I have discussed (PT ), which I take to be a philosophical analysis of action. There is, as far as I can tell, another distinct thesis discussed by linguists. It is not a philosophical thesis at all. It is an empirical hypothesis. I call it the ‘derivation thesis’, in the linguists’ non-logical sense of derivation: (DT) ‘…sentences like [Floyd melted the glass] derive from deep structures like [Floyd caused the glass to melt]’ (Lakoff 1965). My main interest in this discussion is to differentiate (DT) from any philosophical analysis , like (PT ).
As I mentioned before, there is a disagreement amongst linguists about whether the sentence with the transitive verb is basic and the causal sentence with the intransitive form of the verb is derived, or vice versa (or whether they are both generated from a common source). The Derivation Thesis, on which I now wish to comment briefly, holds that the derivation (or generation) is from a sentence about what an agent causes to the sentence about what he does. I have not addressed this view elsewhere in this book. Other than the few remarks I want to make now, I am not competent to enter into this non-philosophical discussion. Unlike the case with (PT ), the relevant literature assumes that the ‘deep-structure sentence’ and the derived sentence are meant to have the same meaning, and that any adverbial qualification in the deep-structure sentence must be accounted for in the derived sentence in a straightforward way.
What I want to do is to review and restate more fully some of the counterexamples to this derivation thesis. The counterexamples are not new (they were discussed extensively by Fodor (1970, 1975, 130–131, especially fn. 23) and by others), but I expand somewhat on their discussion. I follow convention by using (*), in this section of the chapter, to indicate a sentence either of dubious grammatical acceptability or of semantic oddity. Given the counterexamples, and the uncertain status that (DT) has in linguistics, it would be an error to try and support (PT ) by citing as evidence (DT).
The derivation failure of a sentence about what the agent does from a sentence about what an agent causes is often most salient when adverbial modifiers, both to do with time and manner, are added to a sentence. Therefore, these counterexamples are not counterexamples to any form of (PT ), which as stated makes no provision about how to treat adverbial modification on the two sides of (PT ). As far as (PT ) goes: ‘P caused o to Vi iff P Vt-ed o’ could be true, even if ‘P caused o to Vi F-ly iff P Vt-ed o F-ly’ were false. The adverbial modifications that ‘distribute’ in sentences that are equivalent often do so in more complicated ways than any such simple model would allow. Here is an example: ‘A football team wins a match iff it scores more goals than the other side’ states a true equivalence. (It is even an analytical truth about football.) Now add the adverb ‘elegantly’ to the left-hand side: ‘A football team wins a match elegantly’. We cannot ‘plug in’ this adverb in any clear way on the right hand side and still get a true biconditional: Winning elegantly is not the same as scoring goals elegantly, or scoring more elegant goals than the other side. For ‘winning elegantly’ we would need a much more complex sentence on the right-hand side.
So although the counterexamples that follow won’t work as counterexamples to (PT ), as it is currently formulated, I think (and the important point is that the linguistics literature assumes) that they work as counterexamples to (DT), which is about how the meaning of a surface structure sentence gets ‘generated’ from the meaning of a deep structure sentence. Katz’ example of the sheriff and the gunslinger (Katz 1970, 253, fn. 31) is put forward by him in a discussion of (DT). On the view of an interpretative semantics type of grammar that he is discussing and dismissing, ‘kill’ is said to be synonymous (my italics) with ‘cause to die’. At the end, I will also mention a few other counterexamples to (DT) that do not depend on adverbial modification, and these, but only these, might well also be counterexamples to (PT ).
To briefly recap the story I first told in Chap. 3: Henry is keen on Henrietta. Henrietta has a fetish about doors opening too quickly, so she asked Henry to open a door slowly for her, the opening to finish no later than at a specific time. Henry is anxious to comply. Moreover, there are several different processes by which Henry could get the door to start opening slowly (he can choose between various mechanisms that lead to the door opening slowly) and these processes or mechanisms leading up to the commencement of the slow opening of the door are of greater and lesser duration. But Henry is rather forgetful and laid back. He is almost horizontal. He was in no hurry to get things going, dilly-dallying all afternoon, but once he noticed the time ticking by, he needed to select the process that took the least amount of time to start the slow opening of the door.
Assume that (1) is true, given the story just told:
I shall argue in Chap. 7 that in this context, ‘quickly causing’ is really an adjective that modifies a pair of times, t and t*, that between them set the beginning and end points between Henry’s starting to cause the door to open and the door’s starting to open, and that says of them that they are close in time. This is what I called in Chap. 3 the temporal relational use of an adverb of speed, the broad outlines of which were described in that chapter. But in Chap. 7, I will expand on the application of this idea specifically to the case of ‘cause’. But this won’t be important for the point I am making here.
If a sentence with a transitive verb, ‘Vt
’ could be derived from a deep structure, namely, the causative sentence that uses the intransitive form of the verb, ‘Vi
’, then we should be able to derive from (1) either (1a) or (1b) or (1c) (this requirement is accepted, as far as I am aware, by the linguists who discuss the alleged derivation):
(1a) Henry opened the door slowly;
(1b) Henry opened the door quickly.
(1c) Henry opened the door slowly and quickly.
Making it attributive won’t help, because (1c) appears to say that it is both slow and quick and both qua an opening. One might wonder if the placement of the adverb makes a difference:
But (1d) is no better than (1c). (1d) still has the opening being both slow and quick, because there is nothing else for the adverbs to modify (if it is a mode adverb and not a phrase adverb). Adverbs must modify a verb, and there is only one verb in any of the candidate-derived sentences, including (1d), for the two adverbs in (1) to modify. (I will explain in Chap. 7 what adverbial modifications of the verb ‘cause’ amounts to.) For the reader who thinks that (1d) does help in capturing my story, I submit that that is because he reads (1d) as if it is asserting: Henry did something quickly that lead to the door opening slowly. But that of course is not what (1d) says. Neither (1c) nor (1d) makes any sense unless two verbs are at least implicitly understood (say, ‘did’ and ‘open’), so that the sentences are construed as saying something that they don’t explicitly say.
I have some difficulty in deciding whether, on the story told, either (1a) or (1b) is true. If they are not, we have a counterexample to (DT). But even if one of them were true, it would suffer from information loss compared to (1). Since both (1a) and (1b) suffer information loss when compared to the causative from which they are allegedly derived, neither (1a) nor (1b) can be equivalent to, or synonymous with, (1).
The problem is obvious and offers a method for generating many counterexamples to (DT). The causative has two verbs, in the case to hand, ‘cause’ and ‘open’. The allegedly derived sentence has only one verb, ‘open’. If both verbs are adverbially qualified in the causative, the derivation is bound to fail, as there will be information loss in the derivation, since there is only one verb in each of (1a)–(1d).
Other counterexamples use temporal qualification, and they work similarly, whatever view we take about the timing of an action. Both Fodor (1970
, 422–423) and Smith (1970
, 105–106), construct counterexamples using time. Here is one such example, taken from Smith:
Perhaps the secretary lit a 24-hour fuse. There is controversy about whether the secretary burnedt
the documents on Friday or Saturday (I discuss this in the next chapter). But the counterexample works whichever option is chosen, since my argument turns only on information loss and not on any dating discrepancy between the causative and the sentence allegedly derived from it. Documents can burn (completely) only once, so they can’t burn all up on both Friday and Saturday. If (2) were to entail:
The information about something having happened on Saturday is lost,
The information about something happening on Friday is lost.
There are a few more counterexamples, discussed by linguists, that don’t rely on adverbial modification that I want to mention. The two below have an interesting implication, which, as far as I am aware, has not been discussed more generally in the philosophical literature. Unlike the information-loss counterexamples that I have just discussed, these may be counterexamples to (PT ) as well, if they are counterexamples to (DT).
Both Cruse (1972
) and Smith (1970
) provide something like a categorisation of the counterexamples and the rationale behind them. Here are two from Cruse :
(3) John caused the sparks to fly.
(3*) John flew the sparks.
(4) John caused his mirror image to move (by changing his position vis-à-vis the mirror).
(4*) * John moved his mirror image.
(4) depends on ‘move’ taking as its subject something that is not a straightforward physical object. Cruse’s own example uses ‘shadow’, but I have substituted ‘mirror image’ since my intuitions are clearer in the case of a mirror image. One might argue that (4*) is not ungrammatical or nonsensical, but I find that suggestion farfetched. As with all intuitions, intuitions about these matters may differ, so I concede that this would be by itself a weak argument against either (DT) or (PT ). (3) seems to me to be a much stronger example.
What is interesting to my mind about both (3) and (4) is that if the counterexamples work, they work not just because of the verb involved, ‘to fly’ or ‘to move’, but only because and when the verb is combined with one specific kind of object rather than another. One would not get a counterexample with ‘John caused the airplane to flyi’ or ‘John caused his hand to movei’. This suggests that whether or not either derivation works for any specific verb does not depend on the verb alone, but the verb in combination with an object. Even if you thought that ‘he flewt the airplane’ and ‘he movedt his hand’ derived (in either or both the linguists’ ‘generational’ sense or the logical senses) from ‘he caused the airplane to flyi’ and ‘he caused his hand to movei’ respectively, surely you should not also be committed to thinking that ‘he flewt the spark’ and ‘he movedt his mirror image’ could be derived (in either the generational or the logical sense) from ‘he caused the spark to flyi’ and ‘he caused his mirror image to movei’ respectively, although both verbs are ergative verbs.
Compare in the same vein ‘Agent P broket the window’ and ‘Agent P broket the world record’. (The same contrast between these two sentences would arise if ‘break’ were replaced with ‘shatter’.) The first use of ‘break’ has a causative alternate but the second does not. When P breakst the world record, *the world record does not breaki, although it certainly is brokent (the passive form of the transitive verb). This might suggest that even for (PT ), whether (PT ) is true for an ergative verb sometimes depends not just on the verb but also on which object the verb takes. Perhaps a reply to this might be that the verb ‘break’ in these two sentences is ambiguous, and really signifies two different verbs. I think that this is a difficult question and would involve carefully distinguishing between metaphor and change of sense.
But that ambiguity certainly does not seem to be true in the case of ‘fly’. In ‘P caused the airplane to fly’ and ‘P caused the sparks to fly’, ‘fly’ seems to have the same meaning in both sentences, and yet one of the sentences appears to entail a sentence that uses ‘fly’ in the transitive and one clearly does not do so.