Stochastic Models of Executive Control in Monkeys and Humans |
Supported by NSF Grant BCS-0218507 / Collaborative Research in Computational Neuroscience ProgramGordon Logan, Thomas Palmeri, and Jeffrey Schall
|
Gordon Logan | |
Thomas Palmeri | |
Jeffrey Schall | |
Project Summary
The hallmark of primate intelligence is the ability to respond flexibly, focusing on different aspects of the same stimulus situation to produce arbitrary responses. Our project is an interdisciplinary investigation of humans and monkeys to elucidate how executive control over attention, categorization, and response preparation are instantiated in neural processes underlying adaptive behavior. Humans and monkeys perform tasks that require them to make a saccadic eye movement toward (prosaccade) and away from (antisaccade) a target that appears among several distractors in a visual display. Experimental variables are manipulated to selectively influence attention, categorization, and response preparation. The activity in ensembles of neurons are recorded in the frontal lobe of monkeys performing the task. The overt behavior of humans and monkeys and the underlying neural behavior in monkeys are described in terms of a mathematically precise computational theory with three distinct components. (1) an attention component selects behaviorally-relevant target stimuli from a field of distractors, (2) a categorization component selects goal-relevant interpretations of target stimuli, and (3) a response preparation component selects responses necessary to fulfill goals. The theory provides a mathematical description of each component and of their interactions which affords (1) a way to identify each component with experimental operations that selectively influence different aspects of behavior and (2) a way to identify each component with signals produced by neurons with different functional properties. The proposed research provides a novel integration of successful but separate programs of research in computational models of human cognition (Logan and Palmeri) and behavioral neuroscience (Schall).
Single-cell neurophysiology by itself is limited by a lack of linking propositions that bridge the gap between single-cell behavior and psychological states of the cognitive processes that the single cells implement. The cognitive model will provide quantitative descriptions of the computations underlying psychological states as well as quantitative descriptions of the neural processes that underlie them. Computational cognitive modeling by itself is limited by a mimicry problem that stems from an inability to "open the black box" and observe the inner processes that underlie behavior. Computational models with very different underlying processing assumptions often predict the same overt behavior. Our research confronts the mimicry problem by identifying underlying processes with neural behavior, allowing distinctions between models that produce the same overt behavior.
Project Members:
Gordon Logan,
PI
Thomas
Palmeri, co-PI
Jeffrey Schall,
co-PI
Leanne Boucher, Postdoctoral Fellow
Alex Gotler, Postdoctoral Fellow
Corrie Camalier, Graduate Student
Hari Kannan, Computer Programmer
Links to Specific Projects
Under Construction
Department of Psychology, Vanderbilt University
301 Wilson Hall -- Nashville, TN 37203
Phone: 615.322.2874 / Fax: 615.343.8449