Alternative constructions for Romance deadjectival verbs: correlations between eventive and argument structure

MANGIALAVORI RASIA, MARÍA EUGENIA

Ghent

Simposio; NEW WAYS OF ANALYZING SYNTACTIC VARIATION; 2016

Universiteit Ghent

Alternative constructions for Romance deadjectival verbs: relevant correlations between eventive and argument structureCountless proposals have been aimed at elucidating the way in which meaning and morphosyntax of event structure are related (e.g. Jackendoff 1990, Grimshaw1990, Travis 2000, Borer 2005, Hale&Keyser 2002); the general wisdom being that the syntactic projection of arguments correlates non-trivially with event structure. In particular, the causative?inchoative alternation has been the subject of much debate and the source of theoretical innovation (cf. Levin&Rapapport 2005; Chierchia 2004, Reinhart 2002; Ramchand 2008 i.a.). It might also be a case where new theoretical insights can be combined with the use of new technologies providing them with empirical validation.Here, by applying novel theoretical insights on core verbal meaning, we set out to account for certain variability and systematicity of argument structure realizations in relation to the alternatives open by the way a change-of-state[COS] predication can be construed in Romance. Problem►Apart from the alternation between the well-known inchoative/unaccusative and causative/transitive COS structures, Romance renders an additional layout, featuring the Initiator/Causer as sole argument of the deadjectival verb (1). This draws an interesting contrast with languages like English in at least two respects: (i)either participant can be realized in the monoargumental version in Romance; (ii)there is no mechanism guiding the interpretation of the unique DP as Undergoer/Causee. (1) where A=[Causer/Initiator] and B=[Causee/Undergoer]a. El chocolate engorda a los niñosType1:X1-V-X2 ?Chocolate makes kids fat? b. Los niños engordan Type2:X2-V?Kids fatten [get fat/gain weight]?c. El chocolate engordaType3:X1-V?Chocolate fattens [is fattening]? Data&Analysis►A basic rule of event composition states that the event structure of COS verbs combines two basic components: cause and process; the former, when present, causally implicating the latter (cf. Ramchand 2008 i.a. for overview). These subevents are (allegedly) ordered in a hierarchical embedding relation ?in a constructivist approach, participant relations being built up recursively from successively embedded (sub)event descriptions in first-phase/lexical-syntax. However, a closer look on (1)-(3) suggests that the causational and the process components may be equally optional. Hence, we must define whether we are facing a complex structure with unrealized projections/arguments; or if either subevent (with the corresponding projection filled by the DP denoting the subject of the particular subevent) can occur independently, assuming that there is a general combinatorial semantics that interprets the (lexical/first-phase) syntactic structure in a regular and predictable way, and that the semantics of event structure and event participants is read directly off the structure (Ramchand 2008:42), and not directly off information encoded by lexical items.As for the first option, absence of the Theme/Causee/Undergoer could be taken as indication of this eventive portion not being encoded in the verb, but optionally construed as one of many alternatives for deriving a property-denoting root into a verb. Otherwise, cases like ((1)c) would instantiate an argument related to a projection absent in the lexical entry (InitP/VCAUSE), and a process projection lexically specified taking no argument (thus violating the requirement that all subevental projections must have a filled specifier, Ramchand, 2003: 27) . Composite role hypothesis discarded by impossibility of reflexive morphology (2)?and since Romance does not provide us with any overt morphological indicator of such derivational process or eventive complexity?another option would be to evaluate if relevant evidence can be gleaned from the aspectual properties of these constructions, assuming that the absence of an Undergoer/Causee shall correlate with the absence of its corresponding eventive portion (i.e., process/change). Indeed, whereas (deadjectival) COS verbs in English are prototypically argued to be [proc] verbs (Ramchand 2008:91), data indicates that Romance Causer-only alternatives do not comprise a process portion. Corpus and analytic data?comprising nearly 30 deadjectival Spanish verbs and its corresponding Italian/Portuguese/Catalan equivalents tested on specific telicity/durativity diagnostics?show that Type3 is consistently stative: unlike Type1/2, it only occurs in present tense (3) and fails to accommodate duration(framing)/endpoint adverbials (4) and maximality/culminative modifiers (5) ?namely, queries on POS-tagged corpus returned only Type2 hits for the string [NN*][VIS*] (no Type3 hits) and occurrences with relevant adjuncts ([NN*][VP*][*mente].[R*]); surveys on natives report serious oddity or ungrammaticality given these combinations and a non-significant set of exceptions where the construction is tolerated but forcing interpretation of the sole argument as Causee. Interestingly enough, analytic data also showed that proportional modifiers (mucho/bastante) are allowed only if interpreted as a degree scope on the capacity of the causer to determine the corresponding change. The situation holds a relevant correlation between eventive and argument structure: as soon as a causee/undergoer is added (even a generic/defective one), a process (COS) event obtains. In this sense, the additional construction contributed by Romance provides clear evidence in favor of Ramchand?s (2007) intuition about InitP(VCAUSE) being a state. Further tests in languages with more reliable ergativity diagnostics (e.g., auxiliary selection/ne-clitization/passivization) confirm that, unlike Type2 (consistently unaccusative), Type3 show consistent unergative patterns.(2)Las luces (*se) enceguecen.cf. La gente (se) enceguece.the lights SE.REFL go blind the people SE.REFL go blind (3)El chocolate {está engordando/engordó} *(a todo el mundo/a la población).?Chocolate {is making/made} (everybody/the population) fat? (4)El chocolate *(te) engorda {por un tiempo/en pocos días}?Chocolate makes *(you) fat {for a while/in few days}?(5)El chocolate engorda {*totalmente/*completamente/mucho/bastante/seriamente}?Chocolate can make somebody {totally/completely/a lot/enough/seriously} fat? (6)a.#El ruido ensordece ?Noise causes deafness?b.# El aire engorda ?Air causes fatness?Results&Conclusions►Two possible derivations are distinguished: incorporation of the root into (unaccusative V0) en route to the causative verbal head (V0)?i.e., those DVs delivering a COS predicate with a necessary undergoer/causee?and direct conflation of the property-denoting root into the unergative phonologically null verb?i.e., those DVs that only involve a causer and are consistently stative. The operation is relatively free, and restrictions are essentially encyclopedic (the only argument needs to feature some property allowing them to instigate the corresponding change (6)).The alternative syntactic structures may be freely built up by conflation/incorporation instead of requiring lexical tagging?thus, lexical items may not necessarily provided of minimal syntactic information constraining insertion in syntactic structures. The proposed syntax avoids the requirement of finer specification on the roots forcing the nonprojection of a specifier. Finally, the systematicity of the phenomena under discussion consistently points towards structural composition rather than lexical encoding. In this sense, theoretical and methodological innovation can be seen to converge in the detection and analysis of underexplored patterns.Selected references: Jackendoff, R.1990. Semantic structures. Cambridge: MITPressHale, K & Keyser, S J.2002. Prolegomenon to a Theory of Argument Structure, Cambridge: MITPressRamchand, G.C.2008. Verb meaning and the lexicon: a first-phase syntax. Cambridge:CUP