Thesis defense Julian Jonathan Tramper (Donders Series 106)

April 24, 2013

Promotor: Prof.dr. C.C.A.M. Gielen

Feedforward and feedback mechanisms in sensory motor control

This thesis aims at a better understanding of how the brain combines
sensory feedback with feedforward predictions to guide movements.
Therefore, we investigated human motor behavior when subjects were
moving their finger along three-dimensional objects, were playing
video a game, had to steer a randomly moving object to a target, or
were touching surfaces of unknown objects. During these movements, the
brain cannot solely rely on feedback from the sensory system, because
this information is not up-to-date due to sensorimotor delays. In
addition, the brain has to deal with noise in the neural system and
uncertainty about the environment. In order to deal with these
uncertainties, the brain combines sensory feedback with feedforward
predictions.

In this thesis, we showed that during tracing of a complex
three-dimensional path with the finger, a combination of saccades and
vergence eye movements brings the eyes ahead of the finger to explore
the upcoming trajectory. In addition, we found that even smooth
pursuit eye movements could be directed ahead of moving objects,
instead of being driven by the movement of a visual target on the
retina. When subjects were moving their finger along a surface with
their eyes closed, they could not anticipate abrupt changes in surface
curvature. However, they could anticipate smooth changes in curvature
using feedforward predictions of the somatosensory system.

To reveal the neurocomputational mechanisms underlying sensorimotor
behavior, we modeled eye-hand coordination in an obstacle-avoidance
task, taking into account observation noise of the hands and
obstacles, as well as uncertainty about the locations of the objects.
The model correctly predicted the order and location of fixations, and
the hand trajectories. We also demonstrated that models optimizing the
trade-off between effort and performance cannot explain motor behavior
under uncertainty in time-constrained tasks. We postulated a new model
that also took into account the time-integrated position error of the
movement. This model correctly predicted the observed movement
trajectories for different levels of uncertainty