Survey analysts frequently know little about the validity of their questions. They often have no direct evidence that respondents understand the questions being posed and that shared understanding exists between the analyst and the respondent. Using the ‘random probe’ technique developed by Howard Schuman, about 30 questions from the 1984 General Social Survey are examined. Overall there appears to be few validity problems with these questions, but some alternative understandings and problems are examined. The random probes also can be used to gain substantive insight into the issue under investigation and their value in this area is illustrated.