This study aimed to broaden the diagnostic possibilities of lowfield MRI systems (i) by examining the feasibility of functional MRI of human brain activation at 1 T, and (ii) by assessing its reliability in comparison with acquisitions at 3 T. Eight subjects were studied at 1 T and 3T using standard echo-planar-imaging sequences at 3-mm isotropic spatial resolution. Paradigms included silent word generation, sequential finger-tothumb opposition, and passive finger movements. Image post-processing was carried out either with statistical parametric mapping (SPM5, singlesubject and group analysis) or with a two-threshold correlation (TTC, single-subject analysis only) analysis. Single-subject analysis with SPM5 resulted in 3-5 times more activated pixels at 3 T than at 1 T in the examined Broca and sensorimotor regions. By comparison, the TTC single-subject analysis yielded the same amount of activated pixels at 3 T and 1 T. Moreover, this number was identical to that obtained with SPM at 3 T. The group analysis with SPM5 resulted in very similar numbers of activated pixels at both field strengths. The present findings suggest that a field strength of 1 T combined with adequate post-processing allows for reliable functional MRI studies of human brain activation. High-field advantages are therefore best invested in higher spatial resolution.