Identifying the neural mechanisms of perceptual decision-making in a rat auditory task

Abstract

Perceptual decision-making involves accumulation of sensory information over time. The classical view indicates that sensory neurons first transform the physical stimulus into evidence, and then the decision making areas of the brain use this evidence to make a categorical choice. Here, we show that this process is not optimal. In general, rats do not equally weigh the stimulus and they underweight extreme stimuli values. These results indicate that there must be a non-linearity in the stimulus transformation into evidence or that the accumulation of evidence is not perfect. To differentiate which are the underlying mechanisms that are causing these behaviours, we fitted an extended neurobiological model which was composed of a stimulus to evidence and a decision module. Using model comparison methods, we were able to identify that a non-linear stimulus to evidence transformation is relevant to explain the psychophysical kernels of the experimental task, but that time adaptation of the sensory neurons does not have an impact in this case. We also found that, at least for a subset of rats, the non-linear dynamics in the decision module explains the experimental data better than a linear perfect integration model. Our work shows that model fitting is a powerful tool to investigate different brain mechanisms and that non-linear dynamics could be important not only during the accumulation of evidence but also during the stimulus to evidence transformation.