The Costs of Visual Working Memory

Alan Robinson¹, Alberto Manzi² & Jochen Triesch¹

¹Department of Cognitive Science, University of California, San Diego, USA

² Department of Psychology, Second University of Naples,

Italy

VSS 2005 Talk Abstract

While the capacity of visual working memory has been extensively characterized, little work has investigated how occupying visual memory influences cognition and perception. Here we show a novel effect: maintaining an item in visual working memory slows processing of visual input during the maintenance period.

We measured the speed at which subjects could determine the gender of human faces in a dual task paradigm. For the memory task subjects memorized computer generated faces or abstract 3D objects (Fribbles), and then after a 4 second delay, determined if an image was the same or different as the one memorized earlier. For the gender task subjects reported the gender of hairless human faces. The two were combined by inserting the gender task into the delay portion of the memory task. The question was how speed on the gender task would change as a function of the surrounding memory task. In EXPERIMENT 1 gender recognition was slower when another human face had to be maintained in memory than when a Fribble was maintained in memory. In EXPERIMENT 2 we verified that this effect was due to visual memory usage by adding a phonological loop interference task to prevent subjects from using any verbal encoding.

We interpret this effect as interference between memory and perception, caused by the visual similarity between ongoing perceptual input and items already encoded in visual memory. This interference is likely due to a neural overlap in the areas that recognize faces, and the areas that maintain faces in working memory. Thus, using visual memory has perceptual costs, which may explain the limited use of working memory found in research on natural tasks. Thus, everyday behavior may involve a complex trade-off between memory usage and efficient perception.

Example Stimuli

Acknowledgments

We thank the Tarr lab for donating the Fribble stimuli we used in this research, and Tara Sears for running subjects. This work was supported by the University of California Academic Senate, San Diego Division, under grant RC142C-TRIESCH.

Return to publications

Return to Alan's vision page