INTRODUCTION Globally, population surveys on HIV/AIDS and other sensitive topics have been using audio computer-assisted self interview for many years. This interview technique, however, is still new to Vietnam and little is known about its application and impact in general population surveys. One plausible hypothesis is that residents of Vietnam interviewed using this technique may provide a higher response rate and be more willing to reveal their true behaviors than if interviewed with traditional methods. OBJECTIVE This study aims to compare audio computer-assisted self interview with traditional face-to-face personal interview and self-administered interview with regard to rates of refusal and affirmative responses to questions on sensitive topics related to HIV/AIDS. METHODS In June 2010, a randomized study was conducted in three cities (Ha Noi, Da Nan and Can Tho), using a sample of 4049 residents aged 15 to 49 years. Respondents were randomly assigned to one of three interviewing methods: audio computer-assisted self interview, personal face-to-face interview, and self-administered paper interview. Instead of providing answers directly to interviewer questions as with traditional methods, audio computer-assisted self-interview respondents read the questions displayed on a laptop screen, while listening to the questions through audio headphones, then entered responses using a laptop keyboard. A MySQL database was used for data management and SPSS statistical package version 18 used for data analysis with bivariate and multivariate statistical techniques. Rates of high risk behaviors and mean values of continuous variables were compared for the three data collection methods. RESULTS Audio computer-assisted self interview showed advantages over comparison techniques, achieving lower refusal rates and reporting higher prevalence of some sensitive and risk behaviors (perhaps indication of more truthful answers). Premarital sex was reported by 20.4% in the audio computer-assisted self-interview survey group, versus 11.4% in the face-to-face group and 11.1% in the self-administered paper questionnaire group. The pattern was consistent for both male and female respondents and in both urban and rural settings. Men in the audio computer-assisted self-interview group also reported higher levels of high-risk sexual behavior-such as sex with sex workers and a higher average number of sexual partners-than did women in the same group. Importantly, item refusal rates on sensitive topics tended to be lower with audio computer-assisted self interview than with the other two methods. CONCLUSIONS Combined with existing data from other countries and previous studies in Vietnam, these findings suggest that researchers should consider using audio computer-assisted self interview for future studies of sensitive and stigmatized topics, especially for men. KEYWORDS Behavioral research, community surveys, public health surveillance/methods, survey methods, effect modifier, epidemiologic biases, social desirability, HIV/AIDS, Vietnam.