the small effects at each trial combine to demonstrate an important total effect of earlier events, of the memory pool, on judgment.
The results in Table 2.1, and other results in the referenced paper, allow the conclusion that performance in psychophysical tasks depends on the subject's ability to make momentary adaptations to features of the situation--to the response scale, the number of stimuli, trial spacing, difficulty of discrimination, and so on. It is often assumed that learning in psychophysical tasks is rapid, so that performance is essentially stable over the period when measurements are made. This is unrealistic. Rather, there is assimilation on each trial, and this assimilation requires small adjustments to continuously be made to some internal referent scale to maintain consistent responding over trials. Judgment is a very dynamic process. Average results in psychophysical experiments may be described by equations that ignore trial-by-trial events, but the microstructure of how such average data are produced cannot be described so simply.
There are robust and lawful trial-by-trial effects in judgment data. Even judging such simple things as the loudness of a tone or the length of a line involves more than a direct report of each stimulus by the observers. Each perceived stimulus is related to, perhaps assimilated to, an internal scale whenever the person asks "What is this?". The question is answered by comparing this already assimilated stimulus with memories of previous stimuli, with the knowledge or expectation of what stimuli could be presented, and with associated memories of how each event is to be labeled. These activities are revealed in the measures of sequential effects: People prepare for upcoming events in ways that are predictable in terms of the immediate past. Hence, to best predict what response a person will make on any particular trial, account must be taken of the prior trials. If these dynamic effects also occur in tasks other than the psychophysical procedures studied here, then judgments in general are more predictable if the sequence of prior events is known than if it is not.