Adaptation of response transients in fly motion vision. II: Model studies.

The so-called 'Reichardt detector' can successfully account for many properties of fly motion vision. In its simplest form, the signals derived from neighboring image locations become multiplied after a low-pass filter has delayed one of them. This operation is done twice in a mirror-symmetrical form and the resulting output signals become finally subtracted. As predicted by this model, fly neurons respond to a brief motion pulse with a sudden rise in activity followed by an exponential decay. The time constant of this decay has been shown to shorten when tested after presentation of an adapting motion stimulus. In terms of the detector model this inevitably implies that the time constant of the low-pass filter is adapting. Given that, one would expect a concomitant shift of the steady-state response towards higher velocities, which, however, could not be experimentally verified. Here, we show that given a model with an additional temporal high-pass filter in the cross-arms of the detector, only the high-pass filter determines the time course of the impulse response. Assuming consequently that the time constant of the high-pass filter is the locus of adaptation resolves the conflicts mentioned above. Moreover, such an elaborated model with an adaptive time-constant faithfully mimics a particular contrast-dependency of transient response oscillations observed in fly motion sensitive neurons.