Over the study period, significant increases in actual use and social context of use were evident, with 15 percent of 17- to 19-year-olds reporting they had been 'super high or stoned' on cocaine at least once. Path analytic models were used to examine the impact of intentions and social context variables (e.g., peer drug use) on cocaine use over time, as well as to study the effects of earlier use of alcohol and cannabis on subsequent use of cocaine. Self-reported cocaine use was highly predictive of the use of many other drug substances. Early cocaine use had no influence on later variables, including self-reported use. However, intention to use cocaine was significantly correlated with current self-use and significantly predictive of later use by self and peer and adult models. Peer involvement (as models and providers) and self-use in year 4 were significantly related to the frequency of use in year 5. Those who expressed an intention to use cocaine were more pleased with their appearance, less ambitious, less law-abiding, more liberal, and less religious than those with no intention to use cocaine. Psychopathology was not predictive of or influenced by use, but depression was associated with increased use over time. Income from salaries or gifts was predictive of current cocaine use to a small but significant degree. 2 tables, 2 figures, and 25 references. (Author abstract modified)