Measurements of drug use and other illicit or stigmatized behaviors are subject to nontrivial underreporting biases. During in-person surveys, respondents are more likely to report such behaviors when interviewed using techniques that maximize interviewee privacy, e.g., use of paper SAQs and audio-CASI rather than questioning by human interviewers. Until recently, respondents in telephone surveys could not be offered similar privacy. A new technology, telephone audio computer-assisted self-interviewing (T-ACASI) overcomes this limitation of telephone surveys by allowing respondents to respond to a computer. A randomized experimental test of T-ACASI was embedded in the Urban Men's Health Study (UMHS). UMHS surveyed a probability sample of 2,881 men from four United States cities and who reported having sex with men. Respondents interviewed using T-ACASI reported a higher prevalence of drug use and drug-related behaviors than respondents interviewed by human interviewers. However, survey respondents were more likely to break off an interview when the interview was conducted by a T-ACASI computer rather than by a human interviewer.