Forty-one infants with neonatal seizures frequent enough to be captured by randomly recorded routine EEG examinations were studied to determine how often their electrographic seizures were occult (subclinical) and to examine the effects of seizure duration and mental status on their clinical visibility. Seizures were the result of diverse etiologies and most infants had received one or more antiepileptic drug prior to the EEG recordings. The majority of electrographic seizures were occult: only 84 of 393 (21%) were accompanied by distinctive clinical seizure activity; the remaining 79% were occult. There was no significant difference between the duration of EEG seizures with distinctive clinical signs and those without. There was no significant difference in the proportion of occult seizures in neonates with preserved consciousness compared with lethargic or comatose infants. We conclude that unaided visual inspection of infants seriously underestimates true seizure frequency. Long-term EEG monitoring may be necessary in many infants to determine their real seizure frequency and to judge the adequacy of antiepileptic drug treatment.