Anderson Carolyn Jane, Dillon Brian
Department of Computer Science, Wellesley College, Wellesley, USA.
Department of Linguistics, University of Massachusetts Amherst, Amherst, USA.
Open Mind (Camb). 2023 Feb 1;7:31-78. doi: 10.1162/opmi_a_00071. eCollection 2023.
Language use in conversation requires conversation partners to consider each other's points-of-view, or perspectives. A large body of work has explored how conversation partners take into account differences in knowledge states when choosing referring expressions. This paper explores how well findings from perspective-taking in reference generalize to a relatively understudied domain of perspective: the processing of grammatical perspectival expressions like the motion verbs and in English. We re-visit findings from perspective-taking in reference that conversation participants are subject to egocentric biases: they are biased towards their own perspectives. Drawing on theoretical proposals for grammatical perspective-taking and prior experimental studies of perspective-taking in reference, we compare two models of grammatical perspective-taking: a serial anchoring-and-adjustment model, and a simultaneous integration model. We test their differing predictions in a series of comprehension and production experiments using the perspectival motion verbs and as a case study. While our comprehension studies suggest that listeners reason simultaneously over multiple perspectives, as in the simultaneous integration model, our production findings are more mixed: we find support for only one of the simultaneous integration model's two key predictions. More generally, our findings suggest a role for egocentric bias in production for grammatical perspective-taking as well as when choosing referring expressions.
对话中的语言使用要求对话伙伴考虑彼此的观点或视角。大量研究探讨了对话伙伴在选择指称表达时如何考虑知识状态的差异。本文探讨了指称中视角采择的研究结果在一个相对较少研究的视角领域中的推广程度:像英语中的移动动词“come”和“go”这样的语法视角表达的处理。我们重新审视指称中视角采择的研究结果,即对话参与者存在自我中心偏差:他们倾向于自己的视角。借鉴语法视角采择的理论建议和指称中视角采择的先前实验研究,我们比较了两种语法视角采择模型:串行锚定与调整模型和同时整合模型。我们以视角移动动词“come”和“go”为例,在一系列理解和产出实验中测试它们不同的预测。虽然我们的理解研究表明,听众会像在同时整合模型中那样同时对多个视角进行推理,但我们的产出研究结果则较为复杂:我们只发现了同时整合模型两个关键预测中的一个得到了支持。更一般地说,我们的研究结果表明,自我中心偏差在语法视角采择的产出以及选择指称表达时都起作用。