James Dow, an evolutionary anthropologist, has written a piece of software (called Evogod) that he claims suggest how religion has evolved in society. He makes the starting assumption that there are some people who have a genetic predisposition to pass on 'unverifiable information'. On running his software, this trait didn't do much on its own. He then added in a factor representing non-believers who were attracted to what believers were saying. This led to the spread of the unverifiable (or, 'unreal' information, as Dow also calls it).
I guess the conclusion is that the spread of religion wouldn't have occurred if unbelievers ignored the believers. Dow admits the implications of his software modelling are very tentative and in the early stages. Of course, the assumption that unverifiable means unreal is an enormously contentious assumption! The philosophical naturalism that underlies atheistic conclusions about reality and denies the non-existence of anything empirically unverifiable may be inadequate as more and more science is suggesting.
You can read the whole story here in the New Scientist article.