This paper examines respondent reactions to and performance on a CASI (computer-assisted selfinterview) portion of a CAPI (computer-assisted personal interview) survey. We first examine whether there are systematic differences between those who choose to do CASI themselves and those who seek the aid of the interviewer in completing these items. We then explore whether the decision to do CASI has any impact on the quality of data collected. This study appears to be virtually unique in the respect that a record was kept of whether respondents actually completed the CASI items themselves, or had the interviewer assist them. We found that 21% of respondents used some form of interviewer assistance in completing the self-administered items. In most other studies including self-administered portions (whether computer-assisted or paper-and-pencil), there is little discussion of this issue (see for example, Jobe et al., 1994; O'Reilly et aL, 1994; Tumer, Lessler and Devore, 1992). We infer from this that one of three things may have occurred: (a) respondents were pressured into completing the items themselves, (b) interviewers assisted respondents, or (c) these cases were treated as nonresponding units. We suspect that the second option may be likely in many surveys containing selfadministered components. Given that interviewer administration may defeat the purpose of self completion (increased privacy leading to more truthful reporting of highly sensitive behavior), it is important to examine the extent to which this might be happening. The Self Portraits Study, the subject of the present paper, also differed from other self-administered surveys in two other respects. First, many self-administered modules of surveys are designed to elicit reports of highly sensitive and sometimes illegal behaviors, such as drug use and high-risk sexual practices, whereas the Self Portraits CASI questions were of a more general attitudinal nature. Second, many of the CASI applications to date have been on surveys of younger persons, a group that may be regarded as comfortable, or at least familiar, with computer technology. Self Portraits, in contrast, included older persons. Although these differences may limit the generalizability of these results to other CASI studies, they also facilitate certain analyses which might otherwise not be possible. Specifically, we can examine respondent preferences (as evidenced by their behavior) for self-completion versus interviewer-completion of CASI items.