1 Introduction

We humans seem to be capable of a distinctive kind of freedom and control over what we do which is not available to the non-human animals. And we take this distinctive human freedom and control to be crucial to what establishes our moral responsibility for our actions and omissions. In articulating the shape of distinctively human freedom and control, there is a notable tendency to appeal to a notion of self-determination.

There seems to be something compelling about this notion of self-determination as one that can help us to unpack, at least a little, what it is that is distinctive and important in human freedom and control. One can find, across a wide spectrum of accounts of free will, claims asserting that free human action is the exercise of a unique power of self-determination, or that the free human agent determines what they do in some unique manner.Footnote 1 And, since it is commonly assumed that such determination could only be a matter of (efficient) causal determination, it is also commonly suggested that free human action is a matter of the human agent causing what they do in some unique manner.Footnote 2 But there is a question of whether what is distinctive in free human action is really appropriately characterised in terms of either determination of action by the agent or efficient causation of action by the agent. That is, should an account of what is distinctive in free human action include some form of determination or efficient causation of action by the agent, as a necessary and non-trivial component? This is the question to be addressed here.

The issue can be presented as follows. A full account of free human action may be thought to have to include some kind of negative condition—a lack of determination of the action by matters that are beyond the agent’s control, for example—and/or some kind of general ability condition—an ability to do otherwise, for example, however that is to be spelt out.Footnote 3 But it is intuitively compelling that, regardless of whether these are necessary conditions for free human action and how they are to spelt out, there is some further necessary and non-trivial condition—or, perhaps, multiple conditions—regarding the relation between the human agent and their free action. Let us call this an agent-relation condition for free human action. It is such an agent-relation condition that is commonly characterised in terms of determination or causation of the action by the agent.Footnote 4 So we can understand the question that is addressed here as coming to this: is there an agent-relation condition for free human action that is appropriately characterised in terms of determination or causation of the action by the agent? Further questions, regarding negative or general ability conditions for free human action, are here largely put to one side.

My suggestion here is that we arrive at an unorthodox, but compelling, answer to our question by following two major strands of Anscombe’s philosophy: on the one hand, her account of intentional action, primarily in Intention, and, on the other hand, her characterisations of efficient causation and determination in ‘Causality and Determination’. Following these two strands, we find that it is in fact a relation other than efficient causation and determination—the relation of necessitation, which entails neither efficient causation nor determination, and which is also discussed in ‘Causality and Determination’—that we should appeal to in characterising an agent-relation condition for free human action. This conclusion suggests that we should not read too much into the language of self-determination that is so naturally used when unpacking what is distinctive in human freedom and control.

Before outlining the roadmap for the main body of this paper, I’ll briefly lay out here why I suppose that Anscombe’s account of intentional action has significant bearing on the matter of the agent-relation condition for free human action. First off, I assume that free human action is at least intentional. That is, I assume that an action’s being intentional is necessary but not sufficient for its being free and controlled in the distinctively human manner. The contained necessity claim will perhaps engender some doubt, in light of arguments, for example, by Steward (2012, pp. 34–35), that some actions by human agents are free though not intentional. However, such arguments do not unsettle the precise necessity claim assumed here since the notion of freedom targeted in those arguments is one that is applied to an extensive range of non-human animal actions as well as human actions. These arguments do not, then, unsettle the claim that distinctively human free action is necessarily intentional.

Regarding the insufficiency claim, I follow Anscombe in supposing that many cases of non-human animal action, as well as human actions, are appropriately characterised as intentional (I, Sect. 2 p. 5, Sect. 47 pp. 86–7; CD p. 146).Footnote 5 In what follows, I will indicate how this supposition coheres with Anscombe’s general account of intentional action, and, moreover, how it suggests a distinctive form of intentional action that only humans are capable of, which we can call intentional human action. But, insofar as it is supposed that non-human animals and humans alike may act intentionally, an action’s being intentional cannot be sufficient for its being free and controlled in the distinctively human manner.

I will suggest here that the Anscombean account of what is distinctive of intentional human action presents us with at least a partial agent-relation condition for intentional human action as well as free human action. However, the agent-relation condition that is presented does not itself either specify, or obviously rule out, a relation of determination or efficient causation of action by the agent. The presented agent-relation condition is a necessary epistemic relation between the agent and their intentional and free action—the agent must have practical knowledge of what they are doing and why—which leaves open whether there is a further necessary metaphysical relation between the agent and their intentional and free action.

So, ultimately, what I interrogate is whether the full agent-relation condition—the full set of necessary and non-trivial conditions regarding the agent’s relation to their action—for either intentional human action or free human action could amount to the Anscombean practical knowledge condition and a condition of determination or efficient causation of action by the agent. And the answer mooted here is: no. Rather, the full agent-relation condition for both intentional and free human action looks to be, instead, a combination of the Anscombean practical knowledge condition and a certain kind of necessitation by the agent.

In what follows, I begin by presenting the broad outlines of Anscombe’s account of intentional action and how this suggests a certain kind of practical knowledge condition for both intentional and free human action (Sect. 2). Then, before addressing what kind of further agent-relation condition for free human action coheres with the Anscombean practical knowledge condition for intentional and free human action, I turn to Anscombe’s discussion of three candidate relations: efficient causation, determination and necessitation. I present Anscombe’s characterisation of each of these relations in turn (Sect. 3). In the final main section (Sect. 4), I bring the two Anscombean threads together. I argue here that, in light of the Anscombean understanding of these relations, and the Anscombean practical knowledge condition, the full agent-relation condition for both intentional and free human action cannot include either determination or efficient causation of action by the agent; it can only, instead, include a form of necessitation by the agent.

2 Anscombe on intentional action in humans and non-human animals

In this section, I sketch Anscombe’s broad account of intentional action that is developed in Intention. I discuss how the account may apply to certain actions of non-human animals, and how this then indicates a distinctive form of intentional action that only humans are capable of. I suggest that the distinctive form of intentional human action that is indicated amounts to at least a partial agent-relation condition for both intentional and free human action.

2.1 Anscombe’s account of intentional action

The account of intentional action that Anscombe presents in Intention can be roughly summarised as follows: intentional action is action that is done with practical knowledge of what one is doing and why. Leaving aside, for the moment, the matter of what specifically practical knowledge comes to—which concerns the genesis of the knowledge and its relation to its object—we can clarify what it is that is practically known in intentional action on Anscombe’s account. That is, we can specify the relevant sense of ‘why one is doing what one is doing’ which is practically known, along with what one is doing, when doing something intentionally.

The relevant sense of “why …” is given by the fact that doing something intentionally is doing something either (a) because it is a means to some further end(s), or (b) because it is an end in itself (for its own sake).Footnote 6 The end of an action of φ-ing can also be said to be an object of the agent’s intention in φ-ing. And where an action of φ-ing is done as the means for some further end of ψ-ing, ψ-ing can also be said to be the agent’s reason for φ-ing, and the object of the agent's intention with which they φ (I, Sect. 26). By contrast, where an action of φ-ing is done for its own sake, the agent has no reason for φ-ing and no intention with which they φ (I, Sect. 17). In that case, if an agent φ-s “for no particular reason”, this does not imply that there is no end for the sake of which they φ, but rather just that there is no further end: they φ for its own sake (ibid.).

So, for an agent to have practical knowledge of why they are φ-ing, in the relevant sense, is for the agent to have practical knowledge of the end(s) for the sake of which they are φ-ing (which, if they are further ends, are the agent’s reasons for φ-ing). Altogether it can be said that, according to Anscombe’s account of intentional action, an agent a is φ-ing intentionally iff a is φ-ing and has practical knowledge of (i) their φ-ing, and (ii) the end(s) for the sake of which they are φ-ing.Footnote 7

2.2 Practical knowledge and intentional action in non-human animals

As noted in the introduction, Anscombe seems to take her account of intentional action to be one that can be satisfied by non-human animals as well as humans. That is, Anscombe states in Intention that “brutes” can have intentions (I, Sect. 2, p. 5), and that these may be intentions with which they do something, such that they amount to further ends for the sake of which they take means (I, Sect. 47, pp. 86–87). This is illustrated with the example of a cat which is crouching and slinking along with the intention of stalking a bird (ibid.).Footnote 8

Yet, in light of the account of intentional action that I have ascribed to Anscombe, with its requirement of practical knowledge of what one is doing and why, satisfaction of her account may seem to be rather too cognitively and conceptually demanding for non-human animals. In particular, it might be asked how it could be that non-human animals have knowledge of the ends of their actions. Such knowledge might be thought to require the capacity to appreciate reasons-explanations for their own action, and, therefore, sophisticated conceptual understanding which non-human animals lack. After all, doesn’t the capacity to appreciate reasons-explanations for one’s own action turn on both linguistic ability and a sort of self-consciousness beyond anything achievable for non-human animals? Turning the matter around, it might be thought that if a non-human animal, such as a cat, is appropriately said to act intentionally, then this casts doubt on Anscombe’s requirement of practical knowledge for intentional action.

These concerns indicate that the practical knowledge requirement for intentional action, as spelt out so far, is unclear: does practical knowledge of the ends of one’s action necessarily involve understanding a reasons-explanation of one’s action? My suggestion is that Anscombe in fact shows, via her cat example, that the answer to this question is no. That answer is, however, embedded in a complex picture of intentional action and practical knowledge, in which these phenomena take a distinctive, and more conceptually demanding, form in human intentional action.

Let us return, then, to Anscombe’s cat example. This example is presented in a response to a concern closely related to those which I have identified here: how could we legitimately ascribe an intention to an animal that cannot linguistically express knowledge of its own action or intentions? The cat example illustrates a kind of intention-ascription that is legitimate for non-linguistic animals: intention-ascription where “we describe what further they are doing in doing something (the latter description being more immediate, nearer to the merely physical)” (I, Sect. 47 p. 86). That is, the intention ascribed to the non-linguistic animal is an intention with which it is doing something—the object of the intention is the further end for the sake of which the animal is doing something—and the intention’s object is something else that the animal is currently doing: the intention is to ψ, when the animal is ψ-ing as intended.

Accordingly, Anscombe states that “the cat is stalking a bird in crouching and slinking along with its eye fixed on the bird and its whiskers twitching” (ibid.). And the description of the cat as “stalking a bird” is presented as involving an ascription of an intention (to stalk a bird) with which the cat is crouching and slinking along. Anscombe states that “the enlarged description of what the cat is doing is not all that characterises it as an intention (for enlarged descriptions are possible of any event that has describable effects), but to this is added the cat’s perception of the bird, and what it does if it catches it” (ibid.). So, the suggestion is that the intention is appropriately ascribed in virtue of the cat’s clear modulation of its behaviour on the basis of its perception and what it achieves.

For Anscombe, then, a kind of intention-ascription that is legitimate with regards to non-linguistic animals is one that ascribes a further end for the sake of which an animal is doing something, where the further end is something else that the animal is currently doing. And this presents a way of making sense of a non-human animal’s practical knowledge of its ends, without ascribing to it any overly sophisticated conceptual capacities. For it can now be considered that an animal’s practical knowledge of an end of their φ-ing may simply amount to practical knowledge of something else it is currently doing: ψ-ing, insofar as ψ-ing is a further end for the sake of which it is φ-ing. The cat’s practical knowledge of why it is crouching and slinking along, for example, may simply be a matter of practical knowledge of its stalking a bird, as it does so, since its stalking a bird is a further end for the sake of which it is crouching and slinking along. In this way, an animal’s practical knowledge of their ends need be no more conceptually demanding than practical knowledge of what it is intentionally doing. Such practical knowledge may even require no conceptual understanding at all, but rather be a matter of non-conceptual awareness of its ongoing intentional action.Footnote 9 It is, at least, intuitive that a non-human animal will be aware of what it is doing when it is doing something with some purpose. It is intuitive that a cat which is stalking a bird in crouching and slinking along is aware of all of this behaviour.

The suggestion here presents a way for a non-human animal to have practical knowledge of an end for the sake of which it is doing something else, without representing that end as an end for the sake of which it is doing something else.Footnote 10 An animal’s having practical knowledge of the ends that it is currently pursuing does not then require any capacity to appreciate reasons-explanations, or any other implausibly sophisticated conceptual understanding. And that non-human animals can have practical knowledge of the ends of their intentional actions just insofar as this does not involve any representation of their ends as ends is compelling.Footnote 11 It is, in fact, an idea that goes back to Aquinas: a figure who is demonstrably influential on Anscombe's work on intentional action.Footnote 12

Aquinas, writing in terms of voluntary action rather than intentional action, took voluntary action to require knowledge of its end. And, maintaining that non-human animals (irrational animals) as well as humans (rational animals) can act voluntarily, Aquinas put forward that while non-human animals do have some degree of knowledge of the ends of their actions, that knowledge could only be “imperfect”. Imperfect knowledge of an end is defined as consisting “in mere apprehension of the end, without any awareness of its nature as an end or of the relationship of the activity to the end” (Summa Theologiae, IaIIae, Q. 6, A. 3).Footnote 13 In other words, Aquinas proposes that the non-human animal cannot think either of its ends as its ends, or of its means to its ends as means to its ends. But, nonetheless, a non-human animal may have knowledge of what amounts to its means, and of what amounts to its ends.Footnote 14

2.3 Practical knowledge and intentional action in rational human agents

Aquinas also makes a proposal regarding the contrasting form of knowledge of ends that humans (rational animals) may have. They are said to have “perfect” knowledge of the ends of their actions, where perfect knowledge of an end “involves not merely the apprehension of the object which is the end, but an awareness of it precisely qua end, and of the relationship to it of the means which are directed to it” (ibid.). And this does indeed seem to be something that humans (at least those who have reached a certain stage of cognitive development) are distinctively and characteristically capable of. This sort of capacity is exercised and demonstrated whenever a human agent answers a request for a reason for why they are doing what they are doing.

In fact, what Anscombe seems to suggest is that, for a human with the general capacity for reasons-explanations of their own action, insofar as they cannot either identify a reason for their action or recognize that there is no reason for their action (because it is being done for its own sake), they are in that instance not acting intentionally (I, Sect. 17). For this to then be squared with the permissibility of intentional action in non-human animals who lack knowledge of ends as ends, it should be supposed that intentional action takes a distinctive form in humans who have the general capacity for reasons-explanations of their own actions: agents who are rational in this particular sense. That distinctive form of intentional action—intentional human action—is differentiated by its necessary involvement of a form of practical knowledge distinctive to rational human agents: practical knowledge of the end(s) for the sake of which one is doing something as the end(s) for the sake of which one is doing something.Footnote 15

It can then be said that a rational human agent a is φ-ing intentionally iff a is φ-ing and has practical knowledge of (i) their φ-ing, and (ii) the end(s) for the sake of which they are φ-ing as the end(s) for the sake of which they are are φ-ing. For example, a rational human agent who is crouching intentionally in order to hide from the enemy necessarily has practical knowledge of the fact that they are crouching and that they are crouching in order to hide from the enemy (that hiding from the enemy is the end for the sake of which they are crouching).

On this Anscombean (and, indeed, Thomistic) picture of intentional human action vs intentional non-human animal action, intentional human action is distinctive in its dependence on how the rational human agent thinks about the ends that they are pursuing and the means to those ends that they are taking. If a rational human agent does not think of themselves as φ-ing for the sake of ψ-ing, they simply are not intentionally φ-ing for the sake of ψ-ing.Footnote 16 And this is in stark contrast to non-human animal intentional action, on the Anscombean-Thomistic picture: what a cat does intentionally is not dependent on what the cat thinks about its ends, or the means taken to those ends. A cat, after all, does not have any thought at all about its ends as ends, or about its means to those ends as means to those ends. That said, a cat must have some (possibly non-conceptual) awareness of what amounts to their ends and means, including, at least, its on-going purposeful action. And there is, then, a dependence of what the cat is intentionally doing on its awareness of what it is doing. A cat that is not aware of its on-going action of crouching is not crouching intentionally. But, at the same time, the cat need have no thoughts about why it is crouching, for that crouching to be intentional.

2.4 The practicality of practical knowledge

We have now extracted an Anscombean account of distinctively human intentional action (intentional action by a rational human agent): a rational human agent is intentionally φ-ing for the sake of ψ-ing if and only if they have practical knowledge that they are φ-ing for the sake of ψ-ing. It remains to be seen, however, what the practicality of practical knowledge amounts to.

Anscombe introduces the notion of practical knowledge via a contrast to what is labelled as speculative knowledge, which is knowledge that “is derived from what is known” (I, Sect. 48, p. 87). That is, when it comes to speculative knowledge, “the facts, reality, are prior, and dictate what is to be said, if it is knowledge” (I, Sect. 32, p. 57). Practical knowledge is unlike speculative knowledge in this respect. So, practical knowledge is knowledge that is not derived from what is known, and what is known is not prior to the knowledge. Practical knowledge is non-speculative knowledge.

Then, quoting Aquinas, Anscombe also states that “[p]ractical knowledge is ‘the cause of what it understands’” (I, Sect. 48, p. 87).Footnote 17 Anscombe is largely interpreted here as following Aquinas in proposing a relation of formal causation between practical knowledge and what is practically known.Footnote 18 And a formal cause is, following Aristotle, just “the form or the archetype, i.e. the definition of the essence” of what it formally causes (1984, Physics, 194b27-30). So, for practical knowledge to be the formal cause of what is practically known is for what is practically known—what it is that the agent is doing intentionally, and the ends for the sake of which they are doing that—to be essentially (non-speculatively) known by the agent.

This implies, minimally, that the object of practical knowledge is dependent upon its being known in some way other than via derivation from the object (non-speculatively). An agent is only φ-ing for the sake of ψ-ing if the agent has non-speculative knowledge of their φ-ing and ψ-ing. And the agent’s φ-ing and ψ-ing here cannot be prior to their being non-speculatively known by the agent. Part of the story of how this works seems to lie in the idea that practical knowledge is distinctively knowledge “in intention” (I, Sect. 32 p. 57). This implies that the human agent’s non-speculative thought about what they are doing and for what end, which amounts to practical knowledge in the proper circumstances, is founded in their intention. Hereafter I will use the term ‘practical thought’ for such non-speculative thought.

Clearly, then, on the Anscombean account of intentional human action, a rational human agent’s practical thought makes a difference to what it is that they are doing and what their ends are. This may well prompt one to suppose that, given the Anscombean picture, not only is practical knowledge the formal cause of its object, but also practical knowledge—or at least the agent’s practical thought—determines, or is the efficient cause of, its object. This echoes some conceptions of practical knowledge (e.g., Velleman, 1989), as well as Schwenkler’s (2019, pp. 172–179) unorthodox interpretation of Anscombe as holding that practical knowledge is both the formal and efficient cause of its object. One might further suppose that it is via their practical knowledge that the agent themselves determines, or is the efficient cause of, the object of their practical knowledge: their action.

These amount to proposals for a further agent-relation condition for intentional human action—a further necessary and non-trivial condition concerning the relation between the human agent and their intentional action—beyond just the condition of the agent having practical knowledge of what they are doing and for what ends. That is, the account of intentional human action in terms of practical knowledge presents a kind of epistemic agent-relation condition for intentional human action. But the identification of this epistemic agent-relation condition leaves open whether intentional human action must also stand in some specific further, metaphysical, relation to the agent and their practical knowledge or practical thought. The proposals just considered amount to proposals that indeed it must: intentional human action must be determined or efficiently caused by the agent’s practical knowledge or practical thought.

However, these proposals should not be construed as denying that the Anscombean account correctly identifies a sufficient condition for intentional human action. The proposals do not deny that, if a rational human agent has practical knowledge that they are φ-ing for the sake of ψ-ing, then they are intentionally φ-ing for the sake of ψ-ing. Rather, the proposals can be thought of as offering an account of what it takes for a rational human agent to have practical knowledge of what they are doing and for what end, such that they are intentionally doing what they know themselves to be doing.

2.5 Intentional human action and free human action

I will come to assess whether the full agent-relation conditions for intentional human action and free human action could amount to a combination of (a) the agent’s having a distinctive form of practical knowledge of what they are doing and for what ends, and (b) the agent themselves, or their state of practical knowledge or thought, determining or being an efficient cause of what they are doing. And this assessment will be made in light of Anscombe’s discussion of efficient causation and determination in ‘Causality and Determination’.

My suggestion for now is the following. First, insofar as the Anscombean picture of intentional human action is accepted, the full agent-relation conditions for both intentional human action and free human action must at least incorporate the requirement for the agent to have distinctively human practical knowledge of what they are doing and for what ends. Free human action is action by a rational human agent that is at least intentional. Second, the full agent-relation conditions for both intentional human action and free human action do not include any further psychological states besides those built into the practical knowledge condition for intentional human action: the agent’s having ends for their actions (intentions) and having distinctively human thought about those ends, and what they’re doing in pursuit of those ends, which amounts to practical knowledge.

To dig into this claim, we can begin by noting that, given the Anscombean picture of intentional human action, it is clear that there are no further occurrent psychological states that are required for intentional human action. But, similarly, it would seem that there are no further occurrent psychological states that are required for free human action, besides those that are required for intentional human action. While there may be further psychological requirements for free human action (and perhaps also for intentional human action), these should take the form of general ability conditions, rather than occurrent states that would figure in an agent-relation condition.Footnote 19 The free human agent may, for example, necessarily have a distinctive ability to reflect upon their ends and means and change their ends or means to their ends, but it does not seem that for an instance of free human action, the agent necessarily exercises these abilities. Any such distinctive abilities are not then reflected in an agent-relation condition.

In that case, if the full agent-relation conditions for intentional human action and free human action include, as well as the practical knowledge condition, some further relation between the agent and their action, then this can only be a relation that involves the agent, their action, and the psychological states built into the practical knowledge condition. So, if there is a further element in the full agent-relation conditions for intentional human action and free human action, besides the practical knowledge condition, then prima facie plausible candidates are efficient causation or determination: by the agent, their intention, or their practical knowledge or thought, of their action. And prima facie it is also not obvious what other kind of relation could be a plausible candidate. In the following section, however, when I turn to Anscombe’s writing on efficient causation and determination, I identify what is presented as an additional distinct relation: necessitation. This relation, as characterized by Anscombe, does look to be a further plausible candidate, and I ultimately argue that it is the only one of the three candidates which, when understood in line with Anscombe’s characterizations, could figure in the full agent-relation conditions for intentional human action and free human action.

And now a final point to note here is that, given the restriction in what could be a further element in the full agent-relation conditions for intentional human action and free human action—efficient causation, determination, or necessitation, by the agent, their intention, or their practical knowledge or thought, of their action—there is no obvious reason to suppose that there is a difference to be uncovered here between intentional human action and free human action. Indeed, any further element looks like it will just be part of what it takes for a rational human agent to have practical knowledge of what they are doing and for what end, such that they are intentionally doing what they know themselves to be doing. This, at least, will be the defeasible presumption going forward. And, ultimately, I find no basis for giving up the presumption. So I take it, hereafter, that there is a full agent-relation condition that applies equally for both intentional human action and free human action, and it amounts to the practical knowledge condition plus any relation that must hold between the agent and their action for that practical knowledge condition to be satisfied.Footnote 20

3 Anscombe on efficient causation, determination and necessitation

This section presents Anscombe’s characterisation of the constitutive nature of the relations of efficient causation, determination and necessitation in ‘Causality and Determination’.Footnote 21 My focus here is on those aspects of the characterisation that are of particular significance to the question of whether any of these relations figure in the agent-relation condition for intentional human action and free human action. I outline Anscombe’s understanding of efficient causation, determination, and necessitation in turn, and highlight the implications regarding the connections between each of the relations.

3.1 Causality (efficient causation)

While Anscombe is not explicit on this, it is evident that the concept of causality, or causation (she uses the terms ‘causality’ and ‘causation’ interchangeably), that is at issue in ‘Causality and Determination’ is efficient causation. This is made evident in Anscombe’s characterisation of the constitutive nature of causality:

[C]ausality consists in the derivativeness of an effect from its causes ... Effects derive from, arise out of, come of, their causes. This is the core, the common feature, of causality in its various kinds. Effects derive from, arise out of, come of, their causes. For example, everyone will grant that physical parenthood is a causal relation. Here the derivation is material, by fission.

(CD, p. 136).

Anscombe here echoes Aristotle’s differentiation of efficient causation from material, final and formal causation. Aristotle characterises the efficient cause as “the primary source of change or rest” (1984, Physics, 194b30-2) and as “that from which comes the beginning of movement” (1984, Metaphysics, 984a27), evoking the idea of derivation that Anscombe identifies as core to the concept of causality that she is focused on.Footnote 22 Moreover, Aristotle also identifies parenthood as a paradigmatic example of efficient causation, stating that “the father is cause of the child” (1984, Physics, 194b30-2).Footnote 23

It is worth noting that Anscombe’s Aristotelian characterisation of efficient causation is not presented as any kind of reductive analysis. The characterisation is introduced as identifying a fact about efficient causation that is “little attended to, and yet still so obvious as to seem trite” (CD, p. 136; my italics). The obviousness and apparent triteness seem to lie in the circularity contained within the characterisation. A certain kind of derivation—not the sort that is involved in mathematics, at least—is identified as at least necessary for efficient causation, while the concept of the relevant kind of derivation is itself an inherently causal concept.Footnote 24,Footnote 25

It can also be noted here that Anscombe’s characterisation of efficient causation does not specify that the relata of the efficient causal relation must belong to any particular ontological category. Her illustrative example indicates that a substance—a parent—may be an efficient cause of a distinct substance: their child. But it is nowhere suggested that efficient causation is a relation that holds only between substances. Indeed, Anscombe sometimes describes causes and effects as occurring (CD, p. 144), suggesting that events—the kind of entity most felicitously described as occurring—as well as states of affairs, perhaps, may figure as relata of the causal efficient relation. We can, then, assume that Anscombe permits a pluralistic position regarding the relata of the efficient causal relation.

While non-reductive, and permissive of pluralism regarding causal relata, Anscombe’s characterisation of efficient causation in terms of derivation does informatively identify certain non-causal features as essential to efficient causation: in particular, the separateness and priority of causes, with regards to their effects. For, a source—that from which something derives—is necessarily separate from, and metaphysically prior to, that which derives from it. That is, the existence of that which is a source may not depend upon that which derives from it. This form of priority need not imply temporal priority however, and indeed it is not suggested in ‘Causality and Determination’, or elsewhere in Anscombe’s writing, that a cause must have a prior temporal location to its effect.Footnote 26

The characterisation of efficient causation in terms of derivation also lays the groundwork for Anscombe’s suggestion that necessitation does not entail efficient causation, since necessitation does not entail derivation.

[A]n analysis [of causality] in terms of necessity or universality does not tell us of this derivedness of the effect: rather it forgets about that. … [The necessity] does not show us the cause as source of the effect. Causation, then, is not to be identified with necessitation.

(CD, p. 136).

Similarly, Anscombe tells us that efficient causation does not entail necessitation, since an efficient cause might not necessitate its effect: “a non-necessitating cause is …. one that can fail of its effect without the intervention of anything to frustrate it” (CD, p. 144). I will return to this suggestion in my presentation of Anscombe’s characterisation of necessitation. What is more, given the relationship between Anscombe’s notions of necessitation and determination which is to be presented in subsequent sub-sections, it follows that efficient causation similarly does not entail determination. Equally, determination does not entail efficient causation, as I will expand on below.

3.2 Determination

When it comes to characterising determination, Anscombe states the following:

When we call a result determined we are implicitly relating it to an antecedent range of possibilities and saying that all but one of these is disallowed. What disallows them is not the result itself but something antecedent to the result. The antecedents may be logical or temporal or in the order of knowledge. Of the many—antecedent—possibilities, now only one is—antecedently—possible. … In the zygote, sex and eye-colour are already determined. Here the antecedent possibilities are the possibilities of sex and eye-colour for a child; or more narrowly: for a child of these parents. Now, given the combination of this ovum and this spermatozoon, all but one of these antecedent possibilities is excluded.

(CD, p. 141).

These claims are somewhat condensed however, so it’s worth spelling things out further. To begin with, it can be clarified that the notion of antecedence has three different applications in Anscombe’s characterisation of determination:

  1. (1)

    That which determines some result is said to be antecedent to the determined result—let us call this the determining antecedent. The notion of antecedence here implies that that which determines some result is at least non-identical to the determined result: the determining antecedent can be taken as given, while the determining result is not thereby taken as given. We are told that the determining antecedents may be antecedents of three kinds: “logical or temporal or in the order of knowledge”, but I won’t substantially address the differences between these three kinds of determining antecedents here. It would seem that the form of determination is constant, regardless of the kind of determining antecedent. However, it is notable that in identifying kinds of determining antecedents that are not temporal antecedents, Anscombe indicates that determination is a relation that may hold between temporally co-located entities.

  2. (2)

    The range of possibilities, from which all but the determined result are ruled out by the determining antecedent, is also said to be antecedent—let us call this the antecedent range of possibilities. This is an antecedent range of possibilities in the sense that they all amount to possibilities when neither the determining antecedent nor the determined result are taken as given: they are antecedently possible relative to the determined result and the determining antecedent.

  3. (3)

    Furthermore, just one of the antecedent range of possibilities is said to be antecedently possible once the determining antecedent is taken as given, though the determined result is still not taken as given. That which is antecedently possible at this stage is in fact the determined result.

Anscombe’s characterisation of determination can then be understood as amounting to the following: x determines y just if to take x as given is not to thereby take y as given, but taking x as given rules out the possibility of anything other than y from amongst a range of alternatives that are antecedently possible relative to x. It may be noted that, as with causes and effects, Anscombe does not specify any particular ontological category as that to which the relata of the determination relation must belong to. We can then, I think, assume that Anscombe permits a similarly pluralistic position with regards to the relata of the determination relation.

Now, given Anscombe’s characterisation of efficient causation and determination, determination would seem to be something that could hold independently of efficient causation. That is, non-causal determination would seem to be possible. Non-causal determination of y by x would be a matter of taking x as given ruling out the possibility of anything other than y from amongst a range of alternatives that are antecedently possible relative to x, while y does not causally derive from x.Footnote 27 Indeed, it should be unsurprising that determination does not entail efficient causation, as Anscombe understands this, if necessity does not indicate causal derivation. Determination would seem to be a matter of absolute necessity of the determined result, given the determining antecedent. But the relationship between necessitation, as Anscombe understands this, and determination will be explored further below.

3.3 Necessitation

While the title of Anscombe’s lecture, ‘Causality and Determination’, identifies a focus on the relationship between causality and determination, much of Anscombe’s discussion is actually put in terms of causality and necessity or necessitation, rather than determination. It might then be reasonably supposed that Anscombe takes ‘determination’ to be interchangeable with ‘necessitation’. After all, according to general usage of the term ‘necessitate’ in analytic philosophy, necessitation would seem to line up with Anscombe’s characterisation of determination: x necessitates y just when taking x as given rules out anything other than y as possible. And, as noted above, determination does seem to be a matter of absolute necessity of the determined result, given the determining antecedent.

However, Anscombe suggests in CD that there is a concept of necessitation that is weaker than that of determination: a concept of non-absolute necessitation. On this concept of necessitation, x might necessitate y without determining y, although if x determines y, then x also necessitates y. This suggestion emerges in two places. First, Anscombe indicates an ordinary concept of sufficiency, or something’s being enough for something else, such that that which is sufficient for X might be there, and yet X does not happen. This is contrasted with the philosopher’s technical notion of ‘sufficient condition’:

‘sufficient condition’ is a term of art whose users may therefore lay down its meaning as they please. So they are in their rights to rule out the query: ‘May not the sufficient conditions of an event be present, and the event yet not take place?’ For ‘sufficient condition’ is so used that if the sufficient conditions for X are there, X occurs. But at the same time, the phrase cozens the understanding into not noticing an assumption. For ‘sufficient condition’ sounds like: ‘enough’. And one certainly can ask: ‘May there not be enough to have made something happen – and yet it not have happened?’

(CD, p. 135).

This has clear implications for the notion of necessitation, since concepts of sufficiency and necessitation go hand in hand: where x is sufficient for y, x necessitates y, and vice versa. So, the implication is that there is a concept of necessitation such that conditions that necessitate X might be there, but X does not happen.

That Anscombe supposes that there is a concept of necessitation that is weaker than that of determination, as she has defined it, also emerges in her characterisation of necessitating causality, in which it is stated that a necessitating cause need not, by itself, determine the causing of its effect:

[A] cause C is a necessitating cause of an effect E when (I mean: on the occasions when) if C occurs it is certain to cause E unless something prevents it. C and E are to be understood as general expressions, not singular terms. If “certainty” should seem to be too epistemological a notion: a necessitating cause C of a given kind of effect E is such that it is not possible (on the occasion) that C should occur and should not cause an E, given that there is nothing that prevents an E from occurring.

(CD, p. 144).Footnote 28

This definition of necessitating causality clearly connects with Anscombe’s definition of determination; it effectively states that C is a necessitating cause of E iff whenever there is a C, that C, given that nothing prevents there being an E, rules out that the C does not cause an E.Footnote 29 In other words, a necessitating cause need not by itself determine the effect’s being caused; rather it is the conjunction of the necessitating cause and a lack of prevention of its effect which determines the effect’s being caused.

But the relation of necessitation per se that is connected to causality here would seem to be one that is separable from causality, just as determination is. Indeed, as I have noted, Anscombe emphasises that there is nothing in the concept of necessity to indicate causal derivation: the core of causality. We can then extract a characterisation of necessitation per se from Anscombe’s characterisation of necessitating causality, as follows: x necessitates y iff given x, and no prevention of y, it is ruled out that y does not occur. Alternatively: x necessitates y iff x and the lack of prevention of y together determine y.Footnote 30

That said, one might wonder whether it is really possible for there to be non-causal necessitation that does not amount to determination: where there may be prevention of the kind of outcome that is necessitated. One might suppose that where there is scope for prevention of an outcome of some kind that is otherwise necessitated, there is surely an efficient causal process running from the necessitating entity, which leads to an outcome of the necessitated kind as long as it is not interrupted. One might wonder how prevention of something that is otherwise necessitated could be anything other than interruption of such an efficient causal process. Then again, Ancombe’s remarks regarding the independence of necessity and causal derivation invite us to interrogate the basis of these concerns: why should it be assumed that prevention of something that is otherwise necessitated has to take this form?

Prevention of an outcome of kind K that is necessitated by some x may simply be a matter of there being some entity, p,—a preventer—such that p’s presence has the result that there is no K, while the absence of p, together with x, would determine that there is a K.Footnote 31 The K that would be determined need not causally derive from x, and so its prevention need not amount to the interruption of a causal process from x to a K. Anscombe’s remarks point to the coherence of such a notion of non-causal prevention, and so to the possibility of non-causal necessitation that does not amount to determination.Footnote 32

Indeed, such a relation plausibly holds in many mundane cases of constitution that depend upon wider circumstances being a certain way. For example, a particular event of parking on a certain street may constitute an event of breaking the law, but only insofar as certain conditions do not hold: its being a Sunday, the driver having certain special rights, the law being unchanged and so on. Such a case of constitution plausibly involves a relation of non-causal necessitation between the event of parking on that certain street and the event of breaking the law: the latter does not causally derive from the former, which is not a separate and prior source of the latter.Footnote 33 But, the occurrence of the parking event does necessitate the occurrence of a law-breaking event, according to Anscombe’s notion of necessitation. Its being a Sunday, the driver having certain special rights, and the law being unchanged would all make for non-causal prevention of the occurrence of a law-breaking event, given the occurrence of the parking event. And the occurrence of the parking event, together with the lack of prevention of a law-breaking event, determines the occurrence of a law-breaking event.

Ultimately, I will propose, in the following section, that it is the relation of necessitation, as presented here—a relation which entails neither determination nor efficient causation between the necessitans and the necessitandum—which we should appeal to in fleshing out the full agent-relation condition for intentional and free human action. And having now considered Anscombe’s understanding of the three relations of efficient causation, determination, and necessitation, we can now turn to the matter of how these relations might be thought to figure in the full agent-relation condition, together with the practical knowledge condition identified in Sect. 2.

4 Causality, determination and necessitation in free human action

This section evaluates whether the full agent-relation condition for intentional and free human action involves, as well as the practical knowledge condition spelt out in Sect. 2, a relation between the agent and their action that is a matter of (a) efficient causation, (b) determination, or (c) necessitation, as Anscombe understands these relations. As a reminder, a full agent-relation condition for intentional and free human action is a full set of necessary and non-trivial conditions regarding the human agent’s relation to their action. I will suggest that it is not viable that the full agent-relation condition involves either efficient causation or determination, while it does instead feasibly involve the kind of necessitation that Anscombe identifies.

4.1 Causality in free human action

Following Anscombe’s Aristotelian characterisation of efficient causation, efficient causation of action by the agent themselves would be a matter of the agent being the source of their action, which then derives from the agent. And indeed, it has been supposed that, for Aristotle, an action of the distinctively human kind (praxis) is something for which the human agent is the source, and so its efficient cause (see, e.g., Bobzien, 2014, p. 85–86, citing Eudemian Ethics 1222b18-31).Footnote 34

But, even if we suppose that the human agent is the efficient cause of distinctively human action, efficient causation by the agent would nonetheless not be part of what is distinctive of such action. For, if the human agent is held to be the efficient cause of distinctively human action, this will be on account of holding that all actions—in the broadest sense of ‘action’, thereby including, e.g., digestion and reflex action—are efficiently caused by their agents. Thus, Bobzien notes, regarding her interpretation of Aristotle’s account of human action, that “[i]t does not suffice that the agent is the origin and efficient cause of some change. (For the agent is also, say, the origin and efficient cause of her digestion and her reflex actions …)” (Bobzien, 2014, p. 91). Yet, efficient causation of action by the agent would not be merely insufficient for free human action, it would be a trivially necessary condition, if for any action its agent is its efficient cause. In this way, that the human agent is the efficient cause of free human action cannot be part of the agent-relation condition for free human action.

Then again, it might be supposed that there is a distinctive way in which the human agent causes their intentional and free human action, via their practical knowledge or thought, or intention. That is, it might be supposed that efficient causation of action by the agent’s practical knowledge or thought, or intention, is part of the agent-relation condition for free human action.

Let us consider first the idea that practical knowledge is both the formal and efficient cause of intentional human action, as Schwenkler has recently interpreted Anscombe as holding (Schwenkler, 2019, p. 175).Footnote 35 Given Anscombe’s characterisation of efficient causation, for practical knowledge to be the efficient cause of intentional human action would be for the agent’s practical knowledge that they are φ-ing in order to ψ to be the source from which their actually φ-ing in order to ψ derives. And, in that case, the agent’s practical knowledge that they are φ-ing in order to ψ would be separate and metaphysically prior to the agent’s actually φ-ing in order to ψ. However, no state of knowledge is separate and metaphysically prior to its object, such that the object of knowledge could be said to derive from the knowledge of it. While practical knowledge, being non-speculative, does not derive from its object, that is not to say that the object of practical knowledge, or non-speculative knowledge more generally, instead derives from the knowledge of it.

All states of genuine knowledge, including states of non-speculative knowledge, are factive and are thereby dependent for their existence upon their objects. That is, for an agent to know that they are are φ-ing, it must be the case that they are φ-ing. And Anscombe is emphatic that practical knowledge is genuine knowledge, where there is scope for error: were the agent not to be φ-ing, they could not have practical knowledge that they are φ-ing (see I, Sects. 8, 28, 32).Footnote 36 An agent’s practical knowledge is therefore not metaphysically prior to its object in the manner required for the state of knowledge to be an efficient cause of its object, on Anscombe’s understand of efficient causation.

Should we then instead suppose that practical thought, which may or may not amount to knowledge, is the efficient cause of intentional human action?Footnote 37 After all, mere thought is not a factive state and therefore is not dependent for its existence on its object. And practical thought, as an element of human practical knowledge, is necessary for intentional and free human action. However, practical thought that is separate and metaphysically prior to action is not necessary for intentional and free human action. As a result, it cannot be that an agent’s intentionally φ-ing is in all cases efficiently caused by their practical thought that they are φ-ing in order to ψ.

The objection here builds on Anscombe’s explicit rejection of the idea that an agent’s intentionally φ-ing is in all cases efficiently caused by the agent’s intention to φ. Anscombe’s objection is put most precisely in later work, where she states:

The mistake is to think that the relation of being done in execution of a certain intention, or being done intentionally, is a causal relation between act and intention. We see this to be a mistake if we note that an intention does not have to be a distinct psychological state which exists prior to or even contemporaneously with the intentional action whose intention it is. ... That the action ... was intentional, comes out in [the agent’s] explanation, in what he says if someone asks him why.

(Anscombe, 2005, pp. 95–6).

Anscombe’s concern here is that there need not be a psychological state of intending to φ that is separate and metaphysically prior to the agent’s intentionally φ-ing, as would be required for intention to always be an efficient cause—a productive source—of intentional action.

Similarly, we may note that there need not be a psychological state of thought about what one is doing and for what end that is separate and metaphysically prior to the agent’s intentionally φ-ing.Footnote 38 Indeed, this would seem to follow from Anscombe’s point regarding intention, if practical knowledge is taken to be “[knowledge] in intention” (I, p. 57), as Anscombe suggests. If in some case there is no state of intending to φ that is separate and metaphysically prior to the agent’s intentionally φ-ing, then similarly, in that case, there will be no state of practical thought about what one is doing and for what end that is separate and metaphysically prior to the agent’s intentionally φ-ing.

So, neither efficient causation of action by practical thought nor efficient causation of action by intention is a necessary condition for intentional and free human action. And altogether, efficient causation, as Anscombe understands it, does not seem to be the appropriate relation in terms of which we could flesh out the full agent-relation condition for intentional and free human action.

4.2 Determination in free human action

We can now instead consider the idea that part of the agent-relation condition for free human action is a matter of determination of action. As with efficient causation, however, it cannot be that determination simply by the agent themselves is part of the agent-relation condition for intentional and free human action, given Anscombe’s characterisation of determination. The presence of the human agent does not in itself determine the agent’s doing anything in particular, since it does not rule out all but one of a range of antecedent possibilities. It may be considered instead, though, that an agent’s intentional and free human action is distinctively determined by their intention, or their practical knowledge or thought about what they are doing and for what end. In that case, the agent could be said to—in some derivative sense—determine their intentional and free human action via their intention or practical knowledge or thought.

Let us consider the three options here.Footnote 39 On the one hand, it should be clear that neither an agent’s intention to φ nor their practical thought that they are φ-ing, by themselves, rule out that the agent is in fact not φ-ing. There are two sorts of scenarios in which an agent might intend to do something and wrongly think that they are doing it. Anscombe indicates several of these in Intention. Firstly, an agent might not have the ability in the circumstances to do that which they intend and take themselves to be doing in their practical thought: someone might think that they are replenishing the house water supply by operating the water pump connected by a pipe to the house water supply, as they intend, but the water is simply running out of a hole in the pipe. In that case, the agent is not replenishing the water supply by operating the water pump, since this is something that are simply unable to do in the circumstances that they are in (see I, Sect. 31, p 55).Footnote 40 Secondly, an agent may be perfectly able in the circumstances to do that which they intend and think they are doing, but simply end up not doing it: someone might have the practical thought that they are pressing Button A, as they intend, though they are in fact instead pressing Button B, on account of absent-mindedly mixing up Button A and Button B in the moment (see I, Sect. 32, p. 57). In light of these two sorts of scenarios, neither an agent's intention nor their practical thought can be construed as a determining antecedent of what it is that the agent is doing intentionally.

We can turn to, then, the alternative idea that the agent-relation condition for intentional and free human action involves determination of action by the agent’s practical knowledge of what they are doing and for what end. Now, it should be clear that, since states of knowledge are factive, the state of knowledge that p does determine that p. After all, if a knows that p, then this rules out the antecedent possibility of p’s not being the case. In that case, it cannot be denied that an agent’s having practical knowledge determines the object of that knowledge. An agent’s knowledge that they are φ-ing for the end of ψ-ing—whether that knowledge is speculative or non-speculative—determines that they are φ-ing for the end of ψ-ing. But then determination of action by the agent’s practical knowledge of what they are doing and for what end as a condition for intentional and free human action is simply a trivial consequence of the condition of the agent’s practical knowledge of what they are doing and for what end.Footnote 41 We in fact have here no further fleshing out of the agent-relation condition for intentional and free human action.

What is more, I think it is intuitive that having knowledge of what one is doing, and thereby determining what one is doing, is not part of how a human agent controls what it is that they are doing. While it is plausible that control in free human action entails practical knowledge of what one is doing, this may be accounted for as follows: how a human controls what they are doing when acting freely is via their practical thought, founded in their intention. And, if one successfully controls what one is doing—if one manages to be doing what one thinks one is doing—then one’s practical thought is made true, and non-accidentally so, such that it amounts to practical knowledge. On this story, then, practical knowledge of what one is doing is a consequence of successful control of what one is doing. But it is practical thought that is crucial to how the agent exerts control. This role cannot be a matter of practical thought determining or being an efficient cause of what the agent is doing however: these were shown to be dead ends. I suggest that instead of appealing to efficient causation and determination here, we should appeal to Anscombe’s notion of necessitation.

4.3 Necessitation in free human action

My suggestion is that the full agent-relation condition for intentional and free human action involves a necessitation condition given in terms of practical thought. But before spelling out the details of the necessitation condition to be suggested, I will briefly note why a necessitation condition in terms of either practical knowledge or intention will not do. With regards to practical knowledge, it should be clear that insofar as determination by practical knowledge is a trivial consequence of the practical knowledge condition, so too is necessitation by practical knowledge. Necessitation, as Anscombe presents it, after all, is weaker than determination: if x determines y, then x also necessitates y, though x might necessitate y without also determining y. Then, with regards to intention, it does not seem plausible that an agent’s intending to φ by itself necessitates their actually φ-ing. That is, the agent’s intention to φ, together with the lack of prevention of the agent’s φ-ing, does not determine the agent’s actually φ-ing. An agent may intend to do something and then change their mind on the matter before they make a start on acting on their intention. Such a turn of events does not introduce anything that prevents the agent from doing what they had intended to do.

What looks more plausible is that an agent’s practical thought that they are φ-ing for the end of ψ-ing necessitates their actually φ-ing for the end of ψ-ing. If an agent not only intends to φ, but also thinks (practically) that they are φ-ing, for some end or other, then it looks like, if all has gone well, and nothing has got in the way, then they are indeed actually φ-ing for the relevant end. This proposal, then, is that an agent’s practical thought that they are φ-ing for the end of ψ-ing, together with the lack of prevention of the agent’s φ-ing, determines the agent’s actually φ-ing for the end of ψ-ing.

For this proposal to hold water, it must be the case that whenever an agent is not actually φ-ing, despite having the practical thought that they are φ-ing, their φ-ing is prevented. This implies that, in such cases as the button-pressing case presented in the previous sub-section, the agent who does not act in line with their practical thought, despite having the ability to do so in the circumstances, is prevented from doing so. That is, the agent who presses Button B, despite having the practical thought that they are pressing Button A, and being able in the circumstance to press Button A, is prevented from pressing Button A. As I presented this case earlier, the agent’s mistake is attributed to their absent-mindedly mixing up Buttons A and B. Such a mix-up does seem to be appropriately characterised as something that prevents the agent’s pressing Button A.Footnote 42 More generally, it seems appropriate to hold that where an agent is not φ-ing, despite their practical thought that they are φ-ing and their ability to φ in the circumstances, the agent is prevented from φ-ing by a mental lapse of some kind.

Similarly, on the simple necessitation proposal currently being considered, all those conditions that remove an agent’s ability in the circumstances to do what they take themselves to be doing in practical thought are identified as preventing the agent from acting in line with their practical thought. Such conditions include purely circumstantial conditions, such as in the earlier-considered case of water running out of the pipe when someone thinks that they are replenishing the house water supply. This condition removes the agent’s ability only in the given circumstances to replenish the house water supply; the agent’s more general ability to replenish the water supply is not affected.Footnote 43 Such circumstantial conditions seem appropriately classified as preventing the agent from acting in line with their practical thought.

But, of course, there may also be conditions which remove an agent’s more general ability to do what they take themselves to be doing in practical thought. An agent might become ill and lose the muscular strength to pump water forcefully enough to replenish the water supply, for example. Such a condition, on the proposal under consideration, will also be classified as something that would prevent the agent from actually replenishing the water supply, if they have the practical thought that they are replenishing the water supply (while operating the pump weakly). This, however, does not seem obviously appropriate: lack of muscular strength is not obviously something that prevents the agent from replenishing the water supply when they think that they are doing this. More generally, it does not seem wholly appropriate to identify an agent’s lack of general ability to φ—or what underlies such a lack of ability—as preventing the agent from φ-ing when they have the practical thought that they are φ-ing. An agent’s practical thought about what they are doing and for what end is, rather, not by itself enough, even in the sense that Anscombe identifies, with her weak notion of necessitation, for actually doing what one thinks one is doing.

Instead, it is practical thought together with the general ability to do what one thinks one is doing that is enough, in Anscombe’s sense, for actually doing what one thinks one is doing. I suggest that we then take up a slightly more complex necessitation proposal: an agent’s practical thought that they are φ-ing for the end of ψ-ing, only together with their general ability to φ, necessitates their φ-ing. And this necessitation does not imply efficient causation of action by the agent’s practical thought, though it is compatible with such efficient causation. That is, it may be that there are times when the agent’s practical thought about what they are doing, and for what end, takes place separately from their acting in line with their practical thought, such that their practical thought efficiently causes their so acting. But, as previously noted, this need not be the case. And where it is not the case, it may be that the agent’s practical thought that they are φ-ing for the end of ψ-ing, in the context of a general ability to φ, and no prevention of their φ-ing, instead constitutes their actually φ-ing in order to ψ. Such cases of constitution would belong to the class of cases of constitution flagged in Sect. 3.3, where the constitution depends upon wider circumstances being a certain way.

The full agent-relation condition that results from bringing the more complex necessitation proposal together with the practical knowledge condition is the following: for an intentional and free human action of φ-ing for some end of ψ-ing, the agent must have distinctively human practical knowledge of their φ-ing for some end of ψ-ing, where that they are φ-ing for some end of ψ-ing is necessitated by their practical thought that they are φ-ing in order to ψ, together with their general ability to φ. The suggestion is that the necessitation proposal, in terms of practical thought and general ability, specifies a relation between the agent and their action that is necessary for a rational human agent to have practical knowledge of what they are doing and for what end, such that they are acting intentionally and freely. And while the proposal is that it is a relation that must hold for practical knowledge to be in place, it is not a condition that trivially follows from the practical knowledge condition, in the manner that determination and necessitation of intentional action by practical knowledge does. That practical knowledge, and therefore practical thought, is necessary for intentional action does not entail that practical thought is part of a necessitating condition for intentional action.

Given this full agent-relation condition that we have arrived at, while we may not say that the human agent always themselves determines or efficiently causes their intentional and free action, we may say that that the human agent always (derivatively) necessitates their intentional and free human action. The human agent necessitates their intentional and free human action via their practical thought about what they are doing and for what end and their general ability to do what they think they are doing.

5 Conclusion

I have argued that we arrive at a certain full agent-relation condition for intentional and free human action when we follow two threads of Anscombe’s work: her writing on intentional action on the one hand, and her writing on efficient causation, determination, and necessitation on the other. The full agent-relation condition that we arrive at includes the relation of necessitation between the agent’s practical thought, their abilities and their action; it does not feature either efficient causation or determination by the agent, or any state of the agent.

There are three features of this conclusion that deserve some reflection here. First, it should be acknowledged that clearly a lot rests on Anscombe’s account of intentional action and her theorizing of efficient causation, determination, and necessitation. Insofar as any of this is contentious, so, of course, is the proposal for the full agent-relation condition for intentional and free human action. However, I hope that my argument here at least indicates that we should interrogate claims that free human action is distinctively determined or efficiently caused by the human agent.

This leads me to the second feature of the conclusion that I have argued for here, which is its bearing on the language of self-determination, which I flagged at the outset as common to attempts to unpack what is distinctive about free human action. My conclusion here implies that this language is, at least, liable to mislead: we should not to read too much into the language of self-determination, and take it literally. This is not to say that such language is inappropriate, but rather that it simply should not be taken to express a distinctive form of determination or efficient causation of action by the agent. It may instead simply indicate some kind of distinctive influence that the agent has over what they are doing, via their practical thought. Such influence can be captured by the necessitation condition put forward here. Alternatively, it might be supposed that the language of self-determination, as applied to free human action, is instead expressive of something that is to be captured in terms of a negative and/or general ability condition for free human action: self-determined action is perhaps simply action that is not determined by anything beyond the agent’s control, where the agent had the ability to do otherwise.

Finally, it is also worth noting here that my proposed necessitation account of the agent-relation for intentional and free human action does seem to present the free human agent as having rather less control over their action than would be implied by a determination or efficient causation account. After all, on my necessitation account, the free human agent does not causally produce their action in a distinctive way; and nor do they guarantee that they act a certain way. Rather, on my account, the free human agent can only hope that nothing prevents them from doing what they are able to do and take themselves to be doing in practical thought. This is then, perhaps, a lesson in humility regarding the extent of our control.