Anscombe’s book, like Wittgenstein’s Philosophical Investigations, moves back and forth across the topics it covers. This can often be confusing. The following is a reading of the structure of the first half (§§1-31) of Intention.
We speak of intention in at least the following three ways:
1. Intentions to act
2. Intentions in acting
3. Intentional action
Anscombe notes that focusing exclusively on one of these might lead us to make claims about intention which are false of the other two cases.
One response to this is to say that this is not a problem; ‘intention’ might have different senses, like ‘bank’ does. But Anscombe thinks that this move is implausible; clearly the case of ‘intention’ is not much like the case of ‘bank’. The moral of the story seems to be that although intending to act is different from intending something in doing an act, and each are distinct from intentionally undertaking an act, an account of intention ought to explain the very close relations between these notions.
Anscombe begins by focusing on the first kind of intention: intentions to do something. A first problem in giving an account of the nature of intentions to act is to explain what distinguishes intentions from predictions.
One clarificatory point: ‘prediction’ has two meanings here. In one sense, a prediction is a kind of speech act: a claim made about the future. In another sense, a prediction might be unexpressed; in this sense, a prediction is just a belief about the future, whether or not it is voiced. When Anscombe asks about the distinction between intentions and predictions, she is in the first instance asking about the distinction between predictions in the first (verbal) sense and expressions of intention: she is asking about the distinction between two different kinds of utterances. But this is done as a way of getting at the distinction between two different kinds of mental states: predictions in the second, non-verbal sense and intentions to act. It will be useful to reserve ‘prediction’ for the kind of utterance, and so to contrast it (as Anscombe usually does) not with intentions, but with expressions of intention.
As Anscombe says, the distinction between predictions and expressions of intention is intuitively clear. The example of ‘I am going to fail this exam.’ The problem is to provide a theoretical account of this distinction: to say what underlies the intuitive difference.
One possibility is to try to find some grammatical distinction between predictions and expressions of intention. But as Anscombe points out, if ‘grammar’ is taken in a narrow sense, this idea is a nonstarter, since a single sentence can be, in different uses, either a prediction or an expression of an intention.
Differences in justification
A second possibility is to focus not on the grammar of expressions of intention, but rather on the justification that speakers have for those expressions. Anscombe suggests that here we can find a distinction between predictions and expressions of intention: predictions are justified by evidence that the future state of affairs in question will be true, whereas expressions of intention are justified by reasons for thinking that state of affairs attractive (§3).
Recall the example of ‘I am going to fail this exam.’ If this is a prediction, we would expect a request for justification to yield a response like ‘I’ve failed every other one in this class.’ - i.e., some evidence that the state of affairs which is my failing this exam should come to pass. But if this is an expressions of intention, a request for justification might yield a different sort of answer, like ‘I really want to annoy my parents.’ This is not, on the face of it, evidence that the state of affairs in question will come to pass; rather, it is a reason why the agent finds this state of affairs attractive.
(It’s worth noting that requests for justification of expressions of intention can also yield a different sort of answer, e.g. ‘Because I always do what I want.’ Here what is offered does seem to be evidence for the proposition that the state of affairs in question will come to pass, rather than a statement of one’s reasons for bringing it about. But this needn’t be a serious objection to Anscombe’s distinction -- it only makes sense as evidence on the supposition that the agent in question intends to fail the exam.)
It seems that this is a genuine difference; but Anscombe does not press further in this direction. The reason seems to be that the contrast rests on a contrast between evidence (or ‘reasons for belief’) and reasons for action. But the distinction between reasons for action and other uses of ‘reasons’ is a heading which she discusses later in connection not with intentions to act, but with intentional action (§5 ff.).
Nonetheless, the implicit conclusion at this stage of the text seems to be that we cannot give an account of the distinction between intention and prediction in terms of the distinction between reasons for action and justifications for belief. But no argument is given for this view at this stage.
How we identify the intentions of others
Anscombe observes (§4) that, if we are asked to describe what a person is doing, we will usually provide a list of his intentional actions. These are often quite obvious to us, and need no special verification. Further, most of the intentional actions we ascribe to such a person will also give that person’s intentions: if we observe that someone is reading, then it will usually also be true of him that he intended to read, or intends to be reading.
This is not particularly contentious, but it is a bit hard to see how it fits into the line of argument that Anscombe is developing here. It is puzzling for at least two reasons. (i) Anscombe has been discussing intentions to act rather than intentional action or intentions in acting, but here she spends most of the passage, without explaining why, on the latter two topics. (ii) The fact that we can usually tell without a problem what intentional actions someone is undertaking seems on par with the fact that we can distinguish without problem between predictions and expressions of intention. But just as the latter fact does not get us very far in explaining what the distinction between beliefs about the future and intentions is, so this observation about our ascriptions of intentions does not seem to get us very far in explaining what intentions are.
The last several paragraphs of §4 go some distance toward explaining what Anscombe has in mind here. Her first aim is negative: to combat a certain view about the nature of intentions, and their relation to intentional action. She writes:
“...a man can form an intention which he then does nothing to carry out, either because he is prevented, or because he changes his mind: but the intention itself can be complete, although it remains a purely interior thing. All this conspires to make us think that if we want to know a man’s intentions it is into the contents of his mind, and only into these, that we must enquire; and hence, that if we wish to understand what intention is, we must be investigating something whose existence is purely in the sphere of the mind; and that although intention issues in actions, and the way this happens also presents interesting questions, still what physically takes place, i.e. what a man actually does, is the very last thing we need consider in our enquiry. Whereas I wish to say that it is the first.” (§4)
It seems that Anscombe is claiming that we are tempted into thinking of intentions as mental states which are independent of intentional actions, in the following sense: some intentions go on to cause certain intentional actions, but this is not a part of the nature of intentions to act. This is, as she says, a tempting picture precisely because it seems clear that we can have fully formed intentions which do not issue in intentional actions. One of Anscombe’s aims seems to be to undercut this picture.
A few points about this: (i) Anscombe is certainly right that this is a tempting picture; it is endorsed by most philosophers of action, including most causal theorists. (ii) If her aim is to use the way we ordinarily identify the intentions of others to undercut this picture, then it is hard not to be disappointed by the argument. It seems to be a straightforward case of confusing epistemic and metaphysical priorities. Just because we come to know about X’s via coming to know about Y’s, it does not follow that X’s are metaphysically dependent on X’s. (Compare electrons, and the blips they cause on a measuring device.)
A better reading of what’s going on here is that this is meant to add support to an implicit argument in §§2-4 that we cannot understand intentions to act independently of intentional action. This is presumably why, after considering several failed attempts to explain the distinction between expressions of intention and predictions, she moves in §5 to the discussion of intentional action.
If Anscombe is making a positive point in the theory of intention here, then it seems to be a kind of claim about priority: if we want to give an account of intentions to act, then we can only do so via an adequate account of intentional action.
Anscombe begins her treatment of intentional action by proposing the following definition:
|A acts intentionally||
a certain sense of the question ‘Why?’ applies to A’s action
What we need to do is explain what this sense of ‘Why?’ is, and what it means for it to have application.
One easy way to explain the relevant sense of the question is to say that it is a request for the agent’s reasons for action. Then the proposed definition would say that an agent acts intentionally iff the question of what his reasons for acting are has application. The problem, Anscombe seems to think, is that the link between intentional action and action for a reason is too tight for this definition to be illuminating.
But we can still use the notion of a reason for action to get some intuitive grip on the meaning of ‘Why?’ on which Anscombe is trying to focus. So far Anscombe has contrasted reasons for doing something with reasons (evidence) for thinking that a certain proposition is true. Here she also contrasts reasons for action with (mere) causes of events. The questions ‘Why did the leaf fall off the tree?’ and ‘Why did he leave his wife?’ are superficially similar; but, intuitively, this similarity conceals a deeper difference. In the case of the man leaving his wife, we are asking what the man’s reasons for action are; but we are not similarly asking for the leaf’s reasons for falling off the tree. Plausibly, leafs don’t have reasons for falling off the tree. Instead, we are asking for the reason why a certain event occurred, and not some agent’s reason for bringing it about. One intuitive (but potentially misleading) way to frame this difference is to say that in the case of the man leaving his wife, we are asking for reasons for action, whereas in the case of the leaf, ‘Why?’ is a request for causes of an event, not for reasons.
This gives us three senses of the question ‘Why?’:
Anscombe’s claim is that an action is intentional iff the question ‘Why?’ in the third sense has application. Her aim is now to explain this third sense without relying on the intuitive distinction between reasons and causes. She does this, first, by explaining what it takes for the relevant sense of ‘Why?’ not to have application.
There are two ways in which the third sense of the question ‘Why?’, when asked about an agent involved in some action or movement, can fail to have application:
(i) The agent in question does not know that she is involved in that action or movement. (§6)
(ii) The act in question was involuntary. (§7)
The problem with (ii) is that, as with the proposed account of intentional action in terms of reasons for action, the voluntary/involuntary distinction is too closely linked to intentional action for an account of the latter in terms of the former to be informative. So we need to explain the relevant class of involuntary acts in some non-circular way.
Anscombe’s idea (§8) is that this can be done in terms of the notion of knowledge without observation. Roughly, we know something without observation if we know it, and there is no separable sensory event on the basis of which we know it. Some involuntary actions are not known in this way - ‘the peristaltic movement of the gut’ - but we needn’t worry about excluding these, since they are covered by the first way in which the question ‘Why?’ may be refused application. The question is rather: given that the class of acts which are known without observation includes some that are voluntary/intentional and some that are involuntary/nonintentional, how can we explain this distinction?
Anscombe points out that sometimes we know without observation not only what movement or action we are engaged in, but also that the movement was caused by something in particular. Suppose, e.g., that I am sitting at my office desk, and get up and open the door; just before this, I heard someone knock on the door. I cannot, of course, know without observation that someone knocked on the door; I know this only by hearing it. But, Anscombe thinks, given that I know that someone knocked on the door, I can know without observation that I opened the door because someone knocked on it.
Now contrast this with a case of involuntary action, such as the reflex movement of one’s leg after being tapped on the knee. Anscombe thinks that in this case, unlike the case of voluntary actions, I do not know without observation that the doctor’s tapping me on the knee caused my reflex movement. This serves to distinguish involuntary acts (of this type) from voluntary ones.
(Anscombe also restricts the class of involuntary acts in question to acts given a ‘purely physical’ description; but this restriction seems unnecessary. It seems that the only non-intentional acts of which we have non-observation knowledge at all are ‘purely physical’ in the intended sense, since in all such cases the explanation of the non-observational knowledge is proprioception. So all ‘non-physical’ nonintentional acts are already ruled out by the first way in which ‘Why?’ can fail to have application.)
This suggests the following account of the two kinds of cases in which the question ‘Why?’ fails to have application:
(i) Cases in which the agent does not know what he is doing.
(ii) Cases in which the agent does have non-observational knowledge that he is ing, but lacks non-observational knowledge of the fact that such and such is the cause of his ing.
There is, however, reason to think that this does not capture Anscombe’s intention. For consider a case in which an agent knows, but not non-observationally, that she is engaged in some action. Suppose, e.g., that she is in physical therapy and can see in a mirror a machine moving her arm around in circles. We should say that this is a case in which the question ‘Why?’ fails to have application -- after all, her movement is surely nonintentional. But this case is not ruled out by (i) and (ii). It is not ruled out by (i), since she does know what she is doing, and is not ruled out by (ii), since this knowledge is not non-observational.
This indicates that, despite the way that Anscombe puts things in §6, clause (i) should be strengthened to rule out all cases in which the agent lacks non-observational knowledge of the movement or action in which he is engaged. It is plausibly anyway that there is a necessary connection between intentional action and non-observational knowledge, and elsewhere Anscombe seems to rely on such a principle (e.g., the way she rules out the peristaltic movement of the gut as nonintentional in §8). This suggests that the intended sense of the question ‘Why?’ can fail to have application in the following two cases:
(i-a) Cases in which the agent does not have non-observational knowledge of what he is doing.
(ii) Cases in which the agent does have non-observational knowledge that he is ing, but lacks non-observational knowledge of the fact that such and such is the cause of his ing.
We can simplify these two conditions by putting them as follows: the question ‘Why?’ fails to have application to an agent’s ing iff either of the following two conditions are met: (a) the agent lacks non-observational knowledge of the fact that he is ing, or (b) the agent lacks non-observational knowledge of the fact that such and such is the cause of his ing.
Given the claim about intentional action with which we began, namely
|A acts intentionally||
a certain sense of the question ‘Why?’ applies to A’s action
this seems to give us the following account of intentional action in terms of non-observational knowledge:
|A s intentionally||
(i) A knows non-observationally that he is ing, and
One important clarificatory question about this account is, What exactly is non-observational knowledge? The closest thing to a definition of this notion which Anscombe gives us is in §8:
“...a man usually knows the position of his limbs without observation. It is without observation, because nothing shews him the position of his limbs; it is not as if he were going by a tingle in his knee, which is the sign that it is bent and not straight. Where we can speak of separately describable sensations, having which is in some sense our criterion for saying something, then we can speak of observing that thing ...”
There is also an important question about the status of this account. Let’s suppose that Anscombe is right, and that there is a necessary connection between intentional action and non-observational knowledge of the kind outlined in this principle. Then there are two views that we could take of the status of this necessary truth.
On one view, this principle could give an account of the nature of intentional action: of what it is to act intentionally. If this were right, then the fact that we have non-observational knowledge of certain actions and their causes should be prior to, and explanatory of, the fact that those actions are the intentional ones. To me, this seems intuitively wrong. It seems that the order of explanation runs in the opposite direction; if we ask how I can have non-observational knowledge that I am walking to the refrigerator, it seems that the answer will appeal to the fact that I am doing so intentionally. The character of my knowledge of the action seems to be explained by the fact that the action was intentional, rather than the other way around.
A second view of this principle is that by stating a necessary truth about intentional action, it provides some constraints on any proposed account of what it is to act intentionally. Plausibly, an account of the nature of some property should aspire to explain necessary truths involving that property. If so, we should expect a good account of the nature of intentional action to explain the fact that, necessarily, if an agent acts intentionally she has certain kinds of non-observational knowledge. But, on this interpretation, Anscombe’s link between action and the two kinds of non-observational knowledge does not itself amount to an account of what it is to act intentionally.
But this only makes sense, of course, if these are genuinely necessary truths. But there are a number of counterexamples to each which call them into question:
Cases in which an agent intentionally s without knowing that he is ing
1. A professor says, ‘That is the stupidest comment I have ever heard’ to a student in the knowledge that he is saying this. It is true of the professor that he has said something rude; but it may not be true that he knows that he has said something rude. But it still seems true that he intentionally said something rude.
Reply. It is not obvious that the professor has intentionally said something rude. Part of the reason why we want to count his doing so as intentional is that we want to hold him morally responsible for his action. But there may be other ways to accommodate this. Maybe we want to hold people morally responsible in some cases not only for effects which they intentionally bring about, but also for effects which they do not intentionally bring about, but should have known would be effects of their actions but did not. Indeed, it seems intuitively like the professor in the imagined scenario is culpable for different reasons than the professor who is intentionally rude to a student.
2. Someone tries very hard to make ten carbon copies at once, but does not know whether he has succeeded until he checks at the end of his attempt. He makes the ten carbon copies intentionally, but does not have non-observational knowledge that he is doing this.
Reply. This demands a revision of the first of Anscombe’s necessary truths about intentional action. Setiya suggests the following:1
Necessarily, if an agent s intentionally, then there is something that the agent does in the belief that he is doing it.
This seems fine, and is not threatened by either of the above examples. But it is worth noting that this is a successful reformulation only if we give the second of the above interpretations of the importance of Anscombe’s principles. If we are looking around for necessary truths which a theory of intentional action should explain, then Setiya’s weakening of the principle is fine. But if, as Anscombe seemed to be, we are looking for criteria to demarcate the class of intentional actions, then the weakened version of the principle is useless. This is because sometimes we might, e.g., intentionally flip the light switch, thereby alerting the burglar, without alerting the burglar intentionally.
Cases in which an agent intentionally s without knowing why he does so
Consider a case of self-deception, in which I have false beliefs about why I perform a certain action. There is no reason to assume that in such cases I know why I am doing what I do.
I don’t see a way around this. This seems to be enough to show that Anscombe’s way of removing the notion of involuntary action from the definition of intentional action fails.
Cases in which an agent has non-observational knowledge that he is ing
and why he is ing, but does not intentionally
The other direction of the definition of intentional action also fails: I can with the non-observational knowledge that I am ing and that I am ing because p without ing intentionally. Example: falling down when drunk or dizzy. In such cases I might have non-observational (proprioceptive) knowledge that I am falling, and non-observational knowledge that I am falling because I am dizzy; I might, so to speak, feel the dizziness causing the fall.
These three classes of problems for Anscombe’s demarcation of the class of intentional actions seem to show that it is a failure. The main merit of the discussion, if this is right, is her drawing our attention to the fact that it is a necessary truth that if I do something intentionally, then there is something that I am doing in the belief that I am doing it.
Following the discussion of non-observational knowledge, Anscombe moves to a discussion of the distinction between the causes of one’s actions and one’s reasons for acting. She is already committed to their being a distinction here, on the basis of her distinction between three senses of the question ‘Why?’
But though Anscombe begins by talking as though her topic is the relationship between reasons and causes, she is in fact interested only in a very specific kind of cause, which she calls ‘mental causes.’ A mental cause seems, roughly, to be something in an agent’s ‘stream of consciousness’ which leads her to act in a certain way. So a mental cause might be one of two things: a mental state of which the agent is aware (a desire, an urge, a mental image, etc.) or some external event which the agent perceives (a knock on the door, etc.). Anscombe seems to argue for each of the following two claims: (i) some mental causes of actions are not reasons for action, and (ii) some reasons for action are not mental causes of action.
A case to illustrate (i) might be a mental image of a book reminding someone that they’ve accidentally left something at home. The mental image might be a mental cause of the person turning around to go home, but the mental image is not a reason for the person to go home.
Anscombe illustrates (ii) by noting (§11) that some reasons for action might never become conscious, and hence might not be capable of being mental causes (according to her definition of a mental cause). E.g., a desire for apples might make one get up and get some, but this desire may not manifest itself as any particular conscious event which made one pursue this action.
It is worth noting that Anscombe does not seem to argue for the claim that no reason for action is a mental cause. On the contrary, she draws a distinction (§10) between the objects and mental causes of emotions which is supposed to be parallel to the distinction between reasons for and mental causes of actions, and notes that in some cases the object of a feeling is the same as its mental cause. So it would not be surprising if she thought that some reasons for action were also mental causes. (The first paragraph of §14 seems to imply that she does think this.)
It is also worth noting, in light of subsequent developments in the philosophy of action, that Anscombe in these passages does not endorse the thesis often attributed to her, that reasons are not causes of action. Indeed, she does not even endorse the weak thesis that there are some reasons which are not causes of action. All she claims is that some reasons are not mental causes, which is both pretty clearly true and weaker than the claim that some reasons are not causes (since not all causes are mental causes). The text leaves open whether mental states can be causes without being mental causes in her sense.
Anscombe began by discussing intentions to act, and moved from there to a discussion of intentional action. At this point in the text she shifts to discussion of a third ‘kind’ of intention: intentions in action.
The intuitive distinction between intentions to do things and intentions in acting. I intend to write this handout; my intention in writing this handout is to clarify some things in Anscombe’s book. My intention in doing something is closely related to my motive in doing it.
Recall that one of the aims of this book, emphasized since §1, is to get a handle on the relationship between different kinds of intention: intentions to act, intentions in acting, and intentional action. One important question about motive, or intentions in acting, is: what is the relationship between intentions in acting and intentional actions?
Anscombe notes, rightly, that intentions in acting provide a kind of explanation of the action (§12). But she seems to deny that this is a matter of determining, or causing, the actions in question. Rather, she says, knowledge of motives allows us to ‘interpret’ the agent’s actions. But what ‘interpret’ means here is left pretty much open.
A natural thought is that to know an agent’s motives, or intentions in acting, is to know his reasons for action. If so, then Anscombe’s point here seems to indicate that she thinks that reasons cannot be causes; but it is unclear as to whether she again has in mind her restrictive notion of a mental cause. The waters are further muddied by §15, in which Anscombe seems to say that the distinction between reason and cause does not have a point in many cases, where the decision whether to call something a reason or a cause for action seems arbitrary. One plausible interpretation of what she is saying is that there are some clear cases in which we want to say that something is a reason rather than a (mere) cause; and some clear cases in which we want to say that something is a cause rather than a reason; but that sometimes the same thing can be a cause and a reason. This emphasizes that the question whether something is a reason for action is different than the question whether it is a cause, and leaves open whether something can be a reason without being a cause in any sense.
It is also unclear why, if it is true that a reason for acting just is a motive, i.e. an intention in acting, Anscombe does not say so.
Anscombe gives several different versions of the criterion for an act’s being intentional. (§16)
The first is the one mentioned above:
|An act is intentional||
the agent can know that he is performing that act, and why he is performing it, without observation
The second is stated in terms of the special sense of ‘Why?’:
|An act is intentional||
the question ‘Why?’, when it is not a request for evidence or a cause, has application
The third gives a positive rather than a negative characterization of the criteria for this question having application:
|An act is intentional||
the question ‘Why?’ could be answered in any of the
following three ways:
One still might wonder: what is it for a question to have application in a particular case? Anscombe makes clear (§17) that for a question to have application in a case is not for it to have a substantive answer. The question ‘Why?’ may have application to an action even if the response is ‘No reason.’ So Anscombe thinks that actions can be done for no reason and yet be intentional.
The points in §§16-18 are mostly a summary of what has come before. But in §19 Anscombe makes a surprising claim: “an action is not called ‘intentional’ in virtue of any extra feature which exists when it is performed”. Rather, “to call it intentional is to assign it to the class of intentional actions and so to indicate that we should consider the question ‘Why?’ relevant to it.”
It is not easy to know how to interpret this. Of course, to call an action ‘intentional’ is to assign it to the class of intentional actions; but this hardly tells us anything. To call an animal ‘canine’ is to assign it to the class of canines.
I think that the best interpretation of what Anscombe is saying here is to that there are no facts of a certain sort which make certain actions intentional. If this is right, then the claim is about the metaphysical status of intentional actions: about the facts in virtue of which actions are intentional. The question then is: what sort of facts is she opposing here?
The best way to tell this is by looking at her argument in §19 for the conclusion that there are no ‘extra features’ accompanying acts which determine that they are intentional. One interpretation of this argument is as follows:
(i) Suppose (for reductio) that an agent’s bodily movement is an intentional action in virtue of some feature of the agent, present at the time of acting. Call this intrinsic feature ‘I.’ (ii) To say that an action is intentional is to say that there is some description under which it is intentional. So for I to determine that a bodily movement is intentional is for I to determine that that bodily movement satisfy some description. (iii) But intrinsic properties of agents at the time of actions are connected only contingently (it is a ‘mere happy accident’) with the satisfaction of descriptions under which the act is intentional. Hence (iv) these properties cannot be what makes it true that a given bodily movement is intentional.
This is a dense and difficult argument. Two critical points to focus on: (1) Is is true that features of agents are connected only contingently with the (truth of) descriptions under which their acts are intentional? (2) Even if this is true, does it follow that these features cannot be what makes true the description of the act as intentional?
The answer to (1) is, I think, just that it depends which kinds of features on has in mind. It seems clear that Anscombe in this argument is focusing on intrinsic, physical properties of the agent in question. If this is what she has in mind, then it seems right that there is only a contingent connection between these properties and the performance of any given intentional action. Further, (2), if we are looking for an account of what it is to perform a given intentional action, the contingency of this connection seems to be enough to undercut such features serving as the basis for an account of intentional action. (But the argument has limited scope if we restrict the relevant ‘extra features’ to intrinsic properties.)
Suppose that a bodily movement is an intentional act of waving iff it is caused in the right way by the intention to greet someone. Let this intention be the feature I. Can this intention (plus the bodily movement) then have a necessary connection to the performance of the intentional action of waving (or, equivalently, the satisfaction of the intentional act description ‘waving’)?
So far we have some marks of intentional action, some points about intentions to act and intentions in acting, and some points about the relationship between these latter two and intentional action. So far, it seems that Anscombe is treating intentional action as the most fundamental member of this trio; e.g. she left aside the discussion of the distinction between prediction and intention, saying that this was best handled by a discussion of intentional action. This might lead us to think that intentional action is independent of the other two: that we should be able to give a constitutive account of intentional action without saying anything about intentions to act or intentions in acting. §20 is an argument that this is not the case.
The argument proceeds by asking: can we conceive of intentional actions in the absence of intentions to act or intentions in acting? Anscombe presents two arguments that we cannot.
First, we can ask what kind of concept the concept of an intentional action would be in this case. Perhaps, she says, “an action’s being intentional is rather like a facial expression’s being sad”, in the sense that it would be a kind of style of acting, just as sadness is a style of facial expression. That this is a false model of intentional action is shown by the fact that actions can be intentional under some descriptions but not others, whereas a facial expression cannot be sad under one description but not under another.
Second (and more important), if we suppose that there are no intentions in acting, then we have to suppose that the only answer to our special question ‘Why?’ is ‘I just am.’ But this means that we would have no way of distinguishing between intentional actions and those involuntary actions -- such as gasping -- which are known without observation. The distinction between the two was that in the case of the former but not the latter, the agent will know without observation not only what he is doing, but why he is doing it.
(An objection to this line of argument: Anscombe has claimed that there can be intentional actions accompanied by no intention in action, actions for which the answer ‘I just am’ or ‘For no reason’ to the ‘Why?’-question really is appropriate. This is enough to show that someone can perform an intentional action while having no intention in acting. Can we not imagine a possible world in which this is the only intentional action performed? Maybe in such a world no one would have the concept of intentional action, or be able to give criteria for intentional actions; but this is not the same thing as there being no intentional actions. Perhaps there is something incoherent about imagining such a possible world; but it seems that Anscombe needs an argument for this which she does not provide. If such a world is coherent, then Anscombe’s argument fails.)
In §21, Anscombe notes that this dependence of intentional action on intentions in action vindicates a weaker version of the classical thought that ‘human beings must always act with some end in view.’ Though as it stands this appears to be false, we can say that if people did not sometimes act with some end in view - with some intention in acting - then we would not have the concept of intentional action. (But, again one wants to ask, how does this fact relate to the question of whether there would be intentional actions?)
In §22, we get an extended discussion of intentions in action, expressions of which are answers to the question ‘Why?’ which either mention a broader description of an agent’s present activity, or mention a future state of affairs which the agent’s present activity can be thought of as leading toward.
Beginning in §23, Anscombe returns to the topic of intentional action. One theme which has been present throughout Anscombe’s discussion has been the topic of actions ‘under a description’ - she has claimed, e.g., that actions can be intentional under one description but not under another. But this raises at least three questions: (i) When is a description of an action ‘intentional’? (ii) What is the relationship between different descriptions of an action? (iii) When do we have several actions, and when do we have one action described in several ways? Anscombe discusses (i) and (ii) in §23, and (iii) in §§24-26.
(This talk of an action under several descriptions is intended to be analogous to the sense in which a single physical object might fall under several descriptions. My desk might be described as ‘my desk’, ‘the only brown thing in the room’, ’the thing on which my computer sits’, etc. But we only have one object here, described in a variety of ways.)
Suppose that I am typing on my keyboard. Then it might be that all of the following descriptions apply to my action: ‘he is typing’, ‘he is preparing a handout for class’, ‘he is making his computer crash’ (suppose that in a few more keystrokes the computer will crash). All of these sentences might be true, and in that sense they might all be descriptions of my action. But we can ask: under which of these descriptions is my action intentional?
We already have two (equivalent) ways of answering this question. We can say that a description of my action is intentional if I have non-observational knowledge that it applies (along with non-observational knowledge why I am doing such-and-such), or if the question ‘Why?’ in its special sense applies to my action, so described. Both these criteria yield the same answer: my action is intentional under the first two descriptions, but not under the third. The question ‘Why are you making your computer crash?’ is refused application by my response, ‘I didn’t know that I was.’ (But see above for a discussion of the the problems with Anscombe’s way of demarcating the class of intentional actions.)
Anscombe also outlines a way in which different descriptions of an action might be related. Suppose we ask for someone’s intention in doing something; e.g., I am asked why I am typing. I might reply: ‘In order to prepare a handout.’ Then, usually, it will follow that the intention in doing also describes an intentional action: the action of preparing a handout. In such cases we might say that the actions are related in the sense that I do one action by doing the other, or that I do the latter in order to do the other.
Anscombe claims (§26) that in such cases we do not have several actions, but one action under a series of descriptions. In our example, I perform one action, which may be described either as ‘my typing’ or as ‘my preparing a handout’. In such cases we can either say that I have one intention in doing thee act, or several, corresponding to the several descriptions. It sounds natural to say that my intention in typing is to prepare a handout, and that my intention in preparing a handout is to be ready for seminar; but we could also say that my intention in each was to be ready for seminar.
Anscombe suggests that when we have actions related in this way, the application of further description may be explained by circumstances external to the action: in this situation, my typing also counts as my preparing a handout. It is difficult to know here what is supposed to fit into the category of ‘circumstances.’
In a sense, it seems, we have yet to answer the question, ‘What is it for an agent to act intentionally?’ We have given various identifying marks of intentional action; but as noted above, it does not seem right that these marks tells us what intentional action is. Rather, it seems, these provide necessary truths which an account of intentional action should explain.
One view about intentional action is that for an agent to act intentionally is for his action to be produced (caused) by an ‘inner act of the will’ or ‘inner act of intention.’ In this way, intentional action might have two factors: the external bodily movement, and the inner act of intention which produces it.
This is a tempting picture for at least two reasons. (i) We seem to be at a loss for what else might explain the difference between, e.g., my raising my arm and my arm going up. (ii) As Anscombe notes (§29) it seems impossible that we could have non-observational knowledge of some empirical matter of fact, such as the wall’s being painted yellow; so if it is true, as Anscombe has claimed, that we have non-observational knowledge, this non-observational knowledge must not be of any external act, but rather of an interior act of intention which (perhaps) produces such an act. As she puts it, “If there are two ways of knowing there must be two different things known.” (§29)
But Anscombe thinks that this ‘two-factor’ picture of action is misguided. To show this she does two things: (i) argues that any way of developing this picture goes wrong, and (ii) tries to reconcile the falsity of the two-factor theory with the reality of non-observational knowledge. (ii) is really only carried out in the long discussion of practical knowledge and practical reasoning in §§32-45. in §§27-31 she focuses on the negative task. This takes the form of consideration of a series of proposals on how to understand the putative interior acts of intention.
Acts of intention as conscious episodes
One idea is that acts of intentions are conscious episodes, such as saying to myself, ‘I hereby intend to be honest.’ Anscombe gives a Wittgensteinian reply to this move (§27), claiming that we cannot identify acts of intention with these kinds of conscious events, since we can always imagine the conscious event happening without the intention. I might, e.g., say this to myself as a joke.
Acts of intention as non-physical acts of willing
Perhaps, as the early Wittgenstein seemed to think (§29) what we have knowledge of are really just acts of willing which are only contingently connected with any events in the physical world at all. Anscombe is skeptical about these non-physical acts of willing; she asks rhetorically “where is [my intention] to be found? ...what is its vehicle?”
Acts of intention as muscle contractions
A third idea (§30) is that the acts of willing of which we are aware are not mysterious non-physical events, but rather physical events which ‘set things rolling.’ The obvious problem with this is that oftentimes we know what action we are performing, without knowing hardly anything about the physical action by which we accomplish it. Consideration of most skilled activities - shooting a basketball, riding a bike, swimming the butterfly - shows this. We can know what we are doing without having much of an idea how we do it.
A fourth possibility which (for understandable historical reasons) Anscombe does not consider is that acts of intending are events of agents coming to be in certain brain states which, by virtue of their relations to external events and other brain states, qualify as intentions with a certain content. This is a view taken by many contemporary philosophers. There is a substantial body of literature devoted to the question of how, on this conception of mental states, we could have immediate knowledge of what we intend. See especially Sydney Shoemaker, ‘Self-Knowledge and “Inner Sense”: Lecture III: The Phenomenal Character of Experience’, Philosophy and Phenomenological Research, 54:2 (1994).
So much, Anscombe thinks, for attempts to explain non-observational knowledge by claiming that it is knowledge only of a private domain of interior acts of intention. How can we explain this non-observational knowledge, then? Anscombe thinks that the answer is to be found by investigating an under-explored area of ancient and medieval philosophy: practical knowledge, and practical reasoning. She takes this up beginning in §32.