Exactly 20 years ago, I organized a conference that brought together researchers involved in longitudinal studies on drug use and statistical experts in longitudinal data analysis (Kandel, 1978a). My goal was to assess the state of knowledge regarding the predictors and consequences of drug use and the best methods to identify these factors. Richard Jessor was a key participant at the meeting and reported on his classic panel study of drinking, marijuana use, and other problem behaviors among high school and college students (Jessor & Jessor, 1977). The 1970s represented a very exciting period for drug researchers. A number of parallel longitudinal studies had been initiated using comparable samples, methods, and designs. These studies provided unusual opportunities for replication that are all too rare in the social sciences. In an introductory chapter in the volume published after the conference, I summarized in the form of 19 propositions our knowledge regarding the predictors and consequences of drug use (Kandel, 1978b). The title of the chapter, “Convergences in Prospective Longitudinal Surveys of Drug Use in Normal Populations,” reflected the commonality of findings that emerged from the various research groups. The chapter provides a useful benchmark from which to assess how the field has progressed in the intervening two decades. In this chapter, I briefly highlight major new advances. Although the emphasis is on studies based on representative adolescent population samples, epidemiologica research on adults and animal laboratory research are discussed selectively when this work informs on substance abuse in adolescence.