Kill may kill nothing: on intransitivity and causative verbs in English

MANGIALAVORI RASIA, MARÍA EUGENIA; AUSENSI, JOSEP

Ösnabruck

Congreso; Sinn und Bedeutung; 2019

Universitat Ösnabruck

Introduction. We analyze a construction which, to our knowledge, has been largelyoverlooked in the literature, even if it has central implications for the study of argumentstructure/alternations and alternatives in verb formation/derivation in English. This object-lesscausative variant (1) (hereafter, I(ntr)C) features a subject (an inanimate entity) interpreted ascause(r) of a potential change-of-state (COS) whose patient is (semantically and syntactically)unrealized (similar variants reported in Romance/Greek; Alexiadou 2014; M. Rasia 2018).English shows a productive (semi)regularity that deserves to be explored.(1). (a) Smoking kills (b) Alcohol dehydrates (c) Sunlight oxidizes and discolors (d) Shavingcreams irritate (e) Rice constipates (f) Bleach disinfects (g) Normal dryers wrinkle.1. ICs are clear cases of stative (individual-level) predications (ILP). Relevant diagnostics(Maienborn 2004; Rothmayr 2007, i.a.) show that ICs (i) cannot appear in perception reports(2), (ii) present tense does not yield habitual readings (3), (iii) they are infelicitous in what-x-did frames (4) and contexts forcing eventive readings (5) (but see 8) and (iv) cannot be locatedin space. Unlike events, (v) modals (e.g. must) give epistemic (not deontic) readings (7).(2). #I saw smoking kill. (vs. I saw John kill Tom)(3). Smoking (#regularly) kills. (vs. John (regularly) kills animals)(4). #What alcohol did was dehydrate.(5). #What happened was that rice constipated.(6). #Shaving creams irritate in the bathroom.(7). Smoking must kill. (OK Smoking probably has property x | #It is under obligation to kill)2. Like middles, ICs do not license episodic readings?i.e., they do not refer to an actual eventto have occurred; they simply report an (inherent) property of the subject. In this sense, theyinstantiate what Lekakou (2005: 88) calls dispositional generic. In contrast to dispositionalhabituals (8)a, which ?assert the existence of a pattern of regularly recurring events?, the truthof ICs (8)b does not depend on whether there were (burning) events in the past; rather, it is truein virtue of properties inherent to the subject. Conversely, (8)a cannot be true if there have beenno events of John helping people (Krifka et al. 1995). A similar condition draws a key contrastbetween ICs and Levin?s ?intransitives? given by Null/Unspecified Object Alternation (8)c,consonant with data (see (11)-(13) below) supporting a distinct (non-Null/Arb object)structure.(8). a. John helps homeless people.b. Chromic acid burns (that is why it has never been used before).c. This dog bites (#but hasn?t bitten anybody yet). (pro-Arb Object Altern., Levin 1993)3. Middles and ICs share the genericity of an otherwise eventive predicate. Yet, ICs are differentin that the sole DP is not an internal but an external argument. The dispositional property is notattributed to the undergoer of the COS, but to the causer. Consonant with the proposed stativityof ICs, restriction to generic tenses is criterial here: perfect tenses are possible only formaximally different monadic (inchoative) structures (if available: kill), rendering (5)-(9) odd.(9). *Smoking killed/#Alcohol dehydrated/#Shaving creams irritated/#Rice constipated.4. ICs are special in that they denote a stative sort of dispositional causation (Copley 2018). ICsreflect the hallmarks in (10) relating a disposer y (holder of a property), a dispositional state e,a manifestation e′, and a (non-episodic) eventuality description p, with predictable semanticrestrictions (having the relevant property to generate the COS named by 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)5. There seem to be lexical (verb-specific) constraints on the productivity of this construction. Thequestion arises as to why some canonical causative verbs (e.g. break) fail to yield ICs. In thisrespect, we propose that English external-causation change of state verbs are to be split intotwo different classes with regard to the type of causation allowed: causative verbs like those in(1) allow (stative) eventualities that come about as a result of some inherent property (ILPstate) of the cause(r), whereas causative verbs of the sort break or destroy resist such (stative)eventualities (e.g. *Strong winds break/*Earthquakes destroy). Such verbs may resemble coreinternally-caused change-of-state (ICO) verbs (e.g. The flower bloomed/wilted) in that sucheventualities come about as a direct consequence (result) of the internal physical characteristicsof the theme. Since the potential to exert change in ICs is predicated of the causer, and notheme argument figures in the representation, unproductivity in verbs restricted to internal (vs.external) causation falls out (cf. *Fertilizers bloom/okBleach whitens). Such restrictions arelanguage-specific: Romance (e.g. Spanish) freely allows ICs from COS verbs generally,including an important destroy analogue (e.g. La ideología destruye ?Such ideology destroys?.)6. ICs challenge the long-held constraint (Rappaport Hovav & Levin 1998, 2010; Levin &Rappaport Hovav 1995, 2005) that verbs encoding COS disallow constructions where theobject is left unspecified (*John breaks/*killed). In this respect, Levin (2017) argues that if averb encodes a COS predicated of a theme, such theme must be expressed due to the ?patientrealization condition? which arises ?because to know that a state holds requires looking at theentity it holds of? (see Rappaport Hovav 2008 for a similar argument). ICs are counterexamplesto this constraint to the extent that they feature atransitive variants showing that COS verbs canallow object-less constructions with consequent properties, in which the state that holds, holdof a different argument/participant (causer). This upshot is parasitic on 8 below.Analysis. There are important reasons to argue that ICs are not null-object constructions.Notably, they do not allow null-object-oriented depictive predication; adjectival predicates(readily licensed by null/arbitrary implicit arguments, e.g. Il dottore visita [] nudi, ?The doctorvisits [] naked? Rizzi 1996) and null object quantification (e.g. bare molti, Italian) fail (11). InRomance, ne-cliticization and inchoative/passive morphology are also disallowed (12).Further, ICs are productive with unpassivizable verbs (Object-Experiencer statives likesadden) and they do not bind reflexive pronouns (13).(11) *Smoking kills dead/depressed/many. (vs. John cooks healthy/John eats a lot)(12) Fumare (*ne/*si) uccide. (Italian)(13) Bad news sadden (*myself).7. For these empirical, but also for theoretical reasons, ICs seem better analyzed as originalmonadic (intransitive) realizations, where the external-argument-introducing head responsiblefor introducing the causative component is complemented by rhematic information(instantiated by √) specifying the kind of change potentially triggered by the subject (RhemeP,Ramchand 2008, 2013 i.a.) (vP [DP cause/trigger [vINITo, RHEME √]). Regarding the syntax-semanticsinterface and the important body of literature on direct mapping between semantic (event)composition and argument structure realizations, ICs show that non-realization of the internalargument systematically correlates with lack of COS (sub)event instantiation (standardlyattributed to the internal-argument licensing head, Levin & Rappaport 1995, 2005; Hale &Keyser 1993, 2005). This correctly explains the dispositional (non-episodic) causation flavorof ICs, along with the stative (ILP) behavior shown in (1)-(5)?as there is no theme, there isno COS-event encoding component in the semantic nor syntactic composition of the VP.8. ICs challenge important generalizations: they pose a problem for (i) the commonly sharedassumption that the internal argument is a constant/invariable (Hale & Keyser 2002)constituent in the causative alternation; (ii) the prediction that unique arguments in COS verbsare by default interpreted as themes (i.e. Default Linking Rule, Levin & Rappaport Hovav1995, 2005). ICs show that a defective interpretation of unique arguments in causative verbsas cause is possible and natural.Summing up: ICs establish a certain regularity in English, allowing a structure with distinctaspectual and syntactic properties. Lack of eventivity, ILP (stative) predication, restriction togeneric tenses, along with default interpretation of the unique DP of a causative verb as acause(r) rather than as an undergoer (theme) consistently contribute to a distinct, non-episodicpredication basically reflecting dispositional causation (Fara 2001; Copley 2018).