This paper reports the results of a study using cognitive theory to assess an establishment mail survey requesting information on employee "non-wage" compensation, such as bonuses, awards, and commissions. Two methods of evaluation are used in the study, a survey pretest and a Response-Analysis Survey (RAS). Both methods identify problems with respondent comprehension in understanding survey definitions, terms, and instructions. Each method also identifies difficulties with and a lack of establishment records data necessary to complete the survey. These problems lead to omissions in the type and values of requested payments. Recommendations for improvement are discussed based upon information processing theory, focusing on respondent comprehension, retrieval, and respondent motivation.