Stimulus spacing effects in temporal bisection by humans.

Two experiments studied the bisection of sets of stimulus durations (ranging from 200 to 800 msec, or 100 to 900 msec) by normal adults. In Experiment 1, two bisection techniques were used: a "similarity" method, where subjects had to classify each duration in terms of its similarity to the shortest and longest members of the set (which were explicitly identified as Short and Long standards), or a "partition" method, where subjects simply classified each duration as Short or Long, without explicit standards. For different groups, the durations within each set were spaced either linearly or logarithmically between the longest and shortest members. The two bisection methods produced similar results, but psychometric functions (proportion of Long judgements versus stimulus length) were shifted to the left in the logarithmic conditions relative to the linear ones--that is, some shorter durations were classified as Long more frequently in logarithmic conditions. Experiment 2 used arithmetic, but unequal, stimulus spacings between the shortest and longest set members, and the partition method. Sets containing more shorter than longer durations had psychometric functions shifted to the left relative to sets with more longer than shorter durations. The data suggest that judgement of some time value, t, depends on the context in which t appears, not only on its value relative to the longest and shortest set members (as most current theories of bisection propose). A model assuming that judgements of t are based on the relation between t and the arithmetic mean of all the durations in the set fitted data reasonably well in most conditions. This model, furthermore, incorporated decision rules used to account for human performance on temporal generalization tasks, thus promoting theoretical integration of the two sorts of experiments.