Many epidemiologic studies of the relationship between occupation and disease depend on job histories obtained by interview from study subjects. A validation study was undertaken to determine the accuracy of job histories obtained by interview, and to examine whether certain characteristics of respondents and of the study design influence reporting accuracy. For 297 subjects interviewed in Montreal between 1979 and 1981 in the context of a case-control study on occupational factors in cancer, it was possible to compare the names of reported employers with those recorded in the data bank of the government-run quasi-universal Pension Plan. The comparison was carried out year by year for the 13-year period 1966-1978. For the 13 X 297 person-years studied, 82.0% of reports agreed with the records. The extent of concordance did not differ substantially between subgroups defined by age, education level, or social class, nor was there a measurable difference in the degree of concordance between the first half of the 13-year period and the second half. There was some evidence that the three interviewers obtained job histories of varying quality.