Project Summary
Our ability to flexibly select, based on context, the relevant information to form decisions is a fundamental
cognitive process, yet its underlying neural mechanisms are still largely unknown. Most of our knowledge at
single-neuron resolution is derived from a few recent studies in macaque monkeys, where animals were trained
to perform tasks requiring context-dependent selection and integration of sensory information. In these tasks,
feature selection is performed within a sensory modality, and the strength of the evidence can be precisely
titrated by the experimenter. Despite the groundbreaking impact of these studies, so far only a handful of brain
areas have been studied, and no perturbation experiments have yet been performed. To address these issues,
we will train rats to perform context-dependent selection and integration of sensory information, and we will
leverage the powerful experimental tools available in rodents to dissect the neural circuits responsible for this
behavior. First, we will develop a new pulse-based task requiring context-dependent selection and integration of
sensory evidence. In our task, rats will be presented with a train of randomly-timed auditory pulses, where each
pulse varies in its location (right or left) and its tone frequency (high or low). In separate blocks of trials, rats will
be cued to report either the prevalent location of the pulses, or their prevalent frequency. Because in this task
information will be presented to subjects in highly random but precisely known pulses, over the course of
hundreds of thousands of trials we will have access to a highly varied sample of stimuli allowing us to fully
characterize temporal dynamics of the processes involved. Second, we will develop an automated, high-
throughput training pipeline to efficiently train rats to perform the task. Third, we will develop a complementary
automated high-throughput setup for the collection of electrophysiology and optogenetics data during behavior.
Fourth, we will develop a series of novel modeling and statistical analyses to leverage this rich dataset to probe
the underlying computational mechanisms. The proposed approach aims to provide an unprecedented platform
to probe the neural mechanisms underlying flexible decision-making.