Using multiple classification models applied to self-report data on initiation of drug use from nine National Household Surveys on Drug Abuse conducted between 1982 and 1995, this paper shows that the directions of change in period and cohort effects were similar for marijuana and alcohol and for males and females. Period effects--indicative and societal tolerance or support for drug use during 5-year periods between 1961 and 1990--declined between the early 1970s and late 1980s, while cohort effects--indicative of early experiences of birth cohorts favoring drug use--increased. One interpretation is that trends in incidence were determined by two opposing vectors of social forces: Beginning in the 1970s, changes in social policies, values, and drug markets--as reflected in period effects--increasingly acted to reduce incidence, while changes in conditions of childhood socialization--as reflected in cohort effects--increasingly facilitated or encouraged incidence. Especially for marijuana, the increase in cohort effects is larger among females, which gives rise to gender convergence--approximately equal male and female incidence rates for both drugs--by the late 1980s. An innovative method of the paper is the adjustment of incidence rates for reporting bias.