Purchase this article with an account.
Michael Mangini, Michael Villano, Charles Crowell; Visual Short term Memory for One Item. Journal of Vision 2010;10(7):620. doi: 10.1167/10.7.620.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Visual short term memory (VSTM) is a limited capacity system that abstracts visual information from sensory stimulation. Many studies have investigated the storage capacity of this system expressed in the numbers of objects, features, or complexity. Here we investigate the accuracy of visual short term memory for single items. On half of the trials, memory trials, a single face or spatially filtered noise pattern is initially presented. After a one second memory delay, a two-alternative forced choice (2AFC) is presented. On the other half, perceptual trials, participants are presented with a simultaneous 2AFC match to sample task. Both noise and face stimuli are synthesized to contain equivalent low-level visual structure. They have equal spatial frequency profiles and both stimulus classes are generated from the summation of twenty randomly amplified orthogonal linear templates. Results showed participants were more sensitive to face stimuli. The memory delay caused a significant decrease in performance. Interestingly, a significant interaction between the magnitude of the memory decrement and the stimulus type was observed. Specifically, memory trials for noise stimuli showed a larger performance decrement than that observed for face stimuli. These findings suggest that VSTM does not have the capacity to store even a single complex item at the level of detail that is available when an image is present. Significant degradation occurs within a second. Also, the accuracy of VSTM is stimulus specific. Faces seem to be represented by VSTM more efficiently than filtered noise. Because the noise and face stimuli have computationally similar degrees of variation, the differences in performance must be due to internal representation. This suggests that for complex stimuli VSTM likely utilizes previously learned statistical regularities, and is therefore not a general purpose mechanism. Models of stochastic visual memory decay in low-level image space cannot account for our findings.
This PDF is available to Subscribers Only