Philosophers of action very often start with the question: what happens when someone acts? I am going to follow suit. However, I am going to go on to urge that there is a certain kind of answer to that question that is often expected, but that we cannot have, and that we do not need.Footnote ²

Suppose – to start with an unusually simple action – I raise my arm. Two questions immediately arise:

The question that I call (NEC) is a question about what is necessary for me to raise my arm. The question that I call (SUFF) is a question about what, other than that I raise my arm, is sufficient for it to be true that I raise my arm.

Let us set about answering the (NEC) question: what is needed for me to raise my arm? Obviously my material world needs to exist, and the space I move it in, needs to exist. Also my parents needed to have met for me to now be raising my arm. These background conditions are conditions on my acting in virtue of being conditions of more central conditions. What are the more central conditions? I think we can identify four:

It seems unarguable that the Agent Condition is true: we can agree with Gassendi that ‘it is known by the natural light that whatever acts exists’.Footnote ⁴ If I act, I, therefore, exist.Footnote ⁵ I am going to assume for the purposes of this paper that I am this human being, this animal writing. On that assumption, if I raise my arm then that is going to require that this human animal exists.

The Change Condition is also met in the candidate case. If the act at issue is the act of raising my arm then something needs to change: my muscles need to contract, my bones re-orientate, the position of my arm needs to change. Furthermore, acts like the act of raising my arm are necessarily changes; there are no arm raisings which are not changes. Is change always required for action? Could I act – say ‘stand very still’ – and not change anything? It seems that in fact standing very still does require changes – it requires me clenching my muscles and pushing against gravity. Perhaps there are acts such as remaining silent when someone has said ‘speak now or I will take it that you agree to the proposal’ that are not changes. However, we might think that they are not actions either – but rather deliberate omissions with foreseen and intended effects. To the extent that they are actions we can think of them as transitions between similar states – changes in time without change in properties.

Perhaps the more contentious claim here is the Self-change Condition. Why must an action that involves change also involve self-change? The simple thought is that any change I bring about must be a change brought about by my doing something and my doing something is my changing. Take those class of acts that involve changes that are not in the first characterization changes to me: acts such as raising a glass, throwing a ball or switching on a light. Let us call the changes constituting such actions changes to the non-me world. How are changes to the non-me world brought about by me? They are brought about by, myself, changing. If that is right then all actions – even if they involve changes to the non-me world – are also self-changes, or involve self-changes. I don't think I can change anything unless this thing – the animal that I am – changes. Anything else I change, I have to change in virtue of changing. I might wish I could actively change things by just hoping for them but I cannot – I need to make the change that makes the change to the non-me world.

Note that the claim that all actions that involve changes are self-changes does not mean that in all cases of action the object of the action is the self in the same way that the object of an action might be the glass, or ball, or light switch. It is rather the claim that the subject and agent of the action is always the self. This is reflected in the fact that when we specify act types we are not required to specify which subject it is that is to change in the way determined by carrying out the action. We need specify only what kind of action is to be carried out. In deciding whether to F, or to G, or to H, I do not also decide who – me, Harry or someone else – is to F, or to G, or to H. If I am deciding what to do I am already deciding what I, the decider, is to do. It is a given that the subject who determines what she should do is that thing the changes to which are the doing.

The Active Condition is extremely hard to state without falling into heavy theorizing very quickly. Some try to capture the active condition by claiming that I change things on purpose whenever I act. The thought is not that every change I bring about when I act is a change I bring about on purpose, but whenever I act, I act with purpose. Others will try to capture the idea by claiming that all acts are intentional; not that everything done by the agent acting is done intentionally but that everything done by the agent acting is done by the agent doing something intentionally. Others claim that actions are self-changes under the control of the agent. I have tried to stay away from the more or less technical notions of purpose, intention and control and will instead speak simple of self-changes that are up to the agent changing, and then contrast them with self-changes that are not up to the agent.

Consider all the self-changes that an animal undergoes. There are many of them that are not up to the animal: at least not directly. It is not up to the animal to grow: it can eat lots of protein and aid itself in growing upwards, but it cannot execute the growing. Similarly, it is not up to an animal to wrinkle, or shrink or heat up, or digest, or produce insulin. These are changes that the animal can bring about indirectly – but, again, they are not changes that are up to them to execute. In contrast there are a large number of self-changes that are directly up to the animal: the animal can run, jump, move its arms, legs, lips, tongue, eyebrows. Relative to each animal in a context there is a large class of movements, that we can call me-movements, that the animal can execute directly. They are me-movements that are up to the animal that is acting. Philosophers tend to call the relevant class here ‘bodily movements’. This terminology carries with it the unfortunate suggestion that there is on the one hand the agent, and on the other the agent's body that the agent moves when they act – thus making action out to involve some kind of relation between the agent on the one hand and the body to be moved on the other. By calling them me-movements we make it clear that action is a reflexive or self-conscious change – agents change by themselves changing – not by changing themselves. As Gareth Evans puts it, such self-consciousness is manifested in action ‘not in knowing which object to act upon, but in acting. (I do not move myself, I myself move.)’Footnote ⁶

In answer to the initial question: ‘what happens when I act?’ we have the answer: I, myself, change in a way that is up to me. We have given some sense of the category of self-changes that are up to me, and contrasted them with self-changes that are not up to me, but can we say anything more about the difference? What do we mean when we say that someone has brought about a self-change in a way that is up to them?

One way to approach the question is to look more carefully at two cases which on the surface involve a similar change, but where one is up to me, and where the other is not.

Consider the standard case of my moving my arm up. In such a case my arm moves from being down to being up – and it so moving is my moving it up. In contrast consider a case where my arm moves from being down to being up because it was blown up by a strong wind from below. If we are trying to understand what is it for my moving my hand up to be up to me, in contrast to be caused by the wind, we may be led to set ourselves a certain kind of subtraction problem. We may ask Wittgenstein's question ‘What is left over if I subtract the fact that my arm moves up from the fact that I raise my arm?’Footnote ⁷

We may start to note that it is not just that my arm went up when I moved it up – I also wanted it to go up, knew how to make it go up, intended it to go and willed it to go up. We may then, having identified things that do go on when I raise my arm in a way that is up to me, be tempted to make the further move of thinking that it is these extras make it the case that I move my arm in a way that is up to me, in contrast to a case where my arm moves up due to the wind.

We started by asking ‘what is needed for me to raise my arm?’. We have answered: well, one of the things you need is the agent who self-changes – I could not have raised my arm without me. We have also acknowledged that I might need to want my arm to go up, and need to know how to make it go up. Perhaps I also need to intend for it to go up, and to will for it to go up. This list of things that are required for me to raise my arm is very contentious – but for the purposes of this discussion let us assume that we are happy to accept that an instance of each of this list of things needs to be in place when I act – raise my arm, for example.

We have not said anything about whether these things are independent of my arm going up, or whether they are things I can understand without already understanding what it is to move my arm up in a way that is up to me. I have just agreed that we can assume that these things have to be in place when I act. That is, we have assumed that our answer to (NEC) is, at least partly, answered by that list.

The trouble, however, starts when we take two further steps. First, when we take the list that is our best attempt to answer (NEC) as an answer to (SUFF). (SUFF) was the question what is enough for it to be true that I acted. Second, when we take the resulting answer to (SUFF) to be a reductive answer to what my acting is. If we say that not only does the list (of my arm going up, my wanting it to go up by knowing how to make it go up, my intending it etc.) give me an account of what is needed to act, and a list of what is enough to act – it also tells us what my acting is, we have embarked on an attempt to analyse my action in other terms. If the attempt were successful it would imply that the act - my raising of my arm - is not actually a single unified element in my psychological life but is psychologically molecular. It is composed of all, or some, of the ‘more basic’ elements we have on the list we gave in an answer to (NEC).

There are two sort of reasons that might make us worry about whether an answer to (NEC) is going to add up to anything like an answer to (SUFF), or anything like an answer to what an action is that is stated in these more basic terms.

The first sort of worry is a circularity worry. You might think it is true that you need to want, know how to, intend, will, when you act. But what is it you need to want, know how to, intend or will to do? The answer, in our case, is ‘to raise my arm’. But to raise one's arm is the action we are trying to understand, so to know what all those other conditions are, we need to know what an action is. If we are involved in an explanatory project that is going to try to say what an action is, and build it out of the components we identify in our answer to (NEC), those components would need to be understood independently of our knowing what an action is. But it does not, at first sight, look as though they are. To want to raise my arm, know how to raise my arm, intend to raise my arm, to will to raise my arm are all things that may need to be in place when I raise my arm but they do not seem to be things we can understand independently of knowing what it is to raise my arm. As Lavin and Boyle point out, it is a ‘striking fact’ about the reductionist who tries to give an account of what an action is that:

In such re-writing we hide, to some extent, the fact that you already have the action type embedded in the description of what the agent wants when she acts, and thereby hide the fact that we need already to know what is for the agent to raise her arm if we are to know what it is for her to want to raise her arm. The same would be true of attempts to re-write what the agent knows, or intends, or wills.

The second, and more commonly discussed, sort of worry about the attempt at reduction is an insufficiency worry. The insufficiency worry amounts to the worry that one could have a case where you have an agent who wants to do something, knows how to do it, intends to do it, wills to do it, and indeed does it – but where the thing that she does is not an action that was up to her. Here is a case, adapted from a famous case of Davidson's to match the conditions we have identified as candidate states that may be required for action:

Many resourceful philosophers have expended huge effort in trying to give a reductionist theory of action only to face further counterexamples. Davidson himself declared that he ‘despair[ed] of spelling out…the way in which attitudes must cause actions if they are to rationalize the action.’Footnote ⁹

These worries invite the hypothesis that actions are prime: if actions are psychological phenomenon not to be analysed in terms of other psychological attitudes it would be clear why we face the circularity and insufficiency worries.

To assess, and indeed understand, the hypothesis we need to ask what we mean when we claim that some phenomena is prime? The dominant association that philosophers currently have when faced with the claim that some phenomenon is prime is Williamson's claim that knowledge is prime.

‘Prime’ is a word that has two distinct uses, and Williamson picked the word advisedly, meaning to make claims relative to both. However, we need here to untangle those uses if we are to be clear about what we are claiming when we consider whether actions are prime, and what we are not.

What are the two things we might be saying when we say that something is prime?

When we say that X is prime we may mean that X is BASIC in some way and not factorizable into other phenomena. It is prime in the sense that a prime number is prime. It is what it is, and not reducible to another thing. It is important to note that lots of things can be basic.

When we say that X is prime we may also mean that X is FIRST in some ordering. It is prime in the sense of Prime Minister, or prime cut. It is primary. Only ONE thing in a given class or ordering can be first.

I want to pursue the question as to whether we should treat actions as prime, by thinking about whether we should take action to be prime a way that mirrors Williamson's claim that knowledge is prime. So, what is involved in Williamson's claim that knowledge is prime? Williamson claims that knowledge is prime in, more or less, both the senses that we have identified. So, for him, both the following are true:

So, correlatively we may take the claim that action is prime to involve the following two claims:

When deciding which, if any, claims to the effect that action is prime are true it is worth noting just how strong primitiveness claims are when they are of the form ‘X comes first’ rather than ‘X is basic’. For a start claims to the effect that ‘X comes first’ are in competition with each other. If, for example, both action and knowledge are elements of the ordering relative to which we are making a primacy claim it cannot be that both action and knowledge come first: it cannot be that action has primacy over knowledge, in such a way that the latter is to be explained in terms of action and that knowledge has primacy over action, and that action be explained in terms of knowledge. Williamson's claim that knowledge is prime is a claim about both senses. This means that he is committed, not just to the idea that knowledge is basic, but to there being a significant class of phenomenon – evidence, belief, action, justification and so on – which are to be explained in terms of knowledge. This makes it clear straightaway that Williamson would have to deny that action is prime in the sense that action comes first.

Not only are ‘action first’ and ‘knowledge first’ claims in competition with each other, they are independently very ambitious claims. They demand that a whole array of phenomena come to be explained by one primary one. To that extent such claims are claims of systematic philosophy. They are also forms of reductionism about explanatory resources, governed by the idea that philosophical explanations in epistemology or philosophy of mind have to bottom out in a way that refers to just a few basic phenomena. In contrast, claims to the effect that something is basic – and not factorizable in to other more basic phenomenon within a class – are relatively philosophically, and explanatorily, ecumenical.

In suggesting that actions may be prime I mean to be suggesting, not that we take on a philosophical programme in which actions play a primary explanatory role – in contrast to knowledge, evidence, reason, belief or desire and so on – but only that actions may be prime in being basic, and not reducible to other more basic psychological notions. There are interesting questions to raise about the relations between actions and what we want, know how to do, intend or will. The circularity point made above with reference to Boyle and Lavin – that we specify what we want, know how to do, intend or will in terms of the action to which they are directed suggests that action may have a primary explanatory role in relation to a restricted set of phenomenon. However, the task for now is to set out what the claim of basicness amounts to and argue that we have reason to take it to be true.

So, what does the claim of basicness amount to? What are we saying when we say that actions are basic? A claim of basicness of a psychological phenomenon is, for my purposes, to be understood as a claim about the relation of that phenomenon to other personal level psychological phenomenon (PPPs). The class of PPPs is supposed to be the class of familiar psychological states and occurrences that we take to be correctly ascribed to psychological subjects. The class will include desires, beliefs, intentions, perceptions, emotions, suspicions, guesses, judgments, thinkings, deliberatings etc.

When we claim that actions are basic, what we are doing is denying that they are a personal level psychological phenomena that are composed of, or analysed in terms of, more basic PPPs, our bodies and their relations. Within the class of PPPs there will be those that are more basic, and that are not capable of being fully understood with reference to the others. The claim is that actions are more basic. The claim is a relative claim, and indeed not a very strong one in the broader context of philosophical reductionism.

It is important to emphasise that the claim that actions are prime, in the sense of basic, is not a claim to the effect that actions are basic particulars in the final and most fundamental inventory of the universe. It may be that the basic elements of the universe are microphysical particles perhaps not yet identified by any human science. If that is so, then only those particles will be basic in a non-relativized way. In claiming that actions are basic I am not denying that the facts about them will be fixed by the fundamental facts of the universe, if it turns out that there are any. Nor am I claiming that actions are explanatorily or metaphysically independent either of other PPPs, or of other non-PPPs. Facts about whether a creature can act may be importantly interdependent with facts about whether she can intend or desire and so on. Similarly, the actions of an agent bear dependence relations to the movements of its body, to contractions of its, to its neural activities such as its motor commands, and a whole host of sub-personal, physiological and others facts. In claiming that actions are basic, we are certainly making no claim to the effect that an animal could act without the organism and its parts doing various things, and without various low level physiological activities taking place.

The claim that actions are basic is really little more than the claim that actions are (a) psychological phenomena and (b) that, despite the tendency of philosophers to treat them otherwise, they are no less basic than beliefs, desires, intentions, tryings, willings and so on.

Does that mean that if actions are prime there is no work to be done by a philosopher of action? No. There are a number of explanatory tasks for a philosopher of action without her undertaking the task of trying given an account of actions in terms of other PPPs, our bodies, and the relations between them. We can give normal conditions on actions; we can give necessary conditions on actions. We can clarify the nature and consequences of those conditions, and give an account of the relation between actions and other psychological phenomenon. We can learn more about the neural and physiological occurrences and structures that ground or realise our actions. We can try to understand how it is that we can change, and thereby change other objects, in a way that is up to us. We can explore how active reflexive changes are possible. We can explain our particular epistemic relation to actions understood as self-changes that are up to us; we can offer a semantics for action ascriptions that respects the reflexive nature of action. These are all tasks for other occasions. What I want to do now is move beyond setting out what it means to say that actions are basic, and offering the hypothesis that actions are be basic in that sense, and consider a couple of positive arguments in favour of the hypothesis. So far all we have done is appeal to the fact that the task is a tricky one, and stated Davidson's own scepticism and despair in the face of it.

The first argument to consider is one we can call the argument from multiple sources. The reductionist about actions that I have been opposing tends, as we saw, to proceed as follows. They consider a movement a human subject might make – a movement of arm rising, for example. They then note that there are occasions where an arm rising is not an action, because they are not up to the agent. They then ask what pattern of psychological states or occurrences precedes the arm risings in the cases where the arm rising is an action that is up to the subject. The hope is that they can thereby find the difference maker: that they can identify the unique pattern of prior states and occurrences in virtue of which the arm rising is an action that is up to the agent.

A naïve first reaction to this way of proceeding is to point out that when we look at human action we see that the aetiologies, and aims, of human action seem to be as various and unpredictable as the aetiologies of human belief, desire, and intention. We should, therefore, not expect to find a unique kind of aetiology distinctive of those arm risings that are actions, in contrast to those that are mere movements.

Prima facie it seems that I can φ, where φ-ing is something I do intentionally, in very, very different ways with different causal histories and different goals. Take an everyday case of acting: suppose I drink a glass of beer. The psychological reality that can precede my drinking a beer may, it seems, come in many different forms. I might be drinking because I love the taste of beer, I might be drinking automatically from habit, I might be drinking from addiction, I might be drinking because a week ago I planned to have a beer at this time, I might be drinking spontaneously, perhaps out of thirst, maybe out of a desire for the sugar. Perhaps I am drinking for joy, or out of embarrassment, social nerves, or excitement. Perhaps I hate beer, but have resolved to drink so that I confirm to the desires of my host, or to conform to the rule I have set myself to drink at least one beer a week in order that I do my bit to keep hop growers in business. And note that the concern is not just that that there are multiple dispositions, states and occurrences that might be relevant to the fact that I drink beer, but that they may be present at the same time. Even if I am drinking out of thirst, I may also have a desire to fit in, or have planned to have a beer a week ago, or love the sugar, or be resolved to drink one beer a day. Actions spring from many sources of motivation, habit, compulsion and resolve that can all be in place at once, whether or not they are directly operative in the action at issue. Perhaps there will be one source – or one disjoint set – that is the source of the action at any time? Well, it is clear that there may be potential sources that are inoperative – but it is not obvious that we can make good sense of identifying precisely what the psychological source of the action was this time. There may be background motivations, for example, that the action does not seem to be an action for the sake of, but they might be motivations that, had they not been in place, the agent would have not acted on the distinct motivation that lead her to act. So I may not have drunk the beer to fit in, but rather out of my love of beer, but it may be true that I would not have drunk the beer had drinking it, for example, put me at odds with those around me.

Our fixed point seems to be that, whatever the prelude or purpose of an action, we end up with an agent who self-changes in such a way that she moves, and thereby moves her glass towards her mouth pouring the beer into it, and swallowing. We end up with me drinking beer in a way that is up to me. The prospects of giving a tidy account, that is true to the phenomenon, of what kinds of sources, and what kinds of links between those sources and my movement, makes the movement an action of mine look foggy. Of course, we can as Davidson did, try to conceptualise the preceding states and occurrences at a level of abstractness that brings some order to these myriad of sources. We might, for example, look for a way to construe my drinking as drinking rationalised by the belief-pro-attitude pair of (desiring the consequences of drinking, and believing that drinking is a way to secure those consequences). But the prospects of such an attempt are not my primary concern: my primary concern is to point out that the salient thing in common between all these different possible beer drinkings – and indeed in the supposed rationalising reason – is my action of drinking the beer. Without a decisive reason to talk about actions of that kind in other terms, the fact that our actions have multiple sources gives us a good reason to admit them as basic.

The second kind of argument in favour of the basicness of action we can call the argument from explanatory role. The first thing to note is the ubiquity of appeal to actions in explanations. If we look at patterns of explanations for why things are as they are we will find them laced through with appeals to human action. We will also find them laced through with appeals to human wants, knowledge, intentions and movements but not to the same extent and the explanations that appeal to human action cannot in general be inter-substituted for explanations that appeal to other PPPs and bodily movement. I make this point and give an example in O'Brien 2007:

There is nothing particularly special about this case. Explanations will very often appeal at some point of another to an action of another: How come you are late? I went to buy milk on the way home. Why are house prices going down – because more people are selling their homes due to personal debt. Where has the dog gone? He ran out the gate an hour ago. Are you hungry? No, I ate. Why haven't you sent me your paper? I was writing a referee's report.

Perhaps we could in some cases give a sufficing explanation by appealing to what the agent wants, knows, or intends, or wills and how she moves. I am not optimistic about doing so across the board. However, even if I could, the proffered explanation seems to get its target of explanation wrong. If I am explaining my anger at Elmore it is important to record that my anger is with him for doing what he did when what he did was up to him – my concern is not his intention, or his movement, but his action. Human actions – changes that are up to the agent – seem to sit at the centre of our explanatory concerns. This fact is reflected in what is perhaps the central theme of many novels and dramas. Oedipus is punished because he killed the man that was his father. Had he accidently fallen from the chariot, sword raised, desiring his father dead, planning his father's death, and willing him to kill him we would have a comedy on our hands and his punishment would be unintelligible and unspeakable. Hamlet views to seek his revenge on Claudius because he killed his father – not because he wished him dead, caused poison to drip into his ear and whatever else – but because his father died by his action. The lodestone of the story is the act of killing.

Again any force this positive argument has comes not from its having established that it is impossible to do the explanatory work needed if we do not appeal to actions. There may be a sophisticated philosophical framework that will allow us to see how the act of the other is usually our primary concern, even when the act of the other is not basic. Rather, the point is that without serious pressure to give up appealing to actions in our explanations of our interactions with each other we should take actions to be basic elements of our explanatory resources. Our everyday common sense appeal to actions in our explanations are both ubiquitous and central: it would be costly to give them up.

The real question is, given that actions are just as embedded in our explanatory and ascriptive practices as beliefs, intentions and desires, why are philosophers lead so much more easily into thinking that we need to give a reduction of actions in terms of other PPPs and bodily movements?

It is not because a successful reduction has been given that has convinced us that what seemed to be basic turned out not to be. Nor can it simply be because actions are physically realised, whereas beliefs and desires are not: most philosophers of mind will take actions to be no more, and no less, physical that beliefs and desires. They are all as much candidates for being grounded in, realised by, or identical to physical states and occurrences.

One explanation of the tendency to take actions as analysable in terms of supposedly more basic psychological and physical states and occurrences might come from looking at the parallel between knowledge and action. The resistance to taking knowledge to be a basic psychological phenomenon seems to lie in a kind of human individualism about the mental. Knowledge seems to be relational – it relates a subject to the truth – a truth which is usually independent of the individual who knows. Is action similarly relational, and so subject to the same worry, when treated as a basic psychological phenomenon? And if action is relational what is it a relation to?

In Knowledge and its Limits Williamson starts with the suggestion that there is an analogy between knowledge and action – claiming that both need to be treated as key relations between mind and world.Footnote ¹² Williamson goes on to argue there, and in a more recent paper, that this relationality is no impediment to us treating knowledge or actions as mental phenomena. As he puts it in the recent paper:

In this paper Williamson further explores the suggested parallel between the knowledge case and the action case. He sets up the parallel roughly as follows:

Williamson sees the reductionist about knowledge as aiming, and failing, to complete the formula: Knowledge = Belief + Truth + X in such a way that it is true that A knows that P iff A believes P, P, and A satisfies condition X with respect to P. In parallel he sees the reductionist about action as aiming, but failing, to complete the formula: Action = Intention + Success + Y in such a way that it is true that A intentionally φs iff A intends to φ, Aφs, and A satisfies condition Y with respect to A φs.

I have argued, in agreement with Williamson, that the task of trying to complete the action formula is mistaken, and that we have default reasons to take actions to be prime. However, I want to point to important differences between action and knowledge, and suggest that there is a way in which pursuing the analogy between the two may be problematic.

In particular, I want to argue that while we can plausibly construe a subject's knowledge in a given case as a relation between relata – the subject and the fact that P – that exist independently of the subject's knowledge, we cannot construe an action as primarily a relation between an agent and an independent condition. So, to the extent that the objection to treating a mental phenomena as basic, comes from its being understood as a relation between subject and an independent world, actions will be easier to swallow candidates for basicness than knowledge.

Why do I suggest that action may not be as easily thought of as relational in the way knowledge is? Suppose that we react to the dissatisfaction with the reductionist by making the following claims:

To understand the claim being made we need at least understand what a subject is, what a truth, P is, and what a success, one φs, is?

We have assumed for the purposes of this paper that a subject is a human animal.

What of the truth, P? Our choice here is to take knowledge to be a relation between a subject and a proposition that has the property of being true, or taking it to be a relation between a subject and a state of affairs of worldly condition. I will assume that it is agreed between both parties – the reductionist and the non-reductionist – that at least in a particular case of knowing the truth P is to be construed as a particular fact, state of affairs, or worldly condition. On this assumption, in the table above, we are comparing a belief that is true, to a case of knowledge. (P, as it figures in ascriptions of false beliefs with the content P, will be have to construed somewhat differently.) The non-reductionist holds that in a particular case of knowing there is a worldly condition that the knower stands in the knowledge relation to, and that the knower so standing in that relation cannot be reduced to, or explained in terms of, her standing in a belief relation, justification or other relation to that condition. The reductionist denies this.Footnote ¹⁴ If we understood P as, for example, merely a propositional object, or a sentence, that also has the additional property of being true, then we would have little reason to deny that the states of mind involved when a subject stands in the belief relation, and when she stands in the knowledge relation are fundamentally different. They need differ only to the extent that the belief relation can obtain when the propositional object the subject stands in a relation to, is not true, whereas the knowledge relation must meet the extra constraint of the propositional object being true. For, there to be something substantial that the reductionist and the non-reductionist are arguing about: for knowledge to be thought to be fundamentally a different relation to a truth we need to think of it, in a particular case of knowledge, as a relation between a subject and a fact, state of affairs or worldly condition – not a propositional object that also happens to be true.Footnote ¹⁵

So, if P is a truth that a subject stands in relation to when she knows, what, in parallel, does a subject stand in relation to when she acts? The analogy between knowledge and action suggests that we should take an action to be a relation to the success condition appropriate to the action. But how should we understand the success condition such that a subject stands in relation to it when she acts? Williamson suggests that the success condition of an action is that ‘one φs’.Footnote ¹⁶ But now, in order to know what the correlative claim about action is would be I need to know how to understand the condition that ‘one φs’.

First let us suppose – as Williamson's unpacking of the condition suggests – that the success condition that ‘one φs’ is such that if A is acting, then the success condition is that Aφs.

How should we understand Aφs? Williamson's guidance is that ‘the range of the variable ‘φ’ is not limited to paradigmatic actions’ (page 8): there are a range of act types – φings – which A can carry out intentionally or not. It can be true of A that she φs on an occasion when she does not intentionally φ, and also be true of her on another occasion that she φs when she does intentionally φ. However, consider a particular case when A φs intentionally, and suppose that her intentional action is a relation between A and Aφs. Should we understand Aφs in such a case as picking put an action of A's? Given that we know that in that case A φs intentionally it seems natural to do so.

But let's consider more carefully the consequences of taking Aφs to be an action of A's in this framework. In such a case, we are taking the intentional action, A intentionally φs, as a relation between A and Aφs, where Aφs is an action. The question now arises: what is the relation between that action of A's – Aφs – which is the success condition and the intentional action of A's that is the relation between A and the success condition. We have two choices. Either, the actions are the same and A does one thing: Aφs is the same action as the intentional action we are seeking to understand. Or they are different and A does two things: Aφs, but also distinctly, A intentionally φs.

But it cannot be that A does one thing. Aφs cannot be the same action as A intentionally Aφs. A intentionally Aφs cannot be a relation between A and Aφs, as we have assumed, and be identical to Aφs, which is one of its relata. A relation cannot both be one of its relata, and the relation to it.

The other option is that when A acts intentionally, A acts twice. A intentionally φs and Aφs. But that is also absurd. It cannot be that our account of an action as a relation has it that, whenever A acts, she acts twice. Apart from the implausibility of thinking of all action as duplex, we are likely to face regress. Suppose Aφs is itself as action distinct from the action A intentionally φs that it is a relation to. We have to either stop there and say that our account of intentional action as a relation to success conditions depends on accepting a class of actions as those success conditions – which are not themselves to be understood in the same way. But, this would be like suggesting that we had made progress in understanding knowledge by suggesting that we construe ‘A has believed knowledge’ as a relation between A and ‘A knows that P’. More plausibly we will then be led to consider the action ‘Aφs’ – the success condition of A intentionally φs – as itself a relation between A and a success condition, and ask how we should understand that success condition – as Aφs? We will then have a regress on our hands.

I think that these considerations suggest that if the analogy between action and knowledge is to hold, that we cannot without absurdity hold that the success condition Aφs is also an action of A's. We need to take the predicate ‘one φs’ – as it ranges overs the set of particular independent success conditions that actions are supposed to stand in a relation to – to be a non-active predicate – as ‘something that happens with or to A’.

To mirror the situation in the knowledge case we need to be able specify a non-active success condition of the form ‘Aφs’, which both the reductionist and non-reductionist can agree is not itself an action but is an independent success condition for A's action. We would need to establish that there are individual occurrences that: (a) are properly construed as being of the form Aφs, (b) are non-actions (c) which are distinct from, but necessary for, actions of the corresponding type, (d) can plausibly function as something an individual always stands in relation to when she acts intentionally.

I am not confident that this is a task that either needs to be, or can be, successfully completed. It implies – against the grain of the anti-reductionist picture – that we will be able to specify a change A undergoes that is necessary, but distinct from her action, that is of a kind that A could have undergone without acting, and in relation to which we can ask what the difference is when A is related to it when A acts, in contrast to when she does not, but it occurs. It suggests that we will have a class of changes A can undergo – perhaps, changes such as A's arm rising – which are in each case distinct from the change that is A's action – A's raising her arm. The anti-reductionist may complain that when A acts the changes she undergoes just are her actions: A's arm rising is identical with A raising her arm – and to take another view is to take us perilously close to resuming the Wittgensteinian arithmetical task we wanted to move away from.

It is beyond the scope of this paper properly to explore the question of whether we will be able to given a plausible account of non-active success conditions of the form Aφs for our actions.Footnote ¹⁷ Prima facie any success condition of the form Aφs of an action of the form ‘A intentionally φs’ is just the action itself; prima facie the reflexive-changes to A when she acts are just her actions not something that her actions stand in relation to. Again, that is not enough to establish that there is no class of non-active success conditions that the non-reductionist could acknowledge and welcome. The point for now is that it not a given common ground between the reductionist and the non-reductionist that there is a neutral and independently construable relatum – Aφs – that a subject might be argued to stand in relation to when she acts. It is a given common ground that there exists an independent truth, P, the relation to which the reductionist and the non-reductionist disagree about in the case of knowledge.

I want to end by suggesting that we should not be surprised to find a disanalogy here. The breakdown of the analogy between knowledge and action is telling. In the case of knowledge the identity of the thing known of is quite independent of the knowing of it, and the agent who knows. We can therefore quite sensibly think of knowledge as a relation between these independently existing relata. In the case of action, however, the identity of the thing done not independent of the doing of it, and the doing of it is not independent of the thing done: ‘I do what happens’ as Anscombe puts it.

This is not how we consider knowledge. The thing known – what is true – is in general independent of my knowing of it. Of course, if knowing P, in general, was what made P so – as is supposed to be the case for God – then the supposed disanalogy would break down. But so, I think, would the claim that knowledge is a relation between God and the truth: God's knowledge would be the truth. If God's knowing P made P so, then God's knowledge could not be thought of as simply a relation to P – for the relation would have to be such that it produced one of its relata. God's knowledge would instead be both God's creation and the truth. It is interesting to note that this construal explains why a commitment to God's productive knowledge often comes with a commitment to God being that which he produced. If God's knowing creates what is so, and if what God creates is an act of God, and if acts of God are self-changes – as we have argued they are – then the changes God creates must be changes to him – so he must be the truths he creates.

Once we appreciate there is no obvious way of stating the success condition of an action without appealing to the action itself we can come to see why a certain kind of action first philosophy of mind may not suggest itself. Consider the case of knowledge. We start with the following materials – the subject and the truth, P. We propose a knowledge relation between the subject and the truth – and we then note that many weaker relations might relate the subject to the truth – evidential relations, belief relations, justification relations, guessing relations and so on – without the knowledge relation being in place. When we then ask which of these relations is the primary one – which comes first in the order of explanation – it is attractive to see these weaker relations as intelligible in terms of the basic knowledge relation. But now consider the case with action. We cannot obviously suggest starting with the following materials: the agent and the success – and then go on to note the many weaker relations of intention, trying, desiring and so on – relations that might hold without the action relation holding. If the action relation did not hold the success condition would not exist.