Predictions about reward outcomes in rhesus monkeys

Human infants and nonhuman animals respond to surprising events by looking longer at unexpected-than-expected situations. These looking responses provide core cognitive evidence that nonverbal minds make predictions about possible outcomes and detect when these predictions fail to match reality. We propose that this phenomenon has crucial parallels with the processes of reward prediction error, indexing the difference between expected and actual reward outcomes. Most work on reward prediction errors to date involves neurobiological techniques that cannot be implemented in many relevant populations, so we developed a novel behavioral task to assess monkeys’ predictions about reward outcomes using looking time responses. In Study 1, we tested how semi-free-ranging monkeys (n = 210) responded to positive error (more rewards than expected), negative error (less rewards than expected), and number control. We found that monkeys looked longer at a given reward when it was unexpectedly large or...

人类婴儿与非人类动物会对意外事件作出反应：相较于符合预期的情境，它们会更长时间地注视意外情境。这类注视反应为非言语心智能够对可能结果作出预测，并检测自身预测与现实不符的情况，提供了核心认知证据。我们提出，该现象与奖励预测误差（reward prediction error）的加工过程存在关键相似性——后者用于量化预期奖励结果与实际奖励结果之间的差值。迄今为止，绝大多数关于奖励预测误差的研究均采用神经生物学技术，这类技术无法在诸多相关研究群体中实施，因此我们开发了一项新颖的行为任务，通过注视时长反应来评估猴子对奖励结果的预测。在研究1中，我们以210只半自由活动的猴子为被试，测试了它们对正误差（奖励多于预期）、负误差（奖励少于预期）以及数量控制条件的反应。我们发现，当某一奖励的数量超出预期（偏大）时，猴子的注视时长会更长，或者……