Data from: Collective behavior in vertebrates: a sensory perspective

Collective behavior models can predict behaviors of schools, flocks, herds, etc. However, in many cases, these models make biologically unrealistic assumptions in terms of the sensory capabilities of the organism, which are applied across different species. We explored how sensitive collective behavior models are to these sensory assumptions. Specifically, we utilized parameters reflecting the visual coverage and visual acuity that determine the spatial range over which an individual can detect and interact with conspecifics. Using metric and topological collective behavior models, we compared the classic sensory parameters, typically used to model birds and fish, with a set of realistic sensory parameters obtained through physiological measurements. Compared to the classic sensory assumptions, the realistic assumptions increased perceptual ranges, which led to fewer groups and larger group sizes in all species, and higher polarity values and slightly shorter neighbor distances in the fish species. Overall, classic visual sensory assumptions are not representative of many species showing collective behavior and constrain unrealistically their perceptual ranges. More importantly, caution must be exercised when empirically testing the predictions of these models in terms of choosing the model species, making realistic predictions, and interpreting the results.

群体行为模型（Collective behavior models）可用于预测鱼群、鸟群、兽群等集群的行为。然而在诸多场景中，这类模型针对生物体的感知能力（sensory capabilities）做出了不符合生物学现实的假设，且此类假设被跨物种套用。我们探究了群体行为模型对这类感知假设的敏感程度。具体而言，我们采用了反映视觉覆盖范围（visual coverage）与视觉敏锐度（visual acuity）的参数——此类参数决定了个体能够探测并与同种个体（conspecifics）互动的空间范围。我们分别采用度量型群体行为模型（metric collective behavior models）与拓扑型群体行为模型（topological collective behavior models），将常用于模拟鸟类与鱼类的经典感知参数，与通过生理测量（physiological measurements）获取的一系列贴合实际的感知参数进行对比。相较于经典感知假设，贴合实际的假设拓展了个体的感知范围（perceptual ranges），进而在所有受试物种中使得集群数量减少、单体集群规模增大；在鱼类物种中，还带来了更高的极性值（polarity values）与略短的邻体距离（neighbor distances）。总体而言，经典视觉感知假设无法代表诸多具备集群行为的物种，且会不切实际地限制其感知范围。更为关键的是，在基于实证检验这类模型的预测时，需在选择模式物种（model species）、生成贴合实际的预测以及解读实验结果等方面保持谨慎。