Object-based Attentional Modulation of EEG Alpha is Related to Task Difficulty

Abstract

Object-based attention is the selection of visual information from an object or object category. It may be deployed in anticipation of a task-relevant object or as a template during visual search. In this study, we investigated whether object-based attention modulates electroencephalographic (EEG) alpha-band activity across the scalp, analogously to observed alpha modulations in spatial and feature-based attention experiments. If endogenous visual attention operates by a common mechanism throughout visual cortex, alpha modulation should occur whenever selective control is exerted over any visual-cortical activity. To test this hypothesis, we collected EEG data from 10 human participants performing an anticipatory object-based attention task with three categories of objects: faces, places, and tools. These object categories were chosen on the basis of their differentiated cortical representations. We observed reduced reaction time for validly compared to invalidly cued trials across all object conditions, suggesting that our task produced the intended attention effect. Although it is not possible using scalp EEG to unambiguously localize an alpha topography to specific brain regions, large changes in the underlying loci of alpha activity corresponding to effects in face, place, and tool regions would be expected to yield different patterns of alpha over the scalp. Using a random cluster analysis over alpha power, we identified electrodes and time points that differed significantly between object conditions, suggesting that alpha topography is modulated by attention to specific categories of objects. Using support vector machine binary classification as an additional measure of alpha topography modulation, we found above-chance decoding of attend-face anticipatory alpha, but no above-chance classification for attend-scene or attend-tool conditions. The ability of a classification algorithm to detect a systematic alpha topography only among the attend-face data dovetails with variation we observed in the reaction time data across object conditions, with the fastest reaction times in the attend-face condition.