Seminar with Grant Mulliken, Ph.D.
“Online Control Signals for Goal-directed Movements in Posterior Parietal Cortex: Encoding and Decoding”
Thursday, April 26th, 2007, 12:30 PM
McGovern Institute for Brain Research
Seminar Room 46-3189

Grant Mulliken, Ph.D.
Computation and Neural Systems
California Institute of Technology

Online Control Signals for Goal-directed Movements in Posterior Parietal Cortex: Encoding and Decoding

During visually guided movements, primates are able to rapidly and accurately control an online trajectory despite substantial delay times incurred in the sensorimotor control loop. To address the problem of large delays, it has been proposed that the brain uses a forward model of the arm to predict unperceived states of the trajectory. Psychophysical and clinical studies in humans have clearly established a role for the Posterior Parietal Cortex (PPC) in the rapid online updating and correction of movements. However, much less is understood about how PPC neurons continuously encode sensorimotor parameters during online control of a visually guided movement. Using an information theoretic approach, Dr. Mulliken will show that neurons in PPC contain predictive properties while monkeys perform a joystick trajectory task, consistent with this area serving as a forward model for online sensorimotor control. He will report a population of neurons that best encodes recent, current, and upcoming states of the movement angle of the cursor, which reflect a prediction of the cursor state that is neither directly available from sensory feedback nor compatible with outgoing motor commands. He will also report a population of PPC neurons that persistently encode an online estimate of the target direction, which may be used to refine the estimate of the cursor state. Finally, in addition to encoding analysis, Dr. Mulliken tested how well trajectory information can be decoded from neuronal populations in PPC. Offline reconstruction of trajectories accuracy exceeded an r-square of 0.73 while real-time brain control performance was above 70% for 8 random targets.