Regional cerebral blood flow was measured in normal subjects with positron emission tomography (PET) while they performed five different motor tasks. In all tasks they had to moved a joystick on hearing a tone. In the control task they always pushed it forwards (fixed condition), and in four other experimental tasks the subjects had to select between four possible directions of movement. These four tasks differed in the basis for movement selection. A comparison was made between the regional blood flow for the four tasks involving movement selection and the fixed condition in which no selection was required. When selection of a movement was made, significant increases in regional cerebral blood flow were found in the premotor cortex, supplementary motor cortex, and superior parietal association cortex. A comparison was also made between the blood flow maps generated when subjects performed tasks based on internal or external cues. In the tasks with internal cues the subjects could prepare their movement before the trigger stimulus, whereas in the tasks with external cues they could not. There was greater activation in the supplementary motor cortex for the tasks with internal cues. Finally a comparison was made between each of the selection conditions and the fixed condition; the greatest and most widespread changes in regional activity were generated by the task on which the subjects themselves made a random selection between the four movements.