Objective Fusiform face area (FFA) is highly specialized for faces and independent of task modulation.This study was to test the stability of FFA response to face information across 4 fMRI scans spanning one year in the same group of subjects.Methods Block design was adopted and the stimuli included English words, Chinese characters, and faces.General Linear Model (GLM) was performed to define the ROI (region of interested), i.e., the FFA, in each subject.The stability of space coordinates of activation voxels within the ROI was analyzed by within-subject ANOVA.Results The coordinates of the voxels in the FFA ROI were highly reproducible across the 4 scans whether the ROI was defined by face vs.Chinese or face vs.English.Conclusion The position of the FFA measured by fMRI is remarkably stable.It is feasible to use the FFA as a reference region in fMRI studies on the visual system.