Religious Studies, forthcoming
A New Free Will Defense
Alexander R. Pruss
June 25, 2002
One variant of the deductive argument from evil claims that if God is essentially omnibenevolent and essentially omnipotent, then it is logically impossible that God and evil should co-exist. Mackie has argued that considerations of free will are of no help to the theist in refuting the deductive argument form evil, since God can bring about any logically possible state of affairs, and all people always freely doing what is right is plainly a logically possible state of affairs, one that an omnibenevolent deity would a moral obligation to bring about. How, then, is a theist to respond?
A defense is a reply to the deductive problem of evil that argues for the logical possibility of God creating a universe in which at least one evil occurs. Plantinga’s free will defense is a subtle and complicated attempt at answering Mackie by arguing that it is logically possible that even an omnipotent God might be contingently unable to create any significantly free creatures that always do the right thing. Given, however, the great value of significant free will, a God placed in such a predicament could be morally justified in creating significantly free creatures that sometimes go wrong.
On the other hand, Plantinga’s free will defense has the unfortunate feature of relying on F-conditionals, which are subjunctive conditionals of the form “Were someone satisfying P created, then she would freely do A.” But, as Adams has persuasively argued, it is very obscure what if anything such conditionals mean. Adams himself has given a free will defense based on denying that there are F-conditionals, and arguing that prior to deciding what creatures to make, God could not know what creatures would do if created, and hence is to be excused for the evils that they in fact would do.
I shall give a free-will based defense that side-steps the hairy issue of F-conditionals and works even if compatibilistic free will is logically possible. Unlike traditional defenses, I will not be arguing that God would be justified in creating a world that contains an evil, but simply that if God is an essentially omnipotent and omnibenevolent greatest conceivable being, then his nature is not such as to bar him from creating such a world.
By a creature or created being I shall mean a being created by God. Thus, that a creature exists entails that God exists. More generally, a caused being is one that has a cause that brought about its existence. A being essentially has some property if and only if it has it in every possible world in which it exists—i.e., if it would be impossible for the being to lack the property. I take it that God, as traditionally conceived of by Western theism, essentially has the attributes of omnibenevolence and omnipotence, though this assumption will be relaxed in the final section. I shall also take it that God is essentially creator of all contingent beings, other than himself if he is a contingent being. Thus, if God is a contingent being (which I do not believe, but some of my arguments will allow), then in every world in which God exists, God is the creator of all other contingent beings, and if God is a necessary being, then in every world God is the creator of every contingent being. Finally, I will assume that God is a greatest conceivable being, but I will not be assuming that this implies that God is a necessary being.
I will say that a person performs an action significantly freely providing that in this action she is either freely refraining from fulfilling a duty or freely refraining from doing something immoral. A person is then significantly free if she ever performs a significantly free action.
What the deductive argument from evil that I gave says is essentially the following:
(1) The nature of an essentially omnibenevolent and omnipotent God contains a moral principle, which he necessarily acts on as he is omnibenevolent, that would prohibit God from creating a person that does something immoral.
Like Plantinga’s, the argument I shall consider assumes that there is a great value in significantly free acts. In fact, this great value is such that:
God is the greatest conceivable being, and the greatest conceivable being will presumably be capable of creating a significantly free creature—indeed, a being is greater for being able to create a significantly free creature. An alleged moral principle that would make this impossible is simply not an acceptable moral principle for God to be bound by. No free will defense is possible if (2) is false, so in assuming (2), the proposed free will defense is no worse than any competitor.
Although I shall not assume incompatibilism for most of the arguments, nonetheless as William Hasker noted in correspondence, a compatibilist may not see the same kind of deep value in significant freedom as an incompatibilist does. However, it is surely plausible that it is valuable that someone freely refrained from doing what is immoral, since then this refraining is meritorious. One might argue, nonetheless, that there could be some merit even without significant freedom. Although there would then be no merit in a merely dutiful action, there could still be merit in a supererogatory action: one might not be free in fulfilling one’s duty, but one might still be free in exceeding one’s duty. However, a part of the reason we consider supererogatory actions meritorious is that we already think it is good when a person freely fulfills her duty, and so when she freely goes beyond it, we think this all the better. Were we not free in refraining from neglect of our duty, our supererogatory actions would not have this double merit. The starkness of the contrast between the supererogatory action and the immoral action is a part of what gives such value to the supererogatory action.
The structure of the argument from now on will be to argue that a number of different basic theistic ideas about the nature of God, ideas independent of considerations of the problem of evil, each have the property of entailing that:
If this is right, and if at least one of these theistic ideas is something a theist has reason to accept independently of the problem of evil, then by (2) and modus tollens, the theist is within her rights in rejecting, even before learning that there is evil, the principle that God’s essential nature contains a moral principle that would make it logically impossible for God to create a creature that does something immoral. I do not officially endorse all the theistic ideas on the basis of the disjunction of which the overall argument of this paper is run, though they all appear to have a certain plausibility, and their disjunction has even more.
What remains is to argue for (3) and discuss the upshot of this defense. I shall give several arguments for (3) based on independent considerations.
The first two arguments make use of the following premiss which every libertarian will certainly grant, but so will many compatibilists:
The restriction to “caused persons” may seem unnecessary, but will be discussed in Section 6. Once (4) is granted, there are two different ways of showing that traditional theistic ideas about the nature of God, together with (4), entail (3). Both of the ways note that from (4) it follows that:
For suppose that x is a caused person and it is logically impossible that x ever does anything immoral. For a reductio, suppose x has performed a significantly free action, i.e., she has either freely refrained from doing her duty or she has freely refrained from doing something immoral. If she has freely refrained from doing something immoral, then by (4) it was logically possible that she do this immoral action, and so we have a contradiction. If on the other hand she has freely done something immoral, then we have a contradiction to the assumption that it was logically impossible that she does anything immoral. Hence, indeed (5) follows from (4).
Note that in fact (5) may even have some plausibility independently of (4)—someone might think that a contingent person could sometimes commit a significantly free action without it being logically possible for her to have done otherwise, but that for her to have significant freedom she would at some time or another have to have had the logical possibility of doing something immoral.
A significant strand in the theistic traditions holds that God’s existence is logically necessary. For instance, this is entailed by the Catholic dogma of the identity of God and God’s essence, if we take essences to have necessary existence. Alternately, we can consider the arguments given by Findlay in the first part of his atheological argument, where he argued, first, that the theistic tradition needs to hold God to necessarily exist, and, secondly, on rather shaky grounds claimed that there are no necessary beings.
But if God’s existence is logically necessary, then it is logically necessary that every contingent being is a creature of God. Hence, if, as (1) claims, God’s nature contains a moral principle that would make it logically impossible for God to create a creature that does something morally immoral, and if it is logically necessary that every contingent being is a creature of God, then it is logically impossible that a contingent being does anything wrong, and so by (5) it follows that necessarily no contingent being is significantly free. Hence, if (1) holds, it is necessarily the case that no contingent being is significantly free, and (3) has been proved.
However, because the necessity of divine existence is controversial, other arguments for (3) will now be considered.
It is a part of the great theistic traditions that God’s act of creation is something on which our existence utterly depends. One way to explicate this notion is to say that we have creaturehood as an essential property: it would be logically impossible that the same individuals that we are should exist without having been created. If this is true, it is presumably only a special case of a general claim that, necessarily, creaturehood is always an essential property.
A theist can also argue for creaturehood being an essential property apart from reference to the traditions about the utter dependence of us on God. God being essentially omnibenevolent and omnipotent will necessarily design all of his creatures with loving care, a care analogous to that of an artist for his work. But it is an essential property of the work of an artist that it be created by that artist with the kinds of intentions with which it is made. All artwork is an expression of the artist, indeed of the artist considered de re, even if it only expresses that she wishes to efface herself. If a statue just like the David and out of the same materials were to have been made not by Michelangelo but some other sculptor on the same date, that statue would not be the David, just as if Michelangelo had intended to make a statue of David Hume and it turned out to look just like the David, the result would not have been the David. And certainly if the eroding winds and sands were to have shaped that same block of marble into something looking just like the David, the result would not be the David—indeed, it would arguably not be a work of art at all. Likewise, then, if a being is created by God, it is an essential property of that being that it be thus made by God.
Moreover, traditional theism refers to God as not just the efficient cause of all contingent beings, but also as the final cause, the telos for which all creation strives. If God is omnibenevolent, then arguably it is necessarily the case that all beings that he creates will have the glorification of, imitation of and/or union with God as their innate telos. Just as love makes a rigid de re reference to the beloved, so too here “God” rigidly designates the actual creator. However, on Aristotelian grounds, the innate telos of a being is an essential property of that being. Beings are at least partially defined in terms of their telê. (In fact this may be why we may consider works of art to essentially have the property of being made by their artist, for conformity to that artist’s will is a telos of the work of art.) If this is right, and if it is furthermore true that it is impossible for a being to have the glory of, imitation of and/or union with God as an innate telos if God does not exist, then it follows that if a being is created by God, it is impossible that that being exists without God existing. This is particularly clear in the case of persons, and it is only in the case of persons that I will need the essentiality of creaturehood claim. If there is a God, then it is the nature of every person to strive for union with God.
Thus, we have several different considerations, all of which arrive at the conclusion that creaturehood is an essential property, at least in the case of persons. Now then, if the claim (1) holds and “Smith” is a rigid designator of someone who happens to be a created person, then it is logically impossible that Smith does anything wrong. For, being a creature is an essential property of Smith, and hence, necessarily, if Smith exists, she is created by God and, if (1) holds, does no wrong. But if it is logically impossible that Smith does anything wrong, then once again by (4), Smith is not significantly free. Hence we have shown that if (1) holds, then necessarily no created person is significantly free, and so (3) follows.
Both the incompatibilist and many compatibilists will accept:
Of course the incompatibilist and compatibilist will understand the “can” differently. The incompatibilist understands this in terms of a principle of alternate possibilities: if I freely refrain from doing A, then it was not the case that I was causally determined not to do A. The compatibilist, on the other hand, understands this “can” more weakly in the sense of “power”: if I freely refrain from doing A, then I had the power or ability to do A—there is nothing that constrained me from doing A.
But now observe that a part of traditional theism is the idea that God’s existence is something transcendent, entirely out of the reach of our actions. Someone who happens to be a creature, thus, cannot do anything the performance of which would entail that God does not exist. God’s existence, for the theist, sets the range available for creaturely actions. Note that this is a substantial claim about the power of persons—the “cannot” being a “cannot” of lack of power rather than of logical impossibility—rather than the useless de dicto tautology that it is logically impossible that a creature does something whose doing is logically incompatible with God’s existence (since that a creature does something entails that God exists).
If this is right, and if the sense of “cannot” in the claim that a creature “cannot” do anything incompatible with God’s existence matches that of the “can” in (6), we have yet another argument for (3). For, if (1) is true, then that a caused person does something immoral entails that God does not exist, and if someone who is a creature cannot do anything that would entail this, it follows that someone who is a creature cannot do anything immoral. But from this and (6), we conclude that it is impossible that a creature freely refrains from doing something immoral. But by (1), it is also impossible that a creature freely refrains from doing her duty, and so it follows that if (1) is true, then it is impossible that a creature is significantly free, since if a creature is significantly free, at least once she either freely refrains from doing her duty or freely refrains from something immoral.
We say a proposition p is explanatorily prior to q if p serves as part of an explanation of q. Then the incompatibilist will tend to accept the following claim:
(7) Necessarily, if a contingent person freely refrains from doing A, there is no proposition explanatorily prior to the proposition that she refrains from doing A which is logically incompatible with her doing A.
But if someone is created by God, then the proposition that she was created by God is explanatorily prior to all of her actions. First of all, it is prior to her actions because existence is prior to doing, and for a creature, God’s creative act is prior to the existence, whereas explanatory priority is arguably transitive. Secondly, the creator of a person is responsible for instilling in the person her nature, and the nature of a person conditions, though not necessarily determines, all of her acts, and hence the instilling of this nature is explanatorily prior to the actions. But now if claim (1) is true, then the proposition that someone was created by God entails that she does no wrong. Hence, if A is the doing of an immoral action (described as such or in a way that entails that it is such), then the proposition that God created Smith is logically incompatible with the proposition that Smith does A. Next, if the proposition that God created Smith is logically prior to the proposition that Smith refrains from A, it follows by (7) that Smith does not freely refrain from A. Hence, if (1) and (7) are true, a created being cannot freely refrain from doing an immoral action. Moreover, if (1) is true, then a creature cannot freely refrain from doing her duty. Thus, (1) and (7) entail that no creature is significantly free. Therefore, (3) follows, once again.
Recall how Plantinga’s God, prior to creating anybody, knows what a creature would do if created. Richard M. Gale has argued that this entails that God has freedom-canceling control over people’s actions.
Gale argues for this claim as follows. If I press a button which is linked to an indeterministic process which has a certain chance of releasing a poisonous gas that fills a stadium, and if I know ahead of time that pressing the button would release the gas, then it is true to say that I bring about the release of the gas. This would not be true if I did not know the result of the indeterministic process, because the relevant sense of “bring about” is one that is tied to issues of responsibility. Now, a person is not free if all of her actions are brought about by another person. But if God knows ahead of time what we would do, and creates us in this knowledge, then by analogy with my pressing of the button, he brings about all of our actions. And then we are not free, since all of our actions are brought about by another person.
One answer to Gale’s argument is in an Anscombian argument that a person counts as bringing about a foreseen effect if and only if that person intends that effect. (This is obviously tied to the principle of double effect.) One criterion for what is intended is that if I intentionally bring about an effect, then I know by virtue of my intentional knowledge alone that the effect will transpire. But if I merely foresee the effect without intending it, then I must proceed inferentially to know this effect. God, on this argument, foresees how we would choose if created, but does not always intend us to act in this way, and hence does not bring about these actions in an intentional, and hence freedom-canceling, way.
Now, this answer to Gale opens up the way for another argument for (3). First, I need the following premiss:
(8) It is logically impossible that a person x is significantly free and yet another person, y, for every choice of x brings it about that the x makes a right choice or brings it about that x makes a wrong choice.
The compatibilist as well the incompatibilist may well accept this. Control by one person over all of someone’s significantly free actions (as perhaps opposed to control over merely some actions) to an extent that determines what deontic values (permissibility or impermissibility) they fall under, i.e., total control over whether the person does right or wrong, is arguably freedom canceling. Note that just about any free will defense has to assume (8), since if (8) fails, then it is open for God to determine all of a person’s significantly free actions to be right.
But given (8), we can give yet another derivation of (3). Suppose Mackie is right and, as (1) claims, there is a moral principle in God that prohibits him from creating a person who does anything wrong. Recall now Kant’s distinction between acting in accordance with duty and acting out of duty. I act merely in accordance with duty if I do what is dutiful but not because it is dutiful. I act out of duty only if duty is a reason for my action. God being omnibenevolent and hence morally perfect not only fulfills all of his duties, but he acts out of them. Therefore, God intends that his duties be fulfilled, since this fulfillment is a reason for his actions. We have supposed that God is prohibited from creating a person who does anything wrong. Since God acts out of duty and not just in accordance with it, necessarily when he creates a creature, he intends that this be a creature that does nothing wrong. Therefore, God intends that the creature do nothing wrong. But then, by the Anscombian principle, God intentionally brings it about that the creature does no wrong. By (8), it follows that the creature is not significantly free. We have thus seen, once again, that if (1) is true, (2) is not, and have thus given a final argument for (3).
A standard way to object to free-will responses to evil is to claim that according to many of them God is not going to have significant freedom, because God is essentially righteous, i.e., it is logically impossible that God does something immoral. But if God does not have significant freedom, then significant freedom cannot be as valuable as it is claimed to be, since presumably God is not lacking in anything of value.
This is a powerful response, one that can be understood as providing an aporia for the incompatibilist theist or even as an aporia for theists more generally. Since the aporia is largely independent of the problem of evil, a full response would take one outside the subject-matter of this paper. But some brief remarks are in order.
First of all, neither incompatibilist argument in Section 4 nor the freedom-canceling argument in Section 5 yields the conclusion that God lacks significant free will. Both arguments apply only to creatures as they make crucial use of the assumption that creatures are caused to exist by God, and argue that this implies that they could not have significant free will if God could not create a being that does something immoral.
It is also not obvious that the transcendence argument in Section 3 would apply in the case of God, since the crucial assumption there was that creatures in no sense had power over God. In any case, what I am about to say about the arguments in Section 2 will apply to the transcendence argument.
The arguments in Section 2 make use of claim (5) which says that a caused being cannot be significantly free without its being logically possible for her to do something immoral. Would this claim apply in the case of God? It is by no means obvious that it would. If Smith is a caused being and it is logically impossible that Smith does something immoral, then there is a plausible sense in which Smith’s failure to do something immoral is not due to Smith’s merit but due to the antecedent cause of Smith which made Smith exist with the essential nature that precludes Smith from doing immoral deeds, and it is plausible that in that case we should not call Smith “significantly free”. The real cause of Smith’s failure to act immorally is entirely outside Smith in a freedom-canceling way. However, God is an uncaused being. As such, there is no danger of any prior cause claiming credit for God’s not doing immoral things, and thereby taking this credit away from God.
One might insist that God’s nature is responsible for God’s righteousness, and hence God is not responsible for his righteousness and hence is not significantly free. However, it is not immediately clear that it makes sense to talk of a nature being responsible for anything. But even if it does, there is a strong theistic tradition of the doctrine of divine simplicity, which in fact is official dogma for Catholics, according to which God and God’s nature are numerically identical. If this is so, then whatever God’s nature is responsible for, God is responsible for, and so it makes sense to speak of God being significantly free even if because of his nature he cannot do anything immoral. Observe, parenthetically, that one might take this observation to provide reason for a theist to accept the doctrine of divine simplicity, since the doctrine explains how God is responsible for his righteousness.
We have seen that a number of arguments, based on very different premisses, each show that if God were to have the essential property of acting on Mackie’s moral principle that prohibits him from creating a creature that does wrong, then no creature could have significant moral freedom. The great value of significant moral freedom then provides a reason to think that the essential nature of the greatest conceivable being would fail to contain any such principle.
Observe that the arguments based on premisses about divine attributes such as the utter dependence of creatures on God were based on premisses independent of theistic answers to the problems of evil, and hence cannot be said to be ad hoc to the problem of evil. What the arguments have attempted to show is that in claiming that God is barred from creating a being that would go wrong, the arguer from evil is saying something that the theist would be committed to the rejection of even if the theist did not know that in fact there is evil, but merely knew that significant freedom has the value it does. And the special merit of the present defense is that (a) it avoids hairy issues about F-conditionals, and (b) is run from a large disjunctive set of premisses that different people will accept for different reasons.
One might, however, make a more general objection to this defense. The deductive argument from evil that it is opposed to is one that assumes God to be essentially omnibenevolent and essentially omnipotent. But Mackie’s own argument from evil did not assume God to have these properties essentially. Mackie only sought to derive a contradiction between God’s actually being omnibenevolent and omnipotent and there actually existing evil. However, what the present defense shows is that the traditional Western theist based on her background beliefs would have good reason to reject Mackie’s claim that the existence of an omnibenevolent and omnipotent being is incompatible with the existence of evil even prior to learning that there is actually is evil. For the traditional Western theist is committed not merely to God’s being omnibenevolent and omnipotent, but to God’s being essentially such. Moreover, the Western theist has good reason to think that God should be able to make significantly free creatures given the value of them. The above arguments then show that this commits the theist to the claim that God’s essential omnibenevolence and essential omnipotence are logically compatible with God’s creating a universe that contains an evil. But this forces the theist to reject as a part of omnibenevolence any moral principle that prohibits an omnipotent and omnibenevolent being from creating a universe containing an evil.
But perhaps the Western theist should not hold that it is an essential property of God to be omnibenevolent and omnipotent, but only that it is an accidental property? This, however, would not be compatible with other commitments of the Western theist. First of all, plausibly, goodness and power are intrinsic properties, and at least the Catholic theist is committed to God being absolutely simple, and hence in particular to God not having any accidental intrinsic properties. Secondly, many theists accept that God is the greatest conceivable being, and plausibly it follows from this that God is essentially omnibenevolent and omnipotent.
Thirdly, we have a very speculative argument starting with the idea that God is the most perfect possible object for human love. Now, one dimension along which one can measure a love is by looking at its commitment. The more deeply committed a love is, the more deeply true it is that the lover would love the beloved no matter what. Love involves an appreciation of a good. The more permanent, lasting and non-contingent the good, the more committed can the love be insofar as it is an appreciation of this good. Therefore, insofar as love for someone is an appreciation of a good, this love is most strongly committed when this good is one that the beloved has essentially, for then the lover can be committed to have had this dimension of love absolutely no matter what had befallen. Even to imagine a logically possible scenario under which the lover ought not to have had this dimension of love is to weaken the commitment in the love, though of course a human lover recognizes that there could have been contingent circumstances where she, as a matter of fact, would not have loved her beloved. It might seem meaningless to describe a love as having this counterfactual commitment. But I can give meaning to it as follows: In the case of absolute commitment to having loved a person in logically possible counterfactual circumstances C, I am now committed to loving this person should I find out that in fact I am mistaken, perhaps due to hallucinations, in my belief that C does not obtain. And the restriction to logically possible circumstances is not ad hoc, since once one allows logically impossible circumstances, one gets such empty conundrums as whether one is committed to still loving God, understood as a rigidly designating proper name, should one find out that God is not God.
Now, it is true that we may have committed love for a human person who is not essentially good but is only contingently good. But then the love is not absolutely committed in each of its dimensions. While one is committed to loving the beloved as long as she is the same person she is, one is not committed to the dimension of appreciating her goodness as one actually does, because this goodness might cease or might not have been. Thus, insofar as God is supposed to be the most appropriate imaginable object of our love, it is plausible that as many as imaginably possible of those of his attributes which contribute to his lovability should be essential attributes. Since the theist loves God for his omnibenevolence, it is plausible that God’s omnibenevolence should be an essential attribute if this can be imagined, and the presumption is that it can, given that so many theists apparently have done so.
One might try to argue for essential omnipotence on similar grounds, though this case is a little less clear. Or, instead of maximal lovability, one might also bring in awesomeness. The God of monotheism is utterly awesome, and it is plausible that being essentially omnipotent is implicated in this. Alternately, one might argue that God’s sovereignty implies that his free creative action explains all contingent true propositions other than those explained by the free actions of created persons, and so if he is contingently omnipotent, his omnipotence is explained by his having freely brought it about that he is omnipotent, which is logically impossible since arguably any being that can bring it about that it is omnipotent must already be omnipotent.
However, in fact, we do not need God’s essential omnibenevolence and omnipotence to give something like the defense in Sections 3 and 4. In Section 3, the argument was that no creature has the power to do something the doing of which entails that God does not exist. But it is also plausible to say, on exactly the same grounds of divine transcendence, that if God is omnipotent, then no creature has the power to do something the doing of which entails the disjunction that God is not omnipotent or God is not omnibenevolent or God does not exist, and the rest of the argument continues as before to yield the conclusion that there is an incompatibility between the existence of significant free will, essential omnibenevolence, actual omnipotence and Mackie’s principle that an omnipotent and omnibenevolent God does not create a universe in fact containing an evil. In Section 4, the argument adapts, too, to the case of God’s being accidentally omnipotent and omnibenevolent. For God’s power and goodness are explanatorily prior to God’s activity of creation. Thus, that God is omnipotent and omnibenevolent and that God exists is still prior in the order of explanation to creaturely actions, and the argument continues to go through.
Finally, the argument from freedom-canceling control in Section 5 does not in fact make use of either essential omnipotence or essential omnibenevolence, but mere omnipotence and omnibenevolence, both possibly of an accidental sort, as can easily be seen. Thus, if that argument succeeds, it shows that if God is in fact omnipotent and omnibenevolent, and if Mackie’s principle prohibiting an omnipotent God from creating universes containing evil holds, then there is no significant free will, contrary to the theist’s commitment to the possibility of significant free will coexisting with God an omnibenevolent and omnipotent God.
 “Evil and Omnipotence”, Mind 64 (1955); reprinted in Louis P. Pojman (ed), Philosophy of Religion: An Anthology, 3rd edition (Wadsworth, 1998), pp. 186–193.
 Though Mackie’s own deductive argument from evil is subtly different from the one I give above—see Section 7, below.
 See, e.g., Alvin Plantinga, God and Other Minds: A Study of the Rational Justification of Belief in God (Ithaca and London: Cornell University Press, 1990), pp. 131–155.
 Robert M. Adams, “Middle Knowledge and the Problem of Evil”, American Philosophical Quarterly 14 (1977); reprinted in Robert M. Adams, The Virtue of Faith (New York and Oxford: Oxford University Press, 1987), pp. 77–93.
 “God’s Existence Is Necessarily Impossible”, in Louis P. Pojman (ed), Philosophy of Religion: An Anthology, 3rd edition (Wadsworth, 1998), pp. 93–97.
 One may also argue for this by analogy with utterances. Assuming that the meaning of an utterance is part of its identity—if I had intended to say the word “Gift” in German I would have said something different than had I intended to say it in English since the meaning would have been different—and assuming that with respect to the essentiality of origin all utterances are on par, we can argue that being uttered by x is an essential property of an utterance. For it is clear that given the assumption about meaning, the identity of the speaker is essential to the identity of utterance in the case of an utterance of a first-person indexical sentence. If all utterances are the same with respect to the essentiality of the speaker to their identity, it follows that the same is true for non-indexical utterances. It would in fact be rather strange if the utterance “I am human” and the utterance “Socrates is human” were such different items of our ontology that the identity conditions for these utterances were different. Plausibly, all utterances should count as the same kind of item in our ontology, or at least all utterances in a given medium.
 Of course these views of the identity of artwork are controversial. For a brief discussion, see David Carrier, “Art without its artists?”, British Journal of Aesthetics 22 (1982), 233–244. And of course some, as an anonymous referee notes, will take artworks to be Platonic objects. But remember that the main argument of the paper, and indeed the subargument of this section, is run disjunctively: some will find some disjuncts convincing and others will find other disjuncts convincing.
 “Freedom and the Free Will Defense”, Social Theory and Practice 16 (1990), 397–423. See also the discussion of the free will defense in his On the Nature and Existence of God (London: Cambridge University Press, 1991).
 To adapt an example of Gale’s (personal communication), if I know that you are the sort of person who will fulfill a very easy duty I can impose on you, and I impose the duty on you intending and foreseeing that as a result of my imposing it you fulfill it, I have not canceled your freedom in fulfilling it. But the same will not hold, one may argue, if I have like control over all of your significantly free actions.
 Kant interpreters differ on the exact form of the condition. Some insist that duty must be the only reason for the action. Others allow that all that is needed is that duty is a sufficient reason for the action. But they will all agree with the statement that only if duty is at least a reason is the action done out of duty.
 The question depends on whether natures are causes. With Aristotle, I take the answer to be affirmative, and so I cannot use this question as my escape route—instead, I must use the following argument.
 Note the restriction to attributes that could be imagined to be essential. For instance, one might not be able to imagine supererogation to be an essential attribute, and indeed it might cease to be supererogation were it an essential attribute.
 Cf. Jerome Gellman, “Prospects for a Sound Stage-3 of Cosmological Arguments”, Religious Studies 36 (2000), 195–201.
 I am most grateful to Richard M. Gale, William Hasker and David Manley for discussions and comments on earlier versions of this paper. I am also grateful to two anonymous readers for a number of helpful comments and suggestions.