Positive temporal sharp waves (PTS) were studied in electroencephalograms (EEG) of 92 premature infants born either at 31 and 32 weeks of gestational age, recorded during the first week of life. The infants were assigned either to a reference group (asymptomatic) or to one of three pathologic groups (neonatal asphyxia, hypoglycemia or hypocalcemia, or periventricular leukomalacia with rolandic positive sharp waves). Regardless of the group, no significant differences in PTS criteria (morphology, frequency, amplitude, duration and lateralization) were found between 31- and 32-week infants. The PTS observed in 55% of the asymptomatic infants were characterized by low frequency (0.13 +/- 0.12/min), a mean amplitude of 109.8 +/- 25.8 microV and a mean duration of 148.7 +/- 35.4 ms. For PTS recorded in 72 to 75% of pathologic infants, mean duration and amplitude were significantly greater in all groups than in asymptomatic infants, whereas frequency was significantly greater only in the group presenting with asphyxia. Discriminant analysis based on the three PTS criteria (frequency, amplitude and duration) allowed correct classification for only 30 to 54% of infants in the four groups. The frequency of PTS decreased rapidly during the second week of life in asymptomatic infants, but persisted in a larger number of infants in the pathologic groups. This study shows that PTS have no negative significance when they are few in number, short in duration, moderate in amplitude and rapidly regressive, thus probably reflecting the vulnerability of the temporal lobe during the traumatic period of birth. However, they require attention when they are abundant and/or slow, ample or tend to persist. They may thus constitute a nonspecific response to injury to an immature brain.