Over 90 percent of the sample, including nonusers, soft drug users, opioid users, and injectors, were obtained via the same "snowball" of friends and acquaintances. An extensive, structured instrument was used, including pre-existing personality scales, attitudinal information, sociodemographic information, details of family and friends and full details of drug use and crime. Results show that drug use was not a major unique predictor of theft, but it did predict other crimes. However, it did not appear to be the sole and sufficient explanation of any kind of crime, except perhaps fraud. Rather, crime frequency was determined by a combination of friends' behavior, prior criminal experience, and drug use. 6 tables and 28 references (Author abstract modified)