Adaptive gain control during human perceptual choice

Abstract

Neural systems adapt to background levels of stimulation. Adaptive gain control has been extensively studied in sensory systems, but overlooked in decision-theoretic models. Here, we describe evidence for adaptive gain control during the serial integration of decision-relevant information, revealed thorough deviations from optimality. In the experiment observers indicated whether the tilts of a series of visual gratings fell closer to the cardinal axes (0° and 90°) or diagonal axes (45° and -45°). Using a regression based analysis we identified two separate biases: Firstly, observers overweighted evidence that arrived closer in time to the decision (recency bias). Secondly, the impact that each sample wielded over choices depended on its consistency with the previous sample, with more consistent or expected samples wielding the greatest influence over choice. This consistency bias was also visible in the encoding of decision information in pupillometric signals, and in cortical responses measured with functional neuroimaging (fMRI and EEG), whereby the strongest correlation between evidence strength and neurophysiological signal strength occurred for consistent evidence. These data can be accounted for with a new serial sampling model in which the gain of information processing adapts rapidly to reflect the average of the available evidence. Furthermore, this adaptive gain mechanim can be implemented in a biologically plausible population coding model by adjustments to the tuning of neurons coding for expected information. The results are consistent with a dynamic representation of current beliefs that is not reliant on additionally integration stages in which the momentary evidence is evaluated, and is consistent with the rapid influence (<250ms) of prior evidence on future sampling.

Meeting abstract presented at VSS 2014