Signal transformation and noise correlation in the primate dorsal stream during sensory decision-making

Neuroscientists have long sought a link between the activity of single neurons and our thoughts, perceptions and ultimately our mental experiences. As our senses provide the input into the brain, understanding the computations that transform signals along the sensory pathways has remained central to this endeavor. Remarkable progress has been made by studying neural correlates of perceptual decisions in motion-processing and oculomotor areas of the primate brain. In particular, when monkeys indicate their decisions about the direction of motion with eye movements, neurons in the middle temporal area (MT) represent the instantaneous motion evidence and neurons in the lateral intraparietal area (LIP) resemble the integration of motion evidence, effectively transforming the sensory signal into a decision variable. In the main body of this thesis, I describe the results of an effort to measure the sensorimotor transformation between MT and LIP on single trials. First, I describe a motion-discrimination task that is amenable to reverse correlation analysis, allowing the experimenter to measure the temporal dependencies of neural responses and choices on the instantaneous motion energy. I then use a unified statistical framework to analyze simultaneous recordings from both areas during decision-making. Primarily, I found that MT neurons exhibited time-varying sensitivity to motion direction, with important consequences for the behavior and neurophysiology in downstream areas. Individual LIP neurons also carried a signature of an integrated motion signal in their spike rates, however, it was unlikely that this signal results from direct MT input. Finally, I show that a biologically plausible simple decoder can perform as well as the monkey at coarse direction-discrimination task. In the appendix, I describe the results of pharmacological inactivations of MT and LIP and statistical models of single trial dynamics in LIP that were performed in collaboration with fellow graduate students, Leor Katz and Kenneth Latimer, respectively.