How can the difference between intentional actions and mere bodily movements be captured? Many philosophers endorse a version of the causal theory of action (CTA), which, simply put, holds that intentional actions are caused in the right way by psychological states such as belief-desire pairs or intentions, while mere bodily movements are not (e.g., Davidson 1963; Mele 1992). In recent years philosophers have criticized CTA by claiming that certain classes of actions do not fit this picture: Some of our actions are not caused by belief-desire pairs or intentions (e.g., Di Nucci 2011; Pollard 2006; Zhu 2004). Zhu argues that so-called passive actions (Mele 1997) are an example. Passive actions are spontaneous responses to external or other internal factors such as emotions (Zhu 2004, pp. 300, 302). Examples are spontaneously getting up and pacing around the room or cheering when your favorite sports team has won. In some sense agents are passive when they perform these actions, because their bodily movements are not intentionally initiated. But even so, Zhu claims, the agent is in control of what she is doing and she is acting. Similarly, Pollard and Di Nucci maintain that we often act mindlessly: We do not always think, reflect, or pay much attention to what we are doing when we act. We do not need to think or reflect when we brush our teeth, get dressed, or walk to the office (Di Nucci 2011, p. 179; Pollard 2006, p. 57). Often we act automatically or out of habit, and we do not consciously or unconsciously represent our goals (Di Nucci 2013, p. 48). If this is the correct depiction of passive, automatic, and habitual actions they clearly form a challenge for CTA, because these critics suggest that by definition these actions do not have the causal history that would be required for CTA to count them as actions.

Other philosophers have argued that CTA can account for these kinds of actions (see, e.g., Clarke 2010; Mele 1997), but that is not what I will be concerned with in this paper. Rather, my aim is to evaluate the alternative action theory these critics have proposed. Di Nucci (2011), Pollard (2006) and Zhu (2004) use these kinds of actions to argue that a version of  the guidance theory, which can be traced back to Harry Frankfurt (1978), is superior to CTA. According to Frankfurt actions can be distinguished from mere bodily movements because agents have guidance control over their actions. Actions are not necessarily intentionally initiated, but agents are able to intervene on what they are doing if necessary. The aim of Frankfurt and philosophers that defend a similar theory is to account for actions without relying on a causal role for psychological states.

The guidance theory of action has not been critically assessed. With this paper I aim to fill this gap. I will start out with Frankfurt’s proposal for a guidance theory of action in Sect. 1. One of his reasons to argue for a guidance theory seems to be that actions and mere bodily movements differ in themselves; the difference between the two cannot be captured on the basis of their distinct causal histories. Since Frankfurt does not develop his guidance theory in much detail in his paper, I move on to discuss recent versions of the theory in Sect. 2. I argue that in its basic form the guidance theory is too weak: it includes certain bodily movements that someone else makes as actions. I evaluate three conditions that have been added to the basic version of the guidance theory to examine whether this counterexample can be excluded without appealing to causally efficacious psychological states. I conclude that all extended versions of the guidance theory fail to some extent and introduce problems of their own. Therefore, in Sect. 3, I return to Frankfurt’s claim that mere bodily movements and intentional actions differ in themselves. I argue that Anscombe’s (1976) perspective on action offers a way to spell out this inherent difference. Because of this, Anscombean action theory offers an elegant way to account for passive, habitual, and automatic actions. Section 4 concludes the paper.

1 Frankfurt’s Criticism of CTA

In his ‘The Problem of Action’ Frankfurt (1978) criticizes Davidson’s (1963) version of CTA. Davidson argues that intentional actions differ from mere bodily movements because of their causal history: Intentional actions are caused in the right way by reasons, which he takes to be belief-desire pairs. Frankfurt argues that by focusing on something that is prior to the action Davidson diverts attention away from the action itself (p. 157). As a result, the action itself remains ill understood. Furthermore, it is implied that actions and mere bodily movements are in themselves indistinguishable, which is false according to Frankfurt (p. 157). He suggests that we should pay attention to the action itself; that is where the inherent difference between actions and mere bodily movements can be found.

Frankfurt’s alternative proposal is that during the performance of an action the person is in touch with the movements of their body in a certain way (Frankfurt 1978, p. 158). This ‘in touch’ should not be understood as an actual causal relationship, as in CTA, but rather as the possibility to adjust the bodily movements if necessary. Frankfurt claims that the bodily movement is under the agent’s guidance, and is therefore an action, if she can be identified with the causal mechanism that stands ready to affect the course of the movement (pp. 159, 160). Whether adjustments actually need to be made is irrelevant. What matters is that this mechanism would be causally effective were the course of the movement to be jeopardized.

An important question is whether Frankfurt is taking his own criticism of CTA to heart. As I see it Frankfurt makes two important points in his paper: (1) actions and mere bodily movements differ in themselves, and (2) actions are characterized in terms of the agent being prepared to intervene if necessary and in a position to do so more or less effectively. This suggests that he proposes to capture the inherent difference between bodily movements and actions in terms of guidance control. But if the agent is merely in touch with a mechanism that is standing ready to intervene, the distinction between action and mere bodily movements seems to be characterized, like in CTA, in terms of something distinct from the action itself. Even though in Frankfurt’s proposal the action is not understood in terms of something prior to the action as in Davidson’s version of CTA, it seems that the difference between mere bodily movement and action cannot be found ‘in’ the action itself either. The mechanism that distinguishes between mere bodily movements and actions is merely monitoring whether an intervention is called for, and only steps in if necessary. But even without stepping in the bodily movements count as an action. If that is the correct depiction, Frankfurt’s version of the guidance theory seems to suffer from the same objection he formulated in response to Davidson. If the guidance theory does capture the distinction between mere bodily movements and actions this may not a problem. But since Frankfurt does not develop his version of the theory in much detail it is difficult to evaluate it. It remains unclear what he precisely means by the agent being in touch with the bodily movements, how ‘intervene if necessary’ should be spelled out, and how we should make sense of the mechanism that the agent can be identified with.Footnote 1 Because of that in Sect. 2 I will examine more recent version of the guidance theory.

2 Evaluating Guidance Theories

As far as I know Zhu (2004) was the first to argue that certain types of actions give us reason to reject CTA in favor of a guidance theory of action. He claims that even though passive actions, for example cheering when your favorite sports team has won, are not initiated intentionally and the agent never actually has to exercise control, these bodily movements still count as actions. That is because they do take place under the agent’s control: She is ‘in the position to monitor the ongoing process, and is prepared to effectively intervene if necessary’ (p. 300).

One question that arises is what it means to effectively intervene. Is it sufficient that an agent is able to slightly change the course of the movements? It seems to me that merely being able to cheer with one hand instead of two does not suffice. Such an intervention does not seem to provide the agent with control over her cheering. But since Zhu does not spell out what he means by the phrase ‘effective intervention’ it remains unclear what to make of this. Pollard (2006) also states that if an agent has intervention control ‘[she] can stop [herself] from exercising a given habit, or modify it to a greater or lesser extent’ (p. 59). But, again, what kind or amount of modification is required for the agent to be in control? Later in the paper, however, he states that an agent should be able to ‘do something else, or nothing at all’ (p. 60). This suggests that mere modification is insufficient, and that the agent should be able to cease acting. Di Nucci (2008) holds the same view: Being able to intervene should be understood as being able to stop what you are doing. In his words: Even though we can avoid blinking for a while, ‘we can’t avoid blinking for good, and we can’t always avoid blinking; the same way in which we cannot avoid breathing for good (assuming that killing ourselves does not count as a way of controlling our breathing patterns)’ (Di Nucci 2008, p. 26). It is worth pointing out that both Di Nucci (2008, p. 26) and Pollard (2006) emphasize that the agent should be able to intervene directly. This way they can exclude bodily movements over which we have indirect control, for example knee jerks, as actions.

One problem with this proposal is that it excludes bodily movements that we do think of as actions. If someone would ask me to breathe in and out every 10 s, or if I have reason to slow down my breathing, I can do it. Even though I cannot completely stop breathing that does not exclude breathing from the realm of action altogether. This problem, however, is easy to solve. Guidance theorists could simply claim that what matters is whether we have guidance control over the bodily movements under a certain description. They could claim that breathing in and out every ten seconds is an action if the agent can stop doing that, even though she cannot stop breathing altogether. But there is a different counterexample that cannot be dealt with that easily. If bodily movement plus guidance control is all there is for something to count as an action, then also someone else moving my body could count as my action. Consider a case in which a physiotherapist moves my arm up and down. If her grip is not too tight I am able to stop the movement of my arm if necessary. The two conditions for guidance control are met, but someone else moving my arm up and down is a typical example of a bodily movement that is not my action. That implies that in its basic form the guidance theory does not capture the difference between mere bodily movements and actions. The theory needs to be amended to exclude this case. Even though Zhu (2004), Pollard (2006), and Di Nucci (2011) do not discuss a counterexample like this one they all have proposed to extend the theory. In the remainder of this section I will evaluate these additions. The crucial question is whether the counterexample can be excluded without introducing psychological states, i.e., without giving up guidance theory’s unique character.

2.1 Causally Ineffective Intentions

Zhu (2004, pp. 299–300) distinguishes three features of passive actions. Two of those capture the passivity of the action: the action is not initiated intentionally and the agent never actually intervenes. The other feature should capture why it is an action, which is that the bodily movements take place ‘under the agent’s guidance and control’ (p. 300). The result is that Zhu’s proposal is vulnerable to the first counterexample: On the basis of his description of passive actions he cannot make a distinction between a passive action and a physiotherapist making my body move. Later in the paper, however, Zhu does suggest that the agent needs to have an intention for the bodily movements to count as an intentional action, and that the relationship between the intention and intentional action is conceptual instead of causal. He argues that as soon as the agent abandons the intention the intentional action ceases to exist, even if the bodily movements are still taking place (p. 306). For example, an agent has to intend to play a nine-iron shot for the bodily movements to count as the intentional action ‘playing a nine-iron shot’. Thus, we could interpret Zhu’s proposal as follows: For the bodily movements of an agent to be an intentional action, she has to be able and be in a position to intervene and she has to have the intention to perform that action.

But this does not suffice as an antidote to the counterexample. Even if I have an intention to move my arm up and down, as long as the physiotherapist is in fact making my body move there is no good reason to think that it is suddenly something I am doing (cf. Mele 2003, p. 48). What is more, if the agent does need to have an intention, this seems to undermine Zhu’s project. His aim is to give an account of actions that we perform in direct response to an internal or external factor. But if it turns out that we do need to have an intention for it to count as an action, it is no longer clear what the advantage of a guidance theory would be over more recent versions of CTA in which there is room for intentions in action that do not have to be actively formed by the agent (e.g., Searle 1983, Chap. 3). It might be the case that Zhu distinguishes between passive actions and intentional actions, and that he does not think that passive actions are accompanied by an intention. But if that is the view then he still has to find a way to distinguish between the movements that a physiotherapist makes and passive actions.

Perhaps a solution can be found in the work of Pacherie (2014). She uses research on subpersonal action control to argue that we misunderstand the role conscious intentions play. Intentions are often not the initiators of our actions but play a causal role only if necessary, if intervention on this higher level is called for. Most of the time subpersonal motor processes, i.e., inverse and forward models, can take care of the successful execution of the action (Pacherie 2014, p. 38). This way ongoing actions can be corrected and adjusted without a conscious intention playing a causal role. The implication is that on a subpersonal level the action is caused, but personal-level psychological states intervene only if necessary. On such a view the movements the physiotherapist makes would no longer count as an action because my motor processes are not involved, and at the same time the spirit of the guidance theory is saved: We can act without psychological states playing a causal role.

Pacherie (2014) uses an experiment conducted by Fourneret and Jeannerod (1998) to support her proposal. In this experiment subjects had to move a stylus pen on a graphic tablet along a straight line towards a visual target. They could only see the trajectory on a computer screen, not their moving hand. On some trials, small biases were introduced because of which the subjects did not actually move their stylus pen on a straight line, but the line was (slightly) curved. This was not clear from what they saw on the screen—the line still seemed straight—but they of course did have to adjust their hand movements accordingly. If this bias was small, subjects were not aware of the adjustments they were making with their hand and of the fact that there was a discrepancy between their perceptions and their actual movements. It was only with larger biases that subjects became aware of a discrepancy and that they began to use conscious correction strategies to compensate for the bias to reach the target. Only then did subjects report moving their hand (Pacherie 2014, p. 39). These findings suggest that conscious psychological states play a causal role only if necessary, which would imply that the guidance theory is on the right track.

But even though the subjects were not aware of making small adjustments, that does not imply that they were merely observing their hand making the right movements and that the conscious intention was only standing ready to intervene. It may well be the case that they were consciously moving the stylus pen to a visual target. What is more, the adjustments only make sense in light of the intention to move the pen to the visual target. It is this intention that entails that ‘all of the appropriate physical movements fall into place without [the agent] willing them to do so’ as Gallagher nicely puts it (2006, p. 122). Without an ongoing influence of the conscious intention, within this theoretical framework, it seems that no adjustments would be taking place at all. That suggests that a defender of the guidance theory is stuck between a rock and a hard place: either she argues that intentions do not cause actions and is confronted with the example of the physiotherapist, or she ends up with an account in which the intention is causally efficacious and the theory no longer is a guidance theory but a version of CTA.

The defender of the guidance theory could of course argue that this is not a good example and that some actions really are caused only by subpersonal processes, for example the passive actions that Zhu (2004) is interested in.Footnote 2 But this is not unproblematic. For one thing, it introduces the notorious problem of causal deviant chains that CTA suffers from and that guidance theorists wish to avoid (Di Nucci 2011, p. 195; Frankfurt 1978, p. 157; Zhu 2004, p. 302). But there is another issue. One option is that these subpersonal motor processes really are subpersonal in the sense that the agent cannot access their content (see, e.g., Clarke 2010; Valaris 2014 for a discussion). But then on what grounds do we draw the conclusion that the person is acting? The answer that it is caused by subpersonal motor processes is not convincing, because these processes are theoretical notions invoked to explain how agents act. The first step is to establish that what needs to be explained is an action, and then motor processes are invoked to explain it. In other words, it is not the motor processes that define what the agent does as an action; that they play a role in a theory of action explanation is the reason that motor processes are taken to explain action. The other option is to argue that their content is accessible, but then it is unclear why they do not count as unconscious psychological states that cause actions. Taking this route would mean giving up on the defining feature of the guidance theory.

2.2 Knowledge

Di Nucci (2011) argues that the guidance theory needs an additional condition to account for the fact that we act intentionally under certain descriptions while under other descriptions what we do is unintentional. His proposal is that ‘E ɸ-s intentionally iff ɸ-ing is under E’s guidance, where ɸ-ing is under E’s guidance only if the agent can reasonably be expected to know or find out about her movements under the relevant description, ɸ-ing’ (p. 196). This addition does not help to exclude the counterexample, however. I can be expected to know or find out that the physiotherapist is moving my arm up and down (most likely I already know), and her making my body move would therefore count as an intentional action on this view. Again we end up with the wrong result.

There are other reasons to think that this knowledge condition is problematic. Di Nucci’s aim is to account for the fact that we act under certain descriptions but not under others merely in terms of guidance, without introducing psychological states. That is probably the reason why this condition does not state ‘know’, but ‘reasonably expected to know or find out’: It is much harder to defend that an agent currently knows what she is doing without appealing to a psychological state like a belief. And if the agent needs to have such a belief, it would no longer be ‘an account in which guidance is sufficient for intentional action’ (Di Nucci 2011, p. 196). Thus, Di Nucci’s proposal is that a bodily movement counts as an intentional action if the agent can reasonably be expected to know or find out what it is, and she is able to stop doing it directly. Di Nucci (2011, p. 196) gives an example in which he wants to phone Tom but ends up phoning Tim instead, because overnight Tom’s phone number has been assigned to Tim. He dials Tom’s phone number but of course ends up speaking to Tim. Di Nucci argues that phoning Tim was not an intentional action, because he couldn’t have reasonably been expected to know or find out that Tom’s phone number had been reassigned to Tim. Di Nucci claims that what excludes phoning Tim as an intentional action is that he could not be expected to check the number before making the call.

I agree that he phones Tim unintentionally, but the question is whether it is this ‘reasonable expectation’ that is doing the work. Let’s look at a different example, one by Falvey (2000, p. 28). Falvey describes a case in which he intends to make Irish Breakfast tea, but it turns out that he is spooning Darjeeling tea in the teapot instead. Let’s assume that the content is clearly stated on the tins, that he strongly dislikes Darjeeling, and that he often does check whether he took the right tin. In other words, we can reasonable expect him to know or find out that he is putting Darjeeling in the teapot (cf. Di Nucci 2011, p. 196). Furthermore, he has the ability to stop making Darjeeling directly. On the basis of Di Nucci’s conditions, then, we would have to conclude that he was intentionally making Darjeeling tea even before he actually found out or knew that his bodily movements fitted that description. Even though his intention was to make Irish Breakfast, the conclusion would have to be that he was performing a different intentional action. But that is not the right conclusion. As soon as he finds out what he was doing, he would conclude that he was spooning Darjeeling in the pot by mistake (Falvey 2000, p. 28). He would not find out what his intentional action was, but what he was unintentionally doing. Similarly, in Di Nucci’s example it is not the reasonable expectation that is doing the work, but the fact that phoning Tim was not his aim. He made a mistake in phoning Tim, because he wanted to call Tom instead.

A related problem is how in this theory an agent could know or find out what she is doing. Di Nucci should be able to account for the agent’s knowledge, or ability to acquire knowledge, of what she is doing intentionally under a certain description without her psychological states being a source of this knowledge. But then what can her source of this knowledge be? Di Nucci (2008) suggests that this depends on perceptual awareness. By observing that her hand is holding the door handle, the agent can be aware that she is moving the door handle and opening the door. But, as we saw above, observing our bodily movements and the situation we are in might just as well lead us to conclude that we are doing something unintentionally. The person in our example does not conclude when he smells the tea that he was intentionally making Darjeeling. Furthermore, observing your bodily movements and the situation does not always uniquely determine what you are doing, and yet we do know which of the possible actions we are in fact performing. Imagine that a friend is walking on the other side of the street and a taxi is coming that you want to stop. The same bodily movement, your arm and hand going up, might be an intentional action under the description of ‘hailing a cab’ or ‘waiving to a friend’. From the bodily movements and the context alone it cannot be inferred what you were intentionally doing, and yet you know what it is. That we have this kind of knowledge cannot be explained on the basis of Di Nucci’s version of the guidance theory. That suggests that wanting or aiming or intending plays a role in the knowledge we have of our intentional actions as well.

2.3 The Agent’s Nature

Finally, I will examine Pollard’s (2006) version of the guidance theory in more detail. Pollard’s aim is to argue that certain exercises of habits are actions, and that this can be accounted for without any role for psychological states (p. 61). In his view these habitual actions represent ‘a challenge to a prevailing intellectualist philosophy of action which assumes such psychological items must always have an explanatory role’ (pp, 57–58). He does recognize that the ability to intervene is insufficient to secure, in his words, authorship of what the agent does. Because of that, he points out that habits bear a constitutive relationship to agency. When we act out of habit the action is part of our nature: it is part of our personal history, we act without deliberating or a feeling of effort (i.e., automatically), and it is easier for us to do it than not to do it (pp. 58, 60). Pollard argues that by acknowledging that habits are part of our nature authorship comes for free; habitual actions are by definition things that the agent herself does. This, in combination with the ability to stop what you are doing directly, should suffice to pinpoint habitual actions.

It seems that Pollard can exclude the physiotherapist moving my arm as an action. These bodily movements are not part of my nature. But on what grounds can we exactly draw this conclusion? What if I have been visiting my physiotherapist every week for several years? Then it seems that her moving my arm up and down could very well be part of my personal history, take place without deliberation and effort, and be easier to do than not to do. Of course this is not what Pollard means when he talks about my nature and my history. In this example somebody else is doing the moving, and it is not my habit that is doing the work. But strictly speaking Pollard’s description of what it means for something to be part of one’s nature does not exclude this case. We cannot simply presuppose that by using the terms ‘I’ and ‘my’ someone else moving your body is excluded, because we are trying to find out what ‘I’ and ‘my’ mean when it comes to our bodily movements. The question exactly is what it is about actions that they are things that I do, and whether we can make sense of that without a psychological state causing it. The risk of a proposal like Pollard’s is that it sounds right, but that it remains unclear why exactly habitual actions count as my own and, crucially, whether he succeeds in excluding non-actions without appealing to psychological states.

The question indeed is whether he succeeds at doing so. At several points in the paper he does refer to psychological states to distinguish between habitual actions and other habitual behavior. For example, he excludes compulsions and phobias as actions by stating that they ‘are contrary to what the agent herself judges it best to do’ (pp. 59–60) and that they are taken to be ‘inappropriate, often by the agent’s own lights’ (p. 60). Similarly, he claims that the ‘actional status’ of nail biting is in doubt, because ‘[o]ne typically will not have intended to do it, and will have no positive attitude towards it’ (p. 62). Furthermore, in Sect. 3 of the paper he tries to account for the fact that sometimes agents act unintentionally and make mistakes, without invoking intentions. Pollard’s solution is to introduce the notion ‘intrinsic intentionality’: Habitual actions have a certain teleological structure (p. 63). The problem is that it remains unclear how habitual actions come to have this teleological structure. In CTA the intention takes care of that, but what is the alternative that the guidance theory has to offer? And can it be explained without appealing to psychological states? Pollard does not elaborate on this crucial aspect of the notion. We have to conclude that also Pollard has not succeeded in pinpointing the unique character of actions without invoking psychological states.

3 The Inherent Difference: Mind in Action

Zhu (2004) and Di Nucci (2011) take Frankfurt’s criticism that actions should not be analyzed in terms of their causal antecedents to heart. However, these proposals do not take into account the other point that Frankfurt made, namely that actions are inherently different from bodily movements. Although they do not place the defining characteristic of action in its causal history, as the version of CTA that Frankfurt has criticized does, they do not place that characteristic in the action either. That suggests that the guidance theory suffers from the same criticism that laid the ground for its development: We still do not know what the inherent difference between actions and mere bodily movements is. Pollard’s (2006) approach is slightly different, because he does seem to try to capture the inherent difference between actions and mere bodily movements by claiming that habitual actions are part of our nature and have a teleological structure. But he does not spell out these notions, nor does he explain how actions come to have this teleological structure. It cannot be because they are caused by an intention or some other psychological state, because then the basic principle of the guidance theory would be violated.

So, what are the options? One option would be to admit defeat and try to make sense of passive, automatic, and habitual actions within the framework of CTA. But there is another possibility, in which we adhere to Frankfurt’s claim that actions and mere bodily movements differ inherently. This brings me to Anscombe’s (1976) approach to intentional action. She argues that actions cannot be characterized in terms of an external feature, for example in terms of causation by a psychological state. Rather, agents have a distinct kind of knowledge of their intentional actions. They have practical knowledge: direct, non-observational knowledge of what they are doing and why. And it is this practical knowledge that at the same time causes the action: it is ‘the cause of what it understands’ (Anscombe 1976, p. 87). That does not imply that actions are bodily movements plus causation by practical knowledge, in the way proponents of CTA think that actions are bodily movements plus causation by intentions. Practical knowledge does not exist before and is not distinct from the action that it subsequently causes. Rather, it is directly expressed in the action, because of which intentional actions are inherently minded. In Rödl’s (2007) words, an intentional action ‘is a thought that is a movement’ (p. 19).

Practical knowledge is not knowledge of bodily movements; it is knowledge of the means-end structure that the intentional action has. When we act intentionally we do one thing in order to do something else, and because of that we know what we are doing and why. I know that I am crossing the street to go the supermarket, and I know that I am going to the supermarket to buy groceries. And I cross the street, and crossing the street makes sense to me, because of the knowledge of this structure. As Anscombe (1976) puts it, when it comes to intentional action ‘a certain sense of the question “Why?” is given application’ (p. 9). In asking the question ‘Why?’ we move from one intentional action to another (Ford 2015, p. 139). If you ask me why I am crossing the street I have direct knowledge that I do so because I am going to the supermarket. Having practical knowledge means having insight into the means-end structure that is characteristic of action.

Anscombe’s theory has been criticized because of this intimate connection between intentional action and practical knowledge. Her claim is that we do not need perceptual experience to know what we are intentionally doing at all. Philosophers have come up with counterexamples to argue against this crucial part of her theory. They aim to show that we do not always know what we are intentionally doing without observation, for example if we fail to do what we intend to be doing (e.g., Davidson 1978; Paul 2009). Although I cannot defend Anscombe’s theory against such counterexamples in detail, I will point out two things in reply. First, the distinction that Anscombe makes between practical knowledge and contemplative knowledge is worth emphasizing. In the case of practical knowledge, it is the performance’s job to match up with the knowledge, not the other way around (Campbell 2017, p. 5). When the knowledge fails, the mistake is in the performance, not in the judgment of the facts. And second, we should recognize that actions are works-in-progress, and that agents have practical knowledge of the process of reaching a certain end. Thus, even if the agent does not know whether she has or will reach her end successfully or is failing at the moment, she does know what it is she is in the process of doing (Thompson 2011, p. 200).

Much more can be said about Anscombe’s view, but in the remainder of the paper I want to show that on the basis of this view we can make sense of the claim that passive, automatic, and habitual actions do not have a particular causal history and yet are actions. The fact that they do not have the history required in CTA is a reason to take Frankfurt’s insight seriously: it suggests that there is something about the action itself because of which it is an action. Anscombe provides an account of this unique character of actions: When they act agents have practical knowledge of the means-end structure of their action, and it is this knowledge that at the same time causes the action. It is easy to see, then, that Di Nucci’s (2011) and Pollard’s (2006) examples of habitual and automatic actions count as actions: There is no reason to think that agents do not have practical knowledge of walking to the office or getting dressed. They know while they are buttoning their shirt that they are doing so, and that they are doing it in order to get dressed. Because of that, buttoning their shirt is an action, even if the agent did not think, reflect, or form an intention to do it. Furthermore, Anscombe’s theory excludes movements that someone else is making with my body as actions. I do not have practical knowledge of the movements the physiotherapist is making; this is something I infer on the basis of my sensations and observation.

Di Nucci’s (2008) definition of automatic action, however, does suggest that Anscombean action theory falls short. He understands automatic actions as involving a lack of attention of the agent to any aspect—bodily movements, the object of the action, or the plan and goal—of what she is doing. However, this characterisation of automatic actions is so strong that Di Nucci’s own examples do not meet these criteria. There is no reason to think that the agent is not aware of getting dressed. What is crucial, what makes the action automatic, is that the agent does not need to think or deliberate before the action. Also, the agent does not need to think about or reflect on an intention or plan; she only has to know what she is in fact doing. Having practical knowledge and acting automatically or out of habit are perfectly compatible. Agents may not have practical knowledge in all the examples that have been discussed in the literature, however, for example putting on socks in a certain order (Pollard 2006, p. 64). But I do not think it is worrisome if we end up concluding that putting on socks in a certain order is, in most cases, not something an agent intentionally does.

Let’s finally look at Zhu’s examples of passive action. The first example was a person cheering because his favorite sports team has won. One interpretation would be that he is jumping, waving, shouting, and singing in order to cheer, but it seems strange to claim that cheering is his end in doing what he does. His other example was a person suddenly getting up and starting to pace around the room. Also in this case it is not immediately clear that this person is pacing to achieve some end. At first sight, then, these passive actions seem to fall outside of the scope of Anscombe’s approach. But that is not the right conclusion. First, as Thompson (2008, Chap. 8) has emphasized, actions are processes: they consist of phases that form a rational unity. That also holds for relatively simple actions, for example jumping. The agent has to bend her knees and swing her arms for example, and these phases have a means-end structure. The same can be said about the other example. The person is putting one foot in front of the other in order to pace around the room. Also passive actions have a rational structure. Secondly, for Anscombe the issue is whether the question ‘Why?’ has application. Anscombe allows that sometimes there is a null-answer to the question (Ford 2015, p. 133). Even if the person says that she is pacing ‘for no particular reason,’ that does not mean that she lacks practical knowledge of the fact that she is pacing.Footnote 3

4 Conclusion

In this paper I critically reviewed recent versions of the guidance theory. I argued that all versions are problematic, because their proponents have not succeeded in excluding as actions certain bodily movements that are made by someone else without appealing to psychological states as causes. In order to account for passive, automatic, and habitual actions I changed my focus to Frankfurt’s other claim: that intentional actions and mere bodily movements are inherently different. I argued that this inherent difference is captured by Anscombe’s approach to action. Anscombe provides an account of the unique character of actions, and because of that is also able to account for those actions that do not have the causal history that CTA requires them to have.