Data from: Capuchins, space, time and memory: an experimental test of what-where-when memory in wild monkeys

There is considerable controversy about the existence, extent and adaptive value of integrated multimodal memory in non-human animals. Building on prior results showing that wild capuchin monkeys in Argentina appear to recall both the location and amount of food at patches they had previously visited, I tested whether they also track and use elapsed time as a basis for decisions about which feeding patches to visit. I presented them with an experimental array of eight feeding sites, at each of which food rewards increased with increasing elapsed time since the previous visit, similar to the pattern of ripe fruit accumulation in natural feeding trees. Over the course of 68 days, comprising two distinct renewal rate treatments, one group repeatedly visited sites in the feeding array, generating 212 valid choices between sites. Comparison of observations against simulated movements and multinomial statistical models shows that the monkeys' choices were most consistent with dynamic memory for elapsed time specific to each of the eight sites. Thus, it appears that capuchin monkeys possess and use integrated memories of prior food patch use, including where the patch is relative to their current location, how productive the patch is and how long it has been since they last visited the patch. Natural selection to use such integrated memories in foraging tasks may provide an ecologically relevant basis for the evolution of complex intelligence in primates.非人类动物的整合式多模态记忆（integrated multimodal memory）是否存在、其覆盖范围与适应价值，学界至今尚存较大争议。基于此前研究结果——阿根廷境内的野生卷尾猴（capuchin monkeys）能够回忆起此前到访过的取食斑块（feeding patches）的位置与食物量——本研究测试了这些卷尾猴是否同样会追踪并利用距上次到访的流逝时长，作为选择前往哪个取食斑块的决策依据。本研究设置了包含8个取食位点（feeding sites）的实验阵列，每个位点的食物奖励量会随距上次到访的流逝时长增加而提升，这与天然取食树上成熟果实的积累模式相似。实验为期68天，包含两种差异化的更新速率处理条件，一组卷尾猴反复到访该取食阵列中的位点，累计产生212次有效的位点选择行为。将观测结果与模拟移动轨迹及多项式统计模型（multinomial statistical models）进行对比后发现，卷尾猴的选择行为与它们对8个位点各自专属的距上次到访时长的动态记忆（dynamic memory）最为吻合。由此可见，卷尾猴具备并会利用针对过往取食斑块使用情况的整合式多模态记忆，其中包括该斑块相对于当前位置的方位、斑块的食物丰度，以及上次到访该斑块至今的时长。在觅食活动中运用这类整合式多模态记忆所承受的自然选择压力，或许为灵长类复杂智能的演化提供了兼具生态学意义的理论基础。