Purpose: To prospectively determine the repeatability of functional magnetic resonance (MR) imaging brain activation tasks in a group of healthy older male volunteers. Materials and Methods: Local research ethics committee approval and informed consent were obtained. Sixteen men with a mean age of 69 years +/-3 (standard deviation) performed finger-tapping and N-back (number of screens back) working-memory tasks. Each subject underwent MR imaging three times in weekly intervals. Within-subject task repeatability was analyzed in terms of the number of voxels classified as activated (activation extent), the mean activation amplitude, and (for finger tapping) the center of the mass of the activated region. A repeatability index was calculated to compare test-retest repeatability between subjects and between functional MR imaging tasks. Within-session, between-session, and between-subject variability was assessed by using analysis of variance testing of activation amplitude and extent. Results: Nine of the 16 subjects generated useful data at all three MR imaging functional task sessions. At single-subject, single-session analysis, cortical activation was identified in most subjects and at most sessions. The centers of the masses of motor cortex activation were highly reproducible (within 3 mm). Patterns of activation were qualitatively repeatable, but there was substantial variability in the amplitudes and extents of activated regions. Within-session coefficients of variation (CVs) for left-versus right-hand and right-versus left-hand finger tapping were, respectively, 65% and 43% for activation amplitude and 75% and 121% for activation extent. The between-session CVs for activation amplitude were similar to the within-session values, whereas between-session CVs for activation extent were much greater than within-session values, up to 206%. Conclusion: The generally poor quantitative task repeatability highlights the need for further methodologic developments before much reliance can be placed on functional MR imaging results of single-session experiments. (C) RSNA, 2004.