PURPOSE: The authors examine training in the responsible conduct of research and mentoring in relation to behaviors that may compromise the integrity of science. METHOD: The analysis is based on data from the authors' 2002 national survey of 4,160 early-career and 3,600 midcareer biomedical and social science researchers who received research support from the U.S. National Institutes of Health. The authors used logistic regression analysis to examine associations between receipt of separate or integrated training in research ethics, mentoring related to ethics and in general, and eight categories of ethically problematic behavior. Analyses controlled for gender, type of doctoral degree, international degree, and disciplinary field. RESULTS: Responses were received from 1,479 early-career and 1,768 midcareer scientists, yielding adjusted response rates of 43% and 52%, respectively. Results for early-career researchers: Training in research ethics was positively associated with problematic behavior in the data category. Mentoring related to ethics and research, as well as personal mentoring, decreased the odds of researchers' engaging in problematic behaviors, but mentoring on financial issues and professional survival increased these odds. Results for midcareer researchers: Combined separate and integrated training in research ethics was associated with decreased odds of problematic behavior in the categories of policy, use of funds, and cutting corners. Ethics mentoring was associated with lowered odds of problematic behavior in the policy category. CONCLUSIONS: The effectiveness of training in obviating problematic behavior is called into question. Mentoring has the potential to influence behavior in ways that both increase and decrease the likelihood of problematic behaviors.