We examined follow-up data from surveys in 1988, 1992 and 1994 in order to estimate the prevalence and explore the correlates of retest artifact (denial) of drug use among National Longitudinal Survey of Youth respondents who disclosed lifetime cocaine or marijuana use in 1984. In the cocaine use cohort, 42% denied lifetime cocaine use during at least one follow-up wave. In the marijuana use cohort, about 29% denied lifetime marijuana use during at least one follow-up wave. Denial either leveled off (cocaine) or diminished (marijuana) between the second and third follow-up interviews. The most consistent predictors of denial in both longitudinal and cross-sectional models and across substances were race/ethnicity (black informants had increased rates of denial) and marital status (married respondents had increased rates of denial). Other predictors of denial included interviewer characteristics (social attribution), interview mode, and drug salience. The findings with respect to marijuana reporting trends parallel increased willingness of public officials to retrospectively disclose this behavior in the popular press.