A study was presented which compared exposure assessment by an expert team of hygienists and chemists with that done using a job exposure matrix (JEM) in a case referent study of cancer in Montreal. Statistical power was assessed on a substance by substance basis for 160 substances. Subjects were male, aged 35 to 70 years and resident in the Montreal area. There were 3730 cancer patients and 533 referents. A checklist of 302 substances was used for exposure assessment, and a total of 14,760 distinct jobs were identified among all study subjects. Statistical power was determined using parameters of sample size, relative risk, exposure prevalence and statistical alpha level. A hypothetical two fold risk was assigned to the team assessment approach and used for comparison with the JEM approach, with two strategies for defining prevalence employed. For the first strategy, which assigned an exposure prevalence of 4%, statistical power for all 160 substances was 0.801 for the team assessment. JEM assessment produced powers of 0.20 to 0.40 for three substances, 0.40 to 0.60 for 78 and 0.60 to 0.80 for 79. This indicated some misclassification. The second strategy used observed exposure prevalence estimates, and statistical power distributions reflected true prevalence variation as well as misclassification. For both strategies, the loss of power in using the JEM approach was not very large for many substances. The authors conclude that there is considerable loss of statistical power in going from a team assessment approach to the JEM approach for exposure assessment, and they recommend using the data generated for these 160 substances as a guideline for determining feasibility of the JEM approach for a given substance.