A mechanistic understanding of collective escape in starling flocks.

Large flocks of European starlings change shape, size, and internal structure continuously and rapidly when hunted by aerial predators. How their diverse patterns of collective escape emerge is still unknown. Here, we disentangle the collective behavior of starlings combining video footage of flocks pursued by a robotic predator and a data-driven 3-dimentional agent-based model. In vivo, we show that flock members often differ in their evasive maneuvers and that several collective patterns arise simultaneously across a flock. In silica, we identify individual-level rules that lead to dynamics of collective motion and escape similar to real starlings. Our results suggest that the mechanisms underlying starling murmurations depend on how fast escape information propagates through the flock, the relative positions of the escaping individuals to the predator, and the previous state of the flock (hysteresis). Our study highlights the importance of investigating fine-scale dynamics and micro-macro feedback when studying self-organized adaptive systems.