I appreciate Jason Marsh's careful attentionFootnote ¹ to my article ‘Divine hiddenness and the demographics of theism’.Footnote ² As Marsh himself suspects, his Molinist reasoning hadn't occurred to me when I suggested in that article that data reporting the uneven worldwide distribution of theistic belief ‘confound theistic explanations of non-belief in God’. What I meant, of course – but perhaps should have made more explicit – is that the data confound plausible theistic explanations of non-belief in God, since it goes without saying that not all of the infinitely many explanations consistent with a given set of data are plausible explanations of it. Implausible explanations consistent with the data are always plentiful, but we don't let them detain us.

For reasons I'll now give, Marsh's discussion does not change my mind about the lack of theistic explanations that, even in theistic terms, plausibly explain the data. Still less do his Molinist assumptions show that theistic explanations of the data are as plausible, even on their own terms, as naturalistic ones. Whether Marsh wants to claim that they show this or, instead, that they merely open up a potential line of theistic explanation is not clear to me. My aim is to make clear why they won't in fact help the theist.

Why does a population of millions of non-theists persist in Thailand but not in Saudi Arabia? In my article I argued that, even without assuming the truth of naturalism, standard theistic answers to this question are less plausible than standard naturalistic answers: even judged on their own terms, theistic explanations of the geographic lopsidedness of belief look far-fetched compared to naturalistic explanations judged on their own terms. In my canvassing of the theistic options, I did allow the bare possibility that the residents of Thailand, unlike the residents of Saudi Arabia, are especially hampered by epistemic or spiritual defects that explain their failure to believe in the monotheistic God. But I admit I did not regard this possibility as at all likely or as one to take seriously.

Marsh's explanation – his theodicy, really – takes it seriously indeed. The explanation is somewhat complicated, but I think it can be summarized. On Marsh's proposal, there are some people who would refuse to love God during their earthly lives no matter how obvious God made HisFootnote ³ existence and love for them. God therefore uses a ‘grouping strategy’ to save these defective souls, as it were, from themselves: out of benevolence for them, God segregates defective souls among other defective souls, or at least among other non-believers, ‘to keep these individuals innocent for a later time, when they will be in a position truly to love God’ (468). Placing them mainly among other non-believers, God preserves their innocence by removing, or reducing, their ‘opportunity to believe’ in and come to love God, an opportunity they would otherwise become blameworthy in God's eyes for failing to seize (467).Footnote ⁴

As often happens with theodicies, however, Marsh's explanation gives rise to at least as many problems as it solves. I'll quote the two claims on which, as Marsh concedes, his explanation depends (467):

1. (1) There are some persons who would refuse to love God no matter what geographic circumstance they found themselves in.

2. (2) God has middle knowledge.

Claim (2) is supposed to imply that God already knows exactly which persons are referred to in claim (1) even before He freely chooses to create them. A major problem therefore arises right away: these two claims have a chance of explaining the uneven distribution of believers only on an implausible and heterodox view of God's power as the Creator. For surely God did not find Himself stuck with a pre-existing set of people, and then have to decide how to handle the subset of them that He knows will refuse to love Him no matter where He puts them. On the contrary, says orthodox theism, God gets to decide which people, if any, ever exist – and He gets to make that decision even if He confronts an independently existing array of human essences from which He must choose when creating persons.Footnote ⁵

Why, then, should we think that God, when choosing from that presumably infinite array, has to instantiate what I'll call ‘stubborn essences’, essences whose instantiations He knows will refuse to love Him during their natural lives no matter where on earth He instantiates them? (467).Footnote ⁶ Clearly not all essences are stubborn, since clearly not all instantiated essences refuse to love God: according to the major theistic traditions, many human beings have accepted, during their natural lives, God's offer of a loving relationship. In that case, why must God instantiate what appear to be millions or even billions of stubborn essences? Marsh's theodicy invites this question but doesn't answer it.

Even if it should turn out that all human essences suffer from the defect that Alvin Plantinga calls ‘transworld depravity’Footnote ⁷ – a state of affairs for which Plantinga never claims more than logical possibility – it remains obvious that not all human essences suffer from what I'm calling ‘stubbornness’. Again, I've never alleged the logical incompatibility of theism and the lopsided distribution of theistic belief we observe in the world, so it's not relevant whether every human essence might have been stubborn; what is relevant is that clearly not every human essence is stubborn. Why, then, was God bound to instantiate any stubborn essences at all, let alone millions or billions of them?

Marsh's Molinist account confronts a second problem. Naturalistic explanations, such as those offered by social science, cite the undeniable influence of upbringing to explain why children tend to share the religious attitudes of their parents. According to these explanations, children don't come predisposed to hold any particular religious attitudes, but they do come generally predisposed to adopt the attitudes of their parents and to regard their parents as authorities in general. Hence, these explanations predict a high correlation between the religious attitudes, if any, displayed by parents and those displayed by their children, which is exactly what we find.

By contrast, Marsh's account of the grouping strategy rejects this plausible and highly typical causal link in favour of a less plausible one: God causes stubborn essences to be instantiated as the offspring of people who, as it happens, also have stubborn essences. For Marsh, indifference or aversion to the monotheistic God isn't passed culturally from parents to children, as naturalistic explanations claim; rather, the sharing of such indifference or aversion by parents and children results from a common cause, namely, God's prior decision to keep stubbornness ‘in the family’. I can't see how this idea is any more plausible than the notion that smoking doesn't cause lung cancer but instead some unidentified gene causes both lung cancer and the desire to smoke. In some cases of steady correlation between A and B, it is more plausible to conclude that A causes B than to conclude that A and B result from a common cause – especially when we have an obvious candidate for the causal mechanism from A to B, as we do in the case of parents who pass their religion on to their children. Presumably Marsh acknowledges the causal role of parents in transmitting their culture, including the religious beliefs of their culture; in that case it seems like special pleading to deny the causal role of parents in transmitting indifference or aversion to God.

Arrayed against Marsh's claim (1) is social-scientific evidence that human attitudes toward the divine, like other cultural attitudes, are highly plastic and subject to local conditions. To be fair, however, Marsh appears at least somewhat ready to concede such plasticity. In note 5 of his article, he considers a weaker version of (1) that he suspects might suffice for Molinist purposes: (1*) ‘for each individual there is some world in which she would have come to believe in and love God short of experiencing the divine directly, but … these worlds [are] relatively few in number and … none of them [contain] more good-making properties overall than the actual world’ (471). Again, though, why suppose that omnipotent God has to instantiate any, let alone many, of these ‘relatively stubborn’ essences in the first place? Maybe it's logically possible that an omnipotent God found Himself bound to instantiate essences fitting the description in (1*), but we need more than logical possibility in order to ground a plausible explanation of the demographic data.

I'll conclude by commenting briefly on two other questionable claims in Marsh's presentation. First, as Marsh notes, it would seem that God has not done everything possible to safeguard the innocence of those who have stubborn essences, since few if any major cultures in our world are totally devoid of people who claim to believe in and love the monotheistic God. Why would God risk making people with stubborn essences more blameworthy for their non-belief by putting theistic believers in their midst? Marsh answers that a mix of believers and non-believers encourages and sustains belief itself: ‘it seems likely that a certain amount of non-belief in various populations would function to encourage and maintain belief in those populations, such that without it there would have been less, perhaps much less, by way of genuine belief in the world’ (468).

I'm not sure what work the word ‘genuine’ is doing in that sentence, but in any case I see no reason at all to accept Marsh's claim. Whatever the merits of the notion that a pluralistic ‘marketplace of ideas’ helps reveal the truth on some issue, there is no reason to think that the incidence of belief as such benefits from the presence of non-belief. Notoriously, the acceptance of (for instance) well-confirmed scientific theories by people living in modern technological societies has been anything but helped by the presence in those societies of anti-scientific and pseudo-scientific beliefs.Footnote ⁸ Likewise, the incidence and strength of theistic belief in, say, tribal Afghanistan has surely been helped, not hurt, by the near-total absence of atheists there.

Second, in responding to the first of two main objections that he considers in his article, Marsh says that ‘we should never assume of any particular non-believer or group of non-believers that they will remain non-believers throughout their natural lives: we simply lack a God's middle knowledge perspective on these matters’ (469). On this view, apparently, we can't reasonably assume that a staunch atheist like Richard Dawkins won't get religion on his deathbed, or that the Buddhist population of Thailand won't convert entirely to Christianity in the current generation. But such a view seems to me implausibly sceptical. Even if we can't be certain such events won't occur, we are epistemically well-justified in assuming they won't.

Marsh concludes by emphasizing that the demographic problem I identify doesn't show theistic belief to be irrational unless it outweighs ‘all of the inferential and non-inferential grounds in favour of belief in God’ (470). Of course it doesn't, but I never claimed otherwise. All I claim is that the lopsided distribution of theistic belief is less surprising on naturalistic explanations than on theistic ones, a claim that Marsh's theodicy doesn't seem to me to threaten. In fact, it's not clear that my demographic argument needs to rely on a hypothesis even as strong as naturalism. Consider instead a ‘hypothesis of indifference’ that is logically weaker than naturalism but just as unfriendly to theism: ‘There are no gods who care about human welfare’.Footnote ⁹ On Bayesian grounds, the lopsided distribution of theistic belief may confirm that atheistic hypothesis even more than it confirms naturalism: the former hypothesis, being logically weaker, possesses a higher prior probability – and therefore a higher posterior probability on evidence they equally predict. In any case, that lopsided distribution gives theism a lower posterior probability than either of the atheistic hypotheses.

For the sake of argument, I haven't objected to the use that Marsh's explanation makes of the contentious notion of Molinist middle knowledge, although I believe that it diminishes all by itself the plausibility of that explanation. I've concentrated instead on implausible and heterodox features of his explanation even if we grant the possibility of middle knowledge.Footnote ¹⁰