In a stop-reaction-time (stop-RT) task, a subject is presented with a regular, isochronous sequence of brief signals separated by a constant time interval, or stimulus onset asynchrony (SOA). His/her task is to press a response key as fast as possible when the sequence stops. As the sequence Unfolds, an internal representation of the SOAduration builds up. Stop-RT is assumed to be triggered when an internal clock, operating as an "alarm clock," reaches a time criterion. Criterion setting is contingent upon variability in the SONs internal representation. In Experiment lA, stop-RT was measured for isochronous sequences ofbrieftones, light flashes, and also sequences of tones and flashes presented in regular alternation (tone-light-tone ...). Stop-RT was a linear function of SOA duration (ranging from 250 to 1,000msec), regardless of modality, supporting a "central-clock" hypothesis. On the other hand, taken together, the results of Experiments lA, 1B, 2, and 3 suggest that no internal representation of the bimodal (tone-light) SOA of alternating sequences builds up. Indeed, an alternating sequence is physically equivalent to two interlaced isochronous subsequences, one auditory and one visual. So, two internal representations, one for the auditory (tone-tone) and one for the visual (light-light) SOA, could build up, and two time criteria running "in parallel" could thus support stop-RT. To provide a critical test of parallel timing, stop-RT was measured for bimodal 5:3 polyrhythms formed by the superposition of auditory and visual isochronous sequences that had different SOAdurations (Experiment 4). Parallel timing accounted for a large proportion of variance in polyrhythmic stop-RT data. Overall findings can be accounted for by assuming a functional architecture of an internal clock in which pulses emitted by a central pacemaker are available in parallel with two modalityspecific switch-accumulator "timing modules."