According to this brain-imaging study, volunteers presented with negative scenarios (i.e. car crashes, cancer), and asked to estimate the probability of these scenarios happening to them, would only update their beliefs if the actual rate of ocurrence in the population, given to them afterwards, was lower, i.e. more optimistic, than what they had guessed. The more "optimistic" the subjects were, according to a personality test, the less likely they were to update their belief based on more negative information, and the less activity they showed their frontal lobes, indicating that they weren't "paying attention" to the new information.

Sounds like confirmation bias, except that interestingly enough, it's unidirectional in this case. I wonder if very pessimistic people would have the opposite bias, only updating their estimate if the actual probability was higher, or more negative.

Link to article on kurzweilai.

Link to abstract in Nature journal. I can't access the full text.