Twenty-seven school-based studies of psychosocial approaches to smoking prevention are reviewed. Two major approaches are represented: the "social influences" approach and the broader "life/social skills" approaches. The research studies are considered in four "generations": the seminal work by Richard Evans and colleagues at the University of Houston; seven "pilot" studies of improved programs at Stanford, Minnesota, New York, and Washington, with one school or classroom per experimental condition; twelve improved "prototype" studies by these four groups and others, with two or three units randomly assigned to conditions; and six studies in which maximizing internal validity was of prime concern. Reported results were fairly consistent, with each tested program seeming to reduce smoking onset by about 50%. However, none of the pilot or prototype studies considered alone provided easily interpreted results. The major contributions were improved programs and methods. The findings from the fourth generation of studies were more easily interpreted, though only two of them were interpreted with high confidence. It seems that psychosocial approaches to smoking prevention, particularly the social influences approach--fourth generation tests of the broader life/social skills approaches have yet to be reported--are effective, but at this time we know very little about why, for whom, or under what conditions. Suggestions are provided for improved future research.