The types of stimuli loaded in memory can modulate its effects on visual search

Abstract

The effects of memory load in visual search (VS) have shown a diversity of results from the absence through beneficial and detrimental effects of a concurrent memory load in VS performance. One of the hypotheses intended to explain the heterogeneity of results follows the idea proposed by certain models in the context of VS that the contents of working memory (WM) can modulate the attentional processes involved in VS (Desimone & Duncan, 1995; Duncan & Humphreys, 1989). In four experiments, we manipulated the similarity of information maintained in WM and those materials playing the role of target and distractors in the VS task. The results showed a beneficial effect in the first two experiments, where the materials in WM matched the target in VS. However, when they matched the distractors in the attentional task there is no effect in the slope of the search function. Present results strengthen those theories supporting that visual working memory is fractionated to allow for maintenance of items not essential to the attentional task (Downing & Dodds, 2004).

Keywords:

• Memory load
• Selective attention
• Visual search
• Working memory

Notes

¹Smilek et al. (2006) analysed their results employing the so-called Inefficiency Scores (Townsend & Ashby, 1983). The inefficiency scores combine RT and errors in a single measure of search inefficiency by dividing the correct mean RT of each participant by its correct mean proportion. They corrected the RT measure by its appropriate level of accuracy; when accuracy is perfect the inefficiency score equals the mean RT, but as accuracy decreases the inefficiency score increases proportionally to the errors. In order to be sure that any difference between our conclusions and those from them does not depend on using the raw average RT instead of the inefficiency, we redid the RT ANOVAs of Experiment 1 and the following experiments with the inefficiency scores. In all cases we found significant effects of the same factors as with the raw averages of RTs.

²As the p-value for the interaction may be considered as a marginally significant effect (p=.086), we conducted a t-test analysis on the slope values of the single and dual task conditions in order to strengthen the results found for RTs. The results showed that there were no differences between the slopes of the search function in Experiment 3, t(9) = 1.38, p=.20. Moreover, analysing differences between slopes for all experiments we found p<.01 for Experiments 1 and 2A, and p=.40 for Experiment 2B, again strengthening our hypothesis.