Stimulus uncertainty and relative reward rates determine adaptive responding in perceptual decision-making

1 In an ever-changing environment, animals must learn to flexibly select actions based on sensory input and the anticipated positive and negative consequences. This type of adaptive behavior is studied using perceptual decision-making (PDM) tasks that feature block-wise changes in stimulus-response-outcome contingencies. Despite extensive research on PDM, there exists no widely accepted mechanistic model of the decision process that captures trial-by-trial adaptation to contingency changes. To address this gap, we first specified three signal detection theory-based models of adaptive PDM. Next, we identified several scenarios in which these models make diverging predictions. For experimental testing, we subjected rats and pigeons to a two-choice auditory discrimination task comprising a sequence of experimental conditions that differed in their stimulusresponse-outcome contingencies. The contingency manipulations were implemented through the concomitant manipulation of reward probabilities, stimulus presentation probabilities and stimulus discriminability across two stimulus-response categories. We find that both rats and pigeons exhibit condition-specific response biases that increase total reward across the entire range of experimental conditions. However, none of the models were able to fit the choice data across all experimental conditions. Through detailed behavior analysis, we demon-strate that learning is driven by the integration of rewards, but not reward omissions. Moreover, model-based analyses reveal that reward integration is influenced by two additional factors, namely perceptual uncertainty and the alignment of steady-state response ratios to relative (rather than absolute) reward differences between the two choices. A model incorporating these factors accounts well for behavioral data across experimental conditions for both species and connects the arguably most influential framework of perception, signal detection theory, with a learning mechanism operating at the level of single trials which, in the steady state, produces behavior consistent with the generalized matching law from animal learning theory. 2 Author summary Humans and other animals rely on their senses and experience to categorize objects and pursue their goals. For example, a mushroom hunter uses sight, smell and touch as well as knowledge of the local biota to decide whether to pick a particular mushroom. The consequences of erring may be dire – food intoxication if savoring a poisonous exemplar, or a meager dinner if too many palatable mushrooms are rejected. Also, the hunter’s decision may be influenced by ambient lighting conditions or his estimate of how likely it is to encounter poisonous mushrooms in a certain area. Our work is concerned with the algorithms that animals use to make such decisions and how they adapt when circumstances, such as stimulus discriminability, change. We show that mathematical models that incorporate the animals’ uncertainty about the type of stimulus currently being perceived make very similar choices as the animals. Furthermore, we find that the animals balance their choices by considering relative, rather than absolute, reward expectations, reflecting a long-standing principle of animal learning theory. Together, these features collectively allow animals to obtain nearly as many rewards as theoretically possible.