These experiments explore duration perception in vision and audition to find the level of internal noise associated with time perception and the number of samples pooled into a duration estimate. Duration increment thresholds were measured for standard durations of 100,200,400,800 & 1600ms. A spatially distributed array of four LEDs (or loudspeakers) emitted visual (or auditory) signals to define a temporal duration. Randomly staggering signal onsets/offsets added variance (or ‘noise’) to the duration around a standard mean. Increment thresholds for each standard were measured for various temporal noise levels. Threshold change with added duration noise was modeled to estimate the level of internal duration noise and the number of samples being pooled in duration perception. This “equivalent noise� approach estimates internal noise (i.e., thresholds increase when added external noise exceeds internal neural noise) and the number of samples averaged into a single duration estimate (the slope of threshold change vs. added noise once internal noise is overcome). Data for vision and audition were similar. Discrimination thresholds increased with the standard duration (Weber fraction ∼ 15%). Importantly, once external noise exceeded internal noise, thresholds increase in proportion with added noise. This suggests a single timing process with no possibility for pooling estimates to deal with noise. Audiovisual experiments suggest this timing process is supramodal, as performance for visual (or auditory) signals was no better than a mixture of visual and auditory signals. In both modalities, Weber fractions for duration perception with multiple stimuli are relatively poor, reflected by relatively high levels of estimated internal noise. Once added noise exceeds internal noise, discrimination thresholds increase approximately in proportion to added noise. If multiple timers existed, averaging could be used to overcome external noise. The failure to find this suggests the visual and auditory systems cannot access multiple estimates of duration. Overall, the data suggest a single and relatively imprecise supramodal timing process.