Research
As we look at the world around us, we have immediate access to the composition of the visual scene into objects, as well as our relationship in space to those objects.  This natural facility makes it possible to move though the world, catch or avoid moving objects, and base immediate decisions on a detailed understanding of the world around us.  Only secondarily might we note the particular color or composition of particular points in the visual scene, or the angle between two contours resulting from one object partially occluding another, or other low-level visual features of the scene.  Our brain processes vision much differently than computers do: a computer can easily store the hue and luminance of every pixel of an image, but – even with the best available software – it cannot parse an arbitrary natural image into its underlying elements.
 
However, in the absence of larger conceptual theories of how the brain processes information, established techniques have revolved around studying sensory systems’ abilities to represent information -- rather than understand the computation that it performs.  In order to study computation in the brain, it is necessary to both establish larger theories about what is being computed, and design experiments to link these larger theories to observable physiology.  
 
Research in the NeuroTheory Lab is concerned both with developing larger theories of system-level function in the visual and other sensory systems, as well as working closely with neurophysiologists to design and perform experiments that can guide and/or validate these theories.  As a necessary third goal, we also develop new analytical tools to facilitate these new experiments, as well as increase what can be learned from existing experiments.  Work in this lab is divided into four main areas:
 
In order to understand computation in the visual cortex, it must be studied in the context where it performs this computation.  Natural vision is a broadly defined category, but necessarily includes two important elements: (1) visual "features" that are coherent in space and time, and (2) natural time-evolution of the scene due to motion within the scene, motion of the observer, and eye movements. Using naturalistic visual stimuli in experiments brings up many complications due to their correlations and more general spatiotemporal complexity, and require robust theoretical methods to untangle which spatiotemporal features are eliciting neuronal responses.
 
Even when we look at a static image, its projection onto our retina is in constant motion due to eye movements that include saccades and fixational drift.  Counterintuitively, rather than making it harder to see, this lack of stabilization of an image on our retina is an essential part of perception.  In fact, other senses (including sound and active touch) result in naturally time-varying signals, implying that the temporal structure explicitly imparted by the visual system might be a “formatting” in order to enable cortical computation.  We thus hope to gain insight into cortical computation by studying temporal processing in the visual system.
 
A crucial barrier to our ability to characterize more general computation performed in the brain is our reliance on linear methods of characterizing neurons.  While linear methods often provide a great first-order description of neurons early in the visual pathway, computation in the visual system surely occurs through the successive application of non-linear operations.  We are currently development methods to characterize non-linear neural computation based on physiological data.
 
A complementary part in understand the function of the visual system -- and neuron’s role in the larger system-level behavior -- is understanding its structure and connectivity.  Rather than being genetically hard-wired, the cortex follows rules applied at a single-neuron level to determine the correct wiring.  By understanding these rules, we might be able to understand the organization of the cortex on a system-level, and connect this structure to its function.  Likewise, it is likely that changes in connectivity is actually a part of the mature visual system.