An apparent discrepancy exists between the remarkable language competence of the isolated right cerebral hemisphere in some split-brain patients and the lack of language competence in some aphasic patients in whom only the left cerebral hemisphere has been damaged. This has led to a revival of the idea that the dominant cerebral hemisphere inhibits the potential functioning of its partner. We tested this model of interhemispheric inhibition in four patients with unilateral limbic seizures (three left-sided, one right-sided). While seizure activity was monitored with bilaterally implanted stereotactic EEG depth electrodes or bilateral foramen ovale EEG electrodes, the patients were tested in lateralized hemisphere-specific tachistoscopic recognition experiments. In two patients performance was correlated with the electrical pattern during prolonged unilateral subclinical limbic status epilepticus, and in two patients performance and intracranial electrical activity were compared before, immediately after and during the recovery phase of left limbic complex partial seizures with postictal aphasia. Three main findings were obtained. (i) Focal unilateral limbic seizure activity in the depth interferes with cognitive functions, even when unnoticed clinically or undetectable on the surface EEG, provided that epileptiform EEG activity lasts for several seconds. (ii) This interference is specific to the cognitive functions associated with the 'dominant' processing of the discharging hemisphere. (iii) Seizure-induced impairment of the 'dominant' functions of the discharging hemisphere is associated with improvement of performance of the same function by the other hemisphere. Although each patient's clinical and experimental situation differed, the converging results support the hypothesis of hemispheric interaction based upon functional inhibition for verbal processing in patients with a left hemispheric focus. The reverse situation, namely facial processing in a single case with a right hemisphere focus, remains inconclusive.