A review of 30 AIDS behavioral research studies revealed that almost half employed retrospective sexual activity assessment periods of 12 months or longer. The present study examined the correspondence between retrospective reports of sexual activity for 3 overlapping time-frames (past 2 weeks, past 3 months, and past 12 months) for 61 gay men. Adjusted for equivalent 12-month intervals, the reported mean frequencies of sexual practices were considerably higher for shorter recall periods than for longer recall periods, a discrepancy greater than would be expected by normal activity fluctuation. The study also found that behavior reported across shorter retrospective time-frames was more consistent, while reliability for frequently occurring and unsafe sexual activities decreased as the recall period lengthened. Studies utilizing recall periods of 12 months or more may produce data of questionable reliability.