We sought to determine how the pattern of cerebral activation, and in particular in frontal motor areas, during the performance of conditional motor tasks is dependent upon the modality of instruction (visual, auditory, or somatosensory). Regional cerebral blood flow (rCBF) changes with externally instructed movements were also compared with internally generated, self-paced, movements. We used positron emission tomography (PET) with the tracer H2(15O) to measure rCBF in 22 healthy volunteers. External stimuli consisted of the randomized presentation of single or double impulses using a single modality for each condition. In the movement scans, the subjects used the index and middle fingers of their right hands to press a left button for a single and a right button for a double impulse, respectively. In the control scans, subjects were required to covertly distinguish a single from a double stimulus without a motor response. Data were analyzed using conventional subtraction techniques with a statistical threshold of Z > 2.33 with corrections for multiple comparisons. When the activation differences between the three externally instructed movement conditions were statistically compared, nonsignificant trends toward increased rCBF in the sensory cortex of the modality of the cue were observed but no differential activity in cortical motor areas. Internally generated movements, when compared to externally triggered movements, were associated with enhanced activation in bilateral medial and lateral premotor, dorsolateral prefrontal and superior parietal regions, largely confirming previous reports. The data indicate that, on a regional level, modality-specific processing in a conditional motor task does not occur in frontal motor areas and is probably confined to sensory areas.