My research is in the field of computational neuroscience, an interdisciplinary subject that uses computer and mathematical models to understand a variety of brain functions. Experimental neuroscientists have learned a tremendous amount about how individual neurons function and about how pairs of neurons can communicate with each other. However, most brain functions involve the interactions of many thousands (or far more) neurons which are themselves incredibly complex and dynamic units. By modeling and simulating data from neurophysiology experiments, I seek to reveal mechanisms underlying complex brain functions.

I am currently most interested in two questions:

I. How do groups of neurons store short-term memories?
Short-term memory is believed to be maintained by neurons that exhibit a sustained level of activity for many seconds following the removal of a remembered stimulus. A classic example of short-term memory is looking up a phone number in the phone book, remembering it just long enough to reach for the phone and dial it, and then forgetting the number a few seconds later. During the period of remembering the phone number, neurons in the brain are believed to sustain an activity pattern that represents the phone number. When this activity terminates, the memory is gone. I am examining how the biophysical properties of single neurons, the synaptic interactions between neurons, and the patterns of connectivity between networks of neurons contribute to the maintenance of short-term memories.

My present work focuses on the vertebrate brainstem region known as the oculomotor neural integrator. This network is crucial both in producing accurate eye movements and in holding the eyes still when focusing on an object of interest. The integrator network receives signals from other brain areas that encode the desired velocity of impending eye movements. It then converts these desired-velocity signals into desired-eye-position signals that are sent on to the motor neurons that control the tensions of the eye muscles. The network is called an "integrator" because the transformation from eye-velocity-encoding signals to eye-position-encoding signals corresponds to the mathematical operation of integration.

The oculomotor integrator is a model system for studying the mechanisms underlying short-term memory because, in response to a transient eye movement command signal (the stimulus), integrator neurons respond with a sustained change in their activity level (the memory). Because the sustained level of activity is proportional to the intended position of the eyes in their orbit, integrator neurons are said to maintain a "memory of eye position".

Previous modeling of this system has shown how a network of relatively simple model neurons can convert an eye velocity input signal into an eye position output signal. However, these models require extreme fine tuning to perform correctly. I am modeling single neuron properties that lend robustness to the neural integrator network. Synaptic dynamics, dendritic properties, and plasticity mechanisms are being explored as means of reducing the need for precise tuning of network parameters.

II. How do neurons in the sensory systems represent information about the world around us?
Illusions. The study of illusions provides strong constraints on the processing of sensory information: whereas many models might succeed in describing normal sensory function, not all models will produce the errors in perception that underlie illusions. In one project, I am modeling the neural basis of a visual masking illusion. In visual masking illusions, a prior (forward masking) or subsequent (backward masking) stimulus can make a neighboring stimulus appear less visible. Recent neurophysiological recordings of cells in the early visual system have identified a neural correlate of such masking illusions (for a cool demonstration of the illusion I am currently modeling, see http://neuralcorrelate.com/bni_steve.htm#The_Standing_Wave_of_Invisibility). I am currently testing whether classic models of the responses of early visual neurons can explain the neural responses to such illusions and, in cases when they cannot, am proposing modified models that better account for the observed data.

Information processing in the brain. In another project, I am using methods from information theory to understand how information about the sensory world is encoded by and transmitted through the nervous system. I have recently analyzed a simple model of the synaptic transmission process that provides insight into how the temporal dynamics of the synaptic release process affects the flow of information across the synapse. More recently, I have been applying methods of information theory to the analysis of neuronal tuning curves to determine which stimuli are best encoded by sensory neurons.