Abstract
Consider a set of n (0 < n < 9) horizontally-oriented lines in the x−y plane, all of the same standard length (w), except for an uncued target, in a random y position, that is shorter or longer (w +/− Δw). The lines have random offsets along the x axis, sampled from a uniform distribution of width w. Spacing on the y axis is fixed. Without a postmask, there appears to be no capacity limit for deciding whether the target is larger or smaller than the standard; accuracy can be well-modelled by the ‘max’ rule of signal detection theory. Now re-label the x axis as t, so that the horizontal lines represent durations of the stimuli rather than length in space. The standard duration was 2 sec. Accuracy with n=1 was comparable to that of the spatial task, when expressed as a Weber Fraction. However, with multiple stimuli, the task was very hard indeed. Even with n=2, predictions of the max rule fail dramatically. We conclude that observers have access to only a single clock for timing visual stimuli, which is started by a shift of attention, and which stops when attention shifts from the stimulus. A multi-resolution model of timing is proposed to account for the linear increase in the threshold Δw with w.