This work aimed at assessing the validity of job exposure matrix (JEM) for the retrospective evaluation of exposure to polycyclic aromatic hydrocarbons (PAH) within the framework of population-based case-control studies, taking the evaluation of industrial hygiene experts as reference. For this purpose, we used a case-control study for which the different levels of exposure were assessed by such experts after case by case evaluation of all job periods reported by the subjects. The JEM was applied to this set of data so that we had, according to job periods, the experts' evaluation on the one hand, and the JEM evaluation on the other. JEM sensitivity and specificity of the matrix vary widely from 0.13 to 0.96 and 0.58 to 0.99 respectively, depending on whether the experts chose a narrow or wide definition of exposure and on the cutoff point chosen to dichotomize the JEM. We also computed, according to the sensitivity and specificity of the JEM, the odds ratio (OR) and relative efficiency (RE) given by the JEM for several hypothetical OR and frequencies of exposure among the controls. These calculations were made for different definitions of exposure by the experts and different cutoff points for the JEM. The results show a bias in the JEM's evaluation of the OR. In addition, the RE varies widely from very low values to high values (0.05-0.45) depending on the experts' definition of exposure and the cutoff point chosen for the matrix. Note, however, that all these calculations were made taking the experts' evaluation as the reference.(ABSTRACT TRUNCATED AT 250 WORDS)