Since accurate information on sensitive issues is difficult to obtain from persons who might have been involved in a crime, without violating their legal rights, a random response technique has been developed to overcome interview problems related to self-confession. Basically, the technique requires the interviewee to respond, on a prior fixed random basis, to one part of a two-part question. The question contains a 'sensitive' element and a nonsensitive element. The interviewer does not know to which part of the question the interviewee is responding, but the analysis of responses from several respondents will provide a good estimate of the 'sensitive' data values desired. The procedure also avoids the problems of confidentiality of information, since there is no way to link a specific response to a sensitive question with a specific person should the data be required by a court. The study uses an accurate standard (police records of arrests) to compare this technique with the more traditional direct question technique. The random response approach is concluded to be more accurate than the traditional approach; although for a given level of accuracy, the technique requires four times as many respondents. Attached are interview schedules and materials, the interviewer instruction manual, and the codebook for data set. Five notes and 43 references are listed.