Locus of the Stimulus Probability Effect in Item Recognition

Abstract

<p>Part 1 of the present investigation was designed to examine Sternberg's model of information retrieval. In particular, Experiments 1 through 3 focused on the confounding between frequency of occurrences, P, and positive set size, s. These experiments were designed to examine the possibility that the increase in RT with increases in s is, in whole or in part, an effect of variations in the frequency of occurrences of individual positive items. In shot, each experiment examined the memory scanning stage of the linear additive model.</p> <p>In Experiment 1, the two variables, s and P, were unconfounded for some trials by holding constant the frequency of occurrences for one item in each set size. Here, it was found that when P was unconfounded by s, there remained a small but significant effect of s, supporting the conclusion that increases in mean RTs are largely accounted for by the associated decreases in P.</p> <p>To determine whether or not the variable P has additive or interacting effects in the scanning stage, the additive-factors method was employed in the subsequent experiments. Mean RTs were obtained where at least two positive items within each memory set were assigned different P values and these particular values of P were found in all set sizes. In Experiment 2, it was found that P had additive effects when values of P were held constant across s at .25 and . 15. In contrast, in Experiment 3 when P was held constant across s at .25 and .05, it was found that P had interacting effects on the scanning stage, strongly suggesting that the serial and exhaustive scanning model, as proposed by Sternberg, is unable to handle the effects of P.</p> <p>In an attempt to explain some of the features of the data, Stanovich and Pachella's temporal overlap model, Theios et al. 's self-terminating model, and Atkinson and Juola's familiarity model were examined separately. The general features of the data reconciled best with the familiarity model where it is hypothesized that subjects do not always serially and exhaustively scan the memorized list on every trial. The supposition is that repetitions of an item as a probe will result in an increase in its familiarity value, and thereby increase the likelihood of a fast positive response.</p> <p>Working within the basic concepts underlying the familiarity model, Part 2 of this Thesis describes an experiment which examined aspects of repetition that affect the memory scanning stage of the item recognition process. In general, the data revealed that repetition is an important variable since scanning of the memory list seems to be influenced by how often a positive item is probed and by the number of intervening items occurring between consecutive tests of positive item (lag length).</p>