Regional cerebral glucose metabolic rate (GMR) quantified with positron emission tomography (PET) with 18-fluoro-2-deoxyglucose (FDG) was measured twice in 8 young men performing a complex visuospatial/motor task (the computer game Tetris), before and after practice. After 4-8 weeks of daily practice on Tetris, GMR in cortical surface regions decreased despite a more than 7-fold increase in performance. Subjects who improved their Tetris performance the most after practice showed the largest glucose metabolic decreases after practice in several areas. These results suggest that learning may result in decreased use of extraneous or inefficient brain areas. Changes in regional subcortical glucose metabolic rate with practice may reflect changes in cognitive strategy that are a part of the learning process.