The Philosophical Review, LXXXVII, No. 2 (April 1978).

ABILITY AND RESPONSIBILITY

Peter van Inwagen

T here was a time when philosophers would debate the relative

merits of the doctrines of "liberty" and "necessity," or, as we
should say today, debate whether it is more reasonable to believe
in free will or in universal causal determinism. As everyone knows,
the parties to this debate shared a premise: that free will
and universal causal determinism are incompatible. And, as
everyone knows, there arose a philosophical tradition-repre-
sented by Hobbes, Hume, Jonathan Edwards, Mill, and Moritz
Schlick-in which just this premise is denied. Thus the debate
between the libertarians and necessitarians was undercut, and
most of the debates about free will today are, as they have been for
a long time, essentially debates about whether free will and
determinism are compatible or incompatible.
But why should anyone care whether we have free will or
whether determinism is true? The first part of this question is
perhaps easier to answer than the second: we care about free
will because we care about moral responsibility, and we are
persuaded that we cannot make ascriptions of moral respon-
sibility to agents who lack free will. Recently, however,
Harry Frankfurt has denied just this principle, or, at least, a
principle that sounds very much like it, which he calls the Prin-
ciple of Alternate Possibilities.' His formulation of the Principle
of Alternate Possibilities is
PAP A person ismorally responsible for what he has done only
if he could have done otherwise. (p. 829)
If Frankfurt has made out a good case for the falsity of PAP
(and I think he has), then it would seem that he has undercut
the debate between the "compatibilists" and the "incompatibil-
ists" (to use the contemporary jargon) in a way very similar to

' "Alternate Possibilities and Moral Responsibility," Journal of Philosophy,

LXVI (1969), 829-839.
20 1
 P E T E R VAN IN WAGEN

the way in which Hobbes and others undercut the debate between
the libertarians and the necessitarians.
Frankfurt supports his contention that PAP is false by means of
a certain style of counterexample; I shall call counterexamples
in this style, "Frankfurt counterexample^."^ The following Frank-
furt counterexample is due to David Blumenfeld. It is worked
out with rather more concrete detail than any of Frankfurt's
own counterexamples:
Suppose that the presence of a certain atmospheric reaction always causes Smith
to decide to attack the person nearest to him and to actually do so. Suppose
also that he always flushes a deep red when he considers and decides
against performing a n act of violence and that under certain circumstances the
atmospheric reaction is triggered by the appearance of just this shade of red.
Now imagine that on a day on which circumstances are favorable
to the triggering of the reaction, Smith considers whether or not to strike
a person with whom he is conversing, decides in favor of it, and forthwith
does so.3

The general idea behind Frankfurt counterexamples is this.

An agent S is in the process of deciding which of n alternative acts
A,. . . ,A,. .. ,A, to perform. He believes (correctly) that he cannot
avoid performing some one of these acts. He decides to perform,
and, acting on this decision, does perform A,. But, unknown to
him, there were various factors that would have prevented him
from performing (and perhaps even from deciding to perform)
any of A,. . . , A, except A,. These factors would have "come into
play" if he had shown any tendency towards performing (perhaps
even towards deciding to perform) any of A,. . . , A, except A,.
But since he in fact showed no such tendency, these factors re-
mained mere unactualized dispositions of the objects constituting
his environment: they played no role whatever in his deciding to
perform or in his performing A,.
According to Frankfurt, it is evident that in such cases we
should say (i) that Shad no alternative to performing A,, couldn't
have done otherwise than perform A,, and (it) S is nonetheless
responsible for having performed A, (or, at least, if he is not re-
sponsible for having performed A,, this must be due to some factor

O r perhaps they should be called "Fra*kfurt-~ozick" counterexamples.

See n. 2 (p. 835) to Frankfurt's article.
David Blumenfeld, "The Principle of Alternate Possibilities," Journal of
Philosophy, LXVIII (1971), 339-345, n. 3 (p. 341).
 ABILITY AND RESPONSIBILITY

other than his inability to perform any act other than A, for the
reason described).
Now if Frankfurt has indeed shown that PAP is false, this may
be of no great consequence. For it may well be that some trivial
modification of PAP is immune to Frankfurt counterexamples
and that this modified version of PAP entails that if universal
causal determinism and incompatibilism are both true, then all
our ascriptions of moral responsibility are false. Frankfurt argues
that this is not the case, however, and that what one might call
the "correct version" of PAP (that is, the correct principle govern-
ing excuse from responsibility in cases in which alternate possibil-
ities for action are absent) cannot be used to show that deter-
minism and moral responsibility are in ~ o n f l i c t I. ~shall not in
this paper try to determine whether Frankfurt's proposed principle
is true or false, or discuss whether it in fact plays a role in our
deliberations about moral responsibility. I shall instead exhibit
three principles, which, if they are not "versions" of PAP, are at
least principles very similar to PAP, and which do play a role in
our deliberations about responsibility. I shall argue that these
principles are immune to Frankfurt-style counterexamples. (I
shall call counterexamples that are directed against principles
similar to but distinct from PAP, and which are as strategically
similar to Frankfurt counterexamples as is possible, Frankfurt-
style counterexamples. I shall reserve the term "Frankfurt counter-
example" for counterexamples directed just against PAP itself.)
PAP, as Frankfurt formulates it, is a principle about performed
acts (things we have done). In Part 11, I shall consider a principle
about unperformed acts (things we have left undone). In Part 111, I
shall consider two principles about the consequences of what we
have done (or left undone). In Part IV, I shall argue that if these
three principles are true and if a version of incompatibilism
appropriate to each is true, then determinism and moral respon-
sibility are in conflict, even given that PAP is false.

T h e "correct version" of PAP is: "A person is not morally responsible for
what he has done if he did it only because he could not have done otherwise"
( p 838).
 PETER VAN INWAGEN

I1
Consider the following principle (the Principle of Possible
Action):
PPA A person is morally responsible for failing to perform a
given act only if he could have performed that act.
This principle is intuitively very plausible. But the same might
have been said about PAP. Can we show that PPA is false by
constructing a counterexample to it that is like Frankfurt's
counterexamples to PAP? An adaptation to the case of unper-
formed acts of Frankfurt's general strategy would, I think, look
something like this: an agent is in the process of deciding
whether or not to perform a certain act A. He decides not to
perform A, and, owing to this decision, refrains from performing
A.' But, unknown to him, there were various factors that would
have prevented him from performing (and perhaps even from
deciding to perform) A. These factors would have come into play
if he had shown any tendency towards performing (perhaps even
towards deciding to perform) A. But since he in fact showed no
such tendency, these factors remained mere unactualized disposi-
tions of the objects constituting his environment: they played
no role whatever in his deciding not to perform or his failure to
perform A.
Putative counterexamples to PPA prepared according to this
recipe produce, in me at least, no inclination to reject this prin-
ciple. Let us look at one.
Suppose I look out the window of my house and see a man
being robbed and beaten by several powerful-looking assailants.
It occurs to me that perhaps I had better call the police. I reach
for the telephone and then stop. It crosses my mind that if I do

This schema and the instance of it that follows involve the agent's inten-
tionally refraining from performing a given act. Of course not every case in
which we might want to consider holding a n agent responsible for failing to
perform some act is a case in which the agent intentionally refrains from per-
forming that act: he may never even have considered performing that act. This
distinction between two ways of failing to perform a given act is of no impor-
tance for our present purposes. The points made in the text would be equally
valid if we had chosen to examine a case in which the agent fails even to think
of performing the act whose nonperformance we are considering holding him
responsible for.
 ABILITY AND RESPONSIBILITY

call the police, the robbers might hear of it and wreak their ven-
geance on me. And, in any case, the police would probably want
me to make a statement and perhaps even to go to the police sta-
tion and identify someone in a lineup or look through endless
books of photographs of thugs. And it's after eleven already, and
I have to get up early tomorrow. So I decide "not to get involved,"
return to my chair and put the matter firmly out of my mind.
Now suppose also that, quite unknown to me, there has been
some sort of disaster at the telephone exchange, and that every
telephone in the city is out of order and will be for several hours.
Am I responsible for failing to call the police? Of course not. I
couldn't have called them. I may be responsible for failing to
try to call the police (that much I could have done), or for refraining
from calling the police, or for having let myself, over the years,
become the sort of man who doesn't (try to) call the police under
such circumstances. I may be responsible for being selfish and
cowardly. But I am simply not responsible for failing to call the
police. This "counterexample," therefore, is not a counterex-
ample at all: PPA is unscathed.
It is, of course, proverbially hard to prove a universal negative
proposition. Perhaps there are Frankfurt-style counterexamples
to PPA. But I don't see how to construct one. I conclude that
Frankfurt's style of argument cannot be used to refute PPA.

Both PAP and PPA are principles about acts, performed or

unperformed. But, in fact, when we make ascriptions of moral
responsibility, we do not normally say things like "You are
responsible for killing Jones" or "He is responsible for failing
to water the marigolds." We are much more likely to say "You
are responsible for Jones's death" or "He is responsible for the
shocking state the marigolds are in." That is, we normally hold
people responsible not for their acts or failures to act (at least
explicitly), but for the results or consequences of these acts and
failures. What, ontologically speaking, are results or consequences
of action and inaction? What sorts of thing are Jones's death and
the shocking state the marigolds are in? The general terms
"event" and "state of affairs" seem appropriate ones to apply
 PETER VAN INWAGEN

to these items. But what are events and states of affairs? This
question, like all interesting philosophical questions I know of,
has no generally accepted answer. Philosophers do not seem even
to be able to agree whether events and states of affairs are partic-
ulars or universals. In order to avoid taking sides in the debate
about this, I shall adopt the following strategy. I shall state a
certain principle about excuse from responsibility that seems
to me to be a plausible one, provided the events or states of affairs
we hold people responsible for are particulars. And I shall state
a similar principle that seems to me to be plausible, provided
the events or states of affairs we hold people responsible for are
universals. For each of these principles, I shall argue that it cannot
be refuted by Frankfurt-style counterexamples. The first of these
principles (which I shall call principles of possible prevention) is:
PPPl A person is morally responsible for a certain event (par-
ticular) only if he could have prevented it.
This principle is about events; but if we were to examine a prin-
ciple, otherwise similar, about "state-of-affairs particulars" (for
example, the way secondary education is organized in Switzer-
land6)we could employ arguments that differ from the following
arguments only in verbal detail.
What are events if they are particulars? They are items that
can be witnessed (at least if they consist in visible changes in
visible particulars), remembered, and reported.? They are
typically denoted by phrases like "the fall of the Alamo," "the
death of Caesar," "the death of Caesar in 44 B.C.," and "what
Bill saw happen in the garden."' How shall we identify and in-

Perhaps it is debatable whether this pnrase designates a particular.

' I doubt, however, whether they can be anticipated. The objects of antici-
pation and other "future-directed" attitudes are, I think, universals.
Perhaps the last of these phrases could also be used to name a n event-
universal. We seem to be using it this way if we say, "What Bill saw happen in
the garden happens all too frequently." But, I think, we use it to name a partic-
ular when we say, "What Bill saw happen in the garden last night will live
in infamy," or "could have been prevented with a little foresight." The phrases
"the fall of the Alamo" and "the death of Caesar," however, seem to be suited
only for denoting particulars: even if the Alamo had fallen twice, even if Caesar
(like Lazarus) had died twice, we could not say, "The fall of the Alamo has
happened twice" or "The death of Caesar has happened twice." (This is not
 ABILITY AND RESPONSIBILITY

dividuate event-particulars (hereinafter, "events")? Individuat-

ing particulars, whether events, tables, or human beings is
always a tricky business. (Consider the Ship of Theseus.) As
Davidson says,
Before we enthusiastically embrace an ontology of events we will want to think
long and hard about the criteria for individuating them. I am inclined to think
we can do as well for events generally as we can for physical objects (which is
not very well). . . .'
In a paper later than the one this quotation is taken from, David-
son tries to "do as well." He tells us that finding a satisfactory
criterion of individuation for events will consist in providing "a
satisfactory filling for the blank in:
If x and y are events, then x =y if and only if ,,lo

The "filling" he suggests for this blank is (roughly) "x and y have
the same causes and effects." The biconditional so obtained, is,
I have no doubt, true. But this biconditional will not be "satis-
factory" for our purpose, which is the evaluation of PPP1. What
we want to be able to do is to tell whether some event that would
happen if what we earlier called "unactualized dispositions of
the objects constituting the agent's environment" were to come
into play, is the same as some event (the event responsibility for
which we are enquiring about) that actually has happened; that
is, we want to know how to tell of some given event whether it,
that very same event, would (nevertheless) have happened if
things had been different in certain specified ways. (For when
we ask whether an agent could have prevented a certain event E
by doing, say, X, we ;hall have to be able to answer the question
whether E would nonetheless have happened if the agent had done
x.1
To see why Davidson's criterion cannot be used to answer our
sort of question about event-identity, consider the following
formally similar criterion of individuation for persons: "x and

due, or not due solely, to the presence of the definite article in these phrases, for
we can say, "The thing Bill fears most has happened twice.")
' From Davidson's contribution to a symposium on events and event-
descriptions in Fact and Existence ed. by J. Margolis (Oxford, 1969), p. 84.
lo "The Individuation of Events," in Essays in Honor ofCarl G. Hembe1 ed. by
N. Rescher (Dordrecht, 1969), p. 225.
 PETER VAN INWAGEN

y are the same person if and only if x andy have the same blood
relatives (including siblings)." This criterion, while true, does not
help us if we are interested in counterfactual questions about
persons. For, obviously, any given man might have had different
relatives from those he in fact has (he might have had an addi-
tional brother, for example). Davidson's proposed criterion is
of no help to us for what is essentially the same reason: any given
event might have had different effects from the effects it has in
fact had. For example, if an historian writes, "Even if the murder
of Caesar had not resulted in a civil war, it would nevertheless
have led to widespread bloodshed," he does not convict himself
of conceptual confusion. But he is certainly presupposing that
the very event we call "the murder of Caesar" might have had
different effects.
The above considerations are not offered in criticism of David-
son's criterion, which is, after all, true, and may be a very useful
criterion to employ (say) when we are asking whether a given
brain-event and a given mental event are one event or two. But
Davidson's criterion is not the sort of criterion we need. We need
a criterion that stands to Davidson's criterion as "x and y are
the same human being if and only if x andy have the same causal
genesis" stands to the above criterion of personal identity. (I
use "causal genesis" with deliberate vagueness. A necessary condi-
tion for x and y having the same causal genesis is "their" having
developed from the same sperm and egg." But this is not suffi-
cient, or "identical"-monozygotic-twins would be numerically
identical.) This criterion can be used to make sense of talk about
what some particular person would have been like if things had
gone very differently for him.12 Can we devise a criterion for
counterfactual talk about events that is at least no worse than our
criterion for persons? I would suggest that we simply truncate
Davidson's criterion: x is the same event as y if and only if x and
y have the same causes. (Note the similarity of this criterion to
the causal-genesis criterion of personal identity.) I do not know
how to justify my intuition that this criterion is correct, any more

l1 Or so it seems to me. Of course a carte& (for example) will have a dif-

ferent view of the matter.
l2 Cf. Saul Kripke, "Naming and Necessity," in Semantics ofNatural Language
ed, by D. Davidson and G. Harman (Dordrecht, 1972), pp. 312-314.
 ABILITY AND RESPONSIBILITY

than I know how to justify my belief in the causal-genesis cri-

terion. But, of course, arguments must come to an end somewhere.
I can only suggest that since substances (like human beings and
tables) should be individuated by their causal origins, and since
we are talking about events that, like substances, are particulars,
the present proposal is plausible. Moreover, I am aware that this
proposed criterion is vague. It is not clear in every case of, say,
a story about the events leading up to Caesar's murder, whether
it would be correct to say that the murder had "the same causes"
in the story that it had in reality. But I think the notion of same
event is clear just insofar as the notion of same causes is clear. And
this latter notion is surely not hopelessly unclear: if Cleopatra
had poisoned Caesar in 48, then, clearly, there would have
happened an event that has not in fact happened, an event that
it would have been correct to call "Caesar's death," and which
would have had different causes from the event that is called
"Caesar's death." And, just as clearly, we cannot say of the event
we in fact call "Caesar's death," "Suppose it had been caused
four years earlier by Cleopatra's poisoning Caesar in Alexandria."
Moreover, it is hardly to be supposed that we should be able to
devise a criterion that will resolve all "puzzle cases," since we
are unable to devise such a criterion for people, mountains, or
tables. l3
l3 A theory o f event-particulars that is inconsistent with the view presented

in this paper is held b y R. M . Martin and Jaegwon Kim. (See Martin's con-
tribution to the symposium referred to in n. 9 , and, for Kim's latest published
views on events, "Causation, Nomic Subsumption, and the Concept o f Event,"
Journal ofPhilosophy L X X (1973), 217-236). I f we abstract from the particular
twists that each o f these authors gives to his own account o f events, we may say
that, on the "Kim-Martin" theory, the class o f events is the class o f substance-
property-time triples. For example Caesar's death is the triple (Caesar, being
dead, 15 March 44 B.C.). (Strictly speaking, the term "15 March 44 B.C." in
the preceding sentence should be replaced with a term designating the precise
instant at which Caesar died.) A "Kim-Martin" event happens just in the case
that its first term acquires its second term at its third term. However useful
Kim-Martin events may be in certain contexts o f discussion, I d o not think it
is correct to think o f them as particulars. T h e y are, rather, highly specified
universals, just as the property being the tallest man is a highly specified (in fact,
"definite") universal ( c f .n. 20). This property, though only one m a n can have it,
is nonetheless such that it could have been possessed by someone other than the
m a n who in fact has it. Similarly, any Ki'm-Martin event that happens could
have been caused b y quite different antecedent events from those that in fact
caused it. T o suppose that event-particulars have this feature is to violate m y
 PETER VAN IN WAGEN

Let us now return to PPPl. Can we devise a Frankfurt-style

counterexample to this principle? Let us try.
Gunnar shoots and kills Ridley (intentionally), thereby bring-
ing about Ridley's death, a certain event. But there is some factor,
F, which (i) played no causal role in Ridley's death, and (ii)
would have caused Ridley's death fz Gunnar had not shot him
(or, since factor F might have caused Ridley's death by causing
Gunnar to shoot him, perhaps we should say, "if Gunnar had
decided not to shoot him"), and (iii) is such that Gunnar could
not have prevented it from causing Ridley's death except by kill-
ing (or by deciding to kill) Ridley himself. So it would seem that
Gunnar is responsible for Ridley's death, though he could not
have prevented Ridley's death.
It is easy to see that this story is simply inconsistent. What is
in fact denoted by "Ridley's death" is not, according to the story,
caused by factor F. Therefore, if Gunnar had not shot Ridley, and,
as a result, factor F had caused Ridley to die, then there would
have been an event denoted by "Ridley's death" which had factor F
as (one of) its cause(s). But then this event would have been an
event other than the event in fact denoted by "Ridley's death";
the event in fact denoted by "Ridley's death" would not have
happened at all. But if this story is inconsistent it is not a counter-
example to PPPl. And I am unable to see how to construct a
putative Frankfurt-style counterexample to PPPl that cannot be
shown to be inconsistent by an argument of this sort.
Let us now turn to a principle about universals:
PPP2 A person is morally responsible for a certain state of
affairs only if (that state of affairs obtains and) he could
have prevented it from obtaining.14
intuitions (at any rate) about particulars. An additional problem: every Kim-
Martin event is such that there is some particular moment (its third term) such
that the event must happen just a t that moment if it happens at all. But surely
Caesar's death might have happened a t least a few moments earlier or later
than it in fact did, just as a given man might have been born (or even conceived)
a t least a few moments earlier or later than he in fact was.
'Wothing in PPPl corresponds to the parenthetical qualification "that state
of affairs obtains and" in this principle. So far as I can see, to say of a given
event-particular that it "happens" is equivalent to saying that it exists. And,
of course, there exist no events that do not exist. Thus there exist no events that
do not happen. But states of affairs may exist without obtaining, just as proposi-
tions may exist without being true or properties without being instantiated.
 ABILITY AND RESPONSIBILITY

The states of affairs "quantified over" in this principle are uni-

versals in the way propositions are universals. Just as there are
many different ways the concrete particulars that make up our
surroundings could be arranged that would be sufficient for the
truth of a given proposition, so there are many different ways they
could be arranged that would be sufficient for the obtaining of a
given state of affairs. Consider, for example, the state of affairs
that consists in Caesar's being murdered. This state of affairs
obtains because certain conspirators stabbed Caesar at Rome in
44 B.C., but (since it is a universal), it, that very same state of
affairs, might have obtained because (say) Cleopatra had poisoned
him at Alexandria in 48. But this is a bit vague. In order the better
to talk about "states of affairs," let us introduce "canonical"
names for them. Such names will consist in the result of pre-
fixing "its being the case that" (hereinafter, "C") to "eternal"
sentences. l5 Thus a canonical name for the state of affairs referred
to above would be "C (Caesar is murdered)." And let us say that
the result of flanking the identity-sign with canonical names of
states of affairs expresses a truth just in the case that the eternal
sentences embedded in these names express equivalent proposi-
tions, where propositions are equivalent if they are true in just the
same possible worlds. (Hereinafter, I shall assume that every
proposition is equivalent to and only to itself. This assump-
tion could be dispensed with at the cost of complicating the
syntax of the sequel.) A state of affairs will be said to obtain if the
proposition associated with it-that is, the proposition expressed
l5 The choice of eternal sentences as the arguments to which the operator
"C" attaches is made largely for the sake of convenience. If we had chosen in
addition to eternal sentences, noneternal sentences, sentences that can change
their truth-values as time passes, for this purpose, then we should have canon-
ical names for states of affairs that can obtain a t one time and fail to obtain
at other times. If we were to work out a comprehensive and consistent theory of
these entities, we should end up with a theory rather like the theory of "states of
affairs" R. M. Chisholm presents in "Events and Propositions," Noiis, 4 (1970),
pp. 15-24. We might, in fact, say that what we are calling "states of affairs" are
just that subclass of Chisholm's "states of affairs" that he calls propositions. If
we were to interpret PPP2 as involving quantification over all those things
Chisholm calls "states of affairs," then (I claim without argument) we could
nevertheless defend it against Frankfurt-style counterexamples by arguments
essentially the same as those we shall present in the text, but these arguments
would be considerably more complicated. For a discussion of the propriety
of applying the term "universal" to "states of affairs," see n. 20.
 PETER VAN INWA GEN

by the sentence embedded in any of its canonical names-is true. l6

Thus C(Caesar is murdered), C(Caesar is stabbed), and C(Caesar
is poisoned) are three distinct states of affairs, the first two of
which obtain and the last of which does not. To prevent a state of
affairs from obtaining is to prevent its associated proposition
from being true (or to see to it that or insure that that proposition
is not true).
Let us now, so armed, return to PPP2. Can we show that PPP2
is false by constructing Frankfurt-style counterexamples to it?
What would an attempt at such a counterexample look like? Like
this, I think.
Gunnar shoots Ridley (intentionally), an action sufficient for
the obtaining of Ridley's being dead, a certain state of affairs.
But there is some factor, F, which (i) played no causal role in
Ridley's death, and (ii) would have caused Ridley's death if
Gunnar had not shot him (or had decided not to shoot him), and
(iii) is such that Gunnar could not have prevented it from causing
Ridley's death except by killing (or by deciding to kill) Ridley
himself. So it would seem that Gunnar is responsible for Ridley's
being dead though he could not have prevented this state of
affairs from obtaining.
This case seems to show that PPP2 is false. But in fact it does
not. Let us remember that if this case is to be a counterexample
to PPP2 and not to some other principle, some principle involv-
ing particulars, we must take the words "Ridley's being dead"
that occur in it as denoting a universal. What universal? Presum-
ably, C(Rid1ey dies). But while it is indeed true that Gunnar
could not have prevented C(Rid1ey dies) from obtaining, I do not
think it is true that Gunnar is responsible for C(Rid1ey dies). Why
should anyone think he is? Well, Gunnar did something (shoot-
ing Ridley) that was suflicient for C(Rid1ey dies). What is more,
he performed this act intentionally, knowing that it was sufficient
for this state of affairs. This argument, however, is invalid. For
consider the state of affairs C(Rid1ey is mortal). When Gunnar
shot Ridley, he performed an act sufficient for (the obtaining
of) this state of affairs. But it would be absurd to say that Gunnar

l6 O n Chisholm's view (see n. 15), the proposition "associated with" a given

state of affairs just is that state of affairs.
 ABILITY AND RESPONSIBILITY

is responsible for C(Rid1ey is mortal). God, or Adam and Eve

jointly, or perhaps no one at all, might be held accountable for
Ridley's mortality; certainly not his murderer. (Unless, of course,
Ridley would have lived forever if he hadn't been murdered;
let's assume that is not the case.) .
In fact, it is arguable that C(Rid1ey dies) is the very same state
of affairs as C(Rid1ey is mortal). Given our principle of identity
for states of affairs, these "two" states of affairs are one if the two
eternal sentences "Ridley dies" and "Ridley is mortal" express
the same proposition. And what proposition could either of them
express but the proposition also expressed by "Ridley does not
live forever" and "Ridley dies at some time or other"? So, it
should seem, Gunnar is not responsible for C(Rid1ey dies), and
the attempted counterexample to PPP2 fails.
Nor do matters go differently if (somewhat implausibly) we
think of "Ridley's being dead" as denoting some more "specific"
state of affairs, such as C(Rid1ey is killed). If Gunnar is indeed
responsible for C(Rid1ey is killed), we shall nevertheless have a
counterexample to PPP2 only if Gunnar could not have pre-
vented this state of affairs from obtaining. Let us flesh out "factor
F" with some detail to insure that this is the case: suppose there
is a third party, Pistol, who would have killed Ridley if Gunnar
had not: and suppose Gunnar was able to prevent Pistol's killing
Ridley only by killing Ridley himself. By these stipulations, we
insure that Gunnar could not have prevented C(Rid1ey is killed).
But do we, in making these stipulations, absolve Gunnar from
responsibility for this state of affairs, or is his being responsible
for it at least consistent with our stipulations?
I think we absolve him, and that we can stlaw this by an ar-
gument of the same sort as the one we used in connection with
C(Rid1ey dies). Let us first note that we cannot show that Gunnar
is responsible for C(Rid1ey is killed) by pointing out that he did
something logically or causally sufficient for that state of affairs;
for, by the same argument, we could show that he is responsible
for C(Rid1ey is mortal). Now consider the state of affairs-call
it "D"-C(either Pistol or Gunnar kills Ridley). Is Gunnar respon-
sible for D? Note that D would have obtained no matter what
Gunnar had done, just as C(Rid1ey is mortal), C(either 2 2 = +
4 or Gunnar kills Ridley), and C(grass is green or Gunnar kills
213
 P E T E R VAN IN WAGEN

Ridley) would have. These latter states of affairs are obviously

not ones Gunnar is responsible for. Is there some important dif-
ference between them and D in virtue of which Gunnar is respon-
sible for D? There is only one nontrivial difference I can see:
There is no possible world in which Gunnar is responsible for
+
C(either 2 2 = 4 or Gunnar kills Ridley); and while there are
doubtless possible worlds in which Gunnar is responsible for
C(Rid1ey is mortal) and others in which he is responsible for
C(either grass is green or Gunnar kills Ridley), these worlds are
exceedingly "remote" from actuality. l7 But some worlds in which
Gunnar is responsible for D are much "closer" to actuality than
any of these: for example, "close" worlds in which the counter-
factual propositions about Pistol that were built into our example
are false and Ridley would not have been killed if Gunnar had not
shot him. But a miss is as good as a mile; I am arguing only that
Gunnar is not in fact responsible for D.
Now if Gunnar is not responsible for D, then he is not respon-
sible for C(either Pistol or Gunnar or someone else kills Ridley).
And this state of affairs and C(Rid1ey is killed) are one and the
same, since the proposition that either Pistol or Gunnar or some-
one else kills Ridley is equivalent to the proposition that Ridley
is killed. l8
"Worlds (say) in which Ridley would have lived forever if Gunnar'had not
shot him, and worlds in which the color of grass is up to Gunnar.
l8 The editors of T h e Philosophical Review have called my attention to the fact
that the validity of this argument appears to depend on the doubtful assump-
tion that "Gunnar is responsible for x" is an extensional context. But it need
not depend on this assumption. Let us say that in each of the following pairs of
sentences the second sentence is a disjunctive elaboration of the first.

All grass is green.

All grass in London or elsewhere is green.
Ridley is killed.
Ridley is killed by something or other at some time or other at some place
or other.
There is a stack of plates on the table.
There is a stack of plates on the table that contains twelve plates or else
some other number of plates.

Then, I think, a defender of the argument presented in the text need appeal
to no principle stronger than: From 'S is not responsible for C(p)', derive 'S
is not responsible for C(q)', provided p is a disjunctive elaboration of q. For
 ABILITY AND RESPONSIBILITY

In this example, "factor F" involved a second agent who

would have shot Ridley if Gunnar had not. But it would have
made no real difference if we had imagined factor F being such
that it would have caused Ridley's death by "working through"
Gunnar. (See Blumenfeld's counterexample to PAP, quoted
in Part I.) Suppose, for example, that Gunnar decides to kill
Ridley and does so. Suppose that if he had decided not to kill
Ridley he would have flushed red (which he couldn't help) and
that this red flush together with the prevailing atmospheric
conditions would have caused him to decide to kill and, as a result
of this decision, to kill, Ridley. Suppose the presence of these
atmospheric conditions and the effect on him of their copresence
with his flushing red are things he has no choice about. It follows
from these suppositions that Gunnar could not have prevented
C(Rid1ey is killed). But we can show by an argument essentially
the same as the argument we employed in the "Pistol" case that
Gunnar is not responsible for this state of affairs. We proceed by
showing first that Gunnar is not responsible for
K C(Rid1ey is killed by someone who is caused to kill him by
factor F [red flush, atmospheric conditions, and so on] or else
Ridley is killed by someone who is not caused to kill him by
factor F).
This state of affairs plays the role played by C(either Pistol or
Gunnar or someone else kills Ridley) in our demonstration that,
in the "Pistol" case, Gunnar is not responsible for C(Rid1ey is
killed). We cannot say of K what we said of D, and what we could
have said of C(either Pistol or Gunnar or someone else kills
Ridley), that it would have obtained no matter what Gunnar had
done, for it would not have obtained if Gunnar had not shot
Ridley. But we can say of K that it would have obtained no
matter what choices or decisions Gunnar had made, and this seems
to me to entail that Gunnar is not responsible for it. (I owe this

example, from "Henry is not responsible for C(There is a stack of plates on the
table that contains twelve plates or else some other number of plates)" we
derive "Henry is not responsible for C(there is a stack of plates on the table)."
This inference seems to me to be plainly valid, even if we suppose Henry to
be unable to count beyond three and to be ignorant of the logical principle
of Addition.
 PETER VAN IN WAGEN

point to the editors of The Philosophical Review.) The remaining

step in the demonstration consists simply in observing that the
proposition associated with K is equivalent to the proposition
that Ridley is killed and, therefore that K and C(Rid1ey is killed)
are one and the same state of affairs, from which fact we infer
that Gunnar is not responsible for C(Rid1ey is killed). (Or, if this
inference be thought dubious, we can say that the remaining step
consists in observing that the sentence embedded in the displayed
name of K is a "disjunctive elaboration" of "Ridley is killed,"
together with an application of the rule stated in footnote 18.)
If we had chosen to examine instead of C(Rid1ey is killed) some
even more "specific" state of affairs, such as C(Rid1ey is shot to
death at 3:43 PM, 12January 1949, in Chicago), this would have
made no difference to our argument, which in no way depended
on the degree of specificity of C(Rid1ey is killed). An argument
of the same sort could be applied to any attempt at a Frankfurt-
style counterexample to PPP2: the putative counterexample will
not be a counterexample unless it entails that the agent whose
responsibility is in question could not have prevented some given
state of affairs; but if the "counterexample" does indeed have
this feature, then (I claim) we can always find an argument
(sound, I claim), constructed along the lines of the above models,
for the conclusion that the agent is not responsible for that state
of affairs.
The intuitive plausibility of this conclusion can be shown if
we think in terms of the following rather fanciful picture. We are
imagining cases in which an agent "gets to" a certain state of
affairs by following a particular "causal road," a road intentionally
chosen by him in order to "get to" that state of affairs. But,
because this state of affairs is a universal, it can be reached by
various causal roads, some of them differing radically from the
road that is taken. And, in the cases we imagine, all the causal
roads that the agent could take, all that are open to him, lead to
this same state of affairs. Perhaps the point of this fanciful talk
about "roads" will be clearer if we look at the case of an agent
who is unable to prevent a certain state of affairs from obtaining,
where this case involves roads in a literal sense. Suppose Ryder's
horse, Dobbin, has run away with him. Ryder can't get Dobbin
to slow down, but Dobbin will respond to the bridle: whenever
 ABILITY AND RESPONSIBILITY

Ryder and Dobbin come to a fork in the road or a crossroads, it

is up to Ryder which way they go. Ryder and Dobbin are ap-
proaching a certain crossroads, and Ryder recognizes one of the
roads leading away from it as a road to Rome. Ryder has con-
ceived a dislike for Romans and so (having nothing better to do)
he steers Dobbin onto the road he knows leads to Rome, motivated
by the hope that the passage of a runaway horse through the
streets of Rome will result in the injury of some of her detested
citizens. Unknown to Ryder, however, all roads lead to Rome:
Dobbin's career would have led him and Ryder to Rome by some
route no matter what Ryder had done. That is, Ryder could not
have prevented C(Ryder passes through Rome on a runaway
horse). Is Ryder responsible for this state of affairs? It is obvious
that he is not. And it seems obvious that he is not responsible for
this state of affairsjust because he could not have prevented it. I
conclude that Frankfurt-style counterexamples cannot be used
to show that PPP2 is false.
The universals that PPP2 is "about" are states of affairs; but
if we had examined a principle, otherwise similar, about "event-
universals" (for example, "its coming to pass that Caesar dies")
we could have employed arguments that differed from the above
arguments only in verbal detail.
It has been suggested to me19 that these arguments appear
less plausible if one reflects on the fact that essentially similar
arguments could be used to show, for example, that Gunnar did
not bring about C(Rid1ey is killed) or that Gunnar's pulling the
trigger did not cause this state of affairs. It is certainly true that if
the above arguments are sound, then similar arguments can be
used to show that Gunnar did not bring about C(Rid1ey is killed)
and that his bodily movements did not cause this state of affairs
to obtain. But these conclusions appear to me to be simply true.
Let us concentrate on
(1) Gunnar did not bring about C(Rid1ey is killed).
Why should anyone think (1) is false? It would be clearly invalid
to argue that (1) is false since Gunnar did something logically
or causally sufficient for C(Rid1ey ip killed), for by the same argu-

l9 By the editors of The Philosophical Review.

 PETER VAN INWAGEN

ment we could establish the falsity of the (true) proposition that

Gunnar did not bring about C(Rid1ey is mortal). O r consider
the case of Ryder and Dobbin. In turning down a certain road,
Ryder did something causally sufficient for passing through
Rome on a runaway horse, but would anyone want to say that
Ryder brought about the (for him inevitable) state of affairs
C(Ryder passes through Rome on a runaway horse)?
The states of affairs we have been considering are universals.
There are many ways the concrete particulars that make up our
surroundings could be arranged that would be sufficient for their
obtaining. What Gunnar and Ryder can bring about is which of
these possible arrangements of particulars (which murderer,
which road) the universals will be "realized in"; that some arrange-
ment or other of the particulars will realize these universals is some-
thing totally outside their control; it is not something they bring
about. Here is an analogy involving another sort of universal,
properties. Chisel is a sculptor and sculpts the heaviest statue
that ever was or will be, The Dying Whale. Thus Chisel brings
it about that a certain particular, The Dying Whale, exemplifies
the property of being the heaviest statue.20But he does not bring
Perhaps some philosophers would be disinclined to call the property of
being the heaviest statue there ever was or will be a universal, on the ground
t h a t i universal must be "sharable," must be capable of being exemplified by
more than one object. And, for similar reasons, it might be held that what I
have called "states of affairs" are not true universals, since each of them either
obtains or fails to obtain without further qualification, whereas a state of affairs
that was truly a universal should be capable (say) of obtaining in 1943 but not
in 1956 (cf. n. 15), or of obtaining in both Britain and the United States but
not in France. Well, let us say that our "states of affairs" and properties like
being the heaviest statue are, if not "true" universals, a t least cross-world uni-
versals. A property or other abstract object is a cross-world universal if there are
worlds W , and W, such that x falls under it in W , a n d y falls under it in W, and
x # y. (I use the words "fall under" with deliberate vagueness; what "falls
under" a property is whatever has it; what "falls under" a state of affairs is
whatever arrangement of particulars realizes it.) If this usage is a n extension
of traditional philosophical usage, it is a very natural one; I call, e.g., C(Gunnar
kills Ridley) a "universal" because it is not "tied to" any given arrangement
of particulars. I do not pretend that these remarks are very precise. Certainly
the notion of a n "arrangement of particulars" could do with some clarifica-
tion. For example, it is not clear what should be said about states of affairs
that, unlike those discussed above, involve o n h a single particular. (Let us say
that a state of affairs involves a articular if that articular is such that its exis-
tence is entailed by the obtaining of that state of affairs.) Consider, for example,
C(there is such a building as the Taj Mahal). Are there many "arrangements
 ABILITY AND RESPONSIBILITY

it about that this property is exemplified, since, no matter what

he had done, this property would "automatically" have been
exemplified by something or other: he causes something to
exemplify this property, but he does not cause this property to be
exemplified.
In affirming (I), I do not mean to affirm the falsehood
(2) Gunnar did not bring about Ridley's death,
where "Ridley's death" denotes an event-particular (individuated
from other particulars in virtue of having different causal ante-
cedents), one that is also perhaps denoted by "Ridley's death
on Thursday," "the only death Gunnar ever caused," and so on.
Anyone who feels inclined to reject (1) should make sure that this
inclination does not arise from a failure to distinguish between
(1) and (2). To revert to the sculpture example, (1) and (2) stand
to each other roughly as
Chisel did not cause the property of being the heaviest statue
to be exemplified,
and
Chisel did not cause (the particular thing that is) the heaviest
statue to exist,

of particulars" in which this state of affairs could be realized? Tentatively,

I should say Yes. I should think that "the arrangement of particulars that
realizes a given state of affairs" should in general be taken to be an arrange-
ment of a broader class of particulars than those it "involves." For example,
C(there are humans) does not in the strict sense defined above involve you
or me (in fact, no contingent being is such that this state of affairs involves it),
but you and I are, in a very intuitive sense, among those particulars the arrange-
ment of which realizes it. Similarly, though no block of marble is such that
C(there is such a building as the Taj Mahal) involves it-at least on the
assumption that mereological essentialism is false-many blocks of marble
would seem to be among those particulars the arrangement of which realizes
it. O r even if we do not consider parts of the Taj Mahal, we must admit that
the state of affairs we are considering would obtain if the Taj Mahal were
differently placed or differently oriented; and it seems intuitively correct to say
that if the place or orientation of the Taj Mahal were different from what it
in fact is, then C(there is such a building as the Taj Mahal) would be
realized in a different arrangement of particulars.
 PETER VAN ZNWA GEN

stand to each other. The former is, as I argued above, true, and
the latter false.'l
So, it would seem, we are unable to devise a Frankfurt-style
counterexample either to PPPl or to PPP2. If our attempts at
counterexamples looked initially plausible, this, I think, was due
to a confusion. When we hear the Gunnar-Ridley story, it seems
correct to say that it follows from the story that Gunnar is respon-
sible for Ridley's death and that Gunnar could not have prevented
Ridley's death. But "Ridley's death" is ambiguous. If we are
using this phrase to denote a universal, then we may say that
Gunnar could not have prevented Ridley's death, but not that he
was responsible for Ridley's death. If we are using this phrase to
denote a particular, then we may say that Gunnar was responsible
for Ridley's death, but not that he could not have prevented it.
This result might lead us to wonder whether Frankfurt's
counterexamples to PAP rest on a similar confusion. Suppose
we were to split PAP into two principles, one about "act-partic-
ulars" (event-particulars that are voluntary movements of
human bodies) and one about "act-universals" (that is, things
that could be.done by distinct agents, such as murder, prayer,
or killing Jones at noon on Christmas Day, 1953): should we then
see that Frankfurt's alleged counterexamples to PAP depend for
their plausibility on treating one and the same act as a particular
at one point in the argument, and a universal at another?
I do not think that Frankfurt is guilty of any such confusion.
The "acts" that figure in his counterexamples seem to me to be
treated consistently as universals. If this is the case, it raises two
questions. Let us split PAP into two principles as was suggested
in the preceding paragraph: PAP1 (about particulars) and PAP2
(about universals). The first question: If indeed Frankfurt's "acts"
are universals, he is arguing against PAP2; can his argument be
met by considerations like those we raised in defense of PPP2?
The answer seems to me to be No, but I am not at all sure about
I do not mean to give the impression that one never brings about any state
of affairs. For example, (granting the correctness of the Warren Commission
Report), Lee Harvey Oswald brought about C(Kennedy dies on 22 November
1963). But it is not true that Oswald brought about C(Kennedy dies). That
state of affairs was brought about by God or by Adam and Eve or by no one
at all. Moreover, it is true that Oswald brought about the event-particular,
Kennedy's death.
 ABILITY AND RESPONSIBILITY

this. The considerations raised in defense of PPP2 depended

on our having at our disposal a fairly precise notion of "state-of-
affairs universal," and I am not at present able to devise an equally
precise notion of "act-universal" that I find sati~factory.'~ The
second question: what about PAPl? I do not find this question
interesting, since I do not think that "event-particulars that are
voluntary movements of human bodies" are what we hold people
responsible for. I shall not, however, defend this view here. An
adequate defense of it would be fairly complex, and I do not
think my reasons for thinking what I do on this matter are worth
developing merely to establish a negative conclusion.

An adequate construction of such a notion would require the introduction

of a canonical language for act-universals. I am unable to devise a language
for this purpose that comes close to satisfying me. Even without having such
a language at my disposal, however, I think I see a serious obstacle to any
attempt to refute Frankfurt's arguments against PAP2 by raising considerations
like those used to defend PPP2 in the text. Let us suppose that "the act of killing
Ridley" denotes a certain act-universal, an act such that it, that very act, could
be the act of any among a number of agents and be performed under a great
variety of conditions. Consider the following Frankfurt counterexample to
PAP2: Gunnar performs the act of killing Ridley; moreover, if he had decided
not to perform it, some third party, Cosser, would have caused him to perform
it. If we were to try to refute this counterexample by arguments parallel to those
we used in defense of PPP2, we should have to find an act-denoting phrase
that stands to "the act of killing Ridley" roughly as "C(eitker Pistol or Gunnar
or someone else kills Ridley)" stands to "C(Rid1ey is killed)." I am not sure what
such a phrase would look like, but I think something like this:
The act of killing Ridley, either without having been caused to kill Ridley
by anyone, or as a result of having been caused to kill Ridley by Cosser
or someone else.
I am very doubtful whether this phrase makes any sense. T o take a simpler
case, given that there is such an act as eating forbidden fruit, an act one might
perform as a result of one's having been given bad advice, is there such an act
as the act of eating forbidden fruit as a result of having been given bad advice?
I find the notion of such an act difficult to grasp. But if no coherent act-univer-
sal-name can be found to play the formal role played by "C(either Pistol or
Gunnar or someone else kills Ridley)" in our defense of PPP2, then no parallel
argument in defense of PAP2 can be constructed.
These considerations, of course, do not show that Frankfurt's attack on PAP
is successful. They do, however, raise serious doubts about the possibility of
defending PAP against this attack by constructing an argument formally
parallel to our argument in defense of PPP2.
 P E T E R VAN IN W AGEN

We have shown that three principles relating ability and

responsibility cannot be refuted by Frankfurt-style counter-
examples:
PPA A persan is morally responsible for failing to perform a
given act only if he could have performed that act.
PPPl A person is morally responsible for a certain event only
if he could have prevented it.
PPP2 A person is morally responsible for a certain state of affairs
only if (that state of affairs obtains and) he could have
prevented it from obtaining.
Now consider three versions of incompatibilism:
If determinism is true, then if a given person failed to per-
form a given act, that person could not have performed that
act.
If determinism is true, then no event is such that anyone
could have prevented it.
If determinism is true, then if a given state of affairs obtains,
then no one could have prevented that state of affairs from
obtaining.
Obviously, if these three theses are true, then (since PPA, PPP1,
and PPP2 are true) it follows that determinism entails that no
one has ever been or could ever be responsible for any event, state
of affairs, or unperformed act. Moreover if the following schema
R If S is responsible for cPing, then there is some event or state
of affairs for which S is responsible,

I think I am justified in calling these three theses "versions" of a single

doctrine, since, $there were a good argument for any of them, then, I should
think, it could be easily modified to yield a good argument for either of the
others. I have presented arguments for what ?s essentially the first of these three
versions of incompatibilism in "A Formal Approach to the Problem of Free
Will and Determinism," Theoria XL (1974) Part 1, pp. 9-22, and "The Incom-
patibility of Free Will and Determinism," philosophical Studies 27 (1975) pp.
185-199.
 ABILITY AND RESPONSIBILITY

(here "Qing" is to be replaced by any grammatically appropriate

action phrase) is valid, then determinism is (assuming incom-
patibilism) incompatible not only with our being responsible
for the consequences of our acts but for our acts themselves. And
this schema is extremely plausible. I cannot myself conceive of a
case in which an agent is responsible for having performed some
act but is responsible for none of the results or consequences
(either universal or particular) of this act.24
Thus, if all three versions of incompatibilism are true, and if
determinism is true, then there is simply no such thing as moral
responsibility. There is such a thing as moral responsibility only
if someone is responsible for something he has done, or for some-
thing he has left undone, or for the results or consequences of what
he has done or left undone. And the principles for which I have
argued (PPA, PPP1, PPP2, and the validity of schema R) entail
that if incompatibilism is true, then determinism is incompatible
with anyone's being responsible for anything whatever.
Therefore, even if PAP is false,25and even if Frankfurt's "cor-
rect version" of PAP (see footnote 4) cannot be used to show that
determinism and moral responsibility are incompatible, it is
nonetheless true that unless free will and determinism are compat-
ible, determinism and moral responsibility are incompatible.
Thus, Frankfurt's arguments do not, even if they are sound,
rob the compatibilist-incompatibilist debate of its central place
in the old controversy about determinism and moral respon-
sibility. 26

Syracuse University

An obvious argument for the validity of R is this: If someone @s and is

responsible for so acting, then, whatever other events or states of affairs he may
be responsible for, he is at least responsible for its being the case that he @s.But
this argument is unsound. Consider the case (p. 215 above) involving the
counterfactual propensities of atmospheric conditions to cause Gunnar to
decide to kill, and to kill, Ridlcy. I argued that in that case Gunnar is not
responsible for C(Rid1ey is killed). A similar argument could be used to show
that in that case Gunnar is not responsible for C(Gunnar kills Ridley). But it
does not follow that Gunnar is not responsible for killing Ridley. Fo; Gunnar
might have freely decided to kill Ridley and have killed him as a result of this
free decision (and thus be responsible for killing Ridley); nevertheless, I f he
had (freely) decided not to kill Ridley, external factors outside his control
 P E T E R VAN ZNWAGEN

would then have "come into play" and caused him (unfreely, of course) to kill
Ridley. Therefore, while Gunnar is responsible for killing Ridley, he is not
responsible for C(Gunnar kills Ridley freely or Gunnar kills Ridley unfreely)
and hence is not responsible for C(Gunnar kills Ridley). Thus our "obvious"
argument for the validity of R is fallacious.
Nonetheless, R seems to me to be valid. Certainly the case we have just
considered is not a counterexample to its validity. For, in this case, while
Gunnar is not responsible for C(Gunnar kills Ridley), he is responsible for
C(Gunnar kills Ridley without having been caused to do so by atmospheric
conditions). Moreover, he is responsible for the event-particular, Ridley's death.
Of course, if the above arguments are correct, and if determinism and
incompatibilism are both true, then PAP is true: it is vacuously true because
no one, in that case, is responsible for anything he does. Frankfurt, of course,
does not mean to deny that PAP might be, as a matter of contingent fact,
vacuously true.
26 I should like to thank the editors of The Philosophical Review for their care-

..
ful comments on earlier versions of this DaDer. which have led to manv im-
provements. I am especially grateful to them for pointing out to me t h i t an
;rgument I employed was invalid.