PURPOSE: To investigate the efficiency or potency of functional MRI (fMRI) for the determination of language dominance by comparing the results of fMRI with those of the Wada test. MATERIALS AND METHODS: Among 34 patients with intractable seizure who underwent both fMRI and the Wada test, we analyzed the results of 30 (men:women=19:11; mean age=29.7 years). Using echoplanar imaging and the blood oxygen level dependent technique, fMRI was performed using a 1.5 T MR imager with a standard head coil. The language task consisted of two parts: reading words and generating words. For fMRI, a multi-event multi-task paradigm consisting of two sets of activation, rest, and alternative periods was used. Image processing involved the use of the Z test (Z threshold = 1.0 -1.2). To determine the lateralization index, we calculated the activation pixels within the whole frontal cortex., and to ascertain the discrepancy between the two tasks, the clustering grade of activation pixels was measured. After the injection of thiopental, language dominance was determined by means of a modified Wada test. The results of this and the findings of fMRI were compared with the results of Fisher 's exact test (p < 0.05). RESULTS: The correlation indices between the findings of fMRI and the results of the Wada test were 77% for word generation and 50% for reading. The diference was only marginally significant (p = 0.06). For the two tasks, the opposite results were 33% for reading and 3.3% for word generation, and these were significantly different (p < 0.05). The clustering grade for more than one unit was 40% for word generation, and 23% for readings, a difference which was not statistically significant(p> 0.05). CONCLUSION: For the determinaton of language dominance, fMRI showed good correlation with the Wada test. The word generation task was more efficient than the reading task. fMRI which is non-invasive and repeatable, is therefore more efficient and useful than the invasive Wada test.