2.1 Properties of futurates

We saw above that futurates seem to make reference to an eventuality planned by a possibly unexpressed director, as in (15), and are infelicitous when the eventuality is not of the sort that can be planned, as in (16). Exceptions to this are “natural” futurates as in (17), in which there is no director.

Three other properties of futurates are worth mentioning here before we move on to an analysis. First, past progressive futurates permit an extra adverbial which modifies the intention or plan of the director.²

Secondly, in a number of ways futurates behave like present-oriented derived statives (Copley 2008b). For instance, the only epistemic readings of (19a) and (19b) are futurate readings or generic/habitual readings. This intuition is confirmed by the infelicity of (19c), which has no reasonable futurate reading, and by the infelicity of (19d) except in a context where someone has fixed the game. Since statives are compatible with epistemic readings of modals but eventives are not, this is evidence that the futurate is behaving like a stative.

Likewise, futurates in the antecedents of conditionals allow the consequent eventuality to take place before the antecedent eventuality, as shown in (20a) and (20b) (for a similar point, see also Crouch 1994). This behavior is like present statives, as in (21a), not like future-oriented (non-futurate) eventives, as in (21b).

Another way in which futurates behave like present statives is in the fact that the simple present futurate has a feeling of the plan being somehow permanent while the corresponding progressive futurate has a feeling of the plan being somehow temporary, as in (22) and (23) (Dowty 1979). In fact the very temporariness of (23b) rules the sentence out, as it is not a temporarily true fact that the sun is to rise at that particular time tomorrow, but rather a fact of long standing. This pattern is exactly the same, Dowty notes, as the pattern with lexical statives, shown in (24). Compared to the simple present sentence in (24a), the progressive sentence in (24b) conveys that the speaker’s living in Paris is somehow more temporary.

Together these generalizations suggest that there are (at least) two eventualities represented in the structure of futurates, and the higher one is stative. This higher eventuality also represents the director’s plan or intention, when there is a director.

The stativity of the higher, plan-related eventuality in futuruates points to several consequences. First, the idea that futurates are themselves stative, interacting with progressive or imperfective aspect in the way that lexical statives do, suggests that progressive or imperfective aspect is not responsible for their meaning. So, contra Dowty (1977; 1979); Kaufmann (2005); Copley (2008a; 2009), but following Copley (2014), their meaning is not associated with any overt morphology. The lack of morphology is a clue that the logical form of futurates is actually quite simple. If there is to be an unpronounced head carrying the futurate meaning, it should be a relatively simple meaning (we will return to this point below in section 2.5.3 below).

Second, the idea that futurates are themselves stative further suggests that, crosslinguistically, futurate readings should only be possible with aspects that can take statives as their argument. In particular, progressives that disallow statives in general should also disallow futurate readings. As far as I know this is true, though a more rigorous typological search should be done to verify or falsify this prediction.

A third way in which the stativity of futurates is suggestive has to do with the fact that normally futurates are quite bad in the (past) perfective (Vetter 1973), as shown in (25a), although futurate readings of perfectives improve for clauses embedded in sequence of tense structures, as in (25b). Perfective futurates can also be uttered in questions, especially negative questions with do support, as in (25c), and, perhaps just barely, in assertions that are heavily echoic to a previous perfective futurate question.

There is no explanation for these data in the literature that I am aware of. But while we don’t have a reason yet for the judgments in (25), we know that perfective meaning interacts with stative meaning, so we can be hopeful for a compositional explanation.

2.2 The structure of futurates

We have seen that futurates require a future adverbial (at least contextually); they seem to have a non-expressed animate director (absent in the case of natural futurates); and they are themselves stative, preferring imperfective/progressive aspect and resisting perfective aspect.

Given these facts, what is the structure of futurates? Remember that we need to know what the structure is in order for a what-can-cause-what theory to explain the inanimate exceptions to the apparent animacy requirement.

I adopt a modified version of the analysis for futurates given in Copley (2014). There I argued that the combination of a basic verbal structure with two subeventualities (i.e., a subeventuality e that the agent performs and the resulting subeventuality e′) and a future-oriented temporal adverbial such as tomorrow leads to a contradiction because ⟦tomorrow⟧^c (e) is not compatible with a present run time for e (or, if there is a past tense, with a past run time for e).³

Following the idea that any null head must be a semantically simple one, and noting that v heads introduce eventualities (typically a causing sub-eventuality representing what the agent does to cause the result sub-eventuality), I proposed an extra causal head (here notated v_extra) above VoiceP, as in (27), to introduce an extra eventuality argument. This allows a higher causing eventuality, namely e, to hold now, even though its effect, namely the agent’s action e′, is to occur tomorrow. In ordinary futurates, the higher eventuality e represents the director’s plan or intentional state.

The constraints on the causal structure, then, are as in (27): there is a cause and an effect, where p is true of the effect; the cause occurs now while the effect occurs at a future time such as tomorrow. The adverbial is obligatory in futurates because it triggers the temporal contradiction that drives a kind of “reanalysis” from (26) to (27)—this is not to be understood as a true reanalysis in a derivational sense but rather a choice between possible structures, whether for production or interpretation. If there is a future-oriented temporal adverbial, (26) is impossible and (27) must be used instead. It is effectively an implementation of the idea in Kaufmann (2005) that there is a reinterpretation or remedy of such sentences which allows them to be about a plan or a schedule; that is, about an intentional state.

The structure in (27) differs from the structure proposed in Copley (2014) in one respect: the addition of Voice to the syntax (Kratzer 1996; Pylkkänen 2008; Harley 2013), allowing the projection that introduces the agent/causer to be independent of the projection that introduces the causal relation. In Copley (2014), the vP projection introduces both the causal semantics (in the head v) and a null director (in the specifer of vP). However, since the director is always unpronounced, and since in natural futurates there seems to be no director at all, it is simpler to say that futurates have no syntactic realization of the director. We only get an implicit sense of a director because an intention is (almost) the only thing that the extra causing eventuality e can be, and intentions must be held by an animate entity.

We can say that directors are merely implicit in futurates despite an argument from Copley (2009: 54) that directors of futurates are visible in the syntax. That argument goes as follows. Although futurates are possible with animate and inanimate subjects, as shown in (28), there are certain subject-oriented adverbials that seem to force the director to be the subject. For example, as shown in (29a), Andrea must be in charge of her singing tomorrow, and in (29b), where the subject is inanimate, and so cannot be the director, the sentence is infelicitous. This contrast I took as evidence that the director is visible to the syntax, because it seems to be visible to the adverbial, insofar as the adverbial apparently places a requirement on the director.

However, on this account, since both (29a) and (29b) have an (animate) director, it is not really clear why (29b) would be ruled out. So we must retrace our steps, and in fact there is a relatively simple answer. If we understand the adverbial to require its argument to be syntactically provided, i.e., it has to be the subject rather than being contextually provided, the contrast in (29) is entirely expected. Noting that the adverbials in (29) require the subject to control whether and/or how the eventuality occurs, and that the unexpressed director is the one who controls the occurrence of the eventuality in futurates, it follows that the subject must be the director. So the adverbial takes the subject as its argument, and the meaning of the adverbial happens to constrain the subject to be the director; we don’t also need the adverbial to take the director as its argument directly. Therefore there is still no evidence that the director in futurates has a syntactically-visible realization—apart from accidental cases like these where the director happens to have to be the entity referred to by the subject DP. This is good news, since the existence of natural futurates indicates that there need not be a director per se.

We turn now to the question of the causal relation provided by the extra v head in (27): What kind of causation is it? Intuitively it seems that, like other verbal causal relations, this relation must be one of direct causation. The reason is that, whatever the definition of the difference between direct and indirect causation, if indirect causation were allowed for the relation in the denotation of the v head, unplannable eventualities such as rain tomorrow at 6 would be possible. There could, for instance, be something true today that indirectly caused it to rain tomorrow at 6. So, the notion of causation needed for the extra v head is, intuitively, direct causation. This is a welcome result, since it mirrors what happens in lexical causatives.

But let’s think through the consequences of making the extra causal relation one of direct causation. It means that, when there is an intention involved, that intention directly causes the agent’s action. At this point one might object that if the director and the agent are different people, the director’s intention cannot directly cause an agent’s action. At the very least, the agent has to have their own intention to act, and the director also has to convey to the agent what that intention is. Why would not these extra events between the director’s intention e and the agent’s action e′ make the causal relation between e and e′ one of indirect causation?

To answer this question, we must move beyond a merely intuitive understanding of direct causation. As Wolff (2003), among others, points out, certain intermediate events do not “count” as really intervening between cause and effect. For example, it is possible to use a single verb to describe what happens even when the agent uses an instrument. You can, for instance turn on the television by using a remote control, even though there is a causal chain of events that must happen between your depressing the button on the remote and the television turning on. For Wolff, this and similar facts suggest that the intermediate events are behaving not as independent causes, but as enabling events which help a sentient entity realize their intention. Such events do not get in the way of a relation of direct causation between cause and effect.

Similarly, in the case of futurates, the agent’s action would not count as an additional causing event, but merely as an enabling event, since it allows the director’s intention to be realized. This has the curious effect of making the subject of futurates a kind of instrument, conceptually at least, though not grammatically. But this seems correct, as the director is in control of the outcome, rather than the agent who is merely (as proposed by Copley & Harley 2015) the source of the energy that brings about the outcome. The event of communication of the authority’s intention would not count as an intervening cause either, but at most an intermediate enabling event that allows the director’s intention to be realized.

Further confirmation that we are indeed dealing with direct causation comes from the stativity of the intention argument. A requirement for direct causation helps us understand why the highest event argument e must be stative, and this in turn further bolsters the idea that we are indeed dealing with direct causation. In Copley (2014) I concluded that there is a temporal gap between an intentional cause today and its effect tomorrow, and therefore that the kind of direct causation involved (contra Fodor 1970) must not be limited to temporal contiguity. But actually, if e is stative, there is no need for a temporal gap between e and e′. Instead the state e can hold both now and at the time when e happens tomorrow, in a kind of double access (Enç 1987).

Double access is only possible for states, even though one might expect double access to require only a non-instantaneous temporal duration, in which case activities would also be expected to participate in double access. However, in the traditional examples of double access in embedded clauses, only states can participate in double access, not activities, as shown in (30) below.⁴ For an activity predicate to participate in double access, it must be converted to a stative predicate, for example as a generic or a progressive.

Now, given that the highest eventuality e in futurates is interpreted as an intention, it is a state, so it should be able to participate in double access. It also makes sense independently that there is double access of the state to both the time of the intention and the time of the effect e′, because if all else is equal, the director should not have changed their intention, in order for e′ to happen because of their intention. So not only is there is no reason to posit a temporal gap; there is reason not to posit a temporal gap.⁵

A double access requirement along these lines can also explain why past perfective futurates are generally infelicitous, as demonstrated in (25a) above. If double access of the intention to both the present (or past) and the future time is required for direct causation, and if perfective aspect conveys boundedness or closure (e.g. Smith 1991), entailing that the state does not hold after that time, then it follows that (past) perfective futurates should not be possible, because they require the director’s intention to have ended before the outcome is to begin. While interaction of this mechanism with questions as in (25c) and (25d) requires further investigation, it already makes sense that sequence of tense (as in (25b)) improves perfective futurates, insofar as an embedded clause in a sequence of tense structure may not involve a “real” perfective.

Two further questions about the interpretation of the proposed futurate structure in (27) are where the sense of authority comes from, as well as the where sense of settledness comes from. While the usual course of events is for the agent’s own intention to directly cause the actions of which they are an agent, the futurate cases differ only in that it is not specified whose intention directly causes the actions of the agent; it could be either the agent themselves or another entity. I would like to suggest that it is exactly in this case, where a person’s intention is treated as directly causing an agent’s actions, that we interpret as that person having authority over the agent. The sense of ability and control then comes from whatever it is about intentions that gives an intender the ability to control the outcome of their own action (Farkas 1988). This sense of control or ability is thus presupposed, and it in turn gives us the sense of settledness (Condoravdi 2002; Kaufmann 2005; Kaufmann et al. 2006) or plannability (Vetter 1973; Copley 2008a; 2009): Since the intender is presupposed to control whether the eventuality happens or not, and they intend for it to happen, then in the absence of unforeseen intervening events, it will.

2.3 Inanimate “exceptions”: Generalizing from intentions to dispositions

Working from this analysis of ordinary futurates, we must now find a way to generalize this analysis to account for the idiosyncratic inanimate exceptions, i.e. natural futurates. Strictly speaking, what we have been calling inanimate exceptions will not be exceptions at all. Since intentions are crucially involved in the ordinary futurates we have been examining, let’s start from the notion of intention, with the aim of generalizing it to something that does not require an animate or sentient entity.

Intention involves, at the very least, an intender y who holds an intentional state e bearing some intentional relation toward an eventuality description p, such that, if all else is equal, an eventuality e′ of that description occurs. The nature of the intentional relation could be understood as a preferential relation, where the state is some kind of preference for p, as in Heim (1992); or a dispositional relation, where the intentional state is defined as some kind of disposition to cause an eventuality that meets the eventuality description (as done explicitly in Portner 1997; somewhat implicitly in Condoravdi & Lauer 2009; and more explicitly in Condoravdi & Lauer 2012; 2016; Grano 2017).

The existence of natural futurates suggests a dispositional rather than a preferential analysis of intention for futurates. As Copley & Wolff (2014) point out, such cases suggest that the intentionality implicated here does not involve (only) preferences. If it did, then it would be mysterious why some inanimate exceptions are allowed to occur, as inanimate entities do not have preferences. If instead, we view intentions as a kind of disposition, the inanimate exceptions can be those which have a relevant disposition that under certain circumstances causes the eventuality.⁶

I will follow Fara (2001) in thinking of dispositions as essentially habitual or generic in nature: N is disposed to M when C is true just in case N has some intrinsic property in virtue of which N Ms when C. While intentional states are clearly not intrinsic properties, we can still use this idea to sketch the structure of a dispositional intention as in (31):

Of course it would be good to know more about the “certain circumstances”, corresponding to Fara’s “when C”. There is much to be said here about intentions which we will not be able to delve into, but for now we can informally say that someone with an intention toward p acts to cause an eventuality which instantiates p as long as—taking everything relevant they know into account and making a “closed-world assumption” that nothing they haven’t taken into account is relevant, i.e. nothing external intervenes—they expect that this action will cause such an eventuality (see Portner 1997; Kamp 1999/2007; Condoravdi & Lauer 2009; 2012; 2016; Grano 2017). So, in a very informal sense, and without relativizing it to the intender or the speaker, we can say that the certain circumstances are ceteris paribus (“(all) things are equal”) circumstances—nothing unforeseen intervenes.

With this rough definition of intention in mind, we can generalize intentions to a category of ceteris paribus dispositions, which includes physical ceteris paribus tendencies, by not requiring the holder of the state to be animate, nor the disposition to be an intention. So the intender y is generalized to a disposer, to include also inanimate holders of dispositions, and intention is generalized to a (ceteris paribus) disposition, to include also physical tendencies.⁷

A conceptual structure for ceteris paribus dispositions is characterized by the relations in (32) between a disposer (holder of the disposition) y, a dispositional state e, a manifestation e′, and an eventuality description p. These relations are also represented in the graph in (33). There is some redundancy in this set of relations: y being the holder of the state e which causes the eventuality e′ instantiated by p is effectively what it means for y to be disposed toward p. As we will see, however, when we discuss have causatives below in section 3, it will be useful to have a relation between y and p made explicitly available, as in (32d).

The causal relation in (32b) is the “dispositional causation” of the title of this paper; it is “dispositional” merely because the causing relatum is a dispositional state. The “directly” part will not be motivated here in this discussion on futurates, even though, as we saw above, there is reason to think the higher causal relations in futurates is a relation of direct causation. But if v heads generally impose directness, then we cannot detect whether the conceptual structure of disposition also imposes directness, or has a more underspecified relation. In fact, as far as futurates are concerned, the relation in (32b) could even be an entirely underspecified relation between e and e′. The case for direct causation as the relation between e and e′, however, will be motivated below in the discussion on have causatives; see the discussion around (76) below.

But how do ceteris paribus dispositions, or indeed any dispositions, cause their outcomes? This we cannot say. We can only say that they do. For example, in the case of intentions, Copley & Harley (2015) treat it as a law of rational behavior that intentions cause the intended outcome, if nothing prevents them from causing the intended outcome (this is the ceteris paribus condition). A similar law should operate in order to ensure that physical tendencies cause their manifestations.

We can now turn back to the structure of futurates to see how ceteris paribus dispositions help us account for natural futurates. This is summarized in Table 3. The idea is that the compositional semantics evokes, or matches, only certain parts of the dispositional structure, but that evoking one or more parts of it triggers the accommodation of the remaining parts.

From the structure in (27) comes the constraint that e must be a state (because, recall, double access is needed to ensure direct causation), and v^extra provides the constraints that e CAUSE e′ and p(e′). I assume that dispositional causation is generally the way in which states cause things, so the fact that e is a causal state evokes something similar in the conceptual disposition structure; this evocation is indicated by boldface. Likewise, the causal relation provided by the extra v head evokes the causal relation in the dispositional structure. Finally, the instantiation of p by e′ evokes the same in the dispositional structure. The obligatory temporal adverbial in futurates, on the other hand, has no counterpart in dispositional structure.

The dispositional structure in turn has an element that introduces a disposer y (the holder of the dispositional state), which shows up nowhere in the logical form of futurates. For ordinary futurates, y is the director. So, the idea that parts of the dispositional structure that are not evoked are accommodated, gets us the existence of the director, even though, as we have seen, the director does not get a syntactic representation. In the case of natural futurates, y would be the disposer, which would be constrained to be the subject by the world knowledge of what can cause what.

While what-can-cause-what theories such as this one do not make use of syntactically-visible animacy features, the syntax is still important because it closely tracks the compositional semantics and thereby tells us which inanimate exceptions to the apparent animacy requirement will be possible. In the case of futurates, the outcome occurs at a particular future time, so for natural futurates we are constrained to physical dispositions that can directly cause something to happen at this later time, even though not all dispositions themselves are constrained in this way. So we see how the causal structure in (27), constructed in the compositional semantics, combines with world knowledge about what can cause what to result in what looks like an animacy requirement, but which has idiosyncratic inanimate exceptions. Since the causing eventuality needs to be a state which causes, I assume it needs to be a dispositional state, so the question becomes about what kind of dispositions meet the requirements in (33). For natural futurates, what we are looking for is something that fits the causal structure, where e is not an intention but rather whatever disposition that causes the sun to rise at a certain time tomorrow, or the tide to come in at a particular time tomorrow. Since such physical dispositions are rare, so too are natural futurates. On this account, natural futurates are no longer truly exceptions; they are cases of dispositional causation just like the intentional futurates.

2.4 Heading off an objection

Two anonymous reviewers point out that (34) is unexpectedly odd, which poses a prima facie problem for the above analysis. The problem is that a disposition for the sun to rise at a particular precise time tomorrow can also be described as a disposition for the sun to rise at any time tomorrow, so we must find a plausible reason for the infelicity of (34) in order for us to have confidence in the proposed analysis.

The sun of course rises every day, so (34) is not very informative. However, as the reviewers point out, we cannot rely on the uninformativeness of (34) to explain its infelicity, because (35) is equally uninformative, but felicitous.

To explain the infelicity of (34), however, we need only slightly elaborate the analysis, turning to the fact that simple present futurates seem to require focus marking (Copley 2009), while true future statements, including the will statement in (35), do not.

The argument that simple present futurates alone require focus is based on contrasts such as the following. In (36), it is presupposed that there is a plan, and the question is about the content of the plan—depending on the focus, it could be asking whether the plan is for Joe (rather than Mary or Sue) to go skydiving tomorrow, for Joe to go skydiving (rather than going skiing or swimming) tomorrow, or for Joe to go skydiving tomorrow (rather than the next day or next week).

However, the examples in (37) do not need to evoke focus alternatives. The progressive futurate in (37a) can simply ask whether the plan is that Joe goes skydiving tomorrow, and the true future in (37b) can ask whether there’s a future time during tomorrow where Joe (actually) goes skydiving.

Thus, progressive futurates and true futures seemingly do not require focus in their complement.

Now, in order to explain why (34) is infelicitous, the idea is that the utterer of (34), but not the utterer of (35), evokes a disposition that is oriented toward a tenseless proposition which must include focus marking. This proposition will thus be one of the following:

So, (34) feels odd because it raises either an alternative where it is not the sun that rises tomorrow but something else; an alternative where the sun does not rise, but rather, does something else tomorrow; or an alternative where the sun rises on (only) a different day than tomorrow. All of these are absurd alternatives, and that is why (34) feels odd. On the other hand, (35) does not deal in alternatives it only says that there is a future time tomorrow at which the sun rises.

2.5 Competing characterizations

An anonymous reviewer asks whether the notion of predictability might not better characterize the difference between futures and futurates. The idea would be that intended actions, sunrises, and tides are all highly predictable while, e.g., rain at a certain time tomorrow is not, and it would be this difference that distinguishes eventuality descriptions that can occur in futurates from those that cannot occur in futurates.

The question with this story is how to derive infelicity, rather than mere falsity, from the fact of having a low-predictable eventuality described by a futurate. So whatever the nature of futurates, it should require that a low-predictable eventuality described by a futurate is a kind of category mistake, thus deriving the infelicity.

3.1 Have causatives are similar to futurates

We turn now from futurates to have causatives. As we saw in the introduction, have causatives are similar to futurates in several ways. First, they contribute a kind of authority to their matrix subject, which is along the same lines as the authority of the director in futurates. This was shown in the contrast between (49a) and (49b), where Madeleine could be the speaker’s daughter in (49a) but could only be the speaker’s boss in (49b).

We also saw that have causatives have a plannability constraint, requiring a plannable or controllable eventuality, just as futurates do. This was shown in the contrasts in (50), where Obama is understood to be Clinton’s boss.

Finally, we saw that similarly to futurates, there are idiosyncratic inanimate causer exceptions to the apparent animacy requirement.

Further examples of inanimate causer have causatives are shown in (52):

Inanimate matrix subjects apparently can occur in have causatives when the lower subject is animate and the caused eventuality is either a psychological state or a physical state that is indicative of a certain psychological state.⁸ The cause in these cases is a true cause as opposed to Pesetsky’s (1995) “target of emotion/subject matter”, since in (53a), as in a reading of (53b), Bill is angry at something other than the article itself; in fact he is angry at something that the content of the article causes him to be angry at.

The idiosyncratic inanimate exceptions here again point to a what-can-cause-what theory rather than an animacy feature. Note that the inanimate exceptions are not the same as those for futurates; they are idiosyncratic to have causatives, which is what we would expect if futurates and have causatives contribute slightly different structures.

3.2 Have causatives on the model of futurates?

Let’s start, however, from the similarities between have causatives and futurates, and try to get as far as possible on the hypothesis that they have the same structure. Adopting the analysis of futurates developed above, along with Copley & Harley’s (2009) idea that the futurate operator is the same as have, we can start by hypothesizing that have causatives have an extra v, as well as an extra Voice above it which introduces the matrix subject.

Like futurates, have causatives would then have an extra causal v head on top of the usual active Voice projection (if it has the usual active Voice and v projections to begin with; the inanimate exception cases in (51) and (52) do not). A second Voice head on top of the second v head introduces the matrix argument and is realized morphologically by have.

The advantage of this analysis is that it straightforwardly accounts for the similarity to futurates, explaining the data in the previous sub section. As we will see now, however, there are several arguments that point against an extra v head for have causatives. Some of these arguments do not go through, but those that do ultimately disqualify the analysis in (54). Despite this, a dispositional causation account of have causatives will still be warranted, just with a different structure from the one in (54), and with different ingredients of the conceptual dispositional structure left implicit.

3.3 Some existing arguments against an extra v head for have causatives

First, the analysis in (54) conflicts with that of Ritter & Rosen (1993; 1997) and Bjorkman & Cowper (2013). In all of these proposals, the position is that have causatives have a single eventuality, unlike make causatives, which have two eventualities. For example, Bjorkman & Cowper take the impossibility of passivization of a have causative as in (55a) as evidence that the have causative is too small to passivize. Likewise, they take the impossibility of the two adverbials in (55b) as evidence that there cannot be two eventualities.

However, (55b) already has a problem with the by phrase, as shown by the infelicity of (56a) below. The felicitous two-adverbial have causative sentence in (56b) is parallel to felicitous two-adverbial futurate examples as in (56c), as well as (18), above. Note that (56b) and (56c) are in the progressive.

In fact, they have to be in the progressive. As shown in (57), and similar to (55b), both futurates and have causatives become infelicitous in the (past) perfective:

As we saw above in section 2.2, the current proposal for futurates provides an explanation for (57b). This explanation turned on the idea that perfective aspect bounds a state, ensuring that the perfective in (57a) does not allow the double access of the intention needed for futurates. (Recall that the double access is needed to ensure that the director hasn’t changed their mind before the outcome takes place, and at the same time ensure direct causation of the outcome.) So a contradiction ensues for (57a): The throwing of the game has to be directly caused by an eventuality which takes place on Sunday night, and which (because of the perfective) cannot extend into Monday; but it also has to have happened on Monday. It is this contradiction that leads to the infelicity of (57a). This explanation can apply equally to the have causative case in (57b).

In sum, have causatives look exactly like futurates with respect to double temporal adverbials, as would be expected if they had the futurate-like structure in (57). So we must reinterpret what is responsible for the infelicity of (55a) and (55b): It is not the presence of two temporal adverbials. What could be the culprit? Of course by is not unrelated to the passive. Whatever is responsible for the problem with passivization in (55a) may also be responsible for the problem with by in (55b). Therefore, if we can specify that there is no passive Voice that can replace Voice^have, and that by cannot introduce a haver, (55a) and (55b) are not substantive problems for the hypothesized extra v structure in (54).

3.4 New arguments for the lack of v^extra in have causatives

There are, however, other arguments against the extra v analysis (54) for have causatives.

3.5 Explicit and implicit ingredients of dispositions

Having decided on a structure for have causatives, we still need to interpret it to account for the facts, including the similarities between futurates and have causatives discussed in section 3.1. The hope is that the facts can be accounted for by appealing to what-can-cause-what knowledge, resting again on the notion of dispositional causation. Indeed this hope will be justified. But first we need to understand how have causatives can have a disposition without explicitly representing a dispositional state as futurates do.

Recall the ingredients of dispositions: a disposer y, a dispositional state e, a manifestation e′, and a description p. As we saw earlier, these are related variously by several relations: y holds of e, e causes e′ ceteris paribus, e′ instantiates p, and y is disposed toward p. Together the ingredients, related in this way, constitute a dispositional structure.

I propose that futurates and have causatives both evoke dispositions, yet with a difference which will still allow for the similarities between futurates and have causatives. The difference between futurates and have causatives has to do with which of the ingredients of dispositions are explicit—in the sense of “represented in the logical form”, not in the sense of “pronounced”—and which are left implicit. The possibility to evoke ingredients that are left implicit comes, again, from the proposal that the dispositional structure is taken as a whole, so that if one or two ingredients are missing, they are still accommodated.

In the case of futurates, as we have seen, and as is repeated in Table 4, the logical form provides an extra causing eventuality that directly causes a temporally distant result. This extra eventuality must be stative, because the only way to fulfill the direct causation requirement and still hold of the present time is by double access, which is only available for states. So, of the ingredients of the dispositional structure, we get e, e′, and p from the logical form of futurates; specifically, from the denotation of v_extra. The disposer is left implicit, but given that we have evoked a dispositional structure, we fill in the disposer, with the intender (the “director”) in the case of regular futurates, and with the subject in the case of natural futurates.

In have causatives, on the other hand, as shown in Table 5, have relates y to the description p (the predicate relatum). Disposition is one of the possible specifications of this underspecfied relation contributed by have. When in this way a dispositional structure is evoked, the dispositional state e corresponds to the authority or relevant property; and a p eventuality e′ is caused to occur, ceteris paribus. Now it is the disposer y that is explicit, and the causal relation that is implicit—nearly the converse of Table 4. In this way, different logical forms can access the same conceptual structure even though they explicitly evoke different components of it.

3.6 What-can-cause-what for have causatives

It remains to ask what role dispositions play in the what-can-cause-what calculus for have causatives. They play a role in ruling out examples such as (74) and (75) as cases of interpersonal authority (i.e., they permit the director reading where John or the teacher is the creator of a fictional course of events).

An intention (which, again, is a kind of disposition) can apparently directly cause an animate entity to be an agent of another eventuality. This, we said, is what authority is. The example in (74) shows us that, outside of the director reading, an intention cannot directly cause an eventuality that does not have an animate agent, presumably because the only way intentions are causally powerful in the real world is through the actions of animate agents (Folli & Harley 2005). On the director reading, on the other hand, we are in a narrative context that could be anything its creator wants it to be, by fiat of the creator. Therefore, there are no such animacy restrictions for the lower subject on the director reading. Note that the infelicity of the examples in (74) and (75) also supports the case for the causation in dispositional structure to be direct; if it were not direct, this explanation of their infelicity would not go through, and it’s not clear what other explanation could replace it.

One kind of example that might give us pause regarding the constraint in (76) is that in (77), where John is an agent but not a volitional one:

Does the infelicity of (77) mean we need to change (76) to include a notion of volitionality of the lower agent? In fact, it does not. Just like examples where there is no lower agent at all, such as ✍ Madeleine had it rain, this is a case of a description that cannot be part of an intention of Madeleine’s outside of the director reading, since by definition intentions cause, ceteris paribus, a manifestation of the eventuality description. The possible eventualities are limited to those that can be caused, ceteris paribus, by the intention. So all of these cases are ruled out together.

The other main example we need to explain is that of inanimate exceptions such as (78a) and (79a), which can be compared with the minimally different (78b) and (79b).

The contrast within each minimal pair in (78) and (79) is instructive because they are lexically not different at all, though they are different syntactically. John in (79b) bears the AGENT role. But on the assumption that (79a) has a small clause complement, John does not bear an AGENT role.¹² Yet in both, conceptually, John is the source of the energy that causes the laugh(ing). What this underlines is that this “intuitive” notion that only intentions can directly cause agentive events really holds at logical form—so not on whatever lexical-conceptual elements are common to both (79a) and (79b), but only when they are in the particular syntactic structures. So the what-can-cause-what constraint that rules out (79a) is as follows.

What-can-cause-what constraints underline the idea that what-can-cause-what theories depend critically on the syntactic structure. But here we also leave the door slightly open for features. For the agent there is no need for an additional animacy feature. But for the distinction between intention and physical tendency, something is needed. It is not an animacy feature, as animate entities can have dispositions that are not intentions (for example, as in Madeleine had me laughing where she is not intentionally causing me to laugh). But the grammar needs to at least be able to see whether eventualities are intentions or not. Thus to distinguish intentions from other dispositions perhaps a feature such as +MENTAL will ultimately be needed on the state e.