Drosophila vision is an exciting field right now – the latest genetic technologies are enabling rapid progress in understanding vision from biophysics to behavior. Today, Current Biology published our paper showing that a single motion detecting circuit may be sufficient to mediate multiple behavioral responses. This is surprising because in the fly, two important behaviors – object fixation and wide-field stabilization – are widely thought to result from circuitry with elements dedicated to the particular tasks. Contrary to this view, we show that turns toward objects can result from the same motion processing pathway already thought to underlie wide-field stabilization. Consequently, we think this work creates a new "null hypothesis" – subsequent work will need to show behavioral responses to visual objects not predicted by this model to argue for the involvement of additional circuitry. This work also has interesting parallels to the circuitry and behavior of mammalian eye movement.
Our paper is both experimental and theoretical in nature. We performed experiments on genetically modified flies and showed that intact motion-processing T4 and T5 cells are necessary for stripe fixation at high gain and in figure ground discrimination tasks. With this experimental result, we implemented both a simple, phenomenological model and a physiologically more realistic model that predict stripe fixation in which the visual system consists of a single, motion-sensitive pathway. Remarkably, these simple models have no explicit object detection or small-field selectivity. We show that a particular configuration of one of our models is formally equivalent to the classical model published by Poggio and Reichardt in 1973. The models also make non-obvious predictions of stripe-fixation behavior. One such prediction is fixation in front of a moving background, and when we tested this prediction experimentally, we found qualitative agreement between both models and behavioral results on tethered flies.
We hope that by including the full source code of our models - including an online version of the phenomenological model - others can easily reproduce our work and, for example, predict fly physiological and behavioral responses to arbitrary visual stimuli.
The full reference is:
Download and read more about the models from strawlab.org/asymmetric-motion/.