Real-Time Encoding of Motion: Answerable Questions and Questionable Answers from the Fly’s Visual System

Abstract

Much of what we know about the neural processing of sensory information has been learned by studying the responses of single neurones to rather simplified stimuli. The ethologists, however, have argued that we can reveal the full richness of the nervous system only when we study the way in which the brain deals with the more complex stimuli that occur in nature. On the other hand it is possible that the processing of natural signals is decomposable into steps that can be understood from the analysis of simpler signals. But even then, to prove that this is the case one must do the experiment and use complex natural stimuli. In the past decade there has been renewed interest in moving beyond the simple sensory inputs that have been the workhorse of neurophysiology, and a key step in this program has been the development of more powerful tools for the analysis of neural responses to complex dynamic inputs. The motion sensitive neurones of the fly visual system have been an important testing ground for these ideas, and there have been several key results from this work:

1. The sequence of spikes from a motion sensitive neurone can be decoded to recover a continuous estimate of the dynamic velocity trajectory (Bialek et al. 1991; Haag and Borst 1997). In this decoding, individual spikes contribute significantly to the estimate of velocity at each point in time.

2. The precision of velocity estimates approaches the physical limits imposed by diffraction and noise in the photoreceptor array (Bialek et al. 1991).

3. One or two spikes are sufficient to discriminate between motions which differ by displacements in the “hyperacuity” range, an order of magnitude smaller than the spacing between photoreceptors in the retina (de Ruyter van)