Traditional electroencephalography (EEG) produces a large volume display of brain electrical activity, which creates problems particularly in assessment of long periods of intracranial, stereoelectroencephalographic (SEEG) recording. A method for fractal analysis that describes 100 SEEG data points in terms of a single estimate of fractal dimension (1 < FD < 2) is reported; the central processing unit time costs amount to approximately 2 min/Mbyte of input signal (using a Sun SPARCstation LX). The diagnostic sensitivity of this method, applied to quantification and synoptic visualisation of SEEG signals recorded during 35 epileptic seizures in 7 patients, is evaluated. It is found that the method consistently defines ictal onset in terms of rapid relative increase in FD across several channels. Clinically severe seizures are characterised by more intense and generalised ictal changes in FD than clinically less severe events. For all 7 patients, and for 75% of individual seizures, "fractal diagnoses" of anatomically defined ictal onset zone coincided closely with ictal onset zone independently determined by inspection of traditional EEG displays of the same data. We conclude that the method is a computationally feasible way to achieve substantial reduction in the volume of SEEG data without undue loss of diagnostically important information in the primary signal.