Assessments of public perceptions of the characteristics of a nuclear power plant accident and affective responses to its likelihood were conducted 5 months before and 1 month after the Chernobyl accident. Analyses of data from 69 residents of southwestern Washington showed significant test-retest correlations for only 10 of 18 variables--accident likelihood, three measures of impact characteristics, three measures of affective reactions, and hazard knowledge by governmental sources. Of these variables, only two had significant changes in mean ratings; frequency of thought and frequency of discussion about a nearby nuclear power plant both increased. While there were significant changes only for two personal consequences (expectations of cancer and genetic effects), both of these decreased. The results of this study indicate that more attention should be given to assessing the stability of risk perceptions over time. Moreover, the data demonstrate that experience with a major accident can actually decrease rather than increase perceptions of threat.