1.

I start from a generalization about knowing how to do something which Stanley and Williamson themselves endorse. As I see things, Stanley and Williamson's single-minded attention to questions about the syntax and semantics of sentences used to attribute knowledge how ensures that the significance of their own generalization is lost on them. It is given expression by Stanley when he formulates the following schema.Footnote ⁶

(IAK) seems right. One is apt to say that a person could not have done something intentionally if she didn't know how to do it. And a further connection between the ideas of intentionally doing and knowing how to do can also be agreed on all hands.Footnote ⁷ The only things which are candidates for things which a person could know how to do are those that she might intentionally do.

I'll come back to the relation between the schema (IAK) as it stands and the sort of intuitive justification for it that I've just given. But first I bring out its implications for what knowing how to do something may involve.

It is often assumed that the word ‘intentionally’ or ‘intentional’ can serve to pick out what needs to be treated in an account of human agency. A certain line of thought may put this assumption in place – that a particular kind of explanation is proper to cases of human agency, namely reason explanation; and that that which, and only that which, someone has done by virtue of having a reason to do it, she has intentionally done.Footnote ⁸ What (IAK) suggests is that a different line of thought could lead to the same place: that which and only that which someone has done by virtue of knowing how to do it she has intentionally done. Insofar as the two lines of thought converge, knowing how to have done appears to go hand in hand with having had a reason to do. One might account for this by pointing out that when a person's having done one thing rationally explains her having done some other thing, then doing the one was how she did the one. Thus: that she φ-d explains why she Ψ-d, and Ψ-ing was how she φ-d – she φ-d by Ψ-ing.

This little account of the manner in which knowing how to do connects with the having of reasons to do can hardly be adequate as it stands, however. For one thing, it simply passes over the ‘know’ of ‘know how to do’. For another thing, the ‘by’ of ‘φ by Ψing’ falls short of the ‘how’ which introduces the idea of the means employed in getting something done. You may have broken the eggs because you were making a cheese omelette; but you didn't make the omelette simply by breaking eggs. If you made a cheese omelette, then you did so by breaking eggs and whisking them, grating the cheese in readiness, pouring the mix into a hot pan, and … – by doing all of these things, each in turn.

The two points – about knowledge, and about the need to do a series of things in order to get done what one has reason to do – are connected. When someone has φ-d (intentionallyFootnote ⁹ ) by first doing one thing, then another, then a third, a fourth …, the fact that she has φ-d is not a matter merely of her having done each of the things in turn. She knew what sort of things to do because she knew how to φ. Certainly her doing any of the things she did in sequence can be reckoned a part of her φ-ing. But she had no reason to do any of the specific things she actually did except insofar as, in appropriate conjunction with doing other things, doing them would subserve her end.

When one thinks of an agent as setting off with knowledge of how to φ, and as knowing that there are steps to be taken, it becomes easy to imagine that her φ-ing must be a matter of her carrying out a plan or programme whose details were settled at the start. But an agent's question how to do something doesn't lapse as soon as she starts on doing it. It is not as if someone who made a cheese omelette had first thought out how that was to be done, and then, having determined how, set her body onto automatic pilot so that the several intentions she had formed would be executed in turn. No. One's intentions themselves take on new specifications even in the course of their being fulfilled. She knew to get the frying pan out, but in practice how to do so depended upon where it had last been put. She had intended to scatter the cheese she had grated ahead of folding the omelette, but only at the point of introducing the cheese did determinacy attach to how much of the cheese to use. Of course someone who knows how to make an omelette will have no need, in the course of making it, to dwell upon such questions as are implicitly answered by an agent who keeps track. Still, one's knowledge how to make an omelette ensures that one can be alive to what one is doing if ever one engages in omelette making. So the series of particular steps someone took in exercising her knowledge on some actual occasion is not matched to any inflexible routine.

It is hard to say much of a general sort about what goes on when someone puts her knowledge how to do something into practice. And there is a danger, consequent on needing to be explicit about what an agent must have known, to falsify the phenomenology, making it seem that she must consciously have known it all. At any rate, I won't try to say more than starts to show up in the particular mundane example. The point of the example is only to draw attention to the fact that when an agent has done that which she intended to do, a much more detailed story could in principle be told about the knowledge how which actually she had used. For the agent herself, the question what to be doing next is always the question how to proceed as things now stand. In thinking close up about a particular example, one sees how much more has been going on when someone has done something than would come to attention in any explanation of her doing it which would ordinarily be given. (IAK), in portraying knowing how to as co-ordinate with intentionally doing, may remind one of just how much knowledge of how to act is in play in action. In order to bring this out, one needs to advert to the perspective of the one who knew what was going on – who, in exercising her knowledge of how to do what she did, knew what she was doing so long as she was doing it.

I have relied on (IAK), taken from Stanley and Williamson, in order to motivate thinking about the knowledge how that is actually exercised when someone does something. But I should note two discrepancies between Stanley's (IAK) and what I have tried to make from it. They used the schematic letter ‘φ’ whereas I have relied on the ‘something’ present in the ordinary language of ‘do something’.Footnote ¹⁰ And whereas I have stuck to the past tense, for the sake of explaining why one might endorse (IAK) and for thinking about an example, their schema on the face of it is present tensed: ‘She ϕ-s’. Well, I take it that so far as tense is concerned, they actually mean something generic: one imagines the conditional, ‘x intentionally ϕ-s, …’ as falling in the scope of an implicit ‘At any time’. But if so, then the use of a schematic letter in making a connection between ‘intentionally’ and ‘know how to’ might encourage one to focus on some single verb phrase as providing the whole of an answer to the question what someone is now intentionally doing. One would then be bound to fail to appreciate that at any time, the answer to the question what someone is doing may be ‘many things’. She is now pouring egg mix into a pan, and she is now making an omelette, and she is now preparing dinner. She knows how to do all these things if (IAK) is correct. Indeed if she's doing these things now, then there must be other things she knows how to do than any she is now doing.

2.

Stanley ‘s interest in what he calls ‘the folk notion of knowing how to’ leads him to confine his attention to ascriptions of knowledge how. Here are some of the things knowledge of how to do which are ascribed to the agents in Stanley's book: field a fly ball, do the Salchow, swim, ride a bicycle, change a light bulb, get to Boston, grasp a door knob, post a card into a slot.Footnote ¹¹ There is certainly variety here, ranging from kinds of know-how which it requires a skill to possess, kinds which perhaps need no skill but may need to have been learnt, and a more mundane kind which most able-bodied people in our social circumstances can be expected to have, or readily to come to have, without any special instruction. The variety goes hand in hand with differences as to what might be at issue if it were in question whether someone knew how to do something. Depending on the sort of know how in question, different sorts of interest will attach to whether someone has it. If one learns that X knows how to swim, then one comes to know that there are things X might do which a non-swimmer would never do. If one has been told a walkable route to the museum, which no doubt can be told by rehearsing some facts, then knowing already how to walk, one may use what one's been told in getting to the museum oneself. Any ordinary ascription of know how to X is made on the assumption that X has plenty of other knowledge, both of the knowledge-that and the knowledge-how kinds.

An application of (IAK) revealed that in the actual business of getting things done, a person makes use of much more knowledge how than would come to attention in thinking about what it might be said that she knew how to do (§1). And consideration of the actual use of sentences in which it is said that a person knows how to do something, now reveals that an interest in the truths conveyed in such sentences is not in the first instance an interest in knowing what states a person is in.

At one point Stanley asks ‘Why should one expect it to be a virtue of an account of knowing how that it is plausibly taken to be what is expressed by ascriptions of knowing how in natural language?’. He then imagines a challenge from someone who thinks that ‘science could show us that states of knowing how are very different in kind from what ordinary speakers use sentences like ‘Ana knows how to swim’ to express’.Footnote ¹² But in order to put it in question whether an account of knowing how can be confined to uncovering the structure of sentences in which knowledge how is ascribed, there is no need to speculate about whether scientists might postulate a kind of knowing how unfamiliar to the folk. Stanley and Williamson themselves say that ‘intentional actions are … employments of knowledge-how’.Footnote ¹³ If that is right, and if one aims to provide an account of the nature of knowledge-how, then one surely needs to think about its employment. Stanley has a view about its employment. Employing one's knowledge how to do something, Stanley thinks, is a matter of being guided by one's knowledge how to do it. His claim against Ryle is not only that knowing how to do something amounts to knowing a fact, but that someone's acting amounts to her being guided by her knowledge of facts. (See e.g. page 175 of Know How. That we are ‘guided by our knowledge’ of how to do things is a recurrent theme in his and Williamson's paper.)

3.

Let me turn to the argument Stanley and Williamson gave for saying that knowledge how to do something is propositional knowledge. It starts from the thought that the word ‘how’ of ‘know how’, belongs with question words, including ‘when’, ‘where’, ‘what’, ‘who’, … , etc. In such sentences as ‘A knows [how —]’, ‘A knows [when —]’, ‘A knows [where —]’ ... , the ‘[how —]’s and the ‘[when —]’s and the ‘[where—]’s are treated as embedded questions in standard semantics. So the knowledge attributed to A in such a sentence is knowledge of the answer to a question – the question ‘How —?’, ‘When —?’ … A sentence may goes into the slot ‘—’ following the question word (as in ‘She knows how alcohol affects the body’). But in sentences that are of particular interest in a debate about knowing how to, there will be a verb in the infinitive in that slot (as in ‘She knows how to set the thermostat’ and, for that matter, ‘She knows what to do next’).

Where an infinitive comes after the question word, a sentence has to be constructed from the infinitive if something propositional is to be got. That means that the verb ϕ has to be supplied with a subject, and it has to be made finite so as to be predicable of the subject. Take as example ‘Hannah knows how to swim’. Stanley and Williamson use the linguists’ unpronounced pronoun ‘PRO’ as its subject, inheriting its reference from its antecedent (‘Hannah’ here); and they make the verb finite using a modal auxiliary such as ‘could’. Thus for ‘Hannah knows how to swim’ Stanley at one point has ‘Hannah knows how she could swim’, which he says is ‘a natural paraphrase of “Hannah knows how to swim”’.Footnote ¹⁴ The putative paraphrase does away with the infinitive, but it still contains the ‘how’. So there is more work to do in linguistic theory. The idea then is that Hannah's knowing how she could swim is a matter of her being acquainted with some specific way or ways of swimming w, and knowing that w is a way she herself can or could swim – a way that will give her counterfactual success in swimming.Footnote ¹⁵

One might wonder why someone who knows ways of doing something should need, if she is to know how to do it, also to know propositions concerning those ways. It is as if Stanley thought that in addition to knowing ways, one needs to know something else knowledge of which will guide one in doing it. Stanley appears to think that knowing propositions follows somehow from knowing ways. He says:

The ‘hence’ here effects a transition from knowledge of ways of fielding fly balls to knowledge of propositions, or facts, about those ways. In explication, Stanley says:

But why should an explanation be needed of why certain acts constitute exercises of skill?Footnote ¹⁸ If intentional actions are .. employments of knowledge-how (to quote Stanley and Williamson once more), then X's catching a fly ball was X's employment of X's knowledge how to catch a fly ball. No more is needed to see X's catching of a fly ball to be the exercise of a skill. Someone who knows how to do something, whether or not this is something she knows by virtue of having acquired a skill, may, when she is in a position to do it, simply exercise her knowledge how to. She is in no need of further knowledge which will guide her. (I note that ‘exercise’ is the word which Ryle used wherever Stanley would have ‘is guided by’. And Ryle spoke of knowledge how as a capacity. It may be that Stanley's speaking of knowledge how always as a disposition ensures that he has no use for the idea that it may simply be exercised, despite his occasionally allowing that it may be ‘employed’.Footnote ¹⁹ )

4.

When Stanley introduces knowledge of propositions about ways, the idea that the infinitive ‘to ϕ’ of ‘know how to φ’ gives way to a sentence having ‘PRO’ as its subject rather gets lost sight of. But the idea recurs. For in Stanley's theory, the unpronounced element PRO is assumed to be the subject of infinitival clauses generally; Stanley speaks of ‘the widespread consensus that infinitival constructions .. are expressions of de se attitudes’.Footnote ²⁰ Thus a theorist of the linguistics to which Stanley subscribes will say that a ‘PRO’ is present in the underlying syntax not only where an infinitive is preceded by a question word – as in ‘John knows how to hit a ball hard’ or ‘Mary knows when to add the cheese’ – but also where there is a simple infinitive, as in ‘John tried to win the race’ or ‘John wants to become a doctor’. Inasmuch as Stanley and Williamson's treatment view of ‘know how to’ goes hand in hand with a view about sentences containing ‘try to’ and ‘intend to’, it will evidently have repercussions for philosophy of action beyond any I have so far touched on. The repercussions are worth exploring. There is no need to subscribe to any particular theory of infinitives to think that the same sort of construction is found following ‘know how’ and such words as ‘try’, ‘want’ and ‘intend’. Even from the standpoint of superficial syntax, connections between such infinitives are evident. These infinitives can co-occur in sentences, apparently in construction one with another. Consider: ‘Someone who did something that she intended to do knew how to do it.’ Or: ‘If someone knows how to do something, then it will be no accident if her trying to do it should be her actually doing it.’Footnote ²¹

When Stanley argues that a full sentence must stand in place of an infinitival construction following ‘tries’ or ‘wants’ ‘or ‘intends’, he takes ‘try’ as his test case. In accordance with the consensual view, ‘John tried to win the race’ is to be written ‘John tried PRO to win the race’, and according to the propositional theory which Stanley endorses, this is to be understood as saying ‘John stands in the trying relation to a proposition about John’.Footnote ²² Really? It is hard not to notice that ‘stands in relation to’ is mismatched for tense with ‘tried’: perhaps Stanley meant to write ‘John stood in the trying relation to’. But how are we to understand the putative proposition to which John stands, or stood, in relation? It is about John, and it is expressed with a sentence whose subject is ‘PRO’ referring back to John. One wants to know what is predicated in the sentence. The only clue Stanley gives as to how ‘win the race’ is to be inflected for tense or otherwise made finite is given when he treats a different sentence and speaks of ‘assuming that the meaning of the infinitive “to win” … is equivalent to the modal “will”’.Footnote ²³ But suppose now that John has won the race: that which he tried to do, he did. In that case, it would seem that the proposition about John to which John stood in ‘the trying relation’ is one to which he also stood in a ‘doing relation’. But ‘John won the race’ surely does not say that John stands in relation to a proposition about John.

It is not surprising that we are at a loss to know what proposition it could possibly be to which someone stands in relation by virtue of trying, or having tried, to do something. There are no intelligible sentences of the form ‘She is trying that—’ or ‘She tried that—’ from which to seek illumination. It might be thought that ‘try to’ is somehow to be exempted from the usual treatment of infinitives. ‘She intends that—’ is intelligible; and so perhaps is ‘She wants that—’ (or at least ‘She wants it that—’). Perhaps then a philosopher can join the linguists’ consensus about infinitival constructions but set ‘try to’ aside as somehow exceptional.

It is true that English (even if not other languages) allows the construction ‘intends that —’. Still, the view about ‘intending to —’ which is the counterpart view of Stanley's view of trying can hardly be in better standing than Stanley's view of trying. For ‘intend to’ and ‘want to’ share with ‘try to’ the feature that makes a difficulty for the idea that there is some proposition to which one stands in relation by virtue of one's trying to do something. That which one may intend to do or want to do, like that which one may try to do, one may also simply do. Mary, who intends to eat the whole pie, is doing what she intends if she is eating the pie, and she will have done what she intended when she has finished it off. But eating the pie, even eating it until it's finished, is surely not a proposition. (‘Mary ate the whole pie’ expresses a proposition. But that Mary ate the whole pie is not what Mary did.) The fact that one may do what one intends to do apparently makes a problem about taking the ‘to do’ of ‘intend to do’ as holding a place for a proposition.

I think that this is a genuine problem. I want to show that it infects the view of intention held by Williamson. Williamson's view is present in a paper which explores an analogy between knowledge and action by pursuing an analogy between belief and intention. I shan't here be concerned either with the analogies or with Williamson's general view of the practical: I attend exclusively to the account of intention there. The account evidently derives from views about sentence structure found in the sort of contemporary linguistic theory he first applied to the case of ‘knowing how’ in his paper with Stanley. This is the theory which I think can only lead a philosopher of action astray.

5.

On the face of it, Williamson's analogy between belief and intention is hindered by the fact that declarative sentences ordinarily follow the word ‘believe’ but don't ordinarily follow the word ‘intend’. On the face of it, to compare the contents of beliefs with the contents of intentions is not to compare like with like. The point is slightly obscured, thanks to Williamson's giving the name of verb phrases both to phrases that contain finite verbs such as 'melted the butter' and to infinitival verb phrases such as 'to melt the butter'. But however that may be, Williamson thinks that there is no real hindrance. He says that if ‘intend’ is followed by a verb phrase, ‘the verb phrase still needs a subject’. And ‘[I]n “one intends to ϕ”, the unpronounced implicit subject of “to ϕ” is mandatorily co-referential with “one”’. Here word Williamson shows himself as signed up to the consensus that infinitival constructions following such words as ‘intend’ are expressions of de se attitudes. Thus Williamson takes the syntax of the sentence ‘John intends to have a drink’ to be given when it is written ‘Joe_i intends PRO_i to have a drink’. The English version of this is: ‘Joe intends himself to have a drink’.

Williamson acknowledges that one might be bothered by a mismatch between the words we actually use in attributing intentions and those we should use if we stuck to the letter of his account. We simply don't say ‘She intends herself to so-and-so’ when she intends to so-and-so. Williamson's response is to provide an explanation of the difference between the normal sounding ‘Joe intends to have a drink’ and the abnormal sounding ‘Joe intends himself to have a drink’. He says that the difference between them ‘is a non-semantic conversational effect, predictable on quite general grounds… If a speaker envisaged a normal case of Joe intending to have a drink, then going to the trouble of adding the redundant word “himself” would be pointless and prolix, thereby violating Grice's conversational maxim of manner.’ Well, this obviously cuts no ice with someone who questions the syntactic theory, and who thinks that ‘to have a drink’ as it follows ‘intends’ lacks a subject. Doubting that infinitival verb phrases always need a subject, she thinks that ‘himself’ is simply absent from ‘Joe intends to have a drink’.

Williamson speaks of finding no evidence against the ‘assumption that the objects of intentions are as propositional as the objects of belief’. But he needs more evidence in its favour. If the assumption were correct, then not only would a subject be implicit in ‘to have a drink’ but something would be predicated of that subject. Yet Williamson, no more than Stanley, tells us how ‘to have a drink’ is to be made into something predicable of ‘PRO’/‘himself’. A simple way to move from ‘intend to have a drink’ to ‘intend himself that —’ would be to say Joe intends that himself have a drink’. But insofar as ‘have a drink’ is here subjunctive, no proposition is expressed by ‘Himself have a drink’.Footnote ²⁴ Is it then that Joe intends that himself be having a drink, or that himself will be having a drink, or that himself will have had a drink? Less implausible than any of those perhaps is that Joe intends that himself will have a drink. But suppose that Joe has just now finished having a drink: he intended to have one, his intention didn't lapse, and it is just now fulfilled. That which he intended can hardly now be that he himself will have a drink (although I suppose Joe might already have decided to have another drink).

And there is more at issue than whether ordinary ways of talking count against his assumption that intentions have propositional objects. Williamson writes as follows about the success conditions for intentions:

Two points about the ‘to bring it about that p’ here are striking. (1) It is infinitival, so that Williamson has departed from his official view, perhaps because of the difficulty there is about manufacturing a propositional intention out of one expressed with an infinitive. (2) It introduces ‘bring about’, so that ‘to open the door’ is equated with ‘to bring it about that the door is open’. Of course there is something right about what Williamson says. Very certainly if the door is open (at a time) you cannot fulfil an intention you have to open it (then). And if the wind blows the door open, then the wind’s action is very likely to put a stop to your intention. If you intend to open the door, there can be no question of the wind fulfilling your intention. Indeed when Williamson allows that you may or may not do what you intended to do, he seems to recognize that you will fulfil your intention to open the door only if you open the door. And if you are able simply to open the door, then you would have no need to intend to bring it about that the door is open. When it comes to an intention to bring it about that p, it seems impossible to say anything general about what might need simply to be done in order for the intention to have succeeded. The certain thing is that no-one can intend to bring anything about unless she is able to simply do some things intentionally. It seems impossible then that someone should ever have intended to bring anything about unless there were things she could both intend to do and was able simply to do. So the fact that propositions are not the kind of things one is able simply to do matters. It ensures that ‘intend that —’ could not possibly always record what is said with ‘intend to ϕ’.

6.

I want finally to come full circle. I haven't gone into very much detail about the treatment of sentences containing ‘know how’ and ‘intend’ that Stanley and Williamson provide on the strength of current linguistic theories. But I've only meant to put aspects of their treatment in question. In §1, I suggested that when one attends to the knowledge how which is actually exercised in doing something intentionally, one won't think that an account of such knowledge could be got from attending to the grammar of sentences used when knowledge how to do something is ascribed to someone. And so I think it is for an account of intention. Williamson writes ‘Sometimes, even though one ϕs and intends to ϕ, one does not intentionally ϕ. A would-be assassin may accidentally run over and kill the man he intends to kill, without intentionally killing him.’ Indeed. It could be that the explanation why the would-be assassin did not intentionally kill his intended victim is that the knowledge how he exercised when he accidentally ran him over was not knowledge how to kill him. This explanation is in keeping with Stanley and Williamson's own (IAK).