Freedom and Determinism
Freedom and Determinism
Freedom and Determinism
edited by
Joseph Keim Campbell,
Michael O’Rourke,
and David Shier
Editors
Joseph Keim Campbell, Washington State University
Michael O’Rourke, University of Idaho
David Shier, Washington State University
A Bradford Book
The MIT Press
Cambridge, Massachusetts
London, England
© 2004 Massachusetts Institute of Technology
All rights reserved. No part of this book may be reproduced in any form by any elec-
tronic or mechanical means (including photocopying, recording, or information
storage and retrieval) without permission in writing from the publisher.
This book was set in Stone Serif and Stone Sans by Interactive Composition Corpo-
ration and was printed and bound in the United States of America.
Freedom and determinism / edited by Joseph Keim Campbell, Michael O’Rourke, and
David Shier.
p. cm.—(Topics in contemporary philosophy)
“A Bradford book.”
Includes bibliographical references and index.
ISBN 0-262-03319-4 (hc. : alk. paper)—ISBN 0-262-53257-3 (pbk. : alk. paper)
1. Free will and determinism. I. Campbell, Joseph Keim, 1958– . II. O’Rourke,
Michael, 1963– . III. Shier, David, 1958– . IV. Series.
BJ1461.F755 2004
123'.5—dc22
2003066491
10 9 8 7 6 5 4 3 2 1
Contents
4 Trying to Act 89
Carl Ginet
Index 323
Freedom and Determinism: A Framework
For this reason, debates about freedom and determinism center on a set
of distinct, though related, questions.4
1. What is the thesis of causal determinism?
2. What is the nature of moral freedom?
3. Is moral freedom compatible with causal determinism?
4. Is causal determinism true?
5. Are any persons morally free?
The first three questions are conceptual and, thus, the primary focus of
philosophers is directed towards answering those questions. A provisional
answer to (1) was given above, but more precisely, determinism is the con-
junction of the following two theses:
For every instant of time, there is a proposition that expresses the state of
the world at that instant;
If p and q are any propositions that express the state of the world at some
instants, then the conjunction of p with the laws of nature entails q. (van
Inwagen 1983, 65)
Compatibilism Yes
Incompatibilism No
Soft Determinism Yes Yes Yes
Hard Determinism No Yes No
Libertarianism No No Yes
During the last century a few of the above theories began to look rather
archaic. First of all, certain results in quantum mechanics led many philoso-
phers eventually to reject determinism. According to the standard interpre-
tation of quantum theory, there are facts about the sub-atomic level—facts
like this electron has the property of “spin up”—that are not a logical conse-
quence of any conjunction of past facts and laws of nature. Subsequently,
many philosophers have abandoned the positions of soft and hard deter-
minism since they both entail a positive answer to (4), and the acceptance
of determinism.
In addition, philosophers began to question the relevance of the concept
of free will to moral praiseworthiness and blameworthiness. Motivation for
the traditional view of moral freedom stems from an acceptance of some-
thing like the following principle:
The principle of alternative possibilities (PAP): S is morally responsible
for action a only if she has, or had, alternatives to a.
But in an important and influential paper, Harry Frankfurt provides appar-
ent counterexamples to PAP, cases in which agents perform blameworthy
actions even though it appears that they could not have done otherwise.
Frankfurt writes:
Suppose someone—Black, let us say—wants Jones4 to perform a certain action. Black
is prepared to go to considerable lengths to get his way, but he prefers to avoid
Freedom and Determinism: A Framework 5
showing his hand unnecessarily. So he waits until Jones4 is about to make up his
mind what to do, and he does nothing unless it is clear to him (Black is an excellent
judge of such things) that Jones4 is going to decide to do something other than what
he wants him to do. If it does become clear that Jones4 is going to decide to do some-
thing else, Black takes effective steps to ensure that Jones4 decides to do, and that he
does do, what he wants him to do. Whatever Jones4’s initial preferences and inclina-
tions, then, Black will have his way. (Frankfurt 1969, 148–149)
Perhaps Black wants Jones4 to rob a bank, and that Jones4 robs the bank
without the need of Black’s counterfactual intervention. It seems, then,
that Jones4 is morally responsible for his action but that he could not
have done otherwise. Call such examples ‘Frankfurt examples’. To ac-
cept that Frankfurt examples are genuine counterexamples to PAP is to ac-
cept a split between moral freedom and alternatives, for the Frankfurt
examples suggest that one can have the former without having the latter.
This means that the traditional view is wrong about the nature of moral
freedom.6
Lastly, some philosophers have abandoned the traditional view because
of the development of a number of persuasive formal arguments in support
of incompatibilism, presented primarily by Peter van Inwagen (1975, 1983,
1989) and Carl Ginet (1966, 1990). These are all versions of the Consequence
Argument.
If determinism is true, then our acts are the consequences of the laws of nature and
events in the remote past. But it is not up to us what went on before we were born,
and neither is it up to us what the laws of nature are. Therefore, the consequences of
these things are not up to us. (van Inwagen 1983, 16)
where ‘ ’ represents broad logical necessity and ‘→’ represents material im-
plication (1983, 94; 1989, 227). From the above rules and definitions, to-
gether with reasonable assumptions about the past and the laws of nature,
6 J. K. Campbell, M. O’Rourke, and D. Shier
one can show that no one has, or ever had, any choice about whether any
true proposition is true, given determinism. Thus, the Third Argument
apparently establishes incompatibilism.7
The three developments discussed above have led to a number of new
theories in the freedom and determinism debate. Among the more attrac-
tive of these is semicompatibilism, inspired by the rejection of PAP and the
success of the Consequence Argument (Fischer 1994). According to the
semicompatibilist, moral freedom is not essentially linked to alternatives.
The semicompatibilist can agree that free will is incompatible with deter-
minism but also contend that determinism and moral responsibility
are compatible. Moral freedom is best understood in terms of concepts like
guidance control and reasons-responsiveness that rely on features about the
actual causal history leading up to an action, instead of whether or not
there are any alternatives to the action.8
Libertarianism is still widely held.9 Moreover, many incompatibilists
deny the free will thesis altogether. Hard determinism is rarely endorsed,
since few philosophers accept determinism, but in its place are successor
views, like hard indeterminism, according to which incompatibilism is true,
determinism is false, but the free will thesis is false anyway (Pereboom
2001). Most successor views differ from traditional versions of moral ni-
hilism. Traditionalists begin with the Consequence Argument. Hard deter-
minists then argue that determinism is true, and thus all human actions are
unavoidable. Other traditionalists extend the Consequence Argument to
cover cases of indeterminism, as well. For instance, after presenting the Third
Argument one can go on to show that most propositions are unavoidable
even if one assumes the truth of indeterminism (van Inwagen 1983, 1989).
Proponents of successor views often believe that unavoidability is unimpor-
tant to moral freedom, given the Frankfurt examples. Instead, they claim
that moral freedom is essentially linked to origination, and claim that origi-
nation is impossible under the assumption of determinism and unlikely at
best even if indeterminism is true.10
In summary, the majority of contemporary philosophers agree that some
kind of freedom—moral freedom—is required for moral responsibility. But
they differ as to the nature of this freedom as well as some of the other nec-
essary conditions for moral responsibility. Proponents of the traditional
view continue to maintain that moral freedom is just free will, but a variety
of philosophers have rejected the latter notion altogether. This is primarily
Freedom and Determinism: A Framework 7
due to the impact of the Frankfurt examples and formal arguments for
incompatibilism. Moreover, while debate about the compatibility of moral
freedom and determinism is still alive and well, most philosophers have
rejected determinism given quantum mechanics. Gone are the labels of
soft determinism and hard determinism, but the lion’s share of opinions on
the nature of freedom and determinism still fall into three main groups:
libertarians, moral nihilists, and compatibilists, including semicompati-
bilists. We now introduce the essays in this current volume.
As we have seen, there are essentially two projects that are important to
contemporary philosophers working in the area of freedom and determin-
ism: (a) understanding the nature of moral freedom and its relationship
with moral responsibility, and (b) determining whether or not moral free-
dom is compatible with determinism. That is, contemporary philoso-
phers are primarily concerned with questions (2) and (3) from our earlier
list. However, one cannot address these questions without a provisional
understanding of the general concepts involved—determinism and moral
freedom—as well as a clearer sense of the type of events relevant to our con-
cerns here, namely, actions. These presuppositional topics are the subject
of the essays in our first section, “Determinism, Freedom, and Agency.” We
then turn to the nature of moral freedom and responsibility in “The Meta-
physics of Moral Responsibility,” and finally, to questions about the com-
patibility of freedom and determinism in “The Compatibility Problem.”11
which includes the ultimacy condition. But Kane distances himself from
other, more classic libertarian theories, like the libertarian agency theory,
which holds that agent causation is sui generis and not reducible to event
causation.13 Instead, Kane adopts a teleological theory that attempts to
explicate moral freedom in terms of the reasons or purposes of the agent.
Kane closes his essay with a lengthy response to allegations that the no-
tion of indeterministic action is incoherent.
The kinds of things that we are normally held responsible for are actions,
or their consequences, so the problem of freedom and determinism is es-
sentially a problem about reconciling free actions with determinism. Thus,
in addition to delving into the intricacies of the latter view, theorists often
engage in important work on the nature of human action. In “Trying to
Act,” Carl Ginet offers four conditions, each of which he claims is sufficient
for the truth of sentences of the form, ‘S tried to a’. Ginet falls short of of-
fering an analysis since he does not claim that any, or all, of the conditions
are necessary for the truth of such sentences. Still, Ginet notes some im-
portant features that all four conditions share. For instance, in each condi-
tion S tries to a by doing something else which S believes is connected in a
certain way with the possibility of doing a. To try to do a is to do something,
even if one fails to do a, so these are also conditions for the truth of sen-
tences reporting the performance of a certain class of actions.
Establishing (R) is important for at least two reasons. First, it is one of the
most central and universally agreed upon beliefs that we have about moral
freedom. Second, philosophers like Immanuel Kant use (R) as an important
premise in arguments against moral nihilism. After considering and dis-
missing another version of (R), Nelkin concludes that the ‘belief-concept’
reading of it should be accepted. Thus, according to Nelkin, all rational
deliberators essentially believe that their actions are up to them in the
sense that they are accountable for those actions.
10 J. K. Campbell, M. O’Rourke, and D. Shier
agents are morally responsible for their actions in the actual sequence sce-
nario yet these same agents would not have been morally responsible for
their actions were the counterfactual intervention to be enacted. Fischer
and Ravizza are then faced with a dilemma. Either they must concede that
there is more to moral responsibility than an action’s being the result of a
mechanism that is MRR or they must give up their belief that their theory
explains why agents in the Frankfurt examples would not have been
morally responsible for their actions had the counterfactual intervention
taken place. Long also suggests that agents may be morally responsible for
actions even if they undergo external manipulations that are quite severe.
Nomy Arpaly’s paper, “Which Autonomy?” is a general critique of the
concept of autonomy, one of the more popular candidates for moral freedom.
First, Arpaly discusses a variety of divergent accounts of autonomy that
philosophers have offered. This alone poses a problem for such theorists
since the diversity of such accounts has rendered the concept of autonomy
practically meaningless. Arpaly goes on to offer substantive criticisms of
each of these accounts and ends by advising that we might be better off deal-
ing more directly with moral praiseworthiness and blameworthiness and
abandoning the notion of autonomy altogether.
We close the second section with an essay by John Martin Fischer enti-
tled “The Transfer of Nonresponsibility.” Fischer’s contribution is part of a
long-standing debate—one in which he has exerted substantial influence—
on the Principle of the Transfer of Nonresponsibility (Transfer NR), a ver-
sion of which was put forth by van Inwagen. Van Inwagen’s principle is
formally identical to his rule (), noted above, only here ‘Np’ means “p and
no human being, or group of human beings, is even partly responsible for
the fact that p” (van Inwagen 1980). Both rule () and Transfer NR can be
seen as closure principles of unavoidability and nonresponsibility, respec-
tively. Fischer formulates the principle as follows:
Transfer NR: If p obtains and no one is even partly morally responsible for
p and if p obtains, then q obtains, and no one is even partly morally
responsible for the fact that if p obtains, then q obtains, then q obtains,
and no one is even partly morally responsible for q. (Fischer, chapter 9)
Transfer NR has recently come under attack by semicompatibilists, like
Fischer, who argue that the Frankfurt examples are counterexamples to this
principle too. To use an example from Fischer’s essay, suppose that q is the
12 J. K. Campbell, M. O’Rourke, and D. Shier
fact that an avalanche occurs and that the avalanche was started by an
agent, Betty, but that the avalanche would have occurred even if Betty had
not started it due to the fact that a glacier was eroding, p. No one is respon-
sible for the fact that p, and no one is responsible for the fact that if p, then
q, but it seems that Betty is responsible for the fact that q.14 Proponents of
Transfer NR have tried to alter the principle to meet this counterexample.
Fischer considers and responds to a variety of these reformulations. He con-
cludes that the newer versions of Transfer NR cannot be used to support the
incompatibilism of determinism and moral responsibility. At most, he says,
they lead only to a dialectical stalemate.
The principle of epistemic closure: From S knows that p and S knows that
p entails q, deduce S knows that q.
Notes
1. See Aristotle (1963) along with writings by the Stoics and Epicureans in Long and
Sedley (1987). This is disputed by Gier and Kjellberg, chapter 13, who argue that the
problem is a modernist one.
3. See van Inwagen (1998, 365–366). The term ‘metaphysical freedom’ is van
Inwagen’s. The term ‘active power’ comes from Reid (1983).
4. For a related, though somewhat different set of questions and problems, see van
Inwagen (1983, 1–2) and Kane, chapter 3.
5. Following van Inwagen (1983, 8), we use the term ‘free will’ out of respect for tra-
dition. By use of the term we do not suggest that there is some faculty, e.g., the will,
that has the property of being free. To say that S has free will is just to say that S has
alternatives, in the sense noted in this paragraph. This connection between the con-
cepts of free will and alternatives has been accepted by virtually everyone throughout
the history of philosophy. Frankfurt (1969, 1971) remains an exception.
Freedom and Determinism: A Framework 15
6. Not all philosophers agree that the Frankfurt examples undermine PAP. See the
articles in Fischer (1986, part 2), Fischer and Ravizza (1993, part 3), and Kane (2002,
part 5). For discussions of the Frankfurt examples in this volume, see Haji, chapter 6,
Long, chapter 7, Fischer, chapter 9, and van Inwagen, chapter 10.
7. Let ‘P0’ represent a proposition that expresses the state of the world at some time
t0 before any human beings existed, let ‘L’ represent the conjunction of the laws of
nature, and let ‘P’ represent any true proposition. Here is van Inwagen’s Third Argu-
ment (1983, 93–104; 1989, 404–405).
8. Fischer (1982, 1987, 1994) and Fischer and Ravizza (1998). For other theories of
moral freedom along the lines of semicompatibilism, see Lehrer (1980) and the
articles in Fischer (1986, part 1), Fischer and Ravizza (1993, part 2), and Kane (2002,
part 4). This latter section includes discussions of the new compatibilists, who provide
naturalistic conceptions of moral responsibility inspired by P. Strawson (1962).
10. For versions of the successor view, see Kane (2002, part 7), G. Strawson (1986),
Pereboom (2001), and Honderich, chapter 16. The term ‘successor view’ is Kane’s.
See also Kane, chapter 3, for a discussion of origination. Van Inwagen is not a moral
nihilist though he does embrace a related view, called ‘restrictivism’, defined below.
11. The expression ‘compatibility problem’ is from van Inwagen (1983, 2).
12. We thank Bruce Glymour for providing the basis of this paragraph.
13. See Reid (1983), Campbell (1957), Taylor (1963), and Chisholm (1964). For more
recent versions of the theory, as well as criticisms of it, see articles in O’Connor
(1995) and Kane (2002, part 6). Kane, chapter 2, and van Inwagen, chapter 10, con-
tain interesting criticisms of the concept of agent causation from the libertarian
perspective.
15. For compatibilists representing the former approach, see Lewis (1981) and
Lehrer, chapter 2. For compatibilists representing the latter approach, see Moore
(1912) and Lehrer (1976, 1980). Kane (2002, part 4) also contains some recent essays
of interest on compatibilism.
16 J. K. Campbell, M. O’Rourke, and D. Shier
References
Campbell, C. A. 1957. On Selfhood and Godhood. London: George Allen and Unwin.
Chisholm, R. 1964. “Human Freedom and the Self.” The Lindley Lecture. Lawrence,
Kans.: University of Kansas, Department of Philosophy. Reprinted in Pereboom
(1997).
Eliot, T. S. 1943. “Burnt Norton.” Four Quartets. New York: Harcourt, Brace.
. ed. 1989. God, Foreknowledge, and Freedom. Stanford, Calif.: Stanford Univer-
sity Press.
Fischer, J. M., and M. Ravizza, eds. 1993. Perspectives on Moral Responsibility. Ithaca,
N.Y.: Cornell University Press.
. 1971. “Freedom of the Will and the Concept of a Person.” Journal of Philoso-
phy 68: 5–20. Reprinted in Fischer (1986).
Ginet, C. 1966. “Might We Have No Choice?” In K. Lehrer, ed., Freedom and Deter-
minism. New York: Random House.
Kane, R. 2002. The Oxford Handbook of Free Will. Oxford: Oxford University Press.
Lewis, D. 1981. “Are We Free to Break the Laws?” Theoria 47: 113–121. Reprinted in
Lewis (1986). Philosophical Papers, volume 2. Oxford: Oxford University Press.
Long, A. A., and D. N. Sedley, eds. and trans. 1987. The Hellenistic Philosophers:
volume 1: Translations of the Principle Sources. Cambridge: Cambridge University Press.
O’Connor, T., ed. 1995. Agents, Causes, and Events: Essays on Indeterminism and Free
Will. Oxford: Oxford University Press.
Reid, T. 1983. Inquiry and Essays. R. E. Beanblossom and K. Lehrer, eds. Indianapolis,
Ind.: Hackett.
van Inwagen, P. 1975. “The Incompatibility of Free Will and Determinism.” Philo-
sophical Studies 27: 185–199.
John Earman
1 Introduction
The purpose of this essay is to give a brief survey of the implications of the
theories of modern physics for the doctrine of determinism. The survey
will reveal a curious feature of determinism: in some respects it is fragile, re-
quiring a number of enabling assumptions to give it a fighting chance; but
in other respects it is quite robust and very difficult to kill. The survey will
also aim to show that, apart from its own intrinsic interest, determinism is
an excellent device for probing the foundations of classical, relativistic, and
quantum physics.
The survey is conducted under three major presuppositions. First, I take
a realistic attitude toward scientific theories in that I assume that to give an
interpretation of a theory is, at a minimum, to specify what the world
would have to be like in order for the theory to be true. But we will see that
the demand for a deterministic interpretation of a theory can force us to
abandon a naively realistic reading of the theory. Second, I reject the “no
laws” view of science and assume that the field equations or laws of motion
of the most fundamental theories of current physics represent science’s
best guesses as to the form of the basic laws of nature. Third, I take deter-
minism to be an ontological doctrine, a doctrine about the temporal evo-
lution of the world. This ontological doctrine must not be confused with
predictability, which is an epistemological doctrine, the failure of which
need not entail a failure of determinism. From time to time I will comment
on ways in which predictability can fail in a deterministic setting. Finally,
my survey will concentrate on the Laplacian variety of determinism
according to which the instantaneous state of the world at any time
uniquely determines the state at any other time.
22 J. Earman
Suppose that the world is populated with billiard balls. And suppose that
the laws of motion for this world consist precisely of the specifications that
when two balls collide they obey the standard laws of elastic impact and
that between collisions they move uniformly and rectilinearly. And finally
suppose that atomism is false so that billiard balls of arbitrarily small size
can exist. Then à la Zeno we can string a countably infinite number of unit
mass billiard balls in the unit interval. Assume that at t t* all the balls in
this infinite string are at rest and that coming from the right is a cue ball of
unit mass moving with unit speed (see figure 1.1a). In a unit amount of
time an infinite number of binary collisions take place, at the end of which
each ball is at rest in the original position of its left successor in the series
(see figure 1.1b).1 The time reverse of this process has all the balls initially
at rest. Then suddenly a ripple goes through the string, and the cue ball is
ejected to the right. Futuristic Laplacian determinism is violated since it is
consistent with the laws of elastic impact that the string does not self-excite
but remains quiescent for all time (Laraudogoitia 1996).2
(a)
(b)
Figure 1.1
Zeno’s revenge.
What We Have Learned and What We Still Don’t Know 23
One theme that will be sounded again and again in this section is that
classical spacetimes provide unfriendly and even hostile environments for
determinism. A related theme is that determinism in classical physics is
inextricably linked to basic philosophical issues about the nature of space,
time, and motion. To illustrate the latter theme I assert that Laplacian de-
terminism implies that it cannot be the case that both (i) space is ‘ab-
solute’ in the sense that it is a ‘container’ for bodies (e.g., shifting all the
bodies in the universe one mile to the east results in a new state distinct
from the original state), and (ii) all motion is the relative motion of bod-
ies. The argument is simple. Assumption (ii) implies that only relative
particle quantities, such as relative particle positions, relative particle ve-
locities, relative particle accelerations, and so forth, and not absolute posi-
tion, velocity, acceleration, and so forth, are well-defined quantities. The
appropriate classical spacetime setting that supports this conception of
motion has three elements: planes of absolute simultaneity, which reflect
the observer-independent nature of coexistence; a metric (assumed to
be ⺕3) on the instantaneous three-spaces, which measures the spatial
distance between simultaneous events; and a time metric, which measures
the lapse of time between nonsimultaneous events.3 But if (i) is main-
tained in this spacetime setting, not even a weakened form of Laplacian
determinism can hold for particle motions. In coordinates adapted in the
natural way to the spacetime structure, the symmetries of the spacetime
have the form
t → t t const (1b)
*
t=t
Figure 1.2
An apparent violation of Laplacian determinism.
p ∉ D+ (Σ)
t
space invader
Figure 1.3
Domain of dependence in classical spacetime.
the chosen forms of a(t) and R(t) the two solutions will agree for all t t*
but will disagree for t t* since the two solutions (as indicated by the solid
and the dashed world lines of figure 1.2) entail different future positions for
the particles in the container space.
To save determinism one can reject (i) and claim that the alleged viola-
tion of determinism is spurious on the grounds that once the container
view of space is rejected there is no temptation to see the dashed and solid
lines of figure 1.3 as different future histories rather than as different repre-
sentations of the same history. To make good on this point of view it would
have to be shown how to concoct deterministic and empirically adequate
equations of motion that are formulated entirely in terms of relative
particle quantities. As the history of mechanics shows, this is not an easy
row to hoe. But that is a story I don’t have time to recount here.
What We Have Learned and What We Still Don’t Know 25
The alternative way to save determinism is to reject (ii) and beef up the
structure of the spacetime by adding, say, inertial structure to make well-
defined quantities like absolute acceleration. This additional structure
linearizes the transformations (1) to
x → x Rx vt c (2a)
t → t t const (2b)
where is the mass density. The gravitational force acting on a massive test
body moving in the Newtonian gravitational field is proportional to .
Even the weakest form of Laplacian determinism fails because if is a solu-
tion to (3), then so is f, where f(x, t) is any function linear in x. By
choosing f(x, t) to be 0 for all t t* but different from 0 for t t*, we pro-
duce solutions for which the test body feels exactly the same gravitational
force and has exactly the same motion in the past but feels different forces
and, hence, executes different motions in the future. The nondeterministic
solutions can be killed by excluding the homogeneous solutions to (3).
This exclusion amounts to a declaration that the Newtonian gravitational
field has no degrees of freedom of its own and is only an auxiliary device
for describing direct particle interactions. Let us then turn to the pure
particle description of Newtonian gravitation.
Consider a finite number of point mass particles interacting via Newton’s
1兾r 2 force law. Let’s simply ignore problems about collision singularities
by focusing on solutions that are collision-free. Nevertheless, after many
decades of work, it has been established that noncollision singularities can
occur; that is, drawing on the infinitely deep 1兾r potential well, the parti-
cles can accelerate themselves off to spatial infinity in a finite amount of
26 J. Earman
time (Xia 1992). The time reverse of such a process is an example of ‘space
invaders’, particles appearing from spatial infinity without any prior warn-
ing. To put it crudely, you can’t hope to have Laplacian determinism for
open systems, and for the type of interaction under discussion, the entire
universe is an open system.
To save determinism from this threat, three approaches can be taken. The
first, and least interesting, is to impose boundary conditions at spatial
infinity to rule out space invaders. This smacks of making determinism true
by creating a postulate of wishful thinking. The second is to maintain
the idealization of point mass particles while adding to Newton’s 1兾r 2 attrac-
tive force a short-range repulsive force which doesn’t affect predictions for
particles with large spatial separations, but which is such that the total po-
tential well is no longer infinitely deep. One would then have to show that
the combined force law gives rise to a well-posed Laplacian initial value prob-
lem. I am not aware of any results to this effect, but I see no in-principle ob-
stacles to achieving them. The third alternative is to move from point mass
particles to a particle with a finite radius, and to postulate that when two par-
ticles collide they obey, say, the laws of elastic impact. Even if it is assumed
that atomism is true in a form that excludes the Zeno examples of section 1,
this tack can run aground on at least two shoals. (a) If triple collisions occur
the result is generally underdetermined since there will be more unknowns
than there are governing equations. (b) Even if attention is restricted to bi-
nary collisions the uniqueness of solutions can fail if an infinite number of
particles are present in the universe. Turn off the gravitational interactions
of the particles and suppose that they interact only upon contact. Lanford
(1974) constructed a solution4 in which all the particles are at rest for all t t*
but for any t t* all but a finite number of particles are in motion. Thus,
the equations of motion don’t determine whether a quiescent past is to be
extended into the future by a ‘normal’ solution in which the particles con-
tinue to be quiescent or by an ‘abnormal’ solution in which the particles ap-
pear to self-excite. Determinism can be saved either by banning infinities of
particles or by imposing boundary conditions at infinity which prevent a too
rapid increase in the velocities of particles as one goes out to infinity. Either
move smacks of making the world safe for determinism by fiat. A more in-
teresting saving move would be to show that analogues to Lanford’s solution
cannot be constructed if elastic collisions are replaced by a smooth short-
range repulsive force. I am not aware of any results to this effect.
What We Have Learned and What We Still Don’t Know 27
Let us now leave particles to consider pure field theories. A familiar field
equation in Newtonian physics is the Fourier heat equation
⭸
2 ⭸t (4)
D(兺) 艛 D(兺). The idea is that, if the laws of physics are cooperative, the
state on 兺 suffices to fix the state throughout D(兺); but for a point q 僆 D(兺)
it is hopeless to try to determine the state at q from the state on 兺 since
events at q can be influenced by a causal process that does not register on
兺. Using D(兺) we can give a general definition of a Cauchy surface that ap-
plies to general relativistic spacetimes (to be discussed in section 6) as well
as to Minkowski spacetime and also—with suitable allowances—to classical
spacetimes: it is a global time slice 兺 (i.e., a spacelike hypersurface without
edges) such that 兺 is achronal (i.e., is not intersected more than once by
any future directed timelike curve) and such that D(兺) is the entire space-
time M. The t const level surfaces of any inertial time t for Minkowski
spacetime are, of course, Cauchy surfaces. But the level surfaces of absolute
simultaneity of Newtonian spacetime are not Cauchy (see figure 1.3); indeed,
in this case D(兺) 兺 for 兺 as t const.
If determinism fares better in Minkowski spacetime, prediction certainly
does not. The basic problem is that the very null cone structure that helps
to make the special relativistic setting friendly to determinism makes it im-
possible to acquire the information needed for a prediction prior to the oc-
currence of the events to be predicted. To make this precise a few additional
definitions are needed. Define the causal (respectively, chronological ) past of
a point p 僆 M, J(p) (respectively, I(p)) to be the set of all q 僆 M such that
there is a future-directed causal (respectively, timelike) curve from q to p.
Then for a point p 僆 M take the domain of prediction P(p) of p to be the set
of all q 僆 M such that (i) every past inextendible causal curve through q en-
ters J(p), and (ii) I(q) 債 I(p).5 Condition (i) is needed to assure that an ob-
server at p can, in principle, have causal access to all the processes that can
influence the events at q, and condition (ii) is needed to assure that from
the perspective of an observer at p the events to be predicted at q have not
already occurred. The reader can now verify that for any point p of
Minkowski spacetime, P(p) .
simple fact has caused some consternation since it isn’t evident how, con-
sistent with the correspondence principle, chaos can emerge from QM in
some appropriate classical limit (Belot and Earman 1997).
None of the remarks so far touches the core problem of determinism in
QM. But that problem is difficult to discuss, or even to formulate, because
it is bound up with a contentious issue about the nature of quantum ob-
servables; namely, under what conditions do the quantum observables, as
represented by self-adjoint operators on a Hilbert space, take on definite
values? One answer endorsed by standard textbooks on QM is that an ob-
servable A takes the definite value a at t iff the state (t) is an eigenstate of
the operator  with eigenvalue a.6 This eigenvector-eigenvalue rule—or any
value assignment rule which says that an observable has a specified definite
value just in case the state vector has some specified characteristic—
together with deterministic Schrödinger evolution for would mean that
Laplacian determinism holds for quantum systems.7
The well-known difficulty with the eigenvector-eigenvalue value assign-
ment rule is that it implies the unacceptable result that, after an interaction
with an object system, the pointer on the dial of a measuring instrument
that couples to the observable whose value for the object system we wish to
ascertain has no definite value if the initial state of the object system was
a superposition of eigenstates of the observable being measured. One at-
tempt to overcome this embarrassment involves ‘state vector reduction’: at
some juncture during the measurement process, Schrödinger evolution
ceases and the state vector jumps into a simultaneous eigenstate of the
object observable and the macro pointer position observable that serves
as an indicator of the value of the object observable. The violation of the
Schrödinger equation is, of course, a violation of Laplacian determinism;
but statistical determinism is maintained by the assumption that the
propensity of the state vector to collapse into a given eigenstate is governed
by the Born rule probability calculated from the state vector just before
collapse.
The collapse solution to the measurement problem comes at a high cost.
The original theory with an embarrassing consequence has been replaced
by a nontheory: the objection is not simply that the state vector reduction
is a miracle—a violation of the Schrödinger equation—but that since ‘mea-
surement’ is a term of art whose application is not specified by the theory,
exactly when and under what circumstances the reduction takes place is
32 J. Earman
left dangling. Some physicists have proposed to bridge this gap by describ-
ing a mechanism for state vector reduction by means of a nonlinear wave
equation or a stochastic differential equation (see Ghirardi et al. 1986 and
Pearle 1989). The extant state vector reduction schemes are nondetermin-
istic; but it is an open question whether they must have this property.
The major alternative route to a solution of the measurement problem in-
volves a modification of the eigenvector-eigenvalue value assignment rule
rather than of the state vector dynamics. The different paths that branch off
this no-collapse route can lead to quite different conclusions about deter-
minism, as is illustrated by the Bohm interpretation, the modal interpreta-
tions, and the many worlds/many minds interpretations. According to the
first, particles always have definite positions, and these positions evolve
deterministically via an equation of motion that is parasitic on the
Schrödinger evolution of the state vector. Assuming that all laboratory mea-
surements can be reduced to recording positions, the Bohm interpretation
offers a deterministic explanation of how and when other quantum me-
chanical observables, such as spin, take on definite values. It is shown that
the orthodox quantum mechanical probabilities are recovered for all times
on the assumption that the initial probability distribution at t 0 over par-
ticle positions q is equal to the quantum mechanical prescription 冷 (q, 0)冷2.
By contrast, modal interpretations don’t seek to preserve determinism but
settle for the more modest aim of breaking the eigenvector-eigenvalue link
open wide enough to make sure that measurements have definite outcomes.
This aim is accomplished by a value assignment rule which makes use of a
special decomposition of the vector for composite systems (the ‘biorthogo-
nal decomposition’). This decomposition identifies a privileged class of
observables which are said to possess definite values without saying which
particular values are possessed. Some modal theorists want to add to their
value assignment rule a dynamics for possessed values. There is a plethora of
such dynamical schemes (see Dickson 1998). Typically these schemes are not
deterministic. However, a recent proposal by Ax and Kochen (1999) restores
determinism by construing the phase of the state vector as an additional
hidden parameter—the different unit vectors in a ray of Hilbert space are
taken to correspond to different members of an ensemble, and the apparent
indeterminism of QM is due to a random distribution of initial phases.
Many worlds interpretations come in two main versions: the literal ver-
sion according to which a measurement event corresponds to a physical
What We Have Learned and What We Still Don’t Know 33
splitting of the world observer (qua physical object), and the figurative
version according to which it is not the physical world but the mind or
mental state of the observer that splits. On the literal version determinism
fails if there is a fact of the matter about which postmeasurement observer
(qua physical object) is identical with the premeasurement observer. By the
same token, measurement is not an indeterministic process if genidentity
of observers is denied. A parallel conclusion holds for the many minds ver-
sion, with minds in place of bodies. It is hard to take any of this seriously,
not so much because of the metaphysical extravagances, but because it is
not at all evident that the many worlds/many minds interpretation does re-
solve the measurement problem: it doesn’t explain exactly when and under
what circumstances the splitting takes place, and it doesn’t explain why the
splitting takes place in some bases of the Hilbert space but not others.
There is no easy way to summarize the status of determinism in QM, but
two points need to be underscored. The proponents of the Bohm interpreta-
tion claim that it recovers all quantum mechanical predictions, statistical and
nonstatistical, about experiments. If this is correct, it means that in a world of
ordinary nonrelativistic QM the fate of Laplacian determinism would have to
be decided by nonempirical factors. The second remark concerns the claim,
sometimes heard, that determinism is defeated by various no-go results for
hidden variables—for example, those of the Kochen–Specker type and those
that flow from the Bell inequalities. What these no-go results show is that,
consistent with certain plausible mathematical restrictions on value assign-
ments and/or with the statistical predictions of QM, some set of quantum
observables cannot be assigned simultaneously definite values lying in the
spectra of these observables. By itself this hardly defeats Laplacian determin-
ism; nor is it even a terribly surprising conclusion—after all, many classical
quantities are best construed as having a dispositional character in that they
only take on definite values in limited contexts. That it is impossible to view
the values of quantum observables as supervening on more fundamental
quantities whose temporal evolution is deterministic would require a differ-
ent kind of proof, a proof whose existence is highly dubious in view of the
example of the Bohm interpretation.
We saw that moving classical mechanics from the Newtonian setting to
the special relativistic setting improves the fortunes of determinism. The
same cannot be said for quantum mechanics; for the fortunes of deter-
minism in QM are bound by the knots of the measurement problem and
34 J. Earman
the value assignment problem, and in the relativistic setting these knots are
drawn even tighter. State vector reduction is supposed to take place instan-
taneously. Transposing this notion to the relativistic setting seems to
require either breaking Lorentz invariance or relativizing states to spacelike
hyperplanes (Fleming 1996), neither of which is an attractive option. The
relativistic setting is also unkind to the modal interpretation because it is
difficult to reconcile the value assignment rule of this interpretation with
Lorentz invariance (Dickson and Clifton 1998). Nor does the Bohm inter-
pretation fit comfortably with relativistic quantum field theory (QFT) since
the ontology of this theory is best construed not in terms of particles but in
terms of local field observables. The general idea of Bohmian dynamics can
be applied to observables other than particle position. But in the absence of
a demonstration that in measurement contexts the field observables chosen
for Bohmian treatment will exhibit the localization needed to explain mea-
surement outcomes, one of the major reasons for finding the Bohm inter-
pretation attractive has been lost (see Saunders 1999). Furthermore, when
QFT is done on curved spacetime the dynamics for the quantum field may
not be unitarily implementable (see Arageorgis et al. 2002), making it harder
to construct a field theoretic analogue of the Bohm dynamics for particle
position.
In Einstein’s general theory of relativity (GTR) we have to deal not with one
spacetime setting but many since different solutions to the Einstein gravi-
tational field equations (EFE)
1
Rab 2 Rgab Tab (6)
null cones
t
timelike curve
Σ
Figure 1.4
Behavior of null cones in anti–de Sitter spacetime.
Specifying the metric field and its normal derivative on some space-
like slice 兺 does not suffice to determine, via (7), the values of the field at
36 J. Earman
Suppose then we are given a three manifold 兺 and initial data on 兺 for
the source-free gravitational field. Does this data fix an appropriately
unique solution to the source-free field equations (7)? The answer is yes, at
least locally: there exists a unique (up to diffeomorphism) maximal devel-
opment M, gab of the initial data for which 兺 is a Cauchy surface; that is, M,
gab cannot be extended as a solution to the source-free equations in any
way that keeps 兺 a Cauchy surface.11
Extending this local result, which gives no information about the size
of the maximal solution for which 兺 is a Cauchy surface, to a global result—
especially to the very strong global result that requires the unique maximal
development for which the initial value hypersurface is a Cauchy surface
to be maximal simpliciter—can run into various kinds of problems. First,
we might have made a poor choice of the initial value hypersurface; for
instance, we might have chosen 兺 to be the spacelike hyperboloid of
Minkowski spacetime pictured in figure 1.5. Obviously the maximal devel-
opment of the initial data, induced on 兺 by the Minkowski metric, for
which 兺 is a Cauchy surface—namely D(兺)—can be properly extended as a
solution of the source-free field equations. Second, and more interestingly,
there might be no good choice for the initial value surface, as in the case of
an inextendible spacetime, which is a solution of (7) and does not possesses
any global Cauchy surfaces. This disturbing possibility can arise for two rea-
sons. (a) The causal structure might be like that illustrated in figure 1.4. (b) A
singularity might develop in a finite time from the regular initial data, as is
indicated schematically in figure 1.6. Singularities can arise, of course, in
the Newtonian and special relativistic settings, but because of the existence
null
surfaces
D+ (Σ)
Figure 1.5
A bad choice of initial value hypersurface.
38 J. Earman
singularity
D+(Σ)
Σ
Figure 1.6
A naked singularity.
blackhole singularity
interior
ℑ+
ι0
horizon
collapsing
matter
r=0
(a)
singularity
ℑ+
horizon ι0
collapsing
matter
r=0
(b)
Figure 1.7
Weak and strong cosmic censorship in black hole formation.
thus shielded by the one-way causal membrane of the horizon from what-
ever nondeterministic effects emanate from the singularity.13 The strong
version of cosmic censorship conjectures that, generically, naked singulari-
ties do not develop from suitable matter fields, so that even the unfortu-
nate observers who fall into a black hole do not ‘see’ the singularity. The
strong and weak versions are illustrated respectively in figures 1.7a and
1.7b, which represent black hole formation in spherical gravitational
collapse.14 The weasel phrases ‘generically’ and ‘suitable matter fields’ are
essential to warding off potential counterexamples. For instance, the weak
version of cosmic censorship can be violated using collapsing dust matter.
40 J. Earman
But it seems wrong to lay the blame for the resulting naked singularity on
the process of gravitational collapse itself since, even in Minkowski space-
time singularities in dust matter fields can develop from regular initial data.
Thus, ‘suitable’ matter fields should be restricted to the ‘fundamental’
matter fields for which global existence and uniqueness properties hold in
Minkowski spacetime (see section 4). But this restriction is not by itself
enough to save weak cosmic censorship since special initial configurations
of the Klein–Gordon field (certainly a fundamental field) can lead to naked
singularities in gravitational collapse. Making precise the notion of non-
special or generic initial conditions would require either a suitable measure
on the space of solutions to Einstein’s equations (in which case, generic
except for a set of measure zero) or a suitable topology (in which case,
generic complement of a set whose closure has an empty interior). Since
it is not clear what the relevant measure or topology is, the cosmic censor-
ship conjecture is also unclear. Despite these difficulties however, a grow-
ing body of inductive evidence—for example, from the stability of black
holes—suggests that some interesting form of weak cosmic censorship is
probably true (Wald 1998). Opinion on the strong cosmic censorship
conjecture is more mixed, in part because it is harder to state in a form that
is both general and free of obvious counterexamples. Proving a general
form of either the weak or strong version seems to be beyond the capa-
bilities of present techniques for investigating the global existence and
uniqueness of the solutions to nonlinear pdes.
Supposing that cosmic censorship fails, it remains unclear how the for-
tunes of determinism are affected. Classical GTR places no restrictions on
what nondeterministic influences can emerge from the naked singularities.
But this does not mean that anything goes. If we are optimistic, we can
suppose that even at the prequantum level there are additional lawlike
regularities, beyond those codified in Einstein GTR, that govern these
influences. And a quantum theory of gravity may add further restrictions
or, even better, get rid of the singularities altogether.
In closing this section I will note that while the fortunes of determinism
becomes more perilous in passing from special to general relativistic
physics, the fortunes of prediction perk up. For example, in various general
relativistic spacetimes it is possible to have nonempty domains of predic-
tion; indeed, it is possible for there to be points p 僆 M such that P(p) M
(see Hogarth 1993).
What We Have Learned and What We Still Don’t Know 41
ℑ+
Σ2
ι0
horizon
Σ1
r=0
Figure 1.8
Black hole evaporation.
8 Conclusion
One might have hoped that this survey would provide an answer to the
question: If we believe modern physics, is the world deterministic or not?
But there is no simple and clean answer. The theories of modern physics
paint many different and seemingly incommensurable pictures of the
world; not only is there no unified theory of physics, there is not even agree-
ment on the best route to getting one. And even within a particular theory—
say, QM or GTR—there is no clear verdict. This is a reflection of the fact that
determinism is bound up with some of the most important unresolved foun-
dations problems for these theories. While this linkage makes for frustration
if one is in search of a quick and neat answer to the above question, it also
makes determinism an exciting topic for the philosophy of science.
Acknowledgment
I am grateful to Rob Clifton, John Norton, and Laura Ruetsche for a num-
ber of helpful suggestions.
Notes
1. There is, of course, a violation of conservation of energy and momentum in the over-
all process even though both energy and momentum are conserved in each collision.
2. This example also illustrates a violation of even the weakest version of Laplacian
determinism according to which the future state of the world is determined by its
entire past history.
4. Solution in the sense that only binary collisions occur and each such collision
obeys the laws of elastic impact.
5. This is a slight modification of the definition given in Geroch (1977). Other defin-
itions of prediction for general relativistic spacetimes are studied in Hogarth (1993).
6. This rule implies that an observable with a pure continuous spectrum never takes
on a definite value. This awkwardness can be overcome by talking about approxi-
mate eigenstates.
44 J. Earman
9. This is a version of the notorious ‘hole argument’ which led Einstein to abandon
his search for generally covariant field equations from 1913 until late 1915; see
Norton (1987).
10. In 1916 Einstein took the ‘observables’ of GTR (although he did not use this
term) to be spatiotemporal coincidences, such as the intersection of two light rays
(see Einstein 1916 and Howard 1999). In order to be adequate to the content of GTR,
Einstein’s notion of coincidence observables has to be extended to fields.
11. M, gab is said to be a (proper) extension of M, gab iff the latter can be isometri-
cally embedded as a (proper) subset of the former. For the details of the initial value
problem in GTR, see Wald (1984).
12. For a more detailed discussion of spacetime singularities, as well as the cosmic
censorship hypothesis (introduced below), see my (1995).
13. Being visible is not sufficient to make a singularity ‘naked’ in the relevant sense;
for example, the initial big bang singularity in Friedmann-Robertson-Walker space-
times is not ‘naked’ because all of these spacetimes possess Cauchy surfaces.
14. Figures 1.7 and 1.8 use the conventions of conformal diagrams, which preserve
causal relations but distort metrical relations in order to bring infinities in to a finite
distance; namely, null directions lie at 45° with respect to the vertical; ℑ stands for
future null infinity, the terminus of outgoing light rays; and o stands for spacelike
infinity. The center of the spherical symmetry is labeled by r 0.
References
Arageorgis, A., J., Earman, and L. Ruetsche. 2002. “Weyling the Time Away: The
Aon-unitary Implementability of Quantum Field Dynamics on Curved Spacetime.”
Studies in History and Philosophy of Modern Physics 33: 151–184.
Belot, G. and J. Earman. 1997. “Chaos Out of Order: Quantum Mechanics, the Cor-
respondence Principle and Chaos.” Studies in History and Philosophy of Modern Physics
28: 147–182.
What We Have Learned and What We Still Don’t Know 45
Belot, G., J. Earman, and L. Ruetsche. 1999. “The Hawking Information Loss
Paradox: The Anatomy of a Controversy.” British Journal for the Philosophy of Science
50: 189–229.
Dickson, M. 1998. “On the Plurality of Dynamics.” In J. Hellman and R. Healey, eds.
Quantum Measurement: Beyond Paradox. Minneapolis, Minn.: University of Minnesota
Press.
Earman, J. 1992. World Enough and Spacetime: Absolute vs. Relational Theories of Space
and Time. Cambridge, Mass.: MIT Press.
Earman, J. 1995. Bangs, Crunches, Whimpers, and Shrieks: Singularities and Acausalities
in Relativistic Spacetimes. Cambridge, Mass.: MIT Press.
Ghirardi, G. C., T. Rimini, and T. Weber. 1986. “Unified Dynamics for Microscopic
and Macroscopic Systems.” Physical Review D 34: 470–491.
Horowitz, G. T., and R. Myers. 1995. “The Value of Singularities.” General Relativity
and Gravitation 27: 915–919.
Howard, D. 1999. “Point Coincidences and Pointer Coincidences: Einstein and the
Invariant Content of Space-Time Theories.” In H. Goenner, J. Renn, J. Ritter, and
T. Sauer, eds. The Expanding Worlds of General Relativity. Einstein Studies, volume 7.
Boston, Mass.: Birkhäuser.
Johnson, C. V., A. Peet, and J. Polchinski. 2000. “Gauge Theory and the Excision of
Repulsion Singularities.” Physical Review D 61: 086001.
Norton, J. D. 1987. “Einstein, the Hole Argument and the Reality of Space.” In
J. Forge, ed. Measurement, Realism, and Objectivity. Dordrecht: D. Reidel.
Saunders, S. 1999. “The `Beables’ of Relativistic Pilot Wave Theory.” In J. Butterfield and
C. Pagonis, eds., From Physics to Philosophy. Cambridge: Cambridge University Press.
Wald, R. M. 1994. Quantum Field Theory in Curved Spacetime and Black Hole Thermody-
namics. Chicago, Ill.: University of Chicago Press.
Keith Lehrer
It is the power of preference that leads us to think and aver that things are
up to us or in our power. Our preferences are the source of our freedom. But
we must have our preferences because we prefer to have them and not
because they are imposed on us. We have reasons for our preferences. Our
reasons for our preferences must be our reasons because we prefer them to
be. We must be agent and author of our preferences for them to be free. Our
preferences, however, seem to be part of the causal order. So how can we
have our preferences because we prefer to have them? How can things
be reasons for our preferences because we prefer that they be? How can we
be the agent and author of our preferences, when our preferences are part
of the causal order? How, on the contrary, can we have preferences because
we prefer to have them if they are not part of the causal order? When we
reflect on how preferences empower us, we understand that we must em-
power them. But how? These are the questions I shall seek to answer. The
keystone to the answer is a kind of special preference that loops back onto
our preferences including itself.
There is a minimal notion of freedom expressed by ‘can’ and ‘could
have’. It goes this way. I do something freely if and only if I could have
done otherwise. Extending the minimal notion of freedom to preference, I
prefer something freely if and only if I could have preferred otherwise. The
appropriate meaning of ‘could have’ required to articulate our freedom
cannot be assumed to be contained in antecedent usage, however. On the
contrary, we must construct the requisite meaning needed for an adequate
theory of freedom.
Let me return to a simple analysis, recognizing that this is theory
construction, not semantics. We are trying to construct, then, an analysis
of ‘can’ and ‘could have’ that will enable us to build a theory of freedom.
48 K. Lehrer
is one that I have, not because it is manipulated by another, but just because
I prefer to have it. Let us call the preferences we have at various levels con-
cerning doing X my preferences structure for X. Now consider my preference
to have my preference structure for X. Let us call that my power preference
for X. I propose as one condition of freedom of preference that I prefer X
and I have my preference structure for X because I prefer to have it.
There are special features of the power preference and of my having that
preference because I prefer to have it. First, my power preference for having
my preference structure for X is itself a preference in that preference struc-
ture because it is itself a preference concerning X. Thus, it is a preference for
having a preference structure that includes that preference. The preference
refers to itself, as it is a preference for itself as well as other preferences in
the structure. It is, therefore, an unstratified or level-ambiguous preference.
On one standard account of level-ambiguous states, they are semantically
ungrounded (Kripke 1975). Ungroundedness may be considered as a defect
in the truth conditions for a sentence, but it is not a defect of a state
intended to analyze freedom of preference.
Second, the preference ends a possible regress of preferences with a loop.
Theories of higher order preference that appear to require a complete hier-
archy of preferences are often criticized on the ground that the require-
ment is psychologically unrealistic. We have noted that if preferences over
preferences are considered computationally or functionally, it is not clear
that any difficulty in principle results from such requirements. However,
the addition of preferences past the second or third level, though it can be
processed computationally, might lack efficacy. So, there is an advantage in
not requiring a hierarchy of preferences, and use of the power preference
allows us to dispense with that requirement. The power preference ensures
that any preference that one has at any level is one that one prefers to have,
for that is what the power preference is a preference for. It allows, in prin-
ciple, that freedom of preference could result from a first-order preference
for X and the unstratified power preference for X that would, in that case,
be a preference for the first-order preference and for itself.
Moreover, the present account allows for the possibility that preferences
that insure freedom might contain conflict between levels of preference.
Frankfurt’s account and my own earlier account indebted to him require
that the preferences insuring freedom be homogeneous in the sense that
every preference in the structure be a preference for the preference of the
52 K. Lehrer
and preference in terms of the power preference and the primacy condi-
tion, even if such a condition should prove not to be necessary.
Here, then, is a sufficient condition. Let us begin with freedom of prefer-
ence. Suppose I prefer to do something, to write this paper, for example. We
are seeking a sufficient condition for this freedom without worrying about
whether it is necessary. If we add some bells and whistles that are not really
needed for freedom of preference, that will not matter. So let us help our-
selves to a set of conditions that are clearly sufficient, that clearly insure
freedom of preference. Here are the conditions I propose for our considera-
tion. My preference for doing X is free if
1. I have a power preference for X, that is, I would prefer to have the
preference structure for X that I have.
2. The power preference for X satisfies the primacy condition, that is,
I would have the preference structure for X because I would prefer to
have it.
These conditions, though they insure that my preference is not manipu-
lated by another might not provide adequate insurance that I could have
preferred otherwise. So let us add a third condition to ensure that.
but also that he would have it. This treats the very influential Frankfurt
(1969) counterexamples to the principle of alternative possibilities by re-
quiring alternative possibilities for freedom of preference. In these coun-
terexamples, though a person does what he prefers, namely, to do X, and
prefers to prefer to do X and so forth, there is an inactive counterfactual
intervener with knowledge of what the person is about to prefer and the
power, which he would have exercised if necessary, to make the person pre-
fer to do X if the person were about to not prefer to do X. Such a person as
described in the Frankfurt examples who prefers to do X would not have
the preference to do X because he preferred to have that preference when
he was about to not prefer to do X and the intervener caused him to have
the preference to do X. The person would have the preference because the
counterfactual intervener would intervene.
So we have a sufficient condition for freedom of preference when what-
ever preference structure I have or would have concerning doing X is one
I would have because I prefer to have it. Suppose that I do something
because I freely prefer to do it and I would have done otherwise if I had
preferred otherwise which I would also have preferred freely. That is a
sufficient condition for freedom of action. I could have done otherwise. So
from the three conditions formulated above we obtain an account of
freedom of preference and freedom of action.
Are the three conditions offered above compatible with determinism?
They are. The reason is related to the remarks about the primacy of causation
and explanation given above. You may cause me to prefer something with-
out that destroying my freedom of preference if I prefer what you prefer
because I prefer to have that preference structure and my preference for that
preference structure is the primary cause or explanation. Similarly, some
other reason or motive may cause me to prefer something without that
destroying my freedom of preference. Suppose I prefer to respond to that
reason or motive because I prefer to have a preference structure responding
to such reasons or motives. Moreover, suppose my preference for that
preference structure is the primary cause or explanation of it. That ensures
my freedom. Furthermore, the assumption of the primacy of preference in
causation and explanation is compatible with causation and explanation.
These remarks introduce a further complication, and a necessary one in
our theory of what is sufficient for freedom. The complication concerns the
influence of reasons or motives. Any compatibilist account must include a
56 K. Lehrer
The simple answer is that it is caused. I am not sure how. However, I think
that the rough picture is that we at a certain stage have no preferences.
Only impulses and desires, perhaps informed by unreflective beliefs and
impressions, that drive our behavior. We form patterns of action, patterns
of responding to desires and beliefs, at the level of the first-order mind un-
encumbered by the burden of ratiocination and higher-order evaluation.
Then we acquire the higher-order mind, the metamind, that allows us to
represent and think about our desires and beliefs, to prefer the satisfaction
of some desires and not others, to accept some beliefs but not others, as we
sort out the first-order states. At this point, at the point of the intervention
of the metamind, we go through a brief, somewhat confused period, of vac-
illation between mind and metamind, and, finally, we arrive at a state of
preference and acceptance, of systems of reasons supporting preferences
and preferences supporting reasons. That is my conjecture concerning the
transition from mental causation to metamental causation and the free-
dom of preference. It is not more than a conjecture.
The argument from what I have said to the conclusion of compatibilism
is brief. Combine the account of freedom of preference in terms of
the three conditions concerning the power preference stated earlier with
the account of the structure of a system of reasons and the ultrapreference
containing a power preference satisfying those conditions, and you have
sufficient conditions for freedom of preference. A sufficient condition of
freedom of action is easily obtained therefrom. Suppose I do something
because I prefer to do it when my preference is free. Suppose in addition, I
would have done otherwise if I had preferred to and that preference would
have been free. On those suppositions, my action was free. Suppose my ac-
tion and preference satisfy these sufficient conditions for freedom. Is there
anything in these conditions that entails that these structures of preference
that suffice for my freedom are uncaused? Nothing at all. Is there anything
in this account which entails that something must be undetermined, that
there must be some point at which there is no causal explanation for why
I prefer one thing and not another? There need be no such point.
If I were completely rational in my preferences, which I admit without
regret I am not, all my preferences might be responses to reasons that
causally support them. There might be causation all the way around, before
and after my preferences. But some of them would be free. When I prefer
what I do because I prefer to have the preference structure I do, including
Freedom and the Power of Preference 61
because I prefer to have it. I have a sense of agency when I prefer what I
do because I prefer to have that preference and when the effective reasons
for the preference, if any, are effective because I prefer that they be effective.
I am in this view close to the view that Frankfurt (1971) advances when he
emphasized the importance of identifying with a preference for a sense of
agency, but I suggest that my identification with the preference is the result
of having that preference because I prefer to have it and responding to my
reasons for having the preference because I prefer to respond to them.
There is also the issue of control and power. If I am free in my prefer-
ences, then they must be in my control, in my power. Suppose, then, that
my preferences and the system of reasons that supports them are ones
I have because I prefer to have them, and, moreover, are such that if I pre-
ferred not to have them, I would have that preference because I preferred
to have it. Are my preferences in my control? I control them all right.
I have them because I prefer to have them. If I preferred not to have them,
I would have that preference because I preferred to have it. What I prefer is
in my control in the sense that my preferences, whatever they may be, are
ones that I have because I prefer to have them. I am in control. Do I have
power over my preferences? They are in my power. I have the preferences
I do because I prefer to have them. To avoid tedium, I proceed no further
in this line of argument.
Is the theory a theory of volitions? Preferences are like volitions. One prob-
lem with trying to analyze freedom in terms of volitions is that the theory
appears to lead to an infinite regress, when we try to analyze the freedom of
volitions themselves, or, instead, a kind of surd when a philosopher claims
that volitions make us free without need to explain what makes the voliti-
ons free. But the power preference avoids the regress and the surd. What
makes the preference free is that we have the preference structure we do
because we prefer to have it. What makes that preference, the power prefer-
ence, free, is that we have it because we prefer to have it as a preference
included in the preference structure. The power preference, since it is a pref-
erence we have because we prefer to have it, ties freedom in, down, and
together in an explanatory loop.
Finally, the theory avoids a kind of paradox of libertarianism. The para-
dox is that the libertarian insists on the importance of indeterminism and
then faces the objection that the mere lack of causal determination does
not seem to contribute in any positive way to the account of freedom.
Freedom and the Power of Preference 63
of nature, which are part of their definition of laws, then we have no choice
about the outcome of the past conditions and the laws of nature. Or to put
it another way, there is nothing that we can do to render true claims about
the past false or render true claims about the laws false, and, yet, if we can
do otherwise than we do, then either the true claims about the past would
be false or true claims about the laws would be false. Therefore, if we can do
otherwise, we can render either the claims about the past false or the claims
about the laws false. But we cannot do these things. So, if determinism is
true, then we cannot do otherwise than what we do. Such arguments are
very persuasive. How should we reply?
My first reply is that the word ‘can’ is highly ambiguous in a way that
makes it very difficult to argue in a cogent manner about the issue of com-
patibilism formulated in terms of ‘can’. As Moore (1912) noted, there is
clearly a sense of ‘can’ that makes the argument offered sound. I have no
doubt that there is a version of the argument that van Inwagen offers using
‘can’ as he intends so that he can meet the objections of his critics and
argue that if determinism is true, then no one could have done otherwise—
in his use of ‘can’. Similarly, there are other uses, meanings, of ‘can’, some
of which I have proposed, from which compatibilism follows. Examples
include conditional analyses and variations thereof in terms of accessibility
relations to possible worlds, analyzing accessibility in ways that favor com-
patibilism. Formulating the matter in terms of ‘can’ is, I ruefully conclude,
hazardous, and the question is easily begged.
Nevertheless, I do think that the conditions that I have offered are suffi-
cient conditions, not only for freedom, but also for saying that a person
could have done otherwise. So how do we avoid the argument that if de-
terminism is true, then no one could ever have done or preferred to do any-
thing other than what they did or preferred to do? There is a principle that
figures into the argument in one form or another which says that if a
person can do something, and something is a necessary condition for his
doing it (like the falsity of either the past or the laws of nature), then, if the
person can do the thing, then the person can bring about the necessary
conditions for doing it as well. This is plausible. It says that if something is
necessary to doing something, then if you can do the thing in question,
you can bring about the necessary condition. Plausible, but false. One
necessary condition of my writing this paper is the birth of my parents, but
there is nothing I can do to bring about the birth of my parents. Similar
Freedom and the Power of Preference 65
criticisms hold for the general principle that says that if some condition is
necessary for a person doing something, then if the person can do the
thing, the person can bring about the condition.
Now any incompatibilist worth his or her arguments will jump up and
down and reply that it is only conditions that do not exist and which are
necessary for doing something that a person must be able to bring about if
they can perform the action. Again, however, this is false. If the conditions
simply will occur if the person performs the action, that is all that is required.
Now when we turn to what could have been, what a person could have
done, we must say that any condition that is necessary for the person to
have done the action is one that would have occurred. So, as Lewis (1981)
noted, when we say that a person could have done something that she did
not do, then, if determinism is true, we must conclude that if the person
had done what she did not do, then the necessary condition of her doing
it, namely, the falsification of the statements about the past or statements
of laws would have occurred. But we cannot conclude that the person
doing otherwise would have brought about the occurrence of that condi-
tion. That would be a mistake. All we can conclude is that one of those
statements would have been false.
Now let us look at the power preference. Suppose I had preferred other-
wise. Moreover, suppose that I had preferred otherwise because I preferred
to have a preference structure containing that preference. Then a necessary
condition for my preferring otherwise is that I not have preferred what I
actually did and that all necessary conditions for that were satisfied. That
includes that either the laws concerning my preference or the past would
have been otherwise. I see no problem with this. It does not follow from
the fact that I could have preferred otherwise that I could have thereby
brought about the necessary conditions for so preferring, and, more specif-
ically, it does not follow that I could have brought it about that the past or
laws would have been otherwise. All that follows is that they would have
been otherwise if I had preferred otherwise.
You might inquire further what would have been otherwise—the past or
the laws? Suppose it is the laws of nature that would have been otherwise.
What that means, of course, is not that something would have been both a
law and false but, rather, that something that was law would not have been
law. Is that strange? Some, van Inwagen (1975) and Ginet (1966), have
suggested that whether something is a law of nature does not depend on
66 K. Lehrer
our choices or preferences, though van Inwagen (1983) has noted the pos-
sibility of voluntaristic laws. Most laws do not concern human choices and
preferences. Such laws, the laws of chemistry, for example, do not depend
on our choices and preferences. However, they are not laws concerning our
choices and preferences, laws of the form, Lc, to the effect that under con-
ditions C a person S will choose or prefer X, for example. Now consider laws
like Lc, supposing there are some. What is the point of saying that they do
not depend on human choices and preferences? They concern our choices
and preferences, after all, and say that under conditions C people choose or
prefer X. Such laws of the form of Lc clearly do depend on our choices and
preferences. If we do not choose or prefer X under conditions C, then Lc
would not be a law. So, there is nothing strange in the idea that if I had
chosen or preferred other than X under C, then Lc would not have been a
law, though, in fact, it is a law.
Now we come to the critical objection, however. If Lc is a law, shouldn’t
we conclude from this that no one could have not chosen or preferred X
when they were in conditions C? No doubt that there is a sense of ‘could
have’ that would warrant this conclusion, but, as I have noted before, we
need the additional premise that when a person could not have chosen
otherwise in this sense, then the person could not have chosen otherwise
in the sense required for the person to be free to have chosen otherwise.
And it is precisely this step I wish to question. Notice, that a law statement
must be true, and there is a sense of ‘could have’ in which it seems obvious
that no one can make what is true false and, thus, that no one could have
made what is a true law statement false. But that sense of ‘could have’ does
not deprive us of freedom because it rests on the assumption that no one
can make something be both true and false, which is a tautology. That does
not preclude us from saying what is required for freedom, namely, that
somebody could have done something that would have made something
false instead of being true, though, in fact, she did not, and it is true that
she did not.
Philosophers have thought, however, that there is a more important
modal notion of ‘can’ connected with laws, a more significant sense of ‘can’
and ‘could have’, that enables us to distinguish laws from mere accidental
general truths. It is, for example, an accidental general truth that everyone
in this room is older than four years old. The accidental character of the gen-
eral truth might be captured by a modality, by saying that someone could
Freedom and the Power of Preference 67
have falsified this general truth by bringing her four-year-old to hear about
freedom and the power of preference, but no one could have done anything
to falsify any law of nature. I have two replies with which I shall conclude.
First, it may be true of some but not all laws of nature, namely those con-
cerned with matters other than human choice and preference, that they
cannot be falsified by human choice or preference, but it does not follow
from this that no law about human choice or preference can be falsified by
human choice or preference.
Second, and more crucially, there is another and, I believe, better way of
distinguishing between laws and accidentally true generalizations than by
appeal to what can and could not be. Laws warrant counterfactual infer-
ence of a kind that is not warranted by accidental generalizations. The
specification of the exact kind of counterfactual inference is a delicate
matter requiring some subtle distinctions, but the general point is clear
enough. The accidentally true generalization that all people in this room
are older than four years does not warrant the counterfactual inference to
the conclusion that if a four-year-old had been brought in to hear about
freedom and the power preference she would have been older than four
years old. Laws by contrast do warrant counterfactual conclusions about
what would have been. So, if Lc is a law and I, in fact, do not choose or pre-
fer X, someone would be warranted in inferring that if I had been in con-
ditions C, which I was not, then I would have chosen or preferred X. The
person would not, however, be warranted in inferring that I could not have
chosen or preferred otherwise than X when I am in C. They could only
infer that I would not.
I conclude with the remark that this conception of a law is one that dif-
fers from what incompatibilists such as Ginet and van Inwagen have as-
sumed about laws in their arguments for incompatibilism. Am I, therefore,
just missing the point of their argument? One motivation for concern
about the problem of freedom and determinism is that it appears that a sci-
entific conception of the world based on the laws of nature and explana-
tion is incompatible with the sort of freedom we believe ourselves to have.
I have intended to allay this concern. The scientific conception of the
world is, I suggest, one including statements of laws that warrant inference
and counterfactual inference about the world. Some philosophers may go
beyond this conception to one affirming a thesis of laws that contain a
kind of necessity that is incompatible with human freedom. If one builds
68 K. Lehrer
Notes
2. Peter van Inwagen and Krister Segerberg both communicated the objection to me
in correspondence. Cf. Peter van Inwagen (1983), chap. 4.
4. I was not able to pin down this interpretation to a specific passage, for his prose
has a charmingly diffuse character, but I am indebted to Sartre (1956).
5. See chap. 4.
7. Originally argued in van Inwagen (1975), and more fully in van Inwagen (1983).
References
Chisholm, R. 1966. “Freedom and Action.” In K. Lehrer, ed., Freedom and Determin-
ism. New York: Random House.
Ekstrom, L. 2000. Free Will: A Philosophical Study. Boulder, Colo.: Westview Press.
Ginet, C. 1966. “Might We Have No Choice?” In K. Lehrer, ed., Freedom and Deter-
minism. New York: Random House.
Freedom and the Power of Preference 69
Kant, I. 1993. Critique of Practical Reason. L. W. Beck, trans. New York: Macmillan.
Lewis, D. 1981. “Are We Free to Break the Laws?” Theoria 47: 113–121.
Sartre, J. 1956. Being and Nothingness. H. Barnes, trans. New York: Philosophical
Library.
Velleman, J. D. 1992. “What Happens When Someone Acts.” Mind 101: 461–481.
3 Agency, Responsibility, and Indeterminism: Reflections
on Libertarian Theories of Free Will
Robert Kane
past philosophical debate about this question has focused on the question
of whether determinism is compatible with ‘the condition of alternative
possibilities’ (which I call AP)—the requirement that the free agent ‘could
have done otherwise’. As we all know, most arguments for and against in-
compatibilism revolve around AP in one way or another. I don’t think these
debates are unimportant. Far from it. I have engaged in them myself and
they continue. But I think that exclusive focus on AP and ‘could have done
otherwise’ has led philosophers to ignore other important facets of the
debate about Compatibility and Significance questions. AP alone, I argue,
provides too thin a basis on which to rest the case for incompatibilism.
Debates over different interpretations of ‘can’, ‘power’, ‘ability’, and ‘could
have done otherwise’ are symptoms of this fact.
To see what I mean by saying AP provides too thin a basis for resolving
the Compatibility question, consider certain kinds of acts I will call Austin-
style examples (after J. L. Austin who, among others, introduced examples
of these kinds into free will debates in Austin 1956). Austin’s own example
of this type concerned a three foot putt he was trying to make in a golf
match, but missed by chance. Another example of this kind I have used is
this: suppose I approach a coffee machine intending to press the button for
black coffee, but have a brain cross as I stand before the machine and press
the button for coffee with cream instead by mistake. Imagine in both these
cases, Austin’s missing the putt and my pressing the wrong button, some
genuine indeterminacy was involved in the efferent neural pathways of his
arm and the impulses of my brain such that they were genuinely undeter-
mined. They were still actions we performed, as Austin notes of examples
of this type, but unintentionally, nonvoluntarily, by accident or mistake. In
each case, we could have done otherwise (made the putt and pressed the
right button since the outcomes were undetermined). So in each case we
have actions that are (i) such that we could have done otherwise (AP) and
(ii) that are undetermined. Yet neither is a free action in any normal sense
of the word since neither was under our control.
Now imagine a world full of Austin-style examples of this type; a world
in which indeterminism is prevalent and agents can often do otherwise in
an undetermined way in the Austin style—by accident, mistake, uninten-
tionally or inadvertently. But imagine also that in this undetermined
world, God has set all agents’ wills (their reasons, motives, intentions
and purposes) so that they desire and intend to do only what God has
72 R. Kane
predetermined. Austin might miss his putt, but his wanting and trying to
make it rather than miss was preset by God. I might press the wrong button
on the coffee machine but my wanting and intending to press the right
button was predetermined by God. And so it is with all agents for all their
acts in this world. I would argue that this is a world in which agents lack free-
dom of will, even though they are often capable of doing otherwise (have
AP) and many of their acts are undetermined. They are not the sources
of their own wills. All the ‘will-setting’ in this world is done by God and
none by the agents themselves.
This shows that being able to do otherwise (AP) even if the doing other-
wise is undetermined is not sufficient for free will (even if these two condi-
tions of AP plus indeterminism should be necessary). This is one way of
showing why AP is too thin a basis on which to rest the case for incompat-
ibilism. The agents must be able to will otherwise as well as do otherwise.
What we want for free will is not merely to be able to do otherwise but to
be able to do otherwise (or more generally, to act either way) voluntarily,
intentionally, and rationally, rather than one way voluntarily, intention-
ally, and rationally and the other only unintentionally, involuntarily, and
by mistake or accident, as in the Austin-style examples. I call these require-
ments of more-than-one-way voluntariness, intentionality, and rationality
the ‘plurality conditions’ for free will. By satisfying these conditions in a
manner that was undetermined, agents would be engaged in their own
‘will-setting’; setting their wills (intentions and purposes) would not all be
done by God or nature, society, or something else over which they had no
control. AP is not unimportant for free will but it is important because of
these strong requirements of ‘plurality’ and ‘will-setting’ for full-fledged
free will. I have therefore argued in my writings that more attention should
be given to plurality conditions and will-setting in discussions of Compat-
ibility and Significance questions rather than to AP alone.
But plurality conditions and will-setting are not enough. There is yet a
third condition that has to be considered besides AP to get a fully adequate
picture of the Compatibility debate. I have also argued that in the long
history of the free will debate one can find another criterion fueling incom-
patibilist intuitions that is even more important than AP, though compar-
atively neglected. I call it the condition of ultimate responsibility or UR. The
basic idea is this: to be ultimately responsible for an action, an agent must be
responsible for anything that is a sufficient reason (condition, cause, or
Agency, Responsibility, and Indeterminism 73
motive) for the action’s occurring.2 If, for example, a choice issues from, and
can be sufficiently explained by, an agent’s character and motives (together
with background conditions), then to be ultimately responsible for the
choice, the agent must be at least in part responsible by virtue of choices or
actions voluntarily performed in the past for having the character and mo-
tives he or she has now. Compare Aristotle’s claim that if a man is responsi-
ble for wicked acts that flow from his character, he must at some time in the
past have been responsible for forming the wicked character from which
these acts flow (Aristotle 1985).
This UR condition accounts for the ‘ultimate’ in the original definition of
free will: “the power of agents to be the ultimate creators and sustainers of
their own ends or purposes.” Now UR does not require that we could have
done otherwise (AP) for every act done of our own free wills—thus vindicat-
ing philosophers such as Frankfurt, Fischer, and Dennett, and others who
insist that we can be held morally responsible for many acts even when we
could not have done otherwise.3 But the vindication is only partial. For UR
does require that we could have done otherwise with respect to some acts in
our past life histories by which we formed our present characters. I call these
‘self-forming actions’, or SFAs. Consider Dennett’s comments on Martin
Luther (1984, 133). When Luther finally broke with the Church in Rome, he
said “Here I stand, I can do no other.” Suppose, says Dennett, at that mo-
ment Luther was literally right. Given his character and motives, he could
not then have done otherwise. Does this mean he was not morally respon-
sible for his act, or not acting of his own free will? Dennett says “Not at all.”
In saying “I can do not other,” Luther was not disowning responsibility, but
taking full responsibility for acting of his own free will. So ‘could have done
otherwise’, or AP, is not required for moral responsibility or free will.
My response is to grant that Luther could have been responsible for this
act, even ultimately responsible in the sense of UR, though he could not
then have done otherwise and even if his act was determined. But this
would be so to the extent that he was responsible for his present motives
and character by virtue of earlier struggles and self-forming choices (SFAs)
that brought him to this point in his life where he could do no other. Often
we act from a will already formed, but it is ‘our own free will’ by virtue of
the fact that we formed it by other choices or actions in the past (SFAs) for
which we could have done otherwise. If this were not so, there is nothing
we could have ever done to make ourselves different than we are—a
74 R. Kane
omitted either was, or causally contributed to, E’s occurrence and made a
difference to whether or not E occurred; and (U) for every X and Y (where
X and Y represent occurrences of events and/or states) if the agent is
personally responsible for X, and if Y is a sufficient reason (condition,
cause, or motive) for X, then the agent must also be personally
responsible for Y.
are faced with difficult tasks to which we have aversions. In all such cases,
we are faced with competing motivations and have to make an effort to
overcome temptation to do something else we also strongly want. There is
tension and uncertainty in our minds about what to do at such times, I sug-
gest, that is reflected in appropriate regions of our brains by movement
away from thermodynamic equilibrium—in short, a kind of ‘stirring up of
chaos’ in the brain that makes it sensitive to microindeterminacies at the
neuronal level. The uncertainty and inner tension we feel at such soul-
searching moments of self-formation is thus reflected in the indeterminacy
of our neural processes themselves. What is experienced internally as
uncertainty then corresponds physically to the opening of a window of
opportunity that temporarily screens off complete determination by influ-
ences of the past. (By contrast, when we act from predominant motives or
settled dispositions, the uncertainty or indeterminacy is muted. If it did
play a role in such cases, it would be a mere nuisance or fluke, as critics of
indeterminism contend.)
When we do decide under such conditions of uncertainty, the outcome
is not determined because of the preceding indeterminacy—and yet it can
be willed (and hence rational and voluntary) either way owing to the fact
that in such self-formation, the agents’ prior wills are divided by conflict-
ing motives. Consider a businesswoman who faces such a conflict. She is
on her way to an important meeting when she observes an assault taking
place in an alley. An inner struggle ensues between her conscience, to stop
and call for help, and her career ambitions that tell her she cannot miss this
meeting. She has to make an effort of will to overcome the temptation to
go on. If she overcomes this temptation, it will be the result of her effort,
but if she fails, it will be because she did not allow her effort to succeed.
And this is due to the fact that, while she willed to overcome temptation,
she also willed to fail, for quite different and incommensurable reasons.
When we, like the woman, decide in such circumstances, and the indeter-
minate efforts we are making become determinate choices, we make one
set of competing reasons or motives prevail over the others then and there
by deciding.
Now let us add a further piece to the puzzle. Just as indeterminism need
not undermine rationality and voluntariness, so indeterminism in and of
itself need not undermine control and responsibility. Suppose you are try-
ing to think through a difficult problem, say a mathematical problem, and
Agency, Responsibility, and Indeterminism 79
which of a set of outcomes is going to occur before it occurs, that one does
not control or determine which of them occurs, when it occurs. When the
above conditions for SFAs are satisfied, agents exercise control over their
future lives then and there by deciding.
Indeed, they have what I call ‘plural voluntary control’ over the options
in the following sense: they are able to bring about whichever of the op-
tions they will, when they will to do so, for the reasons they will to do so,
on purpose rather than accidentally or by mistake, without being coerced
or compelled in doing so or willing to do so, or otherwise controlled in
doing or willing to do so by any other agents or mechanisms. I show again
in Kane (1996) that each of these conditions can be satisfied for SFAs as
conceived above.6 The conditions can be summed up by saying, as we
sometimes do, that the agents can choose either way, at will.
Note also that this account of self-forming choices amounts to a kind of
‘doubling’ of the mathematical problem. It is as if an agent faced with such
a choice is trying or making an effort to solve two cognitive problems at
once, or to complete two competing (deliberative) tasks at once—in our
example, to make a moral choice and to make a conflicting self-interested
choice (corresponding to the two competing neural networks involved).
Each task is being thwarted by the indeterminism coming from the other,
so it might fail. But if it succeeds, then the agents can be held responsible
because, as in the case of solving the mathematical problem, they will have
succeeded in doing what they were knowingly and willingly trying to do.
Recall the assassin and the husband. Owing to indeterminacies in their
neural pathways, the assassin might miss his target or the husband fail to
break the table. But if they succeed, despite the probability of failure, they
are responsible, because they will have succeeded in doing what they were
trying to do.
And so it is, I suggest, with self-forming choices, except that in the case of
self-forming choices, whichever way the agents choose they will have suc-
ceeded in doing what they were trying to do because they were simultane-
ously trying to make both choices, and one is going to succeed. Their failure
to do one thing is not a mere failure, but a voluntary succeeding in doing
the other. Does it make sense to talk about the agent’s trying to do two
competing things at once in this way, or to solve two cognitive problems at
once? Well, we now know that the brain is a parallel processor; it can simul-
taneously process different kinds of information relevant to tasks such as
82 R. Kane
You may find this interesting and yet still find it hard to shake the intuition
that if choices are undetermined, they must happen merely by chance—
and so must be ‘random’, ‘capricious’, ‘uncontrolled’, ‘irrational’, and all
the other things usually charged. Such intuitions are deeply ingrained. But
if we are ever going to understand free will, I think we will have to break old
habits of thought that support such intuitions and learn to think in new
ways. The first step is to question the intuitive connection in most people’s
minds between ‘indeterminism’s being involved in something’ and ‘its
happening merely as a matter of chance or luck’. ‘Chance’ and ‘luck’ are
terms of ordinary language that carry the connotation of ‘its being out of
my control’. So using them already begs certain questions, whereas ‘inde-
terminism’ is a technical term that merely precludes deterministic causa-
tion, though not causation altogether. Indeterminism is consistent with
nondeterministic or probabilistic causation, where the outcome is not in-
evitable. It is therefore a mistake (alas, one of the most common in debates
about free will) to assume that ‘undetermined’ means ‘uncaused’.
Here is another source of misunderstanding. Since the outcome of the
businesswoman’s effort (the choice) is undetermined up to the last minute,
we may have the image of her first making an effort to overcome the
temptation to go on to her meeting and then at the last instant ‘chance
takes over’ and decides the issue for her. But this is misleading. On the view
I proposed, one cannot separate the indeterminism and the effort of will,
Agency, Responsibility, and Indeterminism 83
so that first the effort occurs followed by chance or luck (or vice versa). One
must think of the effort and the indeterminism as fused; the effort is inde-
terminate and the indeterminism is a property of the effort, not something
separate that occurs after or before the effort. The fact that the effort has
this property of being indeterminate does not make it any less the woman’s
effort. The complex recurrent neural network that realizes the effort in the
brain is circulating impulses in feedback loops and there is some indeter-
minacy in these circulating impulses. But the whole process is her effort of
will and it persists right up to the moment when the choice is made. There
is no point at which the effort stops and chance ‘takes over’. She chooses as
a result of the effort, even though she might have failed. Similarly, the hus-
band breaks the table as a result of his effort, even though he might have
failed because of the indeterminacy. (That is why his excuse, “chance broke
the table, not me” is so lame.)
Just as expressions like ‘she chose by chance’ can mislead in such con-
texts, so can expressions like ‘she got lucky’. Recall that, with the assassin
and husband, one might say ‘they got lucky’ in killing the prime minister
and breaking the table because their actions were undetermined. Yet they
were responsible. So ask yourself this question: why does the inference ‘he
got lucky, so he was not responsible’ fail in the cases of the husband and
the assassin where it does fail? The first part of an answer has to do with the
point made earlier that ‘luck’, like ‘chance’, has question-begging implica-
tions in ordinary language that are not necessarily implications of ‘indeter-
minism’ (which implies only the absence of deterministic causation). The
core meaning of ‘he got lucky’ in the assassin and husband cases, which is
implied by indeterminism, I suggest, is that ‘he succeeded despite the prob-
ability or chance of failure’; and this core meaning does not imply lack of
responsibility, if he succeeds.
If ‘he got lucky’ had other meanings in these cases that are often associ-
ated with ‘luck’ and ‘chance’ in ordinary usage (for example, the outcome
was not his doing, or occurred by mere chance, or he was not responsible
for it), the inference would not fail for the husband and assassin, as it
clearly does. But the point is that these further meanings of ‘luck’ and
‘chance’ do not follow from the mere presence of indeterminism. The
second reason why the inference ‘he got lucky, so he was not responsible’
fails for the assassin and the husband is that what they succeeded in doing
was what they were trying and wanting to do all along (kill the minister
84 R. Kane
and break the table respectively). The third reason is that when they suc-
ceeded, their reaction was not ‘oh dear, that was a mistake, an accident—
something that happened to me, not something I did’. Rather they endorsed
the outcomes as something they were trying and wanting to do all along,
that is to say, knowingly and purposefully, not by mistake or accident.
But these conditions are satisfied in the businesswoman’s case as well,
either way she chooses. If she succeeds in choosing to return to help the
victim (or in choosing to go on to her meeting) (i) she will have ‘succeeded
despite the probability or chance of failure’, (ii) she will have succeeded in
doing what she was trying and wanting to do all along (she wanted both
outcomes very much, but for different reasons, and was trying to make
those reasons prevail in both cases), and (iii) when she succeeded (in choos-
ing to return to help) her reaction was not ‘oh dear, that was a mistake, an
accident—something that happened to me, not something I did’. Rather
she endorsed the outcome as something she was trying and wanting to do
all along; she recognized it as her resolution of the conflict in her will. And
if she had chosen to go on to her meeting she would have endorsed that
outcome, recognizing it as her resolution of the conflict in her will.
Perhaps the problem is that we are begging the question by assuming
the outcomes of the woman’s efforts are choices to begin with, if they are
undetermined. One might argue that ‘if an event is undetermined, it must
be something that merely happens and cannot be somebody’s choice or
action’. But this assumption is clearly question begging in the extreme, as
one can see by considering that it entails ‘all choices and actions must de-
termined’ simply on the grounds that they are choices or actions. Well,
perhaps indeterminism does not undermine the idea that something is a
choice or action simply, but rather that it is the agent’s choice. But again,
why must that be so? What makes the woman’s choice her own on the
preceding account is that it results from her efforts and deliberation
which in turn are causally influenced by her reasons and her intentions
(for example, her intention to resolve indecision in one way or another).
And what makes these efforts, deliberation, reasons and intentions hers is
that they are embedded in a larger motivational system realized in her
brain in terms of which she defines herself as a practical reasoner and
actor.7 A choice is the agent’s when it is produced intentionally by efforts,
deliberation and reasons that are part of this self-defining motivational
system.
Agency, Responsibility, and Indeterminism 85
Well, then, perhaps the issue is not whether the undetermined SFA is a
choice, or even whether it is the agent’s choice, but rather how much con-
trol she had over it. It may be true, as I argued earlier (in the discussion of
plural voluntary control), that the presence of indeterminism need not
eliminate control altogether. But would not the presence of indeterminism
at least diminish the control persons have over their choices and other ac-
tions? Is it not the case that the assassin’s control over whether the prime
minister is killed (his ability to realize his purposes or what he is trying to
do) is lessened by the undetermined impulses in his arm—and so also for
the husband and his breaking the table? And this limitation seems to be
connected with another problem often noted by critics of libertarian
freedom—the problem that indeterminism, wherever it occurs, seems to be
a hindrance or obstacle to our realizing our purposes and hence an obstacle
to (rather than an enhancement of) our freedom.
There is something to these claims, but I think what is true in them
reveals something important about free will. We should concede that inde-
terminism, wherever it occurs, does diminish control over what we are
trying to do and is a hindrance or obstacle to the realization of our pur-
poses. But recall that in the case of the businesswoman (and SFAs gener-
ally), the indeterminism that is admittedly diminishing her control over
one thing she is trying to do (the moral act of helping the victim) is com-
ing from her own will—from her desire and effort to do the opposite (go to
her business meeting). And the indeterminism that is diminishing her con-
trol over the other thing she is trying to do (act selfishly and go to her
meeting) is coming from her desire and effort to do the opposite (to be a
moral person and act on moral reasons). So, in each case, the indetermin-
ism is functioning as a hindrance or obstacle to her realizing one of her
purposes—a hindrance or obstacle in the form of resistance within her will
that has to be overcome by effort.
If there were no such hindrance—if there were no resistance in her will—
she would indeed in a sense have “complete control” over one of her op-
tions. There would no competing motives that would stand in the way of
her choosing it. But then also she would not be free to rationally and vol-
untarily choose the other purpose because she would have no good com-
peting reasons to do so. Thus, by being a hindrance to the realization of
some of our purposes, indeterminism paradoxically opens up the genuine
possibility of pursuing other purposes—of choosing or doing otherwise in
86 R. Kane
accordance with, rather than against, our wills (voluntarily) and reasons
(rationally). To be genuinely self-forming agents (creators of ourselves)—to
have free will—there must at times in life be obstacles and hindrances in
our wills of this sort that we must overcome.
I think libertarians have traditionally tried to ignore this aspect of inde-
terminism. They knew it was required, but assumed it could be entirely cir-
cumvented by special agencies and would have no real impact on free will.
But hindrances and obstacles and resistance in the will are precisely what is
needed for free will, which, like life itself, exists near the edge of chaos. This
is an aspect of what is called in religious terms the ‘problem of evil’. Recall
Kant’s image of the bird that is upset by the resistance of the air and the
wind to its flight and so imagines that it could fly better if there were no air
at all to resist it (Kant 2003). But of course it would not fly better, but cease
to fly at all. So it is with indeterminism and free will. It provides resistance
to our choices, but a resistance that is necessary if we are to have free will.
I conclude with one final objection. Even if one granted that persons,
such as the businesswoman, could make genuine self-forming choices that
were undetermined, isn’t there something to the charge that such choices
would be arbitrary? A residual arbitrariness seems to remain in all self-
forming choices since the agents cannot in principle have sufficient or
overriding prior reasons for making one option and one set of reasons pre-
vail over the other. This is again one of those truths that tell us something
important about free will. In this case it tells us that every undetermined
self-forming free choice is the initiation of what I have elsewhere called a
‘value experiment’ whose justification lies in the future and is not fully ex-
plained by past reasons. In making such a choice we say, in effect, “Let’s try
this. It is not required by my past, but it is consistent with my past and is
one branching pathway my life can now meaningfully take. Whether it is
the right choice, only time will tell. Meanwhile, I am willing to take
responsibility for it one way or the other” (see Kane 1996, 145–146).
The term ‘arbitrary’ as I have often noted comes from the Latin arbitrium,
which means ‘judgment’—as in liberum arbitrium voluntatis, ‘free judgment
of the will’ (the medieval philosophers’ designation for free will). Imagine
a writer in the middle of a novel. The novel’s heroine faces a crisis and the
writer has not yet developed her character in sufficient detail to say exactly
how she will act. The author makes a ‘judgment’ about that which is not
yet determined by the heroine’s already formed past that does not give
Agency, Responsibility, and Indeterminism 87
unique direction. In this sense, the judgment (arbitrium) of how she will
react is ‘arbitrary’, but not entirely so. It had input from the heroine’s fic-
tional past and in turn gave input to her projected future. In a similar way,
agents who exercise free will are both authors of and characters in their
own stories all at once. By virtue of ‘self-forming’ judgments of the will
(arbitria voluntatis) (SFAs), they are ‘arbiters’ of their own lives, ‘making
themselves’ out of a past that, if they are truly free, does not limit their future
pathways to one. In reply to the charge that they did not have sufficient or
conclusive prior reasons for choosing as they did, they may respond: “True
enough. But I did have good reasons for choosing as I did, which I’m willing
to stand by and take responsibility for. If they were not sufficient or conclu-
sive reasons, that’s because, like the heroine of the novel, I was not a fully
formed person before I chose (and still am not, for that matter). Like the
author of the novel, I am in the process of writing an unfinished story and
forming an unfinished character who, in my case, is myself.”
Notes
2. For a formal statement and defense of this condition, see Kane (1996, ch. 3).
3. For defenses of this claim by these authors see the readings in Kane (2002) by
Dennett, Fischer, and Pereboom.
5. We have to make further assumptions about the case to rule out some of these
conditions. For example, we have to assume, no one is holding a gun to the woman’s
head forcing her to go back, or that she is not paralyzed, etc. But the point is that
none of these conditions is inconsistent with the case of the woman as we have
imagined it. If these other conditions are satisfied, as they can be, and the business-
woman’s case is otherwise as I have described it, we have an SFA. I offer the complete
argument for this in Kane (1996, ch. 8).
6. I show in greater detail that each of these conditions can be satisfied by SFAs in
Kane (1996, ch. 8).
and Owen Flanagan (1992). In Kane (1996, 137–142) I call the realization of such a
system in the brain the ‘self-network’.
References
Austin, J. L. 1956. “Ifs and Cans.” Proceedings of the British Academy 42: 109–132.
Churchland, P. M. 1995. The Engine of Reason, the Seat of the Soul. Cambridge, Mass.:
MIT Press.
Clarke, R. 1993. “Toward a Credible Agent-causal Account of Free Will.” Noûs 27:
191–203.
Dennett, Daniel D. 1984. Elbow Room: Varieties of Free Will Worth Wanting.
Cambridge, Mass.: MIT Press.
Kane, R. 1985. Free Will and Values. Albany, N.Y.: State University of New York Press.
. 1996. The Significance of Free Will. New York: Oxford University Press.
, ed. 2002. The Oxford Handbook of Free Will. Oxford and New York: Oxford
University Press.
Kant, I. 2003. Critique of Pure Reason. Tr. N. K. Smith. Basingstoke: Palgrave MacMillan.
O’Connor, T. 2000. Persons and Causes: The Metaphysics of Free Will. New York:
Oxford University Press.
Spitzer, M. 1999. The Mind within the Net. Cambridge, Mass.: MIT Press.
Velleman, D. 1992. “What Happens When Somebody Acts?” Mind 101: 461–481.
4 Trying to Act
Carl Ginet
1 Preliminaries
First, let me state some preliminary points, before I get to the main ques-
tion about trying to act that I want to talk about. These preliminary points
are not all uncontroversial, and I’ll give a bit of an argument for some of
them, but my purpose in stating them is not to persuade you of them, but
to let you know what I am assuming in the rest of my discussion.
1. I individuate actions in such a way that, for example, my action of turn-
ing on my computer by pressing a button on the keyboard is a causally
complex event that has, as a part, a distinct and more basic action of my
pressing the button (as well as a nonaction part caused by that more basic
action, namely, the computer’s turning on); and that more basic action has
as a proper part a still more basic action of my exerting a certain force with
my hand and index finger; and the latter action has as a proper part a most
90 C. Ginet
of their voluntarily moving their arm where they thought they might not
succeed; it’s perfectly natural for it to be intended and interpreted as quan-
tifying over all past cases of their voluntarily moving their arm. And when
it is so interpreted, we would have no difficulty in taking what the speaker
asserts to be true, even if we know that in a great many of the cases quanti-
fied over the speaker had no thought of the possibility of failing in their
attempt to move their arm.
6. The verb ‘try’ does not signify a special kind of action.3 In particular,
contrary to what some philosophers have claimed,4 it is not the case that
trying is always a most basic action. When we have the thought that S tried
to A, the action we think of as constituting her attempt to A—what she did
that was her trying to A—need not be the most basic action she performed.
This is shown by the acceptability of such an utterance as ‘The same
attempt to turn on the computer I made earlier—namely, my pushing
the “on” button on the keyboard—succeeded this time in turning it on’.
Here the speaker implies that what the attempt consisted in was an action
of pressing a certain button; and such an action is not most basic (because
it includes the more basic action of the agent’s moving his hand and finger
in a certain way).
Let me now turn to the question which it is my main purpose in this paper
to address. What is the mental component of trying to act? It seems evident
that an action can count as the agent’s trying to do a certain thing only if
the agent has an appropriate intention concerning that action. What
counts as an appropriate intention?
It’s fairly obvious that if S intended her action to be an attempt to do a
certain thing, then in that action S was trying to do that thing. That is, a
sufficient condition for the truth of ‘By Z-ing S tried to A’ is the following:
(B1) S believed of her Z-ing that by so acting she might A (that is, there
was a serious chance she would A).
This does not entail S1. This is shown by any case where S knowingly takes
a chance that her Z-ing might bring about certain consequences, but does
not intend by Z-ing to bring them about. For example, Hector might plug in
the new electric heater, knowing that there is a chance that by so doing he
might blow a fuse, but not with the intention of blowing a fuse. Such
examples show also that B1 does not entail that by Z-ing S tried to A. In plug-
ging in the new heater, Hector was not trying to blow a fuse. This nonen-
tailment supports the claim that S1 is a minimally sufficient condition for
the truth of ‘By Z-ing S tried to A’. At any rate, the weaker B1 is not sufficient.
But S1, though minimally sufficient, is not also necessary.5 Indeed, not
even the weaker B1 is necessary. Suppose Sam is confronted with a large log
that he would like to move. Suppose that Sam has no idea whether he can
move the log by just pushing on it but wants to find out. So he pushes on
the log as hard as he can. That is, he tries to move the log by pushing on it,
in order to find out whether he can move it by that means. He cannot
intend that his pushing will move the log, because he does not (yet) believe
that he might move it by pushing on it, that there is a chance that his push-
ing on it would succeed in moving it. For all he yet knows, it may be far too
heavy for him, or any human being, to move it just by pushing on it; and
it would be semantically unacceptable to say of him that, although he does
not know (and does not believe) that it is possible for him to move it by
pushing on it, his intention/aim/purpose in pushing on it is to move it.
So, for S’s Z-ing to count as S’s trying to A, though it is sufficient that S
intended of her Z-ing that by it she would A, it is not necessary. Some other
Trying to Act 93
One can start with the following fairly obvious thought: even if S did not
intend that by her Z-ing she would A, S must have intended that her Z-ing
have some sort of connection with the possibility of her A-ing. S must have
an intention whose content relates her Z-ing to the possibility of her
thereby A-ing in some way or other. The complications come when one asks
what specific content or contents for this intention will make an intention
such that, if S had that intention when Z-ing, then in Z-ing tried to A. There
are, it turns out, several different contents that will do the trick.
One we have already seen is our previous example where Sam’s pushing
on the log counted as his trying to move it, despite his not intending that
his pushing would move it. Though he did not intend to move it, he did in-
tend that his pushing on the log would show him whether he was able to
move it by that act of pushing on it.6 And, it seems clear, it is the presence
of this intention that makes it right to count his pushing as his trying to
move the log. After all, if he had not had any such intention having to do
with the possibility of moving the log, but had instead intended only that
by pushing on the log he would exercise his muscles, then it would clearly
not be right to count his pushing on the log as his trying to move it.
This example shows us a second informative sufficient condition for the
truth of ‘In Z-ing S tried to A’:
(S2) S intended of her Z-ing that by so acting she would find out
whether she could A by that Z-ing.
To have the intention ascribed to her in S2, S must believe that by Z-ing she
might find out whether she can A by that Z-ing. That is, S2 entails
(B2) S believed of her Z-ing that by so acting she would find out
whether she could A by that Z-ing.
But B2 does not entail S2. Suppose that Sam doesn’t know whether he can
move the log by pushing on it and, moreover, he doesn’t care. He pushes
on the log merely to exercise his muscles, but it occurs to him as he begins
to do so that this action might show him whether he can move the log by
94 C. Ginet
Is the disjunction ‘S1 or S2’ necessary for the truth of ‘By Z-ing S tried to A’?
No. Suppose Sam’s five-year-old daughter Suzy comes along and asks her
father to push on the log and move it. Sam, having ascertained that the log
is far too heavy for him to move, nevertheless pushes against it as hard as
he can in order to show Suzy that he cannot move it by pushing on it. It
would be correct to characterize Sam’s pushing on the log as his trying to
move the log, indeed trying as hard as he can to move it. He makes an
Trying to Act 95
attempt to move the log in order to show Suzy that he cannot move it; and, of
course, knowing that he cannot move it, he cannot intend of his pushing
on it that he will thereby move it. To intend that one’s attempt will suc-
ceed, one need not believe that it will succeed, but one must not believe
that it is impossible that it will succeed: one must believe that it has some
chance of success.
From this example we learn a third minimally sufficient informative con-
dition for the truth of ‘In Z-ing S tried to A’, namely, the following:
(S3) S intended of her Z-ing that by so acting she would show R (some
particular person or persons in the context) that she could not A by that
Z-ing.
S3, like S1 and S2, entails a belief of S’s about how her Z-ing is related to the
possibility of her thereby A-ing. S3 entails
(B3) S believed of her Z-ing that she would not A by so acting and
might persuade R that she could not A by that Z-ing.
But B3 does not entail S3. Suppose that while Sam is pushing hard on the
log in order to exercise his muscles it occurs to him that his so doing might
persuade Suzy, who is looking on, that he cannot move the log by pushing
on it. Still it need be no part of his plan that by pushing on the log he will
persuade Suzy of his inability. And if that is so, then, clearly, he is not try-
ing by pushing on the log to persuade Suzy of his inability. This supports a
claim that S3 is minimally sufficient for the truth of ‘By Z-ing S tried to A’.
At any rate, the weaker condition B3 is not sufficient.
Are we finished? Are S1–S3 all of the different minimally sufficient condi-
tions for the truth of ‘By Z-ing S tried to A’? I’m afraid not.
Suppose Sam knows that the log is far too heavy for him to move by push-
ing on it. But Sam is a soldier under the command of Rudy, and Rudy has or-
dered Sam to try to move the log by pushing on it. Sam would very much
prefer that Rudy not find out that he (Sam) cannot move the log by pushing
on it, but the penalty for refusing to obey a command would, in Sam’s view,
be a worse consequence than Rudy’s knowing that Sam cannot move the
log by pushing on it. So Sam pushes on the log as hard as he can in order to
comply with Rudy’s command to try to move it by pushing on it, but very
96 C. Ginet
What belief about her Z-ing does this intention require that S have? Pretty
clearly it requires that
It’s not clear to me, however, that in this case the entailed belief is a weaker
condition than the intention that entails it.
For one thing, it is unclear to me how one could believe that one’s action
complies with a command without intending that it comply with it, given
that ‘complying with a command’ means acting with the intention that
one’s action be what was commanded. Of course, it is possible to believe
that one’s action will satisfy the description given in a command or request
without intending to be complying with that command and even without
intending to be acting in such a way as to satisfy that description. One
might be acting entirely out of other motives.
Consider, for example, the parent of a child who has a game in which she
“requests” the parent to do what she knows the parent is going to do any-
way: “Please eat breakfast this morning.” “Please go to work today.” The
parent knows that his or her actions will satisfy these requests, but to sat-
isfy them is not among the parent’s intentions in so acting.
In this case, though, the parent does intend that her action will satisfy the
description that is in fact given in the child’s request; for example, that it
will be her eating breakfast. But not even this intention need be present in
an agent who believes at the time of her action that in the circumstances
it will satisfy the specification given in a request to her. Suppose Suzy is a
Trying to Act 97
guest for dinner at a friend’s house. She approaches the only remaining
vacant chair at the table with the intention to sit in it. Just as she is about to
sit, Rudy, who is already seated and whom Suzy intensely dislikes, says
“Please sit where I can gaze at your beautiful profile.” As Suzy sits she knows
that her action will satisfy Rudy’s description—it will be sitting where he
can gaze at her profile—but she certainly does not intend of her sitting that
it satisfy that description.
Matters seem different, however, when the description in the command
or request has the form that ‘S try by Z-ing to A’. For example, suppose Suzy
regularly exercises her muscles by pushing against a certain large tree outside
her house. Her daughter knows this, and as Suzy approaches the tree, her
daughter says, “I order you to try to uproot that tree by pushing on it.” Now
at the time of her action Suzy might or might not have the belief she would
express by saying, “In pushing on this tree I am obeying my daughter’s order
to try to uproot it by pushing on it”; but if she does have that belief, then she
believes that in pushing on the tree she is trying to uproot it. And if she be-
lieves that, it is hard to see how she could at the same time lack the intention
in pushing on the tree to try to uproot it. How could she think she is trying
and not intend to be trying? And if she has that intention then she is trying
to uproot it by pushing on it.
There are, of course, other sorts of beliefs entailed by S4 that clearly do
not entail it. For example, S can have the intention attributed by S4 only if
S believes that she is Z-ing. But this belief by itself does not entail that S’s
Z-ing is an attempt to A. So there seems no reason to suspect that S4 is not
a minimally sufficient condition for the truth of ‘By Z-ing S tried to A’.
But, as you have no doubt long since been aware, there is a problem with
S4: it is not as informative a condition as we’d like to have. It employs the
concept we wish to use it to explicate, the concept of trying to do some-
thing. Is there any way we can refine S4 so as to rid it of this feature?
Let us approach this problem by considering the following question: if R
asked S to try to A by Z-ing, what must R have believed about the possibility
of S’s A-ing by Z-ing for this to be a serious request, one that R thought it was
possible for S to comply with? It seems to me, for starters, that R need not
have been joking if R had any one of the following beliefs:
R believed that by Z-ing S might A.
R believed that S believes that by Z-ing she might A.
98 C. Ginet
But these are not the only sorts of beliefs on R’s part that would show that R
thinks it possible for S’s Z-ing to be S’s trying to A. Consider again the situa-
tion where Sam is a soldier under Rudy’s command. Being sadistic and en-
joying for its own sake the exercise of his authority over Sam, Rudy seriously
commands Sam to try to move the log by pushing on it as hard as he can,
even though Rudy does not believe that there is any chance at all that Sam
could move the log by pushing on it or that Sam believes that there is; and,
moreover, Rudy does not believe, or believe that Sam believes, that there is
someone around who needs to be persuaded of the fact that Sam cannot
move the log by pushing on it. (For a less dark story, imagine that Rudy is not
Sam’s commander but instead an eccentric friend who has offered Sam a
large sum of money for just trying to move the log by pushing on it.)
On the other hand, not just any action that Sam could perform is such
that Rudy can seriously command Sam to try to move the log by perform-
ing that action, given that Rudy fails to believe (or believe that Sam be-
lieves) that there is any chance at all that Sam could move the log by that
action or to believe (or believe that Sam believes) that there is someone
around who needs to be persuaded that Sam cannot move the log by
that action. For example, Rudy cannot seriously command Sam to try to
move the log by whistling Dixie, if Rudy does not believe (or believe that
Sam believes) that there is any chance at all that Sam could move the log
by whistling Dixie or that there’s someone around who needs to be per-
suaded that Sam cannot move the log by whistling Dixie. What is the
difference between pushing on the log and whistling Dixie, vis-à-vis the
possibility of moving the log by such actions, that explains this?
The difference, I think, is that, although Rudy does not believe that Sam
can move the log by pushing on it, Rudy does believe (or believes that Sam
believes) that pushing on the log is a sort of action that might, if it were
performed with sufficient strength or skill or in the right manner, succeed in
moving the log; whereas Rudy does not believe (or believe that Sam believes)
that whistling Dixie is a sort of action that might, if performed with suffi-
cient strength or skill or in the right manner, succeed in moving the log.
Trying to Act 99
Note that the belief that pushing on the log is a sort of action that might,
if it were performed with sufficient strength or skill or in the right manner,
succeed in moving the log is entailed by the belief that Sam might move
the log by pushing on it: one who believes the latter believes the former.
The general lesson of these observations is that, when R orders S to try to
A by Z-ing, S can take this order seriously (can suppose that R thinks it is
possible for S to comply with the order) only if S thinks that R believes (or
believes that S believes) that Z-ing is a sort of action by which someone
might succeed in A-ing if they performed it with sufficient strength or skill
or in the right manner, or S thinks that R believes (or believes that S
believes) that by Z-ing S would show someone that S cannot A by Z-ing.
It is not necessary that the command (or request) have been that S try to
A by Z-ing. The command might, for instance, have been just that S try to A
or just that S Z. If the content of Rudy’s order were simply that Sam try to
move the log, and in response Sam pushes on the log as hard as he can with
the intention and belief that his so doing will comply with Rudy’s order
and that it will do so because Rudy’s reason for giving the order was his
belief (or belief that Sam believes) that pushing on the log, if performed in
the right manner with sufficient strength or skill, is a sort of action by
which someone might succeed in moving the log, then it follows that in
pushing on the log Sam was trying to move it.
It is likewise if the content of Rudy’s order were simply that Sam push on
the log. If in response Sam pushes on the log as hard as he can with the
intention and belief that his so doing will comply with Rudy’s order and
that it will do so because Rudy’s reason for giving the order was his belief (or
belief that Sam believes) that pushing on the log, if performed in the right
manner with sufficient strength or skill, is a sort of action by which some-
one might succeed in moving the log, then it follows that in pushing on
the log Sam was trying to move it.
Or suppose that in response to either of those orders Sam pushes on the
log with the intention and belief that his so doing will comply with Rudy’s
order and will do so because Rudy’s reason for the order was that he believes
(or believes that Sam believes) that by pushing on the log Sam would show
someone that Sam cannot move the log by pushing on it. Again it follows
that in pushing on the log Sam was trying to move it.
What if Rudy’s command were that Sam should move the log by pushing
on it and, as before, Sam believes that it is impossible for him to move it by
100 C. Ginet
pushing on it?7 Then Sam cannot intend that his pushing on it will com-
ply with the order. The most he can intend is that it will be as good as he can
do to comply with the order. But if he does intend this latter, then his push-
ing on it will be his trying to move it by pushing on it.
Generalizing, we can say the following: If S believes (and intends) that
her Z-ing will comply with the command, or will be as good as she can do
to comply with it, and it will be so because the giver of the command
believes (or believes that S believes) that Z-ing, if performed with sufficient
strength or skill or in the right manner, is a sort of action by which some-
one might succeed in A-ing then in Z-ing S is trying to A.
And if S believes (and intends) that her Z-ing will comply with the com-
mand, or will be as good as she can do to comply with it, and it will be so
because the giver of the command (or request) believes (or believes that S
believes) that by Z-ing S would show someone that S cannot A by Z-ing,
then in Z-ing S is trying to A.
These conditionals hold whether or not the agent S believes that there is
any chance she might A by her Z-ing, or believes that there is someone
around who needs showing that S cannot A by Z-ing, or that Z-ing, if per-
formed with sufficient strength or skill or in the right manner, is a sort of
action by which someone might succeed in A-ing. Thus, for example, if
Rudy gives Sam the order to try to move the log by whistling Dixie and Sam
thinks Rudy meant the order seriously because Rudy thinks that whistling
Dixie in the vicinity of a large heavy object, if performed in exactly the
right manner, is a sort of action by which someone might succeed in mov-
ing the object, then Sam can believe and intend that by whistling Dixie he
is complying with Rudy’s order, even if Sam does not share Rudy’s belief
about the object-moving power of whistling Dixie.
This result gives us a way to formulate an informative condition that
covers those new cases of trying to A by Z-ing that S4 brings in. The for-
mulation is the following:
(S4) S intended of her Z-ing that it would comply with a command (or
request) to her and S also either (i) intended that it would do so for the
reason that the giver of the command (or request) believes (or believes
that S believes) that Z-ing, if performed in the right manner with
sufficient strength or skill, is a sort of action by which someone might
succeed in A-ing or (ii) intended that it would do so for the reason that
Trying to Act 101
the giver of the command (or request) believes (or believes that S believes)
that by Z-ing S would show someone that S cannot A by Z-ing.
If in her Z-ing S has a complex intention satisfying the specification in
S4, then her Z-ing is her trying to A.
S tried to A if there was an action of S’s, S’s Z-ing, such that either
(S2) S intended of her Z-ing that by so acting she would find out
whether she could A by that Z-ing,
(S4) S intended of her Z-ing that it would comply with a command (or
request) to her and S also either (i) intended that it would do so for the
reason that the giver of the command (or request) believes (or believes
that S believes) that Z-ing, if performed in the right manner with
sufficient strength or skill, is a sort of action by which someone might
succeed in A-ing or (ii) intended that it would do so for the reason that
the giver of the command (or request) believes (or believes that S believes)
that by Z-ing S would show someone that S cannot A by Z-ing.
S1–S4 specify four different intention contents each of which is such that,
if S’s Z-ing is accompanied by an intention with that content, then S’s Z-ing
is her trying to A. As far as I can discern, the only feature these intention con-
tents have in common is that each specifies some relation or other between
S’s Z-ing and the possibility of S’s A-ing. It would be nice in a way—in that it
would be more in accord with our preanalytic hopes—if there were some-
thing more specific tying together the sufficient conditions for trying to act.
102 C. Ginet
Notes
6. I say ‘by that act of pushing on it’, which refers to the particular action he per-
formed, rather than ‘by pushing on it’, which refers to the type of action, because
there is the following possibility (brought to my attention by Richard Reilly): Sup-
pose Sam decides that he will move the log off the driveway later, after lunch, if he
can. Sam knows that, if he can move the log at all by pushing on it, he can move it
all the way off the driveway by (a sustained period of) pushing on it. So he pushes
hard on the log for a few seconds and it moves a little bit, whereby he finds out that
he can move it off the driveway by pushing on it. But he did not try to push it off the
driveway, because it was not his intention in pushing on it to find out whether he
could push it off the driveway by that act of pushing on it.
7. I’m grateful to Harry Silverstein for calling my attention to the import of this
possibility.
References
Dana K. Nelkin
1 Introduction
It is important to determine whether (R) is true, then, for at least two rea-
sons. First, if (R) is true and we are not free, then one of our most central
conceptions of ourselves is both false and inescapable. On the other hand,
if Reid and Kant are correct, and (R) can be shown to be true, then a major
premise can be established in an argument that we are in fact free.3
Is (R) true? In this paper, I will explore two lines of reasoning supporting
the conclusion that it is. I believe that one of these lines is promising, and
will suggest some different ways in which it might be developed. I begin by
offering a more detailed account of the sense of freedom attributed to all
rational deliberators. As might be expected, philosophers have understood
(R) in different ways. In section 2, I examine what I call the ‘indeterministic’
reading, according to which the sense of freedom in (R) is the belief that
one can choose from a set of undetermined alternative courses of action.
Although I argue that, on the indeterministic reading, (R) is false, I believe
the ways in which it has been defended contain important insights. In
section 3, I argue that we should adopt what I call the ‘belief-concept’ read-
ing of (R), according to which the sense of freedom in (R) is the belief that
one’s actions are up to one in such a way that one is accountable for them.
qualifies his claims about the proper objects of deliberation in a way that
suggests that deliberation requires the belief that the object of deliberation
possesses certain properties rather than that the object of deliberation
actually possesses those properties.5 What is essential to deliberation are the
subject’s epistemic attitudes about the object of deliberation, and, in par-
ticular, the belief that any object of one’s deliberation is possible but not
determined.
While the inspiration for the indeterministic reading can thus be traced
to Aristotle, the view has many contemporary adherents, and some have
thought Kant to be another advocate.6 Despite its long history and current
popularity, I claim that the indeterministic reading is false.
Let me begin by setting out one line of reasoning that might be used to
support the indeterministic reading of (R).7
(i) If one deliberates about an action, A, then one must believe that it is
in one’s power to do and to forgo A.
But,
Therefore,
(iii) If one deliberates about A, then one must believe that there exist
no conditions sufficient to render inevitable either A or not-A.8
Premise (i) has seemed to many to be an attractive thesis, but I will present
several thought experiments designed to undermine it. Then I will go on to
challenge what are perhaps even more basic assumptions on which the in-
deterministic view must ultimately rest: assumptions about the very nature
and purpose of rational deliberation.
First, consider the following case: imagine that you know that a brilliant
scientist has the ability to fiddle with your brain in a way that causes you
to act as she wishes you to. You know that she wants you to vote for Gore
over Bush in the upcoming presidential race, and that if you do not decide
to vote as she wishes, she will cause you to vote that way. So, for instance,
you know that if you were to prepare to vote for Bush or otherwise fail to
decide to vote for Gore, the brilliant scientist would cause you to vote
for Gore. It seems to me that you could still evaluate the reasons for voting
108 D. K. Nelkin
for each candidate and decide to vote for Gore on the basis of those
reasons. In this case, contrary to (iii), you know that conditions exist which
are sufficient for your voting for Gore, while you nevertheless deliberate
and decide to vote for him. Further, contrary to (i), you do not believe you
could forgo the action upon which you decide.9
Now one might object that your voting for Gore as a result of deliberation
and decision differs in kind from the action of voting for Gore as a result of
the brilliant scientist’s interference with your brain. One might think that
the actions are of different types, especially if one is convinced that types of
actions are individuated in part by the types of causes they have.10 If this is
so, the objection goes, then you do deliberate about something that you
believe you could do or forgo: a particular type of voting for Gore.
Although this objection has some force, I think that it can be met. For if
you accept the initial description of the thought experiment, what you
evaluate with a view to acting are the reasons for voting for Gore. The
objection, by contrast, has you evaluating reasons for voting for Gore on
the basis of deliberation and decision, where this action is something that
you could do or forgo. Even if this sort of deliberation is possible, it remains
a coherent part of the thought experiment that you deliberate and decide
to vote for Gore. Thus, additional objects of deliberation that might be
regarded as actions to be done or forgone within the thought experiment
do not alter the fact that there is something about which one deliberates
and decides, despite not believing that one could either do or forgo it.
The objection might be pressed in the following way, however: consider
a modified version of the original voting case (the ‘inevitability’ case) in
which you believe it to be inevitable that, no matter what, the scientist will
fiddle with your brain in such a way that you vote for Gore. To ensure that
the case is one in which it is inevitable that your action will have the same
cause in counterfactual circumstances, let it be built into the case that there
is no possibility of overdetermination: you believe that, no matter what,
the changes in your brain induced by the scientist will be the sole cause of
your voting for Gore. In this case, the objection proceeds, surely you can-
not deliberate, and the explanation is that you fail to believe that the
action in question—voting for Gore in a particular way—is one that you
could either do or forgo. If the lack of such a belief is the correct explana-
tion of the failure of deliberation in this case, then such a belief must be a
necessary condition for deliberation. Further, this conclusion shows that
The Sense of Freedom 109
she believes that her action is physically determined.13 The question is, can
such a person deliberate if she knows that such a computer is busy churn-
ing out its predictions? The answer seems to me to be “yes.” I can imagine
that such a computer exists right now somewhere in Siberia, printing out
predictions about my future actions. Even if I believed this to be true, and
so did not believe it possible for me to perform each of a number of alter-
native actions, I would go right on deliberating about all sorts of things.
To the case of the deliberating determinists, the following reply might be
offered. Such people do exist, and while they believe that determinism is
true, they also hold the contradictory belief that determinism is false.14
Thus, although there are those who believe that determinism is true, as long
as they continue to deliberate their behavior manifests the belief that they
can do or forgo the actions about which they deliberate, and thus manifests
the belief that determinism is false. In fact, it appears that the proponent of
the indeterministic reading could account for every purported counterexam-
ple to (i) or (iii) simply by attributing contradictory beliefs to every delibera-
tor who claims to have a belief contrary to that attributed in (i) or (iii).
However, the idea that deliberation manifests either of these beliefs can be
seen to rest on still more fundamental assumptions which can themselves
be questioned, assumptions about the nature and point of rational deliber-
ation. For we can ask: what are the conditions under which we take a piece
of behavior to manifest a belief? The answer is that, in general, we do so
when we think that the behavior is rationalized by a belief, that is, when we
think that the presence of the belief in question is part of the correct expla-
nation of the behavior that makes it intelligible in the context of the agent’s
beliefs, desires and actions.15 For example, consider the case of a store owner
who hires a private detective to watch and report on his cashier’s behavior.
In the absence of a belief that it is likely that the cashier is embezzling
money, the store owner’s action does not make sense in light of his other
actions and attitudes. Thus, this belief rationalizes his hiring of a detective.
Given this general understanding of what it would be for an activity to
manifest a belief, let’s consider why it might be thought that the beliefs
attributed to deliberators by the indeterministic view are manifested in the
activity of deliberation. I will focus directly on the belief attributed in (iii) for
the sake of simplicity, although the reasoning can easily be transposed to
account for the belief attributed in (i), as well. Here, then, is the question: in
what way can the belief attributed in (iii) be said to rationalize deliberation?
The Sense of Freedom 111
Either of two answers might be given, one having to do with the essen-
tial nature of deliberation, the other with its point, or purpose. Richard
Taylor provides material for the first sort of answer. He writes that if an
agent, Adam, believes that conditions already existed sufficient for his
performing the action he will perform in the future, then “he can no longer
deliberate about the matter because . . . he believes it is not up to him what
he does; the matter has already been ‘decided’, one way or the other, and
there is no decision for Adam to make” (Taylor 1964, 77). On one natural
reading of this passage, the purpose of deliberation is to decide upon an
action, where ‘decision’ is constituted by an agent’s closing all but one of
several previously ‘open’ possibilities. It follows that if conditions exist
sufficient for a particular action’s being performed, and hence that there
are no previously open possibilities to be closed or to be chosen to remain
open by the agent, then there can be no true decisions on the part of any-
one. It could then be argued that if an individual did not believe that his
deliberation could succeed in bringing about a decision, then his deliberat-
ing would not make sense. Therefore, the belief that he could make a
decision would rationalize his deliberation, and so that belief would be
manifested in the activity of deliberation. Finally, given the present under-
standing of ‘decision’, it follows that one can only deliberate if one believes
that there are open possibilities for one to close.
The problem with this suggestion is that the understanding of ‘decision’
on which it is based ought to be rejected. For on the view that a decision is
the closing of previously open possibilities, it is not strictly speaking true to
say that someone has reached a decision if, for some reason, it cannot be
carried out. Yet the naturalness of saying that someone made a decision
that could not be carried out suggests that we do not normally think of a
decision as being defined by its effects. Rather, it is more natural to say that
a decision is something like the formation of an intention which, depend-
ing on the circumstances, may or may not be carried out. According to this
latter understanding of ‘decision’, we can successfully decide to do some-
thing even if the decision does not succeed in effecting anything, let alone
in the closing of all but one of several previously open possibilities.
The second way in which the belief attributed in (iii) might be said to
rationalize deliberation involves a somewhat more modest view of the con-
nection between determinism on the one hand, and deliberation and deci-
sion on the other. The idea is that if deliberation and decision are to have a
112 D. K. Nelkin
point at all, then they must succeed in closing all but one of several previ-
ously open possibilities. This view is suggested by Aristotle, who, in arguing
that it would be absurd to think that everything that happens, happens of
necessity, claims that in that case “there would be no need to deliberate or
to take trouble” (Aristotle 1984d, 29 [DI 9, 18b31–2]).16
On this view, one could be said to make a decision even if determinism
were true, although in that case the decision would lack a point. It could
then be argued that if an agent failed to believe that there was a point to
making a decision, then it would not make sense for her to deliberate.
Therefore, deliberation manifests the belief that there is a point to making
a decision, and thus, that there are open possibilities to be closed by the
agent as a result of deliberation and decision.
But is it in fact true that rational deliberators can only take rational
deliberation to have a point if they take it that their decisions result in the
closing of all but one of several previously open possibilities? Of course, I
take it that the original voting case provides a negative answer here. For I find
it perfectly coherent that the agent in that case does not believe that her pur-
pose in deliberating is to change the course of events from what it would oth-
erwise be, or even to find out what she will do, and yet the possibility that
she deliberates is perfectly intelligible. Yet merely appealing to this sort of
case will not convince those who would attribute contradictory beliefs to the
voter. I must say what it is that the case reveals about our attitudes toward the
point of deliberation. And that is that reasons can motivate us to act in a cer-
tain way even if we know that we will act in that way no matter what. The
case shows us that the possibility of acting on the basis of good reasons can
itself be seen to have value, and so, one can take the activity of rational de-
liberation to have a point if one takes it that the activity can result in one’s
acting for good reasons. And the possibility that rational deliberators could
take deliberation to have a point in the absence of indeterminism under-
mines this kind of defense of the indeterministic reading. Thus, we have not
been given sufficient reason to conclude that the belief attributed in (iii) is
manifested by rational deliberation.
Each of the two attempts to provide an argument for the conclusion that
rational deliberation manifests the belief attributed in (iii) relies on a prob-
lematic premise: the first rests on a dubious assumption about the nature of
a decision, the second on a questionable assumption about the point of
deliberation. Of course, there may be reasons other than the two that I
The Sense of Freedom 113
have considered here for accepting the conclusion. However, the burden
rests with those who advocate the indeterministic reading; for the fact
that the purported counterexamples to it can only be resisted at the cost of
attributing contradictory beliefs to otherwise rational people provides
some prima facie reason to reject the view.17
Having made it clear which aspects of the reasoning to the indetermi-
nistic reading I wish to reject, let me emphasize an aspect of the reasoning
that I have not rejected. I believe that the general strategy of showing that
the sense of freedom rationalizes the very activity of rational deliberation
is a promising one, and I will pursue it in section 3. In particular, I will
return to the idea that for rational deliberation itself to be an intelligible
activity, deliberators must take such activity to have a point or purpose.
For even if deliberation can be seen to have a point in the absence of inde-
terminism, it does not follow that it has a point in the absence of freedom.
And if we refrain from prejudging the question of whether freedom
requires indeterminism, it is possible that, in an important sense of ‘free’,
we cannot take our deliberation to have a point if we do not have a sense
that we are free.
from among alternative courses of action’ or that ‘one could have done
otherwise’ or that ‘one is the undetermined source of one’s actions’. Yet,
there have been numerous attempts to provide such conditions that do not
allude to the choosing among alternatives, having the ability to do other-
wise, or being undetermined. I do not mean to suggest that any such at-
tempts have met with complete success, only that they have been genuine
attempts. If the concept of freedom were the notion that one can do oth-
erwise, then it would seem that agreement as to the entailment between
freedom and the ability to do otherwise would be readily forthcoming. The
fact that many who have attempted to provide necessary and sufficient
conditions for freedom have not granted the point suggests that the
concept of freedom is not simply the notion that one can do otherwise.
Further, even if we were to accept the point of the objection for the sake of
argument, we can still accept that the notion I have described is an impor-
tant condition for moral and other sorts of responsibility.21 Thus, if it can
be shown that we must represent ourselves as free in the sense that I have
described, that would be a significant result. Further, if Kant and Reid
are correct, and (R) is the first step in a sound argument that we are free
(in the sense specified in (R)), then that, too, would be a significant result.
Finally, the concept I have described satisfies widely accepted constraints
on the notion of freedom that is at the heart of the free will debate. For
example, philosophers often claim that the notion of freedom in which
they are interested is a notion that applies to persons, but not to many
nonhuman animals who are thought not to be persons.22 That is, there is a
sense in which the actions of humans can be up to them, or be their own,
in a way that the actions of a spider cannot. Since the notion of account-
ability might be thought to be limited in its applicability in such a way as
to exclude many nonhuman animals, the concept that I have described
matches this widely accepted feature of freedom. The readings I will con-
sider in what follows take ‘free’ in the sense specified in (R) to pick out the
concept of freedom. Hence, they are ‘conception-neutral’.23
does not follow from this reading that all rational deliberators can articulate
their commitment without considerable probing and reflection; but it does
follow that they actually represent themselves as free. In this respect, it is
like the indeterministic reading of (R). In what follows, I call the conception-
neutral reading that understands ‘sense’ in this way, the ‘belief-concept’
reading. In the final subsection, I consider a weaker reading of (R) according
to which the sense of freedom is a rational commitment that one is free.
Roughly, one is rationally committed to a proposition when reflection and
recognition of the features of one’s own mental states and reasoning are suf-
ficient for one’s believing that proposition. But one need not actually be-
lieve that one is free in order to be rationally committed to the proposition.
I call the reading of (R) that incorporates both the conception-neutral reading
of ‘free’ and the rational commitment reading of ‘sense’, the ‘commitment-
concept’ reading. Although I believe that there is good reason to accept the
belief-concept reading, it is true that even weaker premises are needed to
support the commitment-concept reading, and thus, that the latter requires
less in the way of defense. At the same time, as I will explain, both readings
can do powerful explanatory and justificatory work.
her purpose to act for good reasons. In order to do so, she must take it that
her deliberation can be efficacious and that there are reasons on the basis
of which she should act. Hence, as I argued earlier, she must have a sense
that she is free. The fact that one can imagine a rational deliberator taking
her activity to have any of a number of different points does not show that
rational deliberators can avoid taking their activity to have the point of
acting on the basis of good reasons.28
Further, in addition to consulting our intuitions about the nature of
rational deliberation in the abstract (as I have so far been urging us to do),
it is useful to consider in more concrete terms the question of whether
rational deliberators must take deliberation to have a particular purpose. To
this end, consider our own behavior when we believe that there are no
good reasons for acting in a certain way. For example, suppose you are at an
ice cream shop, having decided to buy an ice cream cone. You believe that
all the flavors are equally good. You don’t generally deliberate in such cir-
cumstances, but simply pick a flavor at random. On my view, the reason for
your lack of deliberation in such a case is that you see no point in deliber-
ating. We see no point in engaging in the evaluation of reasons for acting
when we think that there is no reason that we should act in a particular
way. We are not in a position in which we could act for good reasons in
choosing a flavor of ice cream, hence we see no point to deliberating, and
so do not engage in it.
This thought experiment—together with the investigation into the
nature of goal-directed intentional activity I sketched earlier—strongly sug-
gests that the very activity of rational deliberation manifests the sense that
there is a point or purpose to the activity. Further, these considerations
suggest that rational deliberation manifests the sense that the point of
one’s activity is to decide and ultimately to act on the basis of good reasons.
At this point, a second objection might naturally arise: even if one has a
conception of the purpose of one’s activity, one need not believe that one’s
purpose can be achieved in order to be rational in engaging in it. In what
follows, I defend the idea that one must believe that one has the general
ability to succeed in order to have a conception of one’s purpose in
performing goal-directed intentional actions. There is a long history of
controversy on this and related questions, and I cannot do full justice to it
here.29 But I will suggest one approach to it in the context of defending a
key premise in the reasoning for the belief-concept reading. Let me begin
120 D. K. Nelkin
by emphasizing that I am not relying on the claim that one must believe
one can succeed on every occasion. Rather the claim is that one must be-
lieve one can succeed in typical circumstances when one attempts the rele-
vant type of action. This point might be sufficient to defuse the present
worry about the belief-concept reading. For as I hope to show, those on
both sides of the debate over the connection between intentional action
and belief might be able to find agreement once this point is noted.
In thinking about the relation between intentional action and belief,
intuitions about examples play an important role. Here is an example that
provides support for the idea that one must believe one can achieve one’s
purpose, when one has a conception of it in acting intentionally: I cannot
intentionally engage in any activity that might be described as either flying
or trying to fly, no matter how much I desire to and how much I flap my
arms. The natural explanation seems to be that I do not believe it possible
for me to fly. In fact, I believe that I lack the general ability to fly. I can
imagine circumstances in which someone else tries to fly. But these
circumstances include delusions on the part of the person making the
attempt. If someone were to believe that his physical abilities were very
different from what they are, or that the laws of aerodynamics or gravity
were different from what they are, then he could try to fly. (Of course, this
would require a great number of changes in one’s belief system and proba-
bly a great deal of irrationality.) Here, it seems that what explains the
difference between this person and me is the difference between our beliefs
about what flapping our arms could possibly accomplish.
Those who argue against a necessary connection between belief and
intentional action offer examples on the other side. Kirk Ludwig offers this
one, among others30: A friend insists that I can make a basket from half-
court. I believe it impossible for me to make a basket from that distance,
and set out to show him that even if I try as hard as I can, I still can’t do it.
I try, and, amazingly, make the basket. This appears to be a case in which I
intentionally make a basket without believing that it is possible. According
to Ludwig, my intentionally making the basket in this case shows that I
also intended and tried to make it before surprising myself with my success.
Since I have argued that rational deliberators engaged in intentional goal-
directed activity have a sense of their purpose, the case seems to be one in
which the sense of purpose is not accompanied by a belief that it can be
achieved.
The Sense of Freedom 121
Reasons might be offered to resist this example and others like it, but
what is more important for our purposes here is the kind of example it is
(and is not). For even those who argue that intentional action does not re-
quire the belief that success is possible agree that there are some things one
cannot intentionally do (or even try to do), like flying or making a basket
from ten miles away. Some explanation for this fact is required.
Ludwig, for example, acknowledges that you could not try to hit a home
run by holding the bat in a “bunt” position.31 Why not? The reason, ac-
cording to Ludwig, is that this stance and the limited motion it permits are
not designed to bring about the end of hitting home runs. In cases like this
in which one intentionally performs an action without believing that one
can succeed, one might recognize that circumstances are special in such a
way that one’s activity cannot succeed. This is a cashing out of the idea that
in order to act intentionally, one must conceive one’s actions as of a type
designed to bring about a certain end.
If this is right, then we have a way to distinguish between the cases of mak-
ing a basket from half-court and swinging “all out” for a home run on the
one hand, and making a basket from Mars and a home run from a “bunt”
position on the other. The view also suggests that we assume that when we
act intentionally our actions are of a type that, under at least some typical
circumstances, can succeed. Further, there is a rationale for this: if one views
a certain kind of activity as one some of whose typical instances are success-
ful in achieving their purposes, then one can have a conception of how they
are successful. This makes it possible for us to envision our own activity of
this type as a way of implementing a plan to achieve its purpose.32
Admittedly, it is difficult to give criteria for ‘type’ of action here. But I
think it is possible to rely on an intuitive idea. I can’t try to fly, or make a
home run from the bunt position, because these are not the kinds of
actions at which I could succeed under anything like normal circum-
stances. I must believe that I have the general ability to succeed, if I have a
sense of my purpose in acting.33
This conclusion is all that is needed in order to defend the belief-concept
reading of (R) from the present objection. For even if one need not believe
that one can succeed on a particular occasion of deliberation, one must be-
lieve that one has the general ability to succeed. One must take it that one’s
activity is successful under at least some typical circumstances. Otherwise,
one won’t be able to view one’s engagement in the activity as counting as
122 D. K. Nelkin
a way of achieving the purpose of adopting good reasons for acting. Thus,
one must believe that one is free in the conception-neutral sense.
As I have argued, this reasoning should persuade those on a wide spec-
trum of views positing particular connections between intentional action
and belief. However, others might worry that, according to this reasoning,
rational deliberation requires too much in the way of conceptual develop-
ment and self-reflection. Young children and nonhuman animals would
appear to be counterexamples to the reasoning behind the belief-concept
reading of (R) since they include rational deliberators who lack mental-
state and related concepts such as ‘action’, ‘reasons’, and ‘responsibility’,
and those who have not reflected on the purpose of their activity. This sort
of worry deserves to be taken seriously.
However, I believe that the appeal of the alleged counterexamples is
undermined by reflection on the nature of the central concepts in the rea-
soning behind the belief-concept reading.34 First, let me emphasize that
rational deliberation as I understand it is itself a very sophisticated cogni-
tive activity: the consideration and evaluation of reasons with a view to
deciding to act, where one’s decision is based on one’s evaluation and
adoption of reasons as one’s reasons for performing an action. Once this is
understood, it becomes difficult to maintain that young children and
nonhuman animals provide counterexamples to the claim that all rational
deliberators must have the concept of reasons, for example. For the cogni-
tive sophistication required to engage in rational deliberation itself would
seem to rule out at least some members of these groups, and, in particular,
the very same members who are excluded from the possession of sophisti-
cated concepts such as reasons.
At the same time, it is important to emphasize once again that the concepts
in question do not include the concepts of moral reasons and moral respon-
sibility, but the concepts of reasons and responsibility in a basic sense. Thus,
the argument for the belief-concept reading of (R) does not presuppose the
possession of any moral concepts by rational deliberators. Recognition of the
robust understanding of rational deliberation articulated above together with
the relatively basic nature of the notions of reasons and responsibility should
dispel the worry raised by the alleged counterexamples that there must be a
faulty step somewhere in the reasoning behind the belief-concept reading.
Finally, even if lingering doubts remain about the strength of the belief-
concept reading of (R), it is possible to adopt a weaker conception-neutral
The Sense of Freedom 123
one way of one’s actions being up to one such that one is accountable for
them depends partly on the agent’s ability to respond to good reasons for
acting. But I believe that this is a connection that we should find attractive,
particularly once the connection between obligation by reasons and ac-
countability is noted. For if one’s actions are up to one in such a way that one
is (in a nonmoral sense) accountable for them, then we would seem to have
just what is wanted in a ‘freedom’ condition for moral responsibility.
Finally, even if we set aside the question of whether the concept of free-
dom is the concept of one’s actions being up to one such that one is
accountable for them, the latter concept remains a significant one. For it is
one that provides an important condition for moral and other sorts of
responsibility, and one that we care deeply about.
In sum, the belief-concept reading of (R) faces important challenges. Yet
an intuitively plausible line of reasoning gives it considerable resources
with which to respond. As a result, the belief-concept reading (together
with the weaker commitment-concept reading) remains a promising
interpretation of the widely accepted idea that in virtue of being rational
deliberators we cannot escape the sense that we are free.
4 Conclusion
Both the indeterministic reading and the belief-concept reading of (R) rest
on reasoning that takes the sense of freedom to be a belief manifested by
the very activity of rational deliberation. They diverge in the content of the
beliefs each attributes to rational deliberators. Equally important is the
difference between the notions of rational deliberation employed by each
reading, for it is on this difference that the difference in beliefs ultimately
rests. Unlike the indeterministic reading, the belief-concept reading takes
the essence of rational deliberation to be the attempt to find and adopt
good reasons for acting. By building on the reason-seeking features of
rational deliberation, it is possible to explain why rational deliberators
must have a sense of their actions being up to them in such a way that they
are accountable for them. At the same time, this focus on the reason-seeking
features of rational deliberation makes the belief-concept reading of (R)
particularly well suited to an antiskeptical argument in the spirit of Kant
and Reid. For if rational deliberation is essentially a reason-seeking faculty,
it is tempting to conclude that the simple possession of such a rational
The Sense of Freedom 125
faculty could not be responsible for a false belief. Of course, turning this
thought from a tempting idea into the conclusion of a sound argument
that rational deliberators are free is a project of its own. Yet if the belief-
concept reading of (R) is correct, then we are at least entitled to a key
premise in such an argument.
Acknowledgments
Notes
1. For example, Galen Strawson writes that a free will skeptic who concentrates on
abandoning his “ordinary conception of freedom” may temporarily experience “a
total paralysis of all purposive thought as it is ordinarily conceived and experienced”
(Strawson 1986, 102). It is when we are trying to make decisions (moral and
otherwise) that our belief in our own freedom is most acutely felt. (See p. viii, and
chapter 3.) At the same time, Strawson argues that freedom is impossible and that
our belief is false. It must be noted that Strawson does not accept (R), but rather the
weaker claim that humans who rationally deliberate necessarily believe themselves
free.
Castañeda accepts something like (R), and at the same time accepts the possibility
that the skeptic is right: if so, “the universe is ugly; given the biological and psycho-
logical primacy of practical over contemplative thinking, we are, thus, condemned
to presuppose a falsehood in order to do what we must think practically” (Castañeda
1975, 134).
126 D. K. Nelkin
2. Controversy abounds over just how Kant should be interpreted here. I will not
enter into that controversy here.
3. I explore different ways this argument can be developed in Nelkin (in preparation).
4. See also Aristotle (1984c, 1798 [NE 1139a, 13–14]) for a similar statement.
5. See, for example, Aristotle (1984b, 1941 [EE 1225b, 34–36]) : “. . . nor does he even
choose what is possible, generally, if he does not think it in his power to do or abstain
from doing it.” See also, Aristotle (1984b, 1942 [EE 1226a, 25–26]) : “. . . about these
[things the production of which is not in our power] none would attempt to deliberate
except in ignorance.” These qualifications have suggested to some that what Aristotle
meant in expressing the apparently stronger claim that deliberation requires indeter-
minism was actually the weaker claim that deliberation requires a belief in indeter-
minism. There is some dispute about this, however. See Gail Fine (1981, 572 and
note 10) for a statement of this view, as against Sorabji (1980, 228 and 245–246), who
credits Aristotle with the stronger claim and argues only that Aristotle should have
offered the weaker. See my note 16 for further textual evidence in support of the view
that Aristotle held the weaker thesis.
6. See, for example, Castañeda (1975) who accepts the indeterministic reading. Van
Inwagen accepts something like the indeterministic reading of (R). (See van Inwagen
1983). He argues that deliberators must believe that multiple alternative possibilities
are each within one’s power. However, van Inwagen stops short of claiming that de-
liberators must believe that these alternative possibilities are undetermined. Those
who have understood Kant as advocating the indeterministic reading of (R) include
Castañeda (1975, 134) and Thomas Hill (1985, 16–17).
7. See Richard Taylor (1964) and van Inwagen (1983, especially 152–161), for simi-
lar lines of argument. Taylor argues that a deliberator must not believe that she
cannot choose among undetermined alternatives rather than that such a deliberator
must believe that she can so choose. Van Inwagen adapts Taylor’s argument in order
to argue for the presence of a belief. As mentioned in my note 6, van Inwagen argues
that all deliberators necessarily believe that multiple alternative possibilities are
within their power.
8. This is not the only line of reasoning that might be used in support of the inde-
terministic reading of (R). One might begin with the claim that one can deliberate
only if one believes oneself to be the ultimate source of one’s actions, and infer that
one must believe determinism to be false in order to do so. Although I believe the
line of reasoning set out in the text to be the one most often deployed, much of what
I go on to say addresses this second line, as well.
11. See Randolph Clarke (1992), for a similar line of reasoning against the indeter-
ministic reading.
12. It might also be objected that while the original voting case shows that (i) and
even (iii) are false, it is nevertheless a case in which the agent believes that there is
something she can do or forgo (e.g., voting for Gore in a certain way), even if this
is not the (primary) object of deliberation. Thus, the case leaves open the possibility
of finding reasons other than (i) through (iii) for accepting (R), including the rea-
soning described in note 7. The case to follow in the text can be used to address this
objection, as well.
13. An added virtue of this case is that it is immune to the sort of objection we saw
earlier concerning action individuation. For if the agent believes that her future
actions are determined by past physical states together with natural laws, then she
believes that conditions are sufficient for her performing a particular action in a
particular way (i.e., with certain causal antecedents).
14. A similar sort of reply is suggested by van Inwagen. However, he offers the reply
on behalf of free will skeptics who deliberate rather than on behalf of determinists
who deliberate. See van Inwagen (1983, 157–158).
16. This passage also provides further support for the claim that Aristotle did not
hold the view that deliberation requires indeterminism. For if he had held that view,
then it would have been natural for him to identify the claim that we could not
deliberate as an absurd consequence of determinism. But in fact, he identifies the
weaker claim that deliberation would have no point as the absurd consequence of
determinism.
17. There are ways to resist the indeterministic reading other than the one I have set
out here. Some have described the phenomenology of deliberation and decision in
a way that undermines the indeterministic reading (see, for example, Mele 1995,
133–136 and Strawson 1986, 115, note 30), and others have argued against the
128 D. K. Nelkin
18. See Honderich (1988) for the view that both compatibilists and incompatibilists
make the mistake of focusing on one family of important human attitudes to the
exclusion of another with which it is inconsistent. I cannot do justice to Honderich’s
rich discussion here. For our purposes here, it is worth noting that it is consistent
with there being a single concept of freedom, and even one to which all rational de-
liberators are in some way committed, that human beings often possess inconsistent
attitudes concerning particular conceptions of freedom.
21. In this connection, see Frank Jackson (1998), who writes: “I find compelling
Peter van Inwagen’s argument that . . . determinism is inconsistent with free will.
What compatibilist arguments show, or so it seems to me, is . . . that free action on
a conception near enough to the folk’s [i.e., common-sense] to be regarded as a
natural extension of it, and which does the theoretical job we folk give the concept
of free action in adjudicating questions of moral responsibility and punishment, and
in governing our attitudes to the actions of those around us, is compatible with
determinism” (44–45). Although the concept that I offer is neutral as between com-
patibilism and incompatibilism, the spirit of Jackson’s point applies to it, as well.
Even if one doubts that the concept I offer is the concept of freedom, one can still ac-
cept that it can do the theoretical work we want it to do, including supporting our
attributions of moral responsibility.
The Sense of Freedom 129
23. Bok has recently defended what might naturally be thought of as an explicitly
compatibilist reading of (R). In particular, she argues that being practical reasoners
gives us reason to regard ourselves as free in a compatibilist sense. We are free in the
relevant sense when we can determine our conduct through practical reasoning, and
we have genuine alternatives among which we can choose, where genuine alterna-
tives are those actions we would perform if we chose to perform them. See especially
Bok (1998, 118–119).
25. Frederick Adams develops a similar idea in Adams (1995, 552). Adams there
argues that all intentional action requires an attempt, and that trying to perform an
action requires the lack of a belief that success is impossible. Thus, it would seem to
follow that intentional action requires the lack of a belief that success is impossible.
Sometimes Adams also seems to endorse the stronger claim that intentional action
requires the belief in the possibility of success (see 553–554, for example). And the
stronger claim fits well with his reasoning that intentional action requires that one
have beliefs about how to achieve one’s end.
26. As Burge has pointed out, one might have “incomplete mastery” of a concept,
have false beliefs about even some of the essential properties of its instances, and yet
have genuine beliefs employing the concept nevertheless. For example, one might
believe that one suffers from arthritis, even if one believes it is not a disease of the
joints. (See Burge 1979.) However, it may be that there are certain true beliefs that
one cannot lack, and still be said to have the concept. The case at hand appears to be
of this kind: it is constitutive of having the concept of having reasons that one
believe one ought to act in certain ways if one has reasons to act.
130 D. K. Nelkin
27. This idea, too, recalls Kant: “That our reason has causality, or that we represent
it to ourselves as having causality, is evident from the imperatives which in all mat-
ters of conduct we impose as rules upon our active powers. ‘Ought’ expresses a kind
of necessity and connection with grounds which is found nowhere else in the whole
of nature . . .” (Kant 1781/1965, 472–473 [A457/B575]).
28. Along lines similar to the objection in the text, it might be argued that in ratio-
nally deliberating, one is sometimes guided only by the purpose of finding ‘the best
thing to do’. Yet, as I argued earlier, if one’s activity toward the goal of finding ‘the
best thing to do’ is to constitute genuine rational deliberation (as opposed to other
activities that might aim at that goal, such as making a sacrifice to the gods), one must
also be guided by the purpose of finding good reasons for acting.
29. For a small sampling of the literature on the connections among intentional
action, intention, and belief, see Davidson (1985) who argues that intentional action
requires the belief that one can succeed, Grice (1971), Harman (1976), and Velleman
(1985) who argue that having an intention requires the stronger belief (or accep-
tance, in the case of Grice) that one will succeed, and Bratman (1986) who argues
that (normally) having an intention and being rational requires that one not believe
that one will not succeed. Many have argued against one or more of these claims. For
example, McCann (1986) argues against all of these claims. Ludwig (1992) and
(1995) also argues against all of these claims, and goes one step further. He defends
the claim that one can be rational in both intending and acting intentionally even
though one believes that one cannot succeed. (I am grateful to Kirk Ludwig for email
correspondence about his view.)
32. Albritton (1985) argues that one can try to do what one believes is impossible,
but, like Ludwig, also recognizes the need to account for cases like the flying case. For
example, he agrees that there are certain things he cannot try to do, including trying
to jump over a building and even trying to do fifty push-ups. His explanation for this
fact is that “in his present cognitive position and state of mind that description of
him [as trying] would be inept whatever he did,” and “It’s that nothing I can think
of to do this evening would be rightly described as trying to jump over this building,
in a straightforward sense, unless, for example, my beliefs were to alter or go very
dim” (245). It seems to me that something important must be contained in the
phrases ‘present cognitive position’ and ‘my beliefs’. It is tempting to take them to
include the lack of belief that success is possible or the lack of belief that one has the
general ability to perform these kinds of actions.
33. One consequence of this reasoning is that there is a certain kind of irrationality
in simultaneously being a rational deliberator and a ‘practical reasons nihilist’ or
even a skeptic about practical reasons. Against this, it might be argued that a skeptic
The Sense of Freedom 131
about reasons could rationally deliberate (and be perfectly rational) simply by seeking
reasons if they happen to exist. In reply, as argued earlier, one could not deliber-
ate if one lacked the belief that one can succeed in at least some typical situations;
otherwise, one would lack a conception of what one was doing in deliberating. Since
deliberation requires such a conception, one must believe that one can sometimes
succeed in finding reasons for acting. Significantly, this leaves open the possibility of
a rational skepticism about the possibility of finding reasons on a particular occa-
sion. See also Burge (1998) for a different argument that finds a similar albeit more
general target in a broad reasons skepticism. Burge argues persuasively that for one
to fully understand reasons, “one must be susceptible to reasons” and “one must
recognize” the effect of reasons on one’s judgments and inferences (250).
34. The initial appeal of the examples also fades upon examination of the growing
body of work on infant and child development. Developmental psychologists are
divided on the question of exactly when many of the relevant concepts, such as ‘self’,
‘goal’, ‘desires’, and ‘desires as reasons for acting’ emerge in the human infant and
child. And while psychologists see the field as one that is need of a great deal more
research (see, for example, Wellman and Iagaki 1997, 2 and Meltzoff et al. 1999, 19),
there is an increasing consensus that such concepts appear much earlier in human
development than was previously thought. (See, for example, Wellman and Iagaki
1997 and the essays in part 2 of Malle, Moses, and Baldwin 2001. For example, it is
argued that by the age of 18 months, many children not only have the mental state
concepts of ‘desire’ and ‘intention’ but also understand that others have desires dif-
ferent from their own, and that others can intend to perform actions even though
they are prevented from performing them (Meltzoff et al. 1999). Thus, there is good
evidence that very young children have rich mental state concepts. Of course, it is
true that even if this evidence were conclusive, it does not provide a positive argu-
ment for the belief-concept reading of (R). At the same time, the available evidence
detracts considerably from the initial appeal of the alleged counterexamples.
35. The objection might be pressed in the following way: consider theoretical
deliberation. When engaged in it, we can suppose that we have a sense that we can
believe or judge for good reasons. But there is nothing like a sense of freedom
associated with theoretical deliberation; to the contrary, we do not choose our judg-
ments as we do many of our actions. So perhaps the sense of freedom as I have char-
acterized it is not really a sense of freedom either. This way of pressing the objection
raises a number of interesting issues regarding the relationship between rational
deliberation and its theoretical parallel. For the objection makes a number of presup-
positions including these: (1) that rational deliberation and its parallel in the realm
of judgment differ in that freedom and ‘chosen-ness’ are associated with the former
but not the latter, and (2) that the two forms of deliberation do not differ in any rel-
evant way in the implications of their respective requirements that agents represent
themselves as capable of responding well to good reasons. Both presuppositions
might be challenged, and I discuss both options in Nelkin (in preparation).
132 D. K. Nelkin
References
Adams, F. 1995. “Trying: You’ve Got to Believe.” Journal of Philosophical Research 20:
563–570.
Aristotle. 1984a. The Complete Works of Aristotle. J. Barnes, ed. Princeton, N.J.:
Princeton University Press.
Bok, H. 1998. Freedom and Responsibility. Princeton, N.J.: Princeton University Press.
Bratman, M. 1986. Intention, Plans, and Practical Reason. Cambridge, Mass.: Harvard
University Press.
Chisholm, R. 1964. “Human Freedom and the Self.” The Lindley Lecture. Lawrence,
Kans.: University of Kansas Department of Philosophy. Reprinted in Watson
(1982).
Clarke, R. 1992. “Deliberation and Beliefs About One’s Abilities.” Pacific Philosophical
Quarterly 73: 101–113.
. 1980. Essays on Actions and Events. New York: Oxford University Press.
. 1971. “Free Will and the Concept of a Person.” Reprinted in Frankfurt (1988).
Hill, T. 1985. “Kant’s Argument for the Rationality of Moral Conduct.” Pacific
Philosophical Quarterly 66: 3–23.
Kant, I. 1781/1965. The Critique of Pure Reason. N. Kemp Smith, trans. New York:
St. Martin’s Press.
Malle, B. and J. Knobe. 2001. “The Distinction Between Desire and Intention: A Folk-
Conceptual Analysis.” In B. Malle, L. Moses, and D. Baldwin, eds., Intentions and
Intentionality. Cambridge, Mass.: MIT Press.
Malle, B., L. Moses, and D. Baldwin, eds. 2001. Intentions and Intentionality.
Cambridge, Mass.: MIT Press.
Mele, A. 1995. Autonomous Agents: From Self Control to Autonomy. New York: Oxford
University Press.
134 D. K. Nelkin
Sorabji, R. 1980. Necessity, Cause, and Blame. Ithaca, N.Y.: Cornell University Press.
Van Inwagen, P. 1983. An Essay on Free Will. Oxford: Oxford University Press.
Velleman, J.D. 1989. “Epistemic Freedom.” Pacific Philosophical Quarterly 70: 73–97.
Watson, G., ed. 1982. Free Will. Oxford: Oxford University Press.
Wellman, H., and K. Inagaki, 1997. The Emergence of Core Domains of Thought: Chil-
dren’s Reasoning About Physical, Psychological, and Biological Phenomena. San Francisco,
Calif.: Jossey-Bass.
Ishtiyaque Haji
1 Introduction
is false.
136 I. Haji
Thesis PR*: Possibly (i.e., it can be the case that), for some agent, S, time, t,
and action, A, if S is morally responsible at t for A, then for some t* later
than t, S does A at t* and S could have avoided doing A at t*.
One might exploit the suggestion in this passage that legally mandated
punishment presupposes a sort of openness of the future to argue against
blame future and, hence, to generate support for blame past. Suppose that
although it is now true that, say, Bob will steal Ann’s car at noon, it is still
‘open’ to Bob, as of now, not to steal at noon. But then, it might be claimed,
Bob is not now deserving of blame for stealing later. For, as of now, there
is no relevant desert base. If Bob is blameworthy for doing something, he is
blameworthy partly in virtue of the fact that he freely does that thing.
But as of now, he has not committed any crime, and as of now, it is open
whether he will commit any crime. So he cannot, as of now, be blamewor-
thy for stealing Ann’s car at noon.
I believe it is some libertarian notion of openness that lies at the heart of
this entire line of reasoning. The notion requires elucidation before this
reasoning can be clarified and evaluated. The basic idea underlying liber-
tarian openness (hereafter ‘L-openness’) is that if it is L-open to an agent
to do something, A, at a time, t, then the agent can do something other
than A at t consistent with the past and the laws of nature remaining un-
changed. Somewhat more rigorously, it is L-open to S to do A at t if and
only if S is not causally determined to do A at t. S is not causally determined
to do A at t when S does A at t if and only if S does A at t, and there is a pos-
sible world that is (a) past-wise indiscernible (with respect to ‘genuine’ or
nonrelational facts) from the actual one until just prior to either the
moment at which S does A, or an earlier moment at which S does some-
thing else, B, such as forming a judgment to A or an intention to A; (b)
governed by the same laws of nature as the actual world, and (c) such that,
in it, S does other than A at t.2
Libertarian Openness, Blameworthiness, and Time 139
The case provides healthy support for the view that Bob is morally respon-
sible for deciding to steal the car (and, subsequently, for stealing the car)
even though he could not have decided (and then done) otherwise. Fur-
ther, the case highlights the distinction between indeterministic decision
making and L-openness. We can assume that it is causally undetermined
that certain of Bob’s nonoccurrent beliefs will enter into his deliberations
about whether to steal the car, and that if the beliefs do come to Bob’s
mind, he will decide (on his own) to steal the car.5 In this sort of case, in-
determinism is located relatively early in the causal pathway of an action at
some point prior to the making of a decision, and the event that is
140 I. Haji
even though it wasn’t settled then, that Rob would push the button at 11:45.
Call this case ‘bomb2’. Yet again it seems that the verdict that, as of 11:30
(t2), Rob is blameworthy for pushing the button at 11:45 (t3) is reasonable.
Not everyone will agree with this verdict. One might contest that, in
this context, ‘as good as sealed’ can only mean ‘nearly sealed’, and so ‘not
sealed’. As long as the action is L-open, it cannot be settled or truly
inevitable but only highly improbable, in which case the agent cannot be
responsible for it until it occurs. In this respect, the case differs from one
in which, say, should one now perform some action, it is inevitable that
an upshot of that action will occur.
In defense of my verdict, compare bomb2 with a modification of Mele
and Robb’s Frankfurt-type case. In the modified case, at t1, Black initiates a
deterministic process P in Rob*’s brain with the intention of thereby caus-
ing Rob* to decide at t2 to press the ‘kill now’ button at t3 on a remote de-
vice, one just like the device that Rob has. Pressing the button will detonate
the bomb in Sid*’s room. P, screened off from Rob*’s consciousness, will
deterministically culminate in Rob*’s deciding at t2 to press the ‘kill now’
button at t3 unless he decides on his own at t2 to press the button at t3 or
is decisionally incapacitated at t2. (Black is unaware that Rob* can decide
on his own at t2 to press the button at t3; he is confident that P will cause
Rob* to decide as he wants Rob* to decide.) It happens that Rob* decides on
his own at t2 to press the button at t3, on the basis of his own indetermin-
istic deliberation. Rob* is resolute about this decision, and once made, no-
body or nothing will intervene. But if he had not just then decided on his
own to press the button, P would have deterministically issued, at t2, in his
deciding to press it. P in no way influences the indeterministic decision-
making process that actually issues in Rob*’s decision.
Now let’s add one more layer of complexity to the case. Imagine that, at
t1, Black also initiated another deterministic process, Q , in Rob*’s brain
with the intention of thereby causing Rob* to press the ‘kill now’ button
at t3. Q, just like P, screened off from Rob*’s consciousness, will determin-
istically culminate in Rob*’s pressing the ‘kill now’ button at t3 unless he
presses it on his own at t3 or is physically incapacitated at t3. (Black is,
again, unaware that Rob* can press the button on his own at t3; he is con-
fident that Q will cause Rob* to act as he wants Rob* to act.) It turns out
that Rob* acts on his own at t3, pressing the button at t3 on the basis of his
own indeterministic intention-forming capacities. We can suppose that,
Libertarian Openness, Blameworthiness, and Time 143
I want to end with the suggestion that the controversy over which of blame
future or blame past is true can be explained partly by endorsement, at
some level, of different conceptions of moral responsibility. Consider, for
example, R. J. Wallace’s Strawsonian account of responsibility (1994, 91).
On Wallace’s view, a person is morally responsible for some action if and
only if it would be appropriate to hold her responsible for that action. As
moral norms of fairness, Wallace contends, set the standards of appropri-
ateness for responsibility ascriptions, this view is to be construed as the
normative one that a person is morally responsible for an action if and only
if it is fair to hold her responsible for that action. Wallace proposes that to
hold someone morally responsible is essentially to be subject to the moral
sentiments of guilt, indignation, or resentment in one’s interactions with
the person. In cases in which the standard excuses such as coercion, phys-
ical constraint, duress, mistake, accident, and inadvertence apply, Wallace
contends that the agent has not violated a moral obligation, and conse-
quently, does not deserve to be subjected to the reactive emotions and the
forms of sanctioning treatment that express them. With the exemptions—
psychopathy, behavior control, stress, insanity, addiction, childhood, hyp-
notism, deprivation, or torture—Wallace submits that the agent has been
deprived of the ability to grasp and apply moral reasons, and the ability to
control her behavior by such reasons. But if holding an agent responsible
involves judging it appropriate to sanction her morally, and to sanction her
Libertarian Openness, Blameworthiness, and Time 145
as a result of the violation of moral obligations for which there are moral
reasons, then it will, once again, be unfair to do this if she lacks the ability
to apply the moral reasons that sustain the obligations violated.
Of pertinence to our present concerns is that, if one subscribes to such a
normative conception of responsibility, then it is quite natural to affirm
blame past. For just as it may be unfair to blame a person for having done
something that isn’t wrong in the first place or (as with prepunishment) for
not yet having committed a crime, so, it may reasonably be contended, it
would be unfair to hold a person blameworthy now for an action that she
has yet to perform. As we rarely know precisely what the future will bring,
there is almost always the chance that the person won’t perform the perti-
nent action. It would, hence, seem unfair to react to this person, in advance
of her performing this action, with resentment or indignation.
John Fischer (1994), in an elegant paper, has recently proposed that the
intuitive picture or conception of responsibility that fuels the view that
alternative possibilities are required for moral responsibility is the concep-
tion that moral responsibility requires the ability to make a sort of difference
to the world. Fischer explains that if one makes a difference in this sense, one
selects which path the world will take among various paths that are gen-
uinely accessible. One’s selection determines which way the world unfolds
and one thereby makes a crucial difference. Again, if one is driven by this
conception of responsibility, then one will be attracted to blame past. A pro-
ponent of this conception of responsibility might sensibly propose that as a
prospective action has still to be performed, its agent has not yet made a dif-
ference regarding the pathway the world may take by selecting that action.
Hence, the agent cannot, prior to the fact, be responsible for that action.
Contrast these two conceptions of responsibility with a third according
to which, roughly, when a person is responsible for what she does, she
discloses what she morally stands for in performing that action. This sort
of self-disclosure view has been proposed, among others, by John Dewey
(1957, 160–161) who wrote:
[W]hen any result has been foreseen and adopted as foreseen [by the agent], such
result is the outcome not of any external circumstances, not of mere desires and
impulses, but of the agent’s conception of his own end. Now because the result thus
flows from the agent’s own conception of an end, he feels himself responsible for it
. . . . The result is simply an expression of himself; a manifestation of what he would
have himself to be. Responsibility is thus one aspect of the identity of character and
146 I. Haji
conduct. We are responsible for our conduct because that conduct is ourselves objectified in
actions.8
Needless to say, this is not the place to decide among the various possible
conceptions of moral responsibility. My proposal is simply that if one finds
the self-disclosure picture of responsibility promising, one may well be
willing to renounce blame past in favor of blame future.
In conclusion, focusing on central cases in which persons are morally to
blame for their actions, a number of people have contended that blame-
worthiness, and responsibility in general, are essentially retrospective. I
have proposed that this view is called into question by Frankfurt-type
examples.9 Puzzling over why so many have been attracted to this mis-
taken view, I have further proposed that proponents of this view may be in
the grips of a certain conception of responsibility, like Wallace’s normative
conception, or the make-a-difference conception, that renders dim the pos-
sibility of prospective blameworthiness. Finally, I have suggested that there
is an alternative, intuitive conception of responsibility—the self-disclosure
conception—which lends support to the contrasting view that persons can
be prospectively responsible for some of their actions.
Acknowledgments
Notes
1. A number of people have endorsed this view or at least something very similar to
it. For example, Wojciech Sadurski (1985, 117) says “desert considerations are always
past oriented. When talking about desert, we are evaluating certain actions which
have already happened.” Brian Barry (1965, 111) says, “Desert looks to the past—or
at most to the present—whereas incentive and deterrence are forward looking no-
tions . . . .” David Miller (1976, 93) claims that “Desert judgments are justified on the
basis of past and present facts about individuals, never on the basis of states of affairs
to be created in the future.”
2. Conditions (a), (b), and (c) are advanced and explained by Peter van Inwagen
(1983, 136).
4. Derk Pereboom (2000, 119–137) has advanced another sort of Frankfurt-type case
involving indeterministic decision making. In his case, the counterfactual intervener
148 I. Haji
makes the agent (in a nondeterministic world) decide to cheat on his tax deduction
if he doesn’t (nondeterministically) do so on his own. The agent’s psychology is such
that a causally necessary (but not causally sufficient) condition for the agent to
decide not to cheat is that a moral reason occurs to him with a certain force. Such a
reason’s occurring to the agent is the cue for intervention.
6. For a defense of this view see, for example, Carl Ginet (1983), John Martin Fischer
(1994) and Peter van Inwagen (1983).
7. What of the following case though? In bomb4, it is open to Job, as of t0, to decide
at t1 to press the ‘kill now’ button at t6 and to press the ‘kill now’ button at t6. If the
button is pressed at t6, Cid will be blown to bits. However, unlike Rob and Rob*, Job
is unsure about whether to kill Cid. Let’s suppose that he first decides at t1, on the
basis of his own indeterministic deliberation, to press the button at t6. A little while
later, he changes his mind. At t3, he decides, again on the basis of indeterministic de-
liberation, not to press the button at t6. But he is still not settled in his mind about
what to do. Just before t6, he decides on his own once again to press the button, and
at t6 he presses the button. Is Job at, say, t2, blameworthy for pressing the button at
t6? I confess to being unsure about this case. I’m inclined to say that he isn’t blame-
worthy at t2. If this verdict is correct, then there is a kernel of truth in the idea that
L-openness is incompatible with blameworthiness for future actions. It’s this: both
Rob and Rob* are resolute about their decision to kill. They successfully execute
every element in their plan to kill without indecision or unsettledness of mind. It is
as if, from their perspective, L-openness does not matter; from their perspective,
every future includes the decision to kill and the subsequent killing. Similar things
are not true of Job.
9. It appears that both Fred Feldman and Michael Zimmerman would agree with the
conclusion that blame past is false. Feldman (1997) rejects the following view, DT: If
at t, S deserves x in virtue of the fact that S did or suffered something at t’, then t’ can-
not be later than t, primarily by appealing to putative counterexamples. He says that
several people, including John Cottingham, have accepted DT or something very
similar. Feldman claims that Cottingham seems to be asserting, among other things,
that I cannot blame you for doing something unless you have already done it. How-
ever, Feldman claims: “If, as a result of your failure to drain the pipes, there is going
to be damage to the plumbing later tonight when the temperature drops, I can blame
you for the damage that is going to occur” (1997, 180–181). This example, though
suggestive, falls short of establishing prospective blameworthiness both for future
actions with respect to which one is directly free, and for noninevitable upshots of
one’s present actions. Michael Zimmerman (1988) asks: if S wills (that is, roughly, de-
cides) at t that e should occur, at what time is it true to the facts that S is blameworthy
Libertarian Openness, Blameworthiness, and Time 149
for willing e at t? The answer, he says, is: at that time at which it is true that (to speak
tenselessly) S wills e at t. Zimmerman adds that if S wills e at t and is culpable for
this, then S is culpable at t and forever thereafter. In addition, Zimmerman says that
“if there can be true propositions about the free volitions and actions of persons in
the future, then we should also admit that persons are forever (period) (that is, not
just forever after, but also forever before) culpable for those volitions for which they
are culpable.” (1988, 2, 40). Zimmerman seems to be suggesting that responsibility is
not essentially retrospective. Even on this tenseless approach, one might still want
an explanation of what conception of responsibility, if any, can accommodate
prospective blameworthiness.
References
Barry, B. 1965. Political Argument. London and New York: Routledge and Kegan Paul.
Dewey, J. 1957 (1891). Outlines of a Critical Theory of Ethics. New York: Hillary House.
Haji, I. 1998. Moral Appraisability: Puzzles, Proposals, and Perplexities. New York:
Oxford University Press.
Mele, A. 1995. Autonomous Agents: From Self-control to Autonomy. New York: Oxford
University Press.
Rachels, J. 1991. “What People Deserve.” In John Arthur and William H. Shaw, eds.,
Justice and Economic Distribution. Englewood Cliffs, N.J.: Prentice-Hall.
150 I. Haji
Sadurski, W. 1985. Giving Desert Its Due: Social Justice and Legal Theory. Dordrecht:
Reidel.
van Inwagen, P. 1983. An Essay on Free Will. Oxford: Oxford University Press.
Wallace, R. 1994. Responsibility and the Moral Sentiments. Cambridge, Mass.: Harvard
University Press.
Todd R. Long
examples and threatens a certain appealing intuition that some people have
about the relation between moral responsibility and external manipulation.
In Responsibility and Control, F & R argue that, in the free will debate, guid-
ance control is the freedom-relevant condition that is sufficient for moral
responsibility.2 Guidance control does not require the sort of regulative con-
trol that depends upon the existence of genuine alternative possibilities.
Although regulative control often goes hand in hand with guidance control,
F & R hold that moral responsibility requires only guidance control. Thus, we
are morally responsible for our actions so long as we have guidance control,
even if we lack regulative control. Pointing out that guidance control is the
sort of control we find illustrated in the Frankfurt-type actual-sequence ex-
amples, F & R conclude that guidance control is what ultimately grounds our
ascriptions of moral responsibility. This is a familiar compatibilist claim.
What sets F & R’s theory apart is that F & R ascribe guidance control to an
agent insofar as that agent’s deliberative mechanism is appropriately respon-
sive to reasons. Hence, their theory hinges on a satisfactory account of the de-
liberative mechanism operant in agents. I will return to mechanisms shortly.
Another significant feature of F & R’s notion of guidance control is their
distinction between moral responsibility, on the one hand, and praisewor-
thiness or blameworthiness, on the other. F & R characterize their notion of
moral responsibility as a “Strawsonian view” that “allows for moral re-
sponsibility for ‘morally neutral’ behavior. For instance, one can be morally
responsible for simply raising one’s hand” (Fischer and Ravizza 1998, 8). All
that is required for moral responsibility is that one be “an appropriate
candidate for the reactive attitudes” (Fischer and Ravizza 1998, 7). So, one
can be morally responsible for an action without being an actual target of
a reactive attitude. Nevertheless, moral responsibility is connected to praise
and blame; after all, one would not even be a candidate for praise or blame
unless one were morally responsible, in F & R’s sense, for one’s actions. We
can summarize the key point here by saying that although guidance con-
trol (and, thus, moral responsibility) does not necessarily imply praisewor-
thiness or blameworthiness, the latter do imply the former.3
F & R refine their notion of guidance control by connecting it to the notion
of reasons-responsiveness.4 Their idea is that we are morally responsible for an
Moderate Reasons-responsiveness 153
F & R hold that MRR is sufficient for moral responsibility.7 Claims of suffi-
ciency are customarily tested with specific cases. Any case in which an
agent’s mechanism is MRR but the agent is not morally responsible for an ac-
tion issued in by that mechanism will count as a counterexample to F & R’s
claim of sufficiency. I will provide test cases in section 3. But first a word
about mechanisms.
Here is the crucial idea that F & R want us to get: any MRR mechanism that
operates to issue in my actions is always a different mechanism from any
one that would operate if my brain were monkeyed with to the point that I
am no longer morally responsible for my actions. Hence, a brain-implanted
mechanism that issues in an action would always constitute a different
mechanism from an ordinary practical reasoning mechanism that issues in
an action.
Moderate Reasons-responsiveness 155
So far, so good. However, F & R also think that there is a close connection
between the distinction we’ve just noted and Frankfurt-style examples.
Specifically, they hold that the mechanism operant in an actual-sequence
case is always to be distinguished from the mechanism operant in an
alternative-sequence case. Consider Fischer’s discussion of such cases in
The Metaphysics of Free Will:
In a Frankfurt-type case the kind of mechanism that actually operates is reasons-
responsive, although the kind of mechanism that would operate in the alternative
scenario is not. In the Frankfurt-type case discussed above (in which Jones votes for
Clinton on his own and Black does not actually intercede), Jones’ action issues from
the normal faculty of practical reasoning, which we can reasonably take to be
reasons-responsive. But in the alternative scenario, a different kind of mechanism
would have operated—one involving direct electronic stimulation of Jones’ brain.
And this mechanism is not reasons-responsive. Thus, the actual-sequence mecha-
nism can be reasons-responsive, even though the agent is not reasons-responsive.
(He couldn’t have done otherwise.) . . . Clearly, . . . it is crucial to distinguish between
the kind of mechanism that operates in the actual sequence and the kind of mecha-
nism that operates in the alternative sequence (or sequences).8 (Fischer 1995, 163)
I want to make two points about Fischer’s discussion here. First, let’s consider
a couple of features of this alternative-sequence case: (1) There is brain ma-
nipulation that leads to a particular outcome; and (2) The agent would have
done otherwise had there been no manipulation. Now, one might think that
what underlies our intuition that the agent lacks moral responsibility in
such a case is some kind of manipulation principle that addresses those
features. As we have seen, F & R hold that a brain-implanted mechanism
issuing in an action is a different mechanism from an ordinary practical
reasoning mechanism. F & R say that this is just intuitively the case, but per-
haps we can explain this intuition by drafting a manipulation principle that
addresses the two factors we just noted about alternative-sequence cases:
MAN: If a person S performs an action A at time tn only because another
agent B, acting completely unbeknownst to S, has directly manipulated
S’s brain earlier at t1 then S is not morally responsible for performing A.9
We might think that F & R would happily endorse such a principle, for it
helps to explain the intuition that a person is not responsible for an action
if her brain has been directly manipulated in a certain way. More specifically,
it helps to explain the appealing intuitive view that a person is not morally
responsible for performing an action if the action is such that he would have
refrained from performing it if his brain had not been directly manipulated
156 T. R. Long
by an external source. This is, after all, true of Fischer’s Frankfurt-style case,
wherein the agent is morally responsible in the actual sequence but is not
morally responsible in the alternative sequence; and MAN would explain
why Fischer concludes the quote above by saying that it is crucial to distin-
guish the mechanism operant in the actual-sequence case from the mecha-
nism operant in the alternative-sequence case.
This leads me to the second point I want to make. Clearly, it is crucial to
make this distinction only if it is true that the mechanism operant in the
alternative-sequence case is always a different mechanism from the one
operant in the actual-sequence case. But since F & R give us so little to go
on concerning what a relevant mechanism is, or how to differentiate mech-
anisms, I see no reason to think that the mechanism that operates in
the actual sequence of a Frankfurt-type example must be different from the
mechanism that operates in an alternative sequence of a Frankfurt-type
example. As I will show in section 3, we can tell a Frankfurt-style alternative-
sequence story in which the mechanism issuing in the action is the very
same mechanism as the one operating in the actual-sequence case. The
basic idea that I will develop there is that the results yielded by the human
deliberative mechanism will vary depending on its inputs. It is very plausi-
ble to think that differing inputs (e.g., reasons) will yield differing outputs
(e.g., actions). And in their latest account, F & R grant that their concept of
the relevant deliberative mechanism in a given case involves the idea that
the same mechanism may operate on a variety of inputs:
. . . a mechanism’s reacting differently to a sufficient reason to do otherwise in some
other possible world shows that the same kind of mechanism can react differently to
the actual reason to do otherwise. . . . The picture here is of one kind of mechanism
with different “inputs.” (Fischer and Ravizza 1998, 73–74)
is not a good reason to accept his claim. After all, a light can be turned off by
a human hand’s flicking a light switch, and the same light can be turned off
by falling plaster that strikes a light switch. Here we have two different causes
for the light’s going out, but we have one and the same mechanism (the light
switch). In Fischer and Ravizza (1998), F & R provide no additional reasons
for thinking that the two mechanisms cannot be the same; rather, they just
appeal to intuition to have us think that they must be different.10 I contend
that, given the dearth of explanation F & R provide about mechanisms, it
can be just as intuitive to think that some actual-sequence and alternative-
sequence scenarios involve the same, not different, mechanisms.
Hitler. From reliable sources, he hears that Hitler has intentions of building
up a master race of pure-blooded Aryans, driving non-Aryans out of the
country, and perhaps killing those who do not leave. Schmidt develops deep
doubts about Hitler’s moral character.
Schmidt’s role in the decision is to cast the deciding vote, should the
voting members reach a deadlock. The members do reach a deadlock, so
Schmidt is called upon to cast the deciding vote. Schmidt deliberates.
Waiting in the wings is Block, a malevolent being with extraordinary causal
and predictive powers. Should it become clear to him that Schmidt is going
to vote against Hitler, Block will take effective steps to ensure that Schmidt
votes for Hitler. Otherwise, Block will do nothing. So, whatever Schmidt’s
initial preferences and inclinations, Block will have his way.12
We can divide the basic possibilities here into two: Schmidt may deliber-
ate and vote for Hitler. Although this would be a surprising outcome given
Schmidt’s moral character and his serious misgivings about Hitler, this
would be a possible case under F & R’s account of MRR; for F & R want their
theory of moral responsibility to be compatible with cases of ‘weakness of
will’.13 Note that in this case, which I will deem the Block/Schmidt actual
sequence, Schmidt is both morally responsible and blameworthy.
The other basic possibility is that Schmidt is such that, were Block not to
intervene, he would vote against Hitler.14 Now let us consider what course of
action Block will take, if he believes that Schmidt will vote against Hitler, to
ensure that Schmidt acts as Block wants. There may be a multitude of actions
that Block could take; however, as far as I can tell, there are only three types
of action that Block could take. I will apply each of these three possible types
in the following three cases respectively, each of which will be an example of
what I will call an alternative sequence.
Block/Schmidt Case 1: A minute before Schmidt casts his vote, Block
becomes convinced that Schmidt is going to vote against Hitler.15 Block
goes into action: rather than monkeying with Schmidt’s normal delibera-
tive voting mechanism, Block bypasses it. Block manipulates Schmidt in
some way such that Schmidt’s voting for Hitler is not related in any way to
the reasons that Schmidt has for or against voting for Hitler. We might
imagine him doing something directly to Schmidt’s nervous system that
physically causes him to vote for Hitler.
This is, of course, the kind of action that F & R think would occur should
Block believe that he must act; for, in this alternative-sequence case, the
Moderate Reasons-responsiveness 159
mechanism that issues in Schmidt’s voting for Hitler (i.e., the one that Block
uses) is a different mechanism than the one that would issue in Schmidt’s
voting against Hitler. In the actual sequence, Schmidt’s action is issued in by
a deliberative voting mechanism that is MRR. But in this alternative se-
quence, Schmidt’s action is ushered in by a different mechanism, a mecha-
nism that has nothing to do with reasons-responsiveness. As F & R say, we
would hold Schmidt neither morally responsible nor blameworthy for his
voting for Hitler in this case. Furthermore, note that F & R are committed to
thinking that this is the only kind of action that could occur if Block should
intervene; for, as we saw earlier, F & R’s view is that we cannot hold Schmidt
morally responsible in this case precisely because the mechanism that
operates is not the same as the one at play in the actual sequence (which
Schmidt has taken responsibility for).16 However, I see no reason to think
that this type of action is the only possible way for Block to accomplish
his work. Cases 2 and 3 demonstrate other types of action that could get the
job done.
Block/Schmidt case 2: A minute before Schmidt casts his vote, Block be-
comes convinced that Schmidt is going to vote against Hitler. Block goes into
action: he adds new inputs into the very same mechanism that is operant in
the actual sequence. Suppose that these new inputs come in the form of
reasons for voting for Hitler. Block directly feeds into Schmidt’s delibera-
tive voting mechanism enough reasons, or reasons powerful enough, to
ensure that Schmidt will vote for Hitler. Because Block has extraordinary
predictive powers, he knows which reasons will ensure that Schmidt’s
deliberative voting mechanism yields a vote for Hitler, and Block provides
those reasons as inputs for Schmidt’s deliberative voting mechanism.17
I have two things to discuss about this case. First, note that F & R do not
consider this case as a possible way for Block to accomplish his desire. Their
emphatic claim, that whenever there is intervention (in an alternative
sequence) the mechanism is a different one from the mechanism that
would have been employed had there been no intervention, commits them
to thinking that this case is impossible. However, I do not see that this case
is any less possible than the one that they prefer. Second, this situation
prompts the key question: is Schmidt morally responsible for voting
for Hitler? According to F & R’s account of MRR, the answer is “yes,” for
Schmidt deliberates appropriately in this alternative sequence case; that is,
as he comes to make his decision, Schmidt’s deliberative mechanism is
160 T. R. Long
from the early days when Hitler seemed to be a thoroughly good chap (per-
haps Block would also leave in Schmidt’s brain morally neutral memories
involving Schmidt’s associating with Hitler during the past few years).
Thus, the only reasons that Schmidt has to go on as he deliberates are
reasons to vote for Hitler.
Is Schmidt morally responsible for voting for Hitler in this case? Accord-
ing to F & R’s account of MRR, the answer is “yes,” for in coming to make
his decision, Schmidt’s mechanism meets all the conditions of MRR. More-
over, the very same deliberative voting mechanism is operant in this case
as it is in the actual sequence. That mechanism can yield varying outcomes
in response to different reasons. Schmidt exhibits guidance control in that
the mechanism that actually issues in his action is MRR. So, F & R are com-
mitted to thinking that Schmidt is morally responsible, despite Block’s
intervention.
Nevertheless, it may seem that Schmidt is not morally responsible in this
case. Although it is true that Schmidt’s mechanism is MRR, it seems plausi-
ble to suppose that Block has deprived Schmidt of some of his own reasons
in such a way that he is not morally responsible for voting for Hitler.21
Moreover, note that if we are inclined to think that Schmidt is not morally
responsible in this case, it is because we think that MAN (or something very
much like it) is true. On MAN, Schmidt is not morally responsible, since he
was not aware of Block’s manipulation a moment before Schmidt acted,
and he wouldn’t have acted as he did if Block hadn’t intervened.
What resources do F & R have to deal with these last two test cases? The
answer will depend on whether they concede them to be counterexamples.
If they do, then I doubt that they can fix their theory without denying that
MRR is sufficient for moral responsibility. Here’s why: the only way around
my counterexamples seems to involve construing the relevant deliberative
mechanism broadly enough that it includes the inputs (i.e., the reasons) as
part of the mechanism. On this construal, F & R would be entitled to con-
clude that Schmidt is not morally responsible in cases 2 and 3, because
Schmidt has not taken responsibility for a mechanism that includes inputs
that Block has either directly added or excised. This reply may seem
promising, but it turns out to be deadly to their theory; for F & R want a
162 T. R. Long
an action that she wouldn’t have performed had there been no manipula-
tion, then she is not morally responsible for her action. Despite the fact
that MAN may seem intuitively right, further reflection reveals that there is
little to recommend it. Consider Block/Schmidt cases 2 and 3. In each case,
not only did Schmidt meet the conditions of MRR, but he also intuitively
did the morally responsible thing, given the evidence he had to go on. And
if we knew the relevant facts about each case, we might even praise him for
his action, even as we deeply regret or deplore the trick Block played on
him and the consequences that followed. This shows that MAN is false, for
Schmidt did perform an action at time tn only because another agent, act-
ing completely unbeknownst to Schmidt, directly manipulated Schmidt’s
brain earlier at t1; and yet Schmidt is morally responsible for voting for
Hitler, since he acted in consonance with his reasons in an appropriate
way. The moral is that we are morally responsible for our actions when they
are consonant with our reasons in an appropriate way (perhaps, when MRR
is satisfied), but whether we are praiseworthy or blameworthy for our
actions depends on additional facts about the cases. Counterexamples to
MAN are plentiful. Although the intuition behind MAN may be appealing,
it does not survive scrutiny.
After doing this ground clearing, we are in a position to see that F & R’s
theory does get the cases right. It implies that when the deliberative process
as a whole is directly manipulated the agent is not morally responsible for ac-
tions issued in by that process. This is exactly right. The problem is in F & R’s
assumption about how their theory applies to Frankfurt-style examples; for
it does not, nor need not, imply that all alternative-sequence cases eliminate
moral responsibility on the part of the agent.
stands a chance only if we have reason to doubt that Schmidt meets the third
condition of MRR. However, I think it is clear that Schmidt can easily meet
that condition. In my examples I suppose that Schmidt (a) sees himself as the
source of his behavior, which follows from the operation of the relevant
mechanism, that Schmidt (b) believes that he is an apt candidate for the
reactive attitudes (in the manner F & R specify), and that Schmidt (c) views
himself as an agent based on his evidence for (a) and (b). Nothing about
Block’s manipulation prevents Schmidt from meeting those conditions. To
think that Schmidt fails to satisfy those conditions is to suppose that
Schmidt would have to be responsible both for his prior evidence (or beliefs)
and for the way in which he acquired that evidence (or those beliefs). But,
surely this is to add too much to the notion of taking responsibility. Evidence
(and reasons) supporting beliefs come to us from various directions and in
many different ways. Some of it we are quite aware of, and some of it we are
less aware of. Earlier I mentioned that nobody would balk at the suggestion
that Schmidt could have been convinced to vote for Hitler by more ordinary
external means (news reports, personal testimony, etc.). What is the relevant
difference between manipulating the inputs in those ways and manipulat-
ing the inputs directly?
Consider a case in which a group of conniving folk manipulate the inputs
of someone’s deliberative mechanism by affecting the information that per-
son gets. Suppose a group of Young Republicans who work at a news press
manipulate the inputs that Karen’s deliberative voting mechanism receives
by changing the text of Karen’s newspaper every time something good is said
in print about a Democratic candidate. Would this prevent Karen from being
morally responsible for her voting? If you are inclined to answer “yes,” then
you owe us an explanation of why any of us are morally responsible for our
acts of voting. After all, very few of us control the political reports or the po-
litical opinions that are propounded in our newspapers and other media;
thus, most of us do not control the inputs our deliberative voting mecha-
nisms receive. You can bet your bottom dollar that much of the political talk
in newspapers is written with the intent of manipulating us to vote the way
the authors want us to vote. On what basis should we judge ourselves as
morally responsible for our voting but deny that Karen is morally responsible
for her voting? My point is even more convincing when we turn our attention
to advertising. The very livelihoods of advertisers depend on their ability to
manipulate us into buying whatever marketers are selling. Advertising is a
166 T. R. Long
sneaky, subtle business, but are we to suppose that most of us are not morally
responsible for our purchases? Despite the cunning powers of advertisers,
surely most of us are not helpless victims of manipulation. Likewise for
Schmidt in cases 2 and 3; he was no helpless victim. No one caused him to act
by bypassing his deliberative voting mechanism. Schmidt actually weighed
reasons in a regularly reasons-responsive way, and his deliberative voting
mechanism was weakly reactive to reasons; moreover, that mechanism,
which we may assume Schmidt used many times before, was one that he had
clearly taken responsibility for. So, according to MRR, Schmidt is morally
responsible in cases 2 and 3.
F & R want their theory to capture our intuitions about moral responsi-
bility. I grant that judging Schmidt to be morally responsible strains the
prereflective intuition that some people might have about the cases. How-
ever, it is fairly common for us to find that our prereflective intuitions
about a matter are wrong, after reflection on the details; and, I have found
that, after reflection, many people (philosophers and nonphilosophers)
judge the cases as I do. Even so, I am not inclined to hang my case on any
claim about how many people have a particular intuition. Instead, I put
forward the following evidence for your consideration: (1) F & R want vari-
ant mechanisms; that is, they want any MRR mechanism to be able to
operate on a variety of inputs and to produce a variety of outputs; and (2)
F & R explicitly distinguish between moral responsibility, on the one hand,
and praiseworthiness and blameworthiness, on the other. Consideration of
(1) reveals that it will not do to build all the inputs into an MRR mecha-
nism, but the only way to make sense of Fischer’s response to my examples
is to think that Schmidt would have to be responsible for his input acquisi-
tion as well as for the mechanism itself. This amounts to building the
inputs into the mechanism. Consideration of (2) reveals that F & R have
the resources for their theory to get the cases right. Schmidt is morally
responsible for voting for Hitler, although he is not blameworthy. If you are
still unconvinced, suppose that Schmidt walked into the voting room and
murdered the officials who voted against Hitler. I think that we would all
judge him to be morally responsible for the murders, even if we knew every
detail about Block’s manipulation. But, then, what principled reason is
there to deny that Schmidt is morally responsible for his voting?
In summary, given that F & R explicitly distinguish between any MRR
mechanism and its inputs, given that they explicitly distinguish between
Moderate Reasons-responsiveness 167
6 Conclusion
Acknowledgments
Special thanks to Rich Feldman for many fruitful discussions about this
paper. Thanks also to John Martin Fischer, Tom Senor, Michael McKenna,
Joseph Keim Campbell, Allen Plug, Greg Wheeler, Tony Roark, Ken Himma,
and an anonymous referee for this volume, as well as participants in the
Rochester Philosophy Society (2000), the 52nd Northwest Conference on
Philosophy (2000), the American Catholic Philosophical Association,
Western New York (2000), the Society of Christian Philosophers, Eastern
Region (2001), the Philosophy Writing Seminar at the University of
Rochester (2001), and the Inland Northwest Philosophy Conference on
Freedom and Determinism (2001), for helpful comments on this paper.
Notes
1. For some of the works that have influenced my thinking, see Della Rocca (1998),
Mele and Robb (1998), Widerker (1995a, 1995b, 2000), and Vihvelin (1998).
2. See Fischer and Ravizza (1998, chaps. 1–2). See also Fischer (1995, chaps. 7–8).
3. I thank Michael McKenna and Joseph Keim Campbell for bringing to my atten-
tion that F & R also mention an epistemic condition as necessary for moral respon-
sibility. Though providing no analysis of the condition, F & R say that it is meant to
capture the “Aristotelian intuition” that “in order to be praiseworthy or blamewor-
thy, a person must know (or be reasonably expected to know) what he is doing”
(Fischer and Ravizza 1998, 12). To use their example, if you were to run over a kitten
hiding under your car but were ignorant of the kitten’s location, then, intuitively,
you would not be morally responsible for its death. After noting this epistemic
condition, F & R don’t mention it again but rather assume that it is met both in the
examples that they discuss and in their claim that guidance control is sufficient
for moral responsibility. In my cases 2–3 from section 3 below, I also assume that
F & R’s epistemic condition is met. It is possible that some epistemic condition
would make trouble for my test cases, but I find nothing in F & R’s discussion of their
epistemic condition that implies that there is trouble for my cases; moreover, I doubt
that there is a plausible, fully worked-out epistemic condition that would make
trouble for my test cases.
5. See Fischer and Ravizza (1998, chap. 3). I will use MRR to stand for both ‘moder-
ate reasons-responsiveness’ and its cognate form, ‘moderately reasons-responsive’.
Moderate Reasons-responsiveness 169
6. Although this schema does not appear in Fischer and Ravizza (1998), I have done
my best to reproduce faithfully the conditions of MRR as F & R present them. The
schema contains quotations or near quotations of various claims made by F & R that
are meant to explain what they summarize as follows: moral responsibility “requires
that an agent act on a mechanism that is regularly receptive to reasons, some of which
are moral reasons, and at least weakly reactive to reasons” (Fischer and Ravizza 1998,
82), and also that the agent takes responsibility for the mechanism (Fischer and Ravizza
1998, 242–243). Strictly speaking, on F & R’s theory, condition (3) is not a condition
of moderate reasons-responsiveness but is rather an additional necessary condition
for guidance control. However, for ease of presentation I speak of MRR as encompass-
ing all three conditions. This does not affect F & R’s theory in any relevant way.
7. F & R make this clear in numerous passages in Fischer and Ravizza (1998). Here is
a typical one: “Return to our account of moral responsibility for actions. On this
account, an agent is morally responsible for an action insofar as he has guidance
control of it, where guidance control consists in the action’s issuing from the agent’s
own, moderately reasons-responsive mechanism” (Fischer and Ravizza 1998, 230).
As I mention in note 4, F & R make such claims on the assumption that their
epistemic condition is met.
8. I use Fischer’s earlier account here because of the detail. Here is the parallel
account by F & R: “. . . in a Frankfurt-type case . . . the kind of mechanism that actually
operates is reasons-responsive, even though the kind of mechanism that would
operate—that is, that does operate in the alternative scenario—is not reasons-
responsive. . . . But in the alternative scenario, a different kind of mechanism would
have operated—one involving direct electronic stimulation of [the agent’s] brain. . . .
And this alternative sequence mechanism is not reasons-responsive” (Fischer and
Ravizza 1998, 38–39).
9. I thank Tom Senor for calling to my attention, though not necessarily endorsing,
a version of MAN.
10. F & R repeatedly say the following sort of thing: “the actual-sequence mecha-
nism is, intuitively, the agent’s own, whereas the alternative-scenario mechanism is
not” (Fischer and Ravizza 1998, 228).
11. This story is fictional, but it is supposed to express the character of the historical
Hitler.
12. Readers may note that Schmidt and Block are counterparts to Frankfurt’s Jones
and Black, respectively. See Frankfurt (1997, 162).
13. Condition (2) of MRR allows for cases of ‘weakness of will’. See Fischer and
Ravizza (1998, 42–44). In Fischer (1995), Fischer also wanted his theory to be com-
patible with cases of ‘actual irrationality’ (166–168), but it is unclear whether F & R
maintain this claim in Fischer and Ravizza (1998).
170 T. R. Long
14. I put this as I do for two reasons (and I thank Michael McKenna for bringing this
issue to my attention). First, I want to avoid the kind of criticism that David
Widerker and others have brought against designing Frankfurt-type examples that
both grant libertarian-free action and rule out alternative possibilities. Such exam-
ples would have it that the agent at least starts to choose before the intervenor inter-
venes. The concern is that if some kind of nondetermined act is allowed in the setup
of a Frankfurt-type example, then there is no reason to think that the agent
(Schmidt, in my cases) cannot do otherwise, regardless of the actions of the inter-
venor (Block, in my cases). So, as I characterize my cases, I don’t assume libertarian-
free action on the part of Schmidt. This leads to my second point: F & R do not
assume libertarian-free action on the part of agents in Frankfurt-type examples
either. Their view is that such examples show that alternative possibilities are not
required for moral responsibility even if determinism is true. My goal is to follow their
lead to see where it gets us. If it turns out that there is some general problem along
these lines, then my cases might be problematic as well. However, I think I avoid
problems in this neighborhood by setting up the cases as I do; for, all I need is for it
to be the case that every possible world in which Schmidt does not do what Block
wants is a world in which Block intervenes. I need not tell a story about how Block
does this. If someone demanded such a story, then it seems to me that luck on
Block’s part could account for it.
15. Although in all three of my alternative sequence stories Block goes into action
one minute before he thinks Schmidt will cast his vote, nothing significant turns on
Block’s timing.
16. F & R also consider a case in which scientists don’t directly cause the implantee
to perform an action but rather the scientists implant a mechanism that is itself
MRR, which issues in an action. Although this example is closer to my cases 2 and 3,
it is still a case 1 type in virtue of the fact that the mechanism itself is implanted.
18. I am claiming neither that F & R explicitly endorse MAN nor that they are com-
mitted to it; rather, the point of MAN is to provide something other than mere
intuition to explain why it may seem to us that Schmidt is not morally responsible
in this case.
20. Block may, of course, have to remove more than just some of Schmidt’s memo-
ries in order to avoid psychic weirdness in Schmidt, but the story is just that Schmidt
removes whatever in addition to the relevant memories is minimally necessary in
order to ensure a vote for Hitler.
Moderate Reasons-responsiveness 171
21. In order to motivate the worry that Schmidt is not morally responsible in this
case, consider that it works as an amnesia case might. Consider Chuck, who has lost
all memories of the last five years. Chuck has been married for five years, and his
wedding anniversary was yesterday. Although Chuck knows the woman who is his
wife (for he remembers having dated her), he did not know—indeed, could not have
known—that yesterday was his wedding anniversary. So Chuck not only failed to
buy his wife anything for their wedding anniversary, but he also failed to acknowl-
edge their wedding anniversary at all. Assuming that Chuck had not been told earlier
about his impending wedding anniversary, it seems clear that no one, not even
Chuck’s wife, ought to hold him responsible for not doing something special for his
wife on their wedding anniversary (even if she is a bit disappointed). Although
Chuck’s relevant mechanism is MRR, we might think that he is not morally respon-
sible for his lack of appropriate celebratory action, precisely because his amnesia has
deprived him of the use of reasons that he once had. The idea here is that he is in no
way responsible for the loss of those reasons, and so he is not morally responsible for
failing to act appropriately. Compare case 3 to the reasoning in Chuck’s case:
Schmidt is in no way responsible for the loss of his reasons to vote against Hitler, and
so he is not morally responsible for voting for Hitler. However, as I argue in the main
text, Chuck may be morally responsible for whatever actions he does perform, so
long as he acts in an appropriate reasons-responsive way; just note that if he mur-
dered his wife, he would be morally responsible. So long as we are careful to distin-
guish moral responsibility from praiseworthiness and blameworthiness, we should
have no trouble seeing that Chuck is morally responsible for his actions, even
though he is not blameworthy for failing to do anything special for his wedding
anniversary.
22. Fischer made these claims during the question and answer session of my pre-
sentation of a version of this paper at the Inland Northwest Philosophy Conference
on Freedom and Determinism, University of Idaho and Washington State University,
April 2001.
References
Della Rocca, M. 1998. “Frankfurt, Fischer, and Flickers.” Noûs 32: 99–105.
Fischer, J. 1995. The Metaphysics of Free Will: An Essay on Control. Oxford: Blackwell.
Ginet, C. 1998. “Comments on John Fischer, The Metaphysics of Free Will.” Journal of
Social Philosophy 29: 126–134.
172 T. R. Long
Vihvelin, K. 1998. “John Martin Fischer, The Metaphysics of Free Will.” Noûs 32:
406–440.
8 Which Autonomy?
Nomy Arpaly
“When I make a word do a lot of work like that,” says Humpty Dumpty,
“I always pay it extra.” Some words in philosophy have been working over-
time, as, for example, anyone knows who has attempted to explain to a
bewildered student the relations between various ‘externalisms’ and ‘in-
ternalisms’. It has to be admitted, though, that the multitude of meanings
that ‘internalism’ has in our field is not a real problem. Real problems begin
when a term of art performs so many tasks that it becomes at least as
elusive and complex as the natural language terms it was supposed to help
clarify. I suspect that this is the case with the term ‘autonomy’. In philo-
sophical discussions of moral responsibility, attention has shifted some-
what from questions of praise and blame, responsibility and lack thereof, to
questions about autonomy: what it is, whether it exists, and so forth.
I would like to point out that there are at least eight things that we some-
times call ‘autonomy’, and there is a need to tell them apart. In discussion,
we sometimes act as if these eight things were one, or, more often, act as if
intuitions concerning any one of them should be equally relevant to the
agency theorist who is trying to figure out what true autonomy consists in.
The discussion in this paper is meant to raise questions about the general
usefulness of the concept of autonomy to our theorizing about moral
responsibility—to make us wonder if a moral responsibility theorist would
not be better off going back to investigating praiseworthiness and blame-
worthiness directly, ignoring intuitions regarding autonomy. (I do not
intend to criticize or question the use of ‘autonomy’ in political philoso-
phy, normative ethics, or any other context except the context of moral
responsibility and descriptive agency theory.)
Let us look at the various things that ‘autonomy’ means. One of these
things is what I will call, for convenience, agent autonomy. Agent autonomy
174 N. Arpaly
for not explaining why deferential wives are nonautonmous, is most likely
making a sort of category mistake.
An important concept of autonomy that I have not yet mentioned is the
concept of normative, moral autonomy—the one invoked when people ask
to be allowed to make their own decisions and to be free from paternalis-
tic intervention. There are well-known Kantian arguments for a connec-
tion between moral autonomy and agent autonomy, and there is much to
be said about them (I will return to some of them later), but I would like to
point out that the word ‘autonomy’ in ‘respect for autonomy’—an almost
pretheoretical notion in the United States of America—does not simply
mean ‘agent autonomy’. To see this, it suffices to look at a paradigmatic
case of ‘autonomy violation’ in the normative sense. If someone steals
from you, that person violates your autonomy in the normative sense, but,
plausibly, you are not any less of an autonomous agent as a result. Or
imagine that a doctor lies to a patient or withholds information from a
patient about his condition, ignoring the fact that the patient is in the
process of making an important employment decision to which his condi-
tion would be relevant. Many people would say, in this case, that the doc-
tor made it impossible for the patient to make an autonomous decision.
This is, however, not the sense of ‘autonomy’ that autonomous-action the-
orists are looking for. Suppose that the patient did, based on the false or
incomplete information received from his doctor, make an employment
decision. Are we, as agency theorists, to say that his decision was not agent
autonomous? If we were to say that, we would have to hold that any deci-
sion based on mistaken beliefs or incomplete information is a nonau-
tonomous decision. Given that people err frequently and have very
incomplete information about the world, or the next move one’s opponent
plans to make in chess, for that matter, that would again make autonomy
too scarce a commodity, especially if one intends to make it the basis of
moral responsibility.
I have spoken of autonomy as self-control, autonomy as material and
psychological independence, and normative autonomy. David Velleman
makes an additional distinction between autonomy, which is some form of
self-control, and authenticity, or being true to oneself (Velleman 2002). His
distinction is driven by a reading of Frankfurt’s latest view of love, caring,
and volitional necessity. Until about 1987, Frankfurt’s key examples of
nonautonomous agents were examples of people whom we would normally
Which Autonomy? 177
with what one would be tempted to call their ‘real selves’. We can imag-
ine a person who, like the young E. T. A. Hoffman, represses his passion
for literature due to his conviction that he should become “a respectable
citizen.” This person may be very autonomous, but it is natural to say that
he is out of touch with his real self. On the other hand, Harry Frankfurt’s
agents, who surrender to ‘volitional necessity’, may be very true to them-
selves, but they do not control themselves—it may be said that, on the
contrary, their selves control them, as it were. This is not to suggest, by any
stretch of the imagination, the romantic thesis that the autonomous and
the authentic are always in conflict, but only to point out that sometimes
they are, and that therefore they are not one and the same.
Our talk of people being more ‘themselves’ or less so is not all of a piece,
either. Peter Strawson, in “Freedom and Resentment,” discusses contexts
in which we make statements such as “you must excuse her, she is not
herself—she has been under stress” (Strawson 1962). This is a statement
about what Velleman calls ‘authenticity’—the person’s action is claimed to
be less blameworthy because, presumably, it does not reflect her true will
but only the stress she has endured. Quite different is the kind of talk
in which a person reports that she experiences her feelings or desires as
alien, or feels ‘possessed’ or ‘violated’ by them or passive towards them, like
Frankfurt’s enraged man1 who feels as if his anger is an alien intruder and
thus not truly his own. It is the latter kind of talk that gave birth to the idea
of external desires, and while many accounts of external desires have been
given in terms of the structure of the agent’s will, the pretheoretical expe-
rience of ‘passivity’ toward one’s desires often serves as the intuition against
which these accounts are often measured. For example, the fact that some
of us do not experience our vices as alien led Frankfurt to refine his account
of externality so as to allow some unwanted traits (vices) not to be external,
and the fact that some thoughts are experienced as alien has led him to sug-
gest that there are external thoughts as well as external desires—thoughts
toward which we feel ‘passive’ (Frankfurt 1988). Because autonomy is some-
times perceived as the absence of externality, its phenomenological
face, the experience of passivity and alienation with respect to desires, is
often assumed to be relevant to autonomy.
The idea of external desires can be quite strange. If a Victorian lady
experiences her sexual desires as alien, intrusive, ‘not truly her own’, our
natural reaction is to tell her that she is wrong, that these desires are in fact
Which Autonomy? 179
her own, and only the false, asexual self-image that she acquired with her
upbringing makes her experience them as threatening to her integrity as a
person. We would think that she denies her real self. A person who first dis-
covers the presence of homosexual desires in himself is also likely to expe-
rience his desires as alien and he, too, would be urged to ‘accept himself for
who he is’. Similarly, the man whom Frankfurt used as his key example of
a person possessed by an external passion—namely, the man who feels pos-
sessed or violated by his outburst of rage—need not be the same person to
whom we would be inclined to give the Strawsonian excuse of ‘not being
himself’—he may in fact be revealing what we would call his true colors in
his outbursts, however honestly he experiences his anger as ‘not himself’.
It is perfectly possible for a person to wonder what came over him while his
friends tell each other ‘he is very much himself today’. In other words, ex-
periences of ‘being possessed’ are not always an indication that the desire
in question is inauthentic. Nor are they always an indication of lack of
agent autonomy or self control. Some addicts in religious twelve-step pro-
grams experience their sobriety as the will of God taking over, and their
akratic, ego-distonic falling off the wagon as a manifestation of their will-
fulness. Or consider a person who experiences herself primarily as weak-
willed, spineless, lazy, and passive. When this person, because of her lack of
self-control, takes a course of action or inaction which she regards as utterly
impermissible, she does not feel invaded or possessed at all. Instead, she
sadly thinks: Here I go again. I hate myself.2 Imagine now that suddenly, in
an emergency, our lazy person begins to act with admirable self-control. As
her desires obey her reason and her actions follow her decisions, for a
change, she may feel, paradoxically enough, as if somebody—somebody
cool and competent—has taken over her.
So far we have seen autonomy as self-control, autonomy as having power
in the world, autonomy as psychological independence, autonomy as hav-
ing moral rights, and autonomy as authenticity. The foregoing discussion
of the external desires literature shows that we need to introduce another
figure who, until now, has been lurking in the background of our discourse
on autonomy: the figure of the autonomous person as the self-identified
person, the person with a harmonious and coherent self-image who never
experiences her desires as an external threat.3
Also lurking in this background are concepts of what may be called heroic
autonomy, namely, ideals such as Spinoza’s freedom, stoic Apathia’s or
180 N. Arpaly
It is often taken for granted that someone like Beth is not an autonomous
agent—after all, Beth has been brainwashed by the Dean Who Shall Not be
Named. The reaction that we have to Beth’s story, along the lines of ‘surely
she cannot be autonomous if her state of mind has been compelled’, often
serves as an intuition against which to check theories. It is used as such by
proponents of historical views of autonomy, such as Fischer and Ravizza,
Which Autonomy? 183
obvious at all that she lacks autonomy, or that she lacks it more than the
other converts mentioned.
Is present-day Beth autonomous in the sense of agent autonomy or self-
control? We would have to say she is—Mele himself states that she is
the same as Ann in this way. She does things that she reflectively endorses,
and cannot be regarded as ‘wanton’, compulsive, and so forth. Similarly,
present-day Beth is not lacking autonomy in the sense of being particularly
irrational or particularly unresponsive to reasons. It is true that the shift in
her basic worldview was in fact irrational, or unresponsive to reasons, but
again, even very rational people have had such shifts, and that does not
make their current actions less rational or less responsive to reasons. In her
current actions, including staying in the office for long days, Beth may be
as responsive to reasons as Ann is. One might wish to introduce a new
sense of autonomy—a sense in which a person is autonomous only if all of
her convictions are the result of reason-responsive processes or at very least
intentionally induced by her in a self-controlled act of self-indoctrination.
But this would be a heroic notion of autonomy, of little use to the moral re-
sponsibility theorist. If one wishes to maintain that only autonomous agents
are morally responsible, one cannot say that the relevant sense of ‘autonomy’
requires a history that does not include irrational and uninvited changes of
outlook due to growing up, growing old, falling in love, loneliness, parent-
hood, hormones, Prozac, or fatigue, because this would mean that very few
people would be morally responsible at all. At the end of the day, the only
ways in which Beth may be said to be nonautonomous involve the fact that
her convictions are secondhand, the fact that the dean is in the habit of
violating her rights, and the fact that she is not some sort of super-rational
creature who never undergoes irrational or uninvited changes. Yet, neither
independence of mind nor having one’s rights respected nor heroic auton-
omy seemed, prima facie, when considered by themselves, particularly rel-
evant to the question of moral responsibility or lack thereof. Anyone who
wishes to argue that Beth is not morally responsible for her actions would
need to explain why having been irrationally influenced by an evil human
being exempts her from responsibility in a way that having been influenced
in a similar way by some unlucky chance of a force of nature does not.
Is Beth responsible for her actions? I shall not answer this question here.
For now, let me just point out that this is a real question (as Mele acknowl-
edges) and intuitions run both ways, at times disturbed by personal identity
Which Autonomy? 185
concerns. One need only remember that Patty Hearst, after changing her
values as a result of brainwashing, was still convicted for her crimes, to re-
alize that the issue is complex. My point is that the language of autonomy
obscures the complexity of the issue. It is not clear whether Beth is morally
responsible. It seems obvious that in at least two senses, she is not auto-
nomous. By substituting the question ‘Is Beth autonomous?’ for the ques-
tion ‘Is Beth morally responsible?’—treating the questions as if they were
identical—an author may prejudice the reader towards a negative answer to
the question of Beth’s responsibility, turning a complex question into some-
thing trivial.
Let us look at another case study—this time, not a quote from a philoso-
phy book but a story by John Le Carre. Oliver Single is a young lawyer in a
prestigious firm—an heir apparent to his illustrious father, Tyger Single,
who heads the firm. The firm is heavily involved in organized crime. Oliver
had always known that his father makes shady deals, but only gradually
does he learn quite how shady. Even when he knows quite well that his
father performs immoral actions, he still has no plans to leave his post. He
enjoys the money and the prestige, and in addition, he loves his father
deeply and wishes to please him. At a critical moment in the novel, Oliver
defects from the side of his criminal family to the side of the law, in whose
service he remains to perform some courageous, intuitively admirable acts.
Here is how Le Carre describes Oliver’s defection:
It is Midwinter, and Oliver is a little mad. That much he knows about himself, no more.
The Origins of his madness, its causes, duration and degree, are not within his grasp,
not now. They are out there, but for another time, another life, another couple of
brandies. The neon-lit gloom of a December’s night at Heathrow airport reminds
him of a boy’s changing room in one of his many boarding schools. Garish card-
board reindeer and taped carols compound his mood of unreality. Snowy lettering
dangles from a washing line, wishing him peace and joy on earth. Something amaz-
ing is about to happen to him and he is eager to find out what it will be. He isn’t drunk
but technically he isn’t sober either. A few vodkas on the flight, a half of red with the
rubber chicken, a remy or two afterwards have done no more in Oliver’s eyes than
bringing him up to speed with the furor already raging inside him. He has only hand
luggage and nothing to declare except a reckless ferment of the brain, a firestorm of
outrage and exasperation starting so long ago that is origins are impossible to peg,
blowing through his head like a hurricane while other members of his internal congre-
gation stand around in timid twos and threes and ask one another what on earth Oliver is
going to do to put it out. He is approaching signs of different colors and instead of
wishing him peace, joy on earth and goodwill among men, they are requiring him
186 N. Arpaly
to define himself. Is he a stranger in his own country? Answer, yes he is. Has he
arrived here from another planet? Answer, yes he has. Is he blue? Red? Green? His
eye drifts to a tomato-colored telephone . . . It is familiar to him. Perhaps he noticed
it on his way out three days ago and, unknown to himself, recruited it as a secret ally.
Is it heavy, hot, alive? A notice beside it reads “If you wish to speak to a Customs
officer, use this phone.” He uses it. That is to say, his arm reaches out for it unbidden, his
hand grasps it and puts it to his ear, leaving him with the responsibility of what to
say. (Le Carre 1998, 90)
Much of the novel is built upon the reader’s admiration of Oliver and his
defection, and this admiration assumes moral praiseworthiness—hence,
moral responsibility. The sense that Oliver is morally responsible, that his de-
fection is not just an accidental drunk action, comes from the feeling that
his action is, after all, reason-responsive: it is not an accident that the ac-
tion occurs right after his discovery of his father’s worst crimes, it is not an
accident that Oliver has noted the red phone long ago, as if knowing with-
out knowing that some day he’ll use it, and one even gets the impression
that Oliver’s drinking on the airplane is taken as a preemptive tranquilizer
to assist him in the hard task to come. Oliver’s action also appears to be
authentic: something deep and real in him, his moral sensibility, finally re-
volts against his life as his father’s heir. On the other hand, as for self-control,
Oliver does not seem to have much during his action, and Le Carre stresses
again and again the passivity and turmoil which he experiences. He is
alienated from his actions, and does not seem to identify with them. Le
Carre’s description of Oliver’s dark night of the soul is rather credible, and
may ring true to many people who have left a father, a religion, a marriage.
Is Oliver autonomous? Is there any clear meaning to this question—or any
good reason to ask it?
“Of course,” some agency theorists will answer. “After all, there is a rea-
son to ask if Oliver is morally responsible for his action, and the asking if
he is autonomous is simply the same as asking if he is morally responsible.
And may we point out that when we use the word ‘autonomy’, we do not
mean any of eight things you have listed, for by ‘autonomy’ we mean sim-
ply that on which moral responsibility depends, whatever it turns out to be.
Thus, there is also a reason to ask what it is that constitutes autonomy—
this is simply the same as asking what it is that makes an agent morally
responsible.”
But I take myself to have demonstrated why this may not be the best way
to pose the question of moral responsibility, as at least some of the meanings
Which Autonomy? 187
I have listed invariably get smuggled in, mixed together, whenever we use
the word autonomy, thus inviting confusion. Furthermore, there is also
the risk that ‘What is autonomy?’ is a bit too loaded a way to bring about the
question of moral responsibility, focusing our attention on some kinds of
solutions and not others. Perhaps we should simply ask what it is that makes
an agent morally responsible, or praiseworthy, or blameworthy, and give a
much overworked term of art a well-deserved rest.
Acknowledgments
This work, originally written for the Inland Northwest Philosophy Confer-
ence, appears in a modified form as a part of my book Unprincipled Virtue:
An Inquiry Into Human Agency, forthcoming, Oxford University Press.
Notes
4. Scanlon (2002), Damasio (1994), and Audi (1990). Railton’s work is yet unpub-
lished, though it has been presented in many talks.
References
Arpaly, N. 2000. “On Acting Rationally Against One’s Best Judgment.” Ethics 110:
488–513.
Damasio, A. 1994. Descartes’ Error: Emotion, Reason, and the Human Brain. New York:
G. P. Putnam.
Fischer, J. and M. Ravizza. 1998. Responsibility and Control: A Theory of Moral Respon-
sibility. New York: Cambridge University Press.
Le Carre, J. 1998. Single and Single. New York: Charles Scribner’s Sons.
Scanlon, T. 2002. “Reasons and Passions.” In S. Buss and L. Overton, eds., Contours of
Agency. Cambridge, Mass.: MIT Press.
1 Introduction
. . . a man [Green] walks along a beach and, noting that there is a child drowning,
dives into the water and rescues the child. Though Green has had a device implanted
in his brain [by scientists in a research institute in California—one might now say
‘crazed neurophilosophers in La Jolla’], the device does not play any role in Green’s
decision to save the child (and his subsequent action). That is, the device monitors
190 J. M. Fischer
Green’s brain activity but does not actually intervene in it. Let us suppose that this
is because the scientists can see that Green is about to decide to save the child and to
act accordingly [they are morally good, albeit crazed, neurophilosophers]. But let’s
also suppose that the scientists would have intervened to bring about a decision to
save the child if Green had shown an inclination to decide to refrain from saving the
child. That is, were Green inclined to decide on his own not to save the child, the
scientists would ensure electronically that he decide to save the child and also that
he act to carry out this decision. (Fischer 1986, 41)
Thus, semicompatibilism was born. Here I wish to defend the basic intu-
ition, which I still believe is correct, that the principle of transfer of blame-
lessness (or, more broadly, nonresponsibility) is called into question by the
Frankfurt-type cases, and that it cannot be employed in an uncontroversial,
decisive argument against the compatibility of causal determinism and
moral responsibility.
If
(1) p obtains and no one is even partly morally responsible for p; and
(2) if p obtains, then q obtains, and no one is even partly morally respon-
sible for the fact that if p obtains, then q obtains;
The Transfer of Nonresponsibility 191
then
(3) q obtains, and no one is even partly morally responsible for q. (Fischer
and Ravizza 1998, 152)
[In “Erosion”], Betty plants . . . explosives in the crevices of [a] glacier and [intu-
itively speaking, freely] detonates the charge at T1 causing an avalanche that crushes
the enemy fortress at T3 [a result intended by Betty]. Unbeknownst to Betty . . . how-
ever, the glacier is gradually melting, shifting, and eroding. Had Betty not placed the
dynamite in the crevices, some ice and rocks would have broken free at T2, starting
a natural avalanche that would have crushed the enemy camp at T3. (Fischer and
Ravizza 1998, 157)
If
(1) p obtains and no one is even partly morally responsible for p; and
(2) (i) p is part of the actual sequence of events e that gives rise to q at T3,
(ii) p is a part of e that is causally sufficient for the obtaining of q at T3,
and
(iii) no one is or ever has been even partly responsible for (2i and ii);
then
(3) q obtains at T3, and no one is or ever has been even partly morally
responsible for q.4
But a case similar to Erosion, Erosion*, shows Transfer NR to be problematic.
In Erosion*, the conditions of the glacier do actually cause the ice and rocks
to break free, triggering an avalanche that arrives at the fortress precisely at
the same time as the independent avalanche triggered freely by Betty. Each
avalanche is sufficient for the destruction of the fortress, and yet Betty seems
192 J. M. Fischer
The Critique
In an interesting recent paper, Michael McKenna has argued that the criti-
cism of the transfer idea developed above is unsuccessful (McKenna 2001).
The Transfer of Nonresponsibility 193
He says:
As Fischer and Ravizza observe, their reply relies exclusively upon examples that involve
‘two-path’ cases, i.e., cases in which the obtaining of the event is ensured by two dif-
ferent causal pathways to the same event. They do not object to Transfer NR as it applies
to one-path cases, but hold that Transfer NR restricted to one-path cases cannot be
used to show that moral responsibility is incompatible with causal determinism.
This is because causal determinism does not rule out cases of overdetermi-
nation. They therefore hang their criticism of Transfer NR on counterexamples in-
volving two-path cases of simultaneous overdetermination. (McKenna 2001, 45)
But McKenna is not sanguine about the prospects for our approach; he goes
on as follows:
This is not a convincing strategy. No doubt, Fischer and Ravizza are correct that
causal determinism does not rule out overdetermination. But if determinism is true,
then the manner in which the facts of the past and the laws of nature entail one
unique future is not analogous to the manner in which one set of independently
existing causally sufficient conditions (for example, an erosion) ensures a subse-
quent event also ensured by some distinct set of independently existing sufficient
conditions (i.e., Betty’s action). Assuming determinism, the pertinent facts (consist-
ing in the deterministic order of things) are not independent of an agent’s reasons for
action, they constitute them! Therefore, at a deterministic world involving a typical
case regarding a judgment of moral responsibility, the case is relevantly like a one-
path, not a two-path case. (McKenna 2001, 45)
Although we had chastised van Inwagen for attending only to a proper sub-
set of the relevant cases, the one-path cases, it is precisely McKenna’s point
that these are the only cases relevant to causal determinism. Thus, he seeks
to formulate a version of the modal principle that captures a restriction to
one-path cases. Here is a slight reformulation of McKenna’s proposal, Trans-
fer NR :
If
(1) p obtains and no one is or ever has been even partly morally
responsible for p; and
(2) (i) p is part of the actual sequence of events e that gives rise to q at T3;
(ii) p is causally sufficient for the obtaining of q at T3; and any other
part of e that is causally sufficient for q either causes or is caused by p;
and
(iii) no one is or ever has been even partly responsible for (2i and ii);
then
194 J. M. Fischer
(3) q obtains, and no one is or ever has been even partly morally res-
ponsible for the fact that q obtains at T3. (McKenna 2001, 45)5
McKenna points out that Transfer NR does not rule out moral responsibil-
ity in cases of causal overdetermination, even in a causally deterministic
world. But he says that it would be very implausible for Ravizza and me to
hold that an agent can be morally responsible only in cases of simultaneous
overdetermination. Indeed, McKenna credits both Carl Ginet and Eleonore
Stump with pointing out that it would be a significant problem for the
Fischer/Ravizza approach if moral responsibility turned out to be incom-
patible with causal determinism in all one-path cases, which are, presum-
ably, the vast majority of cases (McKenna 2001, 46 and note 39, 51).
A Reply to McKenna
McKenna’s paper contains many insightful and helpful points. He makes
an important observation when he points out that the manner in which
determinism ensures a unique future may well be crucially different from
the manner in which one set of independent causally sufficient conditions
(say, erosion) ensures a subsequent event also ensured by some distinct set
of independently existing sufficient conditions (for example, Betty’s free
action). I will return to this point below, and also in my discussion
of Eleonore Stump’s defense of the transfer of nonresponsibility idea. But
here I wish to disentangle different points. One contention of McKenna is
that the validity of Transfer NR would pose significant problems for a com-
patibilist about causal determinism and moral responsibility (like me).
Granted. But I think a rather different issue is whether Transfer NR can be
employed to establish the incompatibility of causal determinism and moral
responsibility.
Mark Ravizza and I contend that it is futile to seek to restrict the modal
principle to one-path cases (as in Transfer NR ) precisely because this would
render the principle unable to generate incompatibilism about causal deter-
minism and moral responsibility (which is, after all, the conclusion of van
Inwagen’s argument).6 To point out that Transfer NR poses problems for
developing a plausible compatibilist theory of moral responsibility is one
thing; to suppose that it generates a successful argument for incompatibil-
ism is quite another. One possibility is that there are significant problems for
both compatibilism and incompatibilism, as Saul Smilansky has recently
contended (Smilansky 2000).
The Transfer of Nonresponsibility 195
Stump’s Defense
In two separate places, Eleonore Stump has sought to defend the basic idea
of the transfer of nonresponsibility, and thus the direct argument for the
incompatibility of causal determinism and moral responsibility, against the
attacks mounted by Ravizza and me. In the first piece, “The Direct Argu-
ment for Incompatibilism,” Stump agrees that cases such as Erosion and
Erosion* show the inadequacy of Transfer NR and Transfer NR (Stump
2000). But she, like McKenna, complains that these are two-path cases. She
contends that no one-path case can similarly show the inadequacy of
a suitably restricted transfer principle—a principle whose application is
restricted to one-path cases.
In the new piece, “Control and Causal Determinism,” she offers what she
characterizes as a “significant revision” of the argument in her first paper
(Stump 2002, note 16, p. 57). She says: “In their response . . . Fischer and
Ravizza concede the general conclusion of my argument but argue that it
constitutes only a stalemate between the incompatibilist and their posi-
tion. Fischer and Ravizza are not right in this view, as I hope to show in the
revised version of the argument I give in this essay” (ibid.).
But this is a bit puzzling. I do not know what Stump means in saying that
Ravizza and I “concede the general conclusion” of her argument. At the end
of that first paper, Stump says, “Consequently, Fischer and Ravizza have not
succeeded in showing that the direct argument for incompatibilism fails.” I
The Transfer of Nonresponsibility 197
do not think we agreed with this conclusion. We did agree that we could not
decisively show that a transfer principle restricted to one-path cases is unac-
ceptable. But we also pointed out that it is not our obligation decisively to
prove such a principle problematic, in order to show that one need not
accept the argument for incompatibilism. Given that there are considerable
attractions to compatibilism (for example, a compatibilist need not think
that our very basic views of ourselves and others as morally responsible and
as persons are held hostage to a possible scientific discovery by a consor-
tium of scientists that causal determinism is indeed true), and given (what
we argued) that a restricted transfer principle cannot be uncontroversially
established (i.e., that one reaches a dialectical stalemate in seeking to argue
for—or against—such a principle), we concluded that the incompatibilist’s
argument is not persuasive. That is, in light of the fact that its central prin-
ciple is essentially contested and cannot uncontroversially be supported,
we showed that one need not accept the incompatibilist’s argument. We did
think we were showing that the incompatibilist’s argument fails, insofar as
we thought we were showing that it does not succeed within the relevant
dialectic context.
It will be useful to consider Stump’s new formulation of her argument,
and to consider it in light of the discussion of McKenna’s Transfer NR in
the previous section. She proceeds by considering Erosion* and construct-
ing a dilemma. On the first horn of the dilemma, one is to assume that the
world of Erosion* is not causally deterministic and that Betty acts indeter-
ministically (but in a way that supports the judgments that she is acting
freely). Now Stump is willing to concede that this sort of example shows
that Transfer NR, unqualified or restricted, is invalid; but she thinks that
the principle can simply be qualified so as to rule out the counterexample
and still get the desired incompatibilistic result. She says,
. . . Erosion* shows that in cases of overdetermined effects, one of whose causes is an
agent acting indeterministically, Transfer NR fails. But this is not enough to show
that Transfer NR should be rejected. Because Transfer NR is a highly plausible princi-
ple, we can take Erosion* to show just that the principle needs to be restricted, to
exclude such unusual cases as that in which an indeterministic agent is one of the
causes of an overdetermined effect. (Stump 2002, pp. 40–41)
On the other horn of the dilemma, the assumption is that the world of
Erosion* is causally deterministic. But now of course it will be question-
begging, in the relevant dialectical context (in which Transfer NR is being
deployed in order to establish that causal determinism is incompatible
198 J. M. Fischer
Reply to Stump
I would contend that there is trouble for Stump’s incompatibilistic con-
clusion, on either horn of the dilemma. Consider, first, the assumption of
causal indeterminism. Here Stump suggests a restriction of Transfer NR to
one-path contexts. She doesn’t actually present the principle, but if such
a principle can be crafted, it will have exactly the same problems as
McKenna’s Transfer NR . As I explained above, it will not be possible to em-
ploy this sort of one-path modal principle in a general argument for incom-
patibilism about causal determinism and moral responsibility. And if the
single path contains some insuring factor as seems required in order to trig-
ger the application of the modal principle, then it would seem that the
principle is implicitly presupposing causal determinism. If this is so (and it
explicitly is the case on the second horn of Stump’s dilemma), then it is
admittedly impossible, within the relevant dialectical context (in which
the compatibility of causal determinism and moral responsibility is at
issue) to produce a non-question-begging counterexample to Transfer NR.
But I contend that it is equally impossible (within the relevant dialectical
context) to provide examples that give the necessary sort of support to the
principle. That is, I believe that the current dialectical context is what I
have characterized in previous work as a dialectical stalemate (Fischer 1994,
83–85). This is my characterization of a dialectical stalemate:
Frequently in philosophy we are engaged in considering a certain argument . . . for
some claim C. The argument employs a principle P. Allegedly, P supports C. Now the
proponent of the argument may be called upon to support the principle, and he may
do so by invoking a set of examples (or other considerations). Based on these exam-
ples (or other considerations), he argues that the principle and thus also the philo-
sophical claim are to be accepted.
But the opponent of the argument may respond as follows. The examples are not
sufficient to establish the principle P. One could embrace all the examples and not
adduce P to explain them; rather, it is alleged that a weaker principle, P* is all that is
The Transfer of Nonresponsibility 199
decisively established by the examples (or other considerations). Further, P*, in con-
trast to P does not support C. Finally, it is very hard to see how one could decisively
establish P. One reason it is so difficult is that it at least appears that one cannot
invoke a particular example that would decisively establish P without begging the
question in a straightforward fashion against either the opponent of P or the oppo-
nent of C. Further, it also seems that one cannot invoke a particular example that
would decisively refute P without begging the question against the proponent of P or
the proponent of C. These conditions mark out a distinctive—and particularly
precarious—spot in dialectical space. (Fischer 1994, 83)
Return to Erosion*. I would say (to a first approximation, at least) that Betty
is morally responsible for her behavior, and its upshot (the destruction of
the enemy camp at T3), because she engaged in unimpaired practical
reasoning—practical reasoning undistorted by factors that uncontrover-
sially threaten moral responsibility, such as significant delusions, psychoses,
coercion, compulsion, irresistible impulses, clandestine manipulation, hyp-
nosis, subliminal advertising, and so forth. I am inclined to say that she is
morally responsible, in other words, because no uncontroversially responsi-
bility undermining factor impairs (or in any way affects) her deliberations,
her formation of an intention, and her action in accordance with it. This
is so, even though there is a second path to the same upshot that is a no-
responsibility path. The fact that the no-responsibility path is entirely sepa-
rate insulates Betty’s path from it, and allows all parties to the disagreement
200 J. M. Fischer
If
(1) p obtains and no one is even partly morally responsible for p; and
(2) if p obtains, then q obtains, and no one is even partly morally respon-
sible for the fact that if p obtains, then q obtains; and
202 J. M. Fischer
(3) on the actual path that leads from p’s obtaining to q’s obtaining, either
there is no factor that at least prima facie could be thought to ground moral
responsibility, or there is some factor that uncontroversially undermines
moral responsibility (e.g., a factor that distorts or impairs the distinctive
process of human practical reasoning); then
(4) q obtains, and no one is even partly morally responsible for q.
It will sometimes be helpful to refer to a path that meets conditions (1) and
(3) as a “no-responsibility” path, just for simplicity’s sake. Transfer NRC
is a plausible and attractive modal principle that explains why we transfer
nonresponsibility in the clear and uncontroversial cases. For example, it ex-
plains why no one is even partly morally responsible for the fact that John
dies (from a snakebite) on his thirtieth birthday in Snake Bite, and it ex-
plains why no one is even partly morally responsible for the tsunami’s hit-
ting the coast of California in Tsunami. It has this implication in all of the
examples invoked by van Inwagen in an attempt to support Transfer NR.
Note that it does not entail that Betty is not morally responsible in Erosion
or Erosion*. Note, further, that Transfer NRC cannot be employed as part of
a direct argument for the incompatibility of causal determinism and moral
responsibility that would appeal to those not antecedently committed to
incompatibilism. Transfer NRC thus explains all the uncontroversial exam-
ples just as well as (say) Transfer NR (or an otherwise qualified version of
Transfer NR that only applies to one-path cases), but does not help to yield
incompatibilism. It is thus precisely the sort of principle whose availability
shows that the attempt to employ some appropriately restricted version of
Transfer NR to establish incompatibilism leads to a dialectical stalemate.
McKenna is (to some degree) aware of the dialectical difficulties faced by
someone who wishes to employ Transfer NR to argue for incompatibilism
about causal determinism and moral responsibility (McKenna 2001, note
45, 51).9 In response, McKenna says that Transfer NR is not merely ad hoc,
but is independently confirmed by a range of cases: “. . . notice that cases
like Snakebite can be used to confirm the plausibility of Transfer NR
as readily as they can the leaner Transfer NR (ibid.). I grant that these cases
support Transfer NR as readily as they do Transfer NR. But this is not the
issue. The point is that Snake Bite (and Tsunami and all the cases invoked
by van Inwagen and other incompatibilists) are unable to support Transfer
NR over Transfer NRC. So, as with dialectical stalemates in general, there is
a competitor principle—Transfer NRC—which must be considered when
The Transfer of Nonresponsibility 203
6 Conclusion
deterministic context, the agent may act from his own, appropriately
reasons-responsive mechanism. Thus, in my view, in such a context the
agent may exhibit the characteristic sort of control—guidance control—
that grounds moral responsibility, even though he is not morally responsi-
ble for conditions prior to his birth, and he is not responsible for the fact
that those conditions issue in his behavior. In place of Nagel’s metaphysi-
cal depression, I have sought to lift our spirits (if offering the possibility of
freedom and responsibility could do this). Although it might be supposed
that my suggestion pertains more to Pandora than Pangloss, I nevertheless
propose that the glass is half full.
Acknowledgments
Notes
2. van Inwagen presented his principle, and the argument for incompatibilism that
employs it in van Inwagen (1980), reprinted in Fischer (1986, 241–249).
3. I am indebted to Nathan Salmon for pointing out that, if the connective in (2) is
interpreted as a material conditional, then Transfer NR is clearly valid. This is
because any counterexample would have to be a case in which someone is at least in
part morally responsible for q. It would seem to follow that someone is at least in
part morally responsible for ‘q or not-p’. Thus (2) must be false. So any putative coun-
terexample to the principle would be a case in which the second premise would have
to be false, and thus it would not be a genuine counterexample to Transfer NR. I
accept this point, but it also nicely shows why Transfer NR, so interpreted, cannot be
employed in a sound, uncontroversial argument for incompatibilism. This is because
in any “ordinary” context in which a compatibilist is inclined to say that an agent is
morally responsible for q and causal determinism is assumed to be true, the compat-
ibilist will point out that (2) is false, given that the connective is interpreted as a
material conditional. So (given the dialectical situation) the incompatibilist will not
be entitled to apply Transfer NR to get to the incompatibilistic conclusion.
208 J. M. Fischer
4. We consider such a principle in Fischer and Ravizza (1998, 161–163); also, for the
formulation of the principle, I am indebted to McKenna (2001).
5. McKenna credits Al Mele and Ish Haji for help in formulating this principle. I am
not certain that this formulation works to secure the desired restriction to one-path
cases, but I shall accept it for the purposes of this paper.
8. Stump (2002); she goes on to explore whether Ravizza and I have available to us
some reason to call the incompatibilist’s judgment into question, but she concludes
(lamentably for us) that there is no such reason.
9. McKenna also credits Michael Zimmerman for bringing to his attention the
dialectical difficulty in question.
10. As Michael Bratman has pointed out, there is an interesting parallel between my
methodology here and John Perry’s approach to the mind–body problem in Perry
(2001). Here Perry defends what he calls ‘antecedent physicalism’. By this he means
that he finds physicalism antecedently plausible and sees his job as blocking dualis-
tic challenges.
11. For a defense of this view, see Fischer (1994); and Fischer and Ravizza (1998).
12. He also articulates this worry in Thomas Nagel (1985, 126). Here he is explicit
that the challenge appears whether or not causal determinism obtains.
References
Fischer, J. and Ravizza, M. 1998. Responsibility and Control: A Theory of Moral Respon-
sibility. Cambridge: Cambridge University Press.
. 1985. The View from Nowhere. New York: Oxford University Press.
Perry, J. 2001. Knowledge, Possibility, and Consciousness. Cambridge, Mass.: MIT Press.
I can remember very clearly the first time van Inwagen encountered the
problem of free will. In the autumn of 1965 he was talking with a fellow
graduate student at the University of Rochester, one Myles Brand, and made
some remark that presupposed the incompatibility of free will and deter-
minism. Brand told him—second-year graduate student to first-year gradu-
ate student—that most philosophers believed that free will and determinism
were compatible, and outlined some of the currently popular arguments for
that position. As Athena from the head of Zeus, the argument that van
Inwagen was to publish ten years later in “The Incompatibility of Free Will
and Determinism” (1975) sprang from his head pretty much full-grown—
although it made its entrance into the world by way of his ever-active mouth
and not by Athena’s rather more unorthodox route.
The argument had it roots in the following reflections. If free will and de-
terminism coexist, then someone is able to do something not contained in
that one possible future that is consistent both with the past and the laws of
nature. Suppose that Alice, an inhabitant of a deterministic world, is able to
do something she is not in fact going to do; suppose, to be specific, that
although she is going to remain a prisoner, she is able to escape from her
prison. Her ability to escape can be looked upon as an ability to divert the
river of coming events into a channel through which it is not in fact going
to flow; to realize, that is, or to cause to be actual, a possible future that is not
the future that lies before her, to cause to be actual one of those possible
futures in which she escapes. And what would these possible but nonactual
futures be like? Let’s say that the past of a possible but nonactual future is
that possible past that would be the actual past if that future were the actual
future. Any given future in which Alice escapes must either be a future
whose past is the actual past and which is discontinuous with that past, or a
214 P. van Inwagen
future that is continuous with its past but whose past is not the actual past
(a past different from the actual past all the way back to the Big Bang), or,
finally, a future in which the world is governed by laws of nature that are not
the actual laws. But Alice can bring about futures of none of these sorts. If it
is insufficiently evident as it stands that she can bring about none of these
futures, here is an argument. Let ‘Clio’ be a proper name for the actual past—
thus, when I imagine, as I am about to, Alice using the name ‘Clio’ in an-
other possible world, I imagine her referring to the past as it is in our world,
not the past as it is in hers. Similarly, let ‘Nomos’ be a proper name for the ac-
tual laws of nature (which, remember, we are assuming to be deterministic).
It would seem that if Alice is able to escape, she must be able in some sense
to cause the actuality of or bring about or realize a future in which she could
say, and in so saying speak truly, one of the following three things:
There has been a causal break; the present state of affairs is not continu-
ous with the past.
The present is continuous with the past, but that past is not Clio; it is
some other past (a past different from Clio all the way back to the Big Bang).
The laws of nature are not Nomos, but some other set of laws.
And, obviously, Alice is not able to get herself into a future in which she
can say any of these things and be right.
When van Inwagen got around to writing down the argument that had
occurred to him in his conversation with Brand—he first did this in a doc-
toral thesis he wrote under the supervision of Richard Taylor (de jure) and
Keith Lehrer (de facto)—the argument he wrote down did not look much
like the argument I have just set out. (It looked very much like the argu-
ment he would later publish in “The Incompatibility of Free Will and
Determinism.”) Nevertheless, the central idea of both arguments was the
same, and they no doubt stand or fall together.
When van Inwagen had got his first academic job and was trying to publish
this argument, he found it extraordinarily hard to do so. The reason was
simple: the conclusion of the argument was known to be false. The unanimous
position of the referees whose reports were enclosed with the rejection letters
could have been expressed in the following words, which I take from an essay
by Donald Davidson, words as well known as they are ungrammatical:
I shall not be directly concerned with such arguments [i.e., arguments for the incompat-
iblity of free will and determinism], since I know of none that is more than superficially
Van Inwagen on Free Will 215
plausible. Hobbes, Locke, Hume, Moore, Schlick, Ayer, Stevenson, and a host of others
have done what can be done, or ought ever to have been needed, to remove the confu-
sions that can make determinism seem to oppose freedom. (1973, 137)
It’s not that no one was an incompatibilist in those days, of course. The
volume that contained the essay from which my quotation from Davidson
is taken also contained David Wiggins’s “Toward a Responsible Libertari-
anism.” Roderick M. Chisholm was an incompatibilist, as were Richard
Taylor, Carl Ginet, Elizabeth Anscombe, and Peter Geach. Still, it can hardly
be denied that incompatibilists were thin on the ground in the sixties and
early seventies.
Eventually, however, van Inwagen was able to publish two papers in
which he argued for the incompatibility of free will and determinism. (He
has always suspected that “The Incompatibility of Free Will and Determin-
ism,” which had been rejected by many journals, was accepted by Philo-
sophical Studies only because Sellars chose Keith Lehrer as its referee, but he
has never known whether this suspicion was correct.) His other paper on
the subject, “A Formal Approach to the Problem of Free Will and Deter-
minism” (1974), had, in respect of publication, the advantage of making
use of the then new and exciting apparatus of possible worlds. (This paper
was strongly influenced by Montague’s “Deterministic Theories”(1974); the
impetus for its writing was a suggestion of Rolf Eberle’s.)
Van Inwagen summed up his thought on free will in his book An Essay on
Free Will (1983), and has pretty much avoided learning anything about the
problem since—other than by sitting about and thinking it over. (The pub-
lication of this book by Oxford University Press was due to the good offices
of Tony Kenny and Derek Parfit, neither of whom could have had any
sympathy whatever with its content, and van Inwagen has always been
grateful to them for their generosity and has tried to imitate it. He has done
his best to learn from them the hard lesson that a philosophical book that
he regards as thoroughly wrongheaded can nevertheless be a good book.)
Van Inwagen likes to think that this book bears a significant share of the re-
sponsibility for the fact that incompatibilists are now much more common
than they were thirty or forty years ago. In a paper that he read at a con-
ference in the early nineties, van Inwagen made a remark to the effect that
compatibilism was the standard view among philosophers. Michael Slote,
who was in the audience, said that he thought that, on the contrary, in-
compatibilism had become the standard view, or at least the majority view.
216 P. van Inwagen
A few years later, van Inwagen asked Ted Warfield whether he thought that
was right. Warfield, who comes as close as is humanly possible to knowing
what every analytical philosopher thinks about anything, replied that he
thought that the majority of analytical philosophers who had actually
worked on the free-will problem were incompatibilists, and that the major-
ity of analytical philosophers (full stop) were compatibilists.
There was one passage in An Essay on Free Will that, after the publication
of the book, van Inwagen became more and more worried about.1 (These
worries were aggravated by pointed questions from Alex Rosenberg, and
also, curiously enough, by a science-fiction novel by Larry Niven called
Protector—a novel about which van Inwagen is ready to tell you consider-
ably more than you want to know.) Van Inwagen had said in that passage
that although no one was able to render a physical law false, it seemed that
if there were psychological laws, and if we had free will, we had to be able to
render these laws false. But that raises the question: if one has it within
one’s power to render some proposition false, in what sense can that propo-
sition be a law? As a sort of schematic example of a psychological law, van
Inwagen proposed, “No one who has received moral training of type A in
early childhood ever spreads lying rumors about his professional colleagues
in order to advance his career.” He imagined someone asking, “Why does the
pattern of behavior described in this statement occur if people don’t have to
conform to it?” He answered this question as follows: “Perhaps it is just the
people who have received moral training of type A in early childhood who
see the point in not spreading lying rumors about their colleagues.”
He gradually came to see, or to think he saw, that this response to the dif-
ficulty was facile, and that the difficulty he had his finger on was broader
and deeper than the original puzzle about psychological laws. He gradually
came to the conclusion that if one was faced with the necessity of doing
either A or B, and that if one saw every reason to do A and no reason what-
ever to do B, then one would simply not be able to do B. From this
conclusion it was no great leap to the slightly stronger conclusion that, if
one was faced with a choice between A and B, and one was aware of con-
siderations that could be brought in support of both alternatives, and if the
considerations that supported A seemed to one clearly and decisively to
outweigh the considerations that supported B, then one would simply not
be able to do B. Van Inwagen defended, in “When Is the Will Free?” (1989),
the thesis that the general principles about ability that lead philosophers to
Van Inwagen on Free Will 217
driver morally responsible for the pedestrian’s death, and when we consider
the latter case we do not.
The relevant difference is, of course, that the man whose reflexes were
impaired by drink, was, so to speak, able to avoid being unable to avoid hit-
ting the pedestrian, and the sober and alert driver was unable to avoid being
unable to avoid hitting the pedestrian. At the moment he first saw the
pedestrian, the drunk driver was unable to avoid hitting him, but he had
earlier been able, or so we should suppose if we were judge or jury, to avoid
driving with the impaired reflexes that were the cause of his fatal inability
at the time of the accident. Van Inwagen suggested that this sort of case
could serve as a model for the relation between ability and responsibility.
Here is a second case, a case in which the inseverable connection between
ability and responsibility, though it remains unsevered, as inseverable con-
nections do, is stretched considerably further than it is in most philoso-
phical examples concerning moral responsibility and the ability to do
otherwise. Consider a man who is, in middle age, a corrupt politician and
is, owing to his corrupted nature, unable to refuse bribes when he believes
there is no significant likelihood of the bribery coming to light. That is how
he is, but how did he get that way? Suppose the answer is this: as a young
man, he made a certain series of free choices, choices preceded by genuine
deliberation, which collectively had the effect of establishing him in set-
tled and unbreakable habits of venality. Van Inwagen argued—guided, I
suppose, by Aristotle—that this politician can properly be held morally res-
ponsible for the baleful effects on the public welfare of the informal ser-
vices he renders to his political cronies in return for money. And this despite
the fact that he is unable, in middle age, to reject the bribes he is offered. He
can properly be held responsible for, say, the deaths of the four children in
the fire in the building that wasn’t up to code, because he could, as a young
man, have avoided becoming the sort of man who would be unable to resist
the bribe offered by the slumlord who owned the building.
Several philosophers have disputed van Inwagen’s conclusion that the
principles that lead philosophers to incompatibilism entail that free acts, if
they exist at all, are extremely rare, but van Inwagen has never been able to
see any force in their arguments. Although he has published answers to
them (1994), he is of the opinion that no answers were needed; that his
original arguments were untouched by the arguments of his critics. One
philosopher, who generally disagrees with van Inwagen about free will,
Van Inwagen on Free Will 219
Dan Dennett, agrees with van Inwagen that these arguments are unan-
swerable. As Dan put it, referring to “When Is the Will Free?”, “Thank you,
Peter, for the lovely reductio of incompatibilism.”
Now van Inwagen’s arguments for this conclusion, whether they are
good or bad, presuppose that there is an inseverable connection between
moral responsibility and the power to do otherwise, however flexible this
connection may be. The inseverable connection is this: if one is morally
responsible for anything, it follows logically that one has had a free choice
about something. But Harry Frankfurt has presented a famous argument
that some have taken to refute this thesis.2 A significant proportion of van
Inwagen’s work on free will has been devoted to Frankfurt’s argument.
And Frankfurt’s argument is important. If it is indeed true that one might
be morally responsible for various things, despite one’s never having been
able to do otherwise than one has done, then the problem of free will
loses much of its interest—for the simple reason that most people would
find the thesis that we lack free will much less unappealing if this thesis
could be shown not to entail that we can never be held to moral account
for anything.
I have said that Frankfurt’s argument has been taken by some to show
that it is possible for one to be morally responsible for something even
though one has never been able to do otherwise. The actual conclusion of
Frankfurt’s argument, however, is this: the so-called principle of alternative
possibilities is false, or at least not a necessary truth. (I’ll call it the princi-
ple of alternative possibilities. Frankfurt has recently presented an ill-
advised defense of his use of the adjective ‘alternate’ in his name for the
principle. It is, I concede, uncharitable of me to mention this. I’ll attempt
to atone for my lapse by very charitably saying nothing further about it.)
This is the principle of alternative possibilities:
Bill is morally responsible for lying under oath only if he could have
done otherwise.
Van Inwagen has never been able to convince himself that he understands
sentences like ‘Bill is morally responsible for lying under oath’. It has al-
ways seemed reasonably plain to him that what one is morally responsible
for is not one’s acts but the consequences of one’s acts, or, more exactly, certain
of the consequences of one’s acts—for no one would suppose that one
could be responsible for all the consequences of one’s acts. (When I say ‘it
has always seemed plain to him’, I mean that it has seemed plain to him
since he first encountered the idea that what a person is morally responsi-
ble for is the consequences of his acts and not the acts themselves. Van
Inwagen is sure that he first encountered this idea in something written by
P. H. Nowell-Smith; he cannot now remember where it was that Nowell-
Smith made this valuable point.) And the consequences of one’s acts, it
would seem, are members of the same ontological category or categories as
the consequences of anything that takes place within the causal and tem-
poral order; whatever ontological category one thinks the consequences of
a person’s acts should be assigned to, one should assign them to the same
category or categories as the consequences of an earthquake or a scientific
discovery or a rise in the prime lending rate. There would seem to be two
serious candidates for this categorial office: ‘concrete event’ (for example,
Caesar’s death), on the one hand, and something proposition-like on the
other, ‘fact’, perhaps, or ‘state of affairs’ (for example, the fact that Caesar dis-
regarded the soothsayer’s warning, or Caesar’s having chosen to believe that
Brutus’ strong republican sentiments would never overcome his friendship
with and personal loyalty to Caesar).
Van Inwagen on Free Will 221
The point on which he has changed his mind is this: he now thinks that
Rule , as he presented it in An Essay on Free Will and other places, is wrong.
This rule of inference was stated as follows:
p, and no one has, or ever had, any choice about that. If p then q, and
no one has, or ever had, any choice about that. Hence, q, and no one
has, or ever had, any choice about that.
Van Inwagen had thought that this rule was obviously valid, had made use
of it in formulating one version of his argument for the incompatibility of
free will and determinism (his favorite version, the version inspired by Carl
Ginet’s “Might We Have No Choice?” [1966]), had said that, despite its
seeming obviousness, it was nevertheless the weakest link in the chain of
reasoning that led from the assumption of determinism to the conclusion
that no one is ever able to do otherwise. And, although this principle is
used in only one of the three formal versions of the argument he has
presented, van Inwagen is on record as saying that, in his opinion, if Rule
 should turn out to be invalid, it would almost certainly be the case that
the two versions of the argument that do not involve the concept ‘having
a choice’ would also turn out to be invalid.5
Imagine, then, his embarrassment when Tom McKay and David Johnson
(1996) presented a counterexample to Rule —a ‘beta blocker’, to use the
term that some wag (was it John Martin Fischer?) coined for counterexam-
ples to Rule . (Actually, McKay and Johnson presented a counterexample
to a different rule of inference, but since the invalidity of that other rule
implies the invalidity of , they have for all practical purposes presented a
counterexample to . It is not, by the way, entirely clear why they pro-
ceeded in this indirect fashion, since the idea behind their counterexample
to the other rule of inference can easily be adapted to produce a coun-
terexample to .6)
Van Inwagen (2000, 8–9) has presented a revised rule of inference that is
immune to McKay–Johnson-style counterexamples. Finch and Warfield
(1998) have proposed a different sort of revision that is also immune to
these counterexamples, and McKay and Johnson themselves suggested
some revisions that are, I think, workable. Unfortunately, these revised
rules, although they are far from implausible, lack the ‘luminous evidence’
that was a striking, albeit illusory, property of the original Rule . And this
fact can very properly prompt the compatibilist to ask a pointed question
224 P. van Inwagen
along the following lines: “If the apparent intuitive obviousness of the orig-
inal Rule  turned out to be an illusion—since the rule has in fact turned
out to be invalid—how much confidence should we have in the revised ver-
sions? Nemesis, in the form of a counterexample, was all along lying in
wait for Rule ; why should we not take very seriously the epistemic possi-
bility that in some dark corner of logical space, cousins of this Nemesis
patiently await their appointments with the revised rules?” This pointed
question is pointed indeed. Van Inwagen admits that he has no good
answer to this question, and that, in consequence, although he accepts the
revised rule, he assigns its validity a rather lower subjective probability
than the near certainty he once so confidently assigned to the validity
of the original . He does, however, withdraw his assent to a thesis I men-
tioned a moment ago, the thesis that if Rule  should turn out to be invalid,
this would mean that there was almost certainly something wrong with the
other arguments he has given for the incompatibility of free will and
determinism. He withdraws his assent because the counterexamples that
have shown the invalidity of Rule , the McKay–Johnson counterexamples,
depend on some unexpected properties of an English phrase—“has a
choice about”—that played a key role in his formulation of the original
principle. He is inclined to think that the ‘general idea’ behind Rule  was
sound, and that its invalidity stemmed from the fact that certain features of
the English phrase unfit it for the task he assigned to it. When he made the
statement from which he has now withdrawn his assent, he was assuming
that if Rule  were shown to be invalid, this would be because someone had
shown that the general idea behind the principle was fundamentally de-
fective. But the revised versions of the rule appeal to this same general idea;
they are merely (it is hoped) adequate implementations of this idea, imple-
mentations from which a technical defect has been removed.
I have said that in recent times, van Inwagen has changed his mind about
one point and has become increasingly insistent on another. I have dis-
cussed the point on which he has changed his mind. The point on which
he has become increasing insistent is this: free will is a mystery, a ground-
floor, first-water, Colin McGinn–style philosophical mystery. Free will is a
mystery because, although it obviously exists—of course we sometimes
confront a choice between A and B and are, while we are trying to decide
whether to do A or to do B, able to do A and able to do B—it seems to be in-
compatible with both determinism and indeterminism, and thus seems to
Van Inwagen on Free Will 225
And the falsity of any of these propositions would be, in his view, an even
greater mystery than the falsity of the proposition that free will is compat-
ible with indeterminism. Second, there are lots of philosophical mysteries,
most of which have nothing to do with free will. I might cite one or more
of the great philosophical mysteries, such as ‘temporal passage’ or human
consciousness. But I will mention instead two minor mysteries that are no
less mysterious for being minor. I will mention two mysteries that carry
much less emotional weight than the mystery of free will, or are, as one
might say, unconnected with anything we care deeply about—unlike free
will and time and consciousness, which are connected with many things
that we care deeply about: the mystery of vagueness and the mystery of the
Liar. It is obviously true that a currently living American male who is seven
feet four inches tall is tall, and is tall without qualification (i.e., is not
226 P. van Inwagen
important point, to be sure), appreciates the depth and difficulty of the free
will problem, while the majority of van Inwagen’s fellow incompatibilists,
although they are right on a particular point (and a very important point it
is), do not really appreciate the depth and difficulty of the problem.
I want to close by explaining why van Inwagen thinks one important
group of incompatibilists, those who appeal to what is called agent causa-
tion, do not appreciate the depth and difficulty of the problem of free will.
Many philosophers would agree with this judgment for the simple reason
that they think that the concept of agent causation is incoherent, or
think that agent causation is metaphysically impossible. Van Inwagen is in-
clined to agree with them (although he has no firm opinion on this ques-
tion), but he has lately stressed a different point. It is this: suppose there is
nothing conceptually or metaphysically impossible about agent causation;
suppose in fact that agent causation is a real phenomenon and that an
episode of agent causation figures among the antecedents of every volun-
tary movement of a human hand or limb or vocal apparatus. Van Inwagen’s
position is that even if this is so, and even if (as some have argued) we un-
derstand the concept of agent causation at least as well as we understand the
concept of event causation, all this does nothing to diminish the mystery of
free will. I will try to explain why van Inwagen thinks this by considering a
particular human action. Suppose Marie wants to vote in favor of the pro-
posal before the meeting, and that, for this reason, she raises her right hand
when the chair says, “All in favor . . .?” Suppose that one of the causal an-
tecedents of her hand’s rising was a certain event in her brain that was
undetermined by past events, that the state of her body and her immediate
environment at the moment this brain event occurred was causally suffi-
cient for her hand’s rising, that if this event had not occurred, her hand
would not have risen, and that she, Marie, a particular member of the meta-
physical category ‘substance’ or ‘continuant’, was the cause—that is to say,
the agent cause—of that crucial brain event. The friends of agent causation,
if van Inwagen understands them, believe that these suppositions are suffi-
cient for her having freely raised her hand. If that is so, these suppositions
must entail the following proposition: at some moment shortly before
Marie raised her hand, she was able to raise her hand and she was able not
to raise her hand. But van Inwagen doesn’t see why this entailment should
be supposed to hold. In fact, he thinks he sees a good argument for the con-
clusion that it was not up to her whether her hand rose. Suppose God were
228 P. van Inwagen
miraculously to return the world to precisely the state it was in, say, one
minute before Marie raised her hand, and that he then allowed affairs
once more to proceed, without any further miracles. What would happen?
What would Marie do? Well, if her raising her hand was a free act, and if free
will is incompatible with determinism, then we can’t say. We can say only
that she might have raised her hand and might not have raised her hand. If
God were to cause this episode to be thus “replayed” a very large number of
times, it might turn out that she raised her hand in thirty percent of the
replays and refrained from raising it in seventy percent of the replays. This
much is a simple consequence of incompatibilism, and it brings one of the
main reasons philosophers become compatibilists into stark relief. It seems
to lead us inescapably to the conclusion that on each particular replay, what
Marie does on that occasion is a mere matter of chance. And if there are no
replays, if there is only one occasion on which Marie is in this situation, it
seems to lead us just as inescapably to the conclusion that on that one occa-
sion what Marie does is a mere matter of chance. And if it is a mere matter
of chance whether Marie raised her hand, then it cannot have been true be-
forehand that Marie was both able to raise her hand and able to refrain from
raising her hand, for to have both these abilities would be to be able to de-
termine the outcome of a process whose outcome is due to chance. It is true
that we have, by stipulation, inserted into this process, this process whose
outcome is due to chance, an episode of agent causation. But, if I may so ex-
press myself, so what? That doesn’t change the fact that the outcome of that
process was due to chance. If God caused Marie’s decision to be replayed a
very large number of times, sometimes (in thirty percent of the replays, let
us say) Marie would have agent-caused the crucial brain event and some-
times (in seventy percent of the replays, let us say) she would not have.
Surely, then, whether she agent-caused the brain event was a mere matter of
chance? Whether her deliberations were followed by her agent-causing the
brain event was, it would seem, a matter of chance; Marie, therefore, cannot
have been both able to agent-cause the brain event and able to refrain from
agent-causing the brain event, for to have both these abilities would be to
be able to determine the outcome of a process whose outcome was due to
chance—an impossible ability. I conclude that even if an episode of agent
causation is among the causal antecedents of every voluntary human ac-
tion, these episodes do nothing to undermine the prima facie impossibility
of an undetermined free act. Postulating agent causation, therefore, does
Van Inwagen on Free Will 229
Notes
5. At any rate, he is on record as saying things that commit him to this position. In
van Inwagen (1983, 57), he said, “I am quite sure that any specific and detailed
objection to one of the arguments can be fairly easily translated into specific and
detailed objections to the others; and I think that any objection to one of the argu-
ments will be a good objection to that argument if and only if the corresponding
objections to the others are good objections to them.” One of the arguments referred
to as “the arguments” in this quotation is the argument that explicitly appeals to
Rule ; and he has said (van Inwagen 1989, 405) that if that argument is unsound,
its unsoundness must be due to the invalidity of Rule . And he has said (van
Inwagen 1994, 95), “My position is that all (logically adequate) arguments for incom-
patibilism must make some sort of implicit or hidden or covert appeal to [Rule ].”
References
Finch, A. and T. A. Warfield. 1998. “The Mind Argument and Libertarianism.” Mind
107: 516–528.
van Inwagen, P. 1974. “A Formal Approach to the Problem of Free Will and
Determinism.” Theoria 40: 9–22.
. 1994. “When the Will Is Not Free.” Philosophical Studies 75: 95–113.
. 1997. “Critical Notice: John Martin Fischer’s The Metaphysics of Free Will.”
Philosophical Quarterly 47: 373–381.
John Perry
. . . those who accept that responsibility for a situation implies an ability to bring it
about and, perhaps, an ability to prevent it, must explain how agents are able to do
other than they are caused to do. Without it, they can give no defense of their coun-
terexamples. With it, they can be confident that the Consequence Argument, by
itself, is no refutation of their position.
—Tomis Kapitan (2002, 154)
1 Introduction
Compatibilism is the thesis that an act may be both free and determined
by previous events and the laws of nature. I assume that in normal cases
a condition of a person’s performing an act freely is that the person is able
to refrain from performing the act. Thus, I accept that if determinism en-
tails that agents do not have this ability, we must give up compatibilism.
In this paper I try to contribute to the rethinking of compatibilism by dis-
tinguishing between strong and weak accounts of laws and strong and
weak accounts of ability. I argue that compatibilism is a tenable position
when combined with either a weak account of laws, or a weak account of
ability, or both. I shall concentrate on influential arguments for incom-
patibilism due to Peter van Inwagen, often called collectively the ‘conse-
quence argument’.
Some versions of the consequence argument seem to rely only on in-
escapable modal principles. In his excellent review and discussion of these
arguments, Tomis Kapitan concludes that these principles are not so logi-
cally inescapable as to completely trap the compatibilist (Kapitan 2002).
This is not to say van Inwagen’s arguments are fallacious, simply that they
rely on certain principles about causation and ability that have not yet
232 J. Perry
been fully articulated and defended. Kapitan says, just before making the
remarks quoted above, that his assessment provides the compatibilist with
“momentary breathing room at best” (2002, 154). I am trying to take ad-
vantage of the momentary breathing room afforded by Kapitan to explore
and to a certain extent defend options available to the compatibilist.
Using terms I explain below, my position is that van Inwagen’s arguments
do show that the combination of compatibilism with a strong account of
laws and a strong account ability (as I define these terms) is not tenable.
The options, then, are a weak account of laws, a weak account of ability,
or both.
2 Basic Argument
Prior to being given some quite persuasive argument, we do not take doing
something as a necessary condition of being able to do it, or being able to do
something as a sufficient condition for doing it. We do not take a person’s not
doing something as a sufficient condition for their being unable to do it. We ac-
cept the inference from ‘cannot’ to ‘does not’ and from ‘does’ to ‘can’. But we
do not accept the inference from ‘does not’ to ‘cannot’ or from ‘can’ to ‘does’.
If Elwood did eat a cookie at t 1, that would prove that one of the
premises 1, 2, or 3 is false. But the fact that he can eat a cookie does not
show that one of the premises is false. It merely shows that if he did do
that thing that he is capable of doing, one of the premises would be false.
Hence, the truth of the premises rules out his eating a cookie, but not his
having the ability to eat a cookie.
Now suppose we add another premise, to the effect that (1) is not merely
a true universal generalization, but something that follows from the laws of
nature. We derive (1) from premise (0):
3 Some Preliminaries
It might be better to say it be true, using a tenseless form, and in fact I shall
do so from now on, and say things like
If a proposition is ever made true (or made false), then it be true (or be
false).
The fact that a proposition has not yet been made true (or made false) by
events, does not imply that it be not true (or be not false).
Compare:
The fact that Gore has not yet been chosen as our next President does
not imply that he is not our next president.
If Gore ends up being nominated and elected in 2004, then he is our next
president. If I call him ‘our next president’ now, I’m correct if the future
goes one way, incorrect if it goes the other. It is possible to become rather
puzzled by this. How can Gore be our next president now, if it hasn’t been
decided yet? So he must not be our next president. By parity of reasoning,
no one is our next president. That will be a constitutional crisis. But we can
avoid the crisis. Being our next president is a property Gore has if at some
point between now (summer, 2002) and January 2005 he is elected and in-
augurated and Bush hasn’t been replaced in the meantime. It all works out,
as long as we are careful about the difference between being our next pres-
ident, a property Gore may have, and being elected and inaugurated as our
next president, one that he does not yet have as I write this, in 2002. The
fact that lots of propositions be true that have not yet been made true is
sort of like that. It can be confusing. It may well be that from a metaphysi-
cal point of view our two-valued logic of propositions may not be optimal.
Still, if we are careful, things will work out.
It will be useful to have the following locution available:
Perhaps we should simply say ‘make whether P’, but that sounds even worse.
It seems like there are lots of important propositions whose truth value is
established not by being made true by events, but in some other way. For
example, consider Pythagoras’s theorem, the proposition that the square of
the hypotenuse of a right triangle is equal to the sum of the squares of the other
two sides. No event has ever made this proposition true, and none ever will.
It’s not at all like the proposition that Nixon won the 1972 election. There is
236 J. Perry
The proposition that Eisenhower was president in June, 1953 be true because
of events that occurred prior to 1954, including his election in 1952 and
inauguration in 1953.
The proposition that Gore will be president in 2006 be true, if it be, partly
because of events that have yet to occur.
With propositions such as that Pythagoras’s theorem, a quite different kind
of explanation of their truth seems appropriate, and of course there is a lot
of philosophical controversy about what the correct explanation might be.
I’ll simply say that such propositions are not made true by events, and leave
it at that for now.
Finally, and importantly, suppose that a true proposition, that is not
made true by events, together with some other propositions, that have
already been made true by events, entail a proposition that has yet to be
made true. What should we say about that proposition? To return to our ex-
ample, suppose that the laws of nature are not made true by events, and that
these laws, together with propositions made true by events that have
already happened, entail that Elwood will not eat cookies at future time t. I
will say that although the proposition that Elwood will not eat the cookie at
t has not yet been made true, it’s truth value has been settled. The proposi-
tion won’t be made true until the events that the laws of nature and the past
determine have actually occurred: the time t has arrived, and Elwood says
“no” to the proffered cookie, keeps his arms at his sides, and walks away. But
these events were already entailed by a combination of propositions some of
which were already made true and the rest of which aren’t made true by
events at all. So the truth value of the proposition that he would not eat
was, in that sense, already settled, before he refrained from eating.
Compatibilist Options 237
While by 1975 or so this may have seemed like a decree of God to the
Nottingham fans, we’ll assume for now that it was really just a sad but
true generalization. Suppose that in the second game between these two
teams in 1978 they were in a 0 to 0 tie at the end of the game, and then
Manchester won in overtime. In the last second of regulation play, Not-
tingham had a clear and easy shot on Manchester’s goal, that their best
player missed. Nottingham’s worst player was watching a plane overhead
skywriting advertisements for Guinness, and ran into their best player just
as he kicked. I think the Nottingham fans at least would think that the
Nottingham team could have won that game, even though they did not.
And that means they could have made the true generalization G false.
I think it is common sense to suppose that laws are not simply true
generalizations. Suppose that one Nottingham fan says ‘We could have won
that game in 1978 . . .’ Another particularly bitter Nottingham fan says,
No, you are wrong. You have not grasped that it is an unshakable law of
nature that Manchester United beats Nottingham Sherwood. It’s a law of
nature because God decreed that Nottingham would always lose. Laws
of nature are universal generalizations that God issued as fiats during
creation week, and other things that follow from them. For reasons finite
mortals can’t be expected to fathom, he often punishes the virtuous and
rewards the wicked. And in this spirit he decreed that Manchester
United would always beat Nottingham Sherwood. It seemed like
Nottingham could win, but in fact it could not.
This fan’s remark embodies the intuition behind the argument of section 2.
Intuitively, laws are more than true generalizations, and (0) adds something
substantial to the argument. You can make a mere generalization false;
even if no one gets around to doing so, it remains true that someone could
have. But laws are laws. You cannot make a generalization false, if it follows
from the laws of nature. This is the strong conception of laws. But of
course, one can have a strong conception of laws without believing in an
unfathomable God or any god at all.
Let’s remind ourselves of (1):
(1) ∀t ∀x ((x, t) → (x, t 1))
If (1) is true then there have not been, and will not be, times t and individ-
uals x such that (x, t) and ~(x, t 1). And (0) says that (1) is not simply
true, but true according to the laws of nature.
Compatibilist Options 239
The question is, does (1) make (0) true, or does (0) make (1) true? Is (0)
true because (1) has no disconfirming instances? Or does (1) not have dis-
confirming instances because of the truth of (0)—because (1) follows from
the laws of nature? Is the truth of (0) one of those things, like the truth of
Pythagoras’s theorem, that is established by something other than what
happens? Is it the sort of things to which events conform, but do not make
true? Or is it just a sort of fancy way of saying (1)?
To return to our soccer fans. A third soccer fan, also a fan of Hume’s, may
say:
I don’t think (G) is a decree of God. But I agree with you that it is a
law of nature. A law of nature simply is an exceptionless generalization
that we have grown used to so that it shapes our expectations. And (G)
certainly is that.
This remark would express a weak conception of laws. It isn’t quite enough
for (G) to be a law that it is true. More is required: that we use it to form our
expectations. But that’s all. There is no big metaphysical condition, like a
command from God, which is also required. Being a law is just being a true
generalization that we have internalized so that our expectations about the
future are shaped by it. On this conception, it is (1) being true that ex-
plains, or partly explains, that it is a law. (1) Explains (0), not the other way
round.
On either conception of laws, laws will have no disconfirming instances.
On the strong conception, the fact that L is a law explains why events con-
form to it; the truth of the law is due to something other than the lack of
disconfirming instances. If determinism is also true, laws and propositions
about the past not only entail propositions about the future, but also settle
them.
On the weak conception of laws, however, the incompatibilist argument
does not work. (0) adds nothing to the argument that might push the con-
clusion from (4) to (5). A Humean conception might add something to the
requirement for being a law. (1) not only has to be true, but accepted and
used to guide our expectations. But this doesn’t push us from (4) to (5).
One option for the compatibilist, then, is to insist on this very weak,
Humean conception of laws. What we do is up to us; laws are merely those
descriptions of what we do that will end up being true once human actions
are complete. Laws determine, but do not settle. I’ll call this view ‘existen-
tialist compatibilism’.
240 J. Perry
A strong theory of can supports (S), while a weak theory does not. I’ll argue
for a weak theory, and explain why it undercuts van Inwagen’s arguments.
First an analogy.
It’s 1956 and Elwood doesn’t buy a new Edsel. He thinks they are ugly, un-
gainly, and overpriced. He doesn’t want one. So he doesn’t buy one. Now
does it follow that he can’t afford one? Of course not. He may have all the
money he needs, and simply not want one. One question has to do with
what he wants in a car and what he thinks about the Edsel. These facts, what
he thinks about Edsels and what he wants in the way of a car, are pretty
much located in Elwood’s head. At any rate, they are not located down the
street at the bank. But that’s where the facts about how much money he has
in his account, and how much credit the bankers will give him, reside. It
242 J. Perry
may be that Elwood would rather be drawn and quartered or have rats gnaw
out his eyes than buy an Edsel. But those facts about his mind don’t tell us
anything about his bank account. He may be loaded, so he can easily afford
a fleet of Edsels. He can’t buy the car without money or credit, but he can
not buy the car even though he has plenty of money and plenty of credit.
We could put forward a little argument that Elwood won’t buy an Edsel:
(1) Reasonable people don’t buy cars that they think are ugly, ungainly
and overpriced and that they simply don’t want and have no other
reasons to buy (law of nature).
(2) Elwood thinks Edsels are all of those things, and has no other
reason to buy one (fact about Elwood’s mind).
proposition that rules out J having raised his hand at t. I’ll just take Q to be
the proposition that J did not raise his hand at t. Call the laws of nature L
and call the facts up until t that are relevant PF. Assume that we cannot
change the past. Then,
This proposition entails that J does not raise his hand at t. J can render the
proposition Q false by raising his hand at t. If he renders Q false, R be false,
too. But R may have already been rendered false by the time J renders Q
false. This will be the case if J’s mother did not eat a carrot in 1944. In this
case, J will not render R false, even though R entails Q and he renders Q
false. It simply does not follow from the fact that J will render Q false that
he renders false every proposition that entails Q. What does follow is that
there is no true proposition that entails Q.
Principles (i) and (ii) are clearly true, given the coherent concept of
‘render true’ and ‘render false’.
(i) Suppose one does something that renders P true. Then no proposition
that entails the falsity of P be true.
(ii) Suppose one does something that renders P false. Then no proposition
that entails P be true.
(iii)* Suppose one can do something that would render P true. Then no
proposition that entails the falsity of P be true.
(iv)* Suppose one can do something that would render P false. Then no
proposition that entails P be true.
Principles (iii)* and (iv)* simply amount to the principle that there is no
difference between being able to do something and doing it—that can
collapses into does, and does not into cannot.
If I can drink a beer, I can render that I drink a beer true. So, given (iii)*, if
I can drink a beer, no proposition that entails that I don’t drink a beer is true.
248 J. Perry
So if I can drink a beer, that I don’t drink a beer isn’t true (since it entails
itself), so it’s false, so I drink a beer. If I can do it, I do it. Can implies does.
Suppose I don’t drink the beer. Then, that I don’t drink the beer is true.
Then something is true that entails the falsity of that I drink the beer. Then,
by (iv)*, I can’t render it true that I drink the beer. So I can’t drink the beer.
Doesn’t implies can’t.
Such a collapse of ‘can’ into ‘does’ and ‘doesn’t’ into ‘can’t’ is, of course,
just what the incompatibilist wants and the compatibilist needs to avoid. If
we accept (iii)* and (iv)*, the collapse would be completed without any
appeal to determinism at all. But of course there is no reason to accept (iii)*
and (iv)*. On the contrary, it seems quite clear that on the weak conception
of ability, (iii) and (iv) are true instead:
(iii) Suppose one can do something that would render P true. This does not
imply that no proposition that entails the falsity of P is true.
(iv) Suppose one can do something that would render P false. This does not
imply that no proposition that entails P is true.
Suppose J can raise his arm at t, but decides not to. Then that J does not raise
his arm at t is true. This proposition entails itself. So J can raise his arm at t,
even though a proposition that entails that he does not raise his arm at t is
true. So (iii) is correct.
Suppose J can refrain from raising his arm at t, but in fact he raises it.
Then that J raises his arm at t is true. This proposition entails itself. So J can
refrain from raising his arm at t, even though a proposition that entails that
he does raise his arm at t is true. So (iv) is true.
Now let’s return to the crucial part of van Inwagen’s argument:
L that no one who at t 1 really really does not want to raise his hand
in the next instant, raises his hand at t.
Q that J does not raise his hand at t.
This example meets the conditions of van Inwagen’s argument. That is,
PF & L entails Q.
We can certainly accept steps (2) and (3), given our understanding of
‘render Q false’. But step (4) does not follow.
PF is the proposition that at t 1 J really really wants to not raise his
hand in the next instant. So (4) says that if J could render Q false (i.e., if he
could raise his hand at t), then he could render false the proposition:
that L & at t 1 J really really wants to not raise his hand in the next instant.
But there is nothing that J can do at t, the doing of which would make it the
case that it was not true at t 1 that he really really wanted not to raise his
arm at t.
If J does raise his hand at t, that will show, given L, that PF is not true.
However, that will not make PF untrue; it will not render PF untrue. If he
raises his hand at t, that will be because he is in some state at t 1 than re-
ally really wanting not to raise his arm, perhaps in the state of wanting
to raise it. In this case, PF be untrue, but it be untrue because the events at
t 1 made it false, not because of what J does at t.
If we go back to our simple picture of what it is to be able to raise your
hand at t, this should be reasonably clear. There are two facts about J and
raising his hand, with these possible combinations:
Does raise his hand Does not raise his hand
The argument starts with the premises that J does not raise his hand, that is,
he is in cell 2 or cell 4. It then hypothesizes that he can raise his hand,
putting him in cell 2. From this it is supposed to follow that he changes the
past, since the past determines that he will not raise his hand. But it clearly
does not follow, for in cell 2 he does not raise his hand, just as the past
determines will happen.
I conclude, then, that as long as we have a weak, but realistic and com-
monsense concept of ability, we can be determinists and compatibilists,
even if we accept a reasonably strong conception of laws.
250 J. Perry
(␣) p → Np
() From Np and N(p→q) deduce Nq,
where ‘Np’ means ‘p and no one has, or ever had, any choice about
whether p’.
We are thinking of p and q as propositions, and entailment as a relation
between propositions. It seems we should accept
if p entails q then (p → q).
Then if we also accept (␣) and () we’ll have to accept the rule,
() From Np and p entails q deduce Nq.
with the laws of nature, determine what he does, the fact that he will or
will not do something does not preclude his having the ability to refrain or
not refrain from doing it. The weak theorist thinks that a person can have
a choice about something, in the sense that they have the ability to do it or
refrain from doing it, even if that thing is determined by laws of nature that
are established and facts that are already made true. The weak theorist,
then, having rejected (S), need have no qualms about rejecting ().
7 Lewis’s Analysis
In “Are We Free to Break the Laws?” David Lewis distinguishes between the
following claims: ‘I am able to do something such that, if I did it, a law
would be broken’ and ‘I am able to break a law’ (Lewis 2001, 31ff). Suppose
the laws of nature and the history of the world up until time t 2 entail
that I will not take the glass of water at t, but I don’t. Suppose, as Lewis
does, I cannot change the past. There seem two possibilities:
(a) Something happened at t 1 that was contrary to the laws of nature,
that is, a ‘divergence miracle’.
(b) Everything up to and including t 1 was in accord with the laws of
nature, but my action was not.
Lewis thinks the fact that I can take the glass of water implies that I am able
to do something such that, if I did it, a law would have been broken at
some earlier time, but this requires only (a). He does not think I am able to
break a law, which would require (b).
I do not think the compatibilist need suppose that if I were to take the
drink, any laws of nature would ever need to have been broken. There are
auxiliary premises from Lewis’s metaphysics and analyses of causation,
counterfactuals, and the like that lead him to this defense of compatibil-
ism. But compatibilism by itself does not force us to the divergent miracles
defense, and it does not seem to me the most plausible thing to say about
cases in which one has the ability to do differently than one does.
If I had taken the drink, freely and voluntarily, then surely my beliefs and
preferences would have been different than they actually were. The most
likely difference would be that I was thirsty. Assuming determinism, if I
had been thirsty when the drink was offered, then something earlier also
would have been different; perhaps I wouldn’t have taken a drink at the
252 J. Perry
fountain before stepping on the plane, as I did, because the fountain was
broken. And that would mean some earlier state of the fountain and its sur-
roundings had been different. And so on. Tracing the changes back to the
big bang, perhaps it might be a slight difference in the direction in which
one particle began its travels through time. Or perhaps it goes back to a
deistic god creating the initial state of the universe a very little bit differ-
ently. Or perhaps it just goes back, infinitely. Who knows? It’s certainly
amazing and weird and in my opinion somewhat depressing that the trail
of differences that would have led to my being thirsty rather than not
being thirsty should lead back even a couple of thousand years, much less
to the beginning of time, or forever. Still, I can’t see why either (a) or (b) is
required for me to take the glass. Assuming determinism, it follows from
the fact that I can accept the drink and don’t, that I can do something such
that if I did it either the laws of nature or the events up until that time
would have been different than they in fact are. It does not follow that if I
did what I can do any law would thereby be broken, or any divergence
miracle would ever have occurred, or I would have changed the past in any
way. I wouldn’t have had to change the past, because, according to deter-
minism and the laws of nature, if I had been thirsty, the past would have
been different.
8 Conclusion
Acknowledgments
Early versions of this paper were presented at the 2001 Inland Northwest
Philosophy Conference and the Philosophy Department Colloquium at the
Compatibilist Options 253
Notes
1. Let me emphasize that the right and left sides are chosen completely arbitrarily,
simply as a way of easily visualizing the point. This is not an attempt to fit agency
into what is known, or thought, or imagined, or claimed, about the differences be-
tween the right and left side of the brain.
2. Note that even if we adopted the more complicated account of the issues
discussed in section 3, so that propositions were neither true nor false until events
made them so, making a proposition false would not mean changing its truth value
from true to false. If we did things this way, we would have to say that when a set of
premises entail a proposition about the future, the truth of the premises requires that
the proposition will be true, not that it is true.
References
Ekstrom, L. W., ed. 2001. Agency and Responsibility: Essays on the Metaphysics of
Freedom. Boulder, Colo.: Westview Press.
Israel, D., J. Perry, and S. Tutiya. 1993. “Executions, Motivations and Accomplish-
ments.” Philosophical Review (October 1993): 515–540.
254 J. Perry
Kane, R., ed. 2002. The Oxford Handbook of Free Will. New York: Oxford University
Press.
Lewis, D. 1981. “Are We Free to Break the Laws?” Theoria 47: 113–121. Reprinted in
Ekstrom (2001): 30–37. (Page numbers refer to this latter work.)
Richard Feldman
There are striking similarities between the debate about freedom and
determinism and the debate about knowledge and skepticism. In each case,
the issue can sensibly be set up as a puzzle about an inconsistent triad of
propositions, each of which seems to be true. Furthermore, there are posi-
tions one can take about one of these puzzles that have their analogues in
positions one can take about the other. An optimist might think that in-
sights achieved in one area will shed some light on the other. Such a person
is, of course, doubly optimistic, thinking both that there are insights in the
one domain and that these insights might be of value in the other. Some
philosophers have thought that contextualist theories in epistemology
have just this sort of value. John Hawthorne (2001) has recently defended
this line of thought in print. While I agree about the potential value of in-
sights in one domain for the other, I think that contextualism is of limited
value in helping us to think about either freedom and determinism or
about skepticism. I have previously argued for the limited value of contex-
tualism in epistemology (Feldman 1999, 2001). In this paper I will summa-
rize my arguments about contextualism and skepticism and then extend
those arguments to contextualism and freedom.
Each of these three propositions seems, at first glance to be true. (A1) re-
ports the sort of everyday knowledge we all think we have. (A2) is true since
I have no way to know that I’m no brain in a vat. Things could seem just as
they do if I were a brain in a vat.
(A3) requires a bit more explanation. First, the contained propositions in
(A1) and (A2) should be understood in such a way that they are incompat-
ible. That is, being a mere brain entails not having hands. The idea under-
lying (A3) is a principle of closure, roughly to the effect that if I know a
proposition P, and P entails Q, then I know Q. There are many details here
that need attention if a fully adequate closure principle is to be stated.
For one thing, the closure principle should almost surely be restricted to
known entailments. That is, if I know P, and I know that P entails Q, then
I know Q. These details are important, and they may get in the way of gen-
erating any puzzle about skepticism based on the closure principle. How-
ever, the points I want to raise in this paper do not turn on details about the
closure principle, so I will not go into them here.
Stewart Cohen (1987, 1988, 1999) and Keith DeRose (1995), two leading
defenders of contextualist theories in epistemology, set up their problem in
roughly this way. The burden of their theories is to explain what to make
of the seemingly incompatible intuitions that each of (A1), (A2), and (A3)
is true. Cohen explicitly makes the issue one about an inconsistent triad.
DeRose instead formulates the issue as one about what to say in response to
the argument from (A2) and (A3) to the denial of (A1).
There are other ways to set up the problem of skepticism, also based on
inconsistent triads or their analogous arguments. For example, one could
note that with respect to any belief one has, or at least any belief one has
about the external world, there is some possibility of error. One could also
note that a person who allegedly knows something could be in a situation
that is introspectively indistinguishable from the situation of a person who
lacks knowledge. And considerations along these lines could be used to
construct an inconsistent triad. For example:
What should we say about these triads? Which elements should be given
up? As contextualists seem to see things, a central part of solving the skep-
tical problem is figuring out how to answer these questions.
Variants of the puzzles could be developed that did not rely on the univer-
sal generalization in (C2). For example, puzzles could be formulated by
saying of certain apparently free actions that they are causally determined.
Also, the puzzles could be formulated in terms of claims about decisions
rather than actions. Furthermore, analogous puzzles could be set up in-
volving the ability to do otherwise or being responsible for one’s actions
rather than acting freely. I will not consider these variations here. I believe
that remarks similar to those I make below apply to all such puzzles.
Each of (C1), (C2), and (C3) has some intuitive appeal. At least in our
unguarded moments, we all assume that (C1) is true. In what we say, and
what we do, we presuppose that we ourselves, and those with whom we
typically interact, are free agents. And we assume that, at least sometimes,
they could have done other than what they did and that they are respon-
sible for what they have done. (C2) also seems to be true, or at least close
enough to true for present purposes. Human beings are, according to
many, animals that are part of the natural world. Our behavior is deter-
mined, or at least as close to determined as is the behavior of everything
else around us. We are not, according to this common view, exempt from
the causal laws governing the natural world. (C3) reflects the apparent
conflict between (C1) and (C2). It has intuitive appeal, and there are much
discussed arguments, which I will not consider here, that can be given
in support.
So all three propositions seem true. Yet one must be false. Which should
it be? One way to conceive of the problem of freedom and determinism is
as the problem of figuring out what to say about this inconsistent triad of
propositions.
258 R. Feldman
2 Context Sensitivity
The idea of context sensitivity is simple and familiar. It is crucial for present
purposes to distinguish the sort of context sensitivity under discussion here
from a very different sort. One philosopher (Annis 1978) has argued that
for a person to know that some proposition is true requires that the person
be able to answer all the objections to that proposition that are available in
the setting in which the person is located. This view seems to have the im-
plication that someone might know some proposition to be true in one
setting but could lose that knowledge simply by moving to a place where
people have some particularly difficult objections to that proposition. On
this view, the setting, or context, of the putative knower—the subject’s
context—plays a crucial role in determining whether the person has
knowledge. On this view, a person with the same reasons for believing the
same true proposition might in one setting know that proposition, yet not
know it in another. This is not the view under discussion here.
Context sensitivity as it is invoked in recent discussions of skepticism
has to do with the speaker’s context. Allegedly clear examples of this kind
of context sensitivity involve familiar comparative words such as ‘tall’ or
‘wealthy’ or ‘big’. In a conversation about elephants we might say of
Jumbo, the baby elephant, that he is not big, since Jumbo is much smaller
than the average elephant. But when talking about the animals we saw at
the zoo, we might say that Jumbo is quite big, since he is bigger than al-
most all the animals at the zoo. One idea about what goes on here is that
sentences of the form ‘X is big’ are incomplete because they are implicitly
relational. Things are not simply big or not big; they are big or not relative
to some group or class: they are big for an animal at the zoo but not big for
an elephant. But language works in such a way that we do not always have
to mention the reference class explicitly. The prior conversation and other
features of the setting, all making up the context, enable participants to
understand the intended relativizations.
The sort of context sensitivity just described has to do with the speaker’s
context. The point can be illustrated most clearly by noting that two con-
versations about Jumbo might be going on simultaneously. In one, some
elephant experts are discussing the relative size of some elephants. One of
them says that Jumbo is not big. What the expert says is true. At that very
moment, a child and his family are visiting the zoo. The child has tired of
Freedom and Contextualism 259
seeing the little animals and asks if there are any big animals in the zoo.
The father says that there are, since Jumbo is big. This is true too. The
expert’s statement and the father’s statement do not conflict, even though
they are spoken at the same time. And there is nothing peculiar about
Jumbo’s size that plays any role in this. Nor is there anything about per-
spective or opinion that plays a crucial role here. Jumbo, we may assume,
is at the moment some fully determinate size. The predicate ‘big’ truly ap-
plies to a thing of Jumbo’s size in the context in which the father speaks. It
does not apply to a thing of that size in the context in which the expert
speaks.
The lesson to learn from this is a familiar one: the same sentence can be
used in different settings to say different things with potentially different
truth values.
Lewis’s solution, which fits well with his more general view about lan-
guage, is that the stock of possibilities which must be eliminated if we are
to have knowledge varies with context. In effect, his view pins the context
sensitivity of ‘knows’ on the context sensitivity of ‘all’. ‘All possibilities’ do
not include possibilities that we can properly ignore. In ordinary contexts,
we can eliminate all possibilities. This is in part due to the fact that in
ordinary contexts the odd possibilities envisioned by skeptics are properly
ignored and thus are not among ‘all’ possibilities. But in skeptical contexts
there are additional possibilities, and we cannot eliminate them. His view
is that various things that we say have the effect of enlarging the pool of
possibilities. For example, mentioning a possibility makes it not properly
ignored. The sort of contextualism about freedom that Hawthorne discusses
is modeled on this proposal from Lewis.
There is a detail worth mentioning about contextualist solutions to the
skepticism puzzles. I did not say that in ordinary contexts one would ex-
press a falsehood by uttering (A2) or (B2). The reason for this is slightly
tricky. Contextualists do not simply assert that there are differing standards
for the application of the word ‘knows’. They say something about what
brings about the raising of standards. (Curiously, less is said about what
brings them back to normal.) The leading idea is that talking about skepti-
cism and skeptical hypotheses raises the standards. So uttering sentences
like (A2), by mentioning the skeptical hypothesis that I am a brain in a vat
(or my lack of knowledge that I am not a brain in a vat), induces the higher
standards. So, if I utter (A2), then I have raised the standards and made it
true. But I do meet the ordinary standards for the application of the predi-
cate ‘know that I am not a mere brain in a vat’. So, we can say that relative
to the standards determined by ordinary contexts, (A2) is false. But if we
were to say (A2), it would be true. This illustrates the point David Lewis put
colorfully by saying, “Epistemology robs us of our knowledge” (1996, 550).
Freedom and Contextualism 261
Given the structural similarity of the puzzles about knowledge and about
freedom, it is not surprising that contextualist responses to the latter puz-
zles are also possible. In broadest outline, the idea is that the truth condi-
tions for sentences involving the key terms—‘free’, ‘morally responsible’,
‘could have done otherwise’—are context sensitive. In a recent paper,
John Hawthorne has taken this approach to sentences using ‘free’ (2001).
(In an older paper Terrence Horgan 1979 applied it to sentences involving
‘could have done otherwise’. I will not discuss his view here.) Although
Hawthorne makes some remarks about ‘moral responsibility’ he does not
claim that that phrase itself displays context sensitivity, and I know of no
Freedom and Contextualism 263
one who has. In this section I will spell out the key features of Hawthorne’s
response to the puzzle about freedom.
I should note at the outset of my discussion of Hawthorne’s paper that he
concludes by saying that he does not confidently believe the contextualist
analysis he discusses. Rather, he thinks it deserves to be “taken seriously”
and that it is a “worthy competitor to the more standard” views (2001, 77).
It will be easier to proceed here, as Hawthorne does in his paper, as if he
actually does accept the view he describes.
Hawthorne’s proposal is modeled on David Lewis’s version of contextual-
ism in epistemology. The idea is to make utterances of ‘S did x freely’ true in
ordinary contexts but not true in special philosophical contexts. Hawthorne
writes:
When ordinary speakers utter English claims of the form ‘S did x freely’ (and their
synonyms), they frequently speak the truth. But when our sphere of attention is
widened by philosophical inquiry, we are rarely in a position to truly utter the
English words ‘S did x freely.’ (2001, 68)
The idea is that in any context in which we talk about whether an action is
done freely, we are considering only a certain set of potential causal expla-
nations of the action. We say that an action is done freely when we think
that there is no causal explanation of it within that set, and, if we are right
about this, then what we say is true. In ordinary contexts, he suggests, we
properly ignore as causal explainers hidden facts involving the “neurologi-
cal springs of action” and explanations involving the laws of nature oper-
ating on “the distribution of microparticles in the distant past” (2001, 67).
In ordinary contexts the sorts of causal explanations that would preclude
the correct attribution of ‘free’ apparently include causal explanations
involving macro processes occurring outside the individual. If, for example,
I am caused to move to the left by a sudden gust of wind, then I did not
move to the left freely. Ordinarily, this sort of explanation is not properly
ignored. But if there is no such explanation of my move to the left, if the
only causal explanations involve neurological events that we rarely think
of, or complex processes going back to childhood and before, then there is
264 R. Feldman
The contextualist resolution to the puzzle here is quite similar to the contex-
tualist approach to the skepticism puzzle. It says that (C1) is true relative to
some contexts and false relative to others, that (C2) is false in contexts in
which (C1) is true and vice versa, and that (C3) is true in all contexts.
Consider ordinary contexts first. Here, only some sorts of causal explana-
tions are relevant. Some are properly ignored. In such contexts (C1) expresses
a truth. Some actions do not have ordinarily relevant causal explanations.
But in such contexts (C2) is false. It says that all human actions are causally
determined, but in an ordinary context, this is true only if all human actions
have causal explanations among the sorts of explanations relevant in such
contexts. They do not. So (C2) is false in such contexts.
It might be a little better for Hawthorne’s account if we think of (C2) a lit-
tle differently than we have so far. It is not immediately clear that the phrase
‘causally determined’ has the sort of context sensitivity under discussion.
The key idea is that all human actions have a causal explanation involving
factors the agent cannot control. But the domain of potential explanations
varies with context. In some settings, when only ordinary explanations
Freedom and Contextualism 265
count, there is no such explanation for some actions. But when we expand
the domain to include microphysical and ancestral explanations, then every
action does have a causal explanation involving factors beyond the agent’s
control.
Consider next philosophical contexts in which a much larger set of
causal explanations have been made relevant. In such contexts, (C1) ex-
presses a falsehood, on the assumption that there is some causal explana-
tion for all behavior. In such contexts, (C2) is true. Again, (C3) is true.
It may be that uttering a sentence like (C2) has the effect of making all
causal explanations relevant. That is, it may be that merely using a sen-
tence like (C2) puts us into a context in which higher standards for ap-
plying ‘free’ are in place. This, too, is analogous to the epistemic case.
Thinking about knowledge in terms of relevant alternatives makes the
analogy even clearer. Philosophy has the unhappy consequence of ex-
panding alternatives. Attributions of ‘knows’ require that the putative
knower be able to eliminate all relevant alternative possibilities. Thinking
philosophically about knowledge expands the set of alternatives we must
be able to eliminate so much that it includes remote skeptical hypotheses
that we are unable to eliminate. So, in such contexts, attributions of the
word ‘knows’ are typically false. Similarly, attributions of ‘free’ require that
there be no causal explanation of the action in question among the rele-
vant causes. Thinking about determinism can expand the set of relevant
causal explanations to include all the neurological and remote historical
explanations that we typically ignore. There always are causal explana-
tions in this expanded set. So, in such contexts, attributions of the word
‘free’ are typically false. Adapting Lewis’s remark about epistemology, we
can say that the idea is that metaphysics robs us of our freedom.
thing. When what turns on the truth of a sentence increases, there may be
some tendency to be more cautious about asserting it. Thus, it may be that
hearing about skeptical possibilities also has the effect of making us more
cautious about predicating ‘knows’ without a change in the meaning of
the word. In addition, saying that someone knows something suggests
that the person could defend the thing known from objections. (I do not
mean to suggest that knowledge implies having this ability.) By mention-
ing odd scenarios one calls to mind possible objections that may be harder
to respond to. Realization of this could make us less willing to attribute
knowledge. It need not change the truth value of such attributions.
It is also possible that consideration of skeptical hypotheses and argu-
ments affects our willingness to attribute knowledge for a different reason.
Reflecting on those hypotheses and arguments may make salient reasons
for thinking that we lack knowledge, reasons that are temporarily persua-
sive. There are some propositions that at least some of us do not have stable
attitudes about. Perhaps propositions ascribing knowledge are like that, so
we change our minds.
The contextualist’s claim that knowledge sentences are context sensitive
does not strike me as clearly true. There are plausible alternative explana-
tions of our varying inclinations.
about that. I think that the interesting issue concerning skepticism has to
do with whether we actually do meet the ordinary, lower standards for
knowledge. The most powerful skeptical arguments conclude that we do
not meet those standards. It is thinking through that issue that strikes me
as the fundamental challenge posed by skepticism. And contextualism is
entirely irrelevant to that issue. I should be careful about what I mean by
this. Contextualist philosophers may have something to say about what the
ordinary standards for knowledge are and why it is true that we meet them.
That would be of value in responding to skeptical arguments designed to
show that we do not meet the ordinary standards. The fact that there are
some other extraordinarily high standards that we do not meet is simply
irrelevant to this fact.
Here is another way to put the point. One central part of the challenge of
skepticism arises from the need to respond to arguments alleging that our
reasons for believing much of what we ordinarily believe are not very good,
not good enough for knowledge and perhaps not good enough for some
lower level of justification. These arguments prompt us to consider carefully
the reasons we have for believing the things we typically believe. One such
argument asserts that our ordinary beliefs provide no better explanations of
our basic perceptual data than do rival explanations involving dreams,
hallucinations, evil gods, or computer simulations. Related arguments are
based on the claim, endorsed by Lewis, when he asks the rhetorical ques-
tion, “What (non-circular) argument supports our reliance on perception,
on memory, and on testimony?” (1996, 551).
There are in the philosophical literature a variety of responses to the
arguments I’ve just alluded to, none entirely satisfying. The existing views
seem to amount to stipulation that we do know what we think we know,
avowals of faith in the powers of human cognition, questionable appeals to
inference to the best explanation, the assertion that commonsense views
are ‘intrinsically rational’ and so on.
I think that some accounts of knowledge, those that drop justification or
good reasons as necessary conditions of knowledge, also contribute to less
than edifying responses to skepticism. They may make it true that we have
knowledge, but seem to concede that our reasons for our ordinary beliefs
are not as good as we thought they were. To concede this, but nevertheless
insist that we have knowledge is, as I see it, to win the battle but lose the
war. Skepticism is centrally about whether we have good reasons to believe
270 R. Feldman
claims to be free are true. I doubt that they do. Those who get worried
when they encounter arguments for the lack of freedom, such as the conse-
quence argument, should, if contextualists are right, think that the conclu-
sion that argument leads them to is not the denial of what they were
inclined to assert prior to first hearing the argument. But, in my experience
at least, no one does.
Finally, there are alterative explanations of the varying intuitions and
linguistic inclinations contextualists say we have. It is possible that focus-
ing on certain facts or arguments causes varying reluctance to assert the
same thing. It is possible that varying salient considerations make us some-
times believe and sometimes deny the same thing. There is, at least, room
for debate about these matters. We can wonder what best explains these
varying intuitions.
There is, then, room to doubt that there is context sensitivity here at all.
This issue is much more complex than my treatment here touches on. One
complication is that the word ‘free’ is ambiguous as well. There is the sort
of metaphysical freedom philosophers concern themselves with, auton-
omy that some psychologists focus on, as well as political freedom. This
makes things even messier.
But suppose that ‘free’ is context sensitive. Does that help us see our way
through the issues?
No one ever acts freely. No one is morally responsible for their behavior.
No one could have ever done anything other than what they actually did.
Moreover, in ordinary situations when people are not thinking about
272 R. Feldman
processes in their brains that cause their behavior, they are not free. No
one is ever free. Freedom is an illusion.
The points here are exactly analogous to the points made about contextual-
ism and skepticism. If philosophical settings make the standards such that
the sentences ascribing ‘free’ are not true, then in philosophical settings
such as this one, the incompatibilists (and those who deny that we are free)
are right. The philosophical debate, which obviously occurs in philosophi-
cal contexts, is won by the incompatibilists. Indeed, the compatibilists lose
without much of an argument in Hawthorne’s account. They capitulate
from the outset.
control? Is it one that properly grounds assessment of blame and the appli-
cation of punishment? Another set of questions arises for incompatibilists,
especially if the answers to the questions just listed are negative. Is the (or a)
sense of the word ‘free’ such that it can apply to actions that are not causally
determined? Is this sense one that properly grounds ascriptions of responsi-
bility and control, that properly grounds attributions of blame and the
imposition of punishment?
I find little of any help in thinking about these matters coming out
of contextualist considerations. Contextualism, as developed here, appar-
ently yields negative judgments about our freedom in philosophical
contexts. But it says little about why this is true, other than to endorse in-
compatibilist intuitions. It fails to engage considerations that support com-
patibilism. Furthermore, it is not really designed to explain or support an
understanding of ‘free’ that would help incompatibilists. Indeed, as I un-
derstand it, it does not help with this at all, since it does not address
the question of how causally undetermined actions could be other than
random.
7 Conclusion
Part of the problem with contextualism, both with respect to freedom and
with respect to skepticism, seems to me to arise because of the way contex-
tualists pose the problems. The problems are said to be explaining our
seemingly varying intuitions. We sometimes say one thing and we some-
times say another. Perhaps our intuitions do vary, and what we are inclined
to say varies accordingly. If so, these facts warrant explanations. But there
is another fact that warrants attention. It is that in philosophical contexts
in which all the relevant causal (or skeptical) alternatives have been made
salient, there are puzzles about exactly what to say or think concerning our
freedom or our knowledge. There are competing considerations all at one
time, and the hard question is to figure out what to make of them. Con-
textualism, at least as I’ve described it here, does not really address this in
any developed way. It just concedes the issues to the incompatibilists and
the skeptics.
The denials of incompatibilism and skepticism—compatibilism and
fallibilism—are in certain ways analogous. One such way is that their
Freedom and Contextualism 275
advocates are willing to assert the key sentences in the face of all the alter-
natives. When confronted with skeptical possibilities, fallibilists remain
willing to say ‘I know’. Similarly, when confronted with all the possible
causal explanations of behavior, compatibilists are willing to say, ‘I am
free’. They have things to say in support of those attitudes. Contextualism,
as presented here, simply says that in the contexts in which they speak,
they are wrong. Contextualists add that those very sentences express truths
in ordinary contexts, but they have little to say about why that is the case.
I conclude that contextualism may at best provide ways to explain some
varying linguistic dispositions and may contribute to ways to retain the
truth of some ordinary utterances. But it saves our ordinary utterances
at the expense of addressing some of the central philosophical arguments
that lead people to question those utterances in the first place. Words such
as ‘knows’ and ‘free’ may be context sensitive. The philosophical issues
remain unresolved.
Acknowledgments
Notes
1. Hawthorne does not explicitly apply his contextualist solution to the inconsis-
tent triad under discussion here. He does explain the varying truth conditions for
sentences containing the word ‘free’. Some of the details in my account of the
resolution of the puzzle are based on my way of extending his remarks to the puzzle.
3. This depends, of course, on the details of just what the current standards for
knowledge are.
References
Cohen, S. 1987. “Knowledge, Context, and Social Standards.” Synthese 73: 3–26.
DeRose, K. 1995. “Solving the Skeptical Problem.” Philosophical Review 104: 1–52.
Without cognising free will as a philosophical problem, [the Buddha] takes it for
granted that the innate character of each being leaves him the freedom to decide
about the actions that determine his future.
—H. W. Schumann (1973, 53)
There is free action, there is retribution, but I see no agent that passes out of one
set of momentary elements into another one, except the [connection] of those
elements.
—The Buddha1
of this discussion, free will is defined as a power truly our own independent
from causal determinants.
In sections 2–5 we will analyze the issues of this book from the Pali per-
spective. Here we will find a more robust view of the self, best interpreted
in functionalist terms, and also a realist view of causality. We will argue
that these Buddhists would join the compatiblists of this book in their
defense of moral responsibility. When we turn to the Mahayana schools, we
will discover that it is much further removed from the free will–determinism
debate. Mahayanists take the Buddha’s idea of ‘no self’ much more radi-
cally and they also generally reject a realist view of causality.
In the last two sections we will offer two views of Nagarjuna, one from ‘con-
structive postmodernism’ as opposed to ‘deconstructive postmodernism’
with which Nagarjuna is usually associated. (Because of space limitations, we
will just briefly mention Yogacara idealism, the other major Mahayanist
school.) While we would like to support a constructive postmodernist
Nagarjuna, the texts actually do not allow us to do this. This means that while
Pali Buddhists can embrace real agency in the world, Nagarjuna and his
followers appear to argue that any positive view of the self will prevent us
from escaping the cycles of existence. In this view the self is completely
deconstructed and all that we are left with are competing rhetorics of free-
dom and determinism.
We are delighted that we have been invited to contribute to this book
and to offer a Buddhist perspective on these fundamental issues. We believe
that Euro-American philosophers have much to learn from the Asian
philosophical tradition. Many of them are still unaware of the degree of
philosophical sophistication Asian thinkers have attained. Finally, we want
to emphasize that we are only teachers of Buddhism and not experts with
knowledge of Pali or Sanskrit and all their voluminous texts. We rely heav-
ily on the best work in the field and trust that we have used it accurately
and responsibly.
One possible reason for the failure of Asian philosophers to address the is-
sues of this book is the common observation that Asian thinkers, especially
those of the Indian subcontinent, are far more interested in absolute spiri-
tual freedom than freedom of the will. Rather than a freedom somehow
Buddhism and Freedom of the Will 279
related to the world of cause and effect, the Indian yogis appear mostly
concerned with a freedom that transcends the physical world entirely.
With this vision in mind the casual determination of karma is not a prob-
lem at all, as karmic bondage is a state of all beings until final liberation.
Hindu scripture describes the saints as veritable supermen. For example,
the Taittiriya Upanishad tells us that the yogi “attains . . . independent
sovereignty,” and enjoys a bliss that is a billion times greater than that of
the highest gods (1.6, 2.8). In the Maitri Upanishad the ascetic surpasses
Brahman, the Godhead, and “will go [yet further], he [will surpass] the
gods in the realm of divinity. . . .” (4.4). In the Shvetashvatara Upanishad
yogis gain incredible powers: they “shall roll up space as if it were a piece of
leather” (6.20); and a yogi in the Taittiriya Upanishad boasts that “I am the
first-born of the world-order, earlier than the gods, in the navel of immor-
tality. . . . I have overcome the whole world” (3.5). Such a view has been
called ‘spiritual Titanism’, an extreme form of humanism in which humans
take on divine attributes and prerogatives (Gier 2000).
In the Pali texts the Buddha rejects these incredible claims of the Hindu
and Jain yogis. He was particularly critical of their claims to omniscience.
He did, however, embrace their subordination of the gods and the require-
ment that they had to be reincarnated as humans in order to be liberated.
Therefore, the Buddha and most of his followers are not spiritual Titans,
primarily because they rejected the divinity of the saint and sought
Nirvana in this world rather than in some otherworldly domain. The
Buddha also believed that the body was constitutive of personal identity
(Sankhya-Yoga and Jain dualists rejected this) and that the emotions and
senses were not evil. Finally, the Buddha distinguished between Nirvana in
the body and Nirvana at the end of the cycles of existence. Because the
Buddha was a strict empiricist and because final Nirvana was beyond the
experience of anyone, he declined to say anything about it at all. However,
the Buddha and many of his disciples presumably did live in a constant
state of embodied Nirvana and the best one-word definition of this state is
‘freedom’. We will analyze the meaning of this freedom in compatiblist
terms in the fourth section.
There is a deeper and more philosophically interesting reason why the
Buddha would have found free will a nonissue. This may be the same rea-
son that Greeks and Romans generally did not find it a problem.3 It is also
connected to why American pragmatism did not address the issue, and
280 N. F. Gier and P. Kjellberg
parallel is to James’s radical empiricism. James and the Buddha observed that
basic experience does not divide into inner and outer; rather, the inner flows
into the outer and the outer flows into the inner. (One could perhaps read
Hume in the same way.) It is only by some Cartesian method of systematic
doubt that an inner world of ideas and perceptions is separated from an
outer world of physical things. The Buddha and James also claimed that
basic experience does not divide into facts and values, because as the Buddha
said: “What one feels, one perceives; what one perceives one reasons
about.”7 What one feels is obviously filled with values and emotions.
Critics might agree with James and the Buddha that there are no onto-
logical bifurcations in our immediate experience, but they still might ob-
ject that even though their ‘inner’ flows into their ‘outer’, their ‘inner’ does
not flow into another’s ‘inner’. The Buddhists disagree but for a reason that
many Euro-American philosophers will reject outright. The Buddha and his
disciples claimed to have ESP powers and they said that they could read the
contents of other people’s minds. Specifically, the Buddha claimed to be
able to determine the balance on anyone’s karmic mortgage.
Quite apart from the validity of ESP, one could argue that we do in fact
read other people’s minds through their body language, as many people do
in fact carry their emotions on their sleeves. (Is this what Wittgenstein
meant by “meaning is a physiognomy”?) The Buddhists would say that this
minimal capacity that all of us have is simply the potential for anyone
to use the faculty of mindfulness to a very high and sensitive degree. Al-
though usually referred to as a virtue, mindfulness (sati) is more accurately
called the faculty (indriya) by which Buddhists control the six senses (the
mind is included) and their moral development.
3 Buddhist Compatibilism
they empower us. Buddhists would also agree with Lehrer that what we pre-
fer is always our option while what we desire wells up in us involuntarily.
By giving in to her cravings the Buddhist would continue the cycles of
existence forever, something she, or any other clear-minded person, does
not prefer. By choosing meditation and other spiritual disciplines as prefer-
ences, she prepares herself for the ultimate preference: Nirvana and free-
dom from Samsara (the cycle of birth, death, and rebirth, of which Nirvana
is the end). Furthermore, note that this Buddhist is an agent of her prefer-
ences, as Lehrer explains, rather than being just a passive victim of her
desires. Finally, with regard to the problem of preference control, the non-
theist Buddhist does not have to contend with an omnicausal deity, who,
according to Augustine and Luther, empowers us to turn to God as well
as to turn away from divine grace. Most Buddhists also do not have to con-
tend with mad neurophilosophers from Southern California to control
their preferences.
While the issue of free will does not arise in Buddhism, it is indisputable
that it embraces a universal determinism: every effect, without exception,
has a cause. The idea that the will is uncaused or is self-caused violates the
Buddhist principle of interdependent coorigination (prattiyasumutpada):
nothing in the universe can originate itself as substances allegedly do or
the will is said to do. Buddhist causality, however, is seen as a cosmic web
of causal conditions rather than linear and mechanical notions of push-
pull causation. Furthermore, the Buddha claimed that we are morally re-
sponsible only for those actions that we intend. He took strong exception
to the Jain theory that we suffer from accidental karma, such as stepping on
a bug that we do not see. The Jains, another Indian religion contemporary
with Buddhism, charged that the Buddha’s qualified determinism would
lead to antinomianism and ethical subjectivism. Only their strict deter-
minism, they claimed, would maintain objective ethical standards.
The Buddha countered, however, that if the mind is always subject to the
control of the past, then there could never be any liberation, nor could there
be any moral responsibility. The Buddha proposed that the ‘freedom of
the mind’ works in the following way: “If I were to oppose the formation of
the cause of sorrow, by opposing this formation I would become dispassion-
ate. Also, if I were to become evenminded in respect to this cause of sorrow,
if I were to develop this evenmindedness, I would become dispassionate.”8
The Buddha also claimed that personal temperament and circumstances
284 N. F. Gier and P. Kjellberg
qualify the effects of past causes such that there is no strict correspondence
between cause and effect. The most famous example is the analogy of a lump
of salt, first placed in a cup of water (representing the corrupt person) and
then placed in a lake (representing the Buddhist saint). The effect in the for-
mer is obviously different from the latter. This example has become contro-
versial, because the salt is usually seen as an evil intention, having disastrous
consequences in the sinful person but virtually no effect on the saint. The
Jains’ fears appear to be vindicated: can the saint commit a sin and not be
harmed or even blamed for it?
Pali Buddhist ethics can be called a contextual pragmatism, which can be
best explained by the famous motto “They who know causation (prattiya-
sumutpada) know the Dharma.”9 This can be read as they who know their
own causal web of existence know the truth (i.e., the true facts of their
lives) and they will know what to do. The truths we discover will be very
personal truths, moral and spiritual truths that are, as Aristotle says,
“relative to us.” This is not simply a cognitive knowing but a practical grasp
of what is appropriate and what is fitting for us and our surroundings.
Being mindful is deliberately forming preferences over the desires that
might lead us out of our own personal means and the virtue we have de-
veloped there. Like phronesis, Buddhist mindfulness is primarily nonsensu-
ous, correct perception. Note also the Buddha’s ethical naturalism,
outlawed in European philosophy after Hume, and the bold fusion of fact
and value, a topic we will not pursue further.
Both Aristotle and the Buddha thought it was always wrong to eat too
much, but each person will find his or her own relative mean between
eating too much and eating too little. Such a view is not relativistic in the
nonnormative sense, because the principal determinants for eating just
right are primarily objective not subjective. If people ignore these objective
factors—for example, temperament, body size, metabolism, and other
physiological factors—then their bodies, sooner or later, will tell them that
they are out of their respective means. If people are unmindful, they allow
unhealthy eating habits to take them in one direction or another, but the
mindful stay within their respective means. Therefore, the mindful ones
are free, while the ones tending to either extreme are not.
We can now define Buddhist free agents as those who are keenly aware
of the effects their actions have on themselves and others. They are free
from ego attachment and craving either for ascetic deficiency or indulgent
Buddhism and Freedom of the Will 285
this kind of event, taken in the abstract, whereas the Buddhists are inter-
ested in the concrete conditions of this particular concrete event.”14 There-
fore, Conze concludes that “the Buddhist doctrine of the multiplicity of
conditions seems to make a decision on the ‘freedom of the will’ unneces-
sary. If the total number of conditions is unlimited, and most of them are
unknown, it is impossible to say which condition of necessity brings about
which event” (1962, 146 fn.).
We partially disagree with Conze. It is clear that the Buddha did indeed
reject the existence of any self-causing agents, but did affirm that all events
have a multiplicity of conditions. Therefore, Pali Buddhism falls under the
general rubric of compatiblism and ‘soft’ determinism. These Buddhists be-
lieve that we are morally responsible for our own character and intentions,
which although completely conditioned by antecedent events, are nonethe-
less what we truly want and should do. Recall that the Buddha claims
that we are responsible only for those actions that we intend. In Sanskrit
karma means ‘action’, so it follows that Buddhist karma is volitional action
only.
Seeing the law of karma as a psychological law allows us to avoid both
the extravagance and absurdities of the common view of it as a cosmic
law with inscrutable laws of retribution. (It also means that we may also
see, as some Buddhists do, the six realms of existence as a metaphor for
the ‘animal’, ‘angelic’, and ‘demonic’ moments of one single life span.)15
Therefore, the law of karma can be conceived as the rather trivial truth that
all actions have consequences. Returning to causality as conditionality, we
can now state the following conditionals concerning moral responsibility:
“If we act motivated by greed, hatred, or delusion, we are planting the seed
of suffering; but when our acts are motivated by generosity, love, or wis-
dom, then we are creating the karmic conditions for abundance and hap-
piness.”16 For the Buddhists, karma works at two levels—one immediate
and one delayed. In any of our acts we can immediately experience the
results depending on whether they were done, for example, out of love or
hatred (of which only the truly obtuse person will claim to be unaware).
Later on, these seeds of our actions will produce their inevitable fruits and,
following the principle of interdependent coorigination, these fruits will
finally ripen. The ripening of karmic action is a pervasive metaphor in all
Buddhist literature.
288 N. F. Gier and P. Kjellberg
The Buddha’s response to the axial age’s discovery of the self was strikingly
unique: he proposed the doctrine of no-self (anatman), which literally
means no atman, the Hindu soul substance. This conceptual innovation
was so provocative that it was bound to invite misinterpretation, and un-
founded charges of Buddhist “nihilism” continue even to this day. The
Buddha anticipated Hume’s view that the self is an ensemble of feelings, per-
ceptions, dispositions, and awareness (the skandhas) that is the center of
agency and moral responsibility.17 The Buddha’s view, however, is different
from Hume’s (1978), primarily because the Buddha appeared to support real
causal efficacy among internally related phenomena. (We believe that Hume
may have been misled by the current scientific model of externally related
atoms.) While Hume deconstructed any theory of causality, the Buddha
reconstructed causal relations with his theory of interdependent coorigina-
tion. The Buddha agrees with Hume about the absence of causal power but
disagrees with him about the absence of causal relations. As the Pali philoso-
pher Buddhaghosa said: “There is no real production; there is only interde-
pendence.”18 This is the same as saying that causality has been replaced by
conditionality.
The Buddha rejected the soul-as-spiritual-substance view of the
Upanishads, Jainism, and Sankhya-Yoga, and he deconstructed the ‘specta-
tor’ self of these philosophies 2,500 years before recent thinkers dismantled
the Cartesian self. As opposed to strict deconstruction, for example, Pali
Buddhists hold that selves, though neither the same nor different through-
out their lives, are nevertheless responsible for their actions. (Pali Buddhism,
therefore, should be aligned with the school of constructive postmod-
ernism.) These selves are also real in the sense that they are constituted by
relations with their bodies, other selves, and all other entities. This is why the
Pali self should be viewed in relational or process terms rather than the
skeptical implications of the no-self doctrine, which many later Buddhists
supported. The Pali self is relational primarily in the sense of its dependence
on the five skandhas and the internal relations this dependence entails.
Another positive way to express nonsubstantiality is to describe the
Buddhist self as ‘functional’. In fact, each of the skandhas should be seen as
functions rather than entities. On this point, Kalupahana makes good use of
James, who while denying a soul substance, maintained that consciousness
Buddhism and Freedom of the Will 289
there is nothing that stays the same. The traditional argument here
proceeds by elimination: the physical bodies change; feelings, beliefs, de-
sires, and intentions all change; consciousness is intermittent; and our self-
conceptions change over time. None of the things we can point to as the
self remains the same. Therefore the self does not endure. The argument is
similar to the one given by Hume.
Second, the self is not separate from the causes and conditions that give
rise to it. A standard metaphor for this comes from the Dhammapada, a Pali
text from the third century B.C.E. The appearance of a rainbow arises out of
a certain combination of mist and light. Remove either one of these and
the rainbow no longer exists. Similarly, the appearance of a self arises out
of conditions: oxygen, food, parents, and so forth. Without them, there
would be no self. This argument can be made on a general level, as just
done, or on a particular level. You wouldn’t be the person you are if your
family, friends, and acquaintances all weren’t the people they are, if you
hadn’t had the experiences you’ve had, lived in the society you live in, and
so on. If we define the constellation of these conditions as one’s ‘world’, we
can say that the self cannot be separated, either practically or logically,
from the world in which it exists. Again we can say that causality has been
replaced by conditionality.
Third, the self is not independent, which is a rough approximation of
a Sanskrit term svabhava, meaning ‘self-nature’ or ‘own being’. Nagarjuna
uses reductio ad absurdum arguments to demonstrate the incoherence of
attempts to say anything about an independently existing self. If the self
existed on its own, it could not change or stay the same, could not be
unified or composite, could not know or be known, and so forth. And
yet, if we cannot say that the self really exists, by the same token we can-
not say that the self really does not exist. We cannot say anything at all
about it as an independently existing thing. What we are left with is an
empirical or a thoroughly conventional self, depending on the way we
read Nagarjuna.
The Pali versus Mahayana distinction is now not very helpful, so hence-
forth we will distinguish between ‘constructive postmodernism’ (CPM) and
a skeptical ‘deconstructive postmodernism’ (DPM). Many commentators
have interpreted Nagarjuna as anticipating the deconstruction of French
postmodernism.23 We prefer to interpret the Buddha’s philosophical in-
tentions as anticipating CPM, but we are not certain that Nagarjuna can be
292 N. F. Gier and P. Kjellberg
Action depends upon the agent. The producer proceeds being dependent
The agent itself depends on action. on the product,
One cannot see any way And the product proceeds being depen-
To establish them differently. dent on the producer.
The cause for realization is seen in noth-
ing else.
From this elimination [‘giving up’] of In the same way one should understand
[substantial] agent and action, the ‘acquiring’ on the basis of the ‘giving
One should elucidate appropriation up’ of the producer and the product.
[‘acquiring’] in the same way. By means of [this analysis] the product
Through action and agent All and
remaining things should be understood Producer all other things should be
(Garfield). dissolved (Streng).24
Even without knowing Tibetan (the original Sanskrit text has been lost),
we can see that Garfield’s translation is more elegant. Moreover, we can see
Streng’s skeptical conclusion that distinctions between agents, their actions,
and all other things “should be dissolved.” (The Yogacara idealist would say
“dissolved in the Buddha Mind.”) Garfield’s rendering is very different: he
takes Streng’s “giving up” only as denying agents and their actions any self-
being, not rejecting their existence altogether.
We propose that the phrase ‘phenomenal self’ be used for the Pali tradi-
tion and CPM interpretation while reserving ‘conventional’ as a place-
holder term for the deluded self who thinks it lives in a real world of
interdependent things and events. This conventional self exists only by
Buddhism and Freedom of the Will 293
This quatrain can be rendered intelligible from either a CPM or DPM posi-
tion, but other quatrains in this chapter appear to give the former approach
no support:
What language expresses is nonexistent.
The sphere of thought is nonexistent.
294 N. F. Gier and P. Kjellberg
One can salvage the CPM position only by some clever qualifications, and
then only partially. When Nagarjuna speaks of denying existence, he de-
nies only absolute existence, the negation of which is absolute nonexis-
tence. (Nagarjuna agrees with Parmenides in rejecting this as inexpressible
and meaningless.) But from a CPM standpoint, the nonexistence in this
quatrain must be a relative nonbeing that corresponds to the relative being
of interdependent existence that is expressible and meaningful. (A thing’s
relative nonbeing is everything that it is not, which is actually all the other
things that it is dependent on.) The CPM interpretation of this quatrain
runs aground with the conclusion that Nirvana and all things are ‘unarisen
and unceased’. This quatrain appears to undermine the DPM interpreta-
tion as well, because it implies that Nirvana is the only ‘thing’ that has
self-being. Tibetan Buddhists influenced by Yogacara idealism hold that
Nagarjuna intended only to reject the self-being of phenomenal reality, not
the ultimate reality that is the Dharmakaya. This is the cosmic body of the
Buddha into which all selves that reach Nirvana dissolve. Here is one of
their own quatrains:
If emptiness were the method, then
Buddhahood could not be. Since other
Than this cause there would be no other fruit,
The method is not emptiness.25
Not many Mahayana Buddhists, except those in some Zen schools, were
willing to follow Nagarjuna in rejecting all metaphysical views of the
Buddha and his cosmic body. While Nagarjuna’s deconstruction is logically
rigorous, it was obviously not religiously satisfying for most Buddhists.
translations of verses three and ten Nagarjuna appears to reject the exis-
tence of relational existence and the impossibility of anything whatsoever
arising. Here are the two quatrains from A. L. Herman’s revision of
Theodore Stcherbatsky’s translation:
In these four conditions we can find
No self-nature (svabhava)
Where there is no self-existence,
There can be no relational existence (parabhava).
Note here the clear negation of the two principal features of the doctrine
of interdependent coorigination: the denial of relational existence and the
rejection of conditionality. This view of Nagarjuna’s argument anticipates
Bradley’s famous attempt to show that an ontology based exclusively on
internal relations is just as unintelligible as one based on external relations.
We summarize Bradley’s argument as follows: If A is internally related to B
and R is the relation, one would then need R to relate R to A, R that relates
to R to B, R that relates R to R ad infinitum. Bradley’s conclusion is that
either A and B are totally separate or they are identical (1930, 18). Therefore,
a putative realism of distinct but interrelated entities by necessity collapses
into either an absolute monism (the Yogacara position—they would have
loved Bradley) or a denial of any intelligible view of reality at all (the DPM
Nagarjuna). Appealing to Whitehead’s (1978) doctrine of asymmetrical rela-
tions (e.g., that the present is internal to the past but the past remains exter-
nal), one could assume that the Buddha would agree with him, because he did
not explicitly affirm a doctrine of full internal relations as some Mahayanist
Buddhists did. The presence of asymmetrical relations would save Buddhism
from the extremes related above and keep the CPM interpretation alive.
Garfield’s translation of the two passages above allows us to reject the
deconstruction of dependent beings:
The essence of entities is not present in the conditions. If there is no essence, there
can be no otherness-essence. If things did not exist without essence, the phrase
‘When this exists so this will be’, would not be acceptable. (1. 3,10)
We take this to mean that if any being lacks self-existence, then any being
other than it does also. It also follows that if some beings were substances,
Buddhism and Freedom of the Will 297
we could not affirm the Buddhist law of conditionality. The CPM interpre-
tation of Nagarjuna appears to dissolve his doctrine of two truths, which
Garfield formulates as “the conventional truth of the reality and interde-
pendence of all phenomena and the ultimate truth of their emptiness of
inherent existence” (1995, 225). From the standpoint of CPM these two
truths are the same: real interdependence ontologically implies that no
being can have self-existence.
8 Conclusions
The traditional libertarian, who believes with Kant that, in the absence of
freedom, there cannot exist anything of absolute value in the world, may
wonder if there is any role left for freedom in Buddhism. Normally we
assume that there has to be a self or an agent in order for there to be free-
dom, but this is just the presumption the skeptical Nagarjuna questions.
If we cannot call the karmic web free since it lacks a self, by the same
token we cannot call it determined, since nothing outside of it is causing
it. To the extent that people identify a self, that self is determined by
causes outside of it. The more cultivated they become on the Buddhist
model, the less they think this way. The less who thinks this way? A ques-
tion that the European philosopher might ask. Nagarjuna’s answer is no
one, really. The nonpersonal web of causes and conditions sheds the delu-
sion, or, rather, ceases to give rise to it. Thus you get the seemingly para-
doxical lines from, for instance, The Diamond Sutra, that “even though
infinite beings have been saved, none have been saved.”28 Thus while we
would assume that there has to be a self in order for there to be freedom,
Nagarjuna would say that there is freedom only to the extent that there is
not a self. The familiar correlation in the European tradition of selfhood
and freedom is reversed: rather than correlates, selfhood and freedom are
antitheses.
The reason this matters for the skeptical Nagarjuna is the connection
between the self and suffering. The belief in a self that is enduring, sepa-
rate, and independent, gives rise to desires that are ultimately unsatisfiable
precisely because no such self exists. If all talk of the self or anything else is
conventional, and if no one set of conventions is more right than the
others, what then governs the choice among conventions? The First Noble
Truth of Buddhism, enunciated by the Buddha upon his enlightenment, is
298 N. F. Gier and P. Kjellberg
that all egocentric existence suffers. This is not to say that every moment is
agony, which would obviously be false, but that the process of existing,
thinking one is a substantial self, is inherently painful. Even the joyful
moments take their character from the absence of, or absence of awareness
of, discomfort. Our unwillingness to acknowledge this fact and our stub-
born insistence that suffering is avoidable only compound the difficulties.
More particularly, suffering is caused by belief in a self that does not
exist. Thinking in terms of a self that is enduring, separate, and indepen-
dent causes people to search for a happiness they cannot find precisely
because such a self does not exist. The desire to alleviate suffering by
eliminating the belief in a self is the motive force of Buddhism. It is easy to
see how, in the Mahayana tradition, one would not be able simply to alle-
viate one’s own suffering without relying on the notion of self that caused
the suffering in the first place. As a result, motivation must come from the
desire to alleviate the suffering of all sentient beings. This realization led to
the Bodhisattva ideal, enlightened beings, in one sense greater than the
Buddha himself, who vow to delay Nirvana until all other beings have
entered before them.
The silence of all forms of Buddhism on the debate about free will first
stems from its refusal to divide up the world in the modernist ways
discussed above. Second, it is silent because it rejects a substantial self-
determining self in the Pali tradition and any notion of an individual self in
the Mahayana tradition, whether it be Yogacara idealism or Madhyamaka,
the school of philosophers influenced by Nagarjuna. If this is true, then
contemporary interest in the debate must be premised on certain assump-
tions about the self and personhood. We see this clearly in Harry Frankfurt’s
now classic essay where he claims that one of the desiderata of a theory of
free will is that it demarcate the ‘essential difference’ between people and
animals (1971, 6). Buddhists, however, see people on a spectrum of beings,
between animals, who are unable to diagnose the origin of their suffering or
to take steps to alleviate it, and putative deities who, because of their con-
stant state of bliss, lack the incentive to act. Thus, while Frankfurt and the
Buddhists both observe the same phenomenon of first- and second-order
desires, one reason the Buddhists do not assign it the same weight is because
they refuse to distinguish people from animals for moral purposes. Antici-
pating Tom Regan’s approach to animal rights, the Dhammapada assumes
that animals have interests and that “life [is] dear” to them (10:3).
Buddhism and Freedom of the Will 299
Acknowledgments
We would like to thank Joseph Keim Campbell and Michael Myers for very
constructive comments and criticisms, and we are grateful to Michael
O’Rourke and P. J. Ivanhoe for their encouragement.
Notes
3. One could say that the issue was first raised by the Stoics, who criticized Epicurus’s
attempt to introduce indeterminism among the atoms, but what is entirely missing
from Greco-Roman philosophy is the concept of a fully self-determining will.
Furthermore, Epicurus’s indeterminism is no more successful in addressing free will
than those who would now want to appeal to Heisenberg’s indeterminacy principle.
Finally, it is clear that modernist concepts such as autonomy and the separation of
fact and value were indeed anticipated and affirmed by some ancient schools of
thought, such as Sankhya-Yoga, Jainism, Indian and Greek atomism, and Chinese
Mohism.
4. For example, see Hudson (1973, 1985), Gudmunsen (1977), and Katz (1983).
6. See Gier (2000, chap. 2) and the Series Introduction by David Ray Griffin in any
volume of SUNY Press’s Series on Constructive Postmodern Thought.
7. Mahavedallasutta, I.293.
9. Majjhima Nikaya I.190–191, cited in Kalupahana (1976, 64). The pronoun has
been changed to “they” to avoid the exclusive “he.”
10. See Galston (1991, passim, 230), and Galston (1995) for a revised view of the self.
See Macedo (1990, passim, 220–221), for “situated autonomy.”
12. Garfield (2002, 73–76). “All of the standard desiderata of good theories apply—
economy, elegance, predictive power, confirmation, coherence with other theories,
and so on . . . . Then we can pay attention to . . . interdependence, and its multiple,
multidimensional, inter- and intra-level character—and let a thousand entities
bloom, requiring of each only that it genuinely toil and spin, accomplishing some
real explanatory work . . .” (75).
13. Kalupahana (1976, 28). Reading Kapulahana more carefully than her teacher,
Gier’s student Kaylani Merrill shows that Buddhist conditionality has a different
logic. The Buddha used a Pali locative absolute to describe conditionality. The gram-
mar of ‘when that, then this’ is opposed to the ‘if, then’ formula found in European
forms of sentential logic. See Kalupahana (1992, 57).
14. Conze (1962, 149). The difference between abstract and concrete here can be
explained with the distinction between rational or aesthetic order. See Gier (2001).
16. We are indebted to Joseph Goldstein for this statement and this insightful way
of redefining karma. See Goldstein (1999, 291). The definition of karma as volitional
action is not only good Pali Buddhism but it is also the position of the great
Mahayanist philosopher Vasubandhu: “karma is will (cetana) and voluntary action
(cetayita karanam)” (cited in Stcherbatsky 1974, 32).
17. For the best comparative studies of the Buddha and Hume, see Betty (1971) and
Giles (1993).
19. Kalupahana (1987, 20–21). Kalupahana’s Pali has been changed to Sanskrit.
22. See Garfield (2002) for the best comparative analysis of this issue.
23. See Mabbett (1995, 203). For a constructive postmodern interpretation see
Kalupahana (1992, 1986).
References
Conze, E. 1962. Buddhist Thought in India. London: George Allen and Unwin.
. trans. 1975. Buddhist Wisdom Books. London: George Allen and Unwin.
Frankfurt, H. G. 1971. “Freedom of the Will and the Concept of a Person.” Journal
of Philosophy 68: 5–20.
Galston, W. 1991. Liberal Purposes: Goods, Virtues, and Diversity in the Liberal State.
Cambridge: Cambridge University Press.
Garfield, J. L. 1995. The Fundamental Wisdom of the Middle Way. New York: Oxford
University Press.
Gier, N. F. 1983. “Dialectic: East and West.” Indian Philosophical Quarterly 10: 207–218.
Giles, J. 1993. “The No-Self Theory: Hume, Buddhism, and Personal Identity.”
Philosophy East and West 43: 175–200.
Goldstein, J. 1999. “Cause and Effect.” In J. Smith, ed., Radiant Mind: Essential
Buddhist Teachings and Texts. New York: Riverhead Books.
Gomez, L. O. 1975. “Some Aspects of Free-Will in the Nikayas.” Philosophy East and
West 25: 81–90.
Gudmunsen, C. 1977. Wittgenstein and Buddhism. New York: Barnes and Noble.
Hudson, H. 1973. “Wittgenstein and Zen Buddhism.” Philosophy East and West
23: 471–482.
. 1985. “Wittgenstein and Nagarjuna.” Philosophy East and West 35: 157–170.
Hume, D. 1978. A Treatise of Human Nature. New York: Oxford University Press.
Katz, N. 1983. “Nagarjuna and Wittgenstein on Error.” In N. Katz, ed., Buddhist and
Western Philosophy. Atlantic Heights, N.J.: Humanities Press.
. 1986. Nagarjuna: Philosophy of the Middle Way. Albany, N.Y.: SUNY Press.
Keown, D. 1992. The Nature of Buddhist Ethics. New York: St. Martin’s Press.
Mabbett, I. W. 1995. “Nagarjuna and Deconstruction.” Philosophy East and West 45:
203–226.
Nishida, N. 1990. An Inquiry Into the Good. M. Abe and C. Ives, trans. New Haven,
Conn.: Yale University Press.
Verma, S. 1970. The Metaphysical Foundations of Gandhi’s Thought. New Delhi: Orient
Longmans.
Whitehead, A. N. 1978. Process and Reality: An Essay in Cosmology. New York: Free
Press.
14 After Compatibilism and Incompatibilism
Ted Honderich
You can take determinism to be the family of views, a few of them clear and
otherwise conceptually adequate, that our decisions and the like and also
our actions flowing from them are the effects of certain ordinary causal
sequences. These personal events are necessitated by initial and also by sub-
sequent causal circumstances that make up the causal sequences. Has this
determinism been shown to be false?
Whether we actually experience its being false, which is to say become
aware of its falsehood in the course of our deciding and acting, has been dis-
puted. From the seventeenth century right up to an issue of the Journal of
Consciousness Studies the other year (Searle 2000), some philosophers have
said that we do indeed experience the falsehood of determinism. We learn
or somehow get the truth that at least sometimes our decisions and actions
are not ordinary effects.
That is, in deciding and acting we ordinarily have a certain idea of free-
dom, and we also somehow see that we do actually have this freedom—
that we are having it in the course of the deciding and acting. It is part of
our consciousness of pieces of deciding and acting that they could go the
other way instead, given things exactly as they are and were. Rather than
being ordinary effects, the decisions and actions are originated.
To say they are originated is to refer to some special mental or neural
generative activity that takes the explanatory place of ordinary and clear
causation. Often an attempt is made by philosophers to describe this
generative activity itself. It is described as being a matter of extraordinary
or funny causes, even self-causes, or creative endeavors, or teleology, or or-
dinary English verbs oddly understood, or something sui generis, or reasons
306 T. Honderich
in a particular sense of the term where they are not only terms of logical
relations, premises for conclusions, but events—although of course not
ordinary causal events. Events of some kind are needed, of course, for the
explanation of other events—the decisions and actions.1
All of these descriptions of origination are found unclear or factitious by
other philosophers, unsurprisingly. But that is not the end of origination.
To say decisions and actions are sometimes originated can be to say no
more than that our decisions and actions come about in such a way, what-
ever it is, that certain attitudes on our part are in order. These attitudes
include certain hopes for the future and the particular moral approval that
credits us with a particular kind of moral responsibility, and so on. To say
the lesser thing is to say something perfectly intelligible.2 Nor does it
become unintelligible if you go further and say that our decisions are
somehow within our control or up to us, whatever that comes to, and
hence that certain attitudes are in place.3
Against the idea that in deciding and acting we actually experience the
falsehood of determinism, other philosophers have said or contemplated4
that what we ordinarily experience in deciding and acting is something
quite different from an absence of effects. We have the idea that decisions
and the like are effects but not effects of a particular kind—and we can
know this idea fits the facts.
That is, decisions and acts are not effects of compelling, constraining, or
inhibiting causal circumstances—causal circumstances in conflict with our
desires, wills, personalities, or characters. As we discern in these typical
cases, there is nothing that is forcing our decisions on us. There is an ab-
sence of compulsion or the like. We know that in this clear sense the thing
can go the other way. There is the freedom of voluntariness, as distinct from
origination. Determinism is left perfectly possible by it.
Whatever is to be said of this dispute about deciding and acting with
respect to determinism, of which more in a minute, there is a truth about
the rest of our common experience, including our actual experience in
science—and in neuroscience most relevantly. It is that this wider experi-
ence leads us towards the view, a second determinism, that all experienced
events other than decisions and actions are effects of ordinary causal se-
quences. If this is so, we surely have good reason, a strong inferential base,
for taking the same to be true of the decisions and actions.
After Compatibilism and Incompatibilism 307
with determinism, notably the special guilt and failure associated with our
images of origination in connection with our own actions.
None of this is likely to be successful, essentially because of the strength of
our desires. It has seemed to me that success in the project of affirmation will
be owed, in the end, to nothing other than plain and settled belief in
determinism. That remains rare, a lot rarer than contemplating the truth of
determinism—taking it as something in need of being considered. It has
seemed to me that we or most of us, and perhaps more of our successors, will
succeed in the response of affirmation only when we really believe of the
things we dearly want that we do not have them and cannot ever have them.
4 Another Problem
Let me lay out a couple of lines of thought. Both of them do indicate more
of how radical it seems to me we need to be—it isn’t a matter of anything
like more tinkering with origination. Both of them just might be of use,
too, with the seeming contradiction. The first has to do with the nature of
consciousness, the second with causation and explanation.
What is it for you to be aware of this room now? More generally, on
the assumption that consciousness divides into perceptual, reflective, and
affective parts, what is it for you to be perceptually conscious? Two sorts of
general answer are given, and also given with respect to the other two parts
of consciousness.
One sort of answer, if you will allow a quick but enlightening parody, is
that perceptual consciousness is cells. It is neural activity. This is the old
physicalism about the mind of the seventeenth century, a physicalism
that now includes functionalism and cognitive science with philosophical
ambition. The other answer is only more disgraceful in terms of the physi-
calist conventionality of our current philosophy of mind. It is not often
given openly, but is implied by the increasing resistance to the idea that
consciousness is cells. This other answer, not much parodied, is that con-
sciousness is nonphysical stuff in heads.
It is possible to be attracted, as I am, to a quite general physicalism—the
view that all that exists is something close to physical. It is possible to be
attracted too, likely by way of the fact of the subjectivity of consciousness,
to the intolerable idea that perceptual consciousnessness is indeed funny
stuff in heads. You might think this situation is rather like what we have
been considering, attraction to both determinism and to something that so
far has been sunk in indeterminism. It seems to me possible that the situa-
tion with perceptual consciousness will be resolved by a radical view of this
consciousness.
What is it, really, for you now to be aware of this room? It is for the room
in a way to exist. That answer can be shown not to be merely a rhetorical
After Compatibilism and Incompatibilism 317
way of saying no more than that you are aware of the room. It is not a non-
analysis. Rather, the claim that your perceptual consciousness consists in a
kind of existence of a world consists in the claim that there is a certain state
of affairs, certainly not in your head. It is things, reasonably called chairs
and the like, being in space and time and dependent both on another
world, roughly speaking the relevant atoms, and also on your neurons in
particular. The world in question is anterior to the physical world, the one
dependent both on atoms and crucially on all of us, above all our shared
perceptual apparatus.
A few words delivered on the wing are unlikely to persuade you of this
doctrine, ‘consciousness as existence’ (2000, 2001c, 2004b). My first and
lesser aim, as you have heard, is to indicate something of the order of dif-
ferentness of thought that seems needed if we are to make a further and
better escape from three centuries of impasse in the philosophy of deter-
minism and freedom. In the doctrine about perceptual consciousness, a
general physicalism is in a way held onto, and the mystery of nonphysical
stuff in the head absolutely abandoned. But what attracted us to the stuff
in the head, its recommendation, is delivered to us by other means.
Above all, we are offered a real subjectivity, something clear on this
subject. This is your world of perceptual consciousness, different from the
shared physical world. Given the history and state of the philosophy of
determinism and freedom, is it not clear that only so significant a departure
from the cart tracks, and maybe one’s own recent tracks, has a chance?
Might it be that we have a chance of dealing with the seeming contradic-
tion involving determinism and certain attitudes by way of reflection on the
nature of consciousness—and in particular on consciousness as existence?
Searle thinks something like the first thing, and so, after all, if I may be per-
mitted another moment of un-American philosophical activity, there may
be one thing in his paper that is right. Surely no question of freedom could
arise about just exactly a physical world—a world in which we are present
only as conceived in the current physicalism of the philosophy of mind.
To say a word now of a particular idea about determinism and con-
sciousness, the autobiographical feelings of responsibility that persist de-
spite determinism are a matter of a certain individualism. If my feelings of
this kind cannot rest on my having originated my life, so to speak, can they
rest on the fact that my consciousness consists in what can be called the
private ownership of some reality?
318 T. Honderich
Notes
2. Peter Strawson (1962) spoke for many philosophers before and after him when he
spoke of “the obscure and panicky metaphysics” of origination. Galen Strawson
(1986) takes the view, in brief, that we do not need to think about determinism since
there is nothing intelligible in conflict with it in the talk of origination. Father right,
son not so right.
3. There is more on origination as I see it, and on other matters considered in this
paper, from causation onward, in Honderich (2002a) and (2002b). There is a lot more
on almost all relevant matters, for resolute readers, in (1988).
320 T. Honderich
5. For the final version of Prof. Earman’s lecture, see the first paper in this book, “De-
terminism: What We Have Learned and What We Still Don’t Know”. Cf. the various
contributions in Atmanspacher & Bishop (2002). You need not rush to agree, how-
ever, with what may be assumed, that the only conceptually adequate conception of
determinism is in terms of physics.
References
Atmanspacher, H. and R. Bishop. 2002. Between Choice and Chance. Thorverton and
Charlottesville: Imprint Academic.
. 2002. How Free Are You? Second edition. Oxford and New York: Oxford
University Press.
Kane, R. 1996. The Significance of Free Will. New York: Oxford University Press.
, ed. 2002. The Oxford Handbook of Free Will. Oxford and New York: Oxford
University Press.
Nagel, T. 1986. The View From Nowhere. New York: Oxford University Press.
Searle, J. R. 2000. “Consciousness, Free Action, and the Brain.” Journal of Conscious-
ness Studies 7: 3–22 (“Mind the Gap” issue).
Ability, 216–218, 231, 233, 237, Agency/agent, 61–62, 76–77, 87, 174,
241–245, 252 205, 283–285, 288, 292
basic, 242 Alienation, 178, 180
strong account of, 231–232, 237, Alternative possibilities (also,
241, 250 alternatives), 3–6, 111–112, 114,
weak account of, 231–233, 237, 126–129, 143, 152, 189, 217
241–243, 245, 248–252 argument for, 312
Accountability, 114–115, 117–118, condition of (AP), 70–72, 74
123–124 principle of (PAP), 4, 55, 126–127,
Action, 7, ch. 4 passim, 108–111, 113, 219–222
116–117, 119–124, 126–127, 130, Annihilationism, 289
135–137, 174–175, 181, 206, 217, Approval, moral, 306, 308, 310,
220, 234, 241, 246, 277, 287, 292, 314–315
305–306, 308–310 Atman, 88–289
akratic, 174–175, 179 Attitudes
autonomous, 174–175, 181 personal, 313
basic, 90–91, 244 reactive, 152
consequences of, 217, 220–221 Attitudinism, 312, 314, 319
courses of, 106 Austin-style examples, 71–72, 79
free, 217–218, 228, 257, 273, 277, 285 Autobiographical reflection, 314–315
future, 128 Autonomy, 285, ch. 8 passim
intentional, 243 as agency, 174–175
nonakratic, 174 agent, ch. 8 passim
nonautonomous, 175–176 as authenticity, 176–178, 186
repertoire of, 242, 244 heroic, 179, 184
self-forming, 73–74, 77–82, 85–87 normative moral, 176, 183
theory of, 174–175 as psychological independence,
volitional, 287 175–176, 179, 183–184
weak theory of, 250 as reasons-responsiveness, 180–181,
Affirmation, 313–314, 319 184, 186
324 Index
Wanton, 184
Will, 70–71, 84–85, 90, 277, 283, 290
weakness of, 169, 177
Will-setting, 72, 74
World
phenomenal, 315
state of, 109
Zeno, 22–23, 26