The relation between spontaneous abortion (n = 5010), stillbirth without congenital defect (n = 210), and working conditions was analysed in 22,613 previous pregnancies of 56,067 women interviewed, 1982-4, immediately after termination of their most recent (current) pregnancy. The 22,613 previous pregnancies were those in which at time of conception the women were employed 30 or more hours a week. Ratios of observed (O) to expected (E) fetal deaths after allowance by logistic regression for seven non-occupational confounding variables were calculated at four stages of pregnancy in 60 occupational groups and six main sectors for women whose work entailed various physical demands, environmental conditions, and exposure to chemicals. The O/E ratios for abortion were raised in the sales sector (1.13, p less than 0.05) and services sector (1.11, p less than 0.01) and for stillbirth in the sales sector (1.50, p less than 0.1). Substantially increased O/E ratios for late but not early abortion were found in operating room nurses (2.92, p less than 0.05), radiology technicians (3.82, p less than 0.01), and employees in agriculture and horticulture (2.40, p less than 0.05); in all categories the O/E ratio for stillbirth were also raised but only significantly (5.55, p less than 0.01) in the latter group. The O/E ratio for stillbirth was also raised in leather manufacture (3.09, p less than 0.01). In both individual and grouped analysis (the latter undertaken to minimise the possible effect of recall bias) significantly increased O/E ratios for abortion were found in women exposed to various high levels of physical stress, particularly weight lifting, other physical effort, and standing (p less than 0.01). Increased ratios for stillbirth at this level of significance (p less than 0.01) were found for other physical effort and vibration. Noteworthy chemical exposure was identified only in the health, services, and manufacturing sectors; the O/E ratio for stillbirth approached two in women exposed to solvents, almost all in manufacturing (p less than 0.01). In the latter sector exposed to solvents was also associated with an approximately 20% increase in abortion ratio at similar probability level.