The cerebral localization of multiple languages is a topic of active research. This study presents a method for assessing whether partial overlap of active voxels reflects differential language localization, or simply the variability known to occur with multiple runs of the same task in fMRI studies. Two groups of bilingual subjects (early and later learners of L2) performed word fluency and sentence generation tasks in both languages. The degree of separation for regions of activation did not exceed that associated with run-to-run variability for either task or either group. Early bilinguals, however, showed greater total numbers of active voxels than Late bilinguals for both tasks. This effect occurred despite a lack of a behavioral performance differences by the two groups.