiteness and relevant bibliographical references, the reader isreferred to Lyons (1999).

NAPOLI – THE CASE FOR THE PARTITIVE CASE 17

on those verbs whose actional properties are related to the referential, definite and ⁄orquantificational status of the direct object, i.e. verbs which may behave as activities andaccomplishments (cf. Verkuyl 1972; Tenny 1994; Krifka 1998; Beavers 2006; Levin &Rappaport 2006). These verbs have a different reading, atelic or telic, depending on theirpatient argument, which delimits the action if it refers to a specified entity or to a specifiedquantity. Accordingly, verbal phrases like drink wine and eat apples with a non-quantisedobject (a bare mass noun or a bare plural) are atelic, i.e. they correspond to Vendlerianactivities; on the other hand, verbal phrases like eat the apple, drink a cup of wine and eat twoapples with a count singular noun (definite or indefinite) or a quantised plural are telic, i.e.they correspond to Vendlerian accomplishments.

As pointed out in particular by Krifka (1998), this is a characteristic proper ofincremental theme predicates,5 like creation and consumption verbs: the more specific thequantity of the incremental theme is, the more specific is the extent to which the object isaffected by the action, measuring out the event. In other words, in a verbal phrase like eattwo apples the inherent terminal point of the action coincides with the point at which thetwo apples are totally consumed. This is the reason why consumption verbs like ‘to eat’ and‘to drink’ which are quantised turn out to correspond to accomplishments; otherwise theyare activities.6

In the relevant literature, the difference between verbal phrases like eat apples and eat twoapples has been referred to as a difference between unbounded and bounded situations (cf.section 3, and note 8): these labels have been used, for instance, to denote the specific type ofopposition expressed by the Slavic imperfective ⁄perfective aspectual pairs (cf. Bertinetto &Delfitto 2000: 210; Hedin 2000: 228). In Slavic languages like Russian and Polish, wheredefiniteness is not overtly marked by morphological means, the opposition betweenimperfective verbs and perfective (i.e. prefixed) verbs entails a different interpretation of theobject argument in terms of definiteness and quantification. The following Russian example(taken from Di Sciullo & Slabakova 2005: 63) shows that the imperfective form may imply anindefinite reading of the object, as opposed to the perfective form, which implies that theobject is definite and specifically quantised. As a consequence, in (1a), the predicate isunbounded, while in (1b), it is bounded:

(1) a. Ja yel grusi ⁄ tortI eat-IPFV pears ⁄ cake‘I was eating (some) pears ⁄ cake.’

b. Ja s-yel grusi ⁄ tortI eat-PFV pears ⁄ cake‘I ate all the pears ⁄ the whole cake.’

Filip (2005) quotes an interesting case from Polish, which shows how the imperfective aspectin a sentence like (2a) can be ambiguous with respect not only to the interpretation of theobject (as indefinite or definite), but also to the interpretation of the verbal action (ascompleted or not). Conversely, the perfective aspect always implies that the object is definiteand quantised, and the action is completed:

5On the notion of incremental theme see Dowty (1991); cf., more recently, Beavers (2006), Levin and RappaportHovav (2006), Rappaport (2008), Rothstein (2008), and references therein.

6This is a generalisation with a cross-linguistic validity: for a typological analysis of consumption verbs the reader isreferred to Næss (2007: 52–81), which, however, focuses on intransitive uses of ‘to eat’ and ‘to drink’.

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201018

(2) a. On jadx kasze ⁄ oliwkiHe eat-IPFV porridge ⁄ olives‘He was eating porridge ⁄olives ⁄ the porridge ⁄olives ⁄ some of porridge ⁄olives.’‘He ate porridge ⁄olives ⁄ the porridge ⁄olives ⁄ some of porridge ⁄olives.’

b. On z-jadx kasze ⁄ oliwkiHe eat-PFV porridge ⁄ olives‘He ate (up) (all) the porridge ⁄olives.’

As demonstrated by Kiparsky (1998), a distinction similar to the unbounded ⁄boundedopposition expressed by the verbal category of aspect in Russian and Polish is expressed bythe nominal category of partitive in Finnish, as will be illustrated below.

3. THE ACCUSATIVE ⁄ PARTITIVE ALTERNATION

Some languages with a case-marking system show an accusative case alternating with apartitive case or with a genitive case having a partitive function: this alternation, attested inmodern Indo-European languages as Latvian, Lithuanian, Polish and Russian, as well as innon-Indo-European languages as Estonian, Finnish and Basque, has been interpreted as aninstance of the phenomenon of DOM.7 In this section, a survey of some cross-linguistic datais presented which shows the role of the different parameters determining the accusative ⁄par-titive alternation.

I shall begin with Kiparsky’s (1998) analysis of the partitive case in Finnish (cf. also Hopper& Thompson 1980; Heinamaki 1984; Lazard 1994; Vainikka & Maling 1996; Lyons 1999;Sands & Campbell 2001). In this language, which lacks definiteness as a grammatical category,the partitive case has two functions, labelled byKiparsky (1998) aspectual andNP-related: in itsaspectual function, the partitive encodes the object argument of verbs denoting an unboundedevent;8 in its NP-related function, it is associated to quantitatively indeterminate nominalphrases, even though the predicate denotes a bounded event. It must be pointed out that inFinnish some verbs are intrinsically bounded and other verbs intrinsically unbounded. Otherpredicates, however, can be interpreted as unbounded or bounded, depending on the objectargument: these verbs primarily allow the accusative ⁄partitive alternation. They are typicallyincremental theme predicates (cf. section 2.1), including verbs of creation and verbs ofdestruction such as syoda ‘to eat’ (partitive: soi piirakkaa ‘(he) ate pie, some of the pie’;accusative: soi piirakan ‘(he) ate a ⁄ the pie’; cf. Kiparsky 1998: 281).

The aspectual function of the partitive case, as opposed to the accusative case, isexemplified below by means of one of those verbs admitting an unbounded or a boundedreading, i.e. ampua ‘to shoot (at)’ (from Kiparsky 1998: 267):

7Moravcsik (1978) was one of the first scholars who recognised ‘that this marking difference does not correlate withany difference in semantic case function, and that it correlates at least in some cases with one or more of the followingsemantic properties of the verbs or nouns involved: a. the definiteness-indefiniteness of the noun phrase, b. the extentto which the object is involved in the event, c. the completedness versus non-completedness of the event, d. whetherthe sentence is affirmative or negative’ (Moravcsik 1978: 272).

8Kiparsky (1998) did not provide a precise definition of the notion of ‘boundedness’. In general terms, an event canbe defined as ‘bounded’ if it is conceptualised as having a clear boundary. While some scholars, like Jackendoff (1991),emphasise that telic events, having an inherent end point, are always bounded, as opposed to atelic events, otherscholars, like Depraetere (1995) and Kiparsky (1998), argue for a distinction between boundedness and telicity. Inparticular, Kiparksy (1998) noted that telicity, although correlating with boundedness, is not the right semanticcriterion for identifying the conditions under which objects are accusative in Finnish, as documented by the existenceof bounded atelic verbs and telic unbounded verbs.

NAPOLI – THE CASE FOR THE PARTITIVE CASE 19

(3) a. Ammu-i-n karhu-a ⁄ karhu-j-a

shoot-1SG.PST bear-PART.SG bear-PART.PL‘I shot at the (a) bear ⁄ at (the) bears.’

b. Ammu-i-n karhu-n ⁄ karhu-t

shoot-1SG.PST bear-ACC.SG bear-ACC.PL‘I shot the ⁄a bear ⁄ the bears.’

In (3a), the verb takes a partitive object and corresponds to an activity, denoting anunbounded action; in (3b), the object is marked by the accusative and the verb corresponds toan accomplishment, denoting a bounded action (‘to shoot dead’). Moreover, because of itsdouble functions (aspectual or NP-related), the partitive case yields a three-way ambiguitywith a predicate like ampua (from Kiparsky 1998: 268):

(4) Ammuin karhujashoot-1SG.PST bear-PART.PLi. ‘I shot at the bears’ (aspectual partitivity);ii. ‘I shot bears’ (NP-related partitivity);iii. ‘I shot at bears’ (both aspectual partitivity and NP-related

partitivity).

In some contexts the partitive may denote the incompleteness of the action (regardless ofwhether the object is definite or not): the predicate that takes a partitive-marked objectcan have the value of a progressive form. This may be illustrated by a sentence like (5a),which is three-ways ambiguous, whereas the corresponding sentence with the accusative in(5b) implies that the action is completed and the object definite (from Kiparksy 1998:272):

(5) a. Han kirjoitt-i kirje-i-ta

He ⁄ she write-3SG.PST letter-PART.PLi. ‘He wrote (some) letters (…and left)’ (completed process, indefinite NP);ii. ‘He was writing letters (..when I came)’ (incomplete process, indefinite NP);iii. ‘He was writing the letters (…when I came)’ (incomplete process, definite NP).

b. Han kirjoitt-i kirjee-t

He ⁄ she write-3SG.PST letter-ACC.PL‘He wrote the letters (…and left).’

Inherently bounded verbs can take a partitive-marked object if the object is quantitativelyindeterminate (Kiparksy 1998: 281). Moreover, they can receive an imperfective reading withthe partitive case, in order to represent the event as an ongoing activity (from Kiparsky 1998:289):

(6) Tapo-i-n juuri karhu-a

kill-1SG.PST just bear-PART.SG‘I was just killing the bear.’

The partitive is obligatorily used with inherently unbounded verbs and with verbs of emotion(from Sands & Campbell 2001: 290, 303, n. 25):

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201020

(7) a. Rakastan-n han-ta

love-1SG he-PART

‘I love him.’b. Rakast-i-n hane-t kuoliaa-ksi

love-1SG.PST he-ACC death-TRAN

‘I loved him to death.’

The obligatory use of the partitive case in (7a) depends on the fact that ‘the primary case inFinnish for less affected O, or for reduced transitivity, is the partitive’ (Sands & Campbell2001: 290). However, the accusative may occur when the object argument is conceived of asundergoing a change of state (7b).

It is worth pointing to the fact that, as demonstrated by Kiparsky (1998), definiteness is notthe relevant criterion for selecting the partitive case. With inherently unbounded verbs, thedirect object is always ambiguous between a definite and an indefinite reading. Moreover, in anegative clause, the object is marked by the partitive even though it is definite (from Sands &Campbell 2001: 284):

(8) a. So-i-n omena-n

eat-1SG.PST apple-ACC.SG‘I ate the apple.’

b. E-n syo-nyt omena-a

NEG-1SG eat-1SG.PST apple-PART.SG‘I did not ate the apple.’

In other words, boundedness seems to be the relevant parameter. As mentioned in section 2.1,Kiparsky (1998: 280) observed that, as a tendency, a verb denoting an unbounded situationtakes the partitive case in Finnish, whereas it is imperfective in Russian; a verb denoting abounded event takes the accusative case in Finnish, whereas it is perfective in Russian.Boundedness is morphologically marked in both languages, but in the first case themorphological marking is nominal, while in the second case it is verbal.

Also in a language like Hungarian, which has an article system, ‘partitive case expressesgenuine partitivity, not indefiniteness’ (Lyons 1999: 200). As the example in (9) shows, thepartitive explicitly indicates that only a part of the object is affected by the action (fromMoravcsik 1978: 263):

(9) Olvasott a konyvb}olread-3SG the book-PART.SG‘He ⁄ she read some of the book.’

However, what plays a role in determining the distribution of accusative and partitive cases isalso the completed vs. non-completed nature of the verbal action. Verbs denoting a completedaction take the accusative case only (10a), and they do not allow the partitive case (fromMoravcsik 1978: 263):

(10) a. Elolvasta a konyvet

away.read-3SG the book-ACC.SG‘He ⁄ she read the book (and finished it).’

b. *Elolvasta a konyvb}olaway.read-3SG the book-PART.SG

NAPOLI – THE CASE FOR THE PARTITIVE CASE 21

The accusative ⁄partitive genitive alternation is also allowed in Russian, where it is normallylimited to perfective verbs (cf. Forsyth 1970: 91; Dezs}o 1982: 50f.).9 I quote two examples of averb of consumption (from Dezs}o 1982: 57):

(11) a. Petr vypil vino

Petr drink-PFV wine-ACC

‘Peter drank the wine.’b. Petr vypil vina

Petr drink-PFV wine-GEN

‘Peter drank some wine.’

In (11b), the genitive seems to have an indefinite reading, as opposed to the accusative (11a),which denotes a definite referent.

A similar pattern is found in the Germanic languages (cf. Abraham 1997; 2007a; 2007b;Philippi 1997; Leiss 2000; 2007). For instance, in Gothic and in the early stages of German,which lack definiteness as a grammatical category, the interaction of alternating case markingand aspectual morphology was able to effect the interpretation of the object argument asdefinite or indefinite. In particular, as pointed out by Abraham (1997), in Gothic, which hasno articles at all, and in Old High German, which shows a weakly established and non-systematic use of articles, perfective verbs are employed with the accusative case in order tosuggest a definite reading of the object, whereas their occurrence with the genitive case alwaysimplies an indefinite reading. This alternative interpretation is possible only with perfectiveverbs, whereas imperfective verbs always take the accusative case. This state of affairs iscomparable to the accusative ⁄genitive alternation documented in Russian. However, OldHigh German and Russian differ with respect to the interpretation of the accusative case withimperfective verbs: when the predicate occurs as an imperfective form, the accusative may beambiguous between a definite and an indefinite reading in Old High German, whereas anindefinite reading of the direct object is necessarily presupposed in Russian (Abraham 1997:46).

To conclude, even though the existence of a significant correlation between alternating casemarking and the semantic notion of definiteness seems to be proved by data from ancient andmodern languages, obviously this does not mean that in all languages without articles theaccusative ⁄partitive alternation systematically encodes the same range of definite ⁄ indefinitedistinctions as in languages with articles. In other words, this does not mean that thisalternation covers all the pragmatic and semantic functions performed by articles in languageswhich possess them. For example, it would be improper to argue that the accusative ⁄genitivealternation is a device for encoding definiteness ⁄ indefiniteness distinctions in Russian: thegenitive case does not necessarily imply indefiniteness, considering, for instance, that it tendsto be used also for definite nouns, in the scope of negation (Lyons 1999: 201). In this respect,Lyons’ (1999: 201) conclusion seems to be the most reasonable:

the temptation to view it as a unified phenomenon is very strong, despite variation inprecisely what triggers the appearance of the oblique case. But both this variation, and thecomplexity of the trigger within individual languages, make it clear that if we are dealingwith a single phenomenon, then it cannot be correlated directly with the definite-indefinitedistinction, though definiteness does play a part in what is happening.

9‘It is also characteristic that objects in the partitive genitive usually occur with perfective verbs: the total extensionof the action to the object is given expression by both verb and object noun’ (Forsyth 1970: 91). However, it must benoted that this is only a tendency: imperfective verbs can be used with the genitive case in Russian (Greville Corbett,p.c.).

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201022

In sections 5 and 6, I attempt to demonstrate the validity of this conclusion by means of datafrom Ancient Greek. Before trying to do so, I consider some cross-linguistic material whichillustrates the use of the partitive as embedded in nominal constructions, and whichdemonstrates the possibility of a definite or an indefinite reading of partitives.

4. PARTITIVE VERSUS PSEUDO-PARTITIVE CONSTRUCTIONS

At the level of the noun phrase, a distinction should be drawn between ‘partitiveconstructions’ (PCs) and ‘pseudo-partitive constructions’ (PPCs).10 As an example of PCsone can cite noun phrases like a piece of the cake, a cup of that good tea, whereas noun phraseslike a piece of cake, a cup of tea are instances of PPCs. Only in the first two noun phrases(PCs), where a determiner co-occurs with the quantised noun, are we really dealing with a‘part’ of something; in contrast, in the second two noun phrases (PPCs), what is talked aboutis an ‘amount’ of something (Koptjevskaja-Tamm 2001: 523).

It is worth quoting from Koptjevskaja-Tamm (2006), who clearly explains the differencebetween these two types of nominal constructions:

both partitive and pseudo-partitive nominal constructions are noun phrases consisting oftwo nominals, one of which is a quantifier (cup, slice, pile), while the other nominal will becalled ‘‘quantified’’. […] PCs involve a presupposed set of items or a presupposed entity

referred to by one of the nominals (the cake, Mary’s book), and the quantifier indicates asubset or a subpart which is selected from it. In a pseudo-partitive nominal construction(PPC), the same word merely quantifies over the kind of entity (tea, books) indicated bythe other nominal. (Koptjevskaja-Tamm 2006: 218; bold emphasis original)

In the next section, I briefly sumup the different strategies employed across European languagesin order to codify PCs andPPCs, on the basis ofKoptjevskaja-Tamm (2001; 2006; forthcoming).

4.1. A typology of PCs and PPCs in European languages

First of all, it must be emphasised that not all languages explicitly distinguish between PCsand PPCs. The distinction between these two nominal constructions is overtly expressed bymorphological means in some European languages. Among languages with a partitiveconstruction distinct from a pseudo-partitive construction, it is possible to identify two maintypes.

Type I. The distinction between PCs and PPCs is expressed by morphological means, i.e. bycase morphology. In Finnish, as demonstrated by Koptjevskaja-Tamm (2001: 531), thepartitive case occurs in PPCs (12a) and, in some cases, also in PCs, whereas the elative caseoccurs only in PCs (12b):

(12) a. Osta sakki perunoita!buy-2SG.IMP sack-NOM potato-PART.PL‘Buy a sack of potatoes!’

b. Anna minulle pala tasta hyvasta kakusta ⁄ litragive me-ALL bit-NOM this-ELAT good-ELAT cake-ELAT ⁄ litre-NOM

tuoreesta maidosta-si

fresh-ELAT milk-ELAT-2SG.POSS

‘Give me a bit of this good cake ⁄ a litre of your fresh milk.’

10The label ‘pseudo-partitive’ was introduced by Selkirk (1977); see also Lobel (1989).

NAPOLI – THE CASE FOR THE PARTITIVE CASE 23

Type II. In numerous languages (e.g. Danish, Swedish, Norwegian, Icelandic, German,Yiddish, Dutch, Bulgarian, Macedonian, Modern Greek), PPCs are distinguished from PCsin being juxtapositional: this means that PPCs require juxtaposition of quantifier andquantised entities. In some of these languages, the juxtapositional type turns out to be new,and has clearly replaced the more archaic use of the genitive with a partitive function, which isstill characteristic of PCs (Koptjevskaja-Tamm 2001: 554; 2006: 220). This state of affairs isexemplified by Modern Greek, which has developed a juxtapositional construction limited tothe pseudo-partitive type (from Koptjevskaja-Tamm 2006: 220):

(13) ena kilo kafes kostizi epta dolariaone-NUM.NOM kilo-NOM coffee-NOM cost-3SG.PRS seven dollars-ACC

‘One kilo of coffee costs seven dollars.’

Unlike the two types described above, there are languages in which the same structureoccurs in PPCs and PCs. In English, for instance, the preposition of is employed in bothcases. Similarly, in Baltic and Slavic languages (with the exception of Bulgarian andMacedonian), nominal quantifiers always take the genitive case in PPCs and in PCs (fromKoptjevskaja-Tamm 2001: 540):

(14) Latviana. gl�aze t�ejas

glass-NOM tea-GEN.SG‘a glass of tea’

b. gl�aze s��s gars��g�as t�ejasglass-NOM that-GEN.SG good-GEN.SG tea-GEN.SG‘a glass of that good tea’

(15) Russiana. stakan sok-a

glass-NOM juice-GEN

‘a glass of juice’b. stakan von to-go sok-a

glass-NOM there that-GEN.SG.M juice-GEN

‘a glass of that juice’

It is worth adding that, as pointed out by Koptjevskaja-Tamm (forthcoming), theneutralisation or reduction of partitive and pseudo-partitive distinction is ‘geneticallyrestricted to certain groups of Indo-European languages’, i.e. (as seen above) to Baltic andSlavic (apart from Bulgarian and Macedonian) and also to Romance. Moreover, theneutralisation of the distinction is a trait common to ancient and modern Indo-Europeanlanguages.

5. THE PARTITIVE GENITIVE IN ANCIENT GREEK AND ITS PROTO-INDO-EUROPEAN BACKGROUND

Ancient Greek documents the partitive use of the genitive case, as well as its alternation withthe accusative, when marking the object argument of well-defined semantic classes of verbs(cf. Schwyzer & Debrunner 1950: 103; Chantraine 1953: 51; Humbert 1960: 269; Luraghi1988: 180; Dahl & Napoli in preparation). The occurrence of the genitive with a partitivefunction is considered not only to be of Proto-Indo-European origin (cf. Brugmann &

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201024

Delbruck 1911: 583f.; Meillet 1922: 305; Meillet & Vendryes 1924: 508–10) but also to be oneof the most important functions of the genitive: ‘le principal emploi du genitif indo-europeenetait sans doute d’indiquer le tout dont on prend une partie’ (‘the main use of the Indo-European genitive was undoubtedly to denote the whole from which one takes a part’; Meillet1922: 20).

Apart from Ancient Greek, both phenomena, i.e. the partitive use of the genitive and theaccusative ⁄partitive alternation, are well preserved in other Indo-European languages:11 inparticular, in Vedic (cf. Dahl forthcoming; Dahl & Napoli in preparation), and, as shown insection 3, in Gothic and in the oldest stages of German. There are also traces of theaccusative ⁄genitive alternation in early Latin (cf. Bennett 1966: 35; Serbat 1996; Bauer 2007:133–5), where they are limited to technical, non-literary texts, as in the following examplefrom Cato, who is giving culinary instructions:

(16) a. eo aquam additoit-DAT.SG water-ACC.SG add-IMP.FUT

‘Add water to it.’ (Cato, Agr. 37.2.8)b. aquae paulatim addito

water-GEN.SG gradually-ADV add-IMP.FUT

‘Add water gradually.’ (Cato, Agr. 74.1.2)

As stressed by Serbat (1996), the rarity of similar examples, which are not found in theClassical period, led some Latin grammarians to hypothesise that in such cases the deletion ofa nominal element should be postulated.

5.1. The Greek accusative ⁄genitive alternation: partial affectedness and definiteness

In Ancient Greek, different classes of verbs allow alternating case marking. Also among Indo-Europeanists, the accusative ⁄genitive alternation has usually been interpreted in the light ofthe notion of degree of affectedness of the object argument (cf. section 1). In examining theGreek data, two remarks concerning the notion of partial affectedness and the notion ofdefiniteness are in order.

First, from a synchronic point of view, the accusative ⁄genitive alternation cannot always beexplained by invoking the notion of total vs. partial affectedness of the patient argument. Inthis respect, an interesting case study is represented by Greek perception verbs. In Homer, theaccusative and genitive cases have a specific distribution with verbs of perception: a predicatelike akou�o ‘to listen, to hear’ normally takes the accusative of nouns meaning ‘voice’, ‘word’,‘sound’ (e.g. Il. 2. 200; 4. 435; Od. 11. 421; 12. 198; 16. 10; 21. 326), but it takes the genitivewith animate (human and non-human) nouns and pronouns (Il. 19. 256; Od. 1. 370; 6. 325; 19.85). Vedic shows a similar pattern, which seems to be an Indo-European inheritance.12

Nevertheless, in a few Homeric instances (for example Il. 22. 447, 451; Od. 9. 401; 21. 290),perception verbs occur with the genitive of inanimate nouns, like the accusative. Chantraine(1953: 54) explained these occurrences as follows: ‘des termes comme ops, aud��e, etc … peuventdesigner la voix en tant qu’elle emet un son et qui n’est percue que partiellement (genitif) ou leson percu totalement (accusatif)’ (‘words like ops, aud��e, etc … may denote the voice insofar as

11‘The distinction between total and partial objects by means of case alternation (accusative vs. genitive) is attestedin a number of older Indo-European languages; partial subjects are attested too, but to a very limited extent’(Koptjevskaja-Tamm 2001: 562).

12Cf. e.g. Humbert (1960: 272). A different hypothesis was put forward by Schwyzer & Debrunner (1950: 94), whoclaimed that the genitive performs a ‘separative’ function with perception verbs, closer to some uses of the ablativecase.

NAPOLI – THE CASE FOR THE PARTITIVE CASE 25

it produces a sound which is partially perceived (genitive) or one which is totally perceived(accusative)’). However, this is not confirmed by the data, as shown, for instance, by thefollowing passage:

(17) muk�ethmo~u t’ ��ekousa bo~�on aulizomen�a�onlowing-GEN.SG PLE hear-1SG.AOR.IND cattle-GEN.PL be.stalled-GEN.PLoi~�on te bl�ek��en […]sheep-GEN.PL PLE bleating-ACC.SG‘I heard the lowing of the cattle that were being stalled and the bleating of thesheep.’ (Od. 12. 265–6)

In a sentence like (17), where the same verb (��ekousa) takes the genitive and the accusative oftwo inanimate nouns, it is difficult to find any difference in the semantics of the two cases.Chantraine’s (1953) explanation is clearly based on a previous presupposition, the identityof the notion of partitivity with the notion of partial affectedness, for which, however, thereis no evidence in this case. It could be rather hypothesised that with this specific set of verbsthe genitive was originally employed to denote the (animate) source of the sound, becauseof his ⁄her being unaffected by the action. Later, however, the genitive is extended to nounsdenoting the sound itself, as a simple substitute for the accusative, without implying anyparticular semantic interpretation (as in example 17).

Second, as summarised in section 3, according to scholars like Abraham (1997; 2007a;2007b) and Leiss (2000; 2007), in Gothic and Old High German perfective verbs selecting theaccusative case entail a definite reading of the object, whereas they imply an indefinite readingwhen taking the genitive case. Bauer (2007) discusses Abraham’s and Leiss’s results,concluding that the issue of the relationship between case, aspect and definiteness in Proto-Indo-European deserves more research. However, with regard to Ancient Greek, she quotestwo examples of the genitive case as encoding ‘the partitive indefinite direct object’ (Bauer2007: 134); I quote one of them:

(18) xunetrıb�e t~�es kephal~�esbash-3SG.AOR.IND the-GEN.SG head-GEN.SG‘[till he tumbled off and] bashed his head.’ (Ar., Pax 71)

What is remarkable is the fact that Bauer (2007) regards the object argument as ‘indefinite’even though it is accompanied by a definite article (and it is clearly referential). Evidentlyenough, she extends Abraham’s and Leiss’s interpretation of Germanic languages to Greek.However, this kind of approach – i.e. the claim that the partitive genitive always implies anindefinite reading of the object argument, although with an explicit definite mark – does notaccount for the Greek data.

Nearly a century ago Brugmann & Delbruck (1911) suggested the necessity ofdistinguishing between a verbal phrase like pieın oınou ‘to drink wine’ and a verbal phraselike pieın tou oınou ‘to drink from that wine’, where the genitive-marked noun occurs with adefinite article:

Verba, die eine beliebige korperliche Tatigkeit bezeichnen, wie essen, trinken, geben,nehmen, zeigen den Genitiv in Konkurrenz mit dem Akkusativ. Dabei ist das genitivischeSubstantiv entweder ein allgemeiner Begriff, der ohne Begrenzung vorgestellt ist, odereine bestimmt begrenzte Masse, z.B. pieın oınou ‘‘Wein trinken’’ und tou oınou ‘‘von demWein’’. (‘Verbs denoting any physical activity, like to eat, to drink, to give, to take, showthe genitive alternating with the accusative. The genitive-marked noun involved

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201026

corresponds either to a general notion, which is introduced without any limitation, or to aprecisely bounded mass, for example pieın oınou ‘‘drink wine’’ and tou oınou ‘‘from thatwine’’.’) (Brugmann & Delbruck 1911: 585)

Significantly enough, Brugmann & Delbruck (1911) correlated the issue of partitivitywith the issue of definiteness, implicitly identifying two different types of entailmentproduced by the partitive genitive: in the first case, the object argument is interpretable asgeneric, in the second case it corresponds to a specified referent, as the definite articleindicates.

This confirms that the notion of partitivity does not coincide with ‘indefiniteness’ toutcourt. In particular, the assumption that the partitive genitive is always indefinite in AncientGreek, although accompanied by a definite article, is imprecise, because it conflates thedimensions of definiteness and quantification. What the Greek partitive genitive shows isthat a distinction between non-referential indefinite and referential definite partitives existsat the level of the verbal phrase (as well as a distinction between (indefinite) pseudo-partitive and (definite) partitive constructions exists at the level of the nominal phrase:cf. section 4).

6. A SURVEY OF THE ANCIENT GREEK DATA: ‘TO EAT’ AND ‘TO DRINK’

Verbs of consumption like ‘to eat’ and ‘to drink’ typically allow the accusative ⁄genitivealternation in Ancient Greek,13 as well as in Indo-European languages like Vedic, Gothic andOld High German. I quote here an example from Hippocrates, where the aorist of pın�o occurswith the genitive of a mass noun:

(19) pie~in te epı toutoisin oınou

drink-AOR.INF and besides this-DAT.PL wine-GEN.SG‘and besides these things a little wine to drink.’ (Hp., VM 13. 23)

However, a verbal phrase like pie~in oınou is far from being frequently attested in the Greeklanguage: more precisely, as recently pointed out by Riano Rufilanchas (2005), the occurrenceof a consumption verb plus the genitive of a mass noun meaning ‘bread’ or ‘wine’ is not sowidespread (Riano Rufilanchas 2005: 299–300).14 In other words, the example in (19) does notrepresent the typical occurrence of the Greek partitive genitive.

Moreover, on the basis of the traditional interpretation, the partitive genitive is said toimply that a certain quantity is taken from a given object, which as a results is partially

13The goal of this work is neither to investigate the behaviour of all Greek verbs of consumption showingaccusative ⁄ genitive alternation nor to offer a statistical description of this phenomenon. I will examine someoccurrences of the main consumption verbs, i.e. bibr��osk�o, esthı�o, ed�o, efagon, pat�eomai ‘to eat’, and pın�o ‘to drink’,relevant to the aim of discussing the Greek data in the light of the theoretical issues introduced above. The reader isreferred to Riano Rufilanchas (2005) for a richer sample of consumption verbs in Ancient Greek, and for a detaileddescription of their use, on the basis of semantic criteria.

14Riano Rufilanchas (2005: 299) underlines that the partitive genitive is used as an emphatic and marked case (incontrast with the unmarked status of the accusative), linked to some specific semantic constraints: ‘los verbospropiamente de ‘‘comer, beber’’, que deben en principio considerarse come de transitividad alta, admiten laconstruccion con gen. en dos situaciones: I) Con muchas restricciones, cuando designan una parte definida oindefinida de un todo determinado; II) En contextos en los que el rasgo ‘‘transformacion del objeto’’ desaparece delconcepto lexico debido a que el objeto no es un producto que sirva tipicamente para satisfacer la necesidad de comer obeber’ (‘verbs properly meaning ‘‘to eat, to drink’’, which have to be regarded as highly transitive in principle, allowthe construction with the genitive in two contexts: I) With many restrictions, when they designate a definite orindefinite part taken from a specific whole; II) In those contexts in which the feature ‘‘transformation of the object’’fades away the lexical concept because the object is not a product which typically serves to satisfy the need for eatingor drinking’).

NAPOLI – THE CASE FOR THE PARTITIVE CASE 27

consumed; in opposition, the accusative-marked object is represented as totally consumed (cf.e.g. Chantraine 1953: 51; Humbert 1960: 269–70). This explanation is not inconsistent withthe data. Nonetheless, it needs to be refined by taking into account the properties of the verbalphrase, in the light of the distinction between unbounded and bounded actions sketched insections 2 and 3.

As explained above (section 2.1), consumption verbs are incremental theme predicates:this means that the properties of referentiality, definiteness and quantification of the objectargument have an effect on the actional class of the predicate, determining its classificationas an activity or an accomplishment. Considering that, as a rule, consumption verbs arebounded if they are quantised, i.e. if they have an object argument with specific orquantised reference, the hypothesis can be put forward according to which Greekconsumption verbs with a genitive-marked object are unbounded: a genitive-marked objectcannot delimit the action of eating or drinking, because it refers to a generic and non-referential object or to an unspecified quantity of a given substance. In such a sense, aconsumption verb with a partitive genitive is never quantised and, as a result, neverbounded. This may explain why prefixed verbs like kataphage~in and ekpınein do not take agenitive-marked object: Humbert’s (1960: 270) explanation was that they entail ‘laconsommation totale’ (‘total consumption’). In the light of the theoretical issues discussedabove, the point is that such predicates are inherently bounded, corresponding toVendlerian accomplishments.

In the following pages, I will try to show how Ancient Greek provides evidence for theclaims made so far. As is well known, syntactic test for actionality cannot be applied to ourdata, which could be revealing of the unbounded or bounded nature of the verbal phrase (cf.Napoli 2006: 70f.). However, an analysis of the precise context in which the genitive occursmay cast light on this kind of alternating morphology, especially by means of a comparisonwith the occurrences of the accusative case taken by imperfective forms (i.e. present stems)and perfective forms (i.e. aorist stems).

For the purpose of our investigation, Homeric Greek (represented by the Iliad and theOdyssey) and Classical Greek are treated separately. At the earliest stage of the Greeklanguage (i.e. Homeric Greek), definiteness was not systematically expressed by morpholog-ical means. More precisely, Homeric Greek does not attest a well-established system ofdefinite articles. In Classical Greek, on the contrary, the development of a definite articleappears to be complete. This is a necessary premise in order to focus on the differences andsimilarities of these stages of the Greek language.

6.1. Homeric Greek

In the Homeric language, consumption verbs have an object marked by the genitive only ifthey are inflected as aorist or perfect stems. For instance, ed�o and esthı�o ‘to eat’, which areattested only as present stems, do not show accusative ⁄genitive alternation.

As regards consumption verbs with an accusative-marked object, the patient tends to havea generic value if the predicate is inflected as a present, whereas it tends to be implicitly orexplicitly quantised if the verb is inflected as an aorist. This is consistent with the fact that inHomeric Greek the aorist of activity ⁄accomplishment verbs tends to denote an action whichis perfective from an aspectual viewpoint, and telic from an actional viewpoint, theaccusative-marked object usually referring to an entity which is high in the animacy and ⁄ordefiniteness hierarchies (cf. Napoli 2006: 86f.). Consider the following examples (fromNapoli 2006: 112):

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201028

(20) a. oınon men gar pÇne kath��emenos, �ophr’wine-ACC.SG PLE PLE drink-3SG.IMPF sit-PRS.PTCP as long asethel’ autoswant-3SG.IMPF he-NOM.SG‘he sat there and drank wine as long as he would.’ (Od. 20. 136)

b. h��os phato kaı speısas epien meli�edeaso speak-3SG.IMPF and pour-AOR.PTCP drink-3SG.AOR.IND honey-sweet-ACC.SGoınon, aps d’ en chersın eth�eke depas

wine-ACC.SG back PLE PREP hand-DAT.PL give-3SG.AOR.IND cup-ACC.SGkosm��etori la~�onmarshaler-DAT.SG person-GEN.PL‘so he spoke, and pouring a libation, drank the honey-sweet wine, and then gaveback the cup into the hands of the marshaler of the people.’ (Od. 18. 151–2)

In both (20a) and (20b), in which the imperfect and the aorist indicative occur,respectively, the verb pın�o ‘to drink’ takes an accusative-marked mass noun (oınon ‘wine’).What is significant here is the fact that this noun has a generic value in (20a), wherethe verb is inflected as an imperfect; on the contrary, in (20b), where the aorist indicativeoccurs, the same noun is implicitly quantised: one could argue from the context thatin (20b) oınon corresponds to that amount of wine put into a single depas ‘cup’ (verse152).

In examining the use of the partitive genitive in Homeric Greek, the emphasis must be puton the existence of two different orders of occurrences: the object argument may be generic(i.e. non-referential) or it may correspond to a referential definite entity. However, in bothcases, the use of the partitive genitive with consumption verbs, instead of the accusative, couldbe regarded as a strategy for denoting an unbounded action. The two kinds of entailmentmentioned are exemplified below:

1. The genitive denotes a generic object of which one eats or drinks. Some examples are:

(21) a. […] h��os te leonta, hos rha te bebr�ok��oslike PLE lion-ACC.SG that-NOM.SG PLE PLE feed-PRF.PTCP.NOM.SGboos erchetai agrauloio

ox-GEN.SG come-3SG.PRS.IND of.the.farmstead-GEN.SG‘…like a lion that comes from feeding on an ox of the farmstead.’ (Od. 22. 402–3)

b. […] alla sphi dosan l�otoıo pasasthai

but them-DAT.PL give-3PL.AOR.IND lotus-GEN.SG eat-AOR.INF

‘…but [the Lotus-eaters] gave them lotus to eat.’ (Od. 9. 93)c. alla pasasthai an�ochthi tho~�eis epı n�eusın

but taste-AOR.INF command-2SG.PRF.IMP swift by ship-DAT.PLAchaious ⁄ sıtou kaı oınoio […]Achaean-ACC.PL food-GEN.SG and wine-GEN.SG‘but command the Achaeans by their swift ships to taste of food andwine...’ (Il. 19. 160–61)

It is interesting to compare example (21c), where the aorist infinitive pasasthai ‘to taste’ occurswith the genitive of two nouns having a generic meaning, with example (22), where the sameverb takes the accusative of a noun (splanchna ‘innards’) which denotes a specified object,because its referent has been mentioned before:

NAPOLI – THE CASE FOR THE PARTITIVE CASE 29

(22) splanchna d’ ar’ ampeırantes hupeırechon H�ephaıstoio. autarinnards-ACC.PL ple spit-AOR.PTCP hold.over-3PL.IMPF Hephaestus-GEN.SG butepeı kata m~�er’ eka�e kaı splanchn’

when PREP thigh-NOM.PL burn-3SG.AOR.IND and innards-ACC.PLepasanto

taste-3PL.AOR.IND

‘and the innards they spitted and held over the fire of Hephaestus. But when the thighpieces were wholly burned, and they had tasted the innards…’ (Il. 2. 426–7)

2. The genitive is referential and definite, in the sense that it refers to an element alreadyintroduced in the discourse, or inferable from it (so that the hearer can identify it). In thefollowing case, for instance, oınoio ‘wine’ clearly alludes to an indefinite amount of that winecontained into the goblet (aleison, verse 9):

(23) ~�e toi ho kalon aleison anair��esesthai emelle,PLE PLE he handsome goblet-ACC.SG rise-FUT.INF be.on.the.point-3SG.IMPF

chruseon amph�oton kaı d��e meta chersın en��oma,of.gold two-eared and PLE PREP hand-DAT.PL handle-3SG.IMPF

ophra pıoi oınoio [...]MOD drink-3SG.AOR.OPT wine-GEN.SG‘now he, you must know, was on the point of raising to his lips a handsome goblet,a two-eared cup of gold, and was even now handling it, that he might drink of thatwine …’ (Od. 22. 9–11)

The following cases are particularly interesting because they show the aorist pieın ‘to drink’(24) and the aorist phageın ‘to eat’ (25) occurring with the accusative and the genitive of thesame nouns in the same contexts:

(24) a. haımatos ophra pı�o kaı toi n�emertea eıp�oblood-GEN.SG MOD drink-1SG.AOR.SBJV and PLE truth say-1SG.AOR.SBJV‘so that I may drink of the blood and speak the truth to you.’ (Od. 11. 96)

b. […] ho d’ epeı pıen haıma kelainonhe-NOM.SG PLE when drink-1SG.AOR.IND blood-ACC.SG dark

‘when he had drunk the dark blood …’ (Od. 11. 98)

(25) a. t~�on d’ hos tis l�oto~�o phagoi meli�edeathem-GEN.PL PLE whoever lotus-GEN.SG eat-3SG.AOR.OPT honey-sweet-ACC.SGkarpon

fruit-ACC.SG‘and whoever of them ate the honey-sweet fruit of the lotus …’ (Od. 9. 94)

b. m��e p��os tis l�oto~�o phag��onMOD PLE anyone-NOM.SG lotus-GEN.SG eat-AOR.PTCP.NOM.SGnostoio lath�etaihomecoming-GEN.SG forget-3SG.AOR.SBJV‘for fear that perchance anyone should eat of the lotus and forget hishomecoming.’ (Od. 9. 102)

In (24), the noun haıma has a referential value in being mentioned before (Od. 11. 36). Asopposed to the verbal phrase haımatos pieın ‘to drink of the blood’ in (24a), which

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201030

unambiguously indicates that the object is not totally consumed, as is inherent to thefunction of the partitive genitive, the verbal phrase haıma pieın ‘to drink blood’ in (24b)does not give any explicit indication in this regard. It does not mean that Teiresiasconsumed all the blood; some verses later (Od. 11. 153, 228), others drink from that bloodtoo. However, intuitively, in (24b) the temporal and aspectual value of the predicate, whichrepresents the action as completed in the past, leads the reader to recognise that a specificquantity of the blood has been consumed: Teiresias had drunk that quantity of blood whichallowed him to speak the truth. In contrast, the genitive is chosen when the same action isrepresented as potential: the focus is on the action of drinking itself, without anyspecification of the degree of affectedness of the patient argument. Similarly, the action ofeating is represented as completed in (25a), where the accusative is employed, as opposed to(25b), which is the only Homeric occurrence of the aorist phageın ‘to eat’ with the genitive(l�oto~�o ‘lotus’ has a referential meaning, because the corresponding referent has already beenintroduced into the discourse: cf. 21b).

6.2. Classical Greek

As pointed out above, the Homeric state of affairs primarily differs from the situationfound in Classical Greek15 because definiteness ⁄ indefiniteness distinctions were not overtlymarked at the stage of the Greek language represented by the Iliad and the Odyssey.Greek later developed a definite article, whereas it still lacks an indefinite article. As aconsequence, a bare count noun is two-ways ambiguous in principle: it may denote ageneric entity or it may denote a referential indefinite entity (cf. section 2.1).

In Classical Greek, both present and aorist stems can occur with a direct object marked bythe accusative case. Such an object can be referential or non-referential, definite or indefinite,and it can give rise to a bounded or an unbounded reading of the verbal action. The followingexamples illustrate the occurrence of the accusative with the present stem, entailing a genericreading of the direct object:

(26) a. […] ou pınontas ek krith~�on methu

NEG drink-PRS.PTCP.ACC.PL PREP barley-GEN.PL wine-ACC.SG‘…no drinkers of barley-beer.’ (A., Supp. 953)

b. xenizomenoi de pros bıan epınomen ex hualın�onregale-PRS.PTCP PLE PREP force-ACC.SG drink-1PL.IMPF PREP crystalekp�omat�on kaı chrusıd�on akraton oınon h�edungoblet-GEN.PL and gold unmixed wine-ACC.SG fine‘and when they regaled us they forced us to drink fine unmixed wine from goblets ofcrystal and gold.’ (Ar., Ach. 73–5)

When the present stem occurs with a definite object, it generally entails a progressive reading(27a) or a habitual interpretation (27b), as is consistent with its imperfective aspectualcharacter:

(27) a. houtos, tı dr~�ais? ton oınon ekpıneis

you what do-2SG.PRS.IND the-ACC.SG wine-ACC.SG drink-2SG.PRS.IND

15Most of the examples quoted in the present section are taken from Attic prose and poetry (5th and 4th centuriesBC), traditionally considered as representative of ‘Classical Greek’. However, I have also included a few examples frommore recent works, e.g. from Herodotus’ Ionic prose, which cannot properly be regarded as ‘Classical Greek’.

NAPOLI – THE CASE FOR THE PARTITIVE CASE 31

lathrai?on.the.sly-ADV

‘You! What are you doing? Drinking the wine on the sly?’ (E., Cyc. 552)b. xunoikos ~�en moi, toumon ekpınous’

living-NOM.SG be-3SG.IMPF me-DAT.SG the.my-ACC.SG drink-PRS.PTCP.NOM.SGaeı psuch~�es akraton haıma […]all.the.time-ADV life-GEN.SG pure-ACC.SG blood-ACC.SG‘yes, she was living with me, all the time drinking my very life-blood.’ (S., El. 785–6)

As a tendency, the aorist takes a specific or quantised accusative-marked object, which can besingular and definite (28a and 28b), singular and indefinite (29a and 29b), or plural andexplicitly quantised by means of a numeral (30a and 30b). It is self-evident that in such casesthe aorist denotes a bounded action:

(28) a. dokeı gar d��e beltion eınai lousamenon pieın

seem-3SG.PRS.IND.IMPS PLE PLE better be-PRS.INF bathe-AOR.PTCP drink-AOR.INF

to pharmakon

the-ACC.SG poison-ACC.SG‘for I think it is better to bathe before drinking the poison.’ (Pl., Phd. 115a.7)

b. h�e de pardalis hotan phag�ei to

the-NOM.SG PLE panther-NOM.SG after eat-3SG.AOR.SBJV the-ACC.SGpharmakon ho kaleıtai pardalianchespoison-ACC.SG which be.called-3SG.PRS.IND panther’s-bane‘The panther, after eating the poison which is called panther’s-bane …’

(Arist., HA 612a.7–612a.8)

(29) a. ph��eseis <g’>, epeidan ekpı�eis oınou

say-2SG.FUT.IND PLE when drink.off-2SG.AOR.SBJV wine-GEN.SGneou lepast��ennew-GEN.SG cup-ACC.SG‘That’s what you’ll say when you drink off a cup of new wine!’ (Ar., Pax 916)

b. h�os ekpieın kan kulika bouloım�en mıan

MOD drink-AOR.INF and.MOD cup-ACC.SG would.like-1SG.PRS.OPT one‘I would like to drink down a single cup of this wine.’ (E., Cyc. 164)

(30) a. kaı prın se kotulas ekpieın oınou deka

and before you-ACC.SG goblet-ACC.PL drink-AOR.INF wine-GEN.SG ten‘and before you could have drunk ten goblets of wine…’ (Ar., Pl. 737)

b. hekkaıdek’ artous katephag’ h�em~�onsixteen loaf-ACC.PL eat.up-3SG.AOR.IND us-GEN.PL‘he ate up sixteen our loaves.’ (Ar., Ra. 551)

Less frequently, the aorist may take the bare accusative of a noun having a generic value:

(31) a. beltiston h�emın haıma taureion pieın

best us-DAT.PL blood-ACC.SG of.bull-ACC.SG drink-AOR.INF

‘our best course is to drink bull’s blood.’ (Ar., Eq. 83)b. […] ou gar ethel��esei phageın out’ arton oute mazan

NEG PLE want-3SG.FUT.IND eat-AOR.INF NEG bread-ACC.SG NEG bread-ACC.SG‘…she’ll not want to eat bread or cake.’ (Ar., Pax 852–3)

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201032

However, in contrast with the present stem, and consistent with its perfective aspectualnature, the aorist typically alludes to a single occurrence of the action involved. This is, forinstance, the case of (31a), in which the verbal phrase ‘to drink bull’s blood’ alludes to an actof suicide.

With regard to the occurrence of the partitive genitive with consumption verbs, in ClassicalGreek the accusative ⁄genitive alternation seems to conform to the two patterns found inHomeric Greek.

1. The first case is represented by those instances in which the genitive-marked object has ageneric value. More precisely, the focus is on the kind of entity affected by consumption. Inpost-Homeric Greek generics can occur with a definite article, which is, however, notobligatorily: consequently, as already noted at the beginning of this section, generics may alsobe bare. Interestingly enough, when the partitive genitive is used with a generic reading, itnever takes the definite article: a generic reading of the partitive is available only with barenouns. A few examples are in (32):

(32) a. ornithos ornis p~�os an hagneuoi phag��onbird-GEN.SG bird-NOM.SG how MOD be.pure-3SG.PRS.OPT eat-AOR.PTCP.NOM.SG‘if a bird eats of a bird, how can it be pure?’ (A., Supp. 226)

b. kata de Surıan eınaı tı phasi z~�oion hoin PLE Syria be-PRS.INF a say-3SG.PRS.IND beast-ACC.SG whichkaleıtai leontophonon; apothn��eskei gar hobe.called-3SG.PRS.IND lion-killer die-3SG.PRS.IND PLE the-NOM.SGle�on, h�os eoiken, hotan autou phag�eilion-NOM.SG as seem-3SG.PRF.IND when it-GEN.SG eat-3SG.AOR.SBJV‘In Syria they say there is a beast called the lion-killer; for the lion apparently dieswhen it eats of it.’ (Arist., Mir. 845a.28–30)

c. pionta ton anthr�opon auton hauto~udrink-AOR.PTCP.ACC.SG the-ACC.SG person-ACC.SG himself-ACC.SG it-GEN.SGpoieı pr~�oton h�ile�on euthus mallon ��e proteronmake-3SG.PRS.IND ADV jovial ADV more than before‘it [wine] makes the person who drinks of it more jovial than he wasbefore.’ (Pl., Lg. 649.a.–b.1)

The following example is particularly useful to illustrate the function of the genitive as apartitive:

(33) h�e de chel��on�e hotan eche�os phag�ei, epesthıeithe-NOM.SG PLE tortoise-NOM.SG when viper-GEN.SG eat-3SG.AOR.SBJV eat-3SG.PRS.IND

t��en orıganonthe-ACC.SG origanon-ACC.SG‘when the tortoise has eaten some of a viper it eats origanum as well.’

(Arist., HA 612a.24–a.25)

In (33), Aristotle describes a tortoise which eats origanum after having eaten some of a viper:what is focused on by means of the verbal phrase consisting of the aorist phageın ‘to eat’ andthe genitive eche�os ‘viper’ is the unbounded character of the action. In this case, the choice ofthe aorist may have implied a bounded reading of the action, as in the examples in (29), wherea bare accusative with a referential indefinite value occurs, delimiting the action. In (33),however, the tortoise does not eat up the viper: it is arguable from the context that an

NAPOLI – THE CASE FOR THE PARTITIVE CASE 33

unspecified part of the entity mentioned is consumed each time. Indeed, in the same passage,Aristotle explicitly adds that the tortoise eats of the viper again, repeating the same actionseveral times: I quote the translation of Arist., Ha 612a.25–9: ‘and once a man had observed atortoise doing this several times and going back to the viper after each mouthful of origanumhe began stripping the leaves off the origanum’. In conclusion, this example offers evidence ofthe impact of the genitive on the verbal action, i.e. to decrease the telicity of a predicate likephageın.

Apart from Classical Greek, two other instances of the pattern identified above are thefollowing (from prose works by Herodotus and Plutarch):

(34) a. ichthu�on de ou sphi exesti pasasthai

fish-GEN.PL PLE NEG them-DAT.PL may-3SG.PRS.IND.IMPS eat-AOR.INF

‘they may not eat fish.’ (Hdt., 2.37. 18–19)b. en d’ Apoll�onos polei nenomismenon estı

PREP PLE Apollo-GEN.SG town-DAT.SG be.established-3SG.PRF.IND

krokodeılou phageın pant�os hekastoncrocodile-GEN.SG eat-AOR.INF ADV every.person-ACC.SG‘in the town of Apollonopolis it is an established custom for every person withoutexception to eat of a crocodile.’ (Plu., Mor. 371.D.5–6)

2. The partitive genitive may occur in contexts in which it has a referential definite value. Thismeans that it presupposes an identifiable referent, which has already been mentioned, or isinferable from the discourse; as a consequence, the genitive-marked noun is used with adefinite article or with an anaphoric pronoun:

(35) a. pr~�oton o~un skeps��ometha tına tropon diait��esontai hoi hout�ofirst then consider-1PL.AOR.SBJV what manner-ACC.SG live-3PL.FUT.IND they thuspareskeuasmenoi. allo ti ��e s~it�on te poio~untes kaı oınon

provided-PRF.PTCP any.other than bread-ACC.SG and make-PRS.PTCP and wine-ACC.SGkaı himatia kaı hupod��emata?and garment-ACC.PL and shoe-ACC.PL‘first of all, let us consider what will be the manner of life of men thus provided. Willthey not make bread and wine and garments and shoes?’ (Pl., R. 372.a.5–7)

b. eu�och��esontai autoı te kaı ta paidıa,feast-3PL.FUT.IND they-NOM.PL and also the-NOM.PL child-NOM.PLepipınontes to~u oınou

drink-PRS.PTCP.NOM.PL the-GEN.SG wine-GEN.SG‘they will feast with their children, drinking of that wine thereto.’

(Pl., R. 372b.6–372.b.7)c. phobou pharmakon esth’ hostis theos ed�oken

fear-GEN.SG poison-ACC.SG be-3SG.PRS.IND anyone God-NOM.SG give-3SG.AOR.IND

anthr��opois, h��oste hopos�oi pleon an ethel�eiman-DAT.PL MOD as.many.as more MOD like-3SG.PRS.SBJVtis pınein autou […]anyone drink-PRS.INF it-GEN.SG‘there does exist a God who has given a poison of fear to men, such that, the more aman likes to drink of it…’ (Pl., Lg. 647e.1.–e.2.)

In (35b), the genitive occurs with a definite article which has an anaphoric value, referring toan entity previously introduced into the discourse: a group of men is described, who make

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201034

wine by themselves (35a); then, it is assumed that the same men drink of that wine (35b). Inanother example from Plato (35c), the pronoun autou (‘it’) refers to a noun already mentioned(pharmakon ‘poison’). It is interesting to note that in (35b) the genitive occurs with a presentparticiple, suggesting the iterative character of the verbal action, in (35c) it occurs with apresent infinitive, suggesting the habitual character of the verbal action. In Homeric Greek wedo not find instances of the use of the partitive genitive with consumption verbs in the presentstem (cf. section 6.1). It is difficult to say if this difference depends on the restricted amount ofdata from Homeric Greek, or if it really reflects an extension of the types of contexts in whichthe partitive genitive can occur in Classical Greek.

I now examine a case from Plato where the aorist infinitive pieın ‘to drink’ occurs with theaccusative and with the genitive in the same passage:

(36) a. metron men oun ti tou hudatos pasinmeasure-ACC.SG PLE PLE a-ACC.SG the-GEN.SG water-GEN.SG all-DAT.PLanankaıon eınai pieın

necessary be-PRS.INF drink-AOR.INF

‘they were all required to drink a measure of the water.’ (Pl., R. 621a.6–7)b. autos de tou men hudatos k�oluth~�enai pieın

he PLE the-GEN.SG PLE water-GEN.SG prevent-AOR.INF drink-AOR.INF

‘he was not allowed to drink of the water.’ (Pl., R. 621b.4–5)

In (36a), the object argument is referential, definite and specifically quantised by means of thenominal phrasemetron ti tou hudatos ‘a measure of the water’. In (36b), the object is referentialand definite too, but not quantised: in this case, the focus is on the activity of drinking itself (andon the fact that a certainman is not allowed to do it), independently of the degree of affectednessof the object argument.

I cite below four more examples with verbs meaning ‘to eat’; two are taken from ClassicalGreek, two are more recent:

(37) a. eıper etuchon t~�on mel~�on t~�onif happen.to.be-3PL.AOR.IND the-GEN.PL song-GEN.PL the-GEN.PLPhilokleous bebr�okotesPhilocles-GEN.SG munch-PRF.PTCP.NOM.PL‘if they’d been munching on Philocles’ songs.’ (Ar., V. 462)

b. […] �ethel�ese d’ haımatos koino~u pasasthai16

mean-3SG.AOR.IND PLE blood-GEN.SG that-GEN.SG eat-AOR.INF

‘…and he meant to drink the people’s blood.’ (S.,Ant. 201–2)

(38) a. kaı gar Kuaxar�es kaı hoi pareontes daitumonesand PLE Cyaxares-NOM.SG and the be.present-PRS.PTCP this-GEN.PLt~�on kre~�on tout�on epasanto

guest-NOM.PL the-GEN.PL flesh-GEN.PL eat-3PL.AOR.IND

‘Cyaxares and the guests who feasted with him ate of the body’s flesh.’(Hdt., 1.73. 25–6)

b. h�opou gar en toutois tis l�ogos genoitowhenever PLE PREP this-DAT.PL any mention-NOM.SG be-3SG.AOR.OPT

16The primary meaning of this verb is ‘to eat’; however, pat�eomai also occurs with nouns denoting something todrink, as in (37b), and evidently it corresponds to ‘to drink’. In such a sense, we could assume that it is a ‘generic’ingestion verb.

NAPOLI – THE CASE FOR THE PARTITIVE CASE 35

perı Spartiat~�on, oudena dunasthai krupteinPREP Spartiatae-GEN.PL no.one-ACC.SG be.able-PRS.INF conceal-PRS.INF

to m��e ouch h�ede�os an kaı �om~�onthe-ACC.SG NEG NEG pleasantly-ADV MOD also raw-GEN.PLesthıein aut~�oneat-PRS.INF them-GEN.PL‘for whenever among these classes any mention was made of Spartiatae, no one wasable to conceal the fact that he would be glad to eat them raw.’ (X., HG 3.3.6.6.–7.1.)

Examples (35)–(38) demonstrate that the partitive genitive can co-occur with a determinativeelement (such as a definite article or a demonstrative pronoun) in order to denote a referentialdefinite entity. Clearly enough, in cases like these, we could not assume that the genitiveimplies an ‘indefinite’ reading of the object. I will make this point clearer in the next section,where I summarise and comment on the data in more detail.

7. ANALYSIS OF THE ANCIENT GREEK DATA

In interpreting the data discussed so far, it is worth mentioning again the distinction betweenpartitive (PCs) and pseudo-partitive constructions (PPCs) which was introduced in section 4.In such constructions, the ‘partitive’ marker has a different entailment, depending on thepresence of a determinative element (i.e. according to its definite status). However, thedifference between nominal PCs and PPCs was neutralised in ancient Indo-Europeanlanguages, where both constructions are expressed by the genitive.

Focusing on Ancient Greek, if we compare the level of the nominal phrase (i.e. the case ofnominal quantifiers with a genitive-marked noun) with the level of the verbal phrase (i.e. thecase of consumption verbs taking a genitive-marked object), we could assume that thedifferent entailments of nominal PCs and PPCs are reflected also at the verbal level. As seen insection 6, with verbs like ‘to eat’ and ‘to drink’ the partitive genitive can occur in two differenttypes of context, in both Homeric and Classical Greek. It can denote the kind of entity whichis affected by consumption, and, in such a sense, it is the corresponding verbal form of apseudo-partitive nominal construction, lacking any definiteness implication (the objectargument corresponds to a generic non-referential entity); it can denote a non-specifiedquantity taken from a contextually determined, i.e. referential entity. In the last case, thepartitive genitive can be considered as the corresponding verbal form of a partitive nominalconstruction. In this respect, Homeric Greek and Classical Greek do not differ substantially:the main difference consists in the fact that distinctions in terms of referentiality anddefiniteness are explicitly encoded only in Classical Greek, where the partitive genitive mayco-occur with a determinative element such as a definite article or a demonstrative pronoun.Nevertheless, even in Homeric Greek the partitive genitive may be found in contexts where areferential reading of the object is the only one possible.

This demonstrates why it would be inaccurate to regard the partitive genitive as necessarilyimplying an indefinite reading of the object argument in Ancient Greek (as implicitly arguedby Bauer 2007; see section 5 above). In interpreting the occurrences of genitive-marked nounswith consumption verbs, one must distinguish between the substance consumed, which maybe referential or non-referential, definite or indefinite, and the quantity taken from it, which isalways indefinite, in the sense that it is not specified. Nevertheless, this is not equivalent tosaying that the Greek accusative ⁄genitive alternation encodes definiteness ⁄ indefinitenessdistinctions. Obviously, the dimension of quantification and the dimension of (in)definitenessare distinct and coded independently in Ancient Greek, as well as in other languages (cf.

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201036

section 2.1). The Greek partitive genitive with verbs of consumption suggests that thesubstance consumed cannot be quantised precisely: accordingly, it could be said that thepartitive genitive is a marker of ‘non-specified quantification’, rather than a marker ofindefiniteness. This functional property of the partitive genitive explains why it isincompatible with quantifiers, while it is compatible with definite articles. A quantifier issomething which expresses a judgment of quantification about a quantifiable entity (Putzu2001: 23). No precise quantity may be predicated about the entity denoted by the partitivegenitive with verbs of consumption. However, the partitive genitive can signal that anindefinite quantity is taken from an entity which is marked by the definite article because of itsreferential character.

To sum up, while it is true that the Greek partitive genitive indicates that the objectargument is not entirely consumed, what is noteworthy for the interpretation of the verbalphrase is the fact that such an object is non-specifically quantised. This implies that the actionof eating or drinking is necessarily represented as unbounded, if the object is expressed bymeans of a genitive (regardless of whether there is a precise referent for the object itself):17 ascross-linguistic studies have demonstrated, the action of consuming a non-quantised norquantifiable entity, which accordingly cannot measure out the action, has the properties typicalof a Vendlerian activity. Considering that the accusative case may give rise either to a boundedor to an unbounded reading of the consumption verb with which it occurs, as examples insection 6 demonstrate, the genitive is selected to avoid this ambiguity.

Finally, the distribution of the accusative and the partitive genitive in Ancient Greek doesnot depend on the parameter of definiteness, but on the parameter of boundedness, just as inFinnish (cf. section 3 above on Kiparsky’s 1998 analysis). Obviously, this does not mean thatthe two forms of alternation are equivalent: in Finnish, for instance, the partitive maydetermine a progressive reading of the verb, whereas in Ancient Greek this function isperformed by the imperfective, as opposed to the perfective, aspect.

8. CONCLUSION

In the present paper, I have re-examined a case study from Ancient Greek: the accusative ⁄par-titive genitive alternation with verbs of consumption, which belong to that type of predicate(activity ⁄accomplishment) thatmay be interpreted as unbounded or bounded, depending on thepresence of a quantised object. I have tried to show how in both Homeric and Classical Greekthe accusative ⁄genitive alternation does not turn out to be sensitive to definiteness ⁄ indefinite-ness oppositions. It turns out to be sensitive to the quantificational dimension and, moreover, tothe boundedness dimension.

In Ancient Greek the construction with the accusative is interpretable in principle asquantised or not; accordingly, a consumption verb which takes the accusative case isinterpretable as bounded or not. In other words, the accusative can presuppose an objectwhich is quantised, whereas the construction with the partitive genitive simply lacks thispresupposition As a consequence, the partitive genitive always gives rise to an unboundedreading, independently of the referential ⁄non-referential and definiteness ⁄ indefinitenessproperties of the entity consumed. Indeed, the partitive genitive can occur in two differenttypes of contexts: it can denote a generic object, expressing the kind of entity consumed, or itcan correspond to a referential definite entity.

17Early Vedic data concerning the accusative ⁄ genitive alternation with verbs of consumption are consistent with theGreek state of affairs described here (cf. Dahl forthcoming). Moreover, as argued by Dahl & Napoli (in preparation),the fact that the construction with the genitive is attested with reference to both a definite and an indefinite object inEarly Vedic as well as in Ancient Greek could suggest that these two readings were already available in Proto-Indo-European with genitive-marked objects.

NAPOLI – THE CASE FOR THE PARTITIVE CASE 37

To conclude, the status of the object argument in terms of referentiality and definitenessdoes not influence the selection of the accusative versus the genitive neither in Homeric Greek,which lacks definite articles, nor in Classical Greek, which has them. The relevant criterion isthe parameter of verbal boundedness.

Department of LinguisticsUniversity of PisaVia Santa Maria 3656124 Pisa (Italy)Email: [email protected]

References

ABRAHAM, WERNER, 1997. ‘The interdependency of case, aspect and referentiality in the history of German: the case ofthe verbal genitive’, in Ans van Kemenade & Nigel Vincent (eds.), Parameters of Morphosyntactic Change,Cambridge: Cambridge University Press, 29–61.

ABRAHAM, WERNER, 2007a. ‘Discourse binding: DP and pronouns in German, Dutch and English’, in Stark et al.(2007: 21–47).

ABRAHAM, WERNER, 2007b. ‘The discourse-functional crystallization of the historically original demonstrative’, inStark et al. (2007: 241–256).

AIKHENVALD, ALEXANDRA, DIXON, ROBERT M. W. & ONISHI, MASAYUKI (eds.), 2001. Non-canonical Marking ofSubjects and Objects, Amsterdam: Benjamins.

AISSEN, JUDITH, 2003. ‘Differential object marking: iconicity and economy’, Natural Languages and Linguistic Theory21, 435–483.

ANDERSON, JOHN M., 2006. Modern Grammars of Case, Oxford: Oxford University Press.BACH, EMMON, JELINEK, ELOISE, KRATZER, ANGELIKA & PARTEE, BARBARA H. (eds.), 1995. Quantification in Natural

Languages, Dordrecht: Kluwer.BAUER, BRIGITTE, 2007. ‘The definite article in Indo-European: emergence of a new grammatical category?’, in Stark et

al. (2007: 103–139).BEAVERS, JOHN, 2006. ‘Argument ⁄ oblique alternations and the structure of lexical meaning’, Ph.D. thesis, Stanford

University.BENNETT, CHARLES E., 1966. Syntax of Early Latin, vol. 2, Hildesheim: Olms.BERTINETTO, PIERMARCO & DELFITTO, DENIS, 2000. ‘Aspect vs. actionality: why they should be kept apart’, in Osten

Dahl (ed.), Tense and Aspect in the Languages of Europe, Berlin: Mouton de Gruyter, 190–225.BLAKE, BARRY J., 2001 [19941]. Case, Cambridge: Cambridge University Press.BOSSONG, GEORGE, 1985. Empirische Universalienforschung:Differentielle Objektmarkierung in den neuiranischen

Sprachen, Tubingen: Narr.BOSSONG, GEORGE, 1998. ‘Le marquage differentielle de l’objet’, in Jack Feuillet (ed.), Actance et valence, Berlin:

Mouton de Gruyter, 193–258.BOSSONG, GEORGE, 2003. ‘Nominal and ⁄ or verbal marking of central actants’, in Giuliana Fiorentino (ed.), Romance

Objects: Transitivity in Romance Languages, Berlin: Mouton de Gruyter, 17–47.BRUGMANN, KARL & DELBRUCK, BERTHOLD, 1911. Grundriss der vergleichende Grammatik der indogermanischen

Sprachen, 2nd edn, vol. 2: Vergleichende Laut-, Stammbildungs- und Flexionslehre, Strassburg: Trubner.CENNAMO, MICHELA, 2003. ‘(In)transitivity and object marking: some current issues’, in Giuliana Fiorentino (ed.),

Romance Objects: Transitivity in Romance Languages, Berlin: Mouton de Gruyter, 49–104.CHANTRAINE, PIERRE, 1953. Grammaire Homerique, Paris: Klincksieck.COMRIE, BERNARD, 1976. Aspect: An Introduction to the Study of Verbal Aspect and Related Problems, Cambridge:

Cambridge University Press.DAHL, EYSTEIN, forthcoming. ‘Some semantic and pragmatic aspects of object alternation in Early Vedic’, in Johanna

Barðdal & Shobhana Celliah (eds.), The Role of Semantics and Pragmatics in the Development of Case, Amsterdam:Benjamins.

DAHL, EYSTEIN & NAPOLI, MARIA, in preparation. ‘Case as an aspect marker? The partitive case in Indo-European’.DEPRAETERE, ILSE, 1995. ‘On the necessity of distinguishing between (un)boundedness and (a)telicity’, Linguistics and

Philosophy 18, 1–19.DEZS}O, LASZLO, 1982. Studies in Syntactic Typology and Contrastive Grammar, The Hague: Mouton.DI SCIULLO, ANNA MARIA & SLABAKOVA, ROUMYANA, 2005. ‘Quantification and aspect’, in Henk Verkuyl, Henriette de

Swart & Angeliek van Hout (eds.), Perspectives on Aspect, Dordrecht: Springer, 61–80.DIXON, ROBERT M.W., 1994. Ergativity, Cambridge: Cambridge University Press.DOWTY, DAVID, 1991. ‘Thematic proto-roles and argument selection’, Language 67, 547–619.FILIP, HANA, 2005. ‘On accumulating and having it all: perfectivity, prefixes and bare arguments’, in Henk Verkuyl,

Henriette de Swart & Angeliek van Hout (eds.), Perspectives on Aspect, Dordrecht: Springer, 125–148.

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201038

FORSYTH, JAMES, 1970. A Grammar of Aspect Usage and Meaning in the Russian Verb, Cambridge: CambridgeUniversity Press.

GIVON, TALMY, 1978. ‘Definiteness and referentiality’, in Joseph H. Greenberg (ed.), Universals of Human Language,Stanford, CA: Stanford University Press, 293–330.

HASPELMATH, MARTIN, 2001. ‘Non-canonical marking of core arguments in European languages’, in Aikhenvald et al.(2001: 53–83).

HASPELMATH, MARTIN, 2008. ‘Terminology of case’, in Andrej Malchukov & Andrew Spencer (eds.), The OxfordHandbook of Case, Oxford: Oxford University Press, 505–517. Available at: http://email.eva.mpg.de/�haspelmt/publist.html

HEDIN, EVA, 2000. ‘The type-referring function of the imperfect’, in Osten Dahl (ed.), Tense and Aspect in theLanguages of Europe, Berlin: Mouton de Gruyter, 227–264.

HEINAMAKI, ORVOKKI, 1984. ‘Aspect in Finnish’, in Casper de Groot & Hannu Tommola (eds.), Aspect Bound: AVoyage into the Realm of Germanic, Slavonic and Finno-Ugrian Aspectology, Dordrecht: Foris, 153–178.

HOEKSEMA, JACOB (ed.), 1996. Partitives: Studies on the Syntax and Semantics of Partitive and Related Constructions,Berlin: Mouton de Gruyter.

HOPPER, PAUL J. & THOMPSON, SANDRA A., 1980. ‘Transitivity in grammar and discourse’, Language 56, 251–299.HUMBERT, JEAN, 1960. Syntaxe Grecque, Paris: Klincksieck.JACKENDOFF, RAY, 1991. ‘Parts and boundaries’, Cognition 41, 9–45.KIPARSKY, PAUL, 1998. ‘Partitive case and aspect’, in Miriam Butt & Wilhelm Geuder (eds.), The Projection of

Arguments, Stanford, CA: CSLI, 266–307.KOPTJEVSKAJA-TAMM, MARIA, 2001. ‘A piece of the cake and a cup of tea: partitive and pseudo-partitive nominal

constructions in the Circum-Baltic languages’, in Osten Dahl & Maria Koptjevskaja-Tamm (eds.), The Circum-Baltic Languages: Typology and Contact, vol. 2: Grammar and Typology, Amsterdam: Benjamins, 523–568.

KOPTJEVSKAJA-TAMM, MARIA, 2006. ‘Partitives’, in Keith Brown (ed.), Encyclopaedia of Languages and Linguistics,2nd edn, Amsterdam: Elsevier, 2, 18–21.

KOPTJEVSKAJA-TAMM, MARIA, forthcoming. ‘A lot of grammar with a good portion of lexicon: towards a typology ofpartitive and pseudo-partitive nominal constructions’, in J. Helmbrecht, N. Yoko, S. Yong-Min, S. Skopeteas & E.Verhoeven (eds.), Studies in Grammaticalization, Berlin: Mouton de Gruyter. Available at: http://www.ling.su.se/staff/tamm/publ-e03.html

KRIFKA, MANFRED, 1998. ‘The origins of telicity’, in Susan Rothstein (ed.), Events and Grammar, Dordrecht: Kluwer,197–235.

LAZARD, GILBERT, 1984. ‘Actance variations and categories of the object’, in Frans Plank (ed.), Objects: Towards aTheory of Grammatical Relations, London: Academic Press, 269–292.

LAZARD, GILBERT, 1994. L’Actance, Paris: Presses Universitaires de France.LAZARD, GILBERT, 2003. ‘What is an object in a cross-linguistic perspective?’ in Giuliana Fiorentino (ed.), Romance

Objects: Transitivity in Romance Languages, Berlin: Mouton de Gruyter, 1–16.LEISS, ELISABETH, 2000. Artikel und Aspekt: Die grammatischen Muster von Definitheit, Berlin: de Gruyter.LEISS, ELISABETH, 2007. ‘Covert patterns of definiteness ⁄ indefiniteness and aspectuality in Old Icelandic, Gothic, and

Old High German’, in Stark et al. (2007: 73–102).LEVIN, BETH & RAPPAPORT HOVAV, MALKA, 2005. Argument Realization, Cambridge: Cambridge University Press.LEVIN, BETH & RAPPAPORT HOVAV, MALKA, 2006. ‘Constraints on the complexity of verb meaning and VP structure’,

in Hans-Martin Gartner, Sigrid Beck, Regine Eckardt, Renate Musan & Barbara Stiebels (eds.), Between 40 and 60Puzzles for Krifka. Available at: http://www.zas.gwz-berlin.de/publications/40-60-puzzles-for-krifka/

LOBEL, ELISABETH, 1989. ‘Q as a functional category’, in Christa Bhatt, Elisabeth Lobel & Claudia Schmidt (eds.),Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences, Amsterdam: Benjamins, 133–158.

LURAGHI, SILVIA, 1988. ‘The opposition total : partitive and the use of cases with prepositions in Ancient Greek’, in A.Rijksbaron, H. A. Mulder & G. C. Wakker (eds.), In the Footsteps of Raphael Kuhner, Amsterdam: Gieben, 177–192.

LYONS, CHRISTOPHER, 1999. Definiteness, Cambridge: Cambridge University Press.MALCHUKOV, ANDREJ L., 2008. ‘Animacy and asymmetries in differential case marking’, Lingua 118, 203–221.MEILLET, ANTOINE, 1922. Introduction a l’etude comparative des langues indo-europeennes, Paris: Hachette.MEILLET, ANTOINE & VENDRYES, JOSEPH, 1924. Traite de grammaire comparee des langues classiques, Paris: Champion.MORAVCSIK, EDITH A., 1978. ‘On the case marking of objects’, in Joseph H. Greenberg (ed.), Universals of Human

Language, vol. 4, Stanford, CA: Stanford University Press, 249–290.NÆSS, ASHILD, 2007. Prototypical Transitivity, Amsterdam: Benjamins.NAPOLI, MARIA, 2006. Aspect and Actionality in Homeric Greek: A Contrastive Analysis, Milan: Angeli.NAPOLI, MARIA, 2009. ‘Aspects of definiteness in Greek’, Studies in Language 33.3.ONISHI, MASAYUKI, 2001. ‘Introduction: non-canonically marked subjects and objects: Parameters and Properties’, in

Aikhenvald et al. (2001: 1–51).PHILIPPI, JULIA, 1997. ‘The rise of the article in the Germanic languages’, in Ans van Kemenade & Nigel Vincent (eds.),

Parameters of Morphosyntactic Change, Cambridge: Cambridge University Press, 62–93.PUTZU, IGNAZIO, 2001. Quantificazione totale ⁄ universale e determinatezza nelle lingue del Mediterraneo, Pisa: ETS.RAPPAPORT HOVAV, MALKA, 2008. ‘Lexicalized meaning and the internal temporal structure of events’, in Susan

Rothstein (ed.), Theoretical and Crosslinguistic Approaches to the Semantics of Aspect, Amsterdam: Benjamins.RIANO RUFILANCHAS, DANIEL, 2005. ‘La sintaxis de los verbos ‘‘comer’’ y ‘‘beber’’ en griego antiguo: un estudio sobre

el genitivo partitivo’, Emerita 73, 263–302.

NAPOLI – THE CASE FOR THE PARTITIVE CASE 39

ROTHSTEIN, SUSAN, 2008. ‘Two puzzles for a theory of Lexical Aspect: the case of semelfactives and degree adverbials’,in Johannes Dolling, Tatjana Heyde-Zybatowand & Martin Shafer (eds.), Event Structures in Linguistic Form andInterpretation, Berlin: Mouton de Gruyter, 175–198.

SANDS, KRISTINA & CAMPBELL, LYLE, 2001. ‘Non-canonical subjects and objects in Finnish’, in Aikhenvald et al. (2001:251–305).

SAPIR, EDWARD, 1930, Totality, Linguistic Society of America Language Monographs 6, Baltimore, MD: WaverlyPress.

SCHWYZER, EDUARD & DEBRUNNER, ALBERT, 1950. Griechische Grammatik, Munich: Beck.SELKIRK, ELISABETH, 1977. ‘Some remarks on noun phrase structure’, in Peter W. Culicover, Thomas Wason & Adrian

Akmajian (eds.), Formal Syntax, New York: Academic Press, 283–316.SERBAT, GUY, 1996. Grammaire fondamentale du latin: l’emploi des cas en latin, vol. 6.1, Louvain: Peeters.STARK, ELISABETH, LEISS, ELISABETH & ABRAHAM, WERNER (eds.), 2007. Nominal Determination: Typology, Context,

Constraints, and Historical Emergence, Amsterdam: Benjamins.TENNY, CAROL L., 1994. Aspectual Roles and the Syntax–Semantics Interface, Dordrecht: Kluwer.VAINIKKA, ANNE & MALING, JOAN, 1996. ‘Is partitive case inherent or structural?’, in Hoeksema (1996: 179–208).VENDLER, ZENO, 1957. ‘Verbs and times’, Philosophical Review 66, 143–160.VERKUYL, HENK, 1972. On the Compositional Nature of the Aspects, Dordrecht: Reidel.

TEXTS

(All the following texts are published by Harvard University Press.)

AESCHYLUS, Suppliant Maidens, with an English translation by H. W. Smith, 1973 [1st edn 1922].ARISTOPHANES, Acharnanians. Knights, with an English translation by J. Henderson, 1998.ARISTOPHANES, Clouds. Wasps. Peace, with an English translation by J. Henderson, 1998.ARISTOPHANES, The Plutus, with an English translation by B. Bickley Rogers, 1963 [1st edn 1924].ARISTOTLE, History of Animals, volume 11 (Books 7–10), with an English translation by D. M. Balme, 1991.ARISTOTLE, Minor Works. On marvellous things heard, with an English translation by W. S. Hett, 1963 [1st edn 1936].CATO, On Agriculture, with an English translation by W. D. Hooper, revised by H. B. Ash, 1934.EURIPIDES, Cyclops. Alcestis. Medea, with an English translation by D. Kovacs, 1994.HERODOTUS, The Persian Wars, vol. 1 (books 1 and 2), with an English translation by A. D. Godley, 1966 [1st edn

1920].HIPPOCRATES, Ancient Medicine, with an English translation by W. H. S. Jones, 1923.HOMER, The Iliad, with an English translation by A. T. Murray, 1954, revised by W. F. Wyatt, 1999.HOMER, The Odyssey, with an English translation by A. T. Murray, 1919, revised by G. E. Dimock, 1995.PLATO, Laws, vol. 1 (books 1–6), with an English translation by R. G. Bury, 1967 [1st edn 1926].PLATO, Phaedo, with an English translation by H. N. Fowler, 1966 [1st edn 1914].PLATO, The Republic, vol. 1 (Books 1–5), with an English translation by P. Shorey, 1963 [1st edn 1930].PLATO, The Republic, vol. 2 (books 6–10), with an English translation by P. Shorey, 1963 [1st edn 1935].PLATO, Theaetetus, with an English translation by H. N. Fowler, 1952 [1st edn 1921].PLUTARCH, Moralia, with an English translation by F. C. Babbitt, 1969 [1st edn 1936].SOPHOCLES, Antigone, with an English translation by H. Lloyd Jones, 1994.SOPHOCLES, Electra, with an English translation by H. Lloyd Jones, 1997 [1st edn 1994].XENOPHON, Hellenica, vol. 1 (books 1–4), with an English translation by C. L. Brownson, 1985 [1st edn 1918].

TRANSACTIONS OF THE PHILOLOGICAL SOCIETY 108, 201040