This paper first provides a survey of the expanding brain imaging literature in the field of time processing, showing that particular task features (discrete vs rhythmic, perceptual vs motor) do not significantly affect the basic pattern of activation observed. Next, positron emission tomography (PET) data obtained in a timing task (temporal reproduction) with two distinct duration ranges (2.2--3.2 and 9--13 s) are reported. The stimuli consisted of vibrations applied to the subject's right middle finger. When the vibration ended, the subject estimated an interval identical to its length before pressing a response button. The control task used cued responses with comparable intervals and stimuli. The pattern of activation obtained in the timing task as compared to control mainly included areas having attentional functions (the right dorsolateral prefrontal, inferior parietal, and anterior cingulate cortices), and the supplementary motor area (SMA). No significant difference was seen as a function of the duration range. It is argued, firstly, that involvement of the attentional areas derives from specific relations between attention and the temporal accumulator, as described by dominant timing models; and, secondly, that the SMA, or more probably one of its subregions, subserves time processing.