Intransitivity and causation in the division between English vs. Romance

MANGIALAVORI RASIA, MARÍA EUGENIA; AUSENSI, JOSEP

Congreso; Going Romance 2019; 2019

Leiden University

We focus on a construction gone, to our knowledge, largely overlooked in the literature, though it has key implications for the study of argument structure and verb formation/derivation. This object-less causative (I(ntransitive)C) (1) has a subject (inanimate entity) interpreted as cause of a potential change-of-state [COS] with semantically and syntactically unrealized undergoer.(1).a. Smoking kills. b. Bleach disinfects.c. Alcohol dehydrates.d. Rice constipates.e. Shaving creams irritate. f. Sunlight oxidizes and discolors. g. Normal dryers wrinkle.(2).a. Fumar mata. b. La lejía desinfecta. c. El alcohol deshidrata. d. El arroz estriñe. e. Las cremas de afeitar irritan. f. La luz solar oxida y destiñe. g. Las secadoras arrugan.ICs challenge major generalizations: namely, (a) the widely-shared assumption that the internal argument is a constant/invariable constituent in the causative alternation (Hale&Keyser 2002 i.a.); (b) the prediction that unique arguments in COS verbs are by default interpreted as themes (i.e. Default Linking Rule [DLR], Levin&Rappaport 2005). ICs show that a defective cause interpretation of unique arguments in causative verbs is possible/natural. While similar variants are attested in Romance (2) and Greek (Alexiadou 2014; M. Rasia 2018), and even if not fully (freely) productive, English shows systematic patterns deserving to be explored. E.g., we note:1.CROSS-LANGUAGE CONSISTENCIES setting ICs apart (as a sufficiently different construction)A. ICs pattern as stative (individual-level (ILP)) predications. ICs cannot appear in perception reports (3), be located in space (4), license habitual readings (5), are odd in what-x-did frames (6) and contexts forcing eventive reading (7); modals give epistemic (not deontic) readings (8).(3).#I saw smoking kill. (vs. I saw John kill Tom)(see Maienborn 2005, Rothmayr 2007)(4).#Shaving creams irritate in the bathroom. (5).Smoking (#regularly) kills. (vs. John (regularly) kills animals)(6).#What alcohol did was dehydrate.(7).#What happened was that rice constipated. (8).Smoking must kill. (OK Smoking probably has property x | #It is under obligation to kill) B. Consonant with stativity, restriction to generic tenses (shared with middles) is nontrivial (cf. *Smoking killed/#Alcohol dehydrated/#Shaving creams irritated). Still, ICs are different from middles as the sole DP is not an internal but an external argument: the property is not attributed to the undergoer, but to the cause. Yet, like middles, ICs do not entail a deontic reading, but report a property of the subject (cf. Lekakou 2005 dispositional generic). A similar contrast obtains between ICs and Levin?s intransitives given by Null/Unspecified Object Alternation (9)c, in line with data (11)-(13), supporting a distinct (non-Null/Arb object) structure. (9).a. He helps homeless people.b. Chromic acid burns (that is why it has never been used yet).c. This dog bites (#but it hasn?t bitten anybody yet). (pro-Arb Object Altern., Levin 1993) C. The distinct meaning of ICs reflects the definition of dispositional causation qua a predicate that relates a disposer y (holder of a property), a dispositional state e, a manifestation e′, and a (non-episodic) eventuality description p. This captures consequent restrictions noted in ICs (e.h. subject must have the relevant property to produce COS associated with the verb, Fara 2001).(10).Dispositional causation: (a) y is the holder of e, (b) e is a state that directly causes e′ ceteris paribus, (c) e′ instantiates p (d) y is disposed toward p. (Copley 2018: 13) CROSS-LANGUAGE ASYMMETRIES in productivity A. Specific constraints: In Romance, a set of COS verbs (like break) systematically fail to yield ICs. Apparently, the externally-caused COS verbs factor into two classes wrt structures allowed: while COS verbs like (1) allow (stative) eventualities brought about by an inherent property of the causer, verbs like break/destroy resist such eventualities (*These roots break/*Earthquakes destroy). Non-productive IC verbs seem to line up with core internally-caused COS (ICO) verbs (Levin&Rappaport 1995), insofar as the eventuality denoted comes about as a consequence of an internal property of the internal argument (COS undergoer). As in ICs potential to exert COS lies in the causer, and no theme figures in the representation, this asymmetry falls out (cf. *Fertilizers bloom/okBleach whitens). B. However, other class restrictions seem language-specific. While Romance freely allows ICs from classes standardly related to the causative alternation like psych verbs, English (IC unavailable) have to appeal to a stative-attributive ing-predicate (vs. progressive -ing, yielding the anticausative variant instead (cf. La leche engorda ?Milk is fattening (≠is getting fat)?). (11).a. La playa cansa.(lit. *The beach tires)b. Tarantino aburre. (lit. *Tarantino bores)?The beach is tiresome/makes you tired.??Tarantino is boring/makes you bored.?PROPOSAL. ICs are not Null-Object [NO] constructions: ICs do not allow NO-oriented depictives, secondary/adjectival predicates (readily licensed by null/arbitrary implicit arguments, e.g. Il dottore visita [] nudi, ?The doctor visits [] naked?, Rizzi 1996), nor NO quantification (e.g. bare molti, Italian) (12). Ne-cliticization and inchoative/passive morphology (Romance) are also disallowed (13). Further, ICs are possible with unpassivizable verbs (Object-Experiencer verbs like sadden) and fail to bind reflexive pronouns (vs. the anticausative (se-cl) form *(se) atttrista con se stesso). Consistent data comes from Spanish (M. Rasia 2018). ICs are better analyzed as monadic (atransitive) realizations where the external-argument-introducing head responsible for the causative component is merely complemented, not by a theme (undergoer), but by mere rhematic information (RhemeP, Ramchand 2008, 2013 i.a.), √ specifying the COS potentially triggered by the subject (vP [DPCAUSE/TRIGGER [vINITº, RHEME √]). WRT the syntax-semantics interface and direct mapping between semantic (event) composition and argument structure realizations, ICs show that non-realization of the internal argument systematically correlates with lack of COS (sub)event instantiation (attributed to the internal-argument licensing head, Levin & Rappaport 1995/2005; Hale & Keyser1993/2005 i.a.). This captures the non-eventive denotation in ICs (recall (10)), along with noted stative (ILP) behavior (3)-(9): as there is no theme, there is no COS-event-encoding component in the semantic/syntactic makeup of the VP.(12).Smoking kills *depressed/*some. (vs. John cooks/eats healthy/some)(13).a. Fumare (*ne/*si) uccide. (Italian)b. Fumar (*ne/*se) mata. (Catalan)Now, given (2), we suggest that, across languages, (Romance/English), ICs are formed from verbs denoting COS-caused state, rather than (manner of) action (Rappaport&Levin2010 a.m.o.). Apparently, if the verb only encodes a manner of action, and there is no possible state-like interpretation the subject is able to trigger due to some (ILP) property (recall (10)), ICs fail.(14).a.*Esta sustancia asesina.(vs. okEsta sustancia mata)(Spanish)b.*Aquesta substancia assassina. (vs.okAquesta substància mata)(Catalan)?lit. *This substance murders.? (vs. This substance kills)(15).Esto mata/envenena/sofoca ?this kills/poisons/suffocates? | *extingue/*apaga ?smothers?In turn, psych verbs are noted to class as stative in Spanish, hence (11). The analysis also accounts for productive asymmetries within them (Subject-Experiencer type: IC unavailable).To account for verbs (not) allowing ICs, we suggest that roots select for the type of causing eventuality (Beavers&Zubair 2013). Following Beavers&Koontz-Garboden?s (2019), we assume a set of eventualities, factored into events (UE) and states (US). Apparently, roots like √KILL are unspecified for causing eventuality type (i.e. UEJohn killed the man/USJohn?s stupidity killed us/(John?s) Stupidity kills) thus allowing IC; roots like √BREAK select for event (John broke the vase), and hence, no IC (*Strong winds break). In turn, √MURDER?s agentivity disallows US and IC (#John?s stupidity murdered us/#These substances murder).IF CORRECT: ICs challenge the long-held constraint (Rappaport&Levin 2010 a.m.o.) that COS verbs disallow unrealized theme (*John breaks/killed), as they feature atransitive foms showing that COS verbs can allow object-less constructions with consequent properties, wherein the state that holds, holds of a different argument (= causer), an upshot parasitic on the observation on DLR. ICs establish a regularity in English, with important connections in Romance pointing to semantic/syntactic crosslanguage consistencies, along with evidence on semantic properties key to argument realization, raising questions on lexical encoding of relevant features.