PURPOSE: To evaluate the clinical usefulness of functional MR imaging (fMRI) for localization of the cerebral motor and sensory cortices and the language center in patients with complex partial seizure. MATERIALS AND METHODS: A total of 47 fMRIs were obtained in 14 patients (M:F=9:5; age 15 -50 years; 13 right handed and 1 ambidextrous) with complex partial seizure (6 temporal lobe epilepsy, 6 frontal lobe epilepsy, 1 occipitotemporal lobe epilepsy, 1 hemispheric epilepsy). Conventional MR imaging revealed no abnormality in four patients, localized cerebral atrophy in one, hippocampal sclerosis in four, and benign neoplasm in the remaining five. fMRI was performed on a 1.5 T MR scanner (GE Signa Horizon) using gradient-echo single-shot EPI. Nineteen fMRIs were obtained in eight patients who performed the language task, 16 fMRIs in ten who performed the motor task and 12 fMRIs in ten who performed the somatosensory task. The activation task consisted of three language tasks (silent picture naming, word generation from a character, categorical word generation), motor tasks (opposition of thumb and index finger for hand/dorsiflexion or extension for foot), and sensory tasks (passive tactile stimulation of hand or foot using a toothbrush). The data were analyzed using z-score (p<0.05), clustering, and cross-correlation analysis based upon homemade software, IDL 5.1. The success rate for obtaining meaningful fMRI was evaluated and activated regions were assessed on the basis of each fMRI obtained during language, motor, and omatosensory tasks. fMRI findings were compared with those of the Wada test (n = 7) for language lateralization and with invasive cortical mapping (n = 3) for the localization of eloquent cerebral cortex, especially around the central sulcus. RESULTS: The overall success rate of fMRI was 79 % (37/47); success rates of fMRI with language, sensory, and motor tasks were 89 % (17/19), 83 % (10/12), and 63% (10/16), respectively. Areas activated during language tasks (n=17) included the dorsolateral prefrontal cortex (16/17), anterior cingulate gyrus (16/17), posterior parietal area (15/17), Wernicke's area (10/17), and Broca's area (8/17). Regions activated by motor and somatosensory tasks (n=20) included the central sulcus (16/20), prefrontal cortex (15/20), posterior parietal cortex (9/20), supplementary motor area (6/20), and temporal cortex (3/20). The results of fMRI were concordant with the Wada test in six of seven subjects (86%) for language lateralization of left hemispheric dominance and with invasive cortical mapping in two of three patients (67%) for localization of the motor and sensory cortices. CONCLUSION: fMRI was successful in approximately three-quarters of patients with complex partial seizure, but at present appears to be an inadequate alternative to current invasive studies. Further clinical investigation is needed.