Abstract
Previous work has shown that searching for two targets is less efficient than for one, but why? One possibility is that people abandon one object and guide their search as if there were only one target. Another possibility is that both targets are represented and used for guidance, but that they lack the detail of a single-target representation due to visual working memory (VWM) limitations. To disentangle these hypotheses, we monitored eye movements during search for one versus two real-world targets inside of an fMRI scanner. Participants were first shown a preview display consisting of two images, either one target (shown twice) or two different targets, then a search display after a 9-second delay. One of the two images from the preview was always present in the search display, and the task was to look at the target and press a button. Behavioral analyses revealed that search guidance (percentage of trials in which the target was the first object fixated) was lower for two-target trials (47%) compared to one (67%). If this two-target cost was due to participants representing only one of the previewed objects, we would not expect to find a load effect in brain activity. However, fMRI data revealed greater delay period activation in the lateral and dorsal prefrontal and posterior parietal cortices, all regions previously associated with working memory maintenance loads (Leung et al, 2002, 2004). Moreover, we also found delay period activations in both lower- and higher-order visual regions. These results show that both targets are represented and used to guide search, and suggest that the guidance cost associated with two-target search is likely due to a VWM limitation. For one-target search, sufficient features are coded to yield a good guiding representation, whereas for two-target search, VWM limitations prevent the sufficient coding of features, resulting in weaker guidance despite more brain activity in visual areas.
Meeting abstract presented at VSS 2012