Coding of information in models of tuberous electroreceptors.

Weakly electric fish continuously emit a quasi-sinusoidal electric organ discharge (EOD) to probe their near environment (electrolocation). P-type tuberous receptors located on their skin respond to amplitude modulations of the EOD by varying their firing rate. These receptors, and the neuronal circuitry downstream from them, must encode and process low-frequency stimuli due to prey and obstacles and certain communication signals, as well as high-frequency communication signals emitted by other fish. We ultimately seek the biophysics that govern the encoding process, and in particular, the sensitivity to certain stimulus features. Since the pyramidal cells to which these receptors project can also be monitored, studies of weakly electric fish offer a great opportunity for deciphering the encoding/decoding problem. Here we briefly summarize our recent advances on this issue. We then present new results on the encoding properties and relative modeling advantages of two widely used classes of neuron models of electroreceptor activity: a leaky integrate-and-fire dynamical model, and a non-dynamical modulated stochastic point process model. The quality of encoding, based on the stimulus reconstruction method, is assessed as a function of firing rate and stimulus contrast, in the context of bandlimited Gaussian stimuli. Our main conclusion is that the quality of encoding increases strongly with firing rate, but also depends on the actual combination of biophysical parameters that determine this rate.