The authors surveyed 2,010 biomedical trainees in the fall of 1990 at the University of California, San Diego, regarding their perceptions about unethical practices in research and the extent of their training exposure to the ethics of scientific investigation; 549 responded, representing both clinical and basic science departments and including graduate students and postdoctoral fellows in addition to medical students, residents, and fellows. Of the 549 trainees, 129 (23%) responded that they had received no training in research ethics; 195 (36%), that they had observed some kind of scientific misconduct (although not necessarily in the sense of research fraud defined in federal regulations); and 81 (15%), that they would be willing to select, omit, or fabricate data to win a grant or publish a paper. The trainees planning an academic career were more likely to report having been aware of others' scientific misconduct. Reported exposure to ethics training was not associated with a difference in past or potential unethical behavior. The authors conclude that while the apparent ineffectiveness of past ethics instruction does not preclude the possibility that more systematic training may be useful, it does underscore the need to assess the efficacy of training activities.