The time relations of epileptic events have been studied in 3 sets of data: (I) counts of individual epileptiform discharges in twelve 48 h EEG recordings, (IIa) seizure calendars of 30 therapy-resistant outpatients participating in a drug trial, (IIb) seizure calendars of 10 mentally subnormal epileptic patients resident in a long-stay unit. The EEG data I were characterized most often by a Poisson distribution of intervals between discharges and the occurrence of marked periodicities, particularly at night. The periods of rhythmic nocturnal events ranged from 13 to 142 min and did not appear to correspond to the REM/non-REM cycle. In the seizure data IIa and b a Poisson distribution of intervals between events was found in half the patients. Periodicities occurred only in group IIa and did not correspond to weekly or monthly cycles. A stochastic process is considered to be the model which best fits these data.