Employed data from two longitudinal surveys of gay men in San Francisco (a) to examine for cohort (Study 1) and attrition (Studies 1 and 2) bias effects on reported changes in condom use by gay men and (b) to investigate predictors of condom use (Study 2). Substantial increases in condom use were observed, and these changes were unrelated to attrition and cohort bias. In terms of predictors of condom use, men who always used condoms had higher levels of social support from informal sources of help, had more positive expectations that condoms would have positive interpersonal and personal consequences, and were more likely to be HIV positive than men who used condoms occasionally or never. The results are discussed in terms of their implications for HIV-prevention research.