A neural control system promoting vigorous tracking behavior

Abstract

To track or pursue a moving visual object, the brain needs to generate motor commands that continuously steer the object toward the midline of the visual field via feedback control. In any feedback system, low feedback gain produces slow error correction, but high feedback gain risks instability; thus, the choice of feedback gain should ideally adapt to current circumstances1–6. Here we investigate mechanisms that produce adaptive feedback gain, focusing on the visual-locomotor pursuit system that flies use to chase one another7–12. As predicted by classic behavioral observations in houseflies7, we show that the Drosophila pursuit system is divided into two main parallel pathways, which are recruited in response to small and large errors, respectively. Arousal onset preferentially recruits the small-error pathway, and we show this pathway selectively corrects for object motion near the midline. The small-error pathway is specialized for lower gain, higher direction selectivity, and higher sensitivity to the fly’s forward speed. Modeling shows how the different specializations of these two pathways can improve performance: while the large-error pathway enables fast correction of large errors, the small-error pathway avoids oscillations that can result from over-correction of small errors. Together, these results show how specialized parallel pathways for driving the same action can produce adaptive control in neural systems.