Temporal estimation by a model big brown bat

Abstract

The big brown bat, Eptesicus fuscus, uses echolocation to locate prey and displays extraordinary acuity in the perception of temporal cues in acoustic signals. Behaviorally the bat can detect changes at submicrosecond levels but individual neurons in the inferior colliculus (IC) and cortex operate with much less precision. Most of these cells are poor temporal markers with response variation on the order of a few milliseconds and some in tens of milliseconds. A temporal estimator was created incorporating the response properties of recorded neurons and behaviorally appropriate limitations on the number of echolocation emissions. The response of the neurons can be characterized as probability density functions in time and frequency. The characteristics of these neurons were used to create large simulated populations of IC and cortical neurons that show the full range of recorded variation. The connections between these two populations were simulated using a self-organizing neural network. If more than one IC neuron is required to trigger the response of each cortical neuron then the model operates with resolution of microseconds. Manipulating the firing threshold of cortical cells and the relative population sizes influences the errors in target estimation.