Contingency in Classical Creation: Problems with Plantinga's Free-Will Defense

Ben Huff

1. Alvin Plantinga's Free Will Defense is a milestone in Christian apologetics. It introduced much-needed precision into the discussion of the Problem of Evil. It helped to show how rich a Christian's philosophical options are, and to increase respect for Christian theology in general. It is perhaps the most persuasive response to the Problem of Evil from a classical Christian perspective. In these senses it is a practical success. However, as I show below, it fails in its main theoretical goal: Plantinga does not deliver the argument he promises, and there are a number of reasons why it seems no such argument can be available to him.

2. That Plantinga's Free Will Defense fails is of interest to Latter-Day Saints for a few reasons. One is simply that philosophical problems with classical Christian belief increase the comparative plausibility of the LDS position. I find that both the strength and weakness of Plantinga's defense lend plausibility to LDS beliefs in more specific ways. The strength of Plantinga's argument is its concession that there must be contingent limits to God's power. This claim is more obviously at home among LDS beliefs than classical Christian beliefs. The weakness of Plantinga's argument is that he fails to reconcile the degree of limitation he relies on with other key claims of classical Christianity. He shows some contingent limits on God's power are logically necessary, at least on a certain account of freedom, but to solve the problem of evil requires a much greater degree of limitation than is logically necessary. Plantinga fails to show that this greater degree of limitation is compatible with omnipotence. Since this limitation is supposed to follow from features of an uncreate population of abstract entities, one may also wonder whether the account of creation Plantinga relies on in his Free Will Defense is consistent with the classical Christian belief in creation ex nihilo. A thorough examination of these problems yields many results of interest to Latter-Day Saints concerned to understand the implications of the existence of evil for beliefs about God and his relationship to his creatures. In the interest of brevity, in this paper I will focus on showing simply that Plantinga's Free Will Defense fails, leaving an examination of the consequences for another occasion.

Plantinga's Task

3. In The Nature of Necessity and other texts, Alvin Plantinga presents his Free Will Defense as a response to what Robert Adams has called the abstract logical problem of evil: the problem of showing that the propositions

(1): God exists and is omniscient, omnipotent and wholly good.
(2): There is evil.
are logically consistent, without considering the kind or amount of evil.1 In particular, he responds to an atheological argument, advanced by J.L. Mackie, presupposing that logically prior to creation,
L: God, if omnipotent, would have been able to actualize just any possible world.
After defeating this atheological argument by refuting L, Plantinga attempts to preempt a whole class of such arguments by proving that even though God is omnipotent, it is possible that
R: It was not within God's power to actualize a possible world displaying a better mixture of good and evil than the actual world displays.

This would be an important result, because if R is possible even though God is omnipotent, then theists can rationally suppose that there are morally sufficient reasons why God did not create a better world. Yet while his refutation of L is carefully argued, Plantinga's support for possibly-R is unsatisfying.

4. In this paper, I show the weakness of Plantinga's support for the possibility of R. Then I offer reasons to think R is impossible if God is necessarily omnipotent. While many possible worlds are beyond God's power to actualize, I argue that omnipotence, as classically construed, implies limitations to which combinations of possible worlds could be beyond God's power to actualize, and that R entails a breach of these limits.

The Strength of Plantinga's Argument

5. Plantinga presupposes that for an action to be morally good or morally evil, it must be free, and he assumes a contracausal notion of freedom. He also assumes for the purposes of his argument that there exist counterfactuals of creaturely freedom that specify how a given free being would act if created and placed in certain circumstances. If there are no such counterfactuals, then it is not clear that God decides which world becomes actual, and so the question of why he didn't create a better world is of doubtful relevance, as long as he does create free creatures. The counterfactuals of interest refer not to world-bound individuals but to personal essences, abstract entities that are instantiated in various possible worlds. I do not dispute these assumptions in this paper.

6. In Plantinga's Free Will Defense, the main work is in the argument against L, above, also called Leibniz's Lapse. It seems that if a benevolent God could have created a better world than the actual, then he would have, but few people find it plausible that no possible world is better than the actual world. Hence Leibniz's Lapse has problematic implications, and Plantinga is happy to reject it.2

7. To disprove Leibniz's Lapse, Plantinga starts by pointing out that it is incoherent to suppose that God could cause a person P to freely do an action A. For, for P to be free regarding whether to do A or not, P's action must not be fully determined by any external cause, such as God. This is just what contracausal freedom means. Of course, God does actualize a world that includes both states of affairs He causes to obtain and states of affairs that are or follow causally from the free actions of his creatures. Since the world that results in this way is a cooperative effort, we say that God weakly actualizes such a world.

8. Thus in each possible world W, there is a largest state of affairs that God directly causes to obtain, T(W). T(W) specifies the existence of certain creatures, and these creatures' counterfactuals of freedom indicate whether what would counterfactually follow from T(W) is W, or some other world, say W*, such that T(W) = T(W*). If T(W) does not counterfactually imply W, then it is beyond God's power to weakly actualize W. In this case we say W is infeasible, to use Thomas Flint's term. Since God's inability to weakly actualize some possible worlds follows from purely logical considerations, it is compatible with his being omnipotent.

The Deficiency

9. With Leibniz's Lapse refuted, Plantinga seems to think it is easy to see the possibility of one answer to the question of why God did not create a better world, namely

R (reworded): All the possible worlds displaying a better mixture of good and evil than the actual world displays are among the infeasible worlds.
But a careful reading of Plantinga's text yields no argument from the negation of L to possibly-R. In Alvin Plantinga, he acknowledges that an argument is needed, but rather than giving one, he simply refers to having done so in The Nature of Necessity. In The Nature of Necessity, Plantinga acknowledges that to show the error of Leibniz's Lapse, or L, is not enough to "settle the issue in the Free Will Defender's favor"3. To reason directly from (i) the possibility that any better world, considered individually, be infeasible to (ii) the possibility that all better worlds be simultaneously infeasible, would be fallacious. This fallacy is the well-known fallacy of compossibility, reasoning of the form:
(<>A & <>B) => <>(A & B)
Or, in this case,
[<>(W is infeasible)& <>(W* is infeasible) & . . .]
=> <>[(W is infeasible) & (W* is infeasible) & . . .]
10. Knowing this will not do, to show what remains Plantinga explores an idea he calls transworld depravity. Roughly, if a personal essence suffers from transworld depravity, its instantiation sins in every feasible world in which it has morally significant freedom. If every personal essence suffers from transworld depravity then a proposition similar to R holds, namely
R*: It is beyond the power of God to create a world containing moral good but no moral evil.4
11. Yet Plantinga makes no argument for the consistency with God's omnipotence of any essences', or any group of essences' suffering from transworld depravity. Thus he presents no argument for the consistency of R* with God's omnipotence. The discussion of transworld depravity serves merely as an illustration of one case in which R* would hold. Then Plantinga suggests that similar reasoning regarding a property similar to but nastier than transworld depravity would show the consistency of R itself with God's omnipotence. Yet, again he merely asserts the consistency of God's omnipotence with every personal essence's suffering from this property.5

12. Thus Plantinga has not shown that R is possible; he has merely removed one reason to think R impossible, namely L. Technically, this much is enough to show that this Free Will Defense fails in its main theoretical goal. Still, intuitively, it may not be obvious why one should not suppose that R is possible anyway. With L disproved, as far as we know, R could be true. Yet I believe R is neither true nor possible, if God is omnipotent. I offer some reasons why.

Possible Worlds vs. World-Scripts

13. It follows from Plantinga's premises about freedom that many possible worlds are infeasible. On the basis of his argument against Leibniz's Lapse, it appears that for any given possible world including free creatures, that world may be infeasible for all we know. This is not to say, however, that just any class of such worlds might all be infeasible together. Indeed, I argue that certain combinations of possible worlds can't all be infeasible, given God's omnipotence. To explain my argument, I must introduce further technical terms used by Thomas Flint in explaining the Molinist metaphysics that Plantinga's Free Will Defense implicitly assumes.

14. As we saw in refuting L, the truth or falsity of various counterfactuals of creaturely freedom entails the feasibility of certain possible worlds as opposed to others. Any appropriately complete set of counterfactuals of creaturely freedom that God might know to be true, Flint calls a creaturely world-type. In any given possible world, one world-type will be true, and while some worlds may share a world-type, others will have differing world-types. Yet if questions of which worlds are feasible are to have meaning logically prior to God's act of creative will, then some world-type must be actual prior to any particular possible world's being actual. Since counterfactuals of creaturely freedom are independent of God's will, so is the world-type. Thus the feasible worlds are just those worlds that include the actual world-type. In general, the set of all the possible worlds that share one given world-type is called a galaxy.

15. Thus, the truth of one world-type is a state of affairs that identifies a galaxy, a set of possible worlds that are all feasible, or all infeasible, together. I now describe a state of affairs I call a world-script with which to identify a set of possible worlds that cuts across galaxies.

16. The world-script of a given possible world is the largest temporally invariant state of affairs in that world that is not determinate with respect to which personal essences participate in it. A world-script of the actual world would include two sons' being born to a man named Isaac and their being named Esau and Jacob and living the sorts of lives they actually lived, but it would not specify which personal essences are instantiated as these people. More broadly, a world-script describes everything that happens in a world but not who does it or to whom it happens.

17. A world-script plus a mapping assigning each person in the world-script to a personal essence is sufficient to determine a possible world. Thus, associated with one world-script will be a number of possible worlds, each including a different permutation of personal essences who live out the same events. The class of all possible worlds including one world-script, say, C, is a script-class, the C-class. Perhaps most personal essences, left to act freely where Jacob acted freely, would not have done as he did. If so, then most of the possible worlds in the script-class of the actual world are infeasible. It may sound odd to speak of another personal essence who would have done just what Jacob did under the same circumstances, but if it was logically possible for one personal essence, say, Israel, to do it, it is logically possible for someone else to do it to do everything but be Israel. But to say this is to say that there is another personal essence whose instantiation in some possible world does the same things Israel did, under the same circumstances. There may even be other feasible worlds in the script-class of the actual world, other feasible worlds in which the instantiations of a different set of personal essences do all the same things as are done in this world.

18. As in a possible world W there is a state of affairs T(W) which is the largest state of affairs God himself causes to obtain, so in a world-script C there is a settingS(C), the largest state of affairs indeterminate with respect to personal essences' participation that God causes to obtain. Assignment of personal essences to the lives persons lead in a possible world W is included in T(W), so that the setting plus a mapping M(W) of essences to lives is sufficient to determine T(W). Two world-scripts may share the same setting, just as two possible worlds may share the same total divine act of creation. That is, for world-scripts C(W) and C(W*) it may be that S[C(W)] = S[C(W*)]), just as for possible worlds W and W* it may be that T(W) = T(W*). Now, in the case of these possible worlds, it is logically necessary that at least one of W and W* is infeasible, as shown in the refutation of Leibniz's Lapse, although which is infeasible is logically contingent. For world-scripts, however, the matter is different.

19. We will say that a world-script is feasible just when at least one feasible world includes that world-script. Now, given one mapping of essences to lives in a setting S[C(W)] shared by another world-script C(W*), as with possible worlds at most one of C(W) and C(W*) will ensue, depending on the world-type. Yet it is possible that two world-scripts sharing the same setting both be feasible, each with a different mapping of essences to lives. For no matter how many personal essences there may be who would freely choose in one way, for there to be another who would choose differently implies no change in any of the others' counterfactuals of freedom. While a given world-type may imply that W* is infeasible, yet there may be a feasible world with the same world-script as W*, since the same world-script may be actualized with a different mapping of essences to lives. Hence it is possible that every world-script be feasible.

20. The refutation of L hinges on the fact that when a person P is free to do A or ~A, and the fact is P would do A, it's impossible for God to cause P to freely do ~A. Yet if God wants someone to freely do ~A, and P would do A, then God can simply instantiate someone else who would freely do ~A. If God wants to actualize a world-script in which someone freely does ~A, the fact that P would do A is no impediment. Thus, while there is no world-type that would render every possible world feasible, there are world-types that would render every world-script feasible. A galaxy having, for each world-script, at least one feasible world including that world-script, let us call a pan-world-scriptable galaxy. It may be that only in a pan-world-scriptable galaxy can we call God omnipotent in the classical sense. This is an interesting conjecture. In what follows, though, I argue for a weaker claim. Having explained world-scripts, I can present a counterargument to possibly-R.


21. There are any number of possible worlds better than the actual world, even morally flawless worlds. Plantinga has asserted that possibly none of these better worlds were feasible. In a galaxy where no better world than the actual world is feasible, I suggest God's will would be frustrated, in the sense that there would be infeasible worlds he would rather actualize than any feasible world. For, had a different world-type been actual, God could have actualized a world containing a much better mixture of good and evil than the actual world contains.

22. I take it to be one feature or implication of omnipotence, classically understood, that God's will cannot be frustrated. For what contingent thing can oppose him? Of course, since God knows which worlds are feasible prior to creation, He would not will the actualization of an infeasible world, and so in this sense it seems His will would not be frustrated. Yet if God's will includes the specification of what He would do under various alternative circumstances, as on Flint's account it does, then His will would include aspects like, "If I were in a galaxy such that the feasible worlds included both worlds displaying a similar mixture of good and evil to the actual world and worlds displaying a better mixture (and a comparable amount of non-moral good), then I would weakly actualize one of the latter."6

23. In this sense we would see frustration as being implicit in God's will, if no better worlds than the actual were feasible. It seems far-fetched to imagine God's will as being frustrated, but I suggest that this is not a reason to suppose God's will would not be frustrated in galaxies where his creative options are morally bleak. Rather, to preserve classical notions of God we should find an account on which in no galaxy are the feasible worlds all morally bleak so that in every galaxy He is omnipotent and in no galaxy frustrated.

24. A supporter of Plantinga's account might suggest that the very notion of frustration I describe is problematic: that it would be impossible for God's will not to be implicitly frustrated, since necessarily some possible worlds are infeasible. In defense, I argue that at least in a pan-world-scriptable galaxy, God's will would not be frustrated. For, within a pan-world-scriptable galaxy there is, for any possible world outside that galaxy, a world inside the galaxy that differs only by who does what, in the sense of which personal essences perform the actions which occur in that world. All the same things would go on, the same joys and sufferings; all the same stories would be told in the two worlds;moral and aesthetic qualities would be the same. The only reason for God to prefer the one world to the other would be if he wanted this set of persons to go through those experiences and perform those actions rather than that set. This sort of preference, I claim, is one that God cannot have. A precondition of His perfect justice, and therefore of his perfect goodness, is that "God is no respecter of persons" (Acts 10:34)7. To some, the essential identities of possible persons may seem to remain an important ground for prefering one world to another; but to me and to the God of the New Testament, "identity" stripped of any other determinate property is not the sort of thing to support preferences.

25. For God's will to be frustrated is impossible, and I have argued that in a pan-world-scriptable galaxy, God's will would not be frustrated. Thus it is attractive to suppose that every galaxy is pan-world-scriptable. It may be that

O: God is omnipotent only in a pan-world-scriptable galaxy.
O implies that R is incompatible with God's being omnipotent, but given my time limit, I will not argue for O today. For God to be omnipotent may be a stronger condition than for God's will not to be frustrated. Rather than O, I claim that
F: If some infeasible world displayed a better mixture of good and evil than any feasible world, then God's will would be implicitly frustrated.
But R is a case of the antecedent in F, for R states that all the possible worlds better than the actual world are infeasible. Since for God's will to be frustrated is impossible, I conclude that Plantinga's R is impossible, and that the antecedent of F is necessarily false.

26. I have argued that Plantinga fails to support the crucial claim of his Free Will Defense, that possibly-R. I have further argued that not-possibly-R, and that in no galaxy is some infeasible world better than any feasible world, if God is omnipotent in the classical sense. I conclude that Plantinga's Free Will Defense fails and must fail, that the degree of contingent limitation to God's creative options required to solve the Problem of Evil as Plantinga frames it is not consistent with the classical notion of God he means to defend.

27. I suggest Plantinga is right to posit contingent limits to God's creative power, and right that such limits must be rather substantial to account for the degree of evil in the world. The failure of his Free Will Defense does not make just any LDS view seem more plausible in contrast, since some LDS apparently think of God as omnipotent in roughly the classical sense. After Plantinga I think this view is no longer plausible. However, certain LDS accounts do seem more attractive, such as those that see evil as inevitable given the imperfection of uncreate agents whom God can perfect only over time and with their free cooperation. Perhaps distinctively LDS cosmology, such as the belief that these agents can neither be created nor destroyed (cf. D&C 93:29) can support a distinctively LDS notion of omnipotence, of possessing all power, according to which God can be omnipotent and perfectly good and yet there still be evil.


1 James Tomberlin and Peter van Inwagen, eds.,Alvin Plantinga (D. Reidel, 1985) p227; also Alvin Plantinga, The Nature of Necessity (Clarendon Press, 1974).

2 The form of the refutation of Leibniz's Lapse is somewhat different in The Nature of Necessity than in Alvin Plantinga. The fine points do not concern me here, since I do not criticize this argument in itself. In either text, the conclusion is the same.

3 p185

4 In The Nature of Necessity, what I here call R* bears the name R which in this article I use to refer to the proposition Plantinga uses in his Self-Profile in Alvin Plantinga.

5 The Nature of Necessity, p191.

6 Thomas Flint, Divine Providence: The Molinist Account (forthcoming) pp88-90.

7 cf. Romans 2:11, Ephesians 6:9, Colossians 3:25. Whether being no respecter of persons could be argued to be a necessary condition for anyone whatever to be perfectly just, I leave to others to discuss.

Back to Issue 1:1