Michael MulyarMichael Mulyar11/7/05Unaccusativity and Coercion: A Temporal Analysis of the Russian Gen-Neg ConstructionUnaccusative Hypothesis (UH)- All intransitive verbs are either unaccusative or unergative. - Unaccusativity correlates with patienthood; unergativity correlates with agentivity(Perlmutter 1978, Dowty 1991).- Recent work has challenged syntax-semantics isomorphism put forth in UH: External / internal argument distinction not made in generative models (Sorace 2002).- Variable syntactic behavior cross-linguistically motivates semantic classes of unaccusative verbs (Levin and Rappaport Hohav 1995).- Several syntactic tests for unaccusativity have been proposed for Russian (the genitive of negation construction as well as the distributive po- construction). These tests are intended to reveal verbs whose subjects are syntactically internal arguments (e.g. subject-raising verbs in HPSG).Diagnostics of Unaccusativity in Russian- Distributive po-construction - Genitive of negation construction- Semantic tests, including those proposed by Levin and Rappaport Hohav 1995 forEnglish: Immediate Cause (unergativity), Directed Change and Existence.- Unaccusativity sensitive to both telicity and agentivity. Babko-Malaya 1999 relates both notions to argument structure. Unaccusative and reflexive verbs havean NP or bind a trace in Spec[VP]. Unergative verbs do not have an NP in Spec[VP]; the NP is base-generated in Spec[AspP]. Telicity is a feature checked in the VP projection and agentivity is a feature checked by aspect. This is consistent with the observation that variable behavior is found with telic agentive and atelic non-agentive verbs (Sorace 2002 suggests that variable behavior is subject to a semantic hierarchy).(Gradient) Tests for Unaccusativity: The Distributive Construction- Distributive construction (roughly means, objects distributed from a set, i.e. a one-to-one or many-to-one mapping)- Po- construction is impersonal with intransitive verbs. Unaccusativity test (Pesetsky 1982); since an unergative verb lacks an internal argument, an internal argument cannot be raised to subject (as required by distributivity, i.e. an object must be extracted from a set). - The constraint is violable, as will described below; gradient violations can be explained via verb aktionsart.- Examples (po seems to be a quantifier, semantically something along the lines of total of X from set Y):(1) S kazhdogo dereva, padaet / upalo po yabloku.From each tree, is falling / fell po (total of) appleFrom each tree, there falls / fell one apple(2) V kazhdom classe, boleet / bolelo po tri studenta.In each class, there are / there were sick po (total of) three studentsIn each class, there were sick three students(3) Na kazdoi ploschadke, ?/*begaet / ?begalo po sobacke. On each playground, there runs / ran po (total of) dogOn each playground, there runs a dog(Example from Romanova 2004)(4) V kazhdoi pesne, ?/*poet / ?pelo po dva pevca.In each song, there sings / sung po (total of) two singersIn each song, there sing two singersInterpreting the Po- test - (1) and (2) are fully acceptable both in the present and past impersonal construction; (3) and (4) are gradiently unacceptable, with the present impersonal construction basically unacceptable.- Unaccusativity Diagnostic- Gradient acceptability with unergative verbs such as run and sing. Romanova’s example (3) basically unacceptable with present tense.- Gradient result seems to indicate that the construction may be sensitive to factors other than unaccusativity; present tense test, however, suggests thatan activity or imperfective interpretation is not possible. Thus, the event cannot be viewed “from the inside”; wherever, an interpretation is possible, the event is viewed instantaneously.- Event may be viewed as an accomplishment (or most likely an achievement) with a past impersonal construction (e.g. begalo in (3)). - This suggests that an activity may be coerced into an accomplishment or achievement, e.g. in (3) and (4). Coercion less acceptable with a present impersonal, since the present may access an interval of the activity. If a reading is possible, it is only in an accomplishment or an achievement interpretation, i.e. an activity of each dog running cannot be accessed via (3).- If bolet’ (to be sick) is an unaccusative (Schoorlemmer 2002), we would expect it to pass the po- test, since its internal argument may be raised from the distributive phrase. Likewise, if bolet’ is viewed as an accomplishment (i.e. an accomplishment followed by a state of sickness), it should be acceptable in (2). Semantically, this seems to be the case, e.g. (2) does not suggest a continuing state of illness, but only the statement concerning the condition (this might tested by adding a durative expression, which makes (2) basically unacceptable).- Thus, the condition on the test might be stated as unaccusativity or it might be stated as delimitedness. If the latter is more general, an unaccusative reading is arguably coerced. Thus, the constructional view of unaccusativity might be appropriate (Sorace 2002). The Gen-Neg Construction - Analyzed as an existential construction with perspectival structure (Partee / Borschev 2004). Examples (28-29) from Paducheva seem to indicate that Gen-Neg may be used to deny the existence of either concepts or categories. - Example:(5) Zdes’ daze trava ne rosla.Here even grass-NOM not grewEven grass couldn’t grow here (27b, Babby 1980)(6) Zdes’ daze travy ne roslo.Here even grass-GEN not grew-IMPEven grass didn’t grow here- P&B’s (27b) seems to be a stronger version than the GEN-NEG construction in (6). The nominative denies the existence of category, e.g. in the location, while (6) denies only its concept, e.g. does not deny that nothing could grow in the location. Thus, I disagree with Paducheva on the acceptability of (28); it is unusual but OK if used to express that unicorns could not exist.- P&B suggest that quantification may be the appropriate way to state the syntactic condition on Gen-Neg diathesis alternation (e.g. none of objects X might be a predicate for the construction).- P&B: “Most Western Slavists consider Unaccusativity at least a necessarycondition” (1.2). - If (11-12) (Section 1.2) are analyzed as instances of coercion, it is clear that Gen-Neg does not allow activities as arguments. If quantification is the correct description for the