Mathematical Biology Seminar Will Nesse University of Ottawa Centre for Neural Dynamics Wednesday, Sept. 2, 2009 3:05pm in LCB 225 The ABC's of the Neural Language: Independent Statistical Decomposition of Adapting Spike Dynamics Abstract: Extracting sensory information from the environment by neurons is limited by how much information can be captured by neural spike trains. Sensory systems that can encode more information than others have a survival advantage. Several recent studies have established that fine-grained information from the specific temporal structure of sensory spike trains appears necessary for encoding weak, behaviorally important stimuli. Still, how this information is captured by a neuron's biophysical machinery is unresolved. Commonly, spike generation by neurons is modulated by slow activity-dependent currents that act as a negative feedback, retarding subsequent spiking. Such adaptive currents induce correlations and rich temporal relationships in the spike emission. Using mathematical analysis, and preliminary experimental observations by my collaborator Gary Marsat, I show that spike emission governed by realistic adaptation currents leads to a handy probabilistic decomposition of the spike train into statistically independent components. Using this decomposition, we establish novel information-theoretic properties of these naturalistic spike trains. Further, I argue that while realistic correlated spike trains posses more redundant information overall, that the information is in a more accessible form for subsequent stages of sensory processing.