Research InterestsThe main research focus of our lab is how the brain uses information to perform perceptual decisions. We use a combination of psychophysics and computer modeling to explore the information dynamics that contribute to the following issues:
Recognition and Representation
It is well-known that human vision provides a very rich sense of the world around us. And yet as observers, we throw away a significant amount of information, ignoring almost everything that isn't relevant to the task at hand. The question we ask is: For common decisions such as object recognition, how do observers know what to throw away and what to use without first processing everything? Several prevailing theories exist the suggest we are disproportionately sensitive to certain types of information, but how much of this depends on the task and how much is "hard wired" into our representation of object and shape? We address these questions using a novel method we call error-from-sample (aka EFS) that exploits observer errors to estimate how strongly different types of information are represented in the brain, and how much they contribute to recognition.
Neural Network Self-Organization
Of additional interest to our lab is the question of how visual circuitry develops with experience and how this circuitry subserves perceptual decisions. To think about it a different way, perceptual decisions arise from a neural substrate that adapts to both task and experience -- understanding the dynamics of biology can give us critical insight into the dynamics of information utilization. Primary efforts in this direction use computer models to simulate neural network self-organization in response to natural scenes. Following training, networks are run through a battery of visual tasks and performance is compared to human psychophysical data. The strength of these models rests entirely on their biological plausibility so we currently use compartment models based on standard cell dynamics, combined with our own version of a self-organization learning algorithm derived from spike-timing dependent plasticity. The algorithm is constrained by channel density and cell-specific buffering mechanisms and is capable of replicating single-spike and tetanically stimulated plasticity dynamics observed in both pyramidal cells as well as excitatory and inhibitory interneurons.