Three split sample comparisons of interviews using laptop computers and traditional pencil and paper methods were carried out on surveys consisting largely of attitude questions. Two of the experiments were conducted on waves two and three of a panel survey; the third used a fresh sample of respondents. CAPI was found to be acceptable to both interviewers and respondents; no differences in either initial response rates or response following a computer interview were found. The increased length of CAPI interviews on two of the studies was attributed to interviewers' lack of experience. Comparison of responses to attitude questions showed no difference in use of “don't know” or midpoint options, but CAPI respondents were more likely to choose the extreme responses of five and eleven point scales. There was some evidence that CAPI might improve stability of responses over time and none that the technique would reduce it.