OBJECTIVE To determine the value of computerized antepartum measurements of short-term fetal heart rate (FHR) variation and decelerations as a predictor of outcome, and to compare these with measurements of the umbilical artery flow velocity waveform. METHODS Data were collected from 3563 high-risk patients for measurements of FHR variables (15,702 records) and outcome (intrauterine death or metabolic acidemia on delivery). Detailed analyses were made on 89 patients with one or more records in which short-term FHR variation was 3 milliseconds or less. Umbilical artery flow velocity waveforms were measured concurrently on 2441 occasions in 991 patients. RESULTS More than half of the patients with FHR variation of 3 milliseconds or less were identified before 31 weeks. When short-term FHR variation exceeded 3 milliseconds, there were no intrauterine deaths and only one instance of metabolic acidemia on cesarean delivery. When FHR variation fell below 2.6 milliseconds, 34% of the subjects had metabolic acidemia on cesarean or intrauterine death. The appearance or absence of decelerations was an unreliable guide to outcome. The correlation between FHR variation and the umbilical artery flow velocity waveform was low (r = 0.32). The population studied contained some patients without placental vascular disease or, in five with pre-eclampsia, without abnormal umbilical artery velocity waveforms but with grossly reduced FHR variation. CONCLUSIONS Reduction in short-term FHR variation, as measured by computer, reliably predicts fetal outcome and is a more general measure of fetal well-being than are umbilical artery flow velocities.