A major part of understanding any neural system is to comprehend how it manages to perform useful computation in a dynamic world, in which properties of the sensory input may change by orders of magnitude from moment to moment.  Knowing how the retina adapts to light level or how colour perception changes with the prevailing tint of an image are important parts of our current understanding of the visual system.

In general, adaptive mechanisms are a fundamental part of biological sensory systems.  The aim of my work is to create a new framework for understanding adaptation in the temporal properties of the visual system.  I am carrying out an integrated set of electrophysiological and modelling studies which focus on the recently discovered phenomenon of adaptive temporal integration (ATI): motion sensitive cells in the primary visual cortex (V1) and the cortical motion area (V5/MT) adapt their profiles of temporal integration in response to changes in the speed, spatial structure, and contrast of moving visual stimuli.  My work centres on an experimental investigation of ATI, but it includes the development of a modelling framework to facilitate the unification of ATI with related adaptive phenomena in the cortex and the retina.

The temporal properties of receptive fields have received far less attention than spatial properties but are just as critical for understanding cortical function.  Whereas the spatial structure of receptive fields reflects the spread and selectivity of synaptic connections, temporal structure reflects biophysical aspects of neural transmission, such as the temporal dynamics of conductances and time constants of membranes. 

Temporal response properties are important because they determine when a cell will respond and to what epoch in the past the response refers – issues that are fundamental to neural coding.  It is not therefore enough to understand ATI in a qualitative sense from the experimental studies; the final goal is to produce a working model that operates on arbitrary dynamic visual stimuli.  The model should make precise predictions about the timing of individual action potentials and thereby advance our understanding of neural circuitry, neural coding and visual perception.

Current Research Programme

My research programme involves answering the following questions:

• What is the functional cortical circuitry that underlies the perception of visual motion?
• How are the major classes of spiking neurons along the pathway from the retina, through V1 to V5/MT, connected into a functional neural circuit?
• How can this circuit (or circuits) account for the adaptive responses of direction selective neurons in V1 and V5/MT to dynamic visual stimuli?

The answer will be delivered in the form of a publicly available online model that will accept arbitrary visual stimuli and produce neuronal responses like those recorded experimentally.

The work is significant because:

• Understanding functional circuitry is one of the most important endeavours in neuroscience because of the possibility of uncovering computational motifs, or canonical circuits, that may repeat across brain regions.  It addresses the fundamental question, “What is the nature of the neural code?"  It is particularly important to synthesise a functional understanding of cortical architecture in V1, where extensive knowledge of anatomy and response properties help constrain candidate circuits.
• Studying functional circuitry in the motion pathway is important because signals in V5/MT have been and continue to be linked to motion perception.  This presents the opportunity to study how local computations, arrayed throughout the system, can influence perception.
• It is critical to reformulate our notion of visual receptive fields because current models rely too heavily on fixed filters and fail to account for adaptive properties of neuronal responses under changing visual conditions.  This leaves us with no prediction to compare to experimental results for novel stimuli.
• Answering this question with a working model is important because it will provide a wealth of knowledge about the visual system in a usable form.  This benefits researchers who carry out visual psychophysics, functional magnetic resonance imaging (fMRI) and theoretical modelling because they can see directly how their visual stimuli are expected to activate major classes of neurons in the visual system.
• It is important to try this approach now because computers have only recently become powerful enough to simulate, in reasonable time, models of the complexity required to process extended three-dimensional image sequences with biophysically plausible populations of thousands of neurons.
• The model can help experimentalists design more efficient experiments and thereby use fewer monkeys.

Further information can be found at: http://www.physiol.ox.ac.uk/~wyeth/

Wyeth Bair