The multiple injection [15O]water method offers unique opportunities for studying cognitive processing by the human brain. The influence of the duration and temporal placement of an activation task, in relation to the arrival of the radiotracer in the brain, is a fundamental methodologic question for cognitive activation studies. A quantitative positron emission tomography (PET) study of five normal volunteers was performed in which the stimulation consisted of a visual activation task (alternating checkerboard pattern) superimposed on an auditory baseline task (syllable monitoring). Ten injection conditions, with varying duration and timing of the visual activation, were used. Regional CBF (rCBF) in visual cortex was measured quantitatively using the autoradiographic method. A 20-s stimulation, centered on the bolus arrival in the brain, produced significant changes in rCBF. Because varying the duration and timing of the activation task technically violates the temporal homogeneity assumption of the autoradiographic model, a mathematical simulation was formulated to evaluate the potential influence of these variations. Results of the simulation are consistent with the PET data and suggest that activation can be limited to a narrow temporal window centered on the radiotracer uptake. The ability to observe significant changes in rCBF with short stimulation intervals is of particular interest in the use of [15O]water PET for studies of cognitive processes with a short time course.