By the way, nimh, I am curious whether you have run across this in polling:
(Q) Would you consider not voting for Obama because he is black?
(A) I am not racist. It makes no difference that he is black.
(Q) Do you know people (family, neighbors co-workers) who would vote against Obama because he is black?
(A) Oh, yes.
From what I understand it's a way for pollsters to get round any social desirability bias (or whatever the official term for that is) in the responses they get.
Even when talking anonymously to a pollster on the phone, people tend to adjust their answers somewhat toward what they consider the more socially desirable answer. The bigger the taboo, the more people will fudge about their real feelings about it, basically, and instead answer something more like what they think other people would want them to answer.
Eg, if a polling person calls you on the phone and asks, have you ever fantasised about raping goats, or something, you are quite likely to answer "no" even if you have - and even though you know you're talking anonymously.
(An interesting question is whether pollsters like Survey USA who use automated calling (i.e., you dont talk to a live person but to a computer) get around that somewhat, by taking out the human factor from the equation.)
Anyway, admitting that you wont vote for a black person is by now rather taboo in most parts of the country. So the pollster is less likely to get an honest answer on questions like, "would you be prepared to vote for a black candidate". Something they use to get around that and get a more honest assessment of how widespread such reluctance is, is to ask the respondent, "do you know anyone who would not vote for a black person", or something of the sort.