Visual circuitry for distance estimation in Drosophila.

Abstract

Animals must infer the three-dimensional structure of their environment from two-dimensional retinal images. They use visual cues like motion parallax and binocular disparity to judge distances to objects, and studies across several animal models have found and characterized neural signals that correlate with visual distance. However, the causal role of these neurons in distance estimation and the range of their possible neural properties remain poorly understood. Here, we show that both directional and non-directional feature-selective neurons in the Drosophila visual system are involved in estimating distance during free locomotion. We used a high-throughput behavioral assay to perform a targeted silencing screen of visual neurons, and we subsequently characterized distance tuning using in vivo two-photon microscopy, thus linking distance perception directly to neural signals. Silencing the primary motion detectors eliminated distance-dependent behavior, consistent with reliance on motion parallax. Our screen also identified a visual feature-detecting neuron that encodes a non-canonical motion parallax signal: the signal is not direction selective for object or background motion, but it is tuned to the relative speeds of foreground and background, resulting in a signal that can measure relative distance. Our results demonstrate the behavioral roles of direction-selective and distance-tuned neurons in fly distance estimation and provide a framework for considering broader classes of neurons that encode distance through motion parallax.