2.1 LF-Interpretable Categorial Features

Adhering to Distributed Morphology (Halle & Marantz 1993; Marantz 1997; 2000; 2006; Harley & Noyer 1999; Embick 2000; among many others) and drawing on Embick and Marantz’s (2008) Categorization Assumption, that roots can appear in syntactic structures only as the complements of categorizers where they are assigned a category and become related to concepts, Panagiotidis (2009) points out that “‘noun’, ‘verb’ and ‘adjective’ are not categorial labels specified on lexical items in a pre-syntactic lexicon,” rather “roots are inserted bare in syntax and the assignment of roots to categories is a syntactic process” where “categorizers – a nominalizer (n), a verbalizer (v) and an adjectivizer (a) – make them nouns, verbs or adjectives.”^1,2 Since nPs and vPs are phases, once they are constructed, they must receive phonological and semantic interpretations at the interfaces to the Sensory-Motor (SM) and Conceptual-Intentional (C-I) systems (e.g., Chomsky 2001; 2008). Panagiotidis (2009; 2011; 2013; 2015: Ch. 4) proposes that the (first) phase heads n and v are distinct categorizers because they bear different LF-interpretable categorial features, namely [N] and [V].³ He employs Baker’s (2003) idea that category distinctions must correspond to perspectives on (concepts about) the world and argues that these features impose different perspectives, ‘contexts’, on concepts in their complements (i.e., in their phase domains). The interpretive perspective, in Panagiotidis’ analysis, is necessary for the semantically underspecified and impoverished root to be interpreted at the C-I system. The categorial features [N] and [V] on the categorizers n and v persuasively “close off material associated with the root exactly by providing it with a fundamental perspective for the conceptual systems to view it in” (Panagiotidis 2009). [N] and [V], providing sortal and temporal perspectives, respectively, make the categorizers n and v semantically distinct, as described in (1).

Baker (2003: 95) claims that nouns are semantically differentiated from verbs by “criteria of identity, whereby they can serve as standards of sameness.” A criterion of identity marks nouns as kind-denoting and this lexical property gives rise to the idea that nouns encode referential power manifested as a referential index in syntax (Baker 2003: Ch. 3; cf. Wunderlich 1996). Identity is also considered to be a criterion of sortality by Prasada (2008). He notes that sortality incorporates three criteria: application, identity and individuation. The criterion of application “means that the representation is understood to apply to things of a certain kind, but not others. Thus, the sortal DOG allows us to think about dogs, but not tables, trees, wood or any other kind of thing,” the criterion of identity “provides the basis for thoughts like dogs, [which] by virtue of being dogs, remain dogs throughout their existence,” and the criterion of individuation states that “two instances of a kind are distinct because they are the kinds of things they are” (Prasada 2008: 6–8).

To explore what nominality means, Panagiotidis (2009; 2011: 371–372; 2013; 2015: 85–86) diverges from Baker’s (2003) stance of regarding [N] as encoding kinds and employs Prasada’s (2008) criteria of “application” and “identity” to define sortality. Following Acquaviva’s (2009: 1) proposal that nouns are sorts, or kind-level entities, “defining the categories of entities in a speaker’s conceptualization of the world,” Panagiotidis then emphasizes that the sortal interpretive perspective of nominals grows out of kinds.⁴

To argue for the temporal perspective that [V] imposes on the root, following Uriagereka (1999) and referring to Ramchand’s (2008) idea that “VP is the heart of the dynamic predicate, since it represents change through time,” Panagiotidis (2011: 372–373; 2013; 2015: 86–88) notes that verbal constituents are inherently (sub-)eventive by virtue of the temporal perspective contributed by the categorial feature [V] to event structures, and that Voice, Aspect and Tense exclusively combine with verbs because of the ‘extending into time’ perspective that [V] imposes on the root. In §2.2, I provide several pieces of evidence that nominal constituents possess temporal and aspectual interpretations. Also, I argue that including Tense and Aspect as functional projections in the nominal spine not only establishes links for the temporal and aspectual interpretations of noun phrases at the syntax-semantics interface but also explains the fixed order of epistemic and non-epistemic modal adjectives with respect to temporal adjectives (see also Ilkhanipour 2015; 2016). If my arguments are on the right track, then nominal constituents, like verbal constituents, are subject to Voice, Aspect and Tense systems. This ensures that a similar, if not precisely identical, extending-into-time perspective is available in the nominal domain and hence, temporality cannot introduce a plausible distinction between n and v.

2.2 Temporality in the Nominal Domain

Let us begin with Musan’s (1995) observation that noun phrases can be modified by various kinds of temporal expressions, as shown in (2), and that these temporal modifiers can modify nouns denoting life-time or temporary properties, as in (2a) and (2b), respectively.

Noun phrases can also be modified by aspectual adjectives, as shown in (3).

The temporal and aspectual modification of noun phrases, as in (2) and (3), leaves space for the hypothesis that nominals, like verbals, refer to situations that hold at certain times (see also Musan 1999).

Second, the unacceptability of the sentences in (4) can be best justified if we credit the arguments my future job and the present president with some temporal dimension modified by the temporal adjectives future and present (see also Enç 1987).

A relevant discussion of the temporal relation between predicates and their arguments can be found in Musan (1995; 1997; 1999) where existence-implying predicates, including individual-level predicates (e.g., to be from America) and stage-level predicates (e.g., to be happy), are assumed to impose a presuppositional condition on their arguments’ being in existence or alive;⁵ the utterance John is happy, for instance, is acceptable only if John’s lifetime (i.e., the “time” of John’s existence in the present world) maps onto, or at least covers, the time of his being happy. This issue has been pointed out by Musan also in terms of noun phrases receiving temporally dependent or independent readings (Note that presuming a temporal dimension for noun phrases is crucial for Musan’s discussion):

“A noun phrase occurrence is temporally dependent if and only if its situation time has to intersect with the situation time of the main predicate of its clause.

A noun phrase occurrence is temporally independent if and only if its situation time does not have to intersect with the situation time of the main predicate of its clause.” (Musan 1999: 622)

Third, the acceptability of (5b) in contrast to the unacceptability of (5c) with regard to the situation uttered in (5a) indicates that the noun phrase the probable winner is subject to temporality: the winner is the ‘probable’ winner only before the game finishes. As soon as the game is over, the probable winner is not anymore the ‘probable’ winner; she/he is the winner or the loser. Put it precisely, the epistemic adjective probable is only compatible with noun phrases whose existence time is not in the past of the reference time.⁶

So, the modification of noun phrases with temporal expressions, the temporal relation between predicates and their arguments and the incompatibility of epistemic adjectives with noun phrases located in the past of the reference time provide evidence for considering some extending-into-time perspective for nominals.⁷ In the remainder of this subsection, I suggest that Nominal Tense (T_N) and Nominal Aspect (Asp_N) are two functional heads in the extended projection of the nominal spine. I do not attempt to argue that T_N and Asp_N have the same featural specification or semantic properties as their clausal counterparts (Recall that D(eterminer) and its clausal counterpart C(omplementizer) do not have the same syntactic and semantic properties.); however, on the basis of evidence from the location-in-specifier approach to adjectival syntax, the morphological expression of time within the nominal domain, and the fixed order of modal adjectives with respect to temporal adjectives, I indicate that Tense and Aspect do combine with nominals. I argue this even though the combination of temporal markers with nominal constituents results in idiosyncratic, non-compositional readings (see also Alexiadou 2001: 59–66 and Alexiadou 2005 for interesting pieces of evidence for and against Nominal Tense).⁸

In accordance with Cinque’s (1994; 2010) and Scott’s (2002) proposal that adjectives are base-generated in the specifiers (Specs) of distinct functional projections to which they are associated, I consider the location of temporal adjectives (e.g., present and former in (2d)) to be Spec,T_NP and the location of aspectual adjectives (e.g., gradual in (3)) to be Spec,Asp_NP. Nominal Tense and Nominal Aspect are then two functional projections that not only provide the necessary space for temporal and aspectual adjectives but also contribute to the external and internal temporal interpretation of nouns (See also Alexiadou 2001: 51–56 for arguments on AspP in nominals).

Further, according to Lecarme (1996; 2004; 2008), Sadler and Nordlinger (2001) and Nordlinger and Sadler (2003; 2004a; 2004b), among others, nominals are inflected for tense, aspect and mood in a number of languages. Examples (6) and (7) are two instances from Halkomelem and Guaraní. In these languages, the same set of affixes mark tense on nominals and verbal predicates. Nominal past tense, for example, encoding meanings such as ‘former, ex-, late (dead)’ temporally locates the nominal. When used with a possessed inanimate noun, as in (7), the temporal marker indicates that the possession relation was in the past, or that the possessed item has been destroyed (Burton 1997: 67–68).⁹

The morphological expression of time within the nominal domain provides evidence for the idea that Tense does not exclusively combine with verbs (contra Panagiotidis 2011: 373; 2013; 2015: 88).

Moreover, the study of temporality in the nominal domain provides an exciting opportunity to advance our knowledge of the possible orders of adjectives. In her research, Ilkhanipour (2015; 2016) assumes that, through the appearance of, for instance, temporal adjectives, T_N represents the time of existence or occurrence of the modified noun. She argues that much like the position of epistemic and root modals with regard to Tense in the verbal spine, epistemic and root modal adjectives (e.g., probable and reliable, respectively) occupy different positions in the nominal spine with regard to T_N. This hierarchy is illustrated in (8).

Ilkhanipour (2015; 2016) discusses that the Persian adjective qæbli ‘previous’ is ambiguous: it may be interpreted as a temporal or ordinal adjective, the latter associated with the highest functional projection in the nominal domain (see also Scott 2002). She points out that when qæbli ‘previous’ occurs with an epistemic adjective, such as Ɂehtemali ‘probable’, it can be interpreted only in the specifier of the ordinal projection, as in (9a), and not as a temporal adjective, as in (9b). Importantly, the unacceptability of (9b) follows from the fact that the past temporal adjective qæbli ‘previous’ yields the certainty of being a winner and this leaves no way for the probability-denoting adjective Ɂehtemali ‘probable’ to be interpreted.

The idea so far is that Tense, Aspect and Modality are not exclusively verbal projections and that temporality is not the property that distinguishes v from n. Before setting out to discuss the causal nature of v, I wish to address two problematic aspects of Acquaviva’s (2008; 2009) non-temporal account of nouns. Panagiotidis’ depriving n from temporality is rooted in Acquaviva’s proposal that reference to temporal constitution is inessential for nouns (including continuants and occurents). There are two main points to be considered here:

First, Acquaviva (2008; 2009) takes collection, growth and destruction as nominalizations with deverbal interpretation (and not nouns). He claims that while these nominals are morphosyntactically nouns, their lexical semantics is entirely determined by the event- and argument-structure of the verbs to collect, to grow and to destroy. Contra this point of view and following Chomsky (1970; 1995; 2020) and Marantz (1997), I believe that nominals like collection, growth and destruction (as opposed to -ing gerunds) are never verbs at any stage in the derivation (see also Grimshaw 1990; Siloni 1997; Harley & Noyer 2000; Alexiadou 2001; Borer 2003). The difference in the argument structure of growing and growth in (10a–b) can be explained if we consider growing a nominalization, a mixed projection involving both v and n, and growth a noun with √GROW directly merged with n (see §3.1 for an explanation of the grammaticality pattern in (10); and see Schoorlemmer 2001; 2002 for details on mixed projections).

Second, based on examples (11a–b) from Simons (1987), Acquaviva (2009: 2–3) writes that:

“nouns allow speakers to describe events unfolding in time as if they were complete entities that acquire and lose properties over time. We know that argument or wedding describe events, but we can speak of them as if they could undergo changes in time, as in [(11)]:”

The problem here is that Acquaviva takes into account only the ‘existence/change in time’ of noun phrases and neglects their ‘time of existence/change’. While the former allows speakers to see the referent of the noun phrase as a complete entity existing, moving or changing on the timeline represented by the clausal tense (as in (11)), the latter provides a timeline where the entity/situation comes into existence/starts, develops and vanishes. Although the time of existence/change is usually implied and is recovered in the context of use, it may be explicitly stated via temporal and aspectual markers (affixes, adjectives, etc.). The time clash between the time of the occurrence of the subject argument and the time provided by the clausal Tense in (12) signifies the temporal dimension of noun phrases (see also examples (4a–b) above).

These points show, as well put by Acquaviva (2009: 3) himself, the conclusion that “verbal reference has a temporal dimension built in […] [while] nominal reference does not, and can do without such a dimension even when referring to occurrents [l]ike all negative conclusions […] is wide-ranging but also weak. It means that we cannot just apply to nouns the semantic decomposition that works for verbs. It does not mean that event structure never plays any role in the lexical semantics of nouns, but that it is not its fundamental ingredient.” (Italics mine)

In section 3, I will argue that supralexical causation is the necessary interpretive perspective that v imposes on the root.

3.1 Causation: Not in n But in v

To argue that v, unlike n, introduces supralexical causation, particularly when the root does not introduce an external and/or internal cause, I makes use of Marantz’s (1997) categorization of roots, reiterated in (13).

As shown in (14), verbs of the “destroy” class are generally only transitive while nouns of this class can be transitive or intransitive. Verbs of the “grow” class are either transitive or intransitive while the relevant nouns are only transitive, as shown in (15). And verbs of the “break” class are either transitive or intransitive while nouns of the same class do not take any arguments, as shown in (16) (Marantz 1997).

A possible explanation of the (un)grammaticality pattern in (14)–(16) is proposed by Alexiadou (2001: 154–155):

[…] whenever a root denotes an externally caused event such as √DESTROY, the specifier of D can be interpreted as an external causer (agent). On the other hand, roots like √GROW which denote an activity that occurs spontaneously, thus being internally caused, do not permit the external causation interpretation of the specifier of D […]. This accounts for the unavailability of transitive nominalizations with verbs like grow entering the causative alternation. Crucially, √GROW can receive an agentive interpretation only if it is inserted under a syntactic causative head.

What I wish to highlight in the quote above is the role of “a syntactic causative head.” If we admit that the verbalizer v, and not the nominalizer n, assigns the interpretive perspective of (supralexical) causation to the category-neutral root, the (un)grammaticality pattern in (14)–(16) will be justified. This account allows the necessary causal relation to be established for the argument(s) involved in causation externally (as in (15a)), internally (as in (16b)) or both externally and internally (as in (16a)) in the verbal domain. On the other hand, the (un)grammaticality of the nominal use of these roots (i.e., where √P is the complement of the nominalizer n) reveals that n is not capable of providing the root with the supralexical causal meaning; only the arguments involved in the inherent (lexical) causal relation set by the root are semantically licensed; for destruction, these arguments are the enemy and the city, and for growth, it is tomatoes. There is not any internal causal relation in break and hence no argument is licensed in its nominal form.

Another candidate for the determination of causal meaning is Voice (see Alexiadou et al. 2015, for instance). Following a substantial literature, in §3.2 I concede the idea that Voice and v are two separate projections. Based on the theoretical assumptions of Phase Theory and the presence of Voice in the nominal domain, I argue that Voice does not contribute to the causal meaning of the construction.

3.2 Causation: Not in Voice But in v

Voice and v are two distinct projections. Based on the interaction of applicative and causative morphology, the existence of two kinds of causatives, and the interaction of passive and verbalizing morphology in Hiaki, Harley (2013a) argues for a tripartite internal structure of the verb phrases, made up of VoiceP, vP and a lexical projection (√P or VP). She cogently distinguishes the external-argument introducing projection VoiceP (Kratzer 1996), which makes no lexical-semantic contribution, from vP whose head hosts causative and verbalizing morphology (Marantz 1997) (see also Pylkkänen 2002; Collins 2005; Alexiadou et al. 2006; 2015; Alexiadou & Doron 2012; Harley 2013b; Merchant 2013; Ilkhanipour & Sugawara 2016; among many others).

Concerning our discussion here, one may claim that the head responsible for the introduction of supralexical causation is Voice. Two arguments can be put forth against this: first, according to Chomsky’s (2001; 2008; 2013) Phase Theory, once a phase (CP or vP) is constructed, its interior (i.e., its complement) is transferred to the interfaces. After interpretation, this complement material is treated as stored and becomes unavailable for further syntactic computation – it becomes “impenetrable.” However, the edge of the phase, that is, the head with anything merged to it and the specifier(s) can be modified in the next higher phase. Let us consider the vP phase. As soon as vP is fully formed, the complement of v is transferred to the C-I system and becomes unavailable to any higher head, namely Voice; therefore, Voice cannot impose a causal perspective onto the complement of v. This is illustrated in (17) (cf. Travis 2013; for the Voice-v order see Cinque 2013 and Harley 2013a).

A counter-argument may arise here: upon the √-to-v movement (i.e., internal merge), as is the case of incorporating the root into v (see, e.g., Harley 2007), the root escapes the interior and becomes available to Voice, so that Voice can determine the causative meaning of the sentence. This is not tenable: Voice obviously does not play any role in determining the internal cause, as is the case of The glass broke (16b). Moreover, even if the incorporation account is valid, a copy of the root should stay in the interior of the phase and be transferred to the C-I system for the full interpretation of the √P (or VP) event. This interior copy of the root will not be available to Voice and cannot be dedicated a causal perspective from outside of the phase.

The second argument against the causal interpretation of Voice comes from the presence of Voice in the nominal domain. Based on the presence of manner adverbs (that bear a tight relation to VoiceP) and the morphological reflex of the inclusion of Voice in process nominals, Alexiadou (2001: 47–50) concludes that “process nominals contain a domain that shows clear verbal properties.” While I find Alexiadou’s evidence for VoiceP in nominals quite convincing, I do not regard VoiceP as an exclusively “verbal” projection. In fact, Alexiadou does not distinguish Voice from v in her analysis of functional projections in the nominal domain (see Alexiadou 2001: 56–57, 112; Alexiadou 2019; Yatsushiro & Alexiadou 2020) and this is where my account begins to diverge from hers; I believe that Voice and v are two separate projections and that Voice can be present independent of v. If my account is on the right track, it also deviates from the way of the inclusion of Voice in nominals as presented in Borer (2003), Alexiadou et al. (2007), Sichel (2007), Embick (2010), Anagnostopoulou & Samioti (2013) and Bruening (2013). In their analyses, the presence of Voice is dependent on the presence of v, as shown in (18) below. (For earlier studies on the properties of passive nominals, see Cinque 1980; Anderson 1983; Higginbotham 1983; Grimshaw 1990; Giorgi & Longobardi 1991; Picallo 1991; Longobardi 2001; among others).

In lieu of (18), I assume that in the nominal domain, VoiceP is a projection above nP. Consequently, the two different argument structures in the enemy’s destruction of the city (14c) and the city’s destruction by the enemy (14d) mirror those of their sentential counterparts. The Active Voice head (Voice_Act) provides the necessary space (i.e., its Spec) for the external argument, while the Non-Active Voice (Voice_NAct) leaves out the external argument, so that it is introduced in a by-phrase. (19) illustrates the Voice projection in the nominal domain.¹⁰ In (19a), the external argument the enemy is located in Spec,Voice_Act, whereas in (19b), the Non-Active Voice does not project a Spec and hence the external argument is added to the construction as the complement of the preposition by.

Note that the external argument is semantically licensed by the root √DESTROY and the Active Voice only provides a syntactic position for its merger. The ungrammaticality of *John’s growth of tomatoes (15c) and *John’s break of the glass (16c) is not due to the lack of syntactic position but to the deficient semantic licensing; neither the nominalizer n nor the roots √GROW and √BREAK semantically license an external cause.

To sum up, in this section I illustrated that v (or more precisely, [V] on v) encodes causal meaning and that supralexical causation contributes to the semantic licensing of the external and internal arguments in the verbal domain. Different types or flavors of v (i.e., v_CAUSE, v_DO, v_BECOME, v_BE) as proposed by Folli & Harley (2005; 2007) and Harley (2008; 2009), however, may appear to cast doubt on the account presented here: one may assume that only v_CAUSE, or at most v_CAUSE and v_BECOME (and not v_DO and v_BE), encode causation. I will turn back to this after elaborating on the notion of causation, what it denotes and how it is represented in v in section 4.

4.1 What Is Causation?

There are two broad theoretical views of causation:

1. “Make a difference” theories of causation: The theories that are based on this view assume that “causation can be judged without appealing to the particular mechanism relating cause to effect” and claim that “A causes B if A’s occurrence makes a difference to B’s occurrence in one way or another” (Walsh & Sloman 2011: 23, 21). This view underlies counterfactual (e.g., Lewis 1973; 1986; 2000) as well as covariation (e.g., Cheng 1997) and manipulability (e.g., Pearl 2000; Woodward 2004) theories of causation (see Walsh & Sloman 2011: 23–24 for an overview of these theories, and Copley & Wolff 2014: 13–23, and Wolff 2014: §5.3 for problematic phenomena for dependency theories of causation).

2. “Generative” theories of causation: According to this view, “A causes B if some quantity or symbol gets passed in some way from A to B” (Walsh & Sloman 2011: 21). With reference to the actual causal process linking the cause to the effect (or the causer to the causee), the theories that are based on this view propose that “causal relations are not just stipulated but rather represent mechanisms in the world that take input (causes, enablers, disablers, preventers) and generate outputs (effects)” (Sloman & Lagnado 2015: 227).

Briefly, while “make a difference” theories, based on the idea that a cause is something that makes a difference to the effect, disregard how the effect is brought about, “generative” theories of causation (i.e., causal process theories) rely on the idea that causation involves a process of transmission or exchange of some conserved quantity, such as energy, momentum, force or information, along a causal pathway from the causer to the causee (see Shultz 1982; Salmon 1984; 1994; Dowe 2000).

Wolff (2007) notes that “make a difference” theories, dependency models in his terms, cannot distinguish CAUSE from ENABLE, overdeterminate the cause, sometimes identify non-causal factors as causal, and as such, are better viewed of as tests for causation, not accounts of its representation in the mind. “Generative” theories of causation, studied as physicalist models, on the other hand, reduce causal relationships to dynamics (i.e., the invisible properties of an event, namely the underlying energies and forces that give rise to the motions) rather than kinematics (i.e., the visible properties of an event: the shapes, sizes, positions, points of contact, etc.). The dynamics of an event enter directly into the conception of causation as they are central to causation in the actual world, and thus physicalist models put forth a better understanding of causal relations. Physicalist models, however, much like dependency models, do not distinguish causation from other kinds of relationships (Wolff 2007). To solve this problem, Wolff and his colleagues propose the force theory of causation (introduced as the dynamics model of causation), a physicalist model based on Talmy’s (1988) theory of force dynamics (Wolff & Zettergren 2002; Wolff 2003; 2007; 2014; Wolff & Song 2003; Wolff et al. 2005; Wolff et al. 2010).

According to the force theory, causation is associated with configurations of forces and an endstate vector.¹¹ Causal interactions involve an affector and a patient, and are described at two levels of analysis. At the category level, cause-related concepts are specified in terms of three dimensions: the tendency of the patient for the endstate, the presence of concordance (or opposition) between the affector and the patient, and progress toward the endstate. At the computational level, these three dimensions are redescribed in terms of patterns of forces, or vectors. Four types of force vectors are relevant: A represents the direction of the force exerted on the patient by the affector; P represents the direction of the force generated by the patient itself or its resistance to change; O represents the direction of the summation of the remaining other forces acting on the patient; and R represents the direction of the resultant force acting on the patient. In addition to these four forces, vector E specifies the mental representation of the patient’s location with respect to an endstate. Figure 1 is a schematization.

Figure 1

Forces associated with the affector, A, patient, P, and other forces, O, combine to produce a resultant force, R, that is directed toward the endstate, as specified by the position vector, E. (Wolff 2007: Figure 2).

The relationship between the category level and the computational level is as follows. The tendency of the patient for the endstate is the result of the collinearity of P and E, concordance of the affector and the patient is the result of the collinearity of A and P, and progress toward the endstate is the result of the collinearity of R and E (Wolff 2007).

What I wish to subjoin to Wolff and his colleagues’ force theory here is the idea of profiling from Langacker’s Cognitive Grammar. This will be used to differentiate the distinct flavors of v in §4.2. Langacker (1991; 2009) proposes that speakers, employing the conventional conceptualizations of language and cognitive processes (construal in his terms), can have different active characterizations of events including action chains where energy is transmitted from one entity to another. One type of the speaker’s construal is profiling, which allows her/him to choose to give special prominence to (or single out as a kind of focus of attention) certain segments (i.e., substructures, including entities and/or interactions) of an action chain. Profiling is well applicable to the causal apparatus in Figure 1; different patterns of forces or composites of diverse forces and the endstate vector can be profiled, that is, different substructures of the causal apparatus can be chosen by the speaker to receive enhanced prominence. For instance, the speaker can construe the causal scene, profiling P, but not A, and possibly R and O, as is the case for Tomatoes grew (15b) and The glass broke (16b). I will get back to the details of this issue in the next subsection.

How the force theory of causation decomposes CAUSE to a finer level of detail, as explained above, and how speakers construe a causal event or scene in alternative ways give us the opportunity to plausibly account for the causal nature of [V] and to differentiate the four flavors of v. In §4.2 I will deal with these issues.

4.2 A Force-Theoretic Approach to v

Not all theories of causation can address the causal distinction between n and v. In order to accommodate this difference, a theory of causation must be capable of elucidating the ungrammaticality of *John’s growth of tomatoes (15c) and *John’s break of the glass (16c) in contrast to the grammaticality of John grew tomatoes (15a) and John broke the glass (16a). The causer John and the effect grow/break are the only elements necessary for “make a difference” theories to be satisfied. Both of these elements are present in (15c) and (16c), as well as in (15a) and (16a). Thus, causation as defined by these theories cannot distinguish v from n.

If our definition of causation involves a causal mechanism, as “generative” theories do, however, the causal distinction between n and v will be expounded; v encodes the necessary supralexical causal pathway from cause to effect, hence the grammaticality of (15a) and (16a), while n is not able to provide any causal relation supralexically, hence the ungrammaticality of (15c) and (16c). Thus, so far, we can conclude that the categorial feature [V] on v establishes the necessary causal process/mechanism/pathway from cause to effect. However, two questions still remain to be discussed: first, what is the nature of the causal mechanism as encoded in [V]? And second, how can the different flavors of v be distinguished?

To answer these questions, I resort to the force theory of causation as alluded to above. There are two main reasons for pursuing this theory: first, the fine decomposition of the causal mechanism proposed in the force theory gives us the ability to discern how linguistic elements might correspond to units of cognition (here forces and the position vector, as illustrated in Figure 1). Second, the force theory of causation has been employed in the study of language and argued to be more suited to some linguistic data; Copley & Wolff (2014), for instance, argue that a force-theoretic framework offers us a better understanding of three linguistic phenomena: (i) defeasible causation in non-culminating accomplishments, (ii) agency through dispositions, and (iii) representations of causal chains. Copley & Harley (2015) also use the concept of force (i.e., input of energy) to account for the problem of non-culmination. They further apply the force theory to the composition of Vendlerian eventuality types (Vendler 1957) and discuss that stative predicates are predicates of situations, while dynamic predicates are predicates of forces. In the remainder of this section, I submit, along the similar lines as Copley & Harley, that the force theory of causation yields new insight into the causal nature of v and tells us how distinct flavors of v, namely v_CAUSE, v_DO, v_BECOME and v_BE, can be differentiated. The proposal here differs from Copley & Harley’s (2015) study in two respects: first, while Copley & Harley make use of the force theory to decompose basic Vendlerian eventuality types and replace event and world arguments with force and situation arguments, in this paper I use the force theory to differentiate v from n and attempt to apply this theory to distinct flavors of v by employing the idea of profiling. Second, Copley & Harley, employing the tools of formal semantics, come to a type-theoretic distinction between forces and situations at the syntax-semantics interface. Here, I do not use any linguistic tools to describe the nature of [V] on v or to distinguish different flavors of v; understanding dynamic and stative situations as introduced in the force theory of causation in terms of a causal apparatus (Figure 1) and the speaker’s profiling of different substructures of this apparatus are parts of the speaker’s extra-linguistic cognitive capacity that, I attempt to show, provides the bases for differentiating verbal and nominal constituents.

According to Folli & Harley (2005; 2007) and Harley (2009), all flavors of v serve the verbalizing function although they express distinct structural properties and different meanings to do with the initiation, or lack thereof, of the verbal event. The activity-denoting v_DO selects for an agent and can take a nominal complement, while v_CAUSE with a causer external argument requires a small clause complement. v_BECOME and the stative v_BE, on the other hand, select for a small clause complement and do not take any external argument. Harley (2009) also points out that v_CAUSE and v_BECOME, unlike v_DO and v_BE, involve some change of state.

In order to uphold the analysis in which v encodes a causal mechanism and simultaneously make this analysis responsive to different flavors of v, it suffices to highlight the prominence that the speaker chooses to give to specific vectors or particular patterns of forces (in the spatial schema provided by v) by employing distinct flavors of v. While the categorial feature [V] on v introduces the apparatus illustrated in Figure 1 above (i.e., a conventional conceptualization in language), different flavors of v reflect the speaker’s alternate ways of profiling the (causal) scene.

v_CAUSE reflects the speaker’s accentuation of the process of the exchange of energy. At the computational level of analysis, ||A|| and ||P|| (double vertical lines denote magnitude) are greater than zero, but ||R|| and ||O|| can equal zero (see also Wolff 2007). Take, for instance, John grew tomatoes (15a) and John broke the glass (16a). In these sentences, v_CAUSE provides the positive ||A|| which is not available in the roots √GROW and √BREAK. In contrast, the nominal forms *John’s growth of tomatoes (15c) and *John’s break of the glass (16c) lack v, a supralexical causal apparatus, and thus lack the positive ||A|| necessary for the semantic licensing of an external cause (i.e., an affector in the force theory). Full profiling of the causal apparatus is observed for v_CAUSE; the patient’s initial and endstate positions, the affector’s force and its concordance or opposition with the patient’s force and other forces (if any) are all profiled imaging some progress toward the endstate via the transmission of energy from the affector to the patient (i.e., a full-fledged causal mechanism).

v_BECOME reflects the speaker’s focus on the patient’s change toward the endstate. At the computational level of analysis, A is not defined, ||P|| is greater than zero, but ||R|| and ||O|| can equal zero. Consider The glass broke (16b). In this sentence, v_BECOME provides the positive ||P|| which is not available in the root √BREAK (remember from (13) that √BREAK encodes only the result of the change of state). The nominal *the glass’s break (16d), in contrast, lacks the causal apparatus altogether and thus cannot involve the patient-related force (cf. Gärdenfors 2014).

So far, I have illustrated that a force-theoretic definition of causation as encoded in [V] on v sets forth the grammaticality pattern in (15) and (16). As noted above, for v_CAUSE and v_BECOME, it is possible that ||R|| and ||O|| do not equal zero. This is, in fact, the case of verbs of maintaining, such as keep and stay in (20a–b), which are dynamic but do not provoke a change of any kind.

Following Talmy (1988), Copley & Harley (2015) propose that keep and stay (and other verbs of maintaining, like sit, stand, lie, sleep, etc.) involve a force that maintains a property’s truth from one situation to the successor situation, against a tendency otherwise. To put this in terms of the force theory, in keeping and staying, there are v_CAUSE and v_BECOME, respectively, such that ||R|| and ||O|| do not equal zero, but the summation of the forces’ magnitude is zero and the transition is trivial as the endstate is the same as the initial state of the patient (see also Copley & Harley 2015).

Now let us see how the force-theoretic approach to v accounts for v_DO and v_BE. v_DO, as exemplified in (21), incises the initiation of a force on behalf of the affector (cf. Copley & Harley’s (2015) v_OCCUR).

Here, what is crucially profiled is vector A whose magnitude is greater than zero. Other segments of the causal apparatus provided by [V] on v are not involved in typical cases of activity situations. However, this does not mean that other vectors cannot be linguistically activated. See, for instance, (22) where a secondary predicate provokes the endstate vector E.

The stative v_BE, as in (23), makes a prominent reference to the position of the patient, that is to say, the position vector, as defined in the causal apparatus explained above, is profiled. (Note that the term “position” covers not only the patient’s spatial status but also its mental, social, etc. state of being, as well as its property attributions.)

There are two points to be clarified here: first, v_BE is different from the stay-type v_BECOME in that for v_BE other vectors (i.e., force vectors) and their magnitudes are not relevant and thus are not profiled, while for stay the interaction of forces and their total zero magnitude are of paramount importance; for stay there are at least two profiled forces exerted in opposite directions by the patient on the one side and by some other source on the other. The summation of the magnitude of these forces equals zero resulting in the sameness of the patient’s initial and endstate positions. Second, defining v_BE as referencing the stable position of the patient does not make it synonymous to n. [N] on n does not provide the apparatus illustrated in Figure 1 and thus is not capable of referring to the position of the patient in such an apparatus.

To sum up, [V] on v provides a configuration of forces and a position vector, that is, a causal mechanism/apparatus. Different flavors of v then reflect the speaker’s profiling of different substructures of this apparatus, that is, patterns of forces or the position vector. It should be noted that, although similar results can be found in Copley & Harley (2015), according to the present proposal, we can identify different flavors of v at an extra-linguistic level: different flavors of v encode and activate different patterns of forces regardless of the complements they will take at the linguistic level. Copley & Harley’s dual nature of forces as conceptual forces (perceived inputs of energy) and linguistic forces (functions from situations to situations), on the contrary, leads them to differentiate distinct flavors of v (all, except one (v_BE), of which are predicates of forces) only when they are involved in different linguistic contexts, for instance, when they take different types of arguments: v_BECOME is a predicate of forces that takes a predicate of situation, syntactically represented as a small clause, as its arguments, and v_CAUSE (v_APPEAR in Copley & Harley) is another predicate of forces that takes an entity, the internal argument, as its argument, for instance.

The current analysis has another advantage over Copley & Harley’s, allowing us to explain the force-theoretic causal distinction between v and n. Copley & Harley’s (2015) force-based treatment of predicates is a great leap forward in our understanding of the mapping between linguistic semantics and the cognitive system; however, it is limited to the verbal domain. Two issues arise here: first, dealing with schema (i.e., non-sortal) nouns, following Copley & Harley, one finds identical conceptual “net” force and situation in (14a) and (14c), repeated here in (24). Similar event and argument structures are observed in these examples, thus, one expects similar force analyses for the vP in (24a) and the nP in (24b). This is not tenable; from the discussion in previous sections, we know that due to the causal nature of v (and the non-causal nature of n), (24a) and (24b) must have different force-based analyses.

Second, in Copley & Harley (2015: § 4.3) roots are treated as predicates of forces that add information about the nature of the forces introduced by v. Based on this account, √DESTROY in The enemy destroyed the city (14a) and √BREAK in The glass broke (16b) are both predicates of forces (type <f,t>) that modify v_CAUSE and v_BECOME, respectively. If roots are predicates of forces that modify other predicates of forces (i.e., any node of type <f,t>), then we should accept that roots modify some predicate of forces in nominals, namely little n. This is not true; as argued above, n does not provide any configuration of forces and hence cannot be considered to be a predicate of forces.

The analysis in this paper addresses these two issues. The idea is that not only v, but also roots encode some causal apparatus (unlike n). Different root classes, as in (13), provide different patterns of force and/or position vectors. Roots of “destroy” class show a full causal apparatus where ||A|| and ||P|| are greater than zero, but ||O|| can equal zero. Roots of “grow” class encode a configuration of forces where A is not defined, ||P|| is greater than zero, but ||O|| can equal zero. Remember from (13) that √GROW is internally caused. For roots of “break” class only the endstate vector E is relevant. This model serves to explain the external/internal causal pattern of the noun phrase as well as the semantic licensing of the arguments in the nominal domain.

Based on this proposal, the force analyses of the vP in (24a) and the nP in (24b) can be schematized as in (25a) and (25b), respectively. Other forces (vector O) is disregarded here for the sake of ease. As shown in (25a), the causal apparatus of √DESTROY is mapped onto that of v_CAUSE in the verbal domain. Thus, the patient is identified as the city and the endstate is identified as the state of being destroyed. In the nominal domain, as shown in (25b), this mapping does not take place since n does not provide any causal apparatus. Thus, the causal reading of the noun phrase is merely rooted in the root’s force configuration. Note that the empty space for n in (25b) does not mean that n is meaningless, but that it lacks causal meaning.

The mapping in the verbal domain (25a) is not trivial; below, it is shown that the proper mapping of these causal apparatuses is crucial to have well-formed verbal constructions.

An interesting consequence of the view proposed here is that matching the causal apparatus encoded in v with the causal apparatus encoded in the root turns out why not all flavors of v can select all types of roots (e.g., v_BE does not select a root of “destroy” class): all flavors of v select roots with weaker or at most similar causal apparatus compared with theirs. v_CAUSE can select roots of different classes as it has a full causal apparatus itself, whereas v_BECOME selects roots of “grow” or “break” class, but not roots of “destroy” class since the causal apparatus in the latter involves vector A that is not defined in the causal apparatus of v_BECOME. This is schematized in (26).

On the other hand, when a flavor of v selects a root with a weaker causal apparatus (e.g., when v_CAUSE selects √GROW, as shown in (27)), the causal capacity of the phrase is augmented (e.g., an external cause can be expressed when v_CAUSE selects √GROW). This is not the case with nominals, as seen in (15a) and (15c) above. I leave details of this final issue for further research.¹²