A subset of questions from the 1992 Texas Crime Poll, a statewide poll conducted annually by mail, was replicated in telephone interviews that used the Computer Assisted Telephone Interviewing System. The phone survey yielded better participation rates but less complete responses to individual attitude questions than did the mail poll. As expected, the mail survey was less expensive but less efficient than the automated phone survey. The central finding was that all but one of the responses to five attitude questions differed significantly across the surveys. The samples differed in their demographic composition, but this did not explain differences in the substantive findings from the mail and phone surveys. In the context of this study and its findings, the presence of a telephone interviewer may have tempered the responses by phone. This would explain why the phone survey yielded more moderate views on criminal justice than did the self-administered mail questionnaire, which was completed in relative privacy and anonymity. The better completion rate for individual questions by mail is consistent with this speculation. Alternatively, the self-selection of keenly interested respondents to the mail survey may have influenced the observed difference in attitudinal responses across surveys. Directions for further study are discussed. 6 tables and 31 references