Global behavior of neural error correction

Neural information coding is often characterized as noisy and unreliable, typically because spike trains are irregular in appearance and experimental stimuli presented at different times produce different responses. This paper describes ongoing work investigating the effects and significance of errors in spike trains. More specifically, here we test the dependence of error correction in neural coding on a postsynaptic neuron's extant behavior, described in nonlinear dynamical terms. We show that the time for a neuron to recover from an error varies significantly, depending on its preceding stationary behavior and that behavior's position within the cell's global bifurcation structure. This implies a model of neural computation based not solely on attractors or motion within a state space, but rather motion within a global response space.