from G. Strawson Real Materialism and Other Essays OUP 2008; first published 2004
Philosophical Topics
15
Free Agents
1
I N T RO D U C T I O N
Are we free agents? It depends how you understand the word ‘free’. In this paper I
am going to take my bearings from the fundamental sense of the word in which we
cannot be free agents in the absolute way we sometimes suppose because we cannot
be ultimately (morally) responsible for our actions. I will express this by saying that
we cannot be U-free agents. It does not follow from the fact that we cannot be Ufree agents that we cannot have all the freedom of action we can reasonably want,
for we can.
2
T H E B A S I C A RG U M E N T
The short form of the argument that U-freedom is impossible is as follows.
(1) When you act, at a given time t, you do what you do, in the situation in which you find
yourself at t, because of the way you are, at t.
(2) But if you do what you do at t because of the way you are at t, then in order to be ultimately (morally) responsible for what you do, at t (in order to be U-free, at t), you must
be ultimately (morally) responsible for the way you are, at t, at least in certain fundamental,
mental respects.
(3) But to be ultimately morally responsible for the way you are, at t, in certain fundamental
mental respects, you’d have to be causa sui in those respects.
(4) But nothing can be ultimately causa sui in any respect at all (or if God can be, nothing else
can be).
(5) So you can’t be ultimately morally responsible for what you do—you can’t be U-free.¹
I call this the Basic Argument against U-freedom. Here is a variant.
(A) One cannot be causa sui —one cannot be the ultimate, originating cause of oneself.
(B) But one would have to be causa sui, at least in certain crucial mental respects, in order to
be ultimately morally responsible for one’s decisions and actions.
¹ The way you are when acting at t is a function of many things, including of course your
experience of your situation (the reference to a particular time isn’t necessary, but some find it
helpful). It is not just or especially a matter of your character, and the argument has its full force
even for those who question or reject the explanatory viability of the notion of character (see e.g.
Harman 1999, 2000; Doris 2002).
Free Agents
360
(C) So one cannot be ultimately morally responsible for one’s decisions or actions: one cannot be ultimately morally deserving of praise or blame for one’s decisions or actions or one’s
character or indeed for anything else.
These are brief versions of the argument that U-freedom is impossible, and both
premiss (3) in the first version and premiss (B) in the second can be questioned. While
many think both are obviously true, others think they need argument. I have argued
for both in other places,² and will take their soundness for granted in this paper. ‘The
causa sui’, as Nietzsche says,
is the best self-contradiction that has been conceived so far; it is a sort of rape and perversion of logic. But the extravagant pride of man has managed to entangle itself profoundly and
frightfully with just this nonsense. The desire for ‘freedom of the will’ in the superlative metaphysical sense, which still holds sway, unfortunately, in the minds of the half-educated; the
desire to bear the entire and ultimate responsibility for one’s actions oneself, and to absolve
God, the world, ancestors, chance, and society involves nothing less than to be precisely this
causa sui and, with more than Baron Münchhausen’s audacity, to pull oneself up into existence
by the hair, out of the swamps of nothingness . . .³
3
C AU S A S U I
U-freedom is impossible because one of its necessary conditions—being causa sui, at
least in certain respects⁴—is unfulfillable. But the fact that U-freedom is impossible
does not prevent us from enquiring into its conditions—into what it would take to
be a U-free agent. We can specify the conditions of impossible things. We can say
what something has to be like to be a round square: it has to be an equiangular, equilateral, rectilinear, quadrilateral closed plane figure every point on the periphery of
which is equidistant from a single point within it. It is because we know the content
of the concept ⁵ that we know that there cannot be such a thing, and
the same is true of the notion of U-freedom: we couldn’t know that U-freedom was
impossible unless we not only knew at least one necessary condition of U-freedom,
but also knew that that necessary condition—the causa sui condition—was unfulfillable.⁶
So we may enquire into the conditions of an impossible thing, and that is what I
want to do in this paper. Suspending judgement on the question whether it is actually
impossible for a thing to be causa sui, I want to try to state sufficient conditions of Ufreedom, reworking my previous attempt to do this in Part III of my book Freedom
and Belief (1986).
² See e.g. Strawson 1986: ch. 2, 2001 and Essay 13. (2) may also be questioned, of course; see
Strawson 1986: 29.
³ Nietzsche 1886: §21.
⁴ I will usually take the qualification ‘at least in certain respects’ for granted.
⁵ I use small capitals for names of concepts.
⁶ Some say that statements or concepts that are self-contradictory are meaningless, but this
cannot be right, because meaningfulness is a necessary condition of contradictoriness.
Free Agents
361
—Surely a conviction so strong and so central to our conception of ourselves must at
least have an intelligible possibility as its object!⁷
I think the content of the notion of U-freedom or ultimate responsibility can be
vividly characterized, and is in that sense perfectly—fully—intelligible, even though
it turns out not to be a coherent possibility.
One dramatic way to characterize it is by reference to the story of heaven and hell.
Ultimate moral responsibility is responsibility of such a kind that, if we have it, it
makes sense to propose that it could be just to punish some of us with torment in hell
and reward others with bliss in heaven. It makes sense because what we do is absolutely up to us. I stress ‘makes sense’ because one doesn’t have to believe in the story
of heaven and hell in order to understand the notion of ultimate responsibility that
it is used to illustrate. Nor does one have to believe in the story of heaven and hell
in order to believe in ultimate responsibility; many atheists have believed in ultimate
responsibility.
A less dramatic but equally effective way of characterizing ultimate responsibility is
this: ultimate responsibility exists if and only if punishment and reward can be truly
just and fair without having any pragmatic justification.⁸
Many philosophers set their primary focus on the notion of a U-free action, rather
on the general notion of U-freedom or U-free agenthood. There is, however, an important sense in which the task of giving a general account of free agenthood is strictly prior
to the task of giving a general account of free action. This is certainly so in the case of
standard compatibilist accounts of free action that define freedom of action simply as
the absence of certain sorts of constraints on action, for if one defines freedom in this
way by reference to action one needs an account of free agenthood, that is, an account
of what sort of thing is capable of genuine free action when suitably free from constraint. Stones, after all, are not subject to any of the constraints standardly proposed
as restrictions on free action by compatibilists (they are not kleptomaniacs, they do not
suffer from OCD, they do not have first-order wants that they do not want to have),
but they are not ipso facto capable of free action. So all fully spelt out compatibilist Constraint theories of free action have to give an account of free agenthood. They cannot
simply say that freedom is freedom from some specified set of constraints.
It may seem easy to specify what it takes to be a U-free agent: all it requires is that
one be
[0] causa sui
and that one be
[1] an agent.⁹
On this view, an agent’s being causa sui is not only necessary for it to be a U-free
agent but also sufficient. So once one has an account of what it is to be an agent all
⁷ Cf. Nagel 1987: 6.
⁸ Cf. Strawson 1998: 748–9.
⁹ In this paper I will restrict attention to agents that are capable of intentional action, and when
I speak of acts or actions I will mean intentional actions (obviously chemical agents are not agents
in this sense; nor are economic climates that are agents of change).
Free Agents
362
one has to do is to analyse, as far as one can, what is involved in being causa sui. And
here there seems to be little to say. To be the cause of oneself is to be the cause of
oneself. The notion has nothing hidden about it. If it feels obscure, that
is only because it is so evidently paradoxical. The feeling of obscurity does not show
that the concept does not come clear before the mind (the feeling may stem
principally from the fact that the notion presents vividly as something impossible, but
without the obvious visualizable impossibility of ).
On this view, then, the analysis of U-free agenthood is quickly done, once we know
what it is for something to be an agent: we just add [0] to [1]. But it is not so simple,
for three reasons. First, we need to take account of the fact that when we are concerned with U-freedom we are necessarily concerned with what it might be for something to be causa sui with respect to, and so radically responsible for, certain of its
mental characteristics. Thus suppose we have a good account of what it is to be an
agent. It’s not as if adding a completely general account of what it is to be causa sui
(‘It’s really quite simple . . . in the case of stones, for example, it’s just for the stone
to be—truly—the origin or creator of itself ’) will leave us with nothing more to say
about what it is to be a U-free agent. For the way of being causa sui that concerns
us will essentially involve complicated mental things like the capacity to reflect on
the content of one’s own character traits and preferences in such a way as to choose
among them.
Second, it would be helpful to have some further account of what it is to be an
agent, for in the present context being causa sui is merely the differentia that distinguishes U-free agents from all those other members of the species agent that are not
U-free agents.
Third, it is not clear that being causa sui and being an agent are sufficient for being
a U-free agent, even if they are necessary. In Freedom and Belief I argued that even if
one were an agent, and even if one were causa sui in the right kind of way, one might
still not be a U-free agent.¹⁰ Thomas Nagel and Ingmar Persson disputed this claim
in their reviews of Freedom and Belief,¹¹ and I will consider their objections in §16
below.
4
AG E N T H O O D
I take it that a dog—Fido—is capable of performing intentional actions for reasons, and is therefore an agent in the present sense. But no one I know thinks that
Fido can be a U-free agent in the way that many think human beings can be. No one
believes that Fido can be ultimately responsible for what he does in such a way as to
be (without any sort of qualification) fully morally deserving of praise or blame for
his actions. It seems, then, that we need an account of what we share with dogs that
makes both species agents, followed by an account of what it is that differentiates
us from dogs in such a way that our candidacy for U-freedom survives when theirs
lapses.
¹⁰ Strawson 1986: Part III.
¹¹ Nagel 1987: 5–6; Persson 1987: 66.
Free Agents
363
So what is it to be an agent? I will be brief and take it that one is an agent if and
only if one is
[2] capable of forming beliefs of certain sorts
[3] capable of having desires (pro-attitudes) of certain sorts
[4] capable of self-change (e.g. self-movement)
[5] capable of practical reasoning (in a realistically inclusive sense in which dogs and one-yearold children are capable of practical reasoning).
Some may say that this list of conditions contains redundancy, others may say that
it states only necessary conditions, not sufficient conditions (even after it has been
allowed that a being can be an agent, an entity capable of performing intentional
actions, even if it never actually acts). Others again may say that it is circular, because
[5] already relies on an understanding of the notion of an agent. [5] certainly needs
further careful exposition, and there is a great deal to say about all these conditions,¹²
but here I will take a general understanding of what they involve for granted. Readers
may insert their own preferred account of what is minimally sufficient for being an
agent (a being capable of intentional action) at this point. No one, I hope, will want
to say that dogs, dolphins, and one-year-old infants are not really capable of intentional action.
5
SELF-CONSCIOUSNESS
The question is now this: if conditions [1]–[5] state what agenthood is,¹³ what has
to be added to them to get U-freedom? It seems clear that something must be added,
given that dogs and human infants fulfil these conditions but are not U-free agents.
Well, [0], the causa sui condition, is presumably crucial, and this creates a difficulty, because if we want to test the effect of adding [0] to the minimal case in
which [1]–[5] are fulfilled we have to suppose, very artificially, that at some point
in their (individual) pasts dogs and infants have the sophisticated capacities necessary
for becoming causa sui in the required way (for example, the capacity to reflect on
the content of one’s own character traits and preferences in such a way as to choose
among them) although they revert, after becoming causa sui, to being things that evidently do not have such capacities.
Let us nevertheless suppose this for the sake of argument.¹⁴ Are dogs and infants,
if causa sui, then U-free agents? Are [0]–[5] sufficient for U-freedom? One may well
think not; one may feel that there is some other crucial missing condition. And there
seems to be a leading candidate, the condition that one be
¹² The beliefs and desires must be such as to link up in a certain way in the mind of the putative
agent.
¹³ [1] does no work.
¹⁴ Some may think this is all too fantastic to be about what really concerns us when we think
about the problem of free will. I think, on the contrary, that it is crucial to understand that we are
led to exactly such places when we think seriously about the problem.
364
Free Agents
[6] fully and explicitly self-conscious.
Surely any truly U-free agent must be able to explicitly grasp or apprehend itself as
itself, in the way distinctive of and definitive of self-consciousness, in thinking about
itself and its actions—in a way that (we take it) dogs and infants cannot?
It’s far from clear why [6] should be thought to make the crucial difference, when
it comes to something so momentous as U-freedom, but many think it can and
does. They think that adding [6] self-consciousness to [1]–[5] is enough to secure
U-freedom, or at least full moral responsibility, quite independently of [0]. To be a
U-free agent, they may say, is just to be a genuinely self-conscious agent, for a selfconscious agent can really know what it’s doing and that it is itself doing what it’s
doing. [6], they may say, renders [0], the problematic (unfulfillable) causa sui condition, completely unnecessary:
—That’s right. The capacity for fully explicit self-conscious deliberation in a situation
of choice—the capacity to be explicitly aware of oneself as facing choices and engaging
in processes of reasoning about what to do—suffices by itself to constitute one as a Ufree agent in the strongest possible sense; whatever else is or is not the case. Causa
sui be blowed; one’s full self-conscious awareness of oneself and one’s situation when
one chooses simply annihilates any supposed consequences of the fact that one neither
is nor can be causa sui. The mere fact of one’s self-conscious presence in the situation
of choice confers U-freedom on one, and obviously so. One may in the final analysis
be wholly constituted as the sort of person one is by factors for which one is not and
cannot be in any way ultimately responsible, but the threat that this fact is supposed
to pose to one’s claim to U-freedom is simply vapourized by the fact of one’s fully
self-conscious awareness of one’s situation.
I think this view has considerable power and attractiveness. I think it correctly
describes one of the substructures of our deep belief in U-freedom. But I’m sure that
it is not an account of anything that could really constitute U-freedom, because it
ignores [0], the causa sui condition.
What about the converse view that [0], being causa sui, renders [6], being selfconscious, unnecessary? On this view, if Fido could be causa sui in the required way
he would ipso facto be a U-free agent in every sense in which we are, and whatever else
was or was not the case. If Fido once brought it about that he was the way he was,
mentally, in the relevant respects, at some strangely lucid time in his past, then he
would now be a U-free agent, even though he is not—no longer—not self-conscious.
Does anyone believe that self-consciousness is not necessary for true freedom?
Persson does not explicitly endorse such a view, but he does raise a doubt about
whether ‘common sense in its judgements of desert and its adoption of reactive attitudes’ requires self-consciousness on the part of the agent judged.¹⁵
There is a quick reply to this, already touched on. It seems clear that the causa sui
requirement that is supposed to be necessary for U-freedom cannot possibly be fulfilled by an unself-conscious creature. For to fulfil it one must at some time have
consciously and explicitly decided on a way to be, mentally, in certain respects,
¹⁵ 1987: 65.
Free Agents
365
and (roughly) one must then have acted on that decision with success, and have
intentionally brought it about that one is the way one is, mentally, in the relevant
respects.¹⁶ But it is not possible to do this in the required way without being selfconscious, if only because one must grasp that the mental features one contemplates
are one’s own. So fulfilment of [0], the relevant causa sui condition, requires fulfilment of [6], the self-consciousness condition, and cannot render it unnecessary.
The reply to this reply has also been touched on. Suppose that a self-conscious
being, C, has done everything needed to be causa sui: it has intentionally brought it
about that it is the way it is, mentally, in the relevant respects; it has set up its own
practical-moral constitution in such a way it is indeed wholly responsible for that constitution, and in such a way that we judge it to be ultimately morally responsible for
what it does. Now, suddenly, C loses all capacity for self-consciousness, while continuing to be a complex agent (C, we may suppose, is a cousin of Nemo, a being who
is by definition as much like an ordinary adult human being as it is possible to be
without being self-conscious).¹⁷ Does this mean that suddenly C is no longer a Ufree agent at all, no longer ultimately morally responsible for what it does? Anyone
who thinks self-consciousness is necessary for U-freedom will think so, and I agree,
partly for reasons still to come; but Perssonian common sense may think it far from
obvious.
I will return to this idea when considering Nagel’s objection to a different but
related case. For the moment I will assume that [6], self-consciousness, is at least
necessary for U-freedom, in addition to [1]–[5], the basic conditions of agency.
6
S T RU C T U R A L I S M A N D AT T I T U D I N A L I S M
Many agree that self-consciousness is necessary for U-freedom, even if they want
nothing to do with the causa sui condition; they think it necessary for genuine
moral responsibility whether or not genuine moral responsibility requires causa-suiinvolving U-freedom. And they not only think that self-consciousness is the crucial
difference between dogs and ourselves, when it comes to the question of U-freedom
(or to the question of causa-sui-free moral responsibility), or at least that it is one
crucial difference between dogs and ourselves. They also think, much more generally, that whatever the final analysis of U-freedom turns out to be, U-freedom is and
must be simply a matter of having a certain sort of (suitably complex or sophisticated) agentive-cognitive make-up or agentive structure that can be fully described, for
all the purposes of the free will issue, in very general functional-capacity terms like
[1]–[6] above.
I will call this position the Agent-Structuralist position—the Structuralist position,
for short. It is I think both natural and attractive, but I am going to argue that it
is incorrect, and that there are among the conditions of U-freedom conditions that
require that the agent possess certain experiential or as I will say attitudinal properties
¹⁶ See e.g. Strawson 1986: 28–9, 1998: 746.
¹⁷ See Strawson 1986: 19–20.
366
Free Agents
over and above any agent-structural properties that are necessary for U-freedom.¹⁸
This makes me an Attitudinalist, as opposed to a Structuralist, in a sense that I will
explain.
Structuralists can scoop up all the conditions so far considered. They can include
[0], the causa sui condition, in the set of agent-structural conditions along with all
the conditions of self-conscious agenthood. They can treat it as just one more general
agent-structural condition. It’s true that it is a condition of a different sort from the
conditions of self-conscious agenthood already stated ([1] to [6]) in as much as it is not a
requirement that one have some enduring functional capacity but rather a requirement
that one’s overall motivational structure have been generated in a certain way. But Attitudinalists can simply give the causa sui condition to the Structuralists, allowing it to
count as just one more agent-structural condition in addition to the conditions of selfconscious agenthood, while continuing to insist that agent-structural conditions can
never be enough for U-freedom, and that further distinctively attitudinal conditions
must be added to any proposed set of agent-structural conditions.
Structuralists may also propose to include the requirement that one be
[99] subject to determinism
or alternatively
[98] not subject to determinism
among the general agent-structural conditions of U-freedom, and once again Attitudinalists need not object.¹⁹ They can allow conditions like [99] or [98] to be counted as agent-structural conditions in the largest sense (they are clearly not attitudinal
conditions). They can also, perhaps, agree that either [99] or [98] is necessary for
U-freedom—although they are just as likely to think that neither is (I am going to
leave [98] and [99] out of account in what follows). But throughout all this they will
continue to insist that no set of agent-structural conditions can ever be enough for
U-freedom—not even when [0], the causa sui condition, has been included in the set
of agent-structural conditions—because attitudinal conditions are also required.
The Attitudinalists think that the Structuralists are failing to treat the problem of
free will with sufficient generality. The Structuralists, they think, are not sufficiently
aware of how bleak and weird and intuitively unfree creatures that fulfil all the conditions so far mentioned—[0]–[6] ± [98] or [99]—might be; Structuralists profess to
be thinking about the conditions of U-freedom in an entirely general way, but they
are not sufficiently aware of how many special features of ordinary human beings they
are taking for granted and presupposing. The Attitudinalists hold that some of these
features, properly examined, turn out to be conditions of U-freedom, attitudinal conditions that are irreducible to agent-structural conditions.
¹⁸ Frankfurt’s requirement that a free agent must be capable of forming second-order volitions
(Frankfurt 1971) is a straightforward agent-structural condition on the present view of things, and
can be included in the list of conditions of U-freedom by any agent-structuralists who wish to do so.
¹⁹ Several compatibilists have favoured [99] (see for example Hobart 1934), and all incompatibilist
libertarians have favoured [98]. There is a presumption that [0] is incompatible with [99] unless the
agent in question is God or the universe, and many red herrings lie this way.
Free Agents
7
367
THE BELIEF-IN-U-FREEDOM CONDITION
What might these attitudinal conditions be? One proposal is that
[10] believing one is a U-free agent
—conceiving oneself as a U-free agent, experiencing oneself as U-free, figuring oneself as U-free—is itself a condition of being U-free. On this view, U-freedom is not
just a matter of possessing certain agent-structural capacities. A U-free agent must also
see or experience itself, think or conceive of itself, in a certain specific way. It must
have a certain attitude to itself and its agency. It must fulfil the attitudinal belief-inU-freedom condition. (In what follows ‘belief in U-freedom’ refers only to a creature’s
belief in its own U-freedom.)
What does this amount to? The first thing to say, perhaps, is that it is not strictly
necessary to figure oneself as a specifically moral agent, or have any grasp of the notion
of morality at all, in order to fulfil the belief-in-U-freedom condition. We (human
beings) ordinarily associate the idea of U-freedom closely with moral issues—to such
an extent that we may think that to have a sense of oneself as U-free is necessarily to have some grasp of moral matters.²⁰ But this is not so. Self-conscious agents
that face difficult life-determining choices while lacking any sort of conception of
morality can have a sense of U-freedom—of radical, absolute, buck-stopping up-tome-ness in choice and action—that is just as powerful as any sense of U-freedom
grounded primarily in experience of moral requirements. They can fulfil the beliefin-U-freedom condition just as resoundingly as we do.²¹
—Perhaps, perhaps not. But what is there to back the idea that belief in U-freedom
might be a condition of U-freedom?
Well, suppose an agent a fulfils all possible agent-structural conditions of Ufreedom. It performs an action and is in no way physically or psychologically
constrained in any of the ways standardly cited in (for example) compatibilist constraint theories of freedom. Suppose we know all this, and believe that U-freedom
is possible, and take it that a has performed a U-free action. Then we discover
that it really has no sort of conception of itself as a U-free agent, no sense that it
is a U-free agent. Can we still hold it to be truly a U-free agent? Put yourself in
a’s shoes: you act, when you do, with no sense of yourself as a U-free agent. You
have no sense whatever of yourself as something that can be ultimately, radically
responsible for what you do. Is a life spent like that the life of a U-free agent,
something that can be ultimately responsible for what it does, morally deserving
without any sort of qualification of praise or blame or punishment or reward for its
actions? I don’t think so.
²⁰ Kant (1788: 4; Ak. 5. 4) took our knowledge of the moral law to be proof that we were
U-free agents—to be the means by which we could know with certainty that we were U-free—even
though we could not understand how U-freedom was possible.
²¹ See e.g. Strawson 1998: 746. I will take it that a sense of things being truly or radically up to
one is the same as, or sufficient for, a sense of oneself as U-free.
368
Free Agents
Let ‘B(a, p)’ represent ‘a believes that p’, and let ‘Fa’ represent ‘a is U-free’. The
belief-in-U-freedom condition states that
[Fa → B[a, Fa]]
—that if you are U-free then you believe you are—and that this is so because believing you are U-free is partly constitutive of being U-free, and not, say, because being
U-free is a property that is necessarily epistemically evident, like being in intense
pain. It seems paradoxical because it appears to offend against a very basic (even if
not exceptionless) epistemic principle, the principle of independence of belief and thing
believed.²² Suppose a has an ordinary factual belief p.²³ If p is true then it is true in
virtue of the obtaining of some state of affairs S1 . Obviously. The principle of independence states that the state of affairs of a’s having p—call this S2 —cannot itself be
part of S1 . If a has p and p is true in virtue of S1 then S1 cannot include or involve
S2 . S2 , a’s having p, cannot itself be among the truth-conditions of p. In general, a
true belief must be supposed to be a representation of some state of affairs essentially
other than itself; it must be supposed to be something which can always in principle
be subtracted from the world in a way that leaves its object—the state of affairs it is a
belief about—untouched.
There are, certainly, cases which appear to contravene the principle of independence: it is arguable that one makes a promise, or a knight move, or asserts that p, or
obeys an order to do x, or enters into a contract, only if one believes one does; it is
arguable that one is fashionable, or chic, only if one believes one is.²⁴ These seem to
be properties of which it is true to say that there is some kind of awareness condition on
possessing them. Roughly, it seems that you have to be aware that you have the property in order for it to be true that you have the property. And this awareness cannot
be a standard matter of your coming to form the belief that you have the property as a
consequence of the fact that you have it, because the awareness appears to be (partly)
constitutive of your actually having the property.²⁵ These properties, however, are all
properties that one has only by virtue of participating in some conventional activity
or other, and it seems that we can give a satisfactory explanation of their existence by
reference to the existence of the relevant conventions. But the case of U-freedom, we
suppose, is not like that at all. If U-freedom is real, it is certainly not real because of
the holding of some convention, human or otherwise.²⁶
²² The converse form of the general claim, [B[a, φa] → φa], is also interesting.
²³ Here ‘p’ refers to a’s actual mental state of belief (or actually entertained thought) rather than
merely to the content of a’s belief or thought considered independently of a’s mental state. I take
it that thoughts or beliefs considered as actual features of the world can be true or false just as
statements can be.
²⁴ I believe that this is not true of entering into a contract, given UK law. For other cases see
Strawson 1986: 200–25 (for the case of the ‘Mystery Draw’, in which believing you are a winner is
a necessary condition of being a winner, see pp. 207–11).
²⁵ See further Strawson 1986: 201, 303–4.
²⁶ Are there any clear non-conventional cases? Suppose I think (believe) that this very thought
(belief ) is puzzling —this mental particular, call it q, the ‘token’ thought (belief ) I am now
entertaining. In this case q is the thought (belief ) that q is puzzling; that is its content. So q
is a true thought (belief ) (if and) only if q is puzzling. So that in virtue of which q is true
Free Agents
369
This issue can seem very puzzling, but I am going to leave it here for the moment.
I am going to put [10], the belief-in-U-freedom condition, on hold in order to ask
whether there are any other attitudinal conditions of U-freedom. And to do this I
need to go back before going forward.
8
T H E A B I L I T Y- TO - C H O O S E C O N D I T I O N
In §4 I offered a base set of agent-structural conditions on U-freedom. A Ufree agent must obviously be [1] an agent, and must therefore be [2] capable of
forming beliefs of certain sorts, [3] capable of having desires or pro-attitudes, [4]
capable of self-movement or self-change, [5] capable of practical reasoning. I took
it that dogs and humans have all these properties in common, and were on a par
in so far as possession of the basic property of agenthood was concerned. But it
may now be said that the base set of conditions of agenthood—genuine intentional agenthood—is incomplete. It must be supplemented by the condition that
one be
[7] (genuinely) able to choose
that is, genuinely able to entertain and decide between alternative courses of action;
for [2]–[5] do not by themselves guarantee this.
Is this true? It depends on how one understands [5], the practical reasoning condition: on one natural view the capacity for practical reasoning already involves being
able to choose in this way. But one can also allow that [2]–[5] do not or may not guarantee possession of the ability to choose, and add in [7] as an extra (agent-structural)
condition. I will take the second course, because there is no harm in having redundancy in the list of conditions of U-freedom, given that the aim is only to state sufficient conditions.
The basic conditions of genuine agenthood now read as follows. B is an agent if
and only if B is
[2] capable of forming beliefs of certain sorts
[3] capable of having desires or pro-attitudes
[4] capable of self-movement (self-change)
[5] capable of practical reasoning
[7] (genuinely) able to choose.
All conditions of agenthood, including the new ability-to-choose condition, are a
fortiori conditions of U-freedom, and I take it that Fido can fulfil all of them.²⁷
is not fully specifiable independently of mention of q itself. This does seem perfectly possible,
and not in any deep way paradoxical, but it is entirely unlike the supposed case of U-freedom.
Crucially, it is not a case in which the claim is that to have a property you have to believe you
have it.
²⁷ It seems clear that dogs, not to mention other unself-conscious beings like Nemo, can be said
to make choices. For some further considerations in support of this, see Strawson 1986: 141–5.
Free Agents
370
9
RICH A B I L I T Y TO C H O O S E
We do not, however, think that Fido is a serious candidate for being a U-free agent,
by comparison with ourselves; and when we ask what makes the difference it is very
natural to think that one crucial thing he lacks is
[6] full self-consciousness
as defined in §5. Some, as remarked, think that self-consciousness makes all the
difference—they think it suffices all by itself to make us U-free where Fido is not
and cannot be U-free. Others like myself doubt that self-consciousness alone can
make such a dramatic difference; but there are a number of things that can be said
in support of its importance. The intuitive case for its necessity was put on p. 364,
and the introduction of the ability-to-choose condition suggests a different way of
conveying its importance. For it may now be said that the problem with Fido and
other unself-conscious beings like Nemo (p. 365) is that they cannot really choose
between alternatives—not in the strongest sense, not in the way that we can—,
and that this is so precisely because they are not fully self-conscious. Part of what
it is to choose between alternatives in the full sense of the notion of choice is to be
aware (believe) that you are able to choose, and to be aware of this in a fully selfconscious manner. In which case [6], self-consciousness, is a necessary condition
of [7], unalloyed or full ability to choose; so that while we can be truly able to
choose, dogs can’t be. On this view, fully self-conscious awareness of oneself as
able to choose when in a situation of choice doesn’t make the difference between
ability to choose and ability to choose U-freely. Rather, it makes the difference
between being (genuinely or fully) able to choose and not being (genuinely or
fully) able to choose at all.
Well, this can perhaps seem a natural thing to say—although it appears to contravene the principle of independence just as clearly as the paradoxical belief-in-Ufreedom condition does.²⁸ In fact I think it equally natural to say that Fido—and if
not Fido then Nemo—can genuinely choose between alternatives, but for expository
purposes (and because redundancy doesn’t matter when one is attempting to state
sufficient conditions) I am going to allow that
[8] rich ability to choose
is something more than that ability to choose, now reclassified as
[7] basic (genuine) ability to choose
that unself-conscious creatures may be supposed to have; noting that [8] presupposes
[6] self-consciousness
²⁸ It takes time and care to lay out the case fully. On one view, one can’t take it that the notion
of a situation of choice is independently available in the way I assume here, because one’s being
in a situation of choice is itself constituted partly by one’s sense that this is so. I pursue these
complications with great but perhaps overexcited ardour in Strawson 1986: ch. 14.
Free Agents
371
since rich ability to choose is precisely the capacity to be fully self-consciously aware
of oneself grasped as oneself as able to choose when choosing. But it does not simply
follow, from the fact that one is self-conscious and has basic ability to choose, that
one has rich ability to choose, for it is at least conceivable that one’s self-consciousness
may never turn itself in that direction (as it were). So [7] and [6] do not strictly entail
[8], and [8] needs to be listed as an independent and separate condition.
Is [8] an agent-structural condition or an attitudinal one? It doesn’t matter much
how one classifies it. Some may hold that it is best treated as an attitudinal condition because it does not really increase the agent’s agentive capacities in any way:
all it adds to what is already the case, agent-structurally speaking, is awareness that
what is the case is the case; and ability to choose and self-consciousness are already
in the set of agent-structural conditions. Others, however, may think that it is
best seen as an agent-structural condition, in as much as it genuinely enriches the
overall agentive-cognitive complexity of the agent. I think there are worthwhile arguments on both sides, but here I am going to treat [8] rich ability to choose (selfconsciousness-illuminated ability to choose) as a further agent-structural condition
of U-freedom, because I want to concede as much as possible to the Structuralists
in order to highlight the force of the Attitudinalists’ claim that there must also be
allowed to be attitudinal conditions of U-freedom.
Is rich ability to choose really necessary for U-freedom? If you already think that
it is enough to be an agent and fulfil the causa sui condition you may doubt this (in
a Perssonian spirit). But it does seem plausible that we would not judge a being to
be a U-free agent—capable of being ultimately responsible for what it does in such
a way that it can be, without any sort of qualification, morally deserving of praise or
blame or punishment or reward for its actions—if it was not even capable of being
fully self-consciously aware of itself as able to choose when choosing. So I hope you
will go along with conditions [8] (and [6]), at least for the moment.
So this is how things stand. [1]–[5]+[7] give the base set of conditions of agenthood. And if you agree that merely being an agent and being causa sui are not enough
for U-freedom then you will probably also agree that [6] and [8], self-consciousness
and rich ability to choose, are also necessary for genuine U-freedom.²⁹ So the present
proposed set of conditions of U-freedom consists of [1]–[5]+[7], the base set of conditions of agenthood, [6], full self-consciousness, [8], rich ability to choose, and [0],
the causa sui condition: [0]-[8]. The question is: Is this enough? Do we also need [10],
the belief-in-U-freedom condition?
10
S TO L I D U S
Well, one question—an interesting question—is whether [8] already entails
[10] believing you are a U-free agent.
²⁹ Once again it doesn’t matter much if they are not necessary, since we are only trying to state
sufficient conditions of U-freedom.
372
Free Agents
Perhaps to be richly able to choose, fully self-consciously aware of oneself as choosing
when choosing, is necessarily to have the sense that it is truly or radically up to one
what one chooses—in a way that simply amounts to believing you are a U-free agent.
Perhaps an agent that has rich ability to choose is bound to experience itself as Ufree, when choosing, and is therefore (on one reading of Sartre) bound to be U-free,
when choosing. On this view, there is a crucial, overpowering sense in which such a
being, fully self-conscious as it is, can know inescapably what it is doing, in the moment
of choice or action, in such a way that it is for that reason (alone)—and inevitably,
whether it likes it or not—U-free in fact.³⁰ (It is a sense in which Fido and Nemo,
not being self-conscious, do not and cannot know what they are doing.)
Is this right? I don’t think so. I don’t think [8] entails [10]. I am now going to
present some cases in support of this claim, and in order to do so I’m going to put
aside the unfulfillable causi sui condition, because it makes no difference to the cases.
Consider Stolidus. Stolidus is a being of extremely limited conception, a very blinkered personage, an agent who comes to be fully self-conscious while remaining very
stunted in his general conception of things, including in particular his conception of
his own agency. Stolidus is emotionally very dull, mentally very sluggish.³¹ He is, certainly, able to choose between two things X and Y in the way that Fido and Nemo
can, and he can (being self-conscious) come to be fully self-consciously aware of himself as able to choose between X and Y in a particular situation of choice. And he
standardly does come to be so aware of himself; he has rich ability to choose. But he
need not thereby figure himself as U-free, either in the moment of choice and action
or in general; this self-conception is not forced on him; he need not fulfil the beliefin-U-freedom condition.
He is, in a particular situation of choice, fully self-consciously aware of himself as
now able to choose between X and Y; he is aware that he will very shortly opt for and
actually perform either X or Y. But it does not follow that his thought must—ipso
facto—be informed by any sense that it is radically up to him which he does. Aware of
all that he is aware of, he need not and does not turn round upon himself and figure
himself as U-free. And he doesn’t. He simply lacks this view of things.
Well, this is surely possible. But I think it can be hard to see. It is natural for us,
human beings, to think that any genuine, fully self-conscious, explicit sense of oneself as able to choose, when choosing, just is a sense that things are radically up to
one, and so just is a sense of oneself as U-free. And yet I think it is a mistake. One
might restate the point by saying that although Stolidus goes essentially beyond any
unself-conscious (Fido-like or Nemo-like) form of awareness of choice, in being fully
self-consciously aware of himself as now able to choose between X and Y, still the ableto-chooseness part of the content of his state of awareness may remain in essentials
like Fido’s. It is simply not the case that the able-to-chooseness part of the content of a
³⁰ In its strongest form this line of thought dismisses the causa sui condition as superfluous (and
metaphysically preposterous), but for the moment we may suppose that the being in question does
also fulfil the (unfulfillable) causa sui condition.
³¹ Stolidus’ actions are ordinarily of small importance, his thought is mostly concerned with the
more or less immediate present. He moves through time inside a shell of self-concerned, short-term
aims. But these things are not necessary features of the case.
Free Agents
373
state of awareness of choice must transmute into ‘truly-up-to-me-ness’—into belief in
U-freedom—just by coming into the light of self-consciousness. Belief in U-freedom
(experience of oneself as U-free) essentially involves a further perspective, a further
attitude to what is going on, a further way of thinking of oneself that Stolidus does
not have. And it is—I propose—essentially more than the fullest and most explicit
possible self-conscious awareness or conception of oneself as something that is able
to choose what to do; whether this conception is a general standing conception one
has of what sort of thing one is, or an explicit and occurrent apprehension of what
sort of thing one is in the moment of choice. And it is not any sort of agent-structural
condition.
I think our saturated familiarity with our own case, our tendency to think that any
sense of self must be like our own, causes us to elide, and not to notice, a transition.
The natural thought is this: ‘If I am truly fully aware, now, that I am able to choose
what to do, surely I ipso facto have the experience that what I do is truly up to me, that
I am truly (and inescapably) a U-free agent in this situation? Surely there is no further
basic content to my experience of myself as U-free than my full awareness of myself
as truly, fully, able to choose?’
Stolidus suggests that this is not so. He makes the point by virtue of his extremely
limited outlook. Moira, the ‘true no-U-freedom theorist’, will make the same point,
later, by virtue of her unusually inclusive outlook. But I want to delay her entry for
the moment in order to ask whether there are any other attitudinal conditions of
U-freedom that we need to take into account before attempting to face up to the
paradoxical belief-in-U-freedom condition.
11
THEORIA
I think there are. Consider Theoria,³² the spectator subject, who fails to be a Ufree agent because of her overall emotional attitude or identificatory relation to her
agency. She fulfils all the agent-structural conditions on U-freedom listed so far,³³ but
there is something fatally amiss in her attitude to herself as an agent. She is detached
from her motivation in some curious way. She acts, and for reasons that she can give,
but it is as if it is not really she who desires, decides, and acts, but rather as if her
desires and beliefs work it out among themselves beneath her disengaged, spectatorial,
inward gaze.
Theoria resembles Camus’s étranger at his most detached.³⁴ She is somehow disengaged from life, including her own. When l’étranger, Meursault, alludes to one of his
own desires, it is half as if he were recounting a fact about a feature of the world which
is extraneous to him, for he—what he most truly is—seems to be just the detached
reporting self. The desire seems to be something that affects his life rather in the way
that things external to one, details of one’s surroundings, do. And yet it is still his
desire—it is no one else’s. It is, one might say, something he apprehends before it is
³² From θεωρειν, to be a spectator at the games.
³⁴ Camus 1942.
³³ [1]–[8]; [0] has been put aside.
374
Free Agents
something he feels, but that it is his is part of what he apprehends. There is no defect
in his self-consciousness.
The same is true of Theoria. She has no strong particular sense of herself as a
decision-making, self-governing agent. She doesn’t really see herself as the decider
and rational planner of action. Or rather, she does, for there is no defect in her
self-consciousness—her ability to grasp herself as herself—considered as a cognitive
capacity, and she is after all aware of practical-rational calculations going on in her;
but she does so only in some spectatorial manner. She doesn’t feel herself to be an
agent in the definite, vivid, participatory way in which we do. She has no real sense of
expressing herself in action, no sense (such as we ordinarily have) of having a will that
issues in action in such a way that one is responsible for and in some way committed to what one does. She is aware that her actions are actions executed by a body
that is hers, and motivated by reasons correctly identifiable as hers—they are certainly no one else’s, and they do in fact motivate this her body. She is able to give
reasons why she did something, as we do—‘I wanted X and believed doing Y was the
best way to obtain X, given Z’; and it remains correct to say that she decides to do
Y. No one else does, and she is a single psychophysical thing, and she is not schizophrenic. And yet there is something vital missing in the way she wants X, and in her
attitude to what are correctly identifiable as her projects and actions. She never has
the experience of participatory involvement in the mental stages of action-production
that we can have—the experience underlying the sense that it is really I who decide
and am responsible for my actions. And this seems to matter a great deal, when freedom, or at least U-freedom, is in question. Unlike us, she doesn’t feel herself to be the
decider and animator of action, and it seems that she thereby fails to be the decider
and animator of action in the right kind of way for her to be a U-free agent. She is
not ‘identified’ with the process. She is more like someone who watches an actionproducing process—her own—from an internal point of view. There is something
wrong with her attitude, her affective relation, to herself, so that although she is a
self-conscious agent, she is not a U-free agent.³⁵
Imagine a woman who, although she remains outwardly normal and goes about
her business, has entered into an acute state of ‘existential’ crisis, or accidie, or, more
particularly, depersonalization or aboulia. Imagine someone for whom life has lost
its point, and who acts merely mechanically in some sense, although outwardly normal, continuing to execute complicated and calculated actions, continuing with her
job, for example. To us, ignorant of her inner condition, her actions seem to be
performed by a normally free and responsible agent. But this appearance is—I propose—illusory.
Constraint theorists may say that Theoria has as good a claim to be a U-free agent
as any of us—she fulfils all the agent-structural conditions on U-freedom—and
³⁵ On Frankfurt’s theory of freedom it might be thought that Theoria could be ruled out on
agent-structural grounds—on the grounds that she could not form ‘second-order volitions’, i.e.
desires that other, first-order desires that she had should (be such as to) move her to action. But she
can have (the capacity to form) such second-order desires, in just the sense in which she can have
first-order desires; the problem is that she will be spectatorially detached from them as well.
Free Agents
375
that the reason she is not actually free in her actions is that she has a psychological
condition that counts as a freedom-removing constraint. But this is not the right way
to see her. In my story, Theoria is an alien creature, and her experience of agency
is normal among her kind—the Theoreticals. The Theoreticals, like ourselves, have
just one of the many possible kinds of experience of self, agency, and life available to
beings that fulfil the agent-structural conditions on U-freedom.
I conclude that if and in so far as we judge Theoria’s kind of experience of agency
to be insufficient for, incompatible with, U-freedom—and I think we should, even
if she fulfils the causa sui condition—we place an attitudinal condition on freedom,
a condition whose fulfilment involves something essentially more than whatever is
involved in fulfilling the agent-structural conditions. What should I call it? The integration condition?³⁶ The non-alienation condition? I will call it the engagement condition: one must be
[11] engaged
if one is to be a U-free agent. In her detachment Theoria fails the engagement condition. If complete lack of any such sense of engagement could be decisively established
in a human being (here it is ex hypothesi) it might suffice even in a court of law to
absolve from responsibility.
Let me repeat the point that the Theoreticals’ experience of agency is merely different from, and in no way inferior to, or somehow incorrect relative to, our own.
It cannot be held to involve some form of psychological constraint just because it is
different in this way. What Theoria’s case shows, I suggest, is that there can be fully
self-conscious agents as complex as we are who cannot be said to be U-free agents
simply on account of the character of their experiential attitude to themselves and
their agency. It looks as if Theoria will also fail the belief-in-U-freedom condition,
but I do not think one has to show this in order to say what it is about her precludes
her being a U-free agent in spite of fulfilling all the agent-structural conditions.
The general conclusion is that U-freedom cannot consist merely in the fulfilment
of some set of agent-structural conditions: it cannot consist in the possession of some
perhaps as yet undiscovered set of maximally action-enabling cognitive and practical
capacities. For we also require a certain attitude to self and agency, an experiential disposition which it is very hard to specify precisely in a positive way, but which Theoria
at any rate does not have.
12
T H E N AT U R A L E PI C T E TA N S
Suppose we accept [11], engagement, as an attitudinal condition on U-freedom, a
condition that must be added to any plausible set of agent-structural conditions. And
suppose we put aside the controversial belief-in-U-freedom condition. The question
is then whether the engagement condition is the only attitudinal condition that is
needed.
³⁶ I called it this in Freedom and Belief.
376
Free Agents
Imagine an enormously congenial world inhabited by a race of gifted, active
creatures. They fulfil all the agent-structural conditions of U-freedom so far proposed, and the (attitudinal) engagement condition. What is unusual about them is
that they are never undecided in any way. They are fully—richly—able to choose,
but they never hesitate at all about what to do. They never ponder alternatives,
although they are perfectly well aware of them. They never consciously deliberate
about which ends to pursue or about how to pursue them (they have no need to) and
always succeed in doing what they want to do.
These are the Natural Epictetans —never failing, never disappointed in their congenial world, always able to do what they want to do because always wanting to do
only what they are able to do.³⁷ Their experience is radically unlike ours. But they are
not like Theoria. The strangeness of their experience (relative to ours) does not derive
from a failure to fulfil the engagement condition.
The Natural Epictetans take decisions—they are agents responding variously to
circumstances in complex ways, beings who actually act—but they are never undecided. Decision is not something they ever apply themselves to in any way. It is not
something they live through or dwell on or notice as such. Decision does not issue
from previous indecision. It is not a resolution or conclusion of anything. Perhaps
their world is much less complicated than ours, but it need not be.³⁸ Their deciding
is in any case effortlessly smooth. They never wonder or worry ‘Which (action) shall
I perform?’.³⁹ Given their capacities, they are always capable of acting otherwise than
they do in fact act, and they are fully capable of the thought ‘I could do otherwise’.
Any natural Epictetan can fully understand a story in which it is faced with a button,
and knows that pushing or not pushing it in the next twenty seconds will lead to very
different results. But we may suppose that they never in fact think such a thought as
‘I could do otherwise’. And if some of them, musing philosophically, were to think it,
it would have no energy for them, no relations with the notion or experience of difficult choice; for they do not know what it is to face or make such choices. It would in
effect have no more import for them than either ‘I have such and such capabilities’ or
‘I might have been, or wanted, otherwise’.
The Natural Epictetans, then, can never be frustrated. They are never in a situation
in which they think ‘I could have done A (could do A), if it had not been (were not)
the case that C’ with a sense of external constraint or hindrance. If a natural Epictetan
were to encounter constraint or hindrance, in spite of the fabulous congeniality of its
world, so that it found itself unable to do what it had until that moment wanted to
³⁷ Epictetus enjoins one to adapt one’s desires to one’s circumstances, so that one is never
frustrated by them.
³⁸ They may be extraordinarily intelligent, and find the complexity simple. If God existed, and
were an agent, his experience of agency might be like a Natural Epictetan’s.
³⁹ A creature that was never undecided as to what to do might experience freedom =
unhinderedness of action in unconstrained circumstances solely as a result of the contrast afforded to
such free = unhindered action by experience of hindering constraints in other circumstances. And
if it lacked the immediate Epictetan reflex it might experience hindrance in doing what it wanted to
do and think ‘I could do (have done) A if it were not (had not been) for C’. It might even reason as
follows: ‘Since I cannot do A, I will do B’; or again, thinking ahead, ‘If I cannot do A, I will do B’.
But, thinking these things and these things only, it need never have any sense of U-freedom at all.
Free Agents
377
do, then it would by virtue of the immediate Epictetan reflex always immediately cease
to want to do what it found itself unable to do.
External constraint is one thing; indecision is even more important. The Natural
Epictetans are never in a situation in which they might think ‘I could do A, and I
could do B, and I can’t do both, so which shall I do?’. They are fully capable of thinking such a thing, cognitively speaking, but they never do so in fact, given the automatic natural Epictetan nature of their volition.
In one way, then, they enjoy all the freedom that it is possible for any being to have.
And yet there seems to be a sense in which they are not U-free agents at all because
they do not know what U-freedom is. They seem, with respect to the freedom they
enjoy, like creatures that have ears but live in a soundless world. Such creatures experience total silence—they hear it in the sense that we can be said to hear it—but we feel
that they do not know what silence is. And if they do not know what it is then there is
a sense in which they do not and cannot experience it as we who know noise can and
do. In the same way, sighted creatures living in a lightless world may have no concept
of darkness. Living in a uniformly blue world, with no differences of shade resulting
from shadow or distance, they may have no concept of blue or of other colours.
But what do these cases show? If one takes two things A and B (such as freedom
and lack of freedom) which contrast in such a way that there is either A or B but not
both, the complete lack of B should not have as consequence that there is no A. On
the contrary; there is nothing but A. What may lack in such a case, however, is any
experience of the contrast between A and B. And this may lead, as above, to lack of
any explicit awareness or conception of either A or B.
But why should lack of explicit awareness of A be supposed to have as a consequence lack of A itself, as lack of any sense or conception of U-freedom seems
to have as a consequence lack of U-freedom itself? Well, that is the question. But
it does seem to be so. (Compare the sense in which Adam and Eve, when innocent, were not and could not be good.) I do not think one can simply say that the
Natural Epictetans are in fact U-free agents, and that their coming to know this
would be a standard case of a belief being formed in such a way as to represent
or reflect an already existing belief-independent fact. That does not seem true to
how things are. There seems to be a sense in which the Natural Epictetans are
not U-free agents at all, precisely because—it is natural to put it in this paradoxical way—they don’t have a proper grasp of the fact that they are. Even when we
put aside the strong (causa sui requiring) notion of U-freedom for some weaker
notion of freedom we may feel that their sense of freedom lacks some essential
contrastive grounding, either in the form of experience of constraint, coercion and
inability, or, quite differently, and more importantly, in the form of experience
of indecision, of having to make difficult choices between alternatives. Certainly
the Epictetans know that they are fully free agents in the compatibilist sense: they
are utterly unconstrained and fully self-conscious and do what they want to do
and know they do. But if it is claimed that they are in these conditions bound to
figure themselves as U-free, this is far from clear. If there is any sense in which
they experience themselves as U-free, I think it will be like the sense in which the
inhabitants of the silent world experience their world as a silent world.
378
Free Agents
It may again be suggested that they must at least be able to conceive of doing
otherwise, of constraint, and even of experiencing indecision; whereas the inhabitants
of the silent world cannot conceive of noise at all. It may be suggested that ability to
conceive of the possibility of indecision, and indeed of choice-out-of-indecision, or
at least grasp of the notion of options being open to one, is essentially constitutive of
(and so entailed by) rich ability to choose, which the Natural Epictetans possess by
hypothesis. But one may grant this, for the purpose of argument. For it is not plausible that mere possession of the ability to conceive of difficult choice or indecision
without any actual experience of it—nor even any actual thought about the possibility of it—is bound to precipitate belief in U-freedom of the kind we appear to require
of genuinely U-free agents. Having no natural occasion, the Natural Epictetans may
never formulate the possibility of indecision. They are not speculative creatures. They
are without anxiety. Never having had conflicting desires, never having experienced
indecision, an Epictetan never deploys its general conception of itself as capable of
doing many different things in the thought that it is now able to do other than what
it is now about to do in such a way as to encounter the strong, immediate sense of
open choice that is so central to our lives.
In fact it is not particularly important whether an Epictetan ever actually entertains
any thought of indecision, or of the possibility of doing otherwise. What is important is whether it ever dwells on such things, whether they ever matter to it in a certain
way. And it is a consequence of the description of their case that this does not happen.
The Natural Epictetans know they produce the actions they produce, they know they
are causally responsible for them in that sense, but still they lack any sense of themselves as U-free in their actions. Agency, for them, is something like what it is like
for us when we are involved in doing something that we entirely unproblematically
want to do, and that there is only one way of doing, and that we can do without any
difficulty; something that involves performing a long sequence of simple intentional
actions with regard to which we have no sort of thought that there might be alternative ways of doing them. Here there can be a strange resistlessness, no consciousness
of choice or indecision.
The proposal, then, is that the Natural Epictetans aren’t U-free agents although
they are fully self-conscious, engaged agents: they can’t be U-free agents—even if
they fulfil the causa sui condition—because they have no sense of what U-freedom
is. They fail to fulfil condition [10], the belief in freedom condition; they fail to figure
themselves as U-free.⁴⁰ Some may say that to be free in such a way that one doesn’t
even know that one is is the truest freedom of all: ‘the sense of liberty is a message
read between the lines of constraint. Real liberty is as transparent, as odourless and
tasteless, as water’.⁴¹ This is a good thought, but the other thought seems to remain
undiminished: to have no sense of U-freedom at all is not to be a U-free agent.
⁴⁰ They may even have second-order desires, and what Frankfurt calls second-order volitions—desires that their first-order desires be such that they move or will or would move them to
act, all of which are effortlessly satisfied.
⁴¹ Frayn 1974: §85.
Free Agents
379
Here a further point comes into play. If U-freedom requires being causa sui, and
if being causa sui is impossible, then U-freedom is impossible. So any belief that
one is a U-free agent is false. Now it seems that nearly all human beings do believe
they are U-free agents (although many philosophers try to pretend otherwise); and
it may be that they—we—can hardly help this. But it would be very surprising
indeed if all possible self-conscious agents that had rich ability to choose—all possible
self-conscious agents that were able to choose and knew it—were by that fact alone
compelled to believe that they possessed such U-freedom. For it would be very surprising if all possible self-conscious agents, however intelligent and knowledgeable
they were, were compelled to believe something demonstrably false. It may be that
belief in U-freedom is a necessary condition of U-freedom even if U-freedom is
demonstrably impossible (being round is a necessary condition of being a round
square), and even if any belief in U-freedom is demonstrably false. Stranger things
have been the case. But one point about Theoria and the Natural Epictetans that must
now be registered is that whatever their defects—whatever we may think they lack
when we compare them with ourselves—it is most unclear that failure to have a false
belief is ever a defect.
There is perhaps a connection here with accounts of spiritually advanced states of
mind, and, more particularly, accounts of what the experience of choice and agency
is like in such circumstances. Krishnamurti puts it as follows:
You do not choose, you do not decide when you see things very clearly; then you act [in a
way which] which is not the action of will. . . . Only the unintelligent mind exercises choice
in life. . . . A truly intelligent [spiritually advanced] man can have no choice, because his mind
can be aware of what is true, and can thus only choose the path of truth. It simply cannot have
choice. Only the unintelligent mind has free will.⁴²
Saul Bellow has a related thought:
In the next realm, where things are clearer, clarity eats into freedom. We are free on earth
because of cloudiness, because of error, because of marvellous limitation. . . . ⁴³
—Surely such clarity and such intelligence, which presumably rule out belief in Ufreedom, must bring U-freedom with them if anything does? And if so, belief in Ufreedom isn’t after all a necessary condition of U-freedom.
The trouble is that nothing brings U-freedom with it, because U-freedom is impossible. The highest forms of Krishnamurtian intelligence will indeed rule out belief in
U-freedom, because they will involve clear understanding that U-freedom is impossible.⁴⁴ But this will not show that belief in U-freedom is not a necessary condition of
U-freedom.
⁴² Krishnamurti, quoted in Lutyens (1983: 33, 204). In a similar vein, perhaps, Spinoza remarks
that ‘God . . . cannot be said . . . to act from freedom of the will’ (1675: pt. 1, prop. 32, coroll. 1).
⁴³ 1977: 140. I do not think that Spinoza disagrees when he says that freedom is consciousness of
necessity, although he is considering a somewhat different conception of freedom. See S. Hampshire
1972: 198 ff.
⁴⁴ A stage on the way to Nietzsche’s amor fati.
Free Agents
380
Einstein backs up Krishnamurti and Bellow, judging that ‘a Being endowed with
higher insight and more perfect intelligence, watching man and his doings, would
smile about man’s illusion that he was acting according to his own free will’.⁴⁵ The
basic argument that U-freedom is impossible is entirely a priori, but there are also
extremely strong a posteriori reasons for thinking it impossible. It seems unavoidable
if Einstein’s theory of special relativity is anything like correct, for example—a point
that has received little discussion in recent debate about free will.
13
INDECISION
Suppose it granted that the Natural Epictetans are not U-free agents. The question,
then, is what they lack. ‘Well, they cannot be genuine U-free agents because they do
not figure themselves as U-free—they fail [10], the belief-in-U-freedom condition.’
Fine, but I want to continue to put aside the belief-in-U-freedom condition for the
moment. Is there anything else they lack? Is there anything that would tip them over
into fulfilling the belief-in-U-freedom condition?
I do not think there is; but the best candidate, perhaps, is this. If one asks about
the deep sources of belief in U-freedom, experience of indecision seems, as remarked,
far more important than experience of constraint or hindrance. Suppose, contrary to
hypothesis, that the Natural Epictetans experience constraint, and have thoughts of
the form ‘I could have done A, if it had not been for C.’ Suppose further that the
experience of such constraint serves as a contrastive foil against which other, unconstrained actions are experienced by them as free—as free = unhindered. This doesn’t
advance them much, I think, for what they may still lack in this case is any sort of
experience of indecision, of difficult choice. It is that that promises most powerfully
to give rise to experience of things as radically ‘up to one’.⁴⁶
14
MOIRA
No doubt—but the question is now this. Is vivid experience of indecision sufficient
for belief in U-freedom? Does it compel belief in U-freedom? The answer, I think, is
No. When we picture an agent that is fully self-conscious, engaged, richly (fully selfconsciously) able to choose, and that is currently experiencing itself as vividly, even
agonizingly, undecided as to what to do, we may wonder how it can fail to figure
itself as U-free, as able to choose U-freely. It is at this point that we have to reckon
⁴⁵ Einstein 1931. For an excellent presentation of the a-posteriori point see Putnam 1967.
Lockwood (2005: ch. 3) effectively rebuts Putnam’s critics.
⁴⁶ Although it seems that it is precisely because they lack any sort of experience of indecision
that the natural Epictetans fail to figure themselves as U-free, it would be very implausible to
suggest that it is actually impossible to have a sense of oneself as U-free if one does not experience
indecision. There is nothing incoherent about the neo-Epictetans, who are exactly like the natural
Epictetans—they never actually experience indecision—except precisely for the fact that they do
experience themselves as U-free.
Free Agents
381
with Moira, who is a true no-U-freedom theorist. Moira sees that U-freedom requires
being causa sui, she sees that being causa sui is impossible, and she lives in the full
light of this understanding—in spite of being an agent who makes difficult choices
and acts in the world.⁴⁷
Moira is fully self-conscious and she is fully engaged. Experience of desires of the
kind that the engagement condition guarantees, and a strong interest in their fulfilment, are not incompatible with true—lived—espousal of the no-U-freedom position. And we can put her, too, in front of a button in such a way that she cannot fail
to be aware that she faces a choice. But she is so deeply intimate with the thought
that everything that she is is ultimately not self-determined that she has no sort of
sense of herself as U-free when facing the momentous button—as U-free to choose
in such a way as to be ultimately morally responsible for her choice—, although she
now knows that she is able to choose what to do. She may well find herself calculating consequences and very uncertain what to do, but she is not compelled by this into
any sense of herself as U-free in a way she knows to be impossible.
One might say that Moira is a sophisticated fatalist.⁴⁸ I think that it is extremely
hard for us to imagine what it is like to be her. It requires a conception of things
that is way outside the normal human range (even if not unattainable by human
beings). And yet I think that it must be allowed that she could exist, and that rich
ability to choose cannot in and of itself necessitate belief in U-freedom. There is, after
all, a clear sense in which U-freedom is demonstrably impossible, and it would—as
remarked—be very strange if it were metaphysically impossible for there to be an
agent that was, like Moira, both able to comprehend this impossibility and able to
grasp that it was in a situation in which it was able to choose.
This is really the only argument that can be given for Moira’s possibility, but it is
perhaps enough. Here as so often in philosophy argument seems to be no substitute
for imagination directed onto a given description.
Some may still think that [1]–[8]+[11], rich, fully engaged ability to choose, must
simply amount to [10], belief in U-freedom: that the conviction of U-freedom really
must be inescapable in these circumstances in the way Sartre supposed. It is very natural to think this, and yet I think it is unwarranted.⁴⁹ If in the present case we cannot
see the gap between [8] and [10] it may be because we can’t imagine a human being
who is a true no-U-freedom theorist like Moira but is in other respects—including
⁴⁷ Her sisters—Pepromene, Eimarmene, and Chreousa—are true or genuine incompatibilist
determinists of the kind discussed in Freedom and Belief (281–4). They make the present point as
well as Moira, but their position depends on a belief—in determinism—that could be false.
⁴⁸ Naive fatalism holds that there is no point in deliberating or in doing anything because
everything is predetermined (or is the will of God) in such a way that nothing you can do can
change how things will be. It is false because one’s doings and deliberations can change things, being
themselves real parts of the (possibly deterministic) causal process. Sophisticated fatalism doesn’t
make this mistake. It consists in the attempt (or is the result of a successful attempt) to comprehend
fully the fact that one is not and cannot be ultimately self-determined or self-determining. It
seems, though, that if one is an ordinary human being, one simply cannot attain the perspective of
sophisticated fatalism in one’s daily life.
⁴⁹ This, perhaps, is the true—rarely acknowledged—heart of the philosophical problem of free
will.
Free Agents
382
internal phenomenological respects—normal. But it is not surprising that we cannot
imagine this, because it is impossible: if Moira is human, she is not normal at all.
Either we must suppose that her sense of self is not recognizably human, or we must
suppose that it has dissolved away entirely. The first supposition would seem to be a
reasonable one, for she is a self-conscious, embodied being, and can have an unexceptionable grasp of the simple truth that she is a single thing in the world, in addition
to possessing that self-presence of mind, normal in the self-conscious, that involves
a sense of oneself as something that is somehow single just qua mental. But, granted
that she does have a sense of self, this cannot be supposed to fall within the range of
the recognizably human. A basic sense of oneself as U-free when unconstrained, as
possessed of radical ‘up-to-me-ness’, seems indissociable from the ordinary, sane and
sober adult human sense of self, and Moira’s sense of self is not like that at all. It cannot be, given the nature of her experience of choice and agency: she has no trace of
belief in U-freedom.
The line between rich ability to choose and experience of oneself as U-free may
appear a fine one in a book written for human beings, and by one, for we may be
unable to see the distinction easily, and may have difficulty in employing the device
of sympathetic identification to the case. But in a book written for creatures of some
other planet, perfectly fatal creatures of a deeply deterministical persuasion, the difficulty might rather be to explain how any apparently intelligent race could suppose
the line between having and believing oneself to have rich ability to choose (a property widely instantiated, both on Earth and on this other planet) and believing oneself
to be a U-free agent (an impossible property) could be thought to be a fine one.
Even this may be hard for us to accept, for there remains something very powerful about the Kantian or quasi-Kantian idea that any rational agent that is fully selfconsciously conscious of being able to choose cannot but suppose itself to be a U-free
agent by reason of that consciousness alone. But, having considered Stolidus, Theoria
and the Theoreticals, Meursault and his variants, the Natural Epictetans, Moira and
her sisters, I submit that this idea is nevertheless wrong. And for all that I have painted
Moira as not recognizably human, I don’t suppose Einstein differed much from her
in respect of his attitude to free will.
15
M A X I M A L A B I L I T Y TO C H O O S E
—All right. Rich ability to choose as so far defined⁵⁰ may not be sufficient for belief
in U-freedom, but you have not yet described the fullest form of ability to choose:
maximal ability to choose. Maximal ability to choose is indeed sufficient for belief in
U-freedom, and it involves something more than rich ability to choose as currently
defined. For it involves a certain sort of active deployment of the concept of choice. It
emerges as something more than rich ability to choose precisely when someone like
you goes in for thought-experiments designed to track the minimal case, as here.
⁵⁰ ‘The capacity to be fully self-consciously aware of oneself grasped as oneself as able to choose
when choosing’.
Free Agents
383
Maximal ability to choose is nothing special or mysterious. It’s a widely possessed
property. The trouble with Stolidus and the Natural Epictetans is that they don’t really
have—deploy—the concept of choice at all, and therefore do not have maximal ability to choose. As for Moira, either she too doesn’t have it, or she is not what she
claims to be: it is impossible for a being that genuinely possesses and deploys the
concept of choice to be a true no-U-freedom theorist. Fully to possess and deploy the
concept of choice is necessarily to have something more than that full-consciousnesson-the-point-of-action-of-the-fact-of-facing-action-alternatives that has been allowed
to Stolidus and his more enlightened companions. If the concept of choice is properly
active in experience of rich ability to choose then rich ability to choose is maximal ability to choose—which amounts to belief in U-freedom. The concept of choice adds to
the minimal fully self-conscious option-facing awareness allowed even to Stolidus the
essentially belief-in-U-freedom-involving idea that I can (and do—and must) pick the
option I favour, that what happens lies in my hands, is truly up to me, and so on. It is
given in the very concept of choice. So you need to add
[9] maximal ability to choose
to
[8] rich ability to choose
as a condition of U-freedom—or simply replace [8] by [9]. And then you must take
account of the fact that fulfilling [9] is sufficient for fulfilling [10], the belief-in-Ufreedom condition.’
To which the reply is that Moira certainly has and deploys the concept of choice. She
blocks the claim that [9] entails [10] in any form in which it does not beg the question
by simply building belief in U-freedom into any possession of the concept of choice
that is allowed to count as ‘full’ or ‘genuine’.
—Your assertion that she is possible, possessing [9] and lacking [10], also begs the
question.
Well, but I am fond of her; and she is backed by the thought that being causa
sui is provably impossible, and that it is very hard to believe that it could be
metaphysically impossible that any (finite) self-conscious rational agent should
believe—live—the truth.
The first reply, then, is simply that Moira as described does fulfil condition [9], but
not condition [10]. The second, more determined reply is that it still seems wrong
to say that anything at all is lacking, intellectually or cognitively speaking, even in
Stolidus’s case, so far as possession of the concept of choice is concerned. His selfconscious awareness that he faces options is his being aware that he (he) has actionalternatives. He knows what it is to act, and that he can act. He knows, in sum, that
he is able to choose between different courses of action. If we still think he lacks something this is probably because we blend the concept of choice with affective or noncognitive elements that make it hard for us to see that he has all that is cognitively
necessary, so far as possession of the concept of choice is concerned. We build an attitudinal element into what we claim to be a merely agent-structural condition.
The same goes for the Natural Epictetans, whatever the resistlessness of their
agency. We can put them all—Stolidus, Moira, Theoria, the Natural Epictetans—in
Free Agents
384
front of the momentous button. They are intellectually fully aware of the nature of
the situation. It is just that they do not live it in anything like the way we do. They
experience being able to choose, and in a fully self-conscious manner, but they do not
experience themselves as U-free. But this (as remarked) can hardly be a defect, since
U-freedom is impossible.
16
I S B E L I E F I N U - F R E E D O M R E A L LY A C O N D I T I O N
OF U-FREEDOM?
—But you put this impossibility to one side at the beginning of §10, before introducing Stolidus, and it is still to one side. So we may suppose, for purposes of argument, that all these creatures are in fact causa sui in the required way. In which case
Moira is just wrong about the impossibility of being causa sui. She herself is in fact
truly causa sui, although she believes she is not (she has no memory of having done
whatever is necessary to being causa sui ), and has, ex hypothesi, no sense of being Ufree. Surely we must then hold her to be U-free. I will grant for the sake of argument
that no unself-conscious creature is U-free even if it is appropriately causa sui ; I will put
aside Stolidus, Theoria, even, if you like, the resistless Natural Epictetans. But Moira—
entirely self-aware, engaged, maximally able to choose and causa sui . . . surely she is
indeed U-free, ultimately morally responsible, without any qualification whatever, for
her actions? She has, by hypothesis, no sense or belief that she is U-free. But how can
she not simply be wrong, in this case?
This, in effect, is the objection Nagel raised in 1987. After summarizing my position—that if a man never really experienced or believed himself to be U-free then
even if all the other conditions of U-freedom were met (including the causa sui condition), he would still not be U-free and would not be truly or ultimately responsible—he objects to it as follows:
this is a difficult claim to assess, since it involves a counter-possible conditional: but it doesn’t
seem right to me. If true responsibility [U-freedom] were possible, couldn’t someone be
deceived (even self-deceived) about whether he had it? Couldn’t he act with the illusion
that all this was just happening to him, while actually it was his doing, and he was fully
responsible for [U-free in] a choice which saved or failed to save ten other people from
torture?⁵¹
In reply, note first that this description somewhat misrepresents the case, for in not
believing that one is U-free one does not think that what is happening when one acts
is ‘just happening’ to one. One is, like Moira, fully aware when acting, that it is indeed
oneself that is doing what is being done and that one is performing an intentional
action; it is just that one feels oneself to be entirely clear on the point that it is not
possible to be ultimately morally responsible for what one does. Now one is in fact
wrong about this, given the present ‘counter-possible’ hypothesis, and in that sense
⁵¹ 1987: 5–6; Persson 1987 has a similar doubt. The example of torture is used in Freedom and
Belief (p. 103) in an attempt to dramatize the idea that there are situations in which one could not
possibly fail to feel U-free.
Free Agents
385
one is under an illusion, but one’s illusion is not that one’s action is ‘just happening’
to one.
This, though, is a minor misunderstanding, and the core of Nagel’s question is
simply this:
Couldn’t this person act with the conviction that he is not U-free, not able to be ultimately
morally responsible for what he does, while actually being causa sui and so U-free in his choices
and actions?
I have said No. If he really has no sense at all of being U-free, then he cannot be Ufree, even if he is causa sui. But I am not sure what to say if you disagree. Here we
are reaching the end of argument, and it may be best to stop. But we can perhaps
expose a little more structure by considering one more addition to the proposed set
of conditions of U-freedom—which state, so far, that to be U-free one must be
[0] causa sui
[1] an agent
[2] capable of forming beliefs
[3] capable of having desires (pro-attitudes)
[4] capable of self-movement (self-change)
[5] capable of practical reasoning (in the inclusive sense noted on p. 363)
[6] fully self-conscious
[7] possessed of basic ability to choose (p. 369)
[8] possessed of rich ability to choose (p. 370)
[9] possessed of maximal ability to choose (p. 382)
[10] possessed of the belief or sense that one is U-free
[11] engaged (in the special sense defined on pp. 373–5).
Consider, then, the proposal that one must be
[12] possessed of the conception of U-freedom
whether or not one applies it to oneself and thereby fulfils condition [10].⁵² [12] is
entailed by [10],⁵³ but does not entail it, so we can consider a Moira who fulfils [12]
but not [10], and a Moira who fulfils neither [10] nor [12].
The point of doing so is simply that some may be prepared to allow that [12] may
be a condition of U-freedom, while denying, with Nagel, that [10] is. If Moira fulfils [12] but not [10], they may say, the error she makes about herself, in failing to
recognize that she possesses a property of which she has a perfectly good grasp, simply
does not remove her from the company of the U-free. Even if she has no sort of belief
that she is U-free, she is nonetheless capable of being ultimately morally responsible—without any qualification whatever—for what she does, given that she fulfils
⁵² To possess the conception of U-freedom is not to have thought the notion through in such a
way as to see that it is impossible. One may possibly have done this but few of those who actually
possess it have done so.
⁵³ Just as [2]–[7], at least, are entailed by [1].
Free Agents
386
[12]. But if she fails [12], and a fortiori [10]—if she has no sort of conception of
U-freedom, and a fortiori no belief that she herself is U-free—then even if she is causa
sui she cannot really be counted among the company of the U-free.
You may be wondering why [12] on its own should make such a difference. I sympathize—and although I’m prepared for the sake of argument to let [12] classify as
a merely cognitive requirement, a merely agent-structural condition of U-freedom, it
is arguable that the cognitively most powerful minds won’t actually be able to understand the conception of U-freedom that is in question (for them it will be as difficult
as visualizing a round square is for us). Which leads me to wonder whether grasp of
the conception isn’t essentially partly an attitudinal matter, in some covert way, and
not a merely cognitive matter.⁵⁴
17
C O N C LU S I O N
What should one make of all this? I’m really not sure. I find I still hold the view I
formed in 1979. I think there are ineliminable attitudinal conditions on U-freedom
in addition to structural conditions and, in particular, that if anything is truly U-free
then this is essentially partly because it believes it is. One is U-free, if one is, partly
because one sees oneself and one’s action in a certain way—as U-free. The proposal
is that this is an outright metaphysical condition of U-freedom, a constitutive condition of U-freedom, although it is an attitudinal condition, and a paradoxical one at
that, given that it offends against the principle of independence.⁵⁵ U-freedom is not
just a matter of certain practical capacities, on this view. A U-free agent must see or
conceive itself in a certain specific way, and its seeing itself in this way is not a necessary consequence of its possession of any set of abilities or capacities or attitudes that
does not include this way of seeing itself.⁵⁶
⁵⁴ A final proposal might be that one must not only have a general, standing sense that one is
U-free, but must also be [13] capable of having one’s outlook animated or informed by an explicit
sense of oneself as U-free when choosing or acting; this must in fact be one’s standard condition,
when engaging in choice or actions of moment. [13] is triggered by the possibility, so far left open,
that a being might fulfil [10] in general while being somehow incapable of having any trace of any
explicit thought of itself as U-free in the moment of choice or action.
⁵⁵ p. 368. For a suggestion about how to avoid contravention of the principle of independence
see Strawson 1986: ch. 15.
⁵⁶ I’m most grateful to Gideon Yaffe for sending me written comments on this paper after I was
unable to present it at the Werkmeister Conference on Causation and Free Will at Florida State
University in January 2002.