ABSTRACT

Decisions typically require information to be gathered from multiple sources, evaluated and combined. Canonical views in sensory neuroscience argue that human perceptual classification judgments are optimised by lossless integration of reliability-weighted evidence, furnishing choices that match those of an ideal observer. Here, we describe a series of experiments in which participants viewed multiple discrete samples of information, either simultaneously or sequentially, and made a judgment based on their average. We show that the impact (weight) that information carries on choice varies with slow fluctuations in cortical excitability over the parietal cortex, leading to a rhythmic “bottleneck” that gates the passage of information into a accumulator signal in motor beta-band activity. Dividing attention between two streams does not preclude the conversion of perceptual to decision signals in the parietal cortex, but allows information to leak away from the accumulator. Finally, I will describe a computational model of adaptive gain control in human perceptual categorisation, with evidence from behaviour, pupillometry, and functional neuroimaging.

Funded by

European Research Council