In a study of occupational causes of spontaneous abortions, based upon self-reported data and data from the hospital registry, evidence of differential misclassification was noted. Among those exposed a larger proportion of the self-reported spontaneous abortions were identified in the hospital registry, compared with what was found in the control group. This could be due to recall bias of the questionnaire data masking an effect of exposure, or a lower threshold for hospitalization among those exposed vis-à-vis controls, which would exaggerate the effect of exposure, if any. The analysis tended to support the idea of a less accurate recall of spontaneous abortions among controls, especially for abortion that occurred more than 3 years before the questionnaires were sent out. A second questionnaire was sent out to a subset of the participants 3 1/2 years after the first questionnaire. 17% reported fewer spontaneous abortions in this second questionnaire compared with the situation in the first questionnaire, for the period 1973 to 1980.