Home
ACLC PhD project: “Gradience in split intransitivity: a typological investigation” (starting date: 1 December 2005)
By examining the nonhomogenous behaviour of intransitive verbs in various diagnostic contexts (‘unaccusativity tests’) crosslinguistically, I try to reach deeper insight into the two-way relationship between the lexico-semantic parameters of split intransitivity and their morpho-syntactic expression. ‘Split intransitivity’ (also known as ‘unaccusativity’) refers to a fundamental dichotomy operating at the level of intransitive verbs: some of these verbs (commonly labelled ‘unergatives’) pattern with active transitives in their argument selection and morpho-syntactic distribution; some others (the ‘unaccusatives’) take arguments that are similar to transitive objects and also occur in distinct, characteristic configurations. For instance, the French intransitive verb in (1a) selects an agent subject and the perfective auxiliary HAVE, similarly to the transitive verb in (1b), whereas in (1c) the subject is a patient and the verb selects the auxiliary BE, just like in (1d), which contains a passive construction. (1) a. Paul a travaillé toute la journée. (unergative) French Paul has worked all the day ‘Paul (has) worked all day long.’ b. Marie a écouté le discours. (transitive) Marie has listened the speech ‘Marie (has) listened to the speech.’ c. La neige est tombée sur Rennes . (unaccusative) the snow is fallen on Rennes ‘The snow has fallen/fell on Rennes .’ d. Le gâteau est préparé par le cuisinier. (passive) the cake is prepared by the cook Crosslinguistically, split intransitivity seems to occur in a large diversity of particular forms, often converging to a series of patterns due to genetic filiation and areal distribution. For example, the split BE/HAVE perfective auxiliary selection is a family feature, characteristic of some of the West Germanic and Romance languages, as well as most probably an areal feature, since it also characterises non-related Basque. The effects of the split can be visible both in the syntactic and in the morphological structure of a language. They generally have to do with argument identification, quantification, and control, as well as with predicate valence, voice, inherent lexical aspect and temporal-aspectual frame. The split is also restricted by the way verbs cluster in lexical (semantic field) classes. In many cases, split intransitivity is directly reflected in the distinct marking of the core argument (whichin various languages can berealised either by case inflections on the nominals or by pronominal agreement, and often also by word order). In Acehnese, for instance, an Indonesian member of the large Austronesian family, the split is marked both by the choice of pronominal clitics, and by the fixed word order (Durie 1987, Foley 2005). (2) a. geu-poh- geuh (transitive) Acehnese 3 SG :A-hit-3 SG :P ‘he hit him’ b. geu-jak (unergative) 3 SG :A-walk ‘he is walking’ c. rhët- geuh (unaccusative) fall-3 SG :P ‘he is falling’ The criterion which underlies this dichotomy in Acehnese is mainly a matter of argument semantics, since in this language [+volition] arguments are always expressed as agent clitics, whereas nonvolitional arguments are rendered as patient pronominal markers. Semantic features of the predicate play a smaller role in this type of split. In other languages, the role of the aspectual features of the predicate can be just as important as that of argument semantics. While the distinction between bounded and unbounded states of affairs is of little consequence in languages like Acehnese, in many European languages aspectual features, be they inherently characteristic of a verb or compositionally derived, are often crucial in defining the dichotomy between unergative and unaccusative expressions. For example, a Dutch activity motion verb, normally taking an agent argument and selecting HAVE as a perfective auxiliary, must occur with BE if accompanied by a directional prepositional phrase, which induces a telic interpretation to the whole predicate: (3) a. Bram heeft ruim een uur gezwommen. (unergative) Dutch Bram has about one hour swum ‘Bram has swum/swam for about one hour.’ b. Bram is naar de oever gezwommen. (unaccusative) Bram is to the bank swum ‘Bram has swum/swam to the river bank.’ Perfective auxiliary selection in Dutch is determined first of all by the telicity value of the predicate and only secondly by the agentivity of the argument (Zaenen 1993). Because of this ranking of basic features in the intransitivity split, telic Dutch change of place predicates occur, as a rule, with BE, irrespective of the subject agentivity (e.g. hij is gekomen ‘he has come’). No such restriction seems to apply in languages like Acehnese. Split intransitivity mayalso reflect the distributionofpredicates into aspectual classes, and even into semantic field (lexical) classes. In Georgian, for instance, in the aorist series, the only class of verbs whose (third person) subjects receive special ergative marking is that of (controlled and uncontrolled) activities. The Guaraní (a Tupí-Guaraní language from Paraguay ) subject marking distinction operates between states, on the one hand, and events, on the other hand. In the polysynthetic languages of North America the class of bodily activity verbs (containing verbs such as ‘sneeze’ or ‘cough’), with an obviously lower degree of subject agentivity than ‘walk’ or ‘play’, may fall either in the unergative or in the unaccusative type, depending on the feature(s) according to which the split intransitivity cut-off point is set in each language system. ‘Hiccough’ and ‘sneeze’, expressing basically uncontrolled activities, behave as unergatives in Lakhota (Siouan), because in this language the relevant feature in the split is agentivity, but they act as unaccusatives in Central Pomo (Pomoan), where all noncontrol affected arguments are marked as patients (Mithun 1991). The most thought-provoking thing about split intransitivity is not that it takes such diverse forms crosslinguistically (which is more or less to be expected), but that it allows for a lot of language-internal variation. Some of the variation can be explained as resulting from processes of lexicalisation, where the verb comes to have a new meaning, which places it in a lexical class that would normally take the opposite marking. For instance in Mohawk (Iroquoian), the original meaning of ‘ I (patient) will ride (horseback)’ is ‘it will carry me on its back’, which explains the nonagentive agreement marking (Mithun 1991). Variation may also be the effect of borrowing, as is the case of the Guaraní verb avuří ‘be bored’, fromSpanish aburrir (se).Evenif the verbrefers to a state, it is still used with eventive markers, just like all the other verbs in this language that were borrowed from Spanish (Mithun 1991). Other cases of variation reflect more systematic oppositions. There are languages, such as Acehnese, which make extensive use of the alternation between agent and patient argument markers, for instance to indicate that the action is performed willingly, or controllably vs. unintentionally, or uncontrollably (Durie 1987, Foley 2005): (4) a. gopnyan ka-matê (unaccusative) Acehnese 3 SG PERF -die ‘he died’ b. rila ji-matê (unergative) ready 3 SC :A-die ‘he’s ready to go to his death’ In Tsova-Tush (North-East Caucasian), about two thirds of the intransitive verbs can take either of the split intransitivity markings, depending on the degree of control exercised by the argument. However, with some of these verbs variation is exceptional, whereas with some others it is quite flexible. For example, ‘die’, ‘burn’, and ‘become old’ mostly occur with patient marking, while ‘wash’, ‘laugh out once’, and ‘begin’ usually take agent marking. ‘Lose weight’, ‘slip/slide’, ‘be late’ can select either an agent or a patient, and so can ‘sleep’, ‘stand’, ‘sit’, ‘lie’, and ‘hang’ (Holisky 1987, Merlan 1985, Dixon 1994). (Verbs of body posture and position in general can have a dynamic (‘assume position’) or a static (‘maintain position’ or ‘simple position’) interpretation – Rappaport Hovav & Levin 2000). In Central Pomo, affectedness, aspect, and the speaker’s empathy each play their own part in the split marking of the arguments. First of all, inherent states (e.g. ‘live (in aplace)’, ‘be lazy’) usually select agent markers, as opposed to temporary states, such as ‘be sad’, ‘be sleepy’, which associate with patient markers (Mithun 1991). The referent of the argument in a temporary state (involving transience and thus change; a ‘stage level predication’ in Carlson’s (1977) terminology) is perceived as being more affected than the referent indicated by a permanent state argument (‘individual level predication’ – ibid.). Secondly, inchoative constructions (e.g. ‘I have got old’) also select patient markers, since the participant is more affected by the coming to being of a state than by simply being in a state (e.g. ‘I am old’). Thirdly, if the effect of a state or an event on the participant is not important or evident to the speaker (e.g. ‘he feels warm’), the speaker will choose agent marking by default, and patient marking only in order to express empathy with the participant (Mithun 1991). More variation can be observed when studying the mismatching results of unaccusativity tests in particular languages. For instance, it has been noted that in Dutch there are (low agentivity) verbs which do pass the perfective auxiliary selection test as unergatives (they consistently select HAVE), but fail to pass the impersonal passive test (Perlmutter 1978, Zaenen 1993): (5) a. De man heeft gebloed. (unergative) Dutch the man has bled. ‘The man (has) bled.’ b. * Er werd (door de man) gebloed. there PASS.PAST.SG (by the man) bled. ‘There was bled (by the man).’ Similarly, there are some (low telicity) Dutch verbs which occur by rule with the auxiliary BE, but do not allow their past participles tobeusedprenominally unless their degree of telicity is increased compositionally (e.g. * de gebleven man ‘the remaining man’ vs. de hier gebleven / weggebleven familie ‘the family remaining here / away’). The presence of this kind of variation indicates that each and every split intransitivity diagnostic test in a specific language might be organized on different parameters and constraints and for this reason it is bound to give distinct results, picking out a different list of verbs / predicates. Interestingly, however, some verbs seem to be less prone to variation, being more ‘prototypically’ unergative or unaccusative, respectively, than others. Across languages, highly agentive atelic activities (e.g. ‘play’, ‘walk’) appear to occur most consistently in the unergative class, whereas uncontrolled telic events and states, with highly affected subjects, (e.g. ‘fall’, ‘be broken’, ‘be afraid’) most often fall into the class of unaccusatives. The class assignment of other types of verbs / predicates to the unergative or unaccusative class is less predictable and it may depend on the distinct parameter setting specific to every language. Unaccusativity has often been presented as a binary phenomenon, but such an approach has been faced with difficulties in accounting for the considerable variation that occurs both cross- and intra-linguistically. Recently, attempts have been made to analyse it as a gradient phenomenon (e.g. Sorace 1993, 2000; Foley 2005; etc.). Sorace has proposed an intransitive verb hierarchy, based on the variation in perfective auxiliary selection in languages like Italian and French. Since telicity plays a crucial role in this unaccusativity test, Sorace’s hierarchy spans from the high telicity end (verbs expressing change of location and of state)to the low telicityend(controlled processverbs). (6) Sorace’s (Perfective) Auxiliary Selection Hierarchy change of location (e.g. come) selects BE (least variation) change of state continuation of state existence of state uncontrolled process controlled process (motional) controlled process (non-motional) (e.g. work) selects HAVE (least variation) Foley’s hierarchy, on the other hand, (inspired by Dowty 1991) is based on variation in argument semantics. For this reason, it ranks volition as the highest agentive (‘actor’) feature and change of state as the lowest feature (specific of a patient, or ‘undergoer’). Since telicity does not play any role in this hierarchy, agentive change of location verbs (such as ‘come’) are not predicted to fall in the unaccusative category. (7) Foley’s Macro-Role Hierarchy Actor: volitional performer causing an event or change of state sentience movement stationary causally affected incremental theme Undergoer: undergoing a change in state Both hierarchies seem to be very useful for the description of various languages, synchronically as well as diachronically. Sorace’s hierarchy, for instance, has been employed to account for the evolution of the perfect auxiliary selection in some Romance and Germanic languages. Foley, on the other hand, has used his hierarchy to analyse a series of Austronesian language systems. Split intransitivity has been under the intense scrutiny of linguists representing a large palette of theoretical approaches for more than a quarter of a century, including defendants of relational grammar, generative grammar, role and reference grammar, etc. However, no general agreement seems to have been reached on either the semantic properties, or on the structural configurations characterising the split. Apparently, it involves and connects the morpho-syntax and the semantics of both the argument and the predicate, at a lower lexical level, as well as a higher configurational level, with essential implications to the discourse pragmatics. It is aterrainwhere all the levelsof the language meet and make their own contribution to encoding the linguistic expression. This makes it describable as either an upwards projection, from the lexicon to syntax and discourse, or as a downwards projection, from discourse and syntax into to the lexicon, though I believe that choosing any of these perspectives exclusively is prone to tilt the analysis to a biased view of things. The category of voice is a relevant example, as it shows crosslinguistically that the same information on the discourse prominence of an argument can be just as well coded in the split argument marking, or in the inherent or derivative verbal morphology, or in the syntactic structure, or quite often in an intricate combination thereof.References
Carlson, G. (1977). Reference to Kinds in English. Ph.D. Diss., UMass, Amherst. Published (1980). New York : Garland . Dixon , R.M.W. (1994). Ergativity. Cambridge : CUP. Dowty, D. (1991). ‘Thematic proto-roles and argument selection’. Language 67: 547-619. Durie, M. (1987). ‘Grammatical relationsinAcehnese’.StudiesinLanguage 11:365-99.