In epidemiologic studies, much of the variation in disease risk estimates associated with occupational pesticide exposure may be due to variation in exposure classification. The authors compared five different methods of using interview information to assess occupational pesticide exposure in a US-Canada case-control study of neuroblastoma (1992-1994). For each method, exposure assignment was compared with that of a reference method, and neuroblastoma effect estimates were calculated. Compared with the reference method, which included a complete review of occupation, industry, job tasks, and exposure-specific activities, the use of occupation-industry groups alone or in combination with general job task information diluted the exposed group by including individuals who were unlikely to have been truly exposed. The effect estimates representing associations between each exposure method and neuroblastoma were different enough to influence the study's conclusions, especially when the exposure was rare (for maternal occupational pesticide exposure, the odds ratio was 0.7 using the reference exposure assessment method and 3.2 using the occupation-industry group exposure assessment method). Exposure-specific questions about work activities can help investigators distinguish truly exposed individuals from those who report exposure but are unlikely to have been exposed above background levels and from those who have not been exposed but are misclassified as exposed because of their employment in an occupation-industry group determined a priori to be exposed.