Stimulus probability and stimulus set size in memory scanning.

Abstract

In many recent studies of speeded scanning of immediate memory, variations in the size of the positive set (s) were confounded with variations in the probability (P) of the individual items within the positive set: As s increased, P decreased . The present experiment sought to determine whether the effect on RT attributed to s could be accounted for by variations in P. This was accomplished by factorially varying both s and P. Probability effects were confined to items in the positive set and were insufficient to account for the effect of s. The results are discussed in terms of a model in which s and P affect different information-processing stages. The s affects the number of comparisons between the encoded item and the items stored in the memory of the positive set, as proposed by Sternberg. The P affects response selection— information as to the particular digit that was presented is available to the mechanisms for response selection along with the knowledge that there was or was not a match. The response selection mechanisms are assumed to be biased in tune with the P values of the items within the positive set. The number of things that one has to think about and the expectancy as to the likelihood of occurrence of these things— stimulus number and stimulus probability —have long been regarded as fundamental variables in the study of cognition. The common rinding that longer RTs would be produced by an increase in the number of possible stimuli or a decrease in stimulus probability was a result that was compatible with most theories of stimulus recognition. Discriminating among the various theoretical accounts for these effects has been a more elusive task. One class of models holds that variations in stimulus probability and stimulus number affect only a single commodity such as information (in bits) or repetitions. Examples of such models are those that posit