In a 36-month period antepartum testing was performed 5973 times in 2628 women with singleton high-risk pregnancies. The testing scheme involved a modified nonstress test with sound stimulation. Testing was performed twice weekly for patients with diabetes (classes B through R), gestational age exceeding 42 weeks, and documented intrauterine growth retardation, and weekly for other indications. If no spontaneous acceleration was observed within 5 minutes, a single 1- to 2-second sound stimulus was applied to the lower maternal abdomen with an artificial larynx. If necessary, a second sound stimulation was applied within 10 minutes. In addition all patients received ultrasonographic four-quadrant assessment of amniotic fluid volume. The mean testing time was 10 minutes. Only 2% of tests were nonreactive with sound stimulation. Seventeen percent of nonstress tests that were nonreactive with sound stimulation were followed by positive results of a contraction stress test or a biophysical profile score less than or equal to 4. The overall intervention rate was 3%. All fetuses with a single acceleration only eventually met criteria for negative results to a contraction stress test or had a biophysical profile score greater than or equal to 8. There were no unexpected antepartum fetal deaths. Sound-induced accelerations appear to be valid in the prediction of fetal well being, and the use of sound stimulation results in a significant shortening of testing time. Simultaneous assessment of amniotic fluid volume may reduce the risk of fetal death to a negligible level.