Purchase this article with an account.
Margaret C. Jackson, Jane E. Raymond; Visual working memory for faces. Journal of Vision 2004;4(8):394. doi: 10.1167/4.8.394.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
How many faces can be stored in visual working memory (WM)? Although recent studies agree that visual WM may be limited to only 3–4 items, there is debate as to whether this storage limitation applies to visual features or to integrated objects. To address this issue and produce empirical data on visual WM limitations for complex stimuli, we measured capacity for faces using a standard change detection paradigm. If visual WM is object-based and faces are stored configurally, then capacity for faces should be similar to that of simple coloured squares. If face storage is featural and visual WM is feature-based, then capacity for faces should be significantly less. Using the Vogel et al. (2001) paradigm, we presented two successive arrays: a memory array, followed 900 ms later by a test array. (Set size varied between 1 and 10.) Observers (n = 24) were required to detect a change (50% of trials) between the two arrays whilst simultaneously performing a verbal suppression task. In separate blocks (within-subjects), items were (singularly) coloured squares or greyscale faces. Squares were displayed for 100 ms (memory array) and for 2000 ms (test array). To ensure sufficient encoding time, face arrays were displayed for durations derived from each observer's mean RT in a face visual search pre-task. Proportion correct change detection scores were converted to d′ values and plotted as a function of set size. We found that d′ fell significantly more steeply (p < .001) for faces than for squares with increasing set size, indicating reduced WM capacity for faces. From these data, we estimate that visual WM can retain approximately 2 faces worth and 4 squares worth of information (the latter result replicating Vogel et al., 2001). If visual WM capacity has a limit of 4 chunks (Cowen, 2000), these findings can be interpreted to indicate that storage of a face into WM involves about two chunks of information and that, at least for faces, visual WM cannot be object based.
This PDF is available to Subscribers Only