Experiment 1 subject information.

Many animal species live in multi-level societies regulated by complex patterns of dominance. Avoiding competition with dominant group-mates for resources such as food and mates is an important skill for subordinate individuals in these societies, if they wish to evade harassment and aggression. Chimpanzees (Pan troglodytes) are an example of such a species. This study investigated whether chimpanzees could understand the food preferences of their competitors, and make use of this understanding to select non-contested food items. Fifteen chimpanzees were given thorough experience of the differing target preferences of two virtual competitors. In the test, subjects had to select which of the two targets to approach, based on which competitor was present. To choose correctly, they would have to integrate the competitors’ preferences from across disparate observations, and then infer, before the competitor acted, what they would do in a novel situation. We also included a control condition featuring two targets for which subjects had no information about the competitors’ potential biases. The chimpanzees rapidly learned to direct their virtual agent to collect the targets, and some responded with vocalizations and hard knocking against the screen when competitors “stole” targets from the agent the subject was guiding. However, statistical analyses showed that, both at the individual and the group level, they did not succeed in selecting the correct target item at above-chance levels. Additionally, there was no significant difference between their performance in the test and control. We identify theoretical and methodological discrepancies that could explain the contrasting results of this and other studies.诸多动物物种均栖息于由复杂支配等级模式调控的多层级社会（multi-level societies）之中。对于此类社会中的从属个体而言，若要规避同类的骚扰与攻击，学会避免与优势群体成员争夺食物、配偶等资源是一项核心生存技能。黑猩猩（Pan troglodytes）便是这类物种的典型代表。本研究旨在探究黑猩猩是否能够理解其竞争对手的食物偏好，并据此选择无竞争风险的食物品类。研究共纳入15只黑猩猩，使其充分熟悉两名虚拟竞争对手（virtual competitor）的差异化目标偏好。在正式测试环节中，受试个体需根据当前在场的竞争对手类型，选择应趋近的目标物。若要做出正确选择，受试个体需整合此前不同观测场景中获取的竞争对手偏好信息，并在竞争对手采取行动前，推断其在全新情境下的行为倾向。本研究同时设置了对照条件：该条件下存在两类目标物，受试个体无法获知竞争对手对二者的潜在偏好偏向。黑猩猩群体很快便学会操控虚拟代理（virtual agent）以收集目标物；部分个体在竞争对手“抢走”受试所操控代理持有的目标物时，会发出叫声并用力敲击实验屏幕。但统计分析结果显示，无论是个体层面还是群体层面，受试黑猩猩选择正确目标物的概率均未显著高于随机猜测水平。此外，受试个体在测试环节与对照条件下的表现并无显著差异。本研究指出了可能解释本项研究与其他同类研究结果相悖之处的理论与方法层面的差异。