ITHACA, NY: Ernie Pyle – an iconic war correspondent in World War II – reportedly said "There are no atheists in foxholes." A new joint study between two brothers at Cornell and Virginia Wesleyan found that only part of this is true.
A recent analysis of archived World War II surveys of Army Infantry after a battle showed a soldier's reliance on prayer rose from 32% to 74% as the battle intensified.
In the USA right now in peace time only around 3% of people self identify as atheists (and a similar number as agnostics). If the 26% who didn't rely on prayer during intense battle are counted as atheists in foxholes, then the proportion in foxholes is roughly 9 times the proportion in the general population.
But of course non-atheists don't pray constantly or even, necessarily, ever rely on prayer.
The rest of the article reminds us that rejecting the alternative is not the only problem people can have with statistics. Point estimates without standard errors can also be confusing, in particular, when the sample is small, but it isn't clear how small.
"The question is whether that reliance on faith lasts over time," said Craig Wansink, author and Professor of Religion at Virginia Wesleyan College.
To determine this, a second study of 1123 WWII veterans showed that 50 or more years after combat, most soldiers still exhibited religious behavior, but it varied by their war experience. Those facing heavy combat (versus no combat) attended church 21% more often if they claimed their war experience was negative, but those who claimed their experience was positive attended 26% less often. The more a veteran disliked the war, the more religious they were 50 years later.
The self-funded findings, forthcoming in the Journal of Religion and Health, note that no causality is assumed. "We can't claim, for instance, that combat made soldiers religious or, conversely, that religious soldiers hated combat," said Brian Wansink, study co-author and Professor of Applied Economics and Management at Cornell University.
Also, I really can't doubt that the simple heavy combat vs no combat comparison wasn't tried first (it clearly follows from the earlier study). The 21% ad -26% calculations are two of at least three and possibly many. p-levels should be corrected based on the number of hypotheses tested with the safe Bonferroni test of the null the lowest p-value times the number of tests against different alternatives.