Abstract
Searching for a potentially moving person in a dynamic crowd is a common visual task, but little is known about how we do this. Recent work has shown that the frontoparietal attention network, particularly the intraparietal sulcus (IPS), represents the presence of targets during the search for objects in static natural scenes (Guo et al., 2012). Here, we used fMRI and multivoxel pattern analysis to determine whether the IPS contains information about the presence of a target person in a dynamic scene depicting a crowd. In Experiment 1, separate groups of participants watched a series of brief (8.75 s) videos during fMRI and searched for either a particular person or for an object (a skateboard), each of which was present on 50% of the trials. For both people searchers and object searchers, decoding accuracy using BOLD activity extracted from IPS was significantly above chance (person searchers=63% SEM=0.020, p <0.01; object searchers=71% SEM=0.014, p <0.01). In Experiment 2, a single group of observers watched videos and alternated between searching for a person and searching for a skateboard. Across conditions, target presence was again decoded from activity in IPS significantly above chance (person trials=63%, SEM=0.026, p <0.01; object trials=64%, SEM=0.013, p <0.01). Moreover, IPS activity was able to predict the target of the search (person vs. skateboard) significantly above chance (64% correct classification, SEM=0.024, p <0.01). Together, these results show that the IPS plays an important role in visual search by representing the presence of both person and object targets, as well as the identity of the search target.
Meeting abstract presented at VSS 2014