FREE WILL AND
THE KNOWLEDGE CONDITION
by Eddy Nahmias
This dissertation argues that free will is constituted by a set of cognitive abilities and that
free will is threatened not by determinism but perhaps by certain empirical theories. In Chapter
1, I argue that the question of free will should not be focused on the question of whether or not
determinism is true. Rather, we should take a position (compatriotism) which is neutral about
the logical relations between free will and determinism/indeterminism. Indeterminism does not
necessarily threaten free will, and most of the threats thought to be posed by determinism are in
fact distinct from that metaphysical thesis. The remaining threats posed by determinism cannot
be answered by any legitimate libertarian theory.
Instead, I suggest in Chapter 2 that free will should be analyzed in terms of the
Knowledge Condition: our ability to know ourselves—specifically to know our conflicting
motivations, to know which of them we really want to move us, and to know how to act
accordingly. I analyze these cognitive abilities in light of Frankfurt’s theory of identification,
and I discuss the relations between free will, free action, and responsibility.
The sort of knowledge required by this theory of free will may be threatened by empirical
theories about human nature. In Chapter 3, I specifically examine theories and experiments in
social psychology that suggest we have a limited understanding of why we do what we do, and I
offer some responses to these threats.
The knowledge required for free will may also be threatened by certain philosophical
theories about the nature of the human mind. In the Epilogue I discuss eliminativism and
epiphenomenalism as case studies for such threats. I conclude by suggesting that we may make
significant progress on the problem of free will if we come to better understand the nature of
Finally, I offer an Appendix in which I discuss the evolution and development of the
cognitive abilities required for us to have free will, specifically our ability to understand our own
Situating the Free Will Problem
1. The Power Condition and the Knowledge Condition
Most people believe we have free will. They believe that, if we did not have free will, we
would not be responsible for our actions—and perhaps our lives would have no meaning. Some
philosophers, however, argue that there is no such thing as free will; even more argue that, if a
certain metaphysical thesis about the nature of causation (determinism) turns out to be true, then
we do not have free will. Are these philosophers talking about the sort of free will ordinary
people think they have, the sort they want to have for life to be meaningful? I don’t think so. In
this dissertation I will describe some of the features that are essential to most people’s
conception of free will. In this chapter I will argue that the metaphysical thesis of determinism is
not relevant to this conception.
What, then, are we talking about when we say we have free will? I think we are saying
that we possess certain abilities—abilities that are uniquely expressed in humans and that we
exercise with varying degrees of success when we make decisions and perform actions. These
abilities involve knowledge and power. When we act of our own free will, we express these
abilities: we know what it is we really want and what we need to do to satisfy that desire; and we
have the power to act on that desire.
We feel our freedom threatened when we experience a lack of knowledge or power or
both. Often we feel constrained when we cannot get what we want, because some external force
frustrates our ability to act on our desires. You want to leave the room but the door is locked.
You want to get home but the traffic jam blocks your way. Some external forces are other
people: police, parents, and politicians sometimes control us against our will (though they may
also be crucial in securing our freedom). External forces and agents can undermine our power to
get what we want. So can internal forces. Addictions, phobias, compulsive habits, and
unconscious drives may make us act in ways we don’t really want to. Some moods and emotions
make us act “despite ourselves”—the fit of anger, the paralyzing depression, the overwhelming
lust. Hence, both external and internal forces may constrain our power to get what we really
But sometimes we feel our freedom threatened by another type of internal conflict: we
just don’t know what we really want or how to determine what we should do. We may feel
pulled by conflicting desires and thus torn by the realization that we cannot follow two paths at
once; we know we need to choose but we don’t know how to resolve the conflict. Or we may
believe that we should act in one way but feel like acting in another way, and thus not know what
to do. You may experience a conflict because you want both to read and to watch TV and you
don’t know which you want to do more, and you may experience a conflict because you think
you should help your friend move but you feel like going to the lake, and you don’t know
whether to act on your obligations or your inclinations. If you don’t know what you really want
to do, you can’t act on what you really want—even if you had the power to.
Philosophers often analyze free will in terms of the Power Condition. It lies behind the
basic libertarian view requiring that free agents have the power (a) to do what they choose and
also (b) to choose otherwise (under the exact same conditions). And it also represents the basic
compatibilist view of free will: we are free if we have the power to carry out what we want to
do—if our ability to act on our desires or intentions is not constrained (even if we lack the power
to do otherwise in exactly the same conditions). Knowledge, however, is an important, and often
overlooked, component of these abilities.1 To be able to act on what you really want, it helps to
know what you really want, and it helps to know how your desires are related to how you will
act, and to know how particular actions are likely to satisfy what you want. You must have some
understanding of how desires relate to actions and of how actions affect the world.
We often excuse people to the degree that they are ignorant of the relationship between
their intentions and their situation; for example, when a nurse gives a patient mislabeled
medication (though we may blame him if we have reason to believe he should have recognized
the mistake). As Aristotle says of voluntary actions, “the initiative lies with the agent who knows
the particular situation in which the action is performed.”2 Our legal system also recognizes that
a person is not responsible if he “is unable to understand what he is doing and to control his
conduct.” The M’Naghten rule for insanity excuses the accused if he does not “know what he
was doing [or] know that it was wrong.”3 Thus the Power Condition, in addition to involving
control and self-control, already involves significant types of knowledge about the world and
yourself. The motto “knowledge is power” generally refers to the idea that increased
information about yourself and the world increases your ability to satisfy your desires and goals.4
While most philosophers have focused on the Power Condition, I will focus on the
Knowledge Condition. Knowing what we really want has been largely overlooked as a condition
for free will, even though in some ways it is more important than power since it gets the ball
rolling; we cannot try to get what we really want until we know what to aim for and how to act to
get it. The Knowledge Condition involves several cognitive abilities. First, we must have
introspective access to at least some of our most important desires; we must recognize what
desires we have, how they conflict, and how they might motivate us to act. Second, we must be
able to care about how we act and how we are motivated to act; we must be able to identify
ourselves with some of our desires and not others. We must know what we really want. Third,
we must be able to become motivated by those desires we identify with; this will involve a form
of control over our desires that often requires knowledge of their influence on us. These three
abilities, which I will call introspection, identification, and influence, are examined in Chapter 2,
where I expand upon Harry Frankfurt’s theory of free will as I develop my own.
These conditions will seem overly stringent to some philosophers. Surely we often act
freely without introspecting or deliberating about what we want or how to carry out our action. I
agree. I certainly do not want to suggest that we must be hyper-reflective about all of our
decisions and actions in order to be free and responsible. As I will discuss in Chapter 2, agents
may act freely without engaging the abilities discussed above, and they may act of their own free
Consciousness is also an important, and surprisingly overlooked, component of free will. Timothy O’Conner
writes, “It is a remarkable feature of most accounts of free will that they give no essential role to conscious
awareness” (2000: 122). I hope to remedy this problem, though the knowledge conditions I outline do not require
that agents are concurrently conscious of what they are doing (see Chapter 2, section 3).
Nicomachean Ethics 1111a23 (my italics).
Model Penal Code, Article 4, “Responsibility,” in Morris (1961: 423).
Conversely, the first thing many repressive governments do is limit their subjects’ access to information, as when
the Nazis confiscated Jews’ radios in 1935.
will—which I see as requiring more than free action—without engaging those abilities at the
time of the action. The Knowledge Condition for free will does not apply directly to actions but
to agents; it involves cognitive abilities which an agent must possess in order to have free will,
and these abilities may be possessed to different degrees. Furthermore, agents may then exercise
these abilities in different ways, which will result in actions that reflect different degrees of
freedom and responsibility.5 Hence, satisfying the Knowledge Condition is necessary and
sufficient for an agent to possess free will, but (1) since the Knowledge Condition comes in
varying degrees, so too may different agents possess varying degrees of free will6 and (2) since
free will applies to agents, not actions, an agent with free will may exercise it to varying degrees
in different actions. (These relationships will be more fully discussed in Chapter 2, section 7).
Some philosophers, notably libertarians, will see the abilities I associate with the
Knowledge Condition as obviously insufficient for free will. They believe that free will requires
the ability to do otherwise in a strong sense that is incompatible with deterministic causation
(hence, they believe indeterminism of some sort is necessary for free will). Sometimes the term
“free will” as used in philosophical debates is relegated to refer only to this ability to do
otherwise, and all the attention is focused on whether this ability is compatible with
determinism.7 I think this is a mistake. As I will argue in this chapter, the relation of
determinism (and indeterminism) to free will is often presented in a misleading way. I suggest
that, without giving up the language of free will, we should remain neutral on the question of
determinism and focus our attention on more tenable and interesting questions. These questions
will move the free will debate closer to problems in philosophy of mind about the nature of self-
knowledge, consciousness, and reductionism. As I will discuss in Chapters 3 and the Epilogue,
this view of free will also opens it up to empirical and conceptual challenges distinct from—and
more significant than—the threat of deterministic causation.
2. Compatibilism, Libertarianism, and Compatriotism
Historically the philosophical debate about free will has been about the “compatibility
question” and fought between compatibilists and incompatibilists. Incompatibilists believe that
free will is impossible in a deterministic universe—one in which the present state of the universe
is necessary given its past states and the laws of nature (i.e. a description of the present state of
the universe is entailed by a description of the state of the universe at any other time and the laws
of nature). Compatibilists believe free will is possible even if determinism is true. Some
compatibilists (“soft determinists”) believe both that determinism is true and that we have free
will. Some incompatibilists (“hard determinists”) agree that determinism is true but then
As an analogy, we may apply the term “strong” to persons. Possessing strength to some degree will involve being
able to perform certain actions (e.g. lift various amounts of weight). A strong person has these abilities even when
not exercising them, and then she may or may not exercise them to various degrees in performing particular actions,
such as lifting loads.
These degrees of freedom will apply between members within a type of creature, such as human beings (e.g.
children have less free will than normal adults), and also between types of creatures (e.g. humans have more free
will than apes).
Some philosophers have simply given up on the term “free will,” and turned to other concepts such as free action,
choice, agency, autonomy, self-control, self-determination, and moral responsibility (see section 3 below). See
Double (1991) for an argument that the term “free will” does not pick out a natural kind.
conclude that we do not have free will. More commonly, the incompatibilist will argue that we
do have free will, and therefore, determinism must be false. This is the libertarian position.
The compatibility question usually travels via the “could-have-done-otherwise” condition
(CDO); most philosophers agree that, in some sense, to have free will we must be able to do
other than we in fact do.8 Most compatibilists interpret the CDO condition in such a way that it
can be met even if determinism is true. Incompatibilists, however, argue that if determinism is
true, we can never act other than we in fact act. This debate has been (in every sense of the
word) interminable and often seems to reduce to a clash of intuitions or to begging the question
(or both): I will try to avoid this debate wherever possible.9 Instead, I will argue that the
compatibility question, at least as it relates to the thesis of determinism, should not be the central
question in debates about free will.
A well-known incompatibilist, Peter van Inwagen, writes:
There are four possible positions one might take about the logical relations that obtain
among free will, determinism, and indeterminism: (1) Free will is compatible with
determinism and incompatible with indeterminism (sc. of human actions); (2) Free will is
incompatible with determinism and compatible with indeterminism; (3) Free will is
incompatible with determinism and incompatible with indeterminism; (4) Free will is
compatible with determinism and indeterminism. Positions (1) and (2) are the
historically important ones. Position (3) has, to my knowledge, been taken only by C.D.
Broad. Position (4) has, to my knowledge, been taken by no one. (1985: 349)
I think van Inwagen’s categories do not accurately reflect the free will debate. He is right that
position (2) is historically important; it represents all libertarians. But he suggests that position
(1) represents all compatibilists when, in fact, it represents only soft determinists, and only some
of them have explicitly argued that free will is incompatible with indeterminism.10 (Some hard
determinists represented by position (3) also argue for the incompatibility of free will and
indeterminism, claiming that indeterminism entails randomness. I will argue below that this
claim is mistaken.)
More importantly, some philosophers do hold position (4), that free will is compatible
with determinism and with indeterminism—more precisely, that it is possible free will and
determinism are jointly true and it is possible that free will and indeterminism are jointly true.11
Examples include, I believe, Peter Strawson, Harry Frankfurt, and Daniel Dennett, whose
Dennett (1984b) coined the acronym CDO and argues that it is not required for free will in many cases.
Fischer aptly calls the debate a “dialectical stalemate” (1994: 83-85). And Kant, of course, included it as one of
Notably, Hobart (1934), Schlick (1939), Nowell Smith (1948), Ayer (1954), Smart (1961), and Bergmann (1977,
Appendix). Historically, the position is associated with Hobbes, Hume, Mill, Shopenhauer and perhaps the Stoics.
See section 4 below.
Which is not to say it is possible for free will, determinism, and indeterminism to be true together. The way van
Inwagen describes position (4) makes it seem unlikely any philosopher would hold it, since it seems the philosopher
would have to offer positive arguments for each conjunct—that free will is compatible with determinism and that it
is compatible with indeterminism. But below I explain that the position requires the least defense of the four since it
only requires arguing against the necessity claims of the other positions.
analyses of free will do not require the truth of determinism or indeterminism.12 Frankfurt, for
instance, writes, “My conception of freedom of the will appears to be neutral with regard to the
problem of determinism” (1971: 25). Since van Inwagen made his claim, several other
philosophers have followed this trend of remaining agnostic about the relations between free will
and determinism. Fischer and Ravizza (1998), though they use the language of responsibility
and control rather than free will, state: “Indeed, an implication of our approach to moral
responsibility is that our personhood need not be threatened by either the truth of causal
determinism or its falsity” (16). Alfred Mele (1995) using the language of autonomy, takes the
position of “agnostic autonomism”: “agnostics do not insist that autonomy is compatible with
determinism; nor need they insist that we are internally indeterministic . . . in a way of use to
Such philosophers are usually identified as compatibilists since they often offer some
account of how free will is possible even if determinism is true, though they also allow that free
will may be compatible with at least some types of indeterminism. But they usually argue more
emphatically that the question of determinism, at least as we presently understand it, is irrelevant
to free will. Rather, their conceptions of free will are concerned with the specific causal
processes involved in deliberation and action, and they remain neutral about whether such
relations need to be deterministic or need to involve some sort of indeterminism. Fischer and
Ravizza’s account of control, for instance, requires that we “look more carefully at the
characteristics of the actual sequence that leads to the action . . . [which] holds that ascriptions
of responsibility do not depend on whether agents are free to pursue alternative courses of
action” (1998: 37).
The account of free will I develop here also represents position (4). Perhaps van Inwagen
(falsely) thinks this logical space is unoccupied because it, unlike the other three positions, has
not been given a philosophical name. Let me correct this problem. I will name the position
compatriotism. This name is meant to suggest that compatibilists and libertarians should put
aside their debates about determinism and become “compatriots” in analyzing the positive
conditions of a concept of free will worth defending and in battling more pressing and serious
threats to such a view of free will.13 Compatriotism is phonetically similar to compatibilism,
which is appropriate to the extent that it is anti-incompatibilist: it disagrees with (the first
conjunct of) position (2) to suggest that it is possible that free will and determinism are jointly
true [◊(F & D)]. But compatriotism does not hold that free will is incompatible with
indeterminism—so it also disagrees with (the second conjunct of) position (1) to suggest that it is
possible that free will and indeterminism are jointly true
[◊(F & -D)].14
Strawson (1962), Dennett (1984a) and Frankfurt (1969 and 1971). Other philosophers who explicitly or implicitly
accept this position include Watson (1983), Mele (1995), Wolf (1990), Fischer and Ravizza (1998), and Bok (2000).
See also Fischer (1994), who calls his view that causal determinism is compatible with moral responsibility
Strawson (1962) suggests that a “reconciliation” between incompatibilists (whom he calls “pessimists”) and
compatibilists (“optimists”) is possible if we properly amend the optimists’ meager requirements for free will (see
Chapter 2 below).
Notice that if you believe that free will is incompatible with determinism [□ -(F & D)] and you also believe that
free will exists in at least some possible worlds [◊( F)]—notably, ours—then you are committed to the claim that
free will must be compatible with indeterminism [□ ◊ (F & -D)]. That is, the libertarian says, “There is at least one
Instead, compatriotists remain neutral with regard to the compatibility question for four
reasons. First, it is the easiest position to hold—the default position, as it were. Each of the
other three positions involves claims of incompatibility, which involve the modality of necessity.
For instance, position (3) entails, “It is necessarily false that free will and determinism are both
true” [□ -(F & D)] and also, “It is necessarily false that free will and indeterminism are both
true” [□ -(F & -D)]. Hence, assuming the philosopher holding this position accepts that either
determinism or indeterminism must be true, he is committed to the claim that necessarily free
will does not exist (see footnote 14 for the other positions). Such necessity claims are
metaphysically strong and require powerful arguments to back them up. The burden of proof is
on those philosophers who argue for necessary truths and entailments.15 Possibility claims are
weaker and require less support. To hold position (4) is to say it is possible that free will and
determinism are both true and it is possible that free will and indeterminism are both true (which
is not, of course, to say that it is possible that free will, determinism, and indeterminism are all
true together). Some may say such a position is empty—that you have to take a firm stand on the
logical relations between determinism, indeterminism, and free will.16 But compatriotists do
offer arguments against the necessity claims of the other three positions (I will present some of
these arguments below), and they have a great deal to say about the positive conditions required
for free will and the causal relations such conditions involve—they just don’t see such causal
relations as necessarily requiring determinism or indeterminism.
Indeed, a second reason compatriotists remain neutral about the compatibility question is
their belief that we (e.g. philosophers and scientists) do not have clear conceptions of
determinism, indeterminism, or causation—much less, the logical relations between them—and
until we do, we will not understand whether determinism or indeterminism or neither conflicts
with free will (depending on how free will is conceived). Since we are unsure about how to
analyze determinism and indeterminism, we should be unsure about the implications of each
view, and should avoid claims about necessary implications for free will.17 Many philosophers,
including some incompatibilists, say that science might discover whether deterministic or
indeterministic causation in fact occurs in the relevant processes of humans’ coming to act. In
fact, Peter van Inwagen, who strongly believes we have free will, says that if science did
discover that these processes were deterministic, he would have to renounce his argument for
incompatibilism.18 Martin Fischer, who calls such a possibility “metaphysical flip-flopping,”
possible world—namely, our own—in which free will and indeterminism co-exist.” Similarly if you believe, like
the soft determinist: ◊( F) and free will is incompatible with indeterminism [□ -(F & -D)], then you are committed
to the claim that free will must be compatible with determinism [□ ◊ (F & D)]. (These modal entailments use the S5
system.) Compatriotists, who hold position (4) are committed to much weaker modal claims.
See, for instance, Lycan (1987: 10).
This claim may be legitimate if free will if defined in a way that conceptually ties it to determinism or
indeterminism. But such a definitional move, I believe, is ill-founded (see section 4 below).
I believe some incompatibilist intuitions are fueled by a conception of causation left over from classical physics
(the Newtonian picture of mechanistic interaction). The causation suggested by modern physics is very different,
and not just because it suggests irreducibly indeterministic causal relations. It certainly is not best captured by the
image of billiard balls colliding!
Specifically, he would give up the Beta principle. See van Inwagen (1984: 219). This seems to suggest that he
does not accept that necessarily determinism precludes free will [□ -(F & D)].
says, “I do not think that this very important and basic belief [in moral responsibility] should be
‘held hostage’ to esoteric scientific doctrines.”19
Nor should our basic belief in free will be ‘held hostage’ to esoteric philosophical
arguments for incompatibilism. Indeed, the third reason compatriotists remain neutral about the
relation between free will and determinism is that they believe the debates about this relation,
and the related debates about interpretations of the word “can,” have not advanced, in part
because of the ambiguity and disagreements about the terms involved.20 The debates suggest an
antinomy which can only be escaped by, as Quine suggests, “a repudiation of our conceptual
heritage.” Compatriotism represents one such repudiation by arguing the question of
determinism, at least as presented in these debates, is simply irrelevant to free will.
Fourth, and most importantly, compatriotists believe that the central questions about the
nature of free will involve the cognitive abilities it involves—for instance, an agent’s ability to
know what he really wants and to know how to act accordingly (the Knowledge Condition).
Indeed, those philosophers, whether compatibilist or libertarian, who believe—contra position
(3)—that humans do possess free will, share many intuitions about the positive conditions
required for free will. I will lump together all such positive conditions under the name AS
conditions (for “Agent-as-Source”), because they require that agents—their deliberations,
desires, decisions, intentions—are the source of their actions.21 Despite any disagreements about
the threat determinism may pose to free will, most participants in the free will debates put a lot
of effort into analyzing AS conditions and agree on many of them. (No libertarian, after all,
believes indeterminism is a sufficient condition for free will). Furthermore, these philosophers
share views about various other threats to free will (some of which, we will see, are mistakenly
associated with determinism).
Compatriotism locates free will not in some metaphysical ability to do otherwise in the
exact same situation but (where it seems to be located) in our cognitive abilities to know
ourselves and know how to achieve what we want. That is, it situates questions about free will
amidst questions about the nature of the mind and the mind-body and mind-world relations. In
turn, compatriotism sees the most significant threats to free will not in determinism but in certain
conceptions of the mind and of the types of causal processes involved in an agent’s coming to
act, threats that derive from certain philosophical and scientific theories of human nature.
Compatriotists view these threats as more serious and relevant than those thought to be posed by
the metaphysical doctrine of determinism.22 They can then wage a unified battle against these
Fischer (1999: 129). See also Fischer and Ravizza (1998: 253-254). As is clear from my Chapter 3, I certainly do
not think free will must be immune to all scientific threats (but see note 23 below).
This is not to say the arguments have not been original, interesting, and impressive. Compatibilists’ conditional
analyses of “can,” incompatibilist responses, new libertarian theories, and arguments for and against PAP (the
principle of alternate possibilities), including Frankfurt-style examples, have made the late 20th century an exciting
time for free will enthusiasts, but the debate about the compatibility question still seems deadlocked (see Fischer
1999). The more interesting moves, I believe, have been in the compatriotist direction—i.e. theories of free will that
suggest determinism is irrelevant to free will (such as Frankfurt’s).
These conditions are often discussed in terms of “self-control,” “autonomy,” and “self-determination.”
Indeed, determinism is a theory that might be proven true or false by physicists doing experiments that do not
involve humans (so the fate of our free will would be in the hands of a science that need not even consider us). On
Thus, compatriotism fills a logical space in the free will debates which has remained
(1) The free will problem is not essentially about the question of determinism. Indeed, one can
and should remain neutral about whether determinism or indeterminism is necessary for free
In addition to this claim, I believe most of the philosophers I identify with the compatriotist
position share two other views:
(2) Free will is constituted by a set of cognitive abilities.
(3) Most humans possess free will (i.e. possess the cognitive abilities that constitute free will).
The rest of this chapter will continue to argue for the first thesis. Chapter 2 will discuss
some of the cognitive abilities required for free will that have remained largely unexplored
(thesis 2). I will argue for the third thesis to the extent that, in Chapters 3 and 4, I defend my
theory of free will against empirical and conceptual threats. Beyond that, I will defer to a
simplistic argument offered by several philosophers: free will is required for moral
responsibility; we cannot seriously doubt that we (normal humans) have moral responsibility;
therefore, we have free will.23 An even simpler argument might be, “We all know we have free
will, it’s just a matter of defining what it is we all know we have.”24 I believe this is right, and
such a definition will not involve determinism or indeterminism.
3. Is that the free will debate?
“What men have esteemed and fought for in the name of liberty is varied and complex—
but certainly it has never been metaphysical freedom of the will.”
Before I proceed, I should answer a possible objection to the compatriotist position.
Some philosophers (especially incompatibilists) might respond that the position is incoherent
because the free will debate just is (by definition) the debate about the compatibility question. If
you want to argue about other aspects of human freedom, they would claim, then use different
language—talk about autonomy, agency, choice, deliberation, control, or moral responsibility (or
at most free action). And perhaps when you are finished you can link up your discussion to the
question of determinism and free will. In fact, many philosophers have taken this route, and
especially in recent decades, have focused on the concepts of autonomy and moral responsibility,
the other hand, scientific theories about human nature and the mind must take into account the evidence that we at
least seem to be free and responsible (psychology, for instance, must deal with our experience of deliberation and
See, for instance, van Inwagen (1983: 206-7). This argument underlies his willingness to “flip-flop” if we
discover that our universe is, in fact, deterministic. Though I agree that free will is necessary for moral
responsibility, I do not believe it is sufficient (see Chapter 2, section 8).
Which is almost as simple as Samuel Johnson’s famous quip, “Sir, we know our will is free, and there’s an end
Dewey (1957: 303).
hence conceding that, after all, they are not trying to analyze free will.26 I have sympathy with
this move, since changing the name of the debate is one way to change the terms of the debate.
But I think it concedes too much.
First of all, the term “free will” has a history in philosophy that goes beyond the debate
about determinism, even if we take into account theological determinism (God’s foreknowledge)
and logical determinism (fatalism). Most philosophers, including libertarians, spend many more
pages discussing the positive conditions of agency required for free will (AS conditions) than the
negative condition of freedom from deterministic causation. More importantly, the term “free
will” has too much currency in ordinary language to concede that free will is only about the
question of determinism and the ability to do otherwise. Peter van Inwagen disagrees. He
claims that “the term ‘free will’ is a philosophical term of art. . . . If someone uses the words
‘free will’ and does not use them within [the phrase ‘of his own free will’], he is almost certainly
a participant in a philosophical discussion,” and that discussion, he says, is about whether an
agent can act in one way or another—can do otherwise in the exact same situation—an ability
van Inwagen believes requires indeterminism (1989: 220).
I offer several reasons to suggest van Inwagen is mistaken. First, when I ask my students
to talk about what they mean by free will, most talk about our abilities to choose and deliberate,
to be free of constraints and control by others, to be conscious and in control of what we are
doing. Few bring up determinism (though some mention particular causal histories, such as
brainwashing or genetic influences, which they see as threatening free will). They must be
trained to think that universal deterministic causation may threaten free will (though, I admit, it
is not hard to get them to feel the bite of that intuition if the incompatibilist argument is
presented in certain ways).27
Second, the term “free will” is used often by people in the real world and usually without
reference to the question of determinism, as in two recent letters to the editor of The New York
Times: “The Chimps Don’t Have Free Will” (7/7/99) and “Kennedy ‘Curse’ or Needless Risk-
Taking; Limits of Free Will” (7/21/99). When Robert Sapolsky writes in Newsweek, “Who are
we then, and what will happen to our cherished senses of individuality and free will?”, he is
concerned that “our behaviors, thoughts, and emotions are merely the sum of our genes.”28 As I
will discuss below, this concern is distinct from the question of metaphysical determinism, which
he does not discuss. Since “free will” appears in the popular media, especially in the voluminous
discussions of recent scientific findings about human nature from psychology, genetics, and
neuroscience, philosophers should not limit their use of the term to a distinct (and insular)
debate. Indeed, scientists themselves use “free will,” usually to explain how their theories limit
it or show it to be illusory. For instance, Matt Ridley (1999) titles a chapter of his book about the
genome, “Free Will,” and argues that social causes threaten free will as much as genetic causes.
Autonomy is the focus, for instance, of Dworkin (1988), Haworth (1986), and Mele (1995). Fischer and Ravizza
(1998) do discuss the compatibility question a great deal but still call their book Responsibility and Control and
maintain a view they call semi-compatibilism—that even if free will requires the ability to do otherwise, moral
responsibility does not.
In my experience, students feel threatened by determinism mainly because some incompatibilist arguments
present determinism as entailing that the past controls us, that we are predictable, or that our conscious deliberations
play no role in what we do. I discuss these misleading conflations in the next two sections.
Sapolsky, Robert, “It’s Not ‘All in the Genes,’” Newsweek, April 10, 2000, p. 68.
And a series of articles in American Psychologist (July 1999) discusses “recent fundamental
breakthroughs in the understanding of motivations, free will, and behavioral control” (461).29
Furthermore, our legal system, one might say, is predicated on the concept of free will
and frequently makes reference to the term. Rarely, however, does the subject of determinism
come up in court. Instead, we want to know whether the accused knew what he or she was doing
and had the mental capacities to act on such knowledge.30 When defense lawyers use the
insanity defense (or the Twinkie defense or the Steroid Rage defense), they are not claiming their
clients’ crimes were, like all behavior, caused by the past and the laws of nature, but that, unlike
our (normal) behavior, they were caused by a particular, mitigating causal history. Perhaps the
only instance of the question of determinism leaving the sphere of philosophical debates was
Clarence Darrow’s attempt to use it as a legal defense, but even he referred to particular causal
histories, not the general thesis that all behavior is the inevitable result of the past and laws of
nature: “every case of crime could be accounted for on purely scientific grounds if all the facts
bearing on the case were known: defective nervous systems, lack of education or technical
training, poor heredity, poor early environment, emotional imbalance.”31 These are all examples
of unfortunate histories, not necessarily deterministic histories.
Finally, the concept of free will also makes frequent appearances in literature, from the
poetry of Donne and Milton to the works of D.H. Lawrence and Dostoyevsky. These writers
may mention the concept of fate, or refer to our ability to act as we wish, but they rarely mention
causal determinism. Dostoyevsky’s Underground Man defines free will as the ability “to choose
what is contrary to one’s own interests,” to defy Reason. And Milton’s Adam and Eve gain free
will not by eating from the Tree of Indeterminism but by eating the Fruit of Knowledge. When
they become aware of their motives as being good or evil, this knowledge endows them (and
their descendents) with the burdens and blessings of free will and moral responsibility.32
Hence, the question of free will, in philosophy and especially in the real world, is not
just—indeed, not mainly—about the threat of determinism, and it should not now be relegated
within philosophy to that one question. Compatriotists believes that, in understanding the nature
of human freedom, self-determination, and responsibility, the question of determinism is of
The “breakthroughs” discussed are just the sort I will discuss in Chapter 3 as threatening to free will, but not
because they describe deterministic processes, rather because they suggest that “the source of behavioral control
comes not from active awareness but from subtle cues in the environment and from thought processes and
information not readily accessible to consciousness” (461). That is, they suggest we usually act without knowing
why we are acting—a question that is unrelated to whether determinism is true.
And a defendant’s responsibility is usually mitigated if he did not intend and could not reasonably foresee the
deleterious outcomes of his actions—though there are some interesting discrepancies in some cases, where we
punish people for consequences they could not foresee, which point to the philosophical question of “moral luck”
(e.g. Nagel  and Williams ).
Quoted in Ekstrom (2000: 79). She takes this quotation to suggest that “the judgments of ordinary persons are
responsive to the consideration of determinism.” I take it to suggest that such judgments are responsive to
particular causal histories rather than to the problem of universal determinism.
One more (personal) example: when I tell people outside philosophy that I am writing on free will, their ears perk
up, but not because they want to hear about the relationship between free will and some metaphysical thesis about
the nature of causation. They want to know about free will—a term they use and hear about a lot—what it is (the
positive conditions) and what allows us (as opposed to animals, for instance) to have it. If I say I am writing about
agency or autonomy ears perk up much less.
secondary importance to questions about the abilities of agents to deliberate, choose, and act
autonomously. Hence, they certainly accept that discussions about each of the italicized
concepts in the previous sentence are crucial to the free will debate, but they are not willing to
give up their right to the term “free will” in favor of those concepts alone. To do so is to give up
the game to the incompatibilist.33 Indeed, to do so would be like the naturalist giving up talk
about “minds” in response to a dualist who suggests that the essential question in debates about
the mind is whether or not it must be an immaterial substance. Or like the moral subjectivist
giving up talk about “good”’ and “ought” in the face of a moral objectivist’s claim that the
debate about such terms must be about what objective property they refer to.
Hence, I will not give up the rhetorical battle, and I will continue to talk about free will,
with the caveats that the “free” need not be contrasted with deterministic and the “will” need not
suggest some dualistic faculty of mind (or homunculus within the brain). Rather, the will refers
to our motivational structure, and for the will to be free requires that we have certain abilities in
regards to our motivations, notably some knowledge of them and power to influence them.
Indeed, the most important reason to hold tight to the concept of “free will” is that the abilities I
will discuss have to do with our control over our internal motivational structure, our freedom to
influence our will, and not simply our freedom to control our actions (e.g. to be free from
4. Misconceptions about Indeterminism
In this section and the next I will dispel some misconceptions about indeterminism and
determinism that have led philosophers to accept positions (1) and (2) described by van Inwagen.
Specifically, I will counter some of the arguments that indeterminism is incompatible with free
will, then counter some of the arguments that determinism is incompatible with free will. These
tactics do not, of course, entail that free will is in fact compatible with either determinism or
indeterminism. But they provide reasons to reject the necessity claims of incompatibilists and of
those compatibilists who claim that free will requires determinism. Along the way, we will also
see that other issues are more central to the question of free will than determinism and
Some philosophers have argued that free will is incompatible with indeterminism and
hence believe that free will requires determinism.34 If they then want to defend the existence of
free will, as most do, they naturally see this conclusion as a reason to believe there must be
something wrong with incompatibilist arguments and also to develop a conception of free will
that is clearly compatible with determinism.35 Nowell-Smith states, “I could not be free to
choose what I do unless determinism is correct. . . . Freedom, so far from being incompatible
This is one more reason to differentiate compatriotism from compatibilism, since compatibilism is defined in
contrast to incompatibilism, and historically it has focused on defending the possibility of free will in the face of
determinism, hence offering inadequate positive conditions for free will. Peter Strawson (1962) recognized this
deficiency in compatibilist (“optimist”) accounts of free will and suggested refinements that have been further
developed by more recent philosophers (see Chapter 2).
See footnote 10. C.D. Broad, the hard determinist who holds position (3), argues that free will is incompatible
with indeterminism for similar reasons (but concludes that we do not have free will).
This has led to the development of some overly economical accounts of free will—e.g. free will as the ability to
do what you want or as freedom from external constraints. See footnote 33 above.
with causality implies it.”36 Notice the identification of “determinism” with “causality.” These
compatibilists assume that causation requires deterministic laws, such that given the cause, the
effect necessarily follows (i.e. the cause is sufficient for the effect). R.E. Hobart writes, “Thus
power [to choose] depends upon, or rather consists in, a law. The law in question takes the
familiar form that if something happens a certain something else will ensue. . . . It is just because
determinism is true, because a law obtains, that one ‘could have done otherwise’” (1934: 72-74).
That is, for the agent to act freely and responsibly, his action must be sufficiently caused by his
will (e.g. his desires and volitions), and so his act must be determined by his will. In fact, Hobart
suggests that an undetermined action is not an action at all—it is a behavior that happened to the
agent, not something that he did.
This position is predicated on the belief that any lack of determinism involves an absence
of causation such that any undetermined event must be random, or an accident. Hence, Ayer
argues, “Either it is an accident that I choose to act as I do or it is not. . . . if it is not an accident
that I chose to do one thing rather than another, then presumably there is some causal
explanation of my choice: and in that case we are led back to determinism” (1954: 275).37 The
incompatibilist would argue that for a choice (say, my choice to help a friend move or to go to
the lake) to be free, it must be undetermined by prior events. But, according to Ayer, if my
choice is really undetermined, then the event of my choice is an accident and hence out of my
It is important to distinguish between two interpretations of this claim. It could mean
that, if there are not sufficient conditions (e.g. my deliberations and desires) to determine my
choice, then if I choose to help my friend, that action is out of my control and if I choose to go to
the lake, that action is out of my control. That is, if the choice is undetermined, then whichever
action occurs, it is uncaused and hence out of my control (or, as Hobart suggests, it is not even an
action). A second interpretation is that, if my choice is undetermined, then I do not control
which of the two alternatives I choose, but (so long as I have reasons and desires for either
choice) I control and am the cause of my action of helping my friend or of working, whichever
occurs. This second interpretation will be discussed below, but it need not present a threat to the
requirement of a theory of free will that the agent causes his actions—that is, the agent can still
be the source of his actions. The first interpretation does, however, conflict with an agent’s
ability to cause or to control his actions, and this is the view suggested by compatibilists who
claim indeterminism is incompatible with free will. However, it is a misleading interpretation of
Philippa Foot (1957) offers an ordinary language rebuttal of the claim that if an event is
undetermined, this suggests it is accidental or random: “It is not at all clear that when actions or
choices are called ‘chance’ or ‘accidental’ this has anything to do with the absence of causes”;
Quoted in Foot (1957: 96).
Compare Hume: “According to my definitions, necessity makes an essential part of causation; and consequently
liberty [in the libertarian sense], by removing necessity, removes also causes, and is the very same thing with
chance” (Treatise on Human Nature, 407).
Many incompatibilists agree with this claim, such as Richard Taylor (1963: 47). Thus they argue that agent
causation is the only theory that can explain how agents can be undetermined and control their actions (see section 6
rather, she says, accidental actions are unintentional (107). She also argues that an agent’s
motives need not determine his actions.39 If accidental (or random) events are contrasted with
determined events, then if determinism is true, there could be no accidents at all, and if
determinism is not true, then any event with a probability less than 1 would be an accident
(regardless of whether it had a .01 or a .99 chance of occurring).40
Elizabeth Anscombe (1971) also offers a rebuttal of the claim that causation requires
determinism. Causation need not be seen as a logical connection (as Hume “discovered”), and it
need not be seen as a Humean relation of constant conjunction either; some causes may be non-
necessitating—that is, irreducibly statistical or probabilistic: “A non-necessitating cause is then
one that can fail of its effect without the intervention of anything to frustrate it” (101). Such an
idea is anathema to many philosophers, steeped in the principle of sufficient reason, who want
something (some cause) to make the difference whenever there is a difference (an effect). But
Anscombe argues that the burden of proof is on the philosopher who claims it is a necessary
truth that all causes are deterministic.41
Anscombe’s view has been supplemented by positive accounts of probabilistic causation.
Such accounts suggest that some laws of nature are irreducibly statistical; given a particular
situation (set of causes), there are certain probabilities that one of several effects will follow.
One of the proponents of such accounts, Wesley Salmon, explains that the common response that
such probabilities are only epistemological not ontological—that “hidden variables” will explain
away the probabilities—is “a declaration of faith.”42 Probabilistic accounts generally define a
cause as some factor that increases the probability that an effect will occur, in contrast, for
instance, to those accounts that define a cause in terms of sufficient conditions (i.e. given the
cause, cetibus paribus, the effect necessarily follows). So, on a probabilistic account, a virus is a
cause of a disease if, when a person is infected with the virus, she is more likely to contract the
disease than a non-infected person (even if, on some relevantly similar occasion, a person does
not contract the disease).
These causal accounts were developed in part to deal with the revolution in 20th-century
physics launched by quantum mechanics. The classical physics of Newton and Galileo inspired
Laplace and later philosophers to embrace determinism and mechanism in conceiving of
causation.43 The predictions and explanations in quantum mechanics, however, are irreducibly
She argues that motives are not causes at all—they are descriptions of actions and hence analytically, not
empirically (i.e. causally), related to the actions. In general, I will accept the Davidsonian view (1963) that reasons
are causes (see Epilogue).
See Loewer (1996: 100).
Hence, the burden of proof is on the philosopher who supports the conclusion that free will necessarily requires
determinism with the premises that (1) free will requires causation (sc. of human actions) and (2) causation
necessarily requires determinism.
Salmon (1980). He discusses several probabilistic accounts of causation, including those of Hans Reichenbach,
I.J. Good, and Patrick Suppes.
I believe this historical connection between deterministic causation and mechanistic causation has fueled
incompatibilism. It provided (e.g. for philosophers like Descartes) a reason to think that if the mind is part of the
physical universe, it will have to be explained not only in mechanistic terms but also in deterministic terms. So, to
avoid the implications of either mechanism or determinism, the other had to be rejected (e.g. with a theory like
substance dualism). See section 5D below.
indeterministic, at least according to the current “orthodox” interpretation.44 Hence, given a
certain type of event, such as an electron’s being fired at a barrier, there are certain probabilities
that the electron will penetrate or be deflected by the barrier. Each outcome is possible given the
same initial conditions, and each outcome is caused by those conditions. So, quantum theory
offers a model for probabilistic causes, undermining the idea that science requires the concept of
determinism. As Anscombe rightly points out, “It has taken the inventions of indeterministic
physics to shake the rather common dogmatic conviction that determinism is a presupposition, or
perhaps a conclusion, of scientific knowledge. Not that the conviction has been very much
shaken even so” (103).
Some philosophers suggest that, even if quantum events are irreducibly indeterministic,
they cancel out at the macro-level, including the level of causation involved in mental events and
actions. But such a claim also seems to be a “declaration of faith.” Quantum events may—
perhaps through the sensitive dependence on initial conditions suggested by nonlinear
dynamics—“percolate up” to the macro level (for instance, to neuronal activity). And many
philosophers have offered thought experiments in which an indeterministic event, for instance, a
Geiger counter measurement, has large-scale effects, such as the dropping of a nuclear weapon.45
Some philosophers and scientists also believe that indeterministic causation may occur at the
macro level autonomously (i.e. without being based on quantum effects), for instance, in the
processes of natural selection.46 Again, it is not known whether probabilistic causation is
required to explain and predict events only at the quantum level or whether such causation may
be required to deal with “bigger” events as well, including perhaps events in the human brain,
but the possibility should not be dismissed a priori. Probabilistic accounts of causation are
controversial, but so is every other account of causation. The point is that they provide
arguments against the claim that it is necessary that causation requires determinism and that
indeterminism entails randomness. Hence, they provide reason to dismiss the claims of those
compatibilists who argue that free will requires determinism, and hence, there must be something
wrong with incompatibilism.
In fact, some libertarians have made use of probabilistic accounts of causation in order to
develop a theory of human choice and action that is causal but not deterministic.47 For example,
Robert Kane’s (1996) impressive libertarian theory of free will suggests that quantum
indeterminacy is magnified by nonlinear dynamics (chaos theory) to affect neuronal activity in
In effect, conflicts of will . . . stir up chaos in the brain and make the agents’ thought
processes more sensitive to undetermined influences. . . . The result is that [some choices]
are influenced by, but not determined by, past motives and character. The uncertainty
The Copenhagen Interpretation. Other interpretations (Bohm’s) are deterministic, though the predictions remain
fundamentally statistical. See Loewer (1996).
Schroedinger’s cat was perhaps the first such thought experiment.
See Brandon and Carson (1996). Most laws in psychology are presented in statistical terms, but no argument is
provided that these laws are actually deterministic and only epistemically probabilistic.
For instance, Kane (1996), Ekstrom (2000), Ginet (1990). See section 6 below.
and inner tension that agents feel at such moments are reflected in the indeterminacy of
their neural processes” (130).48
So, for instance, when I feel conflicted in my deliberations about whether to help my friend or go
to the lake, it is because I have good reasons for either choice and these reasons are instantiated
in competing neural networks, sensitive to micro indeterminacies. If the reasons are near
equilibrium, the indeterminacy may affect my choice, making it undetermined. But since I have
a reason to help my friend move, then these reasons can be a cause of my action even if my
choice to help him is undetermined (I might choose to go to the lake). So, if I in fact help my
friend, the action is still caused in the appropriate way (e.g. by my desires) to call it an action.
Since I also have a desire to go to the lake, then if I choose to do that, the action is also caused by
my desire, even though the choice is undetermined. My reasons may, in Leibniz’s phrase,
“incline without necessitating” my choice.49 Hence, Hobart’s argument that undetermined
behaviors cannot be actions is mistaken. Whether to call such actions free actions will depend
on conditions other than whether my choice is determined or undetermined.
I will discuss below whether or not such indeterminism helps secure free will, but it
certainly offers a model for undetermined choices that are nevertheless caused by an agent’s
mental states (and can thereby satisfy AS conditions). It undermines the arguments of those
compatibilists who claim that indeterminism entails randomness and thus lack of control by the
agent. Given the arguments of Foot and Anscombe, the positive accounts of probabilistic
causation, current theories in physics and other sciences, and contemporary libertarian accounts
like Kane’s, we may conclude that:
(1) The causation required for accounts of action and free will—for instance, that one’s reasons
cause one’s actions—may be probabilistic (i.e. irreducibly indeterministic).
(2) Since it is not necessary that such causation is deterministic, it is not necessary that
undetermined choices and actions are uncaused and therefore random (as asserted by some
(3) So, it is not necessary that free will is incompatible with indeterminism (at least for the
reasons given by such compatibilists).
(4) Hence, it is possible that free will is compatible with indeterminism.
For all we know, the universe may have a degree of metaphysical “looseness” or
openness without having random “causal gaps.” Events may be caused without being
deterministically caused (i.e. without there being sufficient prior causes). That this middle
ground between deterministic causation and randomness has opened up suggests to the
compatriotist that the concepts of causation, determinism, and indeterminism (and the relations
between them) may be elucidated in a way that will make the age-old debate between
compatibilists and libertarians outdated.
See also Eccles (1995), Thorp (1980), and Ekstrom (1999) for similar accounts.
See also Ginet (1990), chapter 6, for a discussion of anomic reasons explanations in which “the very same
antecedent state of the world could afford a reasons explanation for either of two or more different alternative
5. Misconceptions about Determinism
We have seen that the argument that free will is incompatible with indeterminism is
based on particular views of indeterminism and causation that are implausible and are certainly
not necessary truths. Now I will turn to some of the arguments for the incompatibility of
determinism and free will to show that they too are misconceived. Traditionally incompatibilists
have presented determinism as a monolithic threat to free will: if determinism is true, then
agents cannot do other than they do and hence they cannot have free will (since it requires the
ability to do otherwise). But in fact, determinism is usually presented, implicitly or explicitly, as
a many-headed monster posing a plethora of threats to free will: determinism “hardens” the
universe and prunes any possible alternatives for action; it makes our actions the inevitable result
of physical forces beyond our control; it makes us mere outcomes of our genes and upbringing; it
makes our conscious states epiphenomenal, as if we are turning steering wheels towards roads
we have to go down anyway; it makes us part of a predictable machine such that we could be
manipulated by someone who knew our “program.” Once we separate these images and
arguments, we can see which of them are threats legitimately implied by determinism and which
are not. We can also see if indeterministic theories can help overcome the threats. I will thus
take on the many heads of the “monster of determinism” one by one and, in most cases, sever
them from determinism. I argue that, in the end, the threat of determinism is much diminished
and what is left of it cannot be defeated by any tenable theory of indeterministic causation.
My divide-and-conquer strategy begins by laying out six different threats that
incompatibilists have suggested determinism poses to free will. These six threats are: (1)
external constraint, (2) scientific determinism, (3) predictability, (4) reductionism and
epiphenomenalism, (5) a closed future, and (6) lack of dual control. In this section I will argue
that it is a misconception of determinism to suggest that it implies any of the first four of these
threats to free will. In the next section I will discuss the last two (legitimate) threats of
determinism and suggest that, while the fifth (closed future) may be countered by indeterminism,
it is not what we really want out of free will nor what incompatibilist arguments demand. The
sixth threat (to dual control) can only be countered by a theory, agent causation, which is
Again, the goal of this strategy is not to prove that free will is necessarily compatible
with determinism, but rather to counter many of the arguments that claim determinism is
necessarily incompatible with a legitimate conception of free will (arguments that, I believe,
account for the strong hold—stranglehold?—that incompatibilist intuitions have on us). I will
also show that coherent libertarian theories do not answer the dilemma posed by incompatibilist
arguments. Along the way, it will also become clear that, because the most significant threats to
our ability to be responsible for our actions are not in fact tied to determinism, we should
develop a conception of free will that answers to these threats, and we should leave behind the
A. External Control and Constraint
We legitimately feel that to be free we must be able to act on our desires and choices
without some external force or agent controlling or constraining us. As discussed in section 1,
we want the power to act on what we really want. I am not free to help my friend move if that’s
what I want to do but I’m locked in my room or paralyzed. I am also not free if someone
hypnotizes me to avoid helping my friend or if an evil neurosurgeon “excises” any desire I have
to help. In describing the conflict between determinism and free will, some incompatibilists
depict the past and the laws of nature as if they were analogous to these external constraints.
They personify history and nature to blur the boundary between constraints we all recognize as
exculpating and metaphysical limitations on our ability to change the past or the laws of nature.
Richard Taylor, for instance, writes, “What am I but a helpless product of nature, destined by her
to do whatever I do and to become whatever I become,” and he claims that there is no difference
between “an ingenious physiologist [who] can induce in me any volition he pleases” and
“perfectly impersonal forces” such as deterministic laws (1963: 36). He continues, “Whether a
desire which causes my body to behave in a certain way is inflicted upon me by another person,
for instance, or derived from hereditary factors, or indeed from anything at all, matters not in the
Of course it matters! First of all, there is the ordinary language point that we do
distinguish between control by external agents and other causal histories.50 To suggest that all
causes are coercive undermines our ability to describe only certain causes as coercive. For
instance, contra Taylor’s assertion, we ascribe responsibility very differently to a person who
chooses to steal because someone threatens him (or even because he is “coerced” by his poverty)
than a person who chooses to steal because he wants more stuff. And we do not blame a person
for having a hereditary disease, but we may blame someone (somewhat) for contracting a disease
he knew he could avoid by refraining from risky behaviors.51 To suggest that determinism
entails that all our actions are like coerced actions is to beg the question against compatibilists
since they agree that coercion undermines free will but disagree that the laws of nature are
External agents can coerce us in a way that Mother Nature cannot, because agents have
desires of their own, whereas the laws of nature have no goals or purposes. Indeed, the concept
of coercion seems teleological in a way that would preclude its applicability to anything that did
not have goals—that is, goals it wanted to achieve by coercing someone to do something.52
Furthermore, if we act on other agents’ desires only because they induce or implant them in us
without our knowledge or assent, then it seems more accurate to say we are acting on their
desires, not our own. More precisely, if we can trace the development of the desires that move
us to act directly to other agents’ manipulation (based on their desires), then they should be seen
as the source of our actions. To respond that we might have developed the same desires anyway
is to suggest that, in analyzing free will, different types of causal processes make no difference—
for instance, that learning from your parents about why helping others is a good desire to
cultivate is no different than having a “friend” (who wants help moving) slip you a drug that will
induce that desire.
Ayer (1954) effectively makes this point. Ironically, he does so using ordinary language considerations similar to
those, described above, that Foot uses against Ayer (and others) to show that he mistakenly conflates “accidental”
To make this point is not to suggest that it is easy to ascribe varying degrees of responsibility in these cases.
Determining which sorts of causal histories mitigate one’s freedom and responsibility is the reason for engaging the
question of free will in the first place. I will return to this problem in Chapter 2. Fischer (1994, chap. 1) argues that
there is no in principle difference between constraint by agents and other external forces, yet his theory of
responsibility is based on distinguishing whether the actual causal sequence leading to action involves coercion.
Hence the scare quotes around “coerce” in the example above: poverty often constrains people’s choices but we
personify poverty (endowing it with intentionality) if we suggest it coerces people.
Even most libertarians allow that coercion and compulsion must be marked off from
other types of causation.53 It is a misconception of determinism to suggest that it implies that we
are always controlled by external forces, that the past and the laws of nature are just like agents
who manipulate our desires and actions. If determinism threatens our free will, it is not because
it entails that we are always pushed around or forced to do what we do. And when we are
pushed around or forced to do what we do, it is not because the universe is governed by
B. Scientific Determinism
Some incompatibilists (and compatibilists too) associate determinism with scientific
practice. They then link threats to free will implied by certain scientific theories to the thesis of
determinism. Laura Ekstrom, for instance, calls the thesis that every event is causally
necessitated by a previous event “scientific determinism” (2000: 16), suggesting that
determinism is implied by or required for scientific explanations. Other philosophers suggest
that scientific determinism is just a specific formulation of metaphysical determinism: scientific
discoveries (e.g. about human genetics and neurobiology) simply spell out the deterministic laws
that necessitate our choices and actions. The bony metaphysical thesis of determinism is thus
“fleshed out” with scientific theories about how agents’ actions are in fact causally determined—
by genes or behavioristic reinforcement or neural activity or economic conditions or Freudian
drives, etc. These threats are more alarming than the metaphysical thesis of determinism since
they are more comprehensible and specific (and they get more media coverage). But they are
also more alarming because they far outstrip the threat to our ability to do otherwise; they pose
potential problems for many aspects of our conception of agency, such as our ability to
understand why we do what we do. And often they replace the reasons we offer for our behavior
with causes we don’t (consciously) want to move us, such as Oedipal drives to “marry” our
mothers or genes to “choose” our mates based on their pheromones.
However, it is a mistake to equate determinism with science, and so, the very term
“scientific determinism” is a misnomer. First of all, scientific practice does not require
deterministic causation. After all, it was quantum theory which first developed a coherent model
of indeterministic causation, though the thesis of determinism is often maintained despite claims
of scientific theories such as quantum mechanics. Furthermore, theories in almost every branch
of science use statistical explanations, and while scientists may view such indeterminacies as
epistemological and pursue explanations to eliminate them, they need not commit themselves to
the metaphysical claim that every effect has a sufficient cause.55 Though perhaps inspired by
Newtonian science, determinism is a metaphysical thesis that arose well before the
Enlightenment with the Atomists and the Stoics. Scientific practice may rely on some
See, for instance, Kane (1996: 30) and van Inwagen (1975: 52).
Dennett (1984, esp. chapters 1 and 3) elucidates much more fully on the arguments in this section.
Some philosophers and some scientists nevertheless believe determinism is a necessary condition for scientific
work (see, for instance, James’ comment about psychology that “for her scientific purposes determinism can be
claimed” [cite]). What is necessary, however, is not determinism but some conception of causal connections and
laws. A belief in determinism might, however, serve as a useful motivational tool to keep a scientist looking for
underlying factors when faced with statistical results.
conception of causation and laws (though philosophers have not agreed on which, if any,
conception), but it is not wedded to deterministic causation and laws.
More importantly, the scientific theories that seem to threaten free will (and often
motivate incompatibilist intuitions) need not rely on deterministic causation and, even if they
turn out to be deterministic, that would not be why they were threatening. For instance, if the
truth of radical behaviorism threatens free will, it is not because the theory requires deterministic
causation—it’s laws (if it has any) could be irreducibly probabilistic—but rather because it
denies the existence (and/or causal efficacy) of mental states, such as conscious deliberations,
which are required by most theories of free will.56 If the idea of the Freudian unconscious
threatens free will, it is not because unconscious drives deterministically cause our behavior; it is
because unconscious drives cause our behavior, and because we have no knowledge or control
over our unconscious drives (at least not until, through therapy, we make them conscious).57 If
the truth of sociobiology threatens free will, it is not because the theory claims we are
deterministically caused by our genes or evolutionary history to act in certain ways, but rather
because it suggests severe limits to our abilities to overcome our genetic “programming” for
behaviors such as aggression, altruism, or sexual attraction (again, these “programs” could
involve probabilistic causation).58
Compatibilists can and do fret about these scientific threats as much as libertarians. Both
fret about such threats because they conflict with our abilities to deliberate rationally and act on
our consciously considered desires, not because they are specific examples of deterministic
causal chains. Libertarians too often present psychological theories as particular examples of
universal causal determinism, but the latter is neither necessary nor sufficient for the former.
Conversely, compatibilists sometimes ignore the threats of particular scientific theories of human
nature because they associate them with deterministic causation, which they do not see as a
threat to free will.
If particular scientific theories, such as Freudian psychology, evolutionary psychology
(the new name for sociobiology), or reductive neuroscience, threaten free will, it is not because
their explanations follow the dictum of determinism that every state is necessary given previous
states and the laws of nature. Their explanations may end up being irreducibly probabilistic. But
such indeterminism need not suggest that “anything goes.” Just because the exact time of decay
of a particular atom is indeterministic, this does not entail that the atom is free never to decay
(much less, to jump to the moon). Similarly, if the relationship between genes and certain
behavior in humans is probabilistic, that does not entail that there is no relationship between
them (much less, that a baby with two blue-eyed parents could have brown eyes). In fact,
probabilistic causes can be more limiting than deterministic causes if the alternatives are more
limiting. For instance, a (hypothetical) gene that has a 70% chance of producing early-stage
In fact, behavioristic laws, such as schedules of reinforcement, are almost always presented as probabilistic, but
this does not diminish the worry that behaviorism threatens free will.
See Hospers (1950) for a discussion of the implications of Freudian psychology for free will, implications that are
distinct from the question of causal determinism, though Hospers conflates them.
The term “genetic determinism” adds to the confusion I am describing. If any of our genes determine (causally
necessitate) any of our behaviors—a highly unlikely claim—it is not because the thesis of determinism is true, and
the falsity of the thesis of determinism would not entail that our genes cannot substantially influence many of our
behaviors in a way that conflicts with our conception of free will.
Alzheimer’s before the age of fifty and a 30% chance of producing late-stage Alzheimer’s after
the age of sixty is more limiting (e.g. to our survival) that a gene that has a 100% chance of
producing late-stage Alzheimer’s.59
Scientific theories conflict with free will, to the extent they do, not because of the type of
causal laws they invoke (deterministic or indeterministic) but because of the content of the
theory (i.e. the specific causal stories they suggest). They are threatening to the extent that they
undermine AS conditions. Isiah Berlin writes, “If social and psychological determinism were
established as an accepted truth, our world would be transformed more radically than was the
teleological world of the classical and middle ages by the triumphs of mechanistic principles or
those of natural selection.”60 Again, linking the word “determinism” to social and psychological
sciences is misleading. But Berlin is absolutely right to suggest that such sciences could
radically transform our “notions of choice, of responsibility, of freedom.” They would threaten
such notions, for instance, if they eliminated teleological concepts, such as desires and goals,
from human deliberation and action. Berlin concludes, “There is, as yet, no need to alarm
ourselves unduly.” True, but this is not because the human sciences have turned out to be
indeterministic; it is because they are increasingly recognizing the need to explain, not explain
away (or eliminate), the purposeful aspects of our mental life.
In order to combat the specific threats to free will from scientific theories, compatriotists
do not see indeterminism as the savior. Rather, they focus their arguments on developing
accounts of free will that are consistent with good scientific theories but that offer alternatives to
interpretations of those theories that undermine our freedom. (This will be the approach I take in
Chapter 3 when I defend free will against experimental work in social psychology that suggests
we do not understand the reasons we act the way we do.) Here, I have shown that there is no
reason to link the truth of threatening scientific theories with the truth of determinism. If the
metaphysical thesis of determinism threatens free will, it is not because it is inextricably linked
to scientific theories that threaten our free will. This head can be severed from the “monster of
Incompatibilists often portray determinism as a threat to free will by suggesting that it
entails the predictability of our choices and actions. Laplace’s demon personifies the connection
between determinism and prediction: “Given for one instant an intelligence which could
comprehend all the forces by which nature is animated and the respective situation of the beings
who compose it . . . nothing would be uncertain and the future, as the past, would be present to
its eyes.”61 Such an intelligence (or God or perhaps a supercomputer) could know every total
state of the universe if it knew any total state of the universe along with the laws of nature, as
Mele (1995: 188) offers another example: “King George’s advisors might be able to steer him in the direction of
a handful of options—any of which would serve their purposes, and none of which conduces to George’s aims—
without being able to ensure that he selects their most preferred option.” This undetermined manipulation is more
constraining on George’s ability to get what he wants than if George were causally determined to choose his
Berlin (Four Essays on Liberty, p. 113) quoted in Ayer and O’Grady (1992: 60).
Laplace (1951: 4). Some philosophers (e.g. Karl Popper) define determinism in terms of predictability by science.
I think this is a mistake for several reasons; for instance, quantum mechanics, though indeterministic, offers the best
predictions for certain phenomena (see also Goldman [1970: 171]).
long as those laws were deterministic. Given that our choices and actions are events in the
universe, it would be able to predict them as well.62
This idea is troubling not only because it suggests our future actions could be known
before they happened, but also because it threatens our sense of individuality and control. We
want to believe we know ourselves (at least in some ways) better than anyone else could. And as
Alasdair MacIntyre writes (in reference mainly to psychology and social sciences), “success in
explaining and predicting can never be divorced from success in manipulating and controlling”
(1957: 241). For instance, the external agents described earlier (such as the evil neurosurgeon)
could control and coerce us much more effectively if they could predict how we will act. Anti-
utopian novels like Brave New World (and Walden Two if you see it as anti-utopian) scare us
because they present the possibility that humans can be controlled in predictable ways. The
contentment of the people in the stories does not outweigh our sense that they lack freedom.63
However, there are several reasons to distinguish determinism from predictability. First
of all, indeterminism does not entail lack of predictability. If the best available theory in a
particular domain invokes probabilistic causation, as in quantum mechanics, it does so because
the probabilities involved offer the best predictions. Randomness may be appropriately
contrasted with predictability, but as we have seen, indeterminism need not mean randomness. If
the best theories of human behavior turn out to be probabilistic, that might limit the precision of
some predictions, but it would not mean we were unpredictable. Furthermore, unpredictability
of the sort suggested by randomness seems inimical to free will to the extent it entails that agents
cannot predict or control their own behavior (this is one reason some compatibilists argue
determinism is necessary for free will). As we will see, free will requires that agents can know
how their motivations are related to their actions, and this requires some degree of self-
prediction. What we really seem to want is that other agents—or even conceivable agents like
Laplace’s Demon—cannot predict our behavior as well as, or better than, we can ourselves (as I
will discuss below, we may have what we want, regardless of whether determinism is true).
In any case, determinism may not in fact entail predictability. First of all, determinism
certainly does not entail predictability in practice. Many systems, especially the human brain,
are too complex to predict with complete accuracy given even the most optimistic hopes for
future science and technology. However, there are also reasons to believe determinism does not
entail predictability in principle. First, there is the problem that, despite the truth of
determinism, any predictor that is part of the universe it is describing will be unable to predict its
own future behavior perfectly.64 There are also two speculative scientific theories that suggest
that determinism does not entail predictability.65
This view suggests a connection between causal determinism, theological determinism (the threat of God’s
foreknowledge of our actions), and logical determinism (the threat that propositions about our actions must be true
or false timelessly and so the actions they describe cannot be other than they are).
It should be noted that, while these novels do illustrate certain threats to our free will, it is not because they
present a universe in which determinism holds (or indeterminism for that matter).
Perhaps this is irrelevant if we are worried only about others predicting our behavior, but it may be relevant if we
are worried that determinism undermines the rationale for deliberation. See Dennett (1984a, chapter 5).
Alvin Goldman (1970, chapter 6) also offers some philosophical reasons to “carefully distinguish between
determinism and predictability” (171). His admonition supports my point, though his arguments are directed
specifically to undermine the claim made by “anti-predictionists” that human actions are not determined, because if
they were, they would be predictable, and they are not predictable. Most of Goldman’s arguments are aimed at
First, some interpretations of nonlinear dynamics (chaos theory) suggest that certain
systems, though deterministic, may nonetheless be unpredictable in principle—i.e. not simply
because we lack some relevant information. John Dupre writes of such a system: “no
measurements to a finite degree of precision of parameters at a time would suffice to predict its
state at a future time. Many physical systems, for example those in meteorology, are
hypothesized to obey such functions, and thus to be both deterministic and in principle
unpredictable.”66 The basic idea is that unmeasurable differences lead to measurable differences.
Of course, even if these interpretations turn out to be unfounded and deterministic but complex
events, such as human actions, are predictable in principle, chaos theory still suggests that these
events, though deterministic, are not predictable in the way imagined within the Newtonian
mindset of mechanical systems. The more complex the system, the more difficult it is to take
into account all the minute factors that can make significant differences. Given the complexity
of our brains alone and their sensitivity to external events, predicting our actions will never be
like predicting eclipses, the motions of the planets, or the arc of a cannonball.67
A second theory that suggests a disconnect between determinism and predictability
involves a particular interpretation of quantum mechanics: Bohm’s theory. Contrary to the
“orthodox” (realist) interpretation that suggests subatomic particles really exist in indeterminate
states, Bohm’s theory suggests that a particle is assigned a probability of having a particular
position, but it actually has a determinate position and that position is deterministically caused by
earlier states. That is, the theory suggests both ontological determinism and unpredictability.
This unpredictability is not due to our epistemological limitations; it is due to the nature of the
phenomena being measured. The theory entails that another system (e.g. a measuring device or
human observer) cannot access the (determinate) position of the particle; the information is
limited to probabilities so that predictions are imprecise just as they are under the “orthodox”
interpretation. Barry Loewer suggests that if “some of these uncertainties infect an agent’s
actions [then] even on the deterministic Bohmian theory it would sometimes be impossible to
predict with certainty an agent’s choices.”68
These interpretations of chaos theory and quantum mechanics indicate that it is “a
mistake to assume that, in a deterministic universe, even Laplace’s demon could predict the
evolution of events.”69 They equally indicate that the agent himself would be limited in his
ability to predict (some of) his own actions, at least at the level of precision associated with a
Laplacian demon. In fact, however, we don’t require that level of precision in predicting our
own behavior. We just want to be able to know how we are likely to act given our beliefs and
refuting the claim that human actions are not—or could not be—predictable. I am more concerned with showing
that determinism (not just of human actions) does not entail predictability.
Dupre (1993: 175). See also references in Kane (1996: 231, note 5) and Juarrerro (1999).
Furthermore, predictions made about increasingly distant times in the future will require increasingly more
information. For instance, a prediction about any event 10 years from now would require the predictor to have
information about the state of a sphere 10 light years in diameter (thanks to David Sanford for this point). Unless
the predictor could calculate instantaneously (like God perhaps), the information would become irrelevant by the
time any predictions could be made.
Loewer (1996: 111, note 24); see also p. 99.
Dupre (1993: 3).
desires in a particular situation. And we want our predictions about our own actions to be, in
general, more accurate and extensive than the predictions of others are.
There is good reason to believe that this is—and will remain—true, regardless of whether
or not we are deterministic systems. This is because humans are conscious creatures; we directly
experience mental states that (in some cases) are the causes of our actions. And no one else can
directly experience our mental states. As Owen Flanagan puts it: “The structure of the nervous
system accounts for the happy fact that we each have our own, and only our own, experiences”
(1992: 107). Despite the general acceptance that our conscious mental states are directly related
to (e.g. supervene on, are identical to) physical states of our bodies, an observer of those physical
states is unlikely to be able to make certain predictions about us. An observer of another agent’s
brain states, for instance, will have to make connections and calculations about those states that
the agent, being directly hooked up to them, can perceive directly and immediately. The agent
has a different type of access to and information about some of his own mental states than can be
obtained with information about the physical states of the agent’s brain and body.70 If such first-
personal information about conscious states is, in certain cases, crucial for the agent to know
what he is likely to do, then he should be able to know what he will do in a way that an observer
The significance of this discrepancy of knowledge between subject and observer would
be challenged, however, if our access to our conscious mental states does not offer reliable
information about how we will act. Indeed, certain scientific theories and certain reductive
theories of mind suggest that our knowledge of our own mental states does not map onto the
actual causes of our behavior and does not provide reliable information about our actions. They
suggest, in fact, that knowledge of the physical states that underlie our conscious states and of
the environmental conditions in which the agent behaves are, in all cases, more accurate and
reliable for predicting behavior than the conscious states themselves. But it is important to
remember that these theories (which I will discuss in Chapter 3 and the Epilogue) do not rely on
determinism to be threatening.
I have not suggested that predictability is unrelated to free will. It is, most obviously,
because we do not want to be predictable in a way that makes us controllable by other agents.
Instead, I have suggested that predictability is not necessarily entailed by determinism and that
indeterminism does not entail unpredictability. Furthermore, we should keep in mind that to any
extent a system is unpredictable (due to deterministic chaos or Bohmian theory or due to
indeterministic factors), everyone will be equally “in the dark.” So, the agent herself will be
unable to predict anything about her future mental states or actions that is also unknowable to
This idea is based on so-called perspectivalist views that suggest there is a difference between information gained
through conscious experiences and information gained through observation of the physical states underlying such
experiences (e.g. Jackson’s knowledge argument ).
This argument, too quickly sketched here, will be developed further in the Epilogue. The idea, put simply, is that
the first-person perspective offers a mode of presentation of information that is different than the third-person
perspective. This difference in information available from different perspectives may entail a discrepancy in the
ability to make predictions based on such information. If so, then the conscious agent may be able to make
predictions about the relations between his conscious mental states and his actions that no one else can make as
accurately (or at least as quickly), even if the agent’s brain is a deterministic system, and hence even if the relations
between brain states could be predicted by Laplace’s demon or a well-informed neuroscientist. (I’ll ignore God,
though it is an interesting question to consider whether God can know—experience—what we experience in the way
we do; if so, it raises the interesting theological problem of God experiencing the desire to rape or plunder!).
any observer.72 Especially when we recognize the importance to free will of knowing one’s own
motivational states, the attraction of any unpredictability offered by indeterminism fades
(especially since probabilistic theories may offer the best predictions of our behavior). If we
want to be unpredictable to other agents, we should not look to indeterminism. Instead, we
should look, first, to the fact that we are complex enough that we will remain unpredictable in
practice despite any advances in the human sciences and, second, to the fact that we are
conscious creatures—the way we are hooked up to ourselves likely offers us more knowledge
about (some aspects of) ourselves than can be had by others. In any case, we have seen that
determinism does not necessarily entail predictability; to the extent predictability threatens free
will, it is not because it is part of the “monster of determinism.”
D. Reductionism and Epiphenomenalism
One of the ways incompatibilists make it seem as if the laws of nature constrain us (as
described in subsection A above) is by suggesting that these laws are the laws of physics, and
since we are physical systems, our actions are dictated by the history and laws of the physical
universe. Peter van Inwagen, for instance, defines determinism in terms of “the state of the
entire physical world” and “the laws of physics.”73 Hence, incompatibilists subtly import
reductionism into their arguments for the incompatibility of free will and determinism by
suggesting that all causal interactions could be explained by reference to the laws of physics
alone. Not only do past events sufficiently cause future events, but lower-level events (e.g. what
physics describes) determine higher-level events (e.g. our conscious experiences).
Incompatibilists often go on to describe reductionism in a way that suggests
epiphenomenalism, the idea that our conscious deliberations do not do any causal work since all
the action, so to speak, happens at the lower level. It then looks as though it is a particular
implication of causal determinism that our thoughts and deliberations are epiphenomenal; they
are determined by mechanistic brain events that we cannot consciously influence and these are
ultimately determined by external events over which we have no control.74 As Richard
Swinburne puts it, “an agent would not be morally responsible at all if he was caused necessarily,
predetermined, to try to do what he did, by his brain state, and that in turn by some prior state,
until we come to causes outside the agent’s body and ultimately to causes long before his
This point will arise again when I discuss Kane’s libertarianism. The indeterminism required by his theory
suggests that an agent will not know what she decides until she decides: “the agents will settle the issue of which is
wanted more by deciding” (1996: 133).
van Inwagen (1975: 47), my italics. I do not think this description invalidates his argument for incompatibilism
(which I will discuss below), but it does make it look as if the physical states and laws of the universe control
everything else, including us.
Norman Malcolm (1968) and Charles Taylor (1971) argue that the teleological concepts required for free will are
incompatible with mechanism, and they suggest that mechanism is “a special application of physical determinism”
(Malcolm, p. 127). I agree will the first claim but disagree with the second (see Epilogue).