Does temporal reference memory change dynamically during temporal generalization performance?

Four experiments investigated the idea that temporal reference memory in a temporal generalization task may show dynamic changes from trial to trial even when the standard duration is constant from the procedural point of view. Temporal generalization gradients arising from a constant set of stimuli were shifted by "biased" sequences of comparisons, which apparently moved the contents of temporal reference memory in the direction of the bias. The basic principle was that if a person identified some comparison duration of x ms as being the standard for the task, the value of x affected the effective standard used. This biasing effect was removed by providing participants with accurate feedback, but could be restored by false feedback in the same direction as the bias. A model incorporating the basic idea of dynamic reference memory produced small biasing effects in the appropriate direction, but was not in exact quantitative agreement with the data, suggesting the role of additional processes yet to be discovered.