The capacity of striped patterns (square-wave gratings) to induce paroxysmal EEG activity in a group of pattern-sensitive epileptic patients is shown to depend on: 1. The spatial frequency of the pattern. The optimum spatial frequency lies between 1 and 4 cycles/degree in every patient tested. 2. The orientation of the pattern. Although, for the patient group as a whole, no one orientation is consistently more likely to induce paroxysmal activity than any other, the responses of individual patients can show marked orientation selectivity. 3. The brightness contrast of the pattern. The probability of paroxysmal EEG activity increases dramatically as contrast is increased from 0.2 to 0.4. Red/green gratings at isoluminance fail to induce paroxysmal activity. 4. The size of the pattern. There is considerable variation between patients in the subtense of a centrally-fixated circular pattern necessary to induce paroxysmal activity with a given probability. However, for every patient an increase in the probability of paroxysmal activity from near zero to near unity is effected by an increase in the angular subtense of the pattern by a factor of two. Patterns other than square-wave gratings are capable of inducing paroxysmal activity but, in general, patterns that stimulate more than one orientation system are less epileptogenic. The above findings are compatible with the hypothesis of a seizure trigger in the striate cortex and are incompatible with a trigger confined to the lateral geniculate nucleus of the thalamus. Further evidence against a geniculate trigger is obtained from an investigation of the response of a pattern-sensitive patient to a diffuse (unpatterned) flickering field in which it is shown that the effects of counter-phase interocular flicker implicate binocular mechanisms.