Temporal dynamics of motion and choice signals in area MT

Sensory encoding is the fundamental front-end to our perceptual experience that ultimately guides our thoughts and actions. In neuroscience, vision has been the primary model system of study for how information about our external environment is represented in the brain. In particular, the study of sensory decision-making has often combined psychophysics and electrophysiology in nonhuman primates to elucidate the neural underpinnings of perception and action. The middle temporal (MT) area has been essential to this process by establishing a causal link between neural encoding of visual motion and behavior. In turn, such work has enabled us to further ask how higher-level brain areas interpret information from early sensory stages to inform decisions. This dissertation builds on that rich history by incorporating a reverse correlation motion stimulus with explicitly controlled temporal statistics, and taking a decoding approach to groups of simultaneously recorded MT neurons to ask the following questions: (1) Can temporal weighting behavior be flexibly adapted to match stimulus statistics? (2) To what extent is temporal weighting behavior a consequence of dynamic sensory encoding? (3) Can we use manipulations of temporal weighting strategy to demonstrate a causal role of choice-correlated activity in MT? (4) How can decoding groups of MT neurons to predict direction of motion vs. decisions elucidate a relationship between sensory noise and choice-correlated activity? First, I will demonstrate that temporal weighting strategies can be flexibly adapted to match stimulus statistics. Second, I will show that changes in temporal weighting behavior do not manifest as strategy-dependent sensory reweighting of the response in MT. Next, by comparing the output of population decoders trained to predict direction and choice, I will show a fundamental lack of overlap between the two signals that suggests a feedback origin for choice probability (CP). Finally, I will show a consistently late-biased time course of CP despite clear changes in temporal weighting strategy that solidify it is not a signature of a feedforward role for sensory noise.