Okay, I did some more looking at the word on the GSS WORDSUM vocabulary test that I associate with nuns. Earlier, I showed that there was a Catholic effect where Catholics were more likely to know this word than Protestants, and this difference was a good deal larger than the Catholic-Protestant difference for any of the other words. Two more things about nuns: (1) their female and (2) Sister Act aside, nun awareness in our culture was much greater in the past than today. So maybe the nun-effect is really not just a Catholic thing but a female, Catholic, early-cohort thing. The graph below compares female Catholics born before 1941 (the GSS median) to male Protestants born after 1941. The words below are the four most difficult words (in terms of overall % correct) of the entire GSS sample. (Something to keep in mind, incidentally, is that WORDSUM is a multiple choice test with 5 responses, so if everybody guessed randomly there would be 20% correct, although a substantial number of people simply say "don't know" to items). Check out the graph below and see if the comparison for one of the words looks different from the others.
Update: Does this make any difference? If you just looked at WORDSUM straight up and followed slavishly the standards of conventional significance testing, you would conclude that women did better on average than men (p < .01). If you looked at WORDSUM score excluding the nun-item, you would not conclude there was any difference (p = .23, the magnitude of the coefficient reduced by more than half).
Update, 12/20: I was telling my colleague Bob Hauser about this and told me to send the graph to a couple of people, including Mike Hout at Berkeley (who, like me, is an Indiana alum). Hout wrote back to see he too had looked for and found the Catholic effect on cloistered, but not the interaction.