Ruediger Krahe, Gabriel Kreiman, Fabrizio Gabbiani, Christof Koch, Walter Metzner
Annual Meeting of the Society for Neuroscience, Miami 1999
Pyramidal cells in the electrosensory lateral line lobe (ELL) of the weakly electric fish Eigenmannia process amplitude modulations (AMs) of the fish's own electric field induced by nearby objects or conspecifics. Using a combination of neuroethological and information theoretical methods, it has been shown that single pyramidal cells only poorly encode the exact time course of AMs but rather extract information on upstrokes or downstrokes in amplitude (E-units and I-units, respectively; W.Metzner et al., J.Neurosci. 1998 18:2283-2300). Since the receptive fields of neighboring pyramidal cells overlap, it is conceivable that stimulus properties are encoded by populations of pyramidal cells. We therefore extended our original approach to simultaneous recordings from multiple pyramidal cells while presenting random AMs of the electric field. Linear decoding from the simultaneous spike trains of two neurons only marginally improved the decoding from single neurons, still yielding a poor encoding of the stimulus waveform. To assess the feature-extraction performance of multiple neurons, we constructed the Euclidian classifier for stimulus waveforms preceding bins containing no spikes and stimulus waveforms preceding bins containing various different types of spikes. Spikes coinciding in pairs of E or I-type neurons within a small time window (10 ms) indicated the occurrence of up- or downstrokes of the AMs more reliably than spikes coinciding within larger time windows (25 ms up to 500 ms). Furthermore, these coincident spikes performed better than spikes from bursts of a single unit, which in turn performed better than isolated spikes. These results confirm that electrosensory processing in the ELL does not preserve details of the stimulus time course. Instead, ELL neurons may signal behaviorally relevant features of the stimulus by burst-like spike patterns of single pyramidal cells and by correlated activity accross sets of neighboring neurons. Supported by NSF
Top