Abstract
Studies of object-based attention have shown that when one feature of a multi-feature object is selected, attention tends to spread to other, task-irrelevant features (Ernst et al., 2013). Other evidence suggests that access to WM can be restricted to only relevant features (Serences et al., 2009; Woodman & Vogel, 2008). Another possibility is that all of an object's features are initially encoded, but irrelevant features are removed from WM over time (Logie et al., 2010). We used MVPA trained on time-frequency decomposed EEG data to examine the temporal evolution of neural representations reflecting the encoding and storage of task-relevant and irrelevant features in WM. In different blocks, participants remembered the orientation, color or both orientation and color of a colored, oriented grating. The color and orientation of the grating was randomly drawn from two distinct feature bins on each trial. A trained support vector machine (SVM) classifier was successful at significantly classifying the task-relevant feature dimension (color, orientation, both) across the whole trial interval including the pre-stimulus interval, although classifier accuracy was higher during the encoding and delay intervals. To examine trial-specific activity reflecting storage of the object's features, the classifier was trained to classify what bin the task-relevant and task-irrelevant feature came from. Interestingly, for orientation, the classifier produced reliably above-chance classification across the delay for the task-relevant feature but not the task-irrelevant feature. Importantly, orientation could be accurately classified on trials for which both orientation and color were remembered. Moreover, classifier evidence was much higher for the correct bin when orientation was task-relevant compared to task-irrelevant during encoding. Above-chance classification for color was only present during the initial 500 ms across all conditions. Our results suggest that both task-relevant and task-irrelevant features are initially encoded, but only task-relevant features continue to be actively represented across the delay period.
Meeting abstract presented at VSS 2018