Shadmehr Laboratory
Laboratory for Computational Motor Control · Johns Hopkins University
Kurt Thoroughman

Kurt Thoroughman

PhD (1999)

Research

Kurt was the first graduate student to join the lab. He had just completed his BS in Physics from University of Chicago and enrolled in the Hopkins BME PhD program. He joined in summer of 1995 at a time when the lab consisted of an empty room. He was instrumental in setting up the robotic apparatus and building the laboratory. He transformed the direction of motor learning by inventing the state-space approach to analyze trial-by-trial behavioral data during motor learning. With it, he showed that behavioral data carries some of the signatures of the motor primitives with which the brain builds internal models of action. He also discovered what may be the teaching signal for these internal models: the error-feedback response generated by the reflexes. He completed his PhD in 1999, and subsequently did a postdoc at Brandies University with Prof. Eve Marder. He then became a faculty member at Washington University, where he is currently an Associate Professor of Biomedical Engineering.

Ideas developed by Kurt

  • Providing physiological evidence for internal models in motor learning. Kurt showed that as people learn to move in a novel force field, muscle activity changes in a systematic way consistent with the formation of an internal model of the new dynamics. In particular, learning produced changes in the directional tuning of muscle activity, providing a physiological correlate of the internal model inferred from behavior.
  • Discovering the motor primitives underlying learning. Kurt developed an innovative trial-by-trial analysis in which errors experienced on one movement were used to determine how learning generalized to subsequent movements. This allowed him to infer the underlying basis functions, or motor primitives, from behavior. He found that these primitives have broad, Gaussian-like tuning with respect to movement velocity.
  • Showing that the structure of neural representations constrains what we can learn. The inferred tuning of the motor primitives predicted that some novel dynamics should be much more difficult for people to learn than others. Experiments confirmed these predictions, establishing a link between the representation used by the nervous system and the limits of motor learning. The similarity between the inferred primitives and neural tuning functions also suggested a possible neural implementation of internal models.
  • Selected publications

    Shadmehr R, Thoroughman KA. Learning and memory formation of arm movements. In: Biomechanics and Neural Control of Movement, JM Winters and PE Crago (eds). Springer-Verlag. 2000; pp. 347–353.

    Thoroughman KA, Shadmehr R. Electromyographic correlates of learning an internal model of reaching movements. Journal of Neuroscience. 1999;19:8573–8588.

    Gallery

    ← Return to lab members