A consecutive cohort of 46 very-low-birthweight infants, who had routine electroencephalograms (EEGs) while in the neonatal intensive care unit, were studied. Two infants were lost to follow-up and were excluded, leaving a study population of 44 infants. Their mean birthweight was 945 +/- 166 g, gestation 27.1 +/- 1.7 weeks. Thirteen infants died before discharge. The remaining 31 had a mean corrected age of 26.1 +/- 8.7 months at the time of the last visit. Three groups were distinguished: normal survivors, handicapped survivors and non-survivors. The longest inactive phase encountered in the discontinuous EEG was the yardstick of the study with 3 subdivisions: less than 20 sec. 20-29 sec and equal to or more than 30 sec. The data showed a relative increase in poor outcome with increasing duration of inactivity (P less than 0.05) and, conversely, a favorable outcome with the absence of 20 sec or longer inactivity (P less than 0.001). All infants received therapeutic doses of phenobarbital during the early part of their illness which could have affected their EEG. Mean phenobarbital levels, obtained close to the time of the EEG, however, did not differ significantly and failed to show a significant relationship between the duration of the inactive phase and drug level observed. Inactive phases greater than or equal to 30 sec were more common (P less than 0.01) in infants with intraventricular hemorrhage. We conclude: although full recovery and normal outcome have been documented, prolonged isoelectric phases beyond 30 sec are more common in infants with fatal outcome.