PLANTINGA ON GOD, FREEDOM, AND EVIL

FREDERICK W. KROON

University of Auckland

Introduction Hume once stated the Problem of Evil as follows: Is (God) willing to prevent evil, but not able? then is he impotent. Is he able, but not willing? then he is malevolent. Is he both able and willing? whence then is evil? (Dialogues Concerning Natural Religion, Pt. X)

Hume's intention, evidently, was not to charge the theist with maintaining belief in a God whose ways are fundamentally mysterious, but to drive home the charge that God's existence appears logically incompatible with the manifest existence of evil; that the statements: (1) There is an all powerful, all knowing, wholly perfect being, God and (2) Evil exists appear to be inconsistent with each other. The so-called Argument from Evil, of which Hume's series of questions is one version, is an argument that purports to show that (1) and (2) are true logical contraries, and that, since (2) clearly is true (and is acknowledged by theists to be true), (1) therefore cannot be true. An attempt to defend the theist against the Argument from Evil can take one of two forms. It may simply rebut particular versions of the argument by showing that it has not been demonstrated that there is an inconsistency. Or it may, more courageously, show that it cannot be shown that there is an inconsistency since (1) and (2) are demonstrably consistent. Not a few philosophers adopt the first approach. Alvin Piantinga is one philosopher who over the years has attempted to formulate a definitive version of the second approach. His most comprehensible formulation of Int J Phil Rel 12:75-96 (1981) 0020-7047/81/0122-007.$ $03.30. ~)1981 Martinus N i j h o f f Publishers, The Hague. Printed in the Netherlands.

76 this approach is found in his recent God, Freedom, and Evil I (hereafter called 'GFE'), but there are other formulations in God and Other Minds 2 and in The Nature o f Necessity. 3 These formulations are not identical, but neither are they essentially different in underlying strategy. In both GFE and The Nature o f Necessity, there is a much more explicit emphasis on the role of counterfactuals and greater stress is placed on the fact (if it is a fact) that the logic of counterfactuals requires that there are possible worlds that God cannot actualize, thus indicating a clear contrast between a position like Plantinga's and a Leibnizian theodicy. But while Plantinga's argument does not seem to have undergone essential changes since it first appeared, I do not think that GFE presents the most subtle version of the argument. In fact, the argument in GFE raises a number of serious questions, some of which are peculiar to the GFE formulation of the argument. This, at least, is what I claim in the present paper. The structure of this paper is as follows. Section 1 is devoted to a summary of Plantinga's consistency proof in God, Freedom, and Evil. In section 2, I point to a rather counterintuitive (but, for Plantinga, extremely crucial) modal assumption that Plantinga makes - an assumption according to which 'it is logically possible for P to ~ ' has a different meaning from 'there is a logically possible world in which P ~ ' s ' - and I defend an account of the ascriptions of powers and abilities that makes sense of this assumption. Having shown that Plantinga's consistency proof clears this important hurdle, I argue in section 3 that Plantinga's proof in God, Freedom, and Evil nonetheless fails to demonstrate that perhaps it was not possible for God to actualize worlds without evil, although (as I argue in section 4) a superficially similar consistency proof in The Nature o f Necessity is more successful. Section 5 argues the more general thesis that, even so, all of Plantinga's attempts to show God's existence compatible with that of evil, including the account in The Nature o f Necessity, suffer from a certain lack of generality. It is argued in this section that there are a number of powerful reasons for being dissatisfied with consistency proofs that (like Plantinga's) categorically rule out the possibility that God could have brought about any possible state of affairs he wished. In the concluding section (6), I briefly consider the question of the feasibility of alternative approaches to the problem of evil that do not rule out this possibility.

1. Plantinga's Free Will in Defense in God, Freedom, and Evil In its broadest outlines, at least, the strategy that Plantinga adopts in dealing with the Problem of Evil is clear enough. He points out that in order to rebut the charge that God's existence is incompatible with the

77 existence of evil the theist is required only to prove (1) and (2) above consistent; he is required only to provide a consistency proof. A full theodicy, one that posits reasons a Supreme Being might have for permitting evil to exist, does much more, and, in particular, does much more than is needed to solve the Problem of Evil. Plantinga's consistency proof, which he calls the Free Will Defense, applies most directly to the case of moral evil. He believes that his strategy for dealing with the problem of moral evil can be modified to apply to other arguments from evil (for example, ones which record the existence of natural evil, or the existence of all the evil the world actually contains). Since he spends most of his time on the problem of moral evil, however, we shall do the same. Here, then, is Plantinga's solution to the logical problem of evil. Leibniz, so Plantinga thinks, was guilty of an important mistake when he made the claim that God could have actualized any possible world at all. Once Leibniz made this mistake, he was naturally led to assert that ours must be the best of all possible worlds. Since the claim that God could have actualized any possible world at all is false - Plantinga calls it 'Leibniz's Lapse' - we are free to reject the highly embarrassing conclusion that this highly imperfect world is nevertheless the best of all possible worlds. We are also free to entertain the possibility that perhaps God could not have actualized worlds containing only moral good and no evil and imperfection. This, in fact, is the central lemma on which Plantinga bases his consistency proof. He claims, that is, that it is at least possibly the case that (3) God could not have actualized a world containing moral good and no moral evil. Piantinga formulates the argument as follows in GFE. First, let a maximal world-segment S be a possible state of affairs such that, if S* is any possible state of affairs more inclusive than S, then S* is a (full) possible world. (Think of a maximal world-segment as a possible world with a single hole in it.) Now consider the following property of persons: Definition 1. P is transworld-depraved iff, for every possible world W in which he is free with respect to at least some morally significant actions' and in which he nevertheless always does what is morally required of him, there is an action A which is morally significant for him in W and which respect to which he is free in W, and there is a maximal world segment S which is included in W but which itself contains nothing about the outcome of P's free choice with respect to A (although it contains everything else, including the fact that P is free with respect to A), such that the following counterfactual holds:

78 (c) If S were the case, P would go wrong with respect to A. (Or, using Lewis 's symbolism: S D--- P goes wrong with respect to A.) This property is not an essential property of persons, Plantinga admits, and it is perhaps not even a property that actually holds of anyone. But it is a property that at least possibly holds of all persons, and hence it is at least possibly the case that (4) All persons suffer from transworld depravity. Call (4) the principle of Transworld Depravity. Now (4) is a possibly true statement that, once amended to read that all possible persons suffer from transworld depravity (including persons that God might have created), entails (3), which therefore is also possibly true. The argument goes as follows. Free creatures are necessary to moral value. Now suppose that all possible creatures are transworld depraved. [This is at least possible, by (4)]. Let W be a world in which allthe creatures in W always freely do what is morally right, and let P be any such creature. Since P is transworid depraved, there is an action A with respect to which P is free in W and which is morally significant for P in W, and there is a maximal world segment S included in W, but containing nothing about the outcome of P's free choice regarding A, such that the following counterfactual is true: (c) S []--" P goes wrong with respect to A. Now God can actualize W only by actualizing S and leaving P free with respect to A. But in that case [as counterfactual (c) tells us] P will go wrong with respect to A, and hence the resultant state of affairs will not be W after all. Hence God cannot actualize such morally perfect worlds W. If Plantinga is right, then the following three statements are mutually consistent: (1) God exists, (3) It was not in God's power to create a world containing moral good and no moral evil, and (5) God has created a world containing moral good. But (3) and (5) entail (6) there is moral evil. Thus (1) and (6) [which is a version of the unadorned (2): there is evil] are

79 mutually consistent, and there is no logical difficulty in supposing that moral evil exists and that there is an all-powerful, all-knowing, all-loving God. This is the Free Will Defense in a nut-shell, at least as it applies to sin or moral evil. Plantinga is able to modify the strategy to deal with evil of other kinds as well, however. The case of natural evil - earth-quakes, and so on - perhaps looks most resistant to this kind of treatment, for a natural response in the case of natural evil is to say that an all-powerful Being could necessarily have actualized a possible world without natural disasters. The prevention of natural evil does not involve the muzzling of free agents, unlike, it seems, the prevention and elimination of moral evil. Plantinga's response to this is as follows. It is at least possible, Plantinga claims, that natural evil too is due to free agents - if not the free agency of humans, then to the free agency of Satan and his gang. And if this is at least possible, Plantinga will be able to use an argument much like the previous one to deal with the case of natural evil as well.

2. P o w e r - Ascriptions and the Actual World

Has Plantinga proved his case in GFE? Well, consider the following response. On the standard view, (7) It is logically possible for P to do A means (8) There is a possible world in which P exists and does A, and (7') It is logically impossible for P to do A means (8') There is no possible world in which P exists, and does A. But now we have an objection to Plantinga. As Plantinga admits, there is a possible world W containing God whose free inhabitants never do what is morally wrong. And it is surely clear that in W God does actualize W. So it seems as if God could have actualized possible worlds like W after all, given the analysis of (7) and (7') in terms of (8) and (8'). This points to a certain weakness in Plantinga's consistency proof as it is

80 found in GFE. No reason is given in GFE for doubting the (obvious) analysis of (7) and (7') in terms of (8) and (8'), and if one retains the analysis certain of Plantinga's moves are seen to be invalid ones. In particular, counterfactual conditionals such as those incorporated into the principle of transworld depravity have a contingent, not a necessary status. Why should the truth of some such set of counterfactual conditionals in the actual world, call it (~), determine which possible world God is able to actualize? The most that we can determine from the truth in (~) of a counterfactual (c) S E]--- P (freely) goes wrong with respect to A (where S is a maximal world-segment) is that (9) I f God actualizes a world W in which S is the case and P freely goes right with respect to A, then (c) is not true in that world, and hence W is not (~). But from (9) it cannot be inferred that it was not in God's power to actualize such a world, but at most (if at all" see section 3 below) that it was not in God's power to actualize a world in which S is the case, in which P freely goes right with respect to A, and in which (c) is true. As a matter of fact, however, there is a defense available to Plantinga, one which requires him to reject the analysis of (7) in terms of (8). I shall attempt here to reconstruct this defense, hopefully in a more perspicuous way than has been done by Plantinga himself. Consider locutions like 'it is possible for me to play tennis tomorrow', 'it is in my power to get you a good teaching position', 'I am able to join you for lunch tomorrow, should you want me to join you'. For want of a better term, call these 'personalmodal ascriptions', or 'p-modal ascriptions'. P-modal ascriptions form an interesting class of modal ascriptions, not least because they seem to resist easy translation into the usual possible world parlance. To say that (10) It is possible for me to play tennis tomorrow, is not to say that there are logically possible worlds in which I play tennis tomorrow, or technologically possible worlds in which I play tennis tomorrow. It is, instead, to say that the exercise tomorrow of my ability at tennis is compossible with all the constraints the world actually imposes on the exercise of this ability tomorrow, constraints such as physical laws, limitations of time and space, previous arrangements, and so on. A certain amount of flexibility in the identification of these actual constraints is inevitable and of some significance. One is often able to challenge the truth of a statement such as (11) It is not possible for me to play tennis tomorrow

81 by arguing that certain of the 'constraints' presupposed by the speaker are actually self-imposed 'constraints', within the direct control of the speaker and hence not really constraints at all (cf. 'You don't really have to go to the library tomorrow'). The speaker may then, in turn, point out that these challenged 'constraints' are truly constraints, given the overall system of desires of the speaker ('If I don't go to the library tomorrow, I shall fail my exams'). This flexibility is of some importance too for our understanding of the differences in semantic behaviour among p-modal operators such as 'is able to ...', 'it is possible for P to ...' and 'it is in P's power to ...'. A complete analysis is not the purpose of this paper, but it seems reasonable to claim that the operators 'P is able to ...', 'it is possible for P to ...', and 'it is in P's power to ...' can be graded according to the degree of constraint presupposed. P-modal ascriptions involving the operator 'P is able to ...' generally seem to presuppose more of the constraints the world imposes upon P's actions than do p-modal ascriptions involving the operator 'it is possible for P to ...' or 'it is in P's power to ...' Previous plans, for example, are included in the class of constraints that form the backdrop against which P is able to do certain things ('I am not able to join you for lunch tomorrow because I have made other arrangements'), but such constraints do not feature at all in p-modal ascriptions involving the operator 'it is possible for P to ...'. (While I may not be able to join you for lunch tomorrow, it is certainly in m y p o w e r to join you for lunch tomorrow if the only conditions that constrain me are certain arrangements that can be set aside.) Call this view the 'actual constraints' analysis of p-modal ascriptions. It is easy to see how one might accept this analysis, at least in broad outline, and yet refuse to apply it to God. God, after all, is not constrained by aspects of his creation, unlike his creatures. But saying that God's power is not limited by things of the kind that limit our power (physical laws, and so on) seems quite consistent with applying the 'actual constraints' analysis of 'it is (was) in P's power to do A' to a being like God. To see this, let us generalize the account a little. The actual world being the way it is, certain things are within my direct control while others are not. The former, I suppose, include not just intendings and other 'acts of will', but also actions such as the lifting up of my left arm, and my running a 5-minute mile. In a way, however, it doesn't matter too much how narrowly we construe our direct powers. Any serious account of what agents have it in their power to do will have to acknowledge the place of indirect powers. I can't directly bring about your voting in a certain way in the next General Elections - you are, after all, a free agent, not a zombie - but, if you are easily persuaded, there are things I can do or say which, if done or said, would result in your voting that way in the next elections. If you are easily persuaded, then I can get you to vote a certain way in the next elections, so

82 that if you are easily persuaded it is in my power after all to get you to vote in a certain way: in my indirect power, not my direct power. This suggests a two-level account of what is in P's power, with both steps involving an appeal to 'actual constraints'. We have already seen how a person's direct powers are generally a function of constraints imposed by the actual world. The second level, which requires the application of a certain kind of counterfactual ('if P were to do A, then B would result'), similarly regards the actual world as a system which determines what can, or cannot, result contingently from the performance of certain actions which are within a person's direct control. The existence of this second level makes disputes about what agents can directly undertake (intendings, basic actions, etc.) seem less important, for it seems intuitively likely that there will be basic agreement at least about what agents have it in their power, direct or indirect, to bring about. An 'actual constraints' analysis can now be applied to God in much the same way as we have applied it to humans. According to Plantinga, there are certain states of affairs even God cannot directly bring about, so that his power to bring these about, if he indeed has that power, can only be an indirect power, one whose scope is determined to a certain extent by the way the world actually is. Just as I have no direct control over whether the universe will end with a bang or a whimper, unlike God, so God has no direct control over whether or not I freely keep a certain promise, unlike me. And if there is nothing God can do that would result in my freely keeping a promise I should keep, then he doesn't even have indirect control over whether or not I freely keep this promise. In that case, it is simply not in God's power to actualize a world in which I freely keep all my promises. The only relevant difference between God and his creatures is that, being all powerful, God is constrained by few of the contingencies that constrain his creatures. According to the 'actual constraints' analysis, then we have something like this: (AC) It is in P's power (direct or indirect) to bring about state of affairs A iff there is a state of affairs T which it is in P's direct power to bring about (given the various constraints the actual world inescapably imposes on P) such that: P brings about T I-q--- A (i.e., such that, given the actual world as backdrop, P's bringing about T would lead to A's being the case). I still have some qualms about this formulation. In particular, the counterfactual 'P does T Fq--- A' was chosen to symbolize 'were P to do T, A would result', but the latter locution looks a causal locution, unlike the bare 'P does T [~--" A'. (The difference can be vividly brought out by

83 noting that, on the standard semantic analysis, 'P does T [ ] ~ A' is true whenever P does T and A is the case.) Maybe, therefore, account (AC) should be amended to yield a tighter connection between possible actions on P's part and those possible states of affairs whose actualization results from P's actions. ~ In any case, the 'only if' part of (AC) will be part of any truly comprehensive 'actual constraints' thesis about what agents can bring about; and, as we shall see later, it is the only part of (AC) that can be considered to play a necessary role in Plantinga's consistency proof. What about the locution 'it is logically possible for P to do X?' It is usually thought that if it is logically possible for God to do X then it is also in God's power to do X; which perhaps casts doubt on account (AC) as applied to God. But we can give a similar 'actual constraints' analysis of 'it is logically possible for P to do X'. Sentences of this latter form can, in fact, be understood in at least two different ways, only one of which makes it a p-modal ascription. On the standard reading, the following is true: (S) it is logically possible for P to do A if there is some logically possible world in which P does A This standard reading is certainly appropriate for locutions like: 'it is logically possible for people to grow to a height of 7 feet', for the latter is not a true p-modal ascription, not a true ascription of a kind of power. (It is more like the clearly non-p-modal ascription 'it is logically possible for cities to reach a population of over 20 million inhabitants.) In some cases, however, 'it is logically possible for P to do A' is plausibly read as a p-modal ascription, in which case the following 'queer' analysis, an 'actual constraints' analysis, captures its latter meaning: (Q) it is logically possible for P to bring about A iff there is some action X which is in P's direct control, assuming no non-logical limitations on what is in P's direct control, such that P does X [ ~ A is the case. The two analyses differ. Suppose that God exists in the actual world, and suppose that in the actual world there is nothing God's creatures can possibly do that will let their breaking his moral laws remain unpunished. (I am assuming that God's creatures are necessarily under God's control.) Then (i) it is not logically possible (in the actual world) for a person P to break God's moral code with impunity, even though (ii) there are possible worlds in which P breaks God's moral code with impunity (worlds in which God is more merciful, for example). In this example, (i) is based on the 'actual constraints' analysis (Q), while (ii) is based on the standard analysis (L). The example quite clearly brings out the curiously indexical nature of

84 account (Q). On the 'actual constraints' analysis of 'it is logically possible for P to bring about A', it may be logically possible in world W for P to bring about A even when it is not logically possible in world W ' for P to bring about A (since the constraints on P in W and W ' are different). Using analysis (Q) rather than (L), we are now in a position to agree that if it is logically possible for God to bring about A then it is also in God's power to bring about A. We can agree with this on the grounds that whatever is possibly in God's control is actually in his control, a principle that can be presumed to hold for God, but not for his creatures.

3. Some Problems for Plantinga's Free Will Defense in God, Freedom, and Evil This completes my defense of Plantinga's position. I have tried to show that there is a natural way of understanding locutions like 'it was in P's power to do A' that leads to an 'actual constraints' analysis of locutions like 'it was in God's power to actualize state of affairs S' or 'God could have actualized S'. Such an 'actual contraints' analysis is needed if Plantinga's Free Will Defense is to succeed, for on the standard analysis of a locution like 'God could have actualized S' Plantinga's Free Will Defense seems to fail. This is a point we made earlier, but we are now in a better position to see why this is so. On the standard analysis, the statement: (12) God could have actualized S gets translated as: (13) There is a possible world W in which God actualizes S. The statement (14) God actualizes S in W is, in turn, best translated as: (15) There is a state-of-affairs T which God directly brings about in W, and in W S results from God's doing T. That is, God actualizes S in W if relative to the constraints on God's power in W, God is not only competent to actualize S but, in fact, does actualize S. ('Actualize' here means 'weakly actualize' or 'indirectly bring about'.) On this analysis, however, Plantinga's Free Will Defense fails, because it is

85 necessarily true that, relative to the constraints on God's power in morally perfect worlds, God is entirely competent to actualize such sinless worlds. Although GFE is silent on these points, some of Plantinga's other publications show that he explicitly accepts an 'actual constraints' analysis. The Nature o f Necessity, for example, has the same analysis, merely substituting the phrase 'God strongly (alt: weakly) actualizes W " for 'God directly (alt: indirectly) brings about W " in (AC). So it is not unfair to ask how well this analysis supports Plantinga's claim that it is at least possibly true that God could not have actualized possible worlds containing moral good but no moral evil. Unfortunately, Plantinga's Principle of Transworld Depravity as it is stated in GFE doesn't help us very much here. Consider the claim that a person P is transworld-depraved in the sense specified in GFE. This means, recall, that for any world W in which P is significantly free and always does what is right there is an action A with respect to which P is significantly free in W, and there is a maximal world-segment S included in W but not itself including the result of P's choice with respect to A, such that (c) S is the case [3-* P goes wrong with respect to A. If S is truly a maximal world segment, however, then S contains states of affairs such as P's always doing the morally right thing, apart, possibly, from A; P's being sincerely disposed to do the morally right thing, intending never to deviate from this disposition; and so on. But this being the case it is difficult to see how a counterfactual like (c) could be true in any world W except for the world W* in which S is actual and in which P nevertheless goes wrong with respect to A. The reason for this should be obvious. A counterfactual such as (Plantinga's example): 'Had Curley been offered a bribe of $ 20,000 he would have freely accepted the offer' is true in a world V if, of those worlds which are identical to V up to, and including, the moment of Curley's being offered the bribe, some world V' ' in which Curley freely accepts the bribe is more like V, given Curley's history in V and especially given Curley's character in V, than any world V' in which Curley refuses the bribe. Since the dispositional and historical facts built into maximal world-segments of worlds in which P always does what is right contain no facts that reflect any weakness in P's character, it is difficult to make sense of the claim that a counterfactual like (c) is true in any world V in which S is not actual. Now it is true that, since it is possible that both S is actual and P goes wrong with respect to A, it is also possible that (c) is true; and this, it may be claimed, is enough for Plantinga. But it isn't, of course. Plantinga's point is that it is at least possibly the case that each counterfactual of type (c) is true [not just that each counterfactual of type (c) is possibly the case], with 'S' ranging over suitable maximal worldsegments of worlds in which P always does what is right; and this is not possible if a counterfactual of type (c) is only true in worlds in which the segment S is actual (since distinct S's are incompatible).

86 But worse is to come. Because Plantinga's counterfactuals mention maximal world-segments, his argument in GFE seems to be guilty of the counterfactual fallacy of transitivity. Plantinga thinks that it follows from the truth of (c) S m ~ P goes wrong with respect to A that (d) God brings it about that S U]--, P goes wrong with respect to A. But (d) does not follow from (c). S, being a maximal world segment, may contain states of affairs that God cannot directly bring about (states of affairs such as Curley's freely refraining from accepting a certain bribe), and hence, if God is able to actualize S at all under these circumstances, he can only do it indirectly. By the 'actual constraints' analysis (AC), it is in God's power to bring about S indirectly if there is some state of affairs T which God can directly bring about such that (e) God does T ~

S.

To derive (d) from (c) and (e) is impossible, however, without committing the fallacy of transitivity,' the fallacy of using the invalid inferential scheme: p [3--} q, q [~--} r, therefore p []--} r. At this point, Plantinga may respond that we have lost sight of the nature of the S's that feature in the relevant counterfactuals (c) and (e). They are supposed to be maximal world-segments, possible worlds with single holes in them, and hence a maximal world-segment S that leaves the result of one of Curley's free choices undetermined contains either (i) God's doing T or (ii) the complement of this state of affairs. In either case, (c) and (e) logically entail (d). The non-trivial case is that where S contains God's doing T, and in this case the inference is essentially of the form p [~

q,

[] (q D P), q I~-~ r,

.'. p E]-" r,

87 which is a valid scheme in the logic of counterfactuals. This response is, of course, based on a completely unLeibnizian view of possible worlds, since for Leibniz God is not strictly an inhabitant of, and toiler in, any possible world. What God does transcends each possible world and is included in none. Is Plantinga on safe grounds, then, if he bases his argument on his own unLeibnizian understanding of possible worlds? I doubt it. The problem now is that the concept of a maximal world-segment does not seem well-defined. Let W be a world in which P always does what is right, and suppose that, in W, God directly brings about T. Then the counterfactual (f) God does T [3--- P freely goes right with respect to A is true in world W. It is clear that no maximal world-segment of W in which P's decision regarding A is left undetermined can include both God's doing T and (f)'s being the case, but it is far from clear where this leaves us. Why can't we have a maximal world-segment S of W that leaves P's decision regarding A undetermined, includes (f)'s being the case, and omits God's doing T? If this is a maximal world-segment, however, then Plantinga is still in trouble, for the argument from (e) and (c) to (d) is again guilty of the fallacy of transitivity under these circumstances.

4. A Reformulation of Plantinga's Free Will Defense These troubles arose because Plantinga worked with counterfactuals whose antecedents mention maximal world-segments and hence, in some cases, the existence of moral biases. Once we accept an 'actual constraints' analysis of what is in God's power we have a much more direct way of doing what Plantinga wants. Instead of using counterfactuals whose antecedents mention maximal world-segments, we should use counterfactuals whose antecedents make n o mention of biases towards the good or the results of other free choices, but only of things which are within God's direct control, counterfactuals like: (c') Even if God had brought it about that P was reared in a socially and materially enriched environment, P would still have told lies, broken promises, and so on. On the surface, such counterfactuals are credible, unlike the corresponding counterfactuals in GFE, because they might be true in virtue of P's sinful dispositions or the depravity of his character, factors over which God has no direct control. And it seems likely that, by making such counterfactuals

88 suitably general, we might finally succeed in showing that it is at least possibly the case that God could not have actualized morally perfect worlds. This is, in fact, precisely what Plantinga does in other statements of his Free Will Defense. In his 'Which Worlds Could God Not Have Created? '8 whose publication actually predates the publication of GFE, Plantinga offers this account of transworld depravity (as applied to persons, not essences): Definition 2: A person P suffers from transworid depravity if for every world W in which P is significantly free and always does what is right, there is a state of affairs T and an action A such that (1) God strongly actualizes (i.e., directly brings about) T in W, and T includes every state of affairs God strongly actualises in W; (2) A is morally significant for P in W; and (c*) If God had strongly actualized T, P would have gone wrong with respect to A. So P suffers from transworld depravity if no direct action on God's part (such as creating P in a socially and materially enriched environment, and so on) would have resulted in P's leading a free, but blameless, existence. The crucial mediating principle Plantinga then employs is, once again, the principle that all possible persons suffer from transworld depravity (as thus understood). The Nature o f Necessity contains the same principle. This principle is then used, as is the corresponding principle in GFE, to prove that it is at least possible that God could not have actualized a world containing moral good but no moral evil. Note, however, that this solution to the problem of evil side-steps the problem about counterfactuals generated by GFE, and does it in the way we suggested. The counterfactuals appealed to in the new principle of transworld depravity are reduced versions of the counterfactuals appealed to in the old version, in so far as their antecedents do not mention full maximal world-segments of worlds W in which a person P never does wrong, but only that part of any such world W which consists of God's strongly actualizing, or directly bringing about, the state of affairs he strongly actualizes in W. These counterfactuals therefore do not have antecedents that include the mention of moral biases, but are simply general versions of a credible counterfactual such as (c') above. And it is not hard to see that, as a result, Plantinga's reformulated proof that possibly God could not have actualized a morally perfect world does not commit the counterfactual fallacy of transitivity (unlike Plantinga's GFE proof)3

5. Some Residual Problems

Should we agree, then, that Plantinga has effectively answered those who, like Hume, see the existence of evil as a logical problem for the theist? I do

89 not think so. My own comments, it should be noted, amount only to a defense of the 'actual constraints' account of powers and abilities that Plantinga needs to use in his consistency proof, and to the observation that Plantinga's new argument in The Nature o f Necessity is formally superior to the old God, Freedom, andEvii argument. But I have had nothing to say about Plantinga's incompatibilism on which the very crucial claim that all possible persons are at least possibly transworld depraved rests, and it is here, perhaps, that Plantinga's argument is most susceptible to attack. Rather than argue for, or against, compatibilism directly, I shall produce two indirect reasons for being less than impressed by the presupposition of incompatibilism in Plantinga's various versions of his Free Will Defense. These two arguments constitute what we might call the strong challenge to Plantinga's crucial lemma that possibly God could not have actualized a normally perfect world; or a world containing less evil than the actual world; and so on. In the sequel, we also produce a weak challenge to Plantinga's lemma. We argue that even if Plantinga is right to embrace incompatibilism there are versions of the problem of evil that, not unreasonably, build into their premise about evil the claim that God could have actualized worlds with a better balance of good over evil, and such versions of the problem appear to resist Plantinga's incompatibilist methodology. My first argument is as follows: Plantinga's 'best' formulation of the principle of transworld depravity claims that transworld depravity is a possible property of all instantiations of creaturely essences, of all possible persons, and not just of all actual persons. But is it really so certain that there can be contingent dispositional truths about the free, uncaused, behaviour of any possible person? Does it really make logical sense to suppose that God, in actualizing the world, is confronted by actual dispositions to (some) evil on the part of each as-yet-uncreated, and perhaps never-to-be-created, possible person? '~ Isn't it really better (logically better) to replace Plantinga's reformulated counterfactuals by slightly weaker counterfactuals that do not make this large modal assumption: I mean counterfactuais like: (c**) God does T ~ --" P goes wrong with respect to A (or: had God done T, P might have gone wrong with respect to A). It can be shown that adopting (c**) instead of (c*) makes no formal difference: it will still be possible to deduce that (possibly) God could not have actualized a morally perfect world, even though (c**) is weaker than (c*)." I suspect that Plantinga's reason for preferring strong counterfactuals like (c*) to weak counterfactuals like (c**) is that an account of freedom according to which possibly both:

90 (i) God does T ~ respect to A

~ P freely goes right with

and (ii) God does T ~ --" P freely goes wrong with respect to A, is inconsistent with yet another feature of God's nature which is traditionally stressed by theists, viz. his foreknowledge. Plantinga, construing omniscience as universal prescience, seems to suppose that God is able to plan his universe in determinate fashion because God necessarily knows, given a state of affairs T, what any possible free agent would do were God to bring about T. The problem with this justification is familiar enough, omniscience thus construed is problematic because it is difficult to reconcile with the theory of knowledge. If foreknowledge is truly knowledge, then God must at least have good reasons for the beliefs - beliefs like (c*) - on the basis of which he decides which world to actualize. His reasons must be sufficient to justify the claim to knowledge. It is difficult to see, however, what reasons God could possibly have for such beliefs. Once we accept incompatibilism, we cannot answer that God's reasons are based on the fact that whatever comes to pass is predetermined by God to come to pass. This reply is available to Leibniz, but not to an incompatibilist like Plantinga. Plantinga is still in a quandary, therefore. The only argument for the possible truth of counterfactuals like (c*) that I am able to think of rests on a philosophically dubious account of divine omniscience. But once we adopt weaker counterfactuals like (c**), we are spiked on the other horn of the dilemma: omniscience conceived as foreknowledge is commonly regarded as one of God's essential attributes, and a consistency proof that refuses to respect this essential attribute consequently does less than we might reasonably have expected. There is a second indirect argument against a consistency proof that presupposes incompatibilism. Consider, again, the structure of my earlier vindication of Plantinga's account of powers. The structure of my argument was as follows. After arguing that an 'actual constraints' analysis works for locutions like 'it is in my power to do A' and 'I could have done A', I followed Plantinga in extending this analysis straightforwardly to the case of God. Since God cannot directly bring about a person's freely undertaking an action A (on Plantinga's incompatibilist understanding of freedom), there are, even in God's case, certain limitations on what God is capable of achieving directly. What God can indirectly bring about is, in turn, a function of crucial dispositional facts about persons that thus form

91 the actual constraints limiting God's indirect power. The resultant account does admit that there is an important difference between the case of God and the case of humans: God's direct powers are so much greater than ours, after all. But the account does not admit any 'in principle' difference between the case of God and the case of humans. God, like his creatures, acts against the background of the actual world. More of the actual world is directly attributable to him than to us, of course, but nonetheless the difference looks, in an important sense, quantitative only. This startingly anthropomorphic conclusion is the price we pay for idealizing the 'actual constraints' analysis for ordinary p-modal ascriptions to yield an 'actual constraints' analysis of statements about God's power. We can see how suspicious we should be of Plantinga's claim that he has conclusively refuted Leibniz's Lapse. Leibniz certainly would not have been able to make sense of God's actualizing the actual world against the background of the actual world itself; or of God's power being to some degree a function of the way the world actually is. Leibniz like Plantinga, agrees that God is in no way responsible for what humans do, but for Leibniz this means something entirely different from what it means for Plantinga. It means that before God ever actualized anything humans and other substances subsisted sub rationepossibilitatis, i.e., as conceptualized possibilities whose existence as possibilia God could not have prevented. This clearly does not mean that certain contingent counterfactuals about humans perhaps constitute the actual constraints against which God actualizes a possible world. For Leibniz, there just cannot be states of affairs that constitute actual constraints against which God actualizes a possible world, for the possibility of any state of affairs can only be the result of God's free decision to actualize some substance that, until actualized, only subsists sub ratione possibilitatis. ,2 Why, then, does Plantinga continue to think that he has conclusively refuted Leibniz's Lapse? It is, no doubt, because he thinks that it is at least logically possible that humans are free in a very strong, incompatibilist sense that rules out causal pre-determination, so that Leibniz too has to face the fact that there are certain possible states of affairs God cannot strongly actualize, cannot directly bring about. But Leibniz would certainly not grant Plantinga the starting premise that humans are at least possibly free in the strong sense Plantinga favours. To leave a person free in that sense would be, for Leibniz, to bring about the existence of a being whose individual concept is radically incomplete, and this is impossible on Leibniz's understanding of possible substances and possible worlds. Persons are, for Leibniz, rather like characters from a novel. They are certainly free, but not in a transcendent sense of freedom that excludes all causal pre-determination of their actions. Characters from novels certainly do not lead lives that are, in such a transcendent sense, independent of their

92 creator's purposes, yet their actions are nonetheless unproblematically free. It is no doubt tempting to response to this counterchange by decrying Leibniz's metaphysics of possibility. I think, however, that such a response would be short-sighted. There may well be alternative metaphysical grounds, rooted, perhaps, in one's concept of God, for the claim that God's relation to his cosmos is appropriately asymmetric, that we are in no sense partners with God in the actualizing of the actual world. Philosophical Calvinists, of course, maintain that God's relation to his creation is a strongly asymmetric relation of just this type, and Anthony Flew argues in his God and Philosophy 13that any theologically coherent account of God's relation to his creation should entail this kind of strong asymmetry. It is tempting, in this context, to adapt something which Kant said about the teleological argument: just as the kind of being for whose existence the teleological argument gave good reasons was far removed from the ens realissimum whose existence should have been, but (for Kant) could never be, the subject of the rational proofs for God's existence, so for many philosophical theologians, and certainly for Leibniz, the kind of being whose existence Plantinga 'proves' compatible with the existence of evil is far removed from the maximally great being who should have formed the topic of Plantinga's consistency proof. So far we have argued that one important weakness of Plantinga's approach lies in its uncompromising incompatibilism. The feature of Plantinga's approach amounts to a logical weakness in the approach. If our first argument is valid, we have reason to suppose that some of the steps in Plantinga's proof require the truth of compatibilism, and this requirement obviously invalidates the proof. If our second argument is valid, furthermore, Plantinga hasn't yet ruled out the possibility of a Leibnizian God, and hence (i) he hasn't succeeded in producing a proof that is general enough to defuse each problem of evil that needs defusing, and (ii) he also hasn't succeeded in demonstrating that a being who is maximally great but who nonetheless couldn't have actualized some worlds is even possible. (It is clear, I think, that the possibility of Leibniz's God entails the impossibility of this kind of being, and hence the impossibility of Plantinga's God.) These arguments suggest the wisdom of attempts to devise consistency proofs that do not rest on incompatibilism. But there is another, quite different, reason for searching for such alternative consistency proofs. Suppose that, in the face of the arguments above, we grant Plantinga his incompatibilism. In that case we are still able to mount a weak challenge to Plantinga's approach. Plantinga's practice of using mediating principle (such as the principle of transworld depravity) whose possible truth rests on incompatibilism cannot, on the face of it, be extended to deal with all

93 versions of the logical problem o f evil whose premise about evil is a reasonable statement o f the nature and scope o f the evil in the world; and Plantinga's approach is, to that extent, limited in application. Plantinga argues (in GFE and elsewhere) that his strategy can be adapted to deal with versions o f the logical problem o f evil that specify in some detail what kind o f evil there is (e.g., natural evil as well as human sin), and how much there is. But consider the following statement about evil: ( 2 " ) There is natural and moral evil which an all powerful being could have prevented, and whose prevention would have resulted in a better world. This statement seems as reasonable and uncontroversial as any of the other claims about evil Plantinga bravely attempts to prove compatible with God's existence, yet Plantinga's Free Will Defense clearly cannot be adapted to cope with it. A comparison with a recent criticism o f Plantinga's approach on the part o f McClosky may be helpful here. McClosky writes in God and Evil that It is not necessary for the critic to explore whether evil spirits, mischievous fairies, clumsy Homeric gods, or the like, can be shown possiblyto be the cause of certain kinds of evil. This is to rest the theist's case on a purely sceptical, unreal doubt. I' If this claim is intended to demonstrate the invalidity of Plantinga's original consistency proof, it clearly misses the mark. But if, instead, it is intended to make the point that any serious attempt to prove God's existence compatible with the existence o f evil should be able to cope with premises about evil that are paradigmatically reasonable - premises such as 'there is evil which has only natural causes' and 'there is much preventible evil' [cf ( 2 " ) 1 - it seems sound enough. On the face o f it, ( 2 " ) is paradigmatically reasonable in just this sense. It is surely no less reasonable than a premise that describes the variety o f evil in the actual world, and an approach to the problem of evil whose competency does not extend to ( 2 " ) can to that extent be accused o f resting on an unreasonable restriction o f God's powers.

6. Conclusion In this paper, I have tried to defend an 'actual constraints' account of powers, and I have suggested that the weakness o f Plantinga's Free Will Defense lies not in its understanding of powers but in the way it circumscribes these powers. I have defended both a strong and a weak version of

94 the claim that the circumscription of God's powers in Plantinga's approach is unjustifiably severe. Let us now briefly consider the following question: Are there ways of dealing with the problem of evil that avoid these criticisms? That is, are there ways of dealing with the problem of evil which do not presuppose the falsity of compatibilitism and which cope successfully with a premise about evil like (2' ')? The following approach seems to offer some slight hope. Consider the inference: (Pi) God is wholly good ". (P2) It is necessarily true that if God could have actualized a world better than the actual world, he would have. This inference is, I think, fallacious. Rather than rest a consistency proof on the denial of the antecedent in (P2) (Leibniz and Plantinga both do this), we would do well to consider the possibility of consistency proofs that simply deny (P2) itself. One way of breaking the logical link between (PI) and (P2) was assayed by Robert Merrihew Adams in his "Must God Create the Best?" is In his article, Adams argues that on the Christian conception of the moral ideal there is no reason for believing that a wholly good God would have actualized the best possible world (supposing that there is a best world). An alternative way of breaking the link is to accept a principle like: (P3) For each world W, there is a world W I better than W, (a principle that should be acceptable to those who espouse classical utilitarianism, 16 for example), for once (P3) is accepted it seems entirely unreasonable to blame an all-powerful being for actualizing some world W simply on the grounds that there are worlds better than W which the being could have actualized. Unfortunately, however, neither of these methods seems to lead directly to a dissolution of the problem of evil. Many philosophers, perhaps including Plantinga 17 himself, would instead claim that while it is arguable that an all-powerful being does not deserve moral blame for not actualizing the best possible world such a being certainly does deserve moral blame for actualizing a world containing preventible evil. I doubt that this intuition can be exploited by philosophies that accept a principle like (P3) and that also regard good and evil as commensurable principles (in the sense, roughly, that a quantity of evil can always be off-set by a sufficient quantity of good.) To this extent, I do not believe that there can be a problem of evil for one who holds (P3) in conjunction with a privative view of evil, for example; or that there can be a problem of evil for the classical utilitarian

95

(who similarly holds that the negative utility of suffering is off-set by the positive utility of a sufficient amount of good). But philosophies like classical utilitarianism aside, the intuition seems a powerful one, and presents a continuing obstacle to the articulation of hoped-for alternative approaches to the problem of evil.

NOTES

1. 2. 3. 4.

5. 6.

7. 8. 9.

10.

I1. 12. 13. 14. 15. 16.

God, Freedom, and Evil (George Allen and Unwin, 1974). God and Other Minds (Cornell University Press, 1967). The Nature o f Necessity (Oxford: Clarendon, 1974). Some definitions: (i) P is free with respect to an action A iff he is free to perform A and free to refrain from performing A. (ii) A is morally significant f o r P iff it would be morally wrong for P to perform A but morally right to refrain, or vice versa. (iii) P is significantly free (on occasion t) iff (at t) he is free with respect to some action which is morally significant for him. David Lewis, Counterfactuals (Harvard University Press, 1973). This objection against (AC) applies with less force against (AC) if the latter is restricted to the case o f God. It seems safe to assume that, if God exists, then any state of affairs A is actual precisely because God knew that A wouM be the case if he were to undertake a certain course of action T (i.e., God knew that: God does T ~i ~ A), and, in the light o f his preferences, therefore decided to do T. I suspect that this makes the connection between T and A strong enough to warrant saying that A's being actual is a genuine consequence of God's doing T. See Lewis, Counterfactuals, 32ff. Journal o f Philosophy 70 (1973):539-52. For Plantinga's own reformulated proof, see, e.g., The .Nature o f Necessity, 181ff. I have also benefiied from seeing a simpler (unpublished) proof by David Lewis. Lewis, however, disagrees with Plantinga's actual constraints' analysis of what God could, or could not, have done (written communication), and rejects Plantinga's conclusions on that basis. I do not doubt that there can be true counterfactuals about what would happen if all essences o f a certain kind were instantiated. We can, for example, attribute truth to a counterfactual assertion such as 'any possible solid sphere of steel larger than 1 cc in volume would sink if placed in water'. This is because a substance like steel has certain essential properties P / t la Kripke or Putnam, such that it is nomically necessary that a solid sphere larger than 1 cc in volume and composed of a substance that has properties P will sink in water. This kind of account, of course, is not open to Plantinga. On Lewis', but not Stalnaker's, semantics for counterfactuals. See Lewis, Counterfactuals, p. 80. See, e.g., Rescher, The Philosophy o f Leibniz (Prentice-Hall, 1967), pp. 13-16. God and Philosophy (London: Hutchinson, 1966), pp. 45ff. God and Evil (The Hague: Martinus Nijhoff, 1974), p. 112. Philosophical Review 81 (1972):317-32. By classical utilitarianism I mean the theory that advocates maximizing the total happiness (as opposed to the average happiness) of sentient beings. This theory has been advocated by J.J.C. Smart in Smart and Williams, Utilitarianism: For and Against (Cambridge University Press, 1973), sect. 4.

96 17.

Plantinga, in GFE and elsewhere, criticizes a number of different principles about omnibenevolence. One he does not criticize is the principle that a wholly good, all-knowing, and all-powerful being eliminates all evil it can properly eliminate (i.e., can eliminate without causing more evil or getting rid of an outweighing good). Through his Free Will Defense, Plantinga instead attempts to show that it is possible that there is much evil that God cannot properly eliminate.