I just found it and it's mostly what I was remembering. Here's the relevant bit: "She then ran an experiment with Christians and non-Christians volunteers who were asked to solve problems of logic. Christian volunteers reminded about the stereotype that Christianity and science don't mix did worse than Christian volunteers who weren't reminded of the stereotype. This, of course, is classic stereotype threat. She also found that when she gave volunteers a test, Christians did worse when she labeled the test as a test of scientific ability, rather than a test of intuitive ability, even though the test was exactly the same."
The whole thing: