797-HOOP-0-0

Case marking strategies Helen de Hoop & Andrej Malchukov1 Radboud University Nijmegen DRAFT January 2006 Abstract Two strategies of case marking in natural languages are discussed. These are defined as two violable constraints whose effects are shown to converge in the case of differential object marking but diverge in the case of differential subject marking. The strength of the case bearing arguments will be shown to be of utmost importance for case marking as well as voice alternations. The strength of arguments can be viewed as a function of their discourse prominence. The analysis of the case marking patterns we find cross-linguistically is couched in a bidirectional OT analysis. 1. Assumptions In this section we wish to put forward our three basic assumptions: (1) In ergative-absolutive systems ergative case is assigned to the first argument x of a two-place relation R(x,y). (2) In nominative-accusative systems accusative case is assigned to the second argument y of a two-place relation R(x,y). (3) Morphologically unmarked case can be the absence of case. The first two assumptions deal with the linking between the first (highest) and second (lowest) argument in a transitive sentence and the type of case marking. For reasons of convenience, we will refer to these arguments quite sloppily as the subject and the object respectively, although we are aware of the fact that the labels subject and object may not be appropriate in all contexts, dependent on how they are actually defined. In many languages, ergative and accusative case are assigned only or mainly in transitive sentences, while in intransitive sentences ergative and accusative case are usually not assigned. In that sense, we may call ergative and accusative ‘dependent’ cases, following Marantz (1991), since ergative and accusative crucially depend on the presence of another (core) argument (direct object and subject, respectively) in the clause. Unlike nominative case which is closely connected to the grammatical function of subject and which can combine with different thematic roles, ergative and accusative case are thematically more restricted in that they are mainly used for agents and patients, respectively. This cannot be reversed, since agents of intransitive sentences do not get ergative case in many languages, nor do intransitive subjects that fulfill the role of patient get accusative usually (although exceptions exist, as will be exemplified below). That brings us to the third assumption. In Chomsky (1981) and subsequent work, a notion of abstract case (usually referred to as Case with a capital C) is used in connection to the inviolable Case Filter, that requires every lexically realized DP to bear case. Since many languages have little or no case morphology, the postulation of abstract Case guarantees DPs in structural case positions to be saved from the Case Filter. In this article, we take an Optimality Theoretic perspective (Prince and Smolensky 2004). Therefore, we assume that linguistic constraints such as the Case Filter, are violable in nature. Hence, we do not need the postulation of abstract Case. Some level of abstraction may still be needed, however, but in the absence of explicit evidence we simply assume that ‘what you see is what you get’. Thus, 2 we interpret the absence of morphological case-marking as the absence of case (Aissen 1999; Aissen 2003). This also means that what is referred to as nominative or in other contexts absolutive case, can sometimes be viewed as the absence of case. This holds for example for the nominative case-marked subject in Hindi in sentence (4) below, but not for the nominative case-marked subject in Japanese in (5): (4) Raam-∅ ek bakre-ko bec-taa hae. Ram-NOM one goat-ACC selling is ‘He sells the goat.’ (5) Boku -ga tomodati-ni hana-o ageta. I-NOM friend-DAT flowers-ACC gave ‘I gave flowers to my friend.’ In (4) the name Ram is glossed as being in the nominative, despite the fact that there is no such thing as a nominative marker in Hindi, the proper noun in (4) is in its unmarked, uninflected form. By contrast, there is a clear case marker, viz. –ko, on the direct object in (4). The uninflected word form is called nominative, but in fact, nominative case in Hindi can be viewed as the absence of case. By contrast, the nominative case in Japanese is expressed by a real case marker, and therefore we cannot equate nominative case with the absence of case in Japanese, unless the case marker gets dropped which is often the case in colloquial speech (Fry 2001; Lee 2002). In ergative case systems, the absolutive case is often unmarked and we will similarly assume that in the absence of morphological case marking, absolutive case is in fact the absence of case as well. This is illustrated in sentence (6) from Yup’ik. 3 (6) Angut-em tangrr-aa arnaq-∅. Man-ERG sees woman-ABS ‘The man sees the woman.’ 2. Case marking: two basic functions In this section we argue, following functional-typological insights, that two basic functions of case marking can be distinguished, the identifying function and the distinguishing one (Mallinson and Blake 1981; Kibrik 1985; Comrie 1989; Song 2001). The identifying strategy makes use of case morphology to encode specific semantic/pragmatic information about the nominal argument in question. We say that case morphology is used to identify semantic or pragmatic properties. Lexical (inherent, oblique) as well as semantic cases are obvious examples of the identifying strategy (Butt and King 2003; Butt and King 2004). A clear example are the locative cases in Finnish: (7) talossa house-INESSIVE ‘in the house’ talosta house-ELATIVE ‘out of the house’ talon house-ILLATIVE ‘into the house’ talolla house-ADESSIVE ‘on the house’ talolta house-ABLATIVE ‘off the house’ talona house-ESSIVE ‘as a house’ The lexical cases above clearly contribute to the meaning of the noun phrase, comparable with the meanings the prepositions in the English translations contribute. The identifying function of case is not restricted to lexical or semantic cases, however. In fact, structural or 4 grammatical cases identify certain semantic/thematic properties to a certain degree as well. For example, accusative case in direct object position can be argued to identify patienthood. Dative case in some languages is a structural case as well, yet it is clearly associated with thematic roles such as goal and experiencer. Ergative case is in many languages associated with ‘true’ agents. In Manipuri, for example, the ergative case on the agent in (8) marks high agentivity (the agent is in control, volitional), while a decrease in agentivity is signalled by the lack of ergative case in (9) (Bhat and Ningomba 1997). (8) əy-nə tebəl-də theŋŋi. I-ERG table-LOC touched ‘I touched the table (volitionally).’ (9) əy tebəl-də theŋŋi. I table-LOC touched ‘I touched the table (involuntarily).’ In Manipuri, all and only true agents get ergative case. Thus, we can say ergative case in Manipuri identifies agentivity. We introduce a general constraint which states that ergative case identifies strong subjects (which we will write as A). (10) IDENTIFY A/ERG: Ergative case identifies strong subjects (A ↔ ERG). Clearly, the constraint IDENTIFY must be conceived of as a family of constraints and not a sole constraint. The notion strength will be elaborated upon in section 3. The distinguishing strategy is a more specific strategy that is used for distinguishing between the two core arguments of a transitive clause, i.e. the subject and the object. The 5 intuition behind the distinguishing function is quite clear. When a two-place predicate R(x,y) is used to describe an event involving two participants, usually an agent and a patient, it is of utmost importance to avoid ambiguity as to which noun phrase corresponds to the first argument x (the agent) and which to the second argument y (the patient). For this purpose, case can be used to mark one of the arguments. If one argument is case marked, this already suffices for the purpose of disambiguation. Thus, from the distinguishing perspective, there is no need to case mark both arguments. Neither would it be necessary to case mark the one and only argument of a one-place (intransitive) predicate. Indeed, it has been argued that in many nominative-accusative case systems only the y is case marked (with accusative case) while the x remains morphologically unmarked. This view is in accordance with our assumption (3) presented above. When nominative case is the unmarked (uninflected) case form, we interpret it as the absence of case. Similarly, in pure ergative-absolutive systems only the x is case marked, while the y remains morphologically unmarked (absolutive). The only argument of an intransitive verb is unmarked as well, and although it is labeled absolutive or nominative, it can often be seen as the absence of case as well. While Manipuri is an example of a radically identifying language, there are also languages which can be characterized as radically distinguishing. In Awtuw the object is obligatorily marked with accusative case if the object is equally high or higher than the subject in the animacy hierarchy (Feldman 1986): (11) Tey tale-re yaw dæli 3FS woman-ACC pig bit ‘The pig bit the woman.’ 6 (12) Tey tale yaw dæli. 3FS woman pig bit ‘The woman bit the pig.’ In Fore, another Papuan language, it is the subject that is marked with ergative case if the object is higher in the animacy hierarchy than the subject (Scott 1978): (13) Yagaa-wama wá aegúye. pig-ERG man hit ‘The pig hits the man.’ (14) Yagaa wá aegúye. pig man hit ‘The man hits (or kills) the pig.’ In the Fore example in (14), the man is higher in the animacy hierarchy than the pig, and that is why man is interpreted as the subject, even though the canonical SOV word order is overruled. If the speaker wants to express that the pig hit the man, then the subject needs to be explicitly marked as the subject (as in (13)). The distinguishing function of case can be characterized as a global constraint like in Fore (i.e., the relative animacy of subject and object are measured) or as a local one as in classical cases of the markedness effects in differential object marking, where for instance the animacy or definiteness of the object is evaluated independent of the animacy or definiteness of the subject (see de Swart 2003 for more discussion and more examples of global distinguishability). To put the general motivation behind this type of case marking, whether locally or globally applied, in a constraint (de Swart 2003; de Hoop and Lamers to appear): 7 (15) DISTINGUISHABILITY: The two arguments of a transitive clause should be distinguishable. Case marking is a way to distinguish between the subject and the object and hence to satisfy the above constraint. However, if the subject and object are otherwise distinguishable (like when in Fore the subject outranks the object in animacy), then case marking is not necessary to satisfy DISTINGUISHABILITY. But if the object is more ‘subject-like’ (absolutely or relatively), i.e., if it equals the (general or actual) subject in animacy/definiteness, the subject and the object can no longer be distinguished on the basis of these animacy/definiteness properties. In order to satisfy the constraint DISTINGUISHABILITY and to avoid potential ambiguity, case marking can apply. Cross-linguistically, a merely distinguishing function of case is rare. This could be explained by the fact that there are alternative strategies for disambiguation between the two arguments of a transitive predicate, such as for instance the use of subject agreement, word order restrictions, context and/or intonation (Keenan 1978; Bouchard 2001; de Hoop and Lamers to appear). For example, when in Fore the two arguments are equal in animacy, solely word order determines what is the subject and what is the object: the first noun phrase will then be interpreted as the subject Obviously, the identifying and the distinguishing function are not orthogonal, but overlap considerably. In fact, if case is systematically used to identify the subject or the object in a transitive clause, then of course we get differentiation ‘for free’. Therefore, case systems that are completely based on the function of identification must be richer in case morphology than the mainly distinguishing ones. But as we will see, both functions are needed to account for the various case marking patterns we find across languages. 8 3. Case and Optimality Theory In the previous section we have introduced two basic case marking constraints, DISTINGUISHABILITY and IDENTIFY, that we will use in the remainder of this article to account in an Optimality Theoretic fashion for cross-linguistic case marking patterns. But first, we will briefly discuss how our account relates to some previous Optimality Theoretic analyses of case marking. A well studied paradigm of differential case marking is differential object marking: in many languages DPs higher in animacy or definiteness are marked, while those lower may be not (Bossong 1985; Aissen 2003). For example, in Turkish specific objects are case marked with accusative case, while non-specific objects remain without case (Enç 1991): (16) Ali bir kitab-ı aldı. Ali one book-ACC bought ‘Ali bought a certain book.’ (17) Ali bir kitap aldı. Ali one book bought ‘Ali bought some book or other.’ An explanation of differential object marking in terms of distinguishability is straightforward and elegant (Aissen 1999, 2003). In a canonical transitive construction, the object is lower than the subject in animacy/definiteness, and thus when the object is animate/definite it is marked (for an object) which means it should be (case-)marked. Aissen crucially combines markedness of form with markedness of meaning within her constraints. Thus, she uses constraints such as “Avoid a specific object which is not case marked” (*Oj/Indef Spec & 9 *ØC) and which is universally ranked higher than “Avoid a non-specific object which is not case marked” (*Oj/Indef Nspec & *ØC). By inserting an economy constraint such as “Avoid case marking” (*StrucC) in between these two constraints, Aissen accounts for the Turkish pattern in which specific indefinite objects are case marked while non-specific indefinite objects are not. The ranking Aissen proposes for Turkish differential object marking is therefore: *Oj/Indef Spec & *ØC >> *StrucC >> *Oj/Indef Nspec & *ØC. We will briefly discuss three problematic aspects of Aissen’s approach. The first problem, pointed out by de Swart (2003), is that we encounter cases where distinguishability is globally evaluated, rather than locally. In those cases, markedness applies to the relation between the subject and the object in that the subject should outrank the object in animacy, and not just to the features of the object (inanimate) or the subject (animate) itself. De Swart discusses the following two examples from Malayalam. In this language, usually only animate direct objects are case marked. This could straightforwardly be accounted for in Aissen’s framework by a constraint ranking such as “Avoid an animate object which is not case marked” (*Oj/Animate & *ØC) >> “Avoid case marking” (*StrucC) >> “Avoid an inanimate object which is not case marked”(*Oj/Inanimate & *ØC). However, when both the subject and the object are inanimate ánd there is a danger of ambiguity, the inanimate object gets case marked. This is illustrated below (Asher and Kumari 1997): (18) Kappal tiramaalakae bheediccu. ship waves-ACC split “The ship broke through the waves.” (19) tiramaalaka kappaline bheediccu. waves ship-ACC split “The waves split the ship.” 10 Thus, the inanimate objects in (18) and (19) are case marked, but only in sentences like these, in order to distinguish them from the inanimate subjects. However, this cannot be accounted for in terms of constraints such as Aissen’s. In her framework we can only rank the economy constraint above the constraint that requires marking of inanimate objects and then no inanimate objects would be case marked, or we can rank it below, in which case all inanimate objects would be case marked. Hence, unlike local differential case marking (e.g., case marking of all and only animate objects) instances of global differential case marking (case marking of the object depending on the properties of the object and the subject) cannot be accounted for by Aissen’s approach (see section 4 below for further discussion of differential case marking in Fore and Awtuw, driven by global distinguishability). The second problem for Aissen’s analysis of differential case marking, pointed out by de Swart (2003) and Woolford (2001; Woolford 2004), is that DISTINGUISHABILITY would predict that differential subject marking would mirror differential object marking in the sense that inanimate and indefinite subjects (which are more ‘object-like’) would be case-marked. This prediction is actually borne out in some languages, such as Qiang (a Tibetan language), where the subject in a transitive clause only takes agentive case when it is inanimate (Lapolla 2003). (20) MoVu-wu qa datuəZ. wind-AGT 1sg knocked-down ‘The wind knocked me down.’ But in fact, such examples are rare. More often, DISTINGUISHABILITY effects in subject case marking manifest themselves in a split between nouns and pronouns. In many split-ergative languages, (first/second person) pronouns do not get ergative case marking when they are the 11 subject of a transitive verb, while nouns do (Silverstein 1976; Aissen 1999). By contrast, (first/second person) pronominal direct objects receive accusative case marking while nominal objects do not. A well-known example of this pattern is Dyirbal (Dixon 1979): Case marking in Dyirbal 1st, 2nd person pronoun 3rd person pronoun noun Transitive subject ∅ -ŋgu -ŋgu Transitive object -na ∅ ∅ Table 1: Dyirbal Although such a pattern can readily be explained in terms of (local) DISTINGUISHABILITY, since first and second person DPs more often occur as subjects than third person DPs, it does not account for the rarity of the animate/inanimate subject split. This was also pointed out by de Hoop and Narasimhan (de Hoop and Narasimhan 2005; de Hoop and Narasimhan submitted) in relation to the case alternations found in Hindi. While the object case alternation in Hindi correlates with animacy and/or specificity of the object (accusative case marks an animate or specific object), the subject case alternation is mainly triggered by the perfectivity of the transitive sentence. That is, ergative case is assigned in the context of a perfective verb phrase. Nominative case is the unmarked (morphologically zero) case which functions as the elsewhere case for both the subject and the object (see the sentences below). (21) Wo ek laD.kaa / ek laD.ke-ko dekhtaa hae. he one boy / one boy-ACC seeing is ‘He sees a boy / the boy.’ 12 (22) Us-ne ek laD.kaa / ek laD.ke-ko dekhaa. he-ERG one boy / one boy-ACC saw ‘He saw a boy / the boy.’ In Aissen’s (1999, 2003) approach, someti