1 To appear in R. Truswell (ed.): Oxford Handbook on Event Structure. Oxford University Press, Oxford. Nominals and Event Structure Friederike Moltman CNRS-IHPST and NYU Final version 1. Events, verbs, and deverbal nominalizations Events have come to play an important role in natural language semantics. While events have been taken to be involved in a wide range of constructions, they most obviously play a role as referents of nominalizations of verbs. It is generally taken for granted that the very same events that verbs describe may act as the referents of NPs with a corresponding deverbal nominalization as head. Thus the very same events described by the sentences in (1a), (2a), and (3a) appear to be what the nominalizations in (1b), (2b) and (3b) stand for: (1) a. John laughed. b. John’s laughter. (2) a. John jumped. b. John’s jump (3) a. John walked. b. John’s walk That the same event is described by the sentence and referred to by the nominalization appears to be supported by the semantic behavior of predicates. In general, it appears, the same predicates can act as adverbials modifying the verb and as predicates predicated of what the nominalization stands for and moreover as adjectival modifiers of the nominalization: 2 (4) a. John laughed intensely. b. John’s laughter was intense. c. John’s intense laughter (5) a. John jumped quickly. b. John’s jump was quick. c. John’s quick jump (6) a. John walked slowly. b. John’s walk was slow. c. John’s slow walk The semantics of nominalizations is closely related to the semantics of adverbial modification, and that needs to be accounted for by any semantic analysis of nominalization based on the semantics of the corresponding verb or sentence. The main question concerning nominalizations and events then is, how does the semantics of deverbal nominalizations relate to the semantics of the verb or the corresponding sentence? That is, how do deverbal nominalizations obtain their referent, given the semantics of the verb, its complements, and its modifiers? I will discuss three approaches to the semantics of event nominalizations (and the related issue of the semantics of adverbials): [1] the Davidsonian account [2] the Kimian (or pleonastic) account [3] the truthmaker account. I will conclude that a combination of the three accounts may be required for the semantics of the full range of event and state nominalizations. In addition, I will present data regarding a distinction between two sorts of event nominalizations for psychological and illocutionary verbs that challenge the received view of the identity of the events described by verbs and by (non-gerundive) deverbal nominalizations, namely a distinction between ‘actions’ and ‘products’ introduced by the Polish philosopher Twardowski (1911). 2. The Davidsonian account of event nominalizations Clearly the most influential semantic account of events described by verbs and their nominalizations is Davidson’s (1967) account, further developed by Higginbotham (1985, 3 2000).1 According to the Davidsonian account, events act as implicit arguments of verbs and adverbials are predicates of the implicit event argument of the verb. A sentence like (7a) then has the logical form in (7b), with existential quantification over events that are to occupy what can be called the implicit Davidsonian argument position: (7) a. John walked slowly. b. e(slowly(e) & walk(e, John)) The main argument Davidson gives for events acting as implicit arguments of verbs and adverbials as predicates of such events is the possibility of adverb-dropping, that is, the validity of an inference from (7a) to (7c): (7) c. John walked. Landman (2000) adds another argument for the Davidsonian view and that is the possibility of permuting adverbial modifiers, as in the valid inference below: (8) John walked slowly with a stick. John walked with a stick slowly. The semantic status of events as implicit arguments of verbs goes along with a particular view about the ontology of events, on which events are primitive entities not to be defined in terms of objects, properties and times (cf. Davidson 1968). Thus, for Davidson, different properties can be used to describe one and the same event and thus won’t be event- constitutive. One and the same event can be described as the rotation of the wheel or as the wheel getting hot, just as one and the same events can be described in both physicalist and psychological terms. Given this view, events could not be conceived as entities strictly dependent on a description, say the content associated with a verb and its arguments. However, conversely, if events are conceived as strictly dependent on objects, properties and times or the content of the verb and its arguments this is still compatible with a Davidsonian view of events as implicit arguments of verbs. In fact, this is Maienborn’s (2007) view of 1 For further developments of the Davidsonian event semantics, see Parsons (1990), Moltmann (1997), Landmann (2000). 4 ’Kimian’ or abstract states, which she considers implicit arguments of stative verbs (Section 6). Given the Davidsonian event semantics, NPs with deverbal nominalizations such as John's walk will simply pick up the implicit event argument of the verb as the referent (cf. Higginbotham 1985, 2000): (9) [John's walk] = e[walk(e, John)] On this account, the implicit event argument of the verb is the very same as the event described by the event nominalization. The formation of the event nominalization thus goes along with a shift in argument structure: the Davidsonian event argument of the verb will become the external argument position of the nominalization, that is, the argument position that will provide a referent for the entire NP of which the nominalization is the head. The Davidsonian account immediately explains why the same expressions that act as adverbials generally appear to be able to act predicates or adjectival modifiers of a nominalization of the verb. In all three cases, on the Davidsonian account, the expression is predicated of the very same events. The Davidsonian account furthermore benefits from a generality of application, allowing a rather straightforward extension to adjectives, though, on my view requiring an enrichment of the ontology so as to include tropes (particularized properties) besides events (Moltmann 2007, 2009, 2013a). Tropes are concrete manifestations of properties in objects, that is, they are properties as particulars, dependent on a particular object as their bearers, rather than properties as universals. It is a general fact that adjectives, to an extent, exhibit the very same alternation of expressions acting as modifiers of the adjective as well as predicates (or adjectival modifiers) of the adjective nominalization. Thus the modifiers of happy in (10a) and pale in (10b) act as predicates of what the nominalizations happiness in (11a) and paleness in (11b) stand for: (10) a. Mary is visibly / profound happy. b. Mary is extremely / frighteningly / shockingly pale. (11) a. Mary’s happiness is visible / profound. b. Mary’s paleness is extreme / frightening / shocking. 5 In the case of adjectives, the implicit arguments should be tropes or particularized properties, rather than events (Moltmann 2009, 2013a).2 That goes along with the view that nominalizations of the sort of Mary’s happiness and Mary’s paleness stand for tropes, not events or states. That is, Mary’s happiness stands for the particular way happiness manifest itself in Mary and Mary’s paleness for the particular manifestation of paleness in Mary. This is the standard view found throughout the literature on tropes, for example in Williams (1953), Campbell (1990), and Lowe (2006). In fact, the view goes back to Aristotle and was common throughout the Aristotelian tradition in medieval and early modern philosophy (way before events gained proper recognition as an ontological category). Why should the implicit argument of an adjective be a trope, rather than an event, or perhaps more specifically a state? That is because a state, as the sort of entity a gerund would stand for, would not have the right properties. For example, ‘Mary’s being happy’ cannot not be profound and ‘Mary’s being pale’ cannot be extreme (Moltmann 2009, 2013a). Formally this then means that (10a) will have the logical form in (12a), with existential quantification over tropes filling in the ‘Davidsonian’ argument position of happy, and (10b) will have the logical form in (12b), where Mary’s happiness is taken to stand for the maximal trope of happiness of Mary.: (12) a. e(happy(e, Mary), & visibly(e)) b. visible(max e[happiness(e, Mary)]) Despite these apparent advantages, the Davidsonian account has also been subject to criticism. A general uneasiness with the account concerns the intuition that events generally should play a role as objects in the semantic structure of sentences only in the presence of nominalizations: as derived objects introduced by nominalizations.3 Positing entities as implicit arguments of all verbs constitutes, on that view, an unnecessary proliferation of entities in the semantic structure of sentences. There are in fact two accounts of event nominalizations that would give justice to that intuition: the Kimian account and the truthmaker account. Let us therefore explore those approaches for the semantics of event nominalizations as well as the related issue of the semantics of adverbials. We will see that 2 For the notion of a trope in contemporary metaphysics see, for example, Williams (1973), Campbell (1990), Lowe (2006), as well as Moltmann (2013a, Chap. 2). 3 Of course setting aside underived event nouns such as fire, war, act, and event. 6 the truthmaker account meets further challenges for the Davidsonian account, namely regarding more complex adverbial constructions. 3. The Kimian account of event nominalizations The Kimian account of event nominalizations is based on the view that events are introduced into the semantic structure of a sentence generally only by the means of nominalizations. This goes along with Kim’s (1976) conception of events, according to which events strictly depend on an individual, a property, and a time, and are introduced as entities by a form of Fregean abstraction. This conception is closely related to the notion of a pleonastic entity of Schiffer (1996, 2003) (which Schiffer (2003) also means to apply to events). Kim’s (1976) original account is a rather simply elaboration of the view, and it has subsequently been adopted and further developed by Bennett (1988), Lombard (1986), and others.4 Kim’s (1976) conception of events consists in the following statement of existence and identity conditions for an event dependent on an object, property, and time, where [d, P, t] is the event dependent on an object d, a property P, and a time t: (13) For individuals d, d’, properties P, P’, and times t, t’, [1] [d, P, t] exists iff P holds of d at t. [2] [d, P, t] = [d', P', t'] iff d = d', P = P', t = t'. The semantics of event nominalizations then appears straightforward, as below, where the denotation of walk, [walk], is taken to be the one-place property of walking: (14) [John’s walk] = e[e = [John, [walk], t]] Events need not strictly be constituted by the entire content of the event description. Kim (1976) in fact draws a distinction between event-characterizing and event-constitutive properties. Event-characterizing properties are merely properties holding of an event constituted on the basis of another, event-constitutive property. If slow is event- 4 The view of events as derived objects has been more popular among philosophers than linguists, with the exception of Chierchia (1984). Linguists generally adhere to the Davidsonian account. 7 characterizing, (15a) has the analysis in (15b); but if slow is event-constitutive, (15a) has the analysis in (15c): (15) a. John’s slow walk b. e[e = [John, [walk], t]] & slow([John, [walk], t])] c. [John, [slowly walk], t]] The availability of event-characterizing properties conveyed by an event description distinguishes descriptions of events from explicit fact descriptions of the sort the fact that S. Whereas adjective modifiers of event nominalizations may just be event-characterizing, all of the content of S in a fact description of the sort the fact that S must be fact-constitutive. This manifests itself in the contrast between (16a), which can be true, and (16b), which can’t: (16) a. John’s slow walk was John’s walk. b. The fact that John walked slowly is the fact that John walked. Events are relatively independent of the description used to refer to them. Facts, by contrast, are entirely reflected in the meaning of explicit fact-referring terms of the sort the fact that S.5 Kim’s account does not explicitly define events in terms of a property, an object, and a time. Rather it gives an implicit definition of events, stating their existence and identity conditions in terms of an object, a property, and a time. In particular, events are not taken to be composed in some way of properties, objects, and times. Kim’s account in fact introduces events by a form of Fregean abstraction (Frege 1884, Dummett 1973, Hale 1987, Wright 1983). Frege’s abstraction principle, given below, just gives identity conditions for objects obtained by the abstraction function g from entities o and o’ that stand in some equivalence relation R: (17) For an equivalence relation R, g(o) = g(o’)  R(o, o’). Thus, Frege introduces directions as entities obtained by abstraction from parallel lines, and 5 For further linguistic support for that view and a semantics for the fact that S, see Moltmann (2013a, Chapt 6). This is the notion of a non-worldy fact, defended by Strawson (1950), as opposed to the notion of a worldly fact, defended by Austin (1979). 8 natural numbers as entities obtained by abstraction from concepts whose extensions stand in a one-to-one correspondence. (17) can naturally be generalized to n-place abstraction functions applying to n objects that stand in respective equivalence relations to each other. Kim’s account of events then introduces events by a three-place abstraction function applying to objects, properties, and times on the basis of the equivalence relation of identity. An object introduced by Fregean abstraction has just those properties specified by the method of introduction. Thus, given (13), events have identity conditions and existence conditions relative to a time, but they won’t have other intrinsic properties. They may, though, act as objects of mental attitudes. This means in particular that events won’t have a part structure, won’t have a spatial location, won’t enter causal relations, won’t act as objects of perception, and won’t have properties of intensity or other measurable properties. This of course is highly counterintuitive. It is certainly part of our notion of an event for an event to have those properties. By contrast, it is part of our notion of a (non-worldly) fact to lack those properties. This difference between events and facts is also linguistically reflected, in the applicability of the relevant predicates:6 (18) a. Mary noticed part of the event. b. ??? Mary noticed part of the fact. (19) a. The meeting was in the room. b. ??? The fact that they met was in the room. (20) a. John’s jump caused the table to break. b. ??? The fact that John jumped caused the table to break. (21) a. John saw Bill’s jump. b. ??? John saw the fact that Bill jumped. (22) a. John’s jump was high. b. ??? The fact that John jumped was high. (23) a. John’s laughter was intense. b. ??? The fact that John laughed was intense. Another interesting difference between facts and events is that facts unlike events do not allow predicates of description: 6 See Vendler (1967), Peterson (1997), and Asher (1993) for similar observations. 9 (24) a. Mary described John laughter / John’s jump. b. ?? John described the fact that John laughed / the fact that John jumped. The reason for this particular difference appears to be that facts are tied to canonical fact- descriptions, but events are not tied to canonical event descriptions. Canonical fact descriptions are of the form the fact that S, descriptions that fully display the nature of a fact and whose content is entirely fact-constitutive. Events do not come with canonical event descriptions, but rather with descriptions that generally do not provide all of the event- constitutive properties or that contain event-characterizing rather than event constitutive parts. Predicates of description in general require that the term used for the object described won’t specify precisely the properties mentioned in the act of describing. The condition is fulfilled in (25a), but not in (25b): (25) a. John described the object: he said it was a book. b. ??? John described the book: he said it was a book. The difference in description (that facts are tied to canonical descriptions, but not events) may explain the applicability of verbs of description. But it cannot be considered the feature distinguishing facts and events, as Bennett (1988) did. The difference in description won’t account of the other differences in properties between events and facts. The distinction between events and facts is clearly an ontological one, not one residing in their description. The main objection to Kim’s account of events has been that it assimilates events to facts. Note that any property, however unspecific or logically complex, can, for Kim, be event- constitutive. Any predicate expressing a non-natural or indeterminate property, any explicitly or implicitly quantified predicate, and any negated or disjunctive predicate can, on Kim’s account, individuate an event (together with an individual and a time). But this is characteristic of facts, not events. Non-specific properties, negation, disjunction, and quantification can be fact-constitutive, but generally not event-constitutive. That is because events as concrete objects must be maximally specific or at least grounded in specific properties. It is the groundedness of events that distinguish events from facts (Moltmann 2007).7 Part of that is also that events unlike facts need to involve particular participants; a 7 The notion of a specific property that I am using is closely related to Armstrong’s (1978) notion of a ‘natural property’ and Lewis (1983) notion of a ‘non-redundant- property. 10 quantifier or disjunction does not suffice for their individuation. This is clearly part of our notions of event and fact. Take the property of making several mistakes (separately) and the property of eating an apple or a pear: (26) a. John made several mistakes. b. John ate an apple or a pear. Then (26a) describes several events, all the events involving a particular mistake, but only one fact, the fact that John made several mistakes. (26b) describe an event that involves either an apple or a pear, but it will describe a fact not involving one particular fruit, but constituted by a disjunction, namely the fact that John ate an apple or a pear Given the Kimian conception of events, one might try to account for the groundedness of events by imposing the restriction that events can be constituted only by fully determinate properties. More adequately, since events generally involve change, events might be conceived as transitions from an object having one determinate property at a time t to the object's having a contrary determinate property at a subsequent time t' ( Lombard 1986). More complex events may then be built from such transitions, either as collections of transitions or as a transitions viewed with a particular, possibly complex property as its gloss, as would be the case with events for which a totality condition is constitutive (Section 3). But if events are introduced by abstraction, even if based on such specific property changes, they will still lack the typical event properties, since the only properties they can have are those stipulated by the strategy of their introduction. Making events be dependent on specific properties will make no difference to the properties events have if events are still introduced by abstraction. Moreover, such a more complex conception of events poses a problem for the Kimian account of event nominalizations. Most verbs in English do not describe the kinds of transitions that could constitute or ground events. In fact, it is hard to find any reasonably simple predicates at all that do. Even such predicates as become soft or turn red, which express a simple property change still involve a non-specific property. Given the range of verbs that can describe events, there are at least five major classes of verbs whose content merely characterizes an event, but would not be fully constitutive of it: [1] verbs involving quantification over kinds of properties: change [2] verbs expressing quantification over spatial positions: move towards, walk [3] verbs expressing quantification over types of actions having a particular of causal effect: disturb, kill