Home Research Papers Courses People Presentations Contact Related Sites

RESEARCH
Our research is currently focused on two important and vigorous areas of computational neuroscience: multisensory integration and motor learning. Our models of multisensory integration are focused on the superior colliculus, which is the first structure in the mammalian brain where there is a major confluence of input from the various sensory systems. Our models of motor control are focused on the cerebellum, which is the main structure responsible for experience dependent plasticity of motor behaviors.
Multisensory enhancement is a property by which the response of a superior colliculus neuron to a stimulus of one sensory modality is augmented by a stimulus of another modality. We have modeled this phenomenon using probability theory. An information theoretic analysis of the model demonstrated that, contrary to intuition, input of another modality does not always increase the amount of information received by a neuron. This may explain why many neurons in the colliculus are unimodal despite the availability of multimodal input. We also demonstrated analytically how simple neural models could implement the probabilistic computations we envision for colliculus neurons. These simple neural models can simulate pharmacological and behavioral observations on superior colliculus neurons.

We have proposed two completely new models of cerebellar learning. In the first, the cerebellum is modeled as a pattern correlator. The model can simulate both the linear and nonlinear aspects of habituation of the vestibulo-ocular reflex. In the second, cerebellar learning is guided by error signals carried over the same neural pathway (the parallel fiber pathway) that carries other signals. The climbing fibers provide only synchronization pulses. The new model provides an alternative to the paradigm in which error signals are carried by climbing fibers. Recent experimental work (by others) reveals a diverse range of possible functions for climbing fibers, and weakens support for the hypothesis that climbing fibers carry error signals. The new model can unify findings on climbing fibers.

Illinois "I"
Copyright 2001 Thomas J. Anastasio, University of Illinois at Urbana-Champaign

Forward all questions and comments to: Thomas J. Anastasio.