Complementary attentional mechanisms for the resolution of representational ambiguity in the human brain

Abstract

When we view cluttered environments containing multiple objects, the neural code for individual items can become ambiguous. This reflects a capacity limit in spatially-tuned visual neurons: when multiple objects fall within the cell receptive field, output cannot be attributed to a single object. Vision models from primate electrophysiology propose this is resolved by attention, which biases competition between object representations. However, evidence in humans is sparse and inconclusive, and reliance on single-cell data limits insight into underlying mechanisms and neurophysiological scope. Here, we use a novel multivariate approach to the analysis of human EEG and concurrent EEG/MRI to address this. First, we test whether attention is recruited by representational ambiguity. Second, we identify the mechanisms that act on representations of attended and unattended objects to resolve ambiguity. Finally, we characterize the millisecond timing and whole-brain action of these mechanisms to identify pervasive effects in semantic and executive brain networks.