This study introduces a direct method of assessing cerebral lateralization for language based on fMRI activation. The method, derived from a voxel-based morphometry study by C. H. Salmond et al. (2000, Hum. Brain Mapping 11, 223-232), bases lateralization on the direct statistical comparison of the magnitude of task-induced activation in homotopic regions of the two hemispheres. Lateralization results obtained with this direct method were compared to those obtained with a widely used method which involves the calculation of a laterality index (LI) based on the number of significantly activated voxels in the inferior frontal gyrus of each hemisphere. In order to compare the validity of the two methods, a covert verb-generation task was performed by eight children with epilepsy whose language lateralization was examined using invasive techniques. Lateralization results derived from fMRI activation showed that the calculation of a LI presented some limitations. Importantly, the LI value was dependent on the activation threshold chosen to calculate that LI. As a consequence, the correlation between the LI and the invasive methods could vary with the chosen threshold. By contrast, the proposed direct method gave some indication of the reliability of the lateralization and provided results that, in all eight children, were consistent with those obtained using invasive techniques. It is suggested that the direct method could be used in future fMRI studies to establish hemispheric lateralization for cognitive functions.