"Adaptive gain control during human decision-making"

The gain of neural responding adapts 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 rapid gain control during a categorisation task in which requires observers to average information occurring a rapid stream. The impact that each sample wielded over choices depended on how surprising their associated decision information was, given the statistics of stimulation, with consistent or expected samples carrying more weight. This bias was visible in the encoding of decision information in pupillometric and electroencephalographic signals. We account for these data with a new serial sampling model in which the gain of information processing adapts to ensure that expected information is more diagnostic of behaviour.