Data from the Montreal survey on occupational factors in pregnancy were used to test the hypothesis that visual display units (VDUs) constitute a hazard to reproduction. Use of a VDU was recorded in 4712 current and 2164 previous pregnancies of women in full time employment at time of conception. After allowance for seven confounding variables, the risk of spontaneous abortion in current pregnancies relative to all working women was 1.19 (90% CI 1.09-1.30) and in previous pregnancies, 0.97. In an analysis by occupational title, in which 60 occupational groups were aggregated into eight categories according to use of VDUs, the relative risk for spontaneous abortion was 1.06 (90% CI 0.8-1.4) in current pregnancies and 1.01 (90% CI 0.7-1.3) in previous pregnancies. This suggests that the small excess of spontaneous abortions among individual women reporting the use of VDUs in current pregnancies may have been due to recall bias. Relative risks for stillbirth, preterm birth, and low birth weight all had 90% confidence limits which included unity. In an analysis of congenital defects the number of pregnancies was increased to include women who worked 15 or more hours a week. In all but one of nine groups of congenital defect examined confidence limits for the relative risk included unity in both current and previous pregnancies. The relative risks for the renal urinary group of defects were raised in both current (1.84, 90% CI 1.07-3.15) and previous pregnancies (1.66, 90% CI 0.82-3.25). There being no prior reason to suspect a causal link with this type of defect, interpretation remains open to question.