Canadian Journal of Philosophy Does Theism Need a Theodicy? Author(s): Richard Swinburne Source: Canadian Journal of Philosophy, Vol. 18, No. 2 (Jun., 1988), pp. 287-311 Published by: Canadian Journal of Philosophy Stable URL: http://www.jstor.org/stable/40231610 . Accessed: 14/06/2014 10:38 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . Canadian Journal of Philosophy is collaborating with JSTOR to digitize, preserve and extend access to Canadian Journal of Philosophy. http://www.jstor.org This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions CANADIAN JOURNAL OF PHILOSOPHY Volume 18, Number 2, June 1988, pp. 287-312 287 DoesTheismNeeda Theodicy? RICHARDSWINBURNE Oriel College University of Oxford Oxford 0X1 4EW Great Britain I To many atheists the existence of evil seems to provide a conclusive argumentagainst the existence of God. God is by definition omnipotent and perfectly good; a perfectly good being will remove evil in so far as he can, an omnipotent being can remove any evil he chooses, so if there is a God there will be no evil, but there is evil, hence there is no God. Theists normally challenge this argument by challenging the premiss that a perfectly good being will remove evil in so far as he can. The theistic defence is usually put as the defence that many evils are logically necessary conditions of greatergoods, and hence a perfectlygood being may allow them to occur in order to bring about the greatergood; so a perfectlygood being may well allow some evils to occur. The purpose of this paper is to investigate whether it is enough for the theist formallyto make this point, in orderto render the argument from evil harmless, or whether evil nevertheless 'counts against' the existence of God, in the absence of a demonstration with respect to each known evil what greater good it serves; whether the 'counts against' can be analysed as 'makes less probable';and to determine at what point evil makes belief that there is a God irrational.However, the defenceneeds to be phrasedmorecarefully.First,for some evils, even if that evil is a necessaryconditionof a greatergood, agents would have no rightever to bringthem aboutfor the sake of the greatergood. A beating may do some child much good, and that amount of suffering might even be logically necessary to effect that particulargood; but, unless I am the child's parent or teacher, I do not have the right This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 288 Richard Swinburne to beat the child for the sake of that good. The evils which the perfectly good being does not attemptto prevent must be ones which he has the right to allow to occur. Second, the logically necessary condition of the greatergood may not be the actual occurrenceof the evils, but some agent P having the power to cause them to occur without being caused to exercise that power or being caused to refrainfrom exercising that power (e.g., P having the power intentionallyto bring about evil, and not being subjectto causes which determine how he will exercise that power; which I shall call his having libertarianfree will). For such a reason a perfectly good being may not attempt to prevent another agent bringing about evil, even though the actualoccurrence of the evil is not a necessaryconditionof the greatergood. And, third, we must avoid phrasingthe defence so as to imply that a perfectlygood being will maximise the quantity of goodness in the world in so far as he can. For, plausibly, there is no maximumto the amount of goodness in the world - however many beings were enjoying the world, the world might be better for the existence of another such being in some distant place or other space. The perfect goodness of an agent is, I suggest, not a matter of his acting so as to maximise the number of good states of affairs, but acting so as to bring about many good states of affairs, so as always to fulfil his obligations, and so as to remove any evil which he can remove without removingan equallygreat (or greater)good. (Fulfillingone's obligations is a matter of performing all positive duties [such as keeping promises] and refrainingfrom performingwhat one does not have the right to do [such as stealing].) That at any rate seems the minimum requirementif an agent is to be perfectlygood; a stricterrequirementwould requirea strongerdefence from the theist. These points being made, the theist can rightly object to the premiss 'a perfectly good being will remove evil in so far as he can' by claiming that all that the atheist is entitled to in this context is 'a perfectly good being will remove evil in so far as he can, unless allowing that evil to occur is something which he has the right to do, and which is such that allowing it or an equallybad evil to occurmakes possible, and is the only morallypermissibleway in which he can make possible, the occurrenceof a state of affairsat least as good as the evil is bad, and he does bring about that occurrence.'I shall in future use the simple phrase which I used before, that a perfectlygood being may allow evils to occur for the sake of a greatergood, but understand by it the careful spelling-out which I have just given. Since all things are possible for an omnipotentbeing, except the logically impossible, the theist's defence in this context is then that God might well allow evil to occur, caused either by himself or by some other agent, in orderto bring about the occurrenceof a good for which allowing it to occur is logically necessary. He might well allow me to This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 289 suffer a hangover after drinkingwhisky, because if he does not allow me to suffer the hangover, it will not be possible for me to have had the libertarianfree will of choosing whether to drinkwhisky and have a hangover, or not to drink whisky and have no hangover, in such a way that my choice either way would be efficacious. Libertarianfree will efficaciousin affecting our own futures is such a good thing that God might allow us the consequences of our bad choices. That there are some evils which a perfectlygood being might allow to occur, even if he could prevent them, seems evident to me and, I suspect, to most atheists. The real issue is: are all the actualevils in the world such that a God could allow them to occur, compatiblywith his perfectgoodness. First, is each kind of evil such that he would have the right to allow it sometimes to occur, even if it serves some great good? Does he have the right to allow children to suffer pain for the sake of some good for others, or even for themselves? And, second, if God has such a right only under certaincircumstances,do those circumstancesobtain?For example,if he has the rightto makechildrensufferonly if there is some balance of pleasure over pain in their lives, is there such a balance? Third,is each evil such that to allow its occurrenceis one way of bringing about a logically necessary condition of a state of affairs at least as good as the evil is bad? Is allowing animal pain a way of bringing about a logically necessary condition of any state as good as the evil is bad?It makes it possible for humans to choose freely (i.e., with libertarianfree will) to show compassion to animals. Perhaps their having this kind of efficaciouschoice is a good as greatas the evil which makes it possible is bad. But, fourth, is allowing the occurrenceof the evil (or the occurrenceof an evil equallybad) the only morallypermissible way of bringing about that logically necessary condition of that good? Is there no less costly way to achievethe goal?God could make human compassionate actions possible by deceiving humans into supposing that animals were suffering when they were not. But, plausibly, such large-scaledeception about the feelings of other creatures,leading to quite unjustified sympathy and compassionate action, would not be morallypermissible. In that case the only morallypermissibleway in which God could bring about a logically necessary condition of compassionate action, viz. animals seeming to suffer, is by actually making them suffer, and that involves theiractualpain or pain equallybad. Finally, does the good state of affairs,which the evil makes possible, actuallyoccur?If there is a God, has he broughtabout, or will he bring about, that good? Suppose that the evil of animal suffering is to be justifiedby the fact that it makes possible the good of men having the opportunityfreelyto show compassiontowardssufferinganimals.God could not make any manfreelyshow compassion - for the well-known This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 290 Richard Swinburne reason that libertarianfree will is incompatiblewith predetermination. Buthe could give men the libertarianfree will, having which is a necessary condition of their using it compassionately. So do they have that free will? In so far as an after-life,or the existence of angels with libertarianfree will of a kind that allows them to choose evil, are necessary for allowing some evil to make possible some good, is there an afterlife, or do such angels exist? I summarize the issue by asking: are all evils in the world such that a 'greatergood' defence works in respect of them? Many an inquirercannot see how a 'greatergood' defence does work with respect to all occurrentevils. Often an evil seems to be such that it would be wrong for God ever (or in existing circumstances)to permit it for the sake of any greatergood; or that it does not make possible any greatergood; or that, if it does, there is a betterway of making that good possible; or that the good for which the evil is a necessary condition does not in fact occur. It will be importantfor me to distinguishamong the inquirer'sdoubts between moral doubts and doubts about contingent non-moral fact, and for this purpose I need to establisha position on the status of moral judgments.11hold that they have truth-value;some are true and some are false. I do not need to argue for that aspect of my position in this context, since anyone who thinks that evil raises for theism the 'problem' which I have described must think this. There could only arise an issue as to whether certainevils were compatible with the existence of a good God if goodness and evil were properties which belonged to persons, actions,and states of affairs,and judgmentswhich affirmed or denied their possession had a truth value. Now the moral goodness (or otherwise) of particularactions is supervenient on non-moralfeaturesof those actions. Thatis, judgments such as 'Youought to fight for France,'It would be good to give money to Oxfam,' derive their truth-value from contingent states of affairs describablein non-moralterms, i.e. (in Mooreanterms)'natural'states. What would make the first true would be such things as you being a French citizen, and Francebeing attackedby an enemy who seeks to rule her; what would make the second true would be such things as that Oxfam feeds the starving in Africabut does not have enough money for this purpose. But once all the naturalcircumstanceswhich make the moraljudgment true are set out, that in those circumstances (fully described)you ought to fight for France,or it is good to give to 1 What follows for the next two pages is a condensed view of the position advocated in my 'The Objectivity of Morality/ Philosophy 51 (1976) 5-20. This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 291 Oxfam, cannot but be true; it is a necessary truth. There could not be a world identical with the actual world in all natural respects but in which different moraljudgments were true (e.g., a world identical with our world in all naturalrespects but in which murderand torture were morallygood actions).No doubtparticularnecessarymoraltruths, such as 'In circumstancesC you ought to fight for France'derive their truthfrom more general moraltruths such as 'Everyoneought to fight for his country when it is attackedby enemies who wish to rule over it,' but my argument does not depend on such derivability.My only point is that, if moraljudgments are true, and those which are contingently true are supervenient on naturalfacts, then the judgments that they hold when those natural facts hold must be necessary. Hence contingent moral truths derive their truth from contingent natural truths and necessary moral truths. The basic moral truths are necessary truths. We saw earlier that, if an inquirerwas to see how a greater good defence worked with respect to some occurrentevil, he would have to understandfive things. The conclusion which we have just reached about the status of moral judgments shows us that these things boil down to three necessarytruthsand two contingenttruths.The inquirer needs to know that the evil E is such that there is a necessary moral truth of the form 'It is not wrong for God to allow E to occur in circumstances Cv' and, second, he needs to know the contingent natural truth that Ca holds (unless, of course, Ca is 'all circumstances,'in which case no second piece of knowledge is necessary). He needs to know the necessary truth that to allow the occurrenceof E entails allowing a logically necessary condition of a state of affairs C2 at least as good as the evil is bad. Fourth, he needs to know the necessary truth that to allow the occurrenceof E (or an evil equally bad) is the only morally permissible way of bringing about that logically necessary condition. And, fifth, he needs to know the contingent natural truth that the good state of affairs C2 in fact occurs. An inquirerneeds to know all these things with respect to the evils of which he has knowledge if he is to see how a greatergood defence works with respect to them. Often a theist cannot see how a greater good defence works with respect to all the evils of which he knows. And not merely so, but it seems often to a theist that there are many evils for which it does not work. Forit seems to him often that for some E the necessary moral truths of the kind described in my first, third and fourth points above are such that, whatever the contingent natural circumstances,it would be incompatiblewith God's perfect goodness that he should allow E to occur; and it seems to him sometimes that, although the necessary moral truths are such that it would be compatiblewith God's perfectgoodness that he should allow E to occur This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 292 Richard Swinburne under certain contingent circumstances C (i.e., if both Ca and C2 occur), it is unlikely that these circumstancesoccur. If the theist is in this position, then eitherit seems to him that there is an evil E which is incompatiblewith the existence of God or it seems to him that there is an evil E which together with not-C entails the non-existence of God, and it seems to him on his evidence unlikely that C obtains, and so unlikely that there is a God. So surely it is not rationalfor him to believe that there is a God. Does he not need a theodicy, a justified account of how such evils do (contraryto appearances) serve a greater good, before he is justified in continuing to believe? II The supreme principle which guides rationalbelief on all matters is surely the principlewhich I have called elsewhere the Principleof Credulity:that, other things being equal, it is rationalto believe that things are as they seem to be. By 'seem' (or 'appear')I mean 'seem epistemically'; the way things seem epistemically is the way we are initially inclined to believe that they are. We find ourselves with involuntary inclinationsto belief;in the absenceof reasonsagainstgoing along with such an inclinationthe rationalman will do so. This must be the starting point for all knowledge; if all beliefs needed to be justifiedby other beliefs before they could be believed with justification,no belief could ever be justified. However, we soon learn to distinguish kinds of belief, and to recognize limits in each kind to the applicationof the Principleof Credulity because it seemsto us (epistemically)right to do so; applying the Principle enables us to see limits to its application.We come to distinguish (perhaps merely in practice, or, if we are philosophers, in a careful verbalizedway) between the realm of the logically necessary and the realm of the logically contingent, and, within each group, between propositions which claim our belief because they purportto represent states of affairsof which we are immediately aware, and those which claim our belief because they entail or are non-deductively supported by propositions of the former type. With respect to the purportedly necessary: if p seems to entail q, or to be compatible, or to be incompatiblewith q, to be a necessary truth, to be a contingent truth, or to be self-contradictory,we ought so to believe - until counter-evidence can be produced. Counterevidence to an initial judgment of necessary truth, such as that p entails q, may be of various kinds. We may, for example, derive from This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 293 many other cases of evident entailmentrules of deductive logic which for many cases lay down when propositions entail each other. These rules may have the consequence that p entails not-q. If it seems to us more evident that the rules are correctand that this consequence follows from them, than that the initial judgment was correct, we must abandon the initial judgment. The realm of the logically necessary comes to us as a realmof things which could not be otherwise,to which our mind has access by mere reflection.How things in this realmseem positively (e.g., that p entails q) and how they seem negatively (e.g., that p does not entailq) would seem equallyto be entitledto our belief. The logicallycontingentcould be otherwise. If, nevertheless,we take how things seem to be as evidence of how they are, that can only be because we believe that in general our inclinationsto belief are sensitive to the way things are. But there can only be this general sensitivity if there is a causal mechanism at work - either a direct causal mechanism, whereby a state of affairs S causes me to believe that S holds; or an indirectmechanismwhereby there is some state of affairs R which causes my belief that S holds and also causes S. If I take how things seem as evidence of how they are because I believe that there is a direct causal mechanism at work, I purport to have a belief justified by experience - by perception or (if the belief is revived belief) by memory. If I take how things seem as evidence of how they are because I believe that there is an indirect causal mechanism at work, we may say (for want of a better word) that I purport to have belief justifiedby indirectperception. How things seem initiallyeither to be necessary truths or to be contingent truths given to the subject by a causal process I shall call his basic beliefs. With apparentcontingenttruths, even more obviously than with apparent necessary truths, basic beliefs ought often to be abandoned under the pressure of counter-evidence. It may seem initially to me that there is an elephant in my back garden, but if everyone else tells me it is not there, and if when I try to touch it my hand passes through where it seems to be, I must abandon my belief. Where basic beliefs are abandoned, they are abandoned under the pressure of stronger beliefs with which the former seem to conflict. These latter may be basic beliefs or, more likely, beliefs generated by deduction or induction from many other basic beliefs. We move by deductive inference to consequences buried within those beliefs. We move by inductive inferencebeyond observationsto claims about the future or distant past or the laws of nature. We do so in ways that initiallyseem right to us. But through knowledge of how we infer and others infer on other occasions, different inferentialsteps seem to us the right ones from those which seemed so at first. Our genetically This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 294 Richard Swinburne inbuilt habits of inference become refined through practice. We may even try to put into words the standardsof inductive inference which we use, and through this process too we refine our practice;more consistent standards may seem the right ones. Beliefs formed by inductive inference from many apparent experiences then come to correct the belief initially formed on the basis of experience. Beliefs formed by inductive inference may come to correctnot just particular'apparent seemings' but whole classes of them. Induction teachesus that whole classes of apparentperceptions- of certainkinds of thing by certain observers in certain circumstances - are unreliable. The latter is a consequence of our general knowledge of how the world works, establishedby induction from a far wider class of apparent perceptions than those whose reliabilityis at stake. How contingent things seem positively to be must be better justification for a claim about how they are than is how things seem negatively a justificationfor a claim about how they are not. 'It seems to me that there is a table in the house' asserted as a reportof an inclination to believe on the basis of experience is better grounds for 'There is a table in the house' than is 'It seems to me that there is no table in the house' grounds for 'Thereis no table in the house.' This is because for the positive judgment to be reliable only one causal chain needs to go from the object apparentlyperceived to the subject who seems to perceiveit. Butif the negativejudgmentis to be reliable,causal chains need to go from all places where the objectmight be to the subject who seems to perceive that the object is not present. Detailed investigationmay show which negativejudgmentsarejustifiablyasserted on the basis of experience;but, even without it, the generalpoint about the extent of causal chains needed for such judgments to be reliable makes them much more shaky than positive judgments.2Detailedem- 2 I wrongly asserted in earlier writing The Existenceof God [Oxford: Clarendon Press 1979] 254f.) that negative seemings are no evidence at all of how things are; I now correct that to an assertion that, as such (in advance of detailed empirical evidence about them), negative 'seemings' on empirical matters give much shakier support to claims about how things are not that do positive 'seemings' to how things are. It remains the case that 'It seems to me that there is a God' is much better grounds for supposing that there is than 'It seems to me that there is no God' is for supposing that there is no God (where both 'seemings' report apparent perceptions of things contingent). For if there is a God, one causal chain from God to me would not be unexpected; but if there is no God, a bundle of causal chains from all the places or states of affairs where there is no God to me is a very dubious large-scale position. This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 295 piricalinvestigationwill refine this initialposition, showing how some negative 'ssemings' are good evidence and others are not. We learn, then, through practice, correctingour initial judgments of how things seem to us, and, correctedby such procedures (which seem to us the right ones to use), things seem to us differently. The Principleof Credulityremainsthe basicmechanismfor passing beyond our mental life to justified judgments about the necessary and contingent world outside, but we learn through practicehow to correctits judgments. It follows from this general accountof rationalbelief that, if it seems to the theist that there is an evil E incompatiblewith the existence of God, he ought so to believe, and so to believe that there is no God. If it seems to him that there is an evil E which, together with not-C, entails the non-existence of God, and it seems to him likely that notC, and so unlikely that there is a God, he ought so to believe. And, since believing p unlikely entails believing not-p more likely that p, it entails believing that not-p. The theist in that situation ought to believe that there is no God. All this - unless he has evidence which seems to him to tell against that conclusion. The onus of proof has passed to the theist; he needs reason for resisting the conclusion. Ill The reason may be of three kinds. It may consist of other reasons for affirmingthat there is a God; or generalreasons for doubtingthe force of this sort of inferenceto his non-existence;or reasons for supposing that a greatergood defence works in respectof the particularevil which seemed initially to show the non-existence of God, in other words, a theodicy. Let us begin with the first reason. Any argumentfor denying a true conclusion must be unsound; and if it seems to a subject strongly enough (as a necessary truth or a deliveranceof experience)that some propositionis true, then any inference to its falsity must be unsound. If it seems to a subjectstrongly enough that there is an externalworld or that 2+2=4, powerfulphilosophicalargumentsto the contrarymust be judged by the subjectto be unsound, even if he cannot put his finger on where they go wrong. Likewise, if it seems to a subject strongly enough (as a basic belief, delivered either by experience or reason) that there is a God, then he is entitled to the belief, whatever else he is initiallyinclined to believe. A purportedlydeductive inferencefrom E to the non-existence of God must start from a false premiss or be invalid. The inductve inferencevia non-C may be justified,but its con- This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 296 Richard Swinburne elusion (however likely it is rendered by E) must be regarded by the believer as false. So much is, I hope, undeniable. But the basic belief will have to be overwhelmingly powerful if it is not to be overcome by an apparentlysimple deductive inference, relying on apparentlyincontrovertiblemoral principles, from a very evident evil. The less sure the thinker is about his moral principles, the less strong does the basic belief need to be. Again, the less confident is the thinkerabout the inferentialprinciplesinvolved in his inference via non-C, the less strong does the basic belief in the existence of God need to be in order to overcome it. Also, in this case, the conclusion of the inductive inference is only that it is likely that not-C. But if it seems to the subject that it is very likely that there is a God, again on balance the evidence against C is outweighed. Even if the belief that there is a God is not a basic belief, as long as it seems to the subject to be supported by evident basic beliefs on the basis of evidently sound inferential principles, it may be strong enough for the subject rationally not to abandon it in the light of counter-argument.Again, it will depend on how evident are the moral and inferentialprinciples involved in the counter-argumentand how likely it seems to the subjectthat they make any state of affairsincompatible, given a certain evil, with the existence of God. Apartfrompositive evidence for the negationof the conclusion,there may be reason for doubting that the principles of inference, and so that any purportedlynecessary truths involved in any argumentfrom evil, are as sound as initiallythey seem to be. In particular,some moral principlefrom which it follows that the existence of God is incompatible with the occurrenceof a certain evil may seem initially evident. But we derive our moral principlesfrom reflectionon particularcases of how humans behave towards each other, judging one kind of behaviour good and another kind wrong, and then apply them to judging what would be good or wrong for God to do. The God whose existence is at stake will, if he exists, be very different in nature from ourselves, and so have kinds of good and evil open to him of which we cannotdream,and understandthe worth of actions(especiallythose open to him alone) with a depth which we cannot. As humans grow up and become more sensitive to the natureand consequences of their actions, they often come to see many kinds of action as good which previously they regarded as bad: e.g., discipline, loyalty, patience, courage and tolerance. But our experience and our reflectionupon it are still so limited that it would not be surprisingif we had yet more to learn about the nature and consequences of actions which would lead to our revising further our judgments about their moral quality. Our extrapolationsfrom human situations to what would be good or This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodiq/? 297 evil conduct for a God must be particularlyshaky. Hence, if there is a God, we would expect to find some evils which, it seems to us initially, he could not allow to occur, compatiblywith his perfect goodness. Theismis in this respect similarto a hypothesis H that some very clever criminaldid the crime;if H is true, one will expect to find some evidence apparentlysuggesting (i.e., evidence which would suggest in the absence of other evidence) that H is false (because that criminal, unlike others, would have planted misleading clues). Although this is so, it remains the case that - barringa consideration to be discussed below - we are only justified in believing that our judgment, that a particularevil is incompatiblewith the existence of God, is erroneousbecausethe resultof inadequatemoralexperience, if on balance all the evidence (including the evidence from evil) suggests that there is a God. If it does, then it makes plausible the view that certain evils are really justified by a 'greatergood defence/ and only seem to count against the existence of God because of the distance in moral understandingbetween ourselves and God. The point about the likely inadequacyof our moral understanding has the consequence that, if the balance tilts Godwards, it tilts more decisively Godwards than it would otherwise do. But otherwise - barringthe considerationbelow - we have no reason to suppose that if our inadequatemoralunderstandingof what a God would do were improved we should come to see the presently puzzling evils as such as God would permit;the improvementin our understandingmight as easily lead us to see existing states of affairswhich we now think of as ones which God would allow as ones which God would not permit. Only if on balance we already have reason to believe that there is a God, do we have reason to suppose that the deficiency of our moral understanding concerns and concerns alone those puzzling evils which would otherwise count against the existence of God. There is a furtherconsiderationwhich, if it held, would reasonably lead us to suppose that improvement in our moral understanding would be in the directionof seeing all present puzzling evils as really such as God would permit. This is that when initiallyit seemed to an inquirerthat many kinds of evil were incompatiblewith the existence of God (straightforwardly,or given furthercontingent naturaltruths which seemed to the inquirerto be highly likely), subsequent reflection, argument, and experience led him to see, one by one, with respect to a numberof such kinds of evil, with apparentclarity,that they were not incompatible with the existence of God; a greater good defence apparentlyworked in respect of them: but the reverse did not happen - the inquirerdid not come to see states of affairswhich he regarded initially as good as apparently incompatiblewith the exis- This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 298 Richard Swinburne tence of God. In Lakatos'terminology,3the theisticresearchprogramme would have proved progressive. That gives grounds to the inquirer to suppose that eventually he will apparentlysee that all kinds of occurrent evil are justified in terms of a greater good defence. Whether a given inquirerhas this kind of evidence of finding more and more evils explicablein terms of a greatergood defence is an empiricalmatter.Certainlysome inquirershave a recordof such progressive understanding;and an inquirerwith such a recordmust rationally be much more doubtfulaboutthe moralprinciplesinvolved in his judging the remainingpuzzling evils to be incompatiblewith the existence of God than he would otherwise be. Depending on the strength of his initial conviction, this evidence of his past recordmay tip the balance in favour of God. I conclude that it follows from the Principleof Credulitythat apparent evils must count against the existence of God, but, even in the absence of a greater good defence in respect of each such evil, other considerations may still outweigh the force of apparent evil, so that on balance it remains rational to believe that there is a God. The provision of a greatergood defence with respect to a given kind of evil is the provision for that evil of a theodicy, in the sense of a reason why (compatiblywith his perfect goodness) God could allow that evil to occur.Barringpoints made above, the theist needs for each kind of evil which seems to him to count againstthe existenceof God a theodicy in this sense; he does not need a theodicy in the sense of God's actual reasons for allowing some evil to occur. We have seen that, barringvery strong evidence of the existence of God, an inquirerdoes at any rateneed a theodicy with respect to some of the evils in the world which seem to him to count against the existence of God, which would lead him to suppose that a theodicy is to be had with respect to other evils. And even that evidence of increasing success of theodicy might be quite insufficientto defuse the power of apparentevil in the absenceof a full theodicy, a theodicyfor all kinds of evil. IV the view that the theist does not need a theodicy of this kind in order rationallyto believe that there is a God has recently been put forward 3 I. Lakatos, 'Falsification and the Methodology of Scientific Research Programmes' in I. Lakatos and A. Musgrove, eds., Criticismand the Growth of Knowledge(Cambridge: Cambridge University Press 1970) This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 299 by Stephen Wykstra4in an interesting article,which seems to suggest that apparentevils do not count at all against theism, independently of whether the balance of evidence actually supports theism or not. Wykstrawrites in criticismof Rowe, who uses in effect the principle which I have calledthe Principleof Credulity.Wykstrais unhappy with the qualificationwhich I originally put on the principle, to the effect that, while how things seem positively is evidence of how they are, how things seem negatively is not. (I have modified this qualification extensively in the present paper.) He suggests instead an alternative qualificationwhich, he claims, is sensitive to the intuitions lying behind my original qualification. Wykstraplaces his alternativequalificationnot on the inferencefrom 'seems' to 'is' but on which assertions of 'seems' are justified. Wykstra calls his limiting principle CORNEA of 'Condition of Reasonable Epistemic Access': On the basis of cognized situation S, human H is entitled to claim "It appears that p" only if it is reasonable for H to believe that, given her cognitive faculties and the use she has made of them, if p were not the case, S would likely be different than it is in some way discernible by her. (85) Thatis, we can only justifiablyassert 'Itappearsthat p' if we have reason to believe that, if not-p, the observed situation would be different in respect of those features which lead us to assert 'It appears that p.' Wykstrathen uses this principle to claim that we cannot claim that 'it appearsthat there are evils which serve no God-justifyingpurpose/ i.e., 'it appearsthat there are evils such that God would not allow them to occur.' He claims that we cannot make this claim because, if there were no evils such that God would not allow them to occur, we would be unlikelyto recognizethis fact;our moralbeliefs being so much shallower than those of God, we are bound to think some states of affairs to be evils serving no greater good, although in fact they do serve a greater good. Wykstra'sCORNEAreminds the readerof a similarconditionin Nozick's account of knowledge.5 It is one of Nozick's conditions for S knows that p that 'if not-p, S would not believe p.' But Nozick is careful to point out that 'ifnot-p'concernswhat would happen in the closest 4 Stephen J. Wykstra, 'The Humean Obstacle to Evidential Arguments from Suffering: On Avoiding the Evils of "Appearance/" InternationalJournalfor the Philosophy of Religion 16 (1984) 73-93 5 R. Nozick, PhilosophicalExplanations(Oxford: Clarendon Press 1981), chapter 3. See especially p.172 and p.199. This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 300 Richard Swinburne worlds in which p is false. It is for that reason that I do know that I am readinga book; even if, were I not reading a book and were a brain in a vat subjectto certainstimuli, I would stillbelieve that I was reading a book;the point is that, if I was not readinga book and otherwisethings were much the same as at present, I would not believe that I was reading a book, and that (in part)is why I know that I am reading a book. Analogously, if CORNEAis to plausible, 'if there were no evils such that God would not allow them to occur'is presumablyto be read as 'in the closest possible world to ours in which there were no evils such that God would not allow them to occur/ viz., if things were different solely in the respect that there were no such evils and, as far as logically possible, otherwise the same. But, that being so, Wykstrais not entitled to claim that, if this was the case, we would not recognize it. For, if the evils were removed from our world which are such that, if there is a God, he would not permitthem, that would of course make no difference if there is in fact a God. But if such evils are removed and there is no God, then this could be expected to make a difference which we would recognize. For there is no reason to suppose that in that case our moral beliefs would be badly deficient in just the kind of way that would lead us to think that there are evils which do not serve a greater good, when in fact all the evils in the world do serve a greatergood. As I wrote earlier, our moral beliefs are just as likely to be at fault in another direction - leading us not to recognize many other evils around us which do not serve a greatergood, while being on the rightlines with respect to our judgments that certainevils serve no greatergood. WhetherWykstra'sclaimthat we wouldn't recognize a differenceif ther were no 'evils which serve no God-justifyingpurpose' depends on whether our world, and so the closest possible world to ours, contains God. In other words, Wykstrahas begged the question. He does so in a way which, when brought to light, reinforces my earlierpoint that whether the Principleof Credulityought to lead us from evils apparentlycounting against the existence of God to the non-existence of God depends crucially on the extent of our other grounds for believing that there is a God. Only to the extent to which he has grounds for believing that there is a God does he have grounds for supposing that we would not notice any difference made by the removal of 'evils which serve no God-justifyingpurpose,' and so for using CORNEAto defuse the force of an argumentfrom apparentevil against the existence of God. V In analysing what the rationalinquirerwill come to believe as a result of it seeming to him that there exist evils which serve no greatergood, This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 301 I was in effectconcernedwith how an inquirermay securewhat I called in Faithand Reasonrational}belief.6 A belief is rational} in so far as (roughly) it fits with and is supported by the subject's other beliefs, by his own standards. A subject's belief that there is a God will be irrational if there are evils apparently (by his own moral principles) incompatiblewith the existence of God, or incompatiblegiven some contingent naturalcircumstanceswhich, by his own standards, very likely exist, unless there is very strong evidence of or supported by basic beliefs, that there is a God. Those concerned for truth will seek to conform their beliefs to how things are, and that minimallyinvolves ensuringthat they believe only those propositionswhich are likely to be true by standardswhich seem to them correct.However, the enquirermust go on to seek more than rationality2-He must seek to believe only those things which are likely to be true, given true standards for assessing the evidence; and his beliefs will be rational2in so far as he does so believe. An inquirer's beliefs are rational2if those are the beliefs which would be reached by an inquirerpossessed of the same empiricalevidence as himself, and having and ableto apply correctlyall correctprinciplesof inference. We distinguish what follows deductively from a premiss from what seems to some individual at some time to followfrom the premiss. If there are correct principles of inference of a non-deductive kind, as we all assume when doing science or history, we can make the same distinction with respect to a far wider class of inferences. Evidence e does in fact 'support'h ('renderh probable,'or however we describe the relationof non-deductivejustification),if a being who applied correctly those correctprinciples of inference would be able to draw out h from e. Since adding necessary truths to premisses adds nothing to what follows from them, a being who can draw any correct conclusion from e can draw any correct conclusion from e conjoined with any necessary truth, including that truth itself. So he would in effect be in possession of all necessary truths (and given the results of an earlier section, that includes the necessary moral truths); so that an agent'sbeliefs are rational2if they are in fact supportedby his evidence (whetheror not he realizesthat they are). He, of course, can only guide his conduct by beliefs which seem to him to be supported by the evidence, but he will seek to make his beliefs rational}, and thence rational2. Now the existence or non-existence of God is an all-importantmatter, and the view formedby an initialjudgment must be examinedfur- 6 (Oxford: Clarendon Press 1981), chapter 2 This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 302 Richard Swinburne ther. We must seek to acquirenew contingentevidence (i.e., new basic beliefs about how things seem contingently to us) and we must test our moralprinciples, and principlesof deductive and inductive inference, through experience, reading, and reflection. In so far as we fail to give enough time to doing this with respect to so importanta matter, our resulting belief will fail to be what I called in FaithandReason rational3,rational4,or rational5.A belief is rational3so long as it results from an investigationwhich the inquirerhimself consideredto be sufficiently lengthy and serious; a belief is rational4so long as it results from an investigationwhich by the inquirer'sown standardswas sufficiently long and serious - whether or not he formed that view about it; a belief is rational5as long as it results from an investigation which by true objective standards was sufficiently long and serious. But, although proper investigation is of great importance, at any given time an inquirer'srationalityconsists in holding such beliefs as are supported by evidence; and that is a matter of rationality! and rationality2.As I wrote, my earlier discussion was concerned with rationalitya. But inquirerswho seek true beliefs will, for that purpose, seek not merelyrational!beliefs but, further,rational2beliefs;will seek to have beliefs which are in an objectivesense supported by their empiricalevidence. If in fact, given contingentempiricalfacts k, some evil e could serve no greatergood, then (given necessary moral truths) (e and k) entails the falsity of h, the hypothesis that there is a God. On evidence (e and k), the belief that not-h will be rational2.If, given contingent empiricalfacts k, some evil e could serve a greatergood only if / certainother naturalcontingent states, holds, if (e and k) make it unlikely that f, then they make it unlikely that h. Again, on evidence (e and k), the belief that not-h will be rational2.(In this case, however, unlike the formercase, [e and k] may hold and h still be true, even if it is rational2to believe otherwise.) Rationality!depends on the subject's estimate of rationality2.A subject'sbeliefs are rational!in so far as by his best estimate (by his own standards, or a weighted average of various standards, in so far as he is uncertainabout them, weighted by the degree of his confidence in each) they are supported by the evidence. It is most unfortunatethat various writers who concern themselves with the rationalityof religious belief do not distinguish between the different kinds of rationality;and some of the disagreements about when religious belief is 'rational'clearly arise from this source. However, most writers who write about the kinds of matter discussed in previous sections are most plausibly interpreted as being concerned with 'rationality!';they are concernedwith whether an inquirer'sbeliefsystem is internally coherent. This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodiqj? 303 VI Is the way evidence functions here adequatelycapturedby the probabilitycalculus,and in particularby Bayes'theorem, which follows from the axioms of the calculus? I believe that it is, on the logical theory of probability,in the sense that the objective'support'or making 'likely' given by evidence to hypothesis obeys the axioms of 'probability'; and so, that a subject'sbelief is rational2in so far as it is rendered'probable' by evidence, by a probabilitywhich obeys those axioms. The logicaltheoryof probabilityis concernedwith the extentto which one proposition(or conjunctionor disjunctionof propositions)renders probableanotherproposition (or conjunctionor disjunctionof propositions). Otheraccountsof 'probability'are not basicallyconcernedwith evidential support. The personalist or subjective theory is concerned with measuresof individualbelief, the propensitytheorywith a 'wouldbe' in nature, and the statisticalor frequency theory with proportions of objects possessing one property among objects possessing another. 'Probabilities'in the senses analysed by one or other of these theories may be used as our evidence or hypothesis when we are concernedto assess the extent of evidentialsupport for some evidence or hypothesis, but for them 'probability'is not the relation of evidential support. Since our concern is with the evidential support provided by certainevidence for the hypothesis of the existence of God, our concern is with 'probability'as analysed by logical theory, with functions of the type P(p/q) measuring how probable evidence q makes hypothesis p. P(p/q) = 1 means that, given q, p cannot but be true. P(p/q) = 0 means that, given q, p cannot but be false. Probabilitylies between 1 and 0. Bayes' theorem, put in these terms, may be expressed as follows: p(h/e.k) = ^^xP(h/k) P(e/k). h is the hypothesis up for investigation, e is the evidence of observation, k is background knowledge. Our evidence may be distributed as we choose between e and k; k may represent mere tautological knowledge or, alternatively,some detailedcontingentempiricalknowledge. P(h/k) is the prior probabilityof h, how likely h was to be true, before we acquirednew evidence e, on backgroundknowledge alone. P(h/e.k) is the posterior probabilityof h, how likely h is to be true in the light of evidence e(as well as backgroundknowledge). P(e/h.k) is the 'predictivepower of h, how likely, if h is true (and k also), it is that e will occur. P(e/k) is the prior probabilityof e, how likely it is This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 304 Richard Swinburne that e will occur, if we do not assume that h is true (but only assume k). The value of P(e/k) is the sum of the probabilitiesof the different routes by which e could occur; viz., P(e/k) = P(e/h.k) P(h/k) + P(e/ha.k)P(ha/k) + P(e/h2.k)P(h2/k)... and so on, where h,hlr h2, etc. are mutually exclusive and exhaustive alternatives. P(e/h.k) / P(e/k) is the 'explanatorypower' of h, how much more likely e is to occur if h is true than it would be if this is not assumed. It will be seen that the lower the prior probabilitiesof the alternativesto h (viz., hy h2, etc.) and the lower their predictive power, the greaterwill be the explanatory power of h. I have argued in many places7that Bayes'Theoremis true in the following sense. If we can give exact numericalvalues to the component probabilities,the theorem states truly the numericalrelationsbetween these. In so far as values can be ascribedto components only within a range (e.g., that P(e/h.k) is greater than 0 but less than P(e/k)), it is possible to find arbitraryvalues within those ranges under which the theorem states correctnumericalrelations. Hence all principlesof comparativeprobabilitywhich are entailedby Bayes'Theoremaretrue, e.g., P(h/e.k) > P(h/k) iff P(e/h.k) > P(e/k). T(h/e.k)' is to be read as 'e confirms h,' that is, increases its probability above its prior probability. Crucialfor the applicationof Bayes'Theorem, especially in the context of assessing theism, is the assessment of the prior probabilityof hypotheses on tautologicalbackgroundevidence, which I call their intrinsicprobability.Although P(h/k), where k is (p v ~p) measuresthis, we are unlikely to get any feel for its value, comparativelyor absolutely, by asking ourselves what we would say about how likely h would be, if we knew nothing except a truth of logic. Rather,the way to go about finding what determines this value is to consider how we assess the posterior probabilitiesof competing hypotheses, relative to each other, when those two hypotheses have equal predictivepower with respect to a body of evidence e, and there is nothing in our empiricalbackgroundknowledge to favour one such hypothesis above another. Considering such situations, we may see that there are two factors which give a value to intrinsic probability,in so far as it has a value. One is content, the other is simplicity. The more content a 7 What follows is the briefest of summaries of a position for which I have argued at length elsewhere. See especially my The Existenceof God (Oxford: Clarendon Press 1979) in particular chapter 3, and An Introductionto ConfirmationTheory(London: Methuen 1973). This summary is, however, necessary in order to defend the view that the existence of apparent evil affects the probability of the existence of God. This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 305 hypothesishas (the more detailedand wide rangingits claims),as such, the lower its intrinsicprobability.The simplera hypothesis - the fewer entities, fewer kinds of entities, mathematicallysimpler modes of behaviour, it postulates - as such, the greater its intrinsic probability. These two factors combine together to determine the intrinsicprobability of a hypothesis. Content, however, becomes of less importance when we are dealing with all-embracingtheories. The great strength of Newton's theory of gravitation, apart from its explanatorypower with respect to a vast range of phenomena, lay in its simplicity - in particularthe mathematicalsimplicity of its three laws of motion and its law of gravitationalattraction.I must make clear, as I did not in earlierwriting, that the 'simplicity'of a proposition h in my sense is a matterof the simplicityin a more normalsense of the simplestproposition logically equivalent to h. (For example, conjoining to h a complicated necessary truth makes a proposition no less simple, and so no less likelyto be true, than the originalh.) By the 'content'of a proposition I understand 'empiricalcontent'; a proposition has greatercontent, the fewer the possible worlds in which it is true. In The ProbabilisticArgumentfrom Evil,'8Alvin Plantingacriticized the view that the occurrenceof evil makes improbablethe existence of God - as 'probability'is understood by logical theory; and, more generally, the view that the probabilitycalculus, interpretedby logical theory, was of any use for assessing the probabilityof any hypothesis. His criticisms of my views contained therein were due to two misunderstandings,one his fault and one mine. He erroneously supposed that I held that simplicitywas the sole determinantof intrinsic probability,and produced knock-down arguments against that view and against a rivalview that content was the sole determinant.These arguments work, but neither of the views advocated was mine, and my view was stated explicitlyin my writing. Plantinga'ssecond criticism was well to the point. He selected pairs of propositions, logically equivalent(and thus of equal content) yet greatly differingin 'simplicity/ and pointed out that it was an axiom of the calculus that logically equivalent propositions always have the same probability.I ought to have said, but did not say (because I did not adequately appreciate), that the 'simplicity'of a proposition h in my sense was to be taken as the simplicity in a more normal sense of the simplest proposition logically equivalent to it. It is a matter of the simplicity of the world, whose existence it entails. 8 Philosophical Studies 35 (1979) 1-53 This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 306 Richard Swinburne These points sufficeto renderharmlessthe purportedlyknock-down arguments contained in Plantinga'spaper. His other arguments concern the difficulty of how we are to go about attributingto propositions intrinsic probabilities.9My claim was, however, only that the principles stated above often yield comparativeintrinsicprobabilities (e.g., show that P^/k) > P(h2/k)or P(ha/k) < 1 or P(h1/k) = 1/2); and that when they do so, we can always then arbitrarilyattributenumericalprobabilitiesin such a way as to conform to Bayes' Theorem and, more generally, to the axioms of the calculus. 9 One such difficulty arises where we have an infinite number of exclusive and exhaustive hypotheses; in many such cases, Plantinga suggests (see his pp. 25-30), it would be intuitively plausible to ascribe equal intrinsic probability to each, but then given the usual spelling out of the axioms of the calculus, the prior probability of each member of such a set will be 0. If the intrinsic probability of a hypothesis is 0, its posterior probability on any evidence will also be 0, and no such hypothesis can be confirmed by evidence. Plantinga suggests that any set of the form: HQ,'there are no A's'; Hv 'there is just 1 A'; H2 'there are just 2 A's' ... and so on, conforms to this pattern. For there is, he claims, nothing to choose between such hypotheses on grounds of content or simplicity. They all have equal content and are equally simple, he suggests; and he gives an example of such a set when 'A' = 'horse/ Plantinga's example is a bizarre one. The hypotheses between which we seek to make a judgment in science, history or other fields of inquiry, have a fuller explicit structure which makes their relative simplicity easier to assess. But if we do take his hypotheses by themselves we find when we look at them more closely that they are committed to more than their form explicitly reveals. Ho is quite different from the others; the postulation of nothing is always simpler than the postulation of something. As regards the others - their explicit claims concern only horses. But (of logical necessity) horses cannot exist on their own. To be a horse, something has to take in food and air, and get rid of waste products. Plausibly too, it has to have a relatively short finite life. The various hypotheses H|, H2 . . . are committed to (and so implicitly postulate) whatever else is necessary for them to be true (e.g., H2 is committed to the existence of food and air for two horses, some causal mechanism which caused the existence of two horses, any other horses which were caused to exist having been eliminated). The simplicity of a proposition is a matter of the simplicity of the world whose existence it entails, and the relative simplicity of the hypotheses about horses turns on the relative simplicity of the further biological assumptions which would need to hold if the theory is to be true. That being so, I suggest that a simpler story can be told of how there comes to be a horse population of modest size on one planet, than of how there comes to be a solitary horse somewhere in the universe, or of how there comes to be a universe swarming with horses (e.g., by means of qualitatively identical horse-generating mechanisms on each of many distant planets). When we give some filling to such bizarre examples, judgments of relative simplicity become plausible, which do not yield Plantinga's negative conclusions. This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodiq/? 307 So Plantinga'sarguments do not succeed in showing that we cannot assess the probabilityof large scale theories on evidence, in accordance with the axioms of the probabilitycalculus, as spelled out by logical theory, using the criteriaof content and simplicity for assessing intrinsicprobability.Thatbeing said, it will be seen that the probabilitycalculus captures more formallyour intuitive judgments about the evidential force in an objectivesense for the hypothesis of theism of the existence of evils which seem to serve no greater good. Some of these latter,as we have seen, we judge to be such as to serve a higher good only if there are entities and processes in the world other than those observable(e.g., humans have free will, or there is an after-life). Let us suppose that we are right in this judgment. Then let e be the evidence of the existence of these evils, k our other evidence about the world, and h the hypothesis of theism. Let f be such entities and processes as we need to postulate if God is to be justified in allowing e. Then whether e disconfirms h, or is irrelevantto h, or even confirmsh depends on variouspossible relationswhich may hold between h, e, f, and k. One theoreticalpossibility is that h may entail f, and f entail e: viz. in virtue of his perfectgoodness God has to bring about an f which includes e. In that case e will always confirmh (the occurrence ofevils will be positive evidence for the existence of God). For P(e/h.k) = l and so P(e/h.k)> P(e/k). Hence P(h/e.k) > P(h/k). However I cannot think of any plausible filling for f which would satisfy this pattern. A second possibility is that while h entails f, f does not entail e. Then e will confirm h if P(e/h.f.k) > P(e/k), disconfirm if P(e/h.f.k) < P(e/k),and be irrelevantif these two probabilitiesareequal. ForP(h.f/e.k) = P(h/e.k) (since if e, f is a necessary condition of h) and P(h.f/k) = P(h/k); and so P(h/e.k) > P(h/k) iff P(h.f/e.k) > P(h.f/k) which will be so iff P(e/h.f .k) > P(e/k). If one supposed that God in virtueof his perfectgoodness would have to createsome creatureswith libertarianfree will to choose between good and evil (f) then this second possibility would apply for e as moral evil. Since it is very probable that such creatureswill sometimes bring about moral evil, for this filling e confirmsh. I do not myself think that God would have to create such creatures;there is good in his doing so and good in his not doing so, and comparisonbetween the two goods seems difficult. A third possibility is that while f entails e, h does not entail f . God in virtue of his perfect goodness does not have to bring about the necessarycondition(f) of his allowinge to occur,but if he does e must occur. In this case P(e/h.f.k) = land so > P(e/k). Hence P(h.f/e.k) > P(h.f/k). P(h.f/e.k) = P(h/e.k). P(h.f/k) < P(h/k), because (h.f) has greatercontent than h (since h does not entail f)- Whether e confirms or disconfirmsturns on whether P(h.f/e.k) exceeds P(h.f/k)by a greateramount than P(h/k) exceeds P(h.f/k). I cannot think of plausible examples of This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 308 Richard Swinburne this third possibility. Almost all plausible fillings for f and e are of a fourth pattern, where h does not entail f, nor does f entail e. There is some state of affairs(e.g. libertarianfree will) which God does not have to bring about but may well choose to, such that if it occurs then it is quite probable(butnot certain)that the e (e.g. moralevil) will occur (the occurrenceof f being a necessary condition of God allowing e to occur).In that case P(h.f/e.k) = P(h/e.k), and P(h.f/k) < P(h/k)for reasons given previously.Then P(h/e.k)will exceedP(h/k)only if P(h.f/e.k) exceeds P(h.f/k)by a greateramount than P(h/k)exceeds P(h.f/k). This will depend on how probableit is that, whether or not there is a God, e will occur; how probable it is that if there is a God, he will bring about f; and how probableit is that f will lead to e. Loosely, one may say that e will confirmh only if it is very probablethat God will bring about f and very probablethat f will lead to e, and not very probable that e will occur if there is no God. Otherwise e will be irrelevantor disconfirmatory.The probabilitycalculus cannot give us exact values here, but it can bring cruciallyto the surface exactly what is at stake. And what is at stake is how unlikely the evils are to occur if there is no God, and how special and peculiar an f we need to postulate in order to save the hypothesis of theism. Otherevils e we judge to be such as could serve no highergood whatever else (other than k, the rest of our knowledge about the world) might be the case. If we are right in this judgement about them, these evils will not merely disconfirmh, the hypothesis of theism, but will conclusivelyfalsifyit. Forsuch evils P(e/h.k) = 0, and so P(h/e.k) = 0.10 The probabilitycalculus will only yield the right answers about the extent of evidential support for a hypothesis, P(h/e.k), given the right values of the othervariableson which it is dependent. In so faras there is doubt about these values, e.g., doubt about whether P(e/h.k) = 0 arising from doubt about the moral principles used to reach this conclusion, there will be doubt about the evidential support for, i.e., the objective probabilityof, h. In this situation the inquirermust utilize his own standards in reaching a judgment about that value (or, if he is uncertainabout his own standards, make an average of its possible values on differentmoral and logical principles, each weighted by the 10 My concern in TheExistenceof Godwas to show that the existence of God was probable(in the objective sense) on generally accessibleevidence, and so that a belief that there is a God was on that evidence rational2.I argued that there were no evils which could serve no higher good, nor even evils which made it improbablethat there was a God or even lowered that probabilityfrom what it was on other evidence. This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 309 strength of his belief that those are the correctprinciples). His belief that h will be rational iff that value exceeds 1/2. We may now restate, in the light of this general understanding of how rationalitya is related to the probabilitycalculus, our earlier account of how a rational belief is to be formed if it seems to us initially that there are evils in the world incompatiblewith the existence of God. If it also seems to us strongly that we have evidence k which entails h, then we are right to conclude that P(h/k) = 1, and so for any e P(h/e.k) = 1, and so any moralprincipleswhich have the consequence that P(e/h.k) = 0 must be mistaken. Conversely, if we are certain of our moralprinciples, and have evidence e, such that it appears to us very strongly that h entails not-e, P(e/h) = 0 and so P(e/h.k) = 0, we must judge that P(h/e.k) = 0, however high (unless k entails h) is P(h/k). The higher is P(h/k), however, the less probable (objectively) it is that an e will be found such that P(e/h.k) = 0; and so, in so faras we have reasonfor believingthatP(h/k)is high, we have every reason for looking at such purported e very carefully, and reflecting on the principles involved. Scientists look with great scepticism at a few purported falsifying experiments of a theory which is otherwise very successful. They rightly suspect a misobservationor a miscalculation. Newton's vastly sucessful theory of gravitationlived for many years quite happily with the fact that it appeared to predict a movement of the moon's apse far slower than that observed; the observation seemed to falsify the theory. After sixty years it was shown to be a miscalculation;what was observed was in fact what was predicted by the theory. Rationalityin religiondemands a similarattitude.A solitary counter-instanceor kind of counter-instanceis rightly not taken seriously; there must be a mistake somewhere. But if P(h/k) is apparently none too high, e is well observed (and the more separateobservations there are of e, the more certain it is that something of this characterhas occurred),and reflectionon the logicalprinciples(including moralprinciples)involved apparentlyshows that P(e/h.k) = 0, then the rational man must judge that almost certainly P(h/e.k) = 0. The only reason for not so arguing with respect to some evil apparently incompatiblewith the existence of God, I claimed earlier, was if theodicy proved to have a good trackrecordin making evils at first sight apparentlyinconsistent with the existence of God not to appear so subsequently. I now put this point in terms of the calculus. Let e be the evil apparentlyincompatiblewith h, the hypothesis of theism, g this recordof the past success of theodicy and k the rest of our knowledge. The value of P(e/h.k) is a necessarytruthand cannot be changed by g; rather,g suggests that its true value is not 0. Let j = 'Whileretaiing our general rationality,and our basic moral concepts (and applying them to normalcases roughly as before), it will through reflection This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions 310 Richard Swinburne come to seem to us stronglythat P(e/h.k) * 0/ Then P(j/e.g.k) is much higher than P(j/e.k) and may come to exceed 1/2 (the record of past performancebeing so impressive despite it initiallyseeming to us that the evil e serves no higher good and so makes it likely that continuing to be generally rational, we shall recognize this fact). If we have reason to believe that sincere reflection on moral principles will lead us to a position where it will seem to us, retainingour generalrationality1 and all our present evidence and applicationof our principles to normal cases, that P(e/h.k) =£0, we ought to believe that that is so. This is a principle for rational1belief, an extension of the Principleof Credulity. It says that how things will appear to us, when we have built upon our earlierreflectionon them (in ways which have given results which appear to us to be correctin other cases) is how we ought to believe that they are. This principle can easily be seen to be correct from a less controversialexample. Suppose that a student discovers a book of purporteddeductiveproofsof some twenty theoremsof some calculus(e.g. geometry). The principlesof the science, the axioms and rules of inference are explained in the introduction.The first ten theorems seem to follow from the axioms, but the subsequent theorems seem not to follow. However, the student works at the 'proofs/ and eventually it seems to him clearthat Theorem11 does follow from the axioms;the proof given turns out to have been rathercondensed. Still, the student can't see how the remainingnine follow; in each case the affirmationof the premisses and denial of the conclusion seem to him consistent. However, he works away and eventually it seems to him clearthat the proofs of Theorem12, Theorem13, Theorem14 and Theorem 15 work. Surely this gradual accumulationof results is grounds for supposing that the remaining proofs also work. It is so because it providesgroundsfor supposing that if the student were to work hard enough at the remaining proofs, he would come to believe (while retaining his general rationality,and building on his previous achievements) that they work. And the belief that he will come to see something while retaining his general rationalityand his application of principles to clear cases is grounds for how things will appear to him when he sees more deeply, and so grounds for believingtrue what otherwise seems false. VII So, yes, in general theism does need a theodicy. A theist needs to discover, with respect to each evil which seems not to make possible any greatergood, how it does make possible a greatergood. Or, in default of that, he needs good grounds from the past success of theodicy in This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions Does Theism Need a Theodicy? 311 apparentlymakingclearto him the point of other evils, for supposing that he could eventually find a theodicy for each evil. Otherwise, evil will count against the existence of God. And if it seems to the theist stronglyon moralgrounds that (whateverthe other contingentcircumstances) present evils serve no higher good, or that factual evidence is againstthe existenceof the entities and processeswhich alone would make present evils serve higher goods, then the theist's counterevidence (in the form of a basic belief that there is a God, or a nonbasicbelief apparentlyrenderedprobableby evidence) will need to be stronger, if he is to be justified in retaining his theistic beliefs. Howver, for myself, I believe that the requiredtheodicy can be provided. ReceivedNovember, 1986 This content downloaded from 195.34.79.101 on Sat, 14 Jun 2014 10:38:16 AM All use subject to JSTOR Terms and Conditions