Between 1986 and 1993, a program of questionnaire design and cognitive research was conducted by the Census Bureau and Bureau of Labor Statistics, to improve labor force measurements in the Current Population Survey. As part of the research program, diagnostic measures for systematically testing and evaluating alternative questionnaire versions were developed and applied. This paper reports results of applying two methods, special follow-up probes and hypothetical vignettes, to the measurement of "work" in the CPS. These measures provided both direct and indirect information about problems of respondent comprehension and reporting errors. In this paper we analyze results using these diagnostic measures to evaluate the effect of questionnaire revisions on reporting of work activities, and we assess the consistency and usefulness of the information provided by alternative diagnostic measures for pretesting and selecting questions.