Abstract
Spatial attention is thought to play an essential role in selecting relevant information and ignoring irrelevant information. But spatial attention is dynamic, constantly shifting and splitting across multiple objects and locations. How can we measure neural representations of visual features under conditions of dynamic attention, and how do these measurements link with behavior? Both fMRI (e.g., Brouwer & Heeger, 2009) and EEG (e.g., Garcia, Srinivasan & Serences, 2013) have recently been used to reconstruct object features. Here we ask whether these reconstruction techniques can be applied to behaviorally-relevant, attended features (from a multi-item display), and whether the quality of these reconstructions are linked to behavior. In an fMRI task, subjects were briefly shown an array of three colored and oriented gratings. Subjects were then asked to report either the color or orientation of the grating at a spatially pre-cued location. To manipulate dynamic attention, some trials included a second spatial pre-cue at a different location, such that subjects had to covertly shift attention and report the features of the object at the new location. In a similar EEG task, two gratings were shown on the screen, and subjects were asked to covertly attend to one of them to detect subtle orientation changes. Using both techniques, we were able to reconstruct the features of the attended item by applying an inverted encoding model (e.g., Sprague & Serences, 2015). In particular, we achieved reliable feature reconstructions only when the feature was relevant to the current task. Moreover, fMRI neutral reconstruction performance was linked to trial-by-trial behavioral errors. These results emphasize the role of focused spatial attention in the feature-binding process and illustrate the potential of these techniques to provide neural measurements of attended feature representations under dynamic conditions without requiring behavioral responses.
Acknowledgement: NIH R01-EY025648(JG), Alfred P. Sloan(JG)