Our work aims to uncover neural algorithms of information processing. We are particularly interested in the way in which neural coding and representation is affected or shaped by the complex statistics of the natural world.

Adaptive coding

Neural systems are context-sensitive: the encoding of a stimulus depends on the temporal, spatial and semantic context within which it is embedded. In many systems, typical features that are represented adjust according to the signal-to-noise ratio, and the coding range adapts to the typical range of stimuli. What are the rules that govern these processes? What mechanisms support it? How long does adaptation take? Why does it take that time?

Multiple timescale adaptation

Different processes encode information at different timescales. Alongside fast adaptive processes that normalize response functions to the scale of the stimulus, there are slower processes that depend on the history of temporal changes in stimulus statistics. We explore this dependence from a variety of perspectives: finding functional descriptions of the behavior, determining underlying mechanism, and deriving normative models that may explain the observed behavior.

From single neurons to networks

What can single neurons compute and how is this governed by the underlying biophysics? We aim to translate complex conductance-based models into functional models that provide an intuitive understanding of the input features that drive neuronal responses.

We have recently shown that during development, single neurons acquire the ability to normalize inputs. How does this property arise from their evolving biophysics? What impact does this change have on the network’s ability to transmit information on different timescales? In collaboration with the Moody lab, we examine the spatial and temporal distribution of intrinsic properties and connectivity in cortex during development, and aim to relate this to single-unit participation in large-scale waves of activity.

Transformation from sensory input to motor output

In collaboration with the Daniel and Dickinson labs, we are interested in how sensory inputs change when animals are engaged in active behavior, and how they are converted into motor commands.

Learning to produce and maintain reliable motor output

Birdsong has become an excellent testing ground for theories of learning. In collaboration with the Perkel and Fee labs, we are investigating how the thalamocortical-basal ganglia circuitry of the song system is able to selectively modulate the variability required for lifelong skill learning and maintenance. In collaboration with the Lois and Gardner labs, we are building models to explore how the robustness of the network is maintained.

Thermosensation in foraging mosquitoes

Foraging mosquitoes find food sources in the environment using thermal and chemical cues that have a complicated intermittent structure due to turbulence. How do insects use this intermittent information to locate sources? With the Daniel, Riffell and Dickinson labs, we study the behavioral and neurophysiological responses of mosquitoes to temporal patterns in heat and CO2 plumes.

From neural firing to behavior in Hydra

The small, transparent cnidarian Hydra is an excellent model system in which to connect neural activation patterns with behavior. Our lab are working on models to understand how neural inputs drive muscle activation and the biomechanics of the body.

Navigation and cognition in virtual reality environments

Primates, like humans, learn to navigate and interact with virtual reality environments. We are collaborating with the Buffalo lab to investigate the behavioral strategies and neural representations of spatial environments as animals forage for reward.