This fine book consists of meticulous scholarship in the history of nearly thirty years' work on God and evil, accompanied by acute, original criticism. As its subtitle indicates, it is focused on the publications of William L. Rowe, the most prominent of recent writers offering ‘evidential’ arguments from evil against the existence of God.

Trakakis's overall conclusion is that Rowe's arguments succeed in establishing a strong prima facie case in favour of atheism (341–342). Unless theists are prepared to modify their traditional claim that God is omnipotent, omniscient, and perfectly (morally) good, their only hope is that evidence can be found in favour of the existence of God that is strong enough to override the case based on evil against it. Trakakis himself holds that such evidence is in fact available, and he describes himself as ‘a theist, albeit a quite tentative one’ (3); but he does not offer any argument for theism in this book.

Trakakis distinguishes three periods in Rowe's work on the problem of evil: early (1978–1986), centred on the well-known paper ‘The problem of evil and some varieties of atheism’ (American Philosophical Quarterly, 16 (1979), 335–341); middle (1988–1995), in which Rowe discusses two particular instances of horrific evil (respectively natural and moral evil); and late (1996–present). Trakakis outlines the main paper of the late period, ‘The evidential argument from evil: a second look’, which first appeared in Daniel Howard-Snyder (ed.) The Evidential Argument from Evil (Bloomington IN: Indiana University Press, 1996), and which he sees as revamping earlier presentations in such a way as to evade various criticisms of them. He does not offer a detailed evaluation of this paper, but concentrates on those of the early and middle periods.

The middle-period argument deals with the case (E1) of a five-year old girl who is severely beaten, raped, and strangled to death, and the case (E2) of a fawn that is caught in a forest fire, is badly burned, and dies after several days of intense distress. Rowe argues: (P) no good state of affairs we know of is such that an omnipotent, omniscient being's obtaining it would morally justify that being's permitting E1 or E2. Therefore (Q) no good state of affairs is such that an omnipotent, omniscient being's obtaining it would morally justify that being's permitting E1 or E2. (According to Rowe, to ‘know of’ a good state of affairs is to conceive of it and to recognize that it is intrinsically good; one need not know that it actually obtains.)

Trakakis spends five chapters carefully assessing the inference from P to Q, in the light of the extensive literature on this matter. Rowe regards P as giving a good reason for Q only if P is conjoined with the following true ‘noseeum assumption’:

• RNA  If there are goods justifying God's permitting horrendous evil then we would probably know of them.

Trakakis thinks that it is reasonable to accept RNA, though he adds that Rowe need not rely on it, but can instead rely on the weaker assumption,

• RNA4  If there are goods justifying God in permitting horrendous evil but we do not know of them then God would make it clear to us that He exists, in such a way that we would not be inclined to infer, from the fact that we do not know of any such goods, that there are none. (191–192).

I am here paraphrasing Trakakis's formulations of RNA and RNA4. As explained above, ‘know of’ is a technical term.

Granting, for the sake of argument, Rowe's working assumption that it is not clear that God exists, attention turns to whether the consequent RNA is true, i.e. to whether we know of goods justifying God in permitting horrendous evils such as E1 and E2. Since theodicies typically purport to identify which actually obtaining goods justify God's permitting horrendous evils, a successful theodicy will support the view that we know of the justifying goods. Trakakis's discussion of theodicies for horrific moral evils – e.g. the suffering and death of the child in the E1 example – leads him to conclude that the prospects look good for a theodicy which combines appeals to morally significant free will, soul-making, and heavenly bliss. He acknowledges, however, that more work needs to be done on the issue.

We have now arrived at chapter 11, which Trakakis begins by saying: ‘the bold (some would say foolish) thesis I wish to defend in this chapter is that theism is wholly incapable of accounting for any instance, kind, amount, or distribution of natural evil’ (275). On the next page, however, he says that

It seems best to understand Trakakis as believing that by fulfilling the latter aim he provides strong support for the thesis initially stated. In n. 52 to chapter 11, Trakakis says that there ‘may be’ a successful theodicy for one variety of horrendous natural evil: epistemic evil, which consists of great suffering caused by non-culpable human false beliefs.

Trakakis considers and rejects three candidate justifications for God's permitting natural evil. The first is John Hick's soul-making theodicy. The second is Richard Swinburne's view that free will requires justified, true beliefs about what consequences would follow from candidate actions, and that this can exist, together with extensive freedom of choice with respect to good and evil only if there has been natural evil. The third candidate justification is Bruce Reichenbach's view that (possible) worlds in which God continually intervenes to prevent natural evils are not a viable divine option, because rational predictions and decisions would be impossible, while worlds containing natural laws which ruled out natural evil would be worse overall than the actual world, because they could not contain significantly free, conscious and sentient creatures. With respect to the first, for example, Trakakis argues that even granted that evil is necessary for the great good of soul-making, moral evil alone would suffice for this purpose.

Trakakis argues that theodicies can at most be expected to explain why there are the various broad kinds of evil we find in the world, and why there is the amount of evil we find in the world. Nevertheless, if currently available theodicies for horrendous natural evil all fail, then surely it follows that no good state of affairs we know of is such that God's obtaining it would morally justify God's permitting E1, the fawn's suffering. If so, and if either RNA or RNA4 is correct, then does it follow that Rowe has a strong prima facie argument against the existence of God?

At the beginning of chapter 12 Trakakis tells is that ‘gratuitous’ evil is ‘evil that is (roughly speaking) avoidable, pointless, or unnecessary with respect to the fulfilment of God's purposes’ (303). As Trakakis explains, Van Inwagen believes that God has good reasons for allowing people to live in a set-up in which they undergo many particular instances of evil which occur for no reason at all. There is also a narrower sense of ‘gratuitous’ evil, to cover all and only actual instances of evil such that there are no actual instances of good which outweigh the evils, and which God could not bring about without having to permit these instances of evil, or others at least as bad. Rowe's strategy throughout his writings on the problem of evil plainly commits him to the view that if God exists then there are no gratuitous evils in the latter sense. Trakakis ends up saying in his chapter 12 that Rowe can and should abandon this view, while replacing it by one formulated in terms of evil-kinds, and another formulated in terms of amounts of evil (309, 317).

On 334 Trakakis declares that ‘some challenges to Rowe's theological premise were considered in chapter 12, but these were shown to be unsuccessful’. This makes it sound as if in response to the ‘challenges’ Rowe had made only some minor adjustments to the wording of one or more premises. This is far from being the case. Shifting the focus from instances to evil-kinds or amounts of evil would be a dramatic change by Rowe, because it would involve a major reformulation of the arguments of his early and middle periods. Readers are given only a faint hint about how the reformulation would proceed.

This book is detailed, insightful, and clearly written. It will be useful to many philosophers working on the problem of evil. In the rest of this review I will confine myself to two matters bearing strategically on the success of Trakakis's case for his overall conclusion.

The first of these concerns Trakakis's idea that Rowe can and should modify his stance by conceding that even if God exists there may be gratuitous instances of evil, while employing a premise formulated in terms of evil-kinds. Here is a candidate formulation of such a premise: ‘If God exists then each instantiated broadly defined evil-kind must be such that God's permitting there to be instances of that kind is logically necessary for there being outweighing instances of good’. Unfortunately, it is not clear what is to count as a broadly defined evil-kind. Take, for example, throbbing pain. If we can and should abandon the proposition that if God exists there are no gratuitous particular instances of evil, why should we adopt any proposition that commits us to saying that if God exists then God's permitting there to be throbbing pain is logically necessary for there being outweighing instances of good? Why shouldn't we be satisfied with the thought, say, that if God exists then God's permitting there to be very unpleasant sensations is logically necessary for there being outweighing instances of good? Indeed, if we can and should abandon the proposition that if God exists there are no gratuitous evils, then why should it not be the case that God exists and there is natural evil yet there being natural evil is not logically necessary for there being outweighing instances of good? It might be required merely that there is a lot of suffering and deprivation (whatever the causes), along with morally wrong actions and morally bad states of mind. There is a gap in Trakakis's case that he needs to fill.

The second important issue is whether Trakakis has given us good reasons for believing that ‘theism is wholly incapable of accounting for … natural evil’. Even if in chapter 11 Trakakis has refuted the three candidate theodicies examined in the main text (and others dismissed briefly in notes), these candidates are not the only ones advanced recently by reputable philosophers. For example, Richard Swinburne, in chapter 9 of Providence and the Problem of Evil (Oxford: Clarendon Press, 1998), and in chapters 10 and 11 of The Existence of God, 2nd edition (Oxford: Clarendon Press, 2004) has a great deal to say about natural evil that Trakakis does not consider. He might reply that the various greater goods that Swinburne cites – such as compassion and its expressions in co-operative effort in the face of hardship – could be obtained by God's permitting moral evil only. Swinburne anticipates and addresses this thought (Providence and the Problem of Evil, 165–167, and The Existence of God, 142–245). Trakakis needs to say a great deal more in order to provide strong support for his claim that theism cannot account for natural evil.