CONICET | Buscador de Institutos y Recursos Humanos

If psych, then stativeThe problem. The aspectual value of psych verbs, and specially of object-experiencer psychverbs (OEPVs), is still a matter of debate. Arad (1998) i.a. proposes 3 types: (1a) stativereading (Economy concerns Mary) which denotes an (individual-level) state; (1b)eventive reading, where a change-of-state arises in the experiencer with no intentionalagent (The noise bothered Mary), and (1c) agentive reading, where an agent intentionallyacts triggering a change of state in the experiencer (John frightened Mary deliberately).We agree with Arad (1998) in that Spanish dative OEPVs (1a) denote (individual level)states. However, following Marín & McNally, 2011), we assume that accusative (nonagentive)OEPVs (1b) denote inchoative states, a particular type of stage-level states. Theexistence of two distinct subtypes of OEPVs (Marín 2011), molestar- ?bother? andpreocupar- ?worry? verbs (2), is central to any analysis, as it defines divergent data sets(3). In essence, both verb types (preocupar-verbs also with animate subjects; molestartypeonly with non-animate subjects) produce inchoative states (Marín, 2011). Withagentive subject, molestar-vebs denote activities instead.(1) a. El ruido le molesta mucho. ?The noice bothers her a lot?b. El ruido molestó a Uma. / El ruido la molestó. ?The noise bothered Uma?c. Uma molestó a Quentin deliberadamente / para que se fuera.?Uma bothered Quentin deliberately/to make him go away?(2) Type 1: molestar ?bother?: accepts agentive subjectsType 2: preocupar ?worry?: do not take agentive subjects(3) Uma molestó/#preocupó a Quentin deliberadamente / para que se fuera.?Uma bothered/#worried Quentin deliberately / to make him go away?.On the one hand, we note that agentive molestar-verbs pass major dynamicity tests. Theyallow the progressive (4a), serve as infinitival complements of perception verbs (4b), andyield habitual reading in the present (4c). On the other hand, agentive molestar-verbs arenot telic. In fact, they are compatible with for x time ?not with in x time? (5a) adverbials,they cannot serve as complements of acabar or terminar ?finish? (5b), and are not allowedin absolute clauses (5c).(4) a. Uma está molestando a Quentin.b. Vi a Uma molestar a Quentin.c. Uma molesta a Quentin habitualmente / cada semana.(5) a. Uma molestó a Quentin #en/durante dos horas.b. #Uma acabó/terminó de molestar a Quentin.c. #Una vez molestado Quentin, ?bother-verbs diverge by yielding dynamic (agent-controlled) events with defeasible(cancellable) root-named states (6) (Koenig & Davis 2001). This explains the availabilityof Failed Attempt readings and should modals (Martin 2016) giving deontic/epistemic[D/E] reading correlated with ±animacy, with D marking an activity that T should perform.(6) Tarantino ha molestado a Uma durante horas, pero ella no se molesta/no se molestó.?Tarantino bothered Uma for hours, but she isn?t bothered / she didn?t bother.?(7) a. Taratino debería molestar a Uma. b. El muro debería molestar a Uma.?Tarantino should bother Uma.? (OKD/??E) ?The wall should bother Uma? (#D/OKE).Based on similar semantic outputs, Martin et al. posit that some verbs (accomplishments),if agentive, yield a nonculminating reading also present in Romance OEPVs. Yet, in (6),does not reveal the cancellation of molestar endpoint/result (which, as an activity does nothave any culmination to cancel), but rather a zero-Cos predication, as data above shows.Analysis. We propose it all follows from constructional constraints shown by each root.We note that some verbs allow two possible constructions (10), but others (preocupartype)have more specific requirements. Molestar-type verbs allow for either (i) anagentive construction with all the expected aspectual properties (animacy, agentcontrolledevent, dynamicity, defeasible psych state) (cf. defeasible causatives, Martin &Piñon 2012); or (ii) true psych constructions, where the psych state cannot be cancelledand stativity holds as expected. This would be directly constrained by the lexical-semanticproperties of the root. We contend the above facts could be readily explained byproposing a fair workload division balancing a nonradical constructional account wherethe two predications decompose into distinct vP configurations and componentscombined with the root with certain (noticeable) lexical-based properties defining twodistinct root types associated with distinct OEPVs generally subsumed.Proposal. Psych roots like preocup- (and their counterparts in other languages) would berealized through ?psych? state roots (Levin 2006 i.a.). For this root type, only the stativeconstruction, with the psych state as core (nondefeasible) part of the verb?s predicationwould be available. For the molest- type, instead, both constructions are available, asthese verbs would denote an act performed with intention of triggering a certain state, butthis state is not part of the core denotation; therefore, a result does not have to truthfullyoccur for the predication to hold. For agentive|eventive predications, an activity type issyntactically and semantically more tenable, leaving the option to a constructionallyadded (not lexically entailed) result. Crucially, OEPVs are predicted to allow thisalternative along with (non)defeasible results (Martin 2016, Martin & Piñon 2012). Thisallows a correlation between aspectual structure and syntactic realization: in eventive(agentive) constructions, arguments are canonical event participants (canonicalsubject/object positions, Arad 1998), whereas the stative (true psych) construction hasneither an agent nor change event (hence, no canonical subject/object). Options dependon whether the agentive construction ((8)i) is available for a verb. In that case, thesyntactic and semantic layout will be congruent with the type of predication/configurationrequired (activity vs. psych), the result dovetailing with the standard description (onlyagentive/eventive constructions are true transitives, all ?special? properties arise only instative [true psych] constructions, Arad 1998 i.a.) and with ACH (9). (10) picks up oursuggested analysis: taking agentive-friendly OEPVs as default activity predications.(8) (i) agentive reading - external argument - canonical object ? psych effects implied (noncore)(ii) stative reading - no external argument - non-canonical object ? psych state entailed (core)(9) AGENT CONTROL HYPOTHESIS (ACH): Zero-CoS construals require the predicate?s externalargument to be associated with agenthood properties. (Demirdache & Martin 2015)(10) HYPOTHESIS 3 (Martin 2016). Agentive ongoing events are ontologically independent of theireffects. Only nonagentive ongoing events entail their effectsResult. Based on data below we contend that (i) purely psych uses are always stative andthe state cannot be cancelled; (ii) agentivity plays a key role only for some verbs, (iii) 2classes (stative and agentive [eventive]) are relevant and sufficiently different. Thisfollows as a result of a key observation on which the central claim builds: the criteriataken by Arad i.a. to operate the differentiation between possible constructions (whetherthere is an agent deliberately doing something to bring about a mental state in theexperiencer; whether there exists a change of (mental) state in the experiencer) are inOEPVs in complementary distribution and fail to coexist as core part of the predication.

enviar mensaje