What kind of movement detector is triggering the landing response of the housefly?

Abstract

As shown before, the latency of the housefly's landing response depends on the conditions of the visual stimulus (Borst 1986). Accordingly, the latency can be used to characterize the movement detection system which is triggering the landing response.

The stimulus was a sinusoidal periodic pattern of vertical stripes presented bilaterally in the frontolateral eye region of the fly. It started to move, simultaneously on either side, from front to back at a given time. The latency of the response was measured by means of an infrared light-beam that was interrupted whenever the fly lifted its forelegs to assume a preprogrammed landing posture (Fig. 1). The latency was found to vary in a range from 60 ms up to several seconds depending on the pattern's spatial wavelength λ, contrast frequency cf and contrast C.

For sufficiently high pattern contrast the optimum of the reaction (minimum latency) is found at spatial wavelengths of 30–40° and contrast frequencies of 8–17 periods/s (Fig. 3a). This is about 2–10 times more than is anticipated from the optomotor response under similar conditions. Evaluation of the optimum contrast frequency cf _OPT at different wavelengths shows that cf _OPT is not independent of λ (Fig. 3b, solid line). The same is true for the contrast dependence of the reaction: reduction of the contrast leads not only to a general decrease in the response amplitudes (prolongation of the latency) (Fig. 4a), but also to a shift of cf _OPT towards lower contrast frequencies (Fig. 4b, solid line).

In the theory of the correlation-type movement detector (Reichardt 1961) which underlies the optomotor response of flies the dependence of cf _OPT on pattern wavelength and/or pattern contrast is not expected under stationary conditions. However, as shown by computer simulation all experimental results can be explained by a homogeneous retinotopic array of correlation movement detectors (Fig. 2) if their response under non-stationary conditions is taken into account. We simply assume that the spatially and temporally integrated output of the movement detectors is evaluated by a threshold device (Fig.5). The correlation-type movement detection in combination with a temporal integrator system predicts the rather complex dependence of the optimum contrast frequency on pattern wavelength and pattern contrast (dashed lines in Fig. 3b and 4b) and provides the missing explanation of the variable latencies of the landing response.

Comparing the parameters of the correlation-type movement detector derived in the present study with those of the optomotor response, the landing response seems to use the same type of movement detection system. To account for the high wavelength optimum, however, the input elements of the movement detection system of the landing response might have an increased visual field (e.g. by pooling neighbouring visual elements) and, accordingly, a reduced visual acuity as compared with the input elements of the optomotor system.