From Determinism and Freedom in the Age of Modern Science, Hook, 1958.

Edwards and Hospers hold that there is an important sense in which we may be said not to be morally responsible for any of our acts or choices. I propose the following as an explicit formulation of their reasoning:

1. If a choice is one we could not have avoided making, then it is one for which we are not morally responsible.

2. If we make a choice under conditions such that, given those conditions, it is (causally but not logically) impossible for the choice not to be made, then the choice is one we could not have avoided making.

3. Every event occurs under conditions such that, given those conditions, it is (causally but not logically) impossible for that event not to occur.

4. The making of a choice is the occurrence of an event.

5. We are not morally responsible for any of our choices.

If we wish to reject the conclusion (5) — and for most of us (5) is difficult to accept — we must reject at least one of the premises.

Premise (1), I think, may be interpreted as a logical truth. If a man is responsible for what he did, then we may say, "He could have done otherwise." And if we may say, "He couldn't help it," then he is not responsible for what he did.

Many philosophers would deny (2), substituting a weaker account of avoidability. A choice is avoidable, they might say, provided only it is such that, if the agent had reflected further, or had reflected on certain things on which in fact he did not reflect, he would not have made the choice. To say of a choice that it "could not have been avoided," in accordance with this account, would be to say that, even if the agent had reflected further, on anything you like, he would all the same have made the choice. But such conditional accounts of avoidability ("An act or choice is avoidable provided only it is such that, if the agent were to do so-and-so, the act or choice would not occur") usually have this serious defect: the antecedent clause ("if the agent were to do so-and-so") refers to some act or choice, or to the failure to perform some act or to make some choice; hence we may ask, concerning the occurrence or nonoccurrence of this act or choice, whether or not it is avoidable. Thus one who accepted (5) could say that, if the agent's failure to reflect further was itself unavoidable, his choice was also unavoidable. And no such conditional account of avoidability seems adequate to the use of "avoidable" and "unavoidable" in questions and statements such as these.

If we accept a conditional account of avoidability, we may be tempted to say, of course, that it would be a misuse of "avoidable" to ask whether the nonoccurrence of the antecedent event ("the agent does so-and-so") is avoidable. But the philosopher who accepts (5) may well insist that, since the antecedent clause refers to an act or a choice, the use of "avoidable" in question is not a misuse.

What, then, if we were to deny (3)? Suppose that some of our choices do not satisfy (3) — that when they are made they are not made under any conditions such that it is (causally) impossible (though logically possible) for them not to be made. If there are choices of this sort, then they are merely fortuitous or capricious. And if they are merely fortuitous or capricious, if they "just happen," then, I think, we may say with Blanchard that we are not morally responsible for them. Hence denying (3) is not the way to avoid (5).

We seem confronted, then, with a dilemma: either our choices have sufficient causal conditions or they do not; if they do have sufficient causal conditions they are not avoidable; if they do not, they are fortuitous or capricious; and therefore, since our choices are either unavoidable or fortuitous, we are not morally responsible for them.

There are philosophers who believe that by denying the rather strange-sounding premise (4) we can escape the dilemma. Insisting on something like "the primacy of practical reason," they would say that since we are certain that (5) is false we must construct a metaphysical theory about the self, a theory denying (4) and enabling us to reconcile (3) and the denial of (5). I say "metaphysical" because it seems to be necessary for the theory to replace (4) by sentences using such terms as "active power," "the autonomy of the will," "prime mover," or "higher levels of causality"—terms designating something to which we apparently need not refer when expressing the conclusions of physics and the natural sciences. But I believe we cannot know whether such theories enable us to escape our dilemma. For it seems impossible to conceive what the relation is that, according to these theories, holds between the "will," "self," "mover," or "active power," on the one hand, and the bodily events this power is supposed to control, on the other—the relation between the "activities" of the self and the events described by physics.

I am dissatisfied, then, with what philosophers have proposed as alternatives to premises (1) through (4) above, but since I feel certain that (5) is false I also feel certain that at least one of the premises is false.

From Agents, Causes, and Events: Essays on Indeterminism and Free Will, O'Connor, Oxford, 1995.

In earlier writings on this topic, I had contrasted agent causation with event causation and had suggested that "causation by agents" could not be reduced to "causation by events." I now believe that that suggestion was a mistake. What I had called agent causation is a subspecies of event causation. My concern in the present study is to note the specific differences by reference to which agent causation can be distinguished from other types of event causation.

We cannot hope to succeed in this task unless we try to cope with the very difficult concept of causation — event causation. And this means, in turn, that we should have a clear conception of the ontological status of events and, in particular, of their relation to attributes or properties and of their relation to individual things.

We begin with the ontological question.

The Nature of States

Events are here construed as being a subspecies of states. The concept of a state is taken as undefined, but it can be clarified in several different ways.

Suppose that you are reading. Then the following entities are involved: (1) that contingent substance which is yourself; (2) that noncontingent thing, which is the property of reading; and (3) that contingent state which is you reading. It will be useful to say that you are the substrate of that state and that the property of reading is the content.

We introduce the following twofold definitional abbreviation:

D1 x is the substrate of the state y, and z is the content of the state y
= Df. y is that state which is x-exemplifying-the-property z.

We may now formulate a general principle, telling us that every state is necessarily such that it has the substrate that it has.

A1 For every x, if there exists the state, x-being-F, then x-being-F is necessarily such that it is a state of x.

From the fact that that state, which is you reading something, is necessarily such that it is a state of you, it does not follow, of course, that you are necessarily such that you are reading something.

Higher Order States and the Concept of an Event

We have assumed that, for every x there is the state x-being-F, if and only if x is F. Our assumptions imply, therefore, that there are infinitely many states. They also imply that there is an infinite hierarchy of states. The hierarchy may be illustrated this way:

(1) x-being-F

(2) (x-being-F)-being-G

(3) [ (x-being-F)-being-G]-being-H

An instance of (1) would be Jones walking. An instance of (2) would be (Jones walking) being strenuous. And an instance of (3) would be (Jones walking being strenuous) contributing causally to (Jones being tired).

We could say that a first-order state is a state that has a non-state as its substrate. Second-order states will have first-order states as their substrates. Second-order states are illustrated by those states that consist of one first-order state contributing causally to another first-order state.

In order to say what an event is, we refer to the concepts of a first-order state and of a second-order state:

D2 x is a first-order state = Df. x is a state of a substance.
D3 x is a second-order state = Df. x is a state of a first-order state.

We are now in a position to characterize the concept of an event.

D4 x is an event = Df. x is either a first-order state or a second-order state.

In some of his earlier writings on the concept of an event, Jaegwon Kim suggests a theory according to which all events would be first-order states. Such a restriction provides no place for those paradigmatic events that consist of one event contributing causally to the occurrence of another event. Examples are the striking of a match contributing causally to the burning of a piece of paper; the treatment of a patient contributing causally to the patient being cured; and the rush of the sea contributing causally to the destruction of the pier. Here we have second-order events that relate first-order events.

Events and the Concept of Causation

Causation cannot be analyzed by reference to the "constant conjunction" of events. Most investigators agree that the concept of causation is nomological. It presupposes the concept of physical necessity, a concept that is usually expressed by reference to "laws of nature."

How are we to interpret "It is a law of nature that if A occurs then B occurs"? Speaking somewhat loosely, we may say that the reference to "a law of nature" is intended to call attention to two types of necessity: that imposed "by logic" and that imposed "by nature." How, then, might one distinguish "laws of logic" from "laws of nature"?

If it is "a law of logic that if A then B," then conceivably a rational being could know a priori, just by reflection, that it must be the case that if A occurs, then B occurs. Some philosophers would say: "Every possible world is such that, if A occurs in that world, then B also occurs in that world." But rational reflection does not suffice to tell us what the laws of nature are.

It is possible that there occurs a conjunction of events A that taken together will constitute a sufficient causal condition of B without logically implying B. The states that would make up such a conjunction are "partial causes," or "contributing causes," of B. One of the most common errors to which discussions of freedom and causation are subject is that of confusing partial or contributing causes with sufficient causal conditions. The contributing causes that make up a sufficient causal condition of an event B need not themselves be sufficient causal conditions of B.

Let us consider an example.

We will define the concept of a sufficient causal condition, not by reference to a set of states or events, but by reference to those properties we have called the contents of the states or events. Thus we may have:

D5 S is a sufficient causal condition of E = Df. S is a set of properties such that the conjunction of its members does not logically imply E; and it is law of nature that, if all the members of S are exemplified by the same thing at the same time, then E will be exemplified either at that time or later.

In referring in the definition to the properties of the thing that undergoes the effect, we do not thereby exclude the properties of other things that happen to be in the environment of the thing that undergoes the effect. The piece of wood bums in part because of the presence of oxygen in the environment. But in that case one of the properties of the piece of wood is that it happens to be in an environment in which oxygen is present.

The realistic view of properties, here presupposed, implies that, for any two properties, P and Q, there is also the property, P-and-Q. It also implies that, for any two properties, there is also the property of having those two properties.

Why say that the effect of the sufficient causal conditions must be exemplified either at the same time as or later than the members of that condition? To say this is simply to say that the effect not precede its cause. The effect, in other words, will not be exemplified before the members of any sufficient causal condition of that effect are exemplified.

We next single out the concept of a minimal sufficient causal condition:

D6 C is a minimal sufficient causal condition of E = Df. C is a sufficient causal condition of E; and no subset of C is a sufficient causal condition of E.

We may speak of a "subset" of a sufficient causal condition, since such a condition, according to our previous definition, is a set of properties.

If your action is a part or member of a minimal sufficient causal condition of an event, then, clearly, the action contributes causally to that event. It is a partial cause of the event (which is not to say, of course, that it is the cause of that event).

D7 That state which is x-being-C contributes causally to that state which is y-being-E = Df. C is a member of a set S of properties that are all exemplified by x at the same time, and S is a minimal sufficient causal condition of E.

What of those situations where the effect is overdetermined? Two marksmen shoot at the victim; they are each successful and the two shots do their work at precisely the same time. Given the one shot, the other shot was not needed to bring about the effect. Both shots, therefore, would not be a part of a minimal sufficient causal condition; yet each contributed. We need not, therefore, revise the definition of a minimal sufficient causal condition.

Freedom and Indeterminism

The concept of being able to undertake is somewhat more broad than that of being free to undertake. It is only when you "could have done otherwise" that your undertaking may be said to be free.

I have not used the expression "free will," for the question of free dom, as John Locke said, is not the question "whether the will be free"; it is the question "whether a man be free."' The question is whether the agent is free to undertake any of those things he does not undertake and whether he is free not to undertake any of those things he does undertake.

Consider the question: Is the person free to bring about what it is that he or she undertakes to bring about? This is not the question with which we have been concerned. But many would have us think that it is. Many philosophers and theologians whose views may seem unduly to curtail our freedom have tried to soften this consequence by redefining the problem of freedom. Thus Jonathan Edwards, using the the verb "to will" where I have used "to undertake," would have us think that the question is this: Is the person free to do what it is that he wills to do? This question is not difficult to deal with. We may answer it affirmatively by pointing out that on occasion people do do the things that they will to do; that is to say, they do bring about what it is that they undertake to bring about. Those who put this question are asking about what Thomas Aquinas called the actus voluntatis imperatus. They are simply asking: Do we ever bring about the things we intend to bring about? But our question might be put by asking: Are we free to will the things that we do will? Thus they have tried to bypass the more fundamental question of the freedom of the .

Objection: "An undertaking that has no sufficient causal condition is completely arbitrary; it is simply a random event for which the agent has no responsibility at all. Hence your proposal implies that we are really not responsible for anything that we do."

From the fact that an undertaking has no sufficient causal condition, it does not at all follow that it is "completely arbitrary" or "random." Nor does it follow that the person has no responsibility for that undertaking. For even if the undertaking has no sufficient causal condition, there are several ways in which other events may contribute causally to that undertaking.

Suppose you are in the middle of a room that has many exits and you hear someone screaming "Fire!" Your hearing the scream may complete a sufficient causal condition for your undertaking to leave the room. But, so far as each particular exit is concerned, there may be no sufficient causal condition for your undertaking to leave by that exit rather than by any of the others. Suppose, then, that you undertake to leave by the exit that is north of you and that you succeed. In this case, your undertaking to leave may have a sufficient causal condition but your undertaking to leave by the northern exit may not. The latter event, although it has no sufficient causal condition, was such that the shout of "Fire!" contributed causally to it.

And there are other ways of contributing causally to an event that has no sufficient causal condition.

An automobile driver with a long trip ahead of him interrupts the trip to get something to eat. of the two available restaurants, he chooses the one that serves alcoholic beverages, not with the intention of getting a drink, but because he thinks the food is better there. He knows full well, however, that he could easily succumb to the temptation to have a drink. Suppose now that he does succumb to that temptation and endeavors to have a drink. Even if this endeavor has no sufficient causal condition, the driver's beliefs, motives, and desires contributed causally to its occurrence. And if the results of that endeavor should be still more drinks and a subsequent serious accident, then one would be completely justified in holding him responsible for that free endeavor and for everything to which it led.

If an agent's undertaking contributes causally to a certain event, then he, the agent, may also be said to contribute causally to that event. Agent causation need not be construed as an alternative to event causation; we may think of it as a subspecies of event causation. For "Agent S contributed causally to so-and-so" may be construed as: "There was a certain thing that agent S undertook and his undertaking that thing contributed causally to the occurrence of so-and-so."