Abstract
Humans can readily distinguish durations of sensory events. However, in contrast to many other basic properties of visual stimuli, such as motion, colour, orientation etc., little is known about the neural encoding of time. Here we present a simple model for encoding the duration of a visual stimulus, using a low pass filter approximating the temporal response of sensory neurons and a step-function nonlinearity (threshold gated output). The time constant (τ) of the low-pass filter determines its response gradient to an input and thus determines the duration between the signal onset and the time at which the threshold is reached and the triggering of an output response. A population of such duration detectors with a range of time constants and fixed thresholds can encode the duration of an input signal in the form of a labelled line set, where the time constant of the most recent detector to reach threshold indicates duration of the signal. This population encodes an on-going estimate of duration since the stimulus onset as well as the final duration at stimulus offset. This mechanism has properties reflecting Weber's law and shows a reduction in its duration estimate when a detector sub-population is reduced in sensitivity (representing adaptation), which corresponds to perceptual effects reported in the literature (Johnston, Arnold, & Nishida, 2006). Although amplitude and frequency of the input signal do not affect duration measures, duration measures in the simplest model increase proportionally with mean input signal intensity. Consequently, we use divisive normalization to keep the mean input signal intensity constant, whilst preserving the signal waveform shape. This work demonstrates that a very basic model, based on the simple temporal response properties of sensory neurons, is capable of encoding an explicit measure of duration and of modelling related psychophysical phenomena.
Meeting abstract presented at VSS 2014