School of Psychology, University of Minho, Braga, Portugal.
PLoS One. 2013 May 21;8(5):e64934. doi: 10.1371/journal.pone.0064934. Print 2013.
Foraging animals typically encounter opportunities that they either pursue or skip, but occasionally meet several alternatives simultaneously. Behavioural ecologists predict preferences using absolute properties of each option, while decision theorists focus on relative evaluations at the time of choice. We use European starlings (Sturnus vulgaris) to integrate ecological reasoning with decision models, linking and testing hypotheses for value acquisition and choice mechanism. We hypothesise that options' values depend jointly on absolute attributes, learning context, and subject's state. In simultaneous choices, preference could result either from comparing subjective values using deliberation time, or from processing each alternative independently, without relative comparisons. The combination of the value acquisition hypothesis and independent processing at choice time has been called the Sequential Choice Model. We test this model with options equated in absolute properties to exclude the possibility of preference being built at the time of choice. Starlings learned to obtain food by responding to four stimuli in two contexts. In context [AB], they encountered options A5 or B10 in random alternation; in context [CD], they met C10 or D20. Delay to food is denoted, in seconds, by the suffixes. Observed latency to respond (Li) to each option alone (our measure of value) ranked thus: LA≈LC<LB<<LD, consistently with value being sensitive to both delay and learning context. We then introduced simultaneous presentations of A5 vs. C10 and B10 vs. C10, using latencies in no-choice tests to predict sign and strength of preference in pairings. Starlings preferred A5 over C10 and C10 over B10. There was no detectable evaluation time, and preference magnitude was predictable from latency differentials. This implies that value reflects learning rather than choice context, that preferences are not constructed by relative judgements at the time of choice, and that mechanisms adapted for sequential decisions are effective to predict choice behaviour.
觅食动物通常会遇到它们要么追求要么跳过的机会,但偶尔也会同时遇到几个选择。行为生态学家使用每个选项的绝对属性来预测偏好,而决策理论家则专注于选择时的相对评估。我们使用欧洲椋鸟(Sturnus vulgaris)将生态推理与决策模型相结合,链接和测试关于价值获取和选择机制的假设。我们假设,选择的价值取决于绝对属性、学习背景和主体的状态。在同时选择中,偏好可能是通过比较使用深思熟虑时间的主观价值来产生的,也可能是通过独立处理每个选择而不进行相对比较来产生的。价值获取假设和选择时独立处理的组合被称为序列选择模型。我们使用在绝对属性上相等的选项来测试这个模型,以排除在选择时建立偏好的可能性。椋鸟通过在两个环境中对四个刺激做出反应来学习获取食物。在环境 [AB] 中,它们随机交替遇到选项 A5 或 B10;在环境 [CD] 中,它们遇到 C10 或 D20。后缀表示获得食物的延迟,以秒为单位。单独观察到的对每个选项的反应潜伏期(Li)(我们衡量价值的指标)的排名如下:LA≈LC<LB<<LD,与价值对延迟和学习背景都敏感一致。然后,我们在不选择的测试中引入了 A5 对 C10 和 B10 对 C10 的同时呈现,用潜伏期来预测配对中的偏好的符号和强度。椋鸟更喜欢 A5 而不是 C10,也更喜欢 C10 而不是 B10。没有检测到评估时间,偏好的大小可以从潜伏期的差异来预测。这意味着价值反映了学习而不是选择背景,偏好不是在选择时通过相对判断构建的,并且适应于序列决策的机制能够有效地预测选择行为。