This study compares ictal semiology, neurological examination and scalp EEG between lateral and mesial occipital epilepsy to assess the contribution non-invasive data make in determining the epileptogenic region within an occipital lobe. We assessed seizure origin in 41 occipital patients as lateral (11 patients), mesial (20) and both surfaces (10) as indicated by subdurally recorded seizures (nine), a lesion whose removal reduced seizure quantity by > or =90% (six), or who met both criteria (26). No aspect of semiology distinguished lateral from mesially originating occipital seizures. A pre-operative visual field deficit appeared in eight (42%) out of 19 testable patients with mesial originating seizures, three (30%) out of 10 patients with both surfaces epileptogenic, but none of the 10 testable patients whose seizures arose only from the lateral surface (P = 0.0373, lateral versus mesial and both surfaces). Although occipital seizures appeared on the majority of the first five scalp EEG recordings in four (36%) out of 11 patients with laterally originating occipital seizures compared with none of 20 patients in whom seizures originated mesially (P = 0.0105), no other scalp EEG feature distinguished seizures from these surfaces. We conclude that subdural electroencephalography is likely to be necessary to delineate the epileptogenic region within an occipital lobe. Nonetheless, focally originating scalp-recorded seizures accurately lateralized the epileptogenic zone in 20 (49%) of our 41 patients compared with only one (2%) which originated contralaterally (P = 0.0001). This relationship held when considering only the first five scalp EEGs: the seizures of 10 patients (24%) appeared ipsilaterally and none contralaterally (P = 0.001). Moreover, interictal occipital (01,2) and posterior temporal (T5, T6) spikes appeared consistently and significantly (P < 0.001) more commonly ipsilateral to epileptogenesis than contralateral using multiple methods of analysis.