Atran describes religion as (1) a community's costly and hard-to-fake commitment (2) to a counterfactual and counterintuitive world of supernatural agents (3) who master people's existential anxieties, such as death and deception. Later in the book, he adds that 1, 2, & 3 lead to (4) they demand ritualistic & rhythmic co-ordination of 1, 2, & 3, such as "communion". He later describes religion (paraphrased) as a thought process which involves the exaggerated use of everyday cognitive processes to produce unreal worlds that easily attract attention, are readily memorable, and are subject to cultural transmission, selection, and survival.
Atran's premise and aim are clear: since religion so described is, cognitively speaking, false, the scientist's job is to explain why it's so powerful and ubitquitous, so that we're better equipped to think about how to supplant it with something that helps us deal with reality just as well and also has the merit of being true. The book has gotten a good deal of press; see Time Magazine's recent cover article The God Gene, which cites Atran and several other authors in a similar vein. All very good for business. But what I want to know is: why should anybody care anymore about this sort of book?
Starting with Lucretius and Democritus in the ancient world, moving through the 19th-century work of Ludwig Feuerbach, and continuing today with the crude polemics of Richard Dawkins, thinkers who exalt natural science as the normative template for all human knowledge have always sought to Bulverize religion by explaining it away in non-religious terms. The difficulty is that the entire exercise is not only dubious by the very epistemic criteria it favors but also question-begging by any broader standard.
It's scientifically dubious because there is no one theory that is both testable in a scientfically respectable sense and sufficient to exclude other scientific explanations. Different thinkers have "explained religion away" in different terms—economic (Marx), psychological (Freud), evolutionary (Atran and Dawkins), and it isn't clear how to sort out the wheat from the chaff. But from the looks of the whole thing, it's not important to today's scientific Bulverizers that any one theory be correct. The important thing is the convergence of otherwise competing theories on the one conclusion they want.
That should tell us right there that we're dealing not so much with science as with metaphysics. It doesn't matter that one can't test these theories the way one can test those of subatomic physics or even of medicine. What matters is that some-or-other version be a plausible corollary of the secular worldview according to which religion, in the supernatural sense, is illusory. That shows that the fundamental question is being begged. If one assumes that supernatural religion is illusory, then it makes sense to assume further that something natural will eventually explain why the illusion is so persistent and extensive. But of course, the defenders of supernatural religion could just as well rejoin that the natural propensity of humans toward supernatural religion is evidence that the natural manifests and points to the supernatural—if, that is, some-or-other version of supernatural religion is true. Both assumptions beg the question if and when made in debate.
If either side wishes to avoid begging the question, what they really need to do is show how, within their own respective worldviews, some proposed explanation of religiosity makes more sense of it than any other explanation. That would require more work and more fairness than polemicists on the scientistic side, at any rate, are likely to show. It is up to defenders of supernatural theism to do better. I haven't studied yet whether they have.