Large-scale surveys increasingly inquire into health-related behaviors such as sexual activity, drug use, and alcohol intake. Many of these behaviors are regarded as sensitive by potential respondents, sometimes leading to refusal to participate in the survey and sometimes to underreporting of what are regarded as undesirable behaviors. As a result, the issue of how to elicit truthful reports of such behaviors has assumed increasing importance. The relatively recent development of audio computer-assisted self-interviewing has received much attention, and this mode of administration offers potential gains in the reporting of a variety of sensitive behaviors and attitudes. A laboratory-based experiment was conducted in an effort to understand the mechanisms through which audio-CASI may improve reporting of socially sensitive information. While this lacks the external validity of an in-home survey, it allows us to isolate the factors that may produce the effects of interest.