In a prospective design, it was predicted that the role of negative mood in precipitating relapses and crises among alcohol-dependent participants would be greater when assessed with retrospective mood ratings than when assessed by prospective mood ratings because of participants' attributional biases. This hypothesis was not supported. Support was found, however, for a relationship between negative bias and the amount of negative mood assessed at the time the retrospective report was given. As has been previously found, the most frequent precipitant of relapses and crises in this study was negative emotional states. Female participants were more likely to report interpersonal and less likely to report intrapersonal determinants than male participants. Minor relapses were more likely to be precipitated by social pressure and less likely to be related to negative emotions than major relapses.