University of Cambridge Neurology Unit, Addenbrooke's Hospital, Cambridge, UK.
Cogn Neuropsychol. 2001 Mar 1;18(2):125-74. doi: 10.1080/02643290125857.
Many cognitive psychological, computational, and neuropsychological approaches to the organisation of semantic memory have incorporated the idea that concepts are, at least partly, represented in terms of their fine-grained features. We asked 20 normal volunteers to provide properties of 64 concrete items, drawn from living and nonliving categories, by completing simple sentence stems (e.g., an owl is __, has , can). At a later date, the same participants rated the same concepts for prototypicality and familiarity. The features generated were classified as to type of knowledge (sensory, functional, or encyclopaedic), and also quantified with regard to both dominance (the number of participants specifying that property for that concept) and distinctiveness (the proportion of exemplars within a conceptual category of which that feature was considered characteristic). The results demonstrate that rated prototypicality is related to both the familiarity of the concept and its distance from the average of the exemplars within the same category (the category centroid). The feature database was also used to replicate, resolve, and extend a variety of previous observations on the structure of semantic representations. Specifically, the results of our analyses (1) resolve two conflicting claims regarding the relative ratio of sensory to other kinds of attributes in living vs. nonliving concepts; (2) offer new information regarding the types of features-across different domains-that distinguish concepts from their category coordinates; and (3) corroborate some previous claims of higher intercorrelations between features of living things than those of artefacts.
许多认知心理学、计算科学和神经心理学方法都将概念至少部分地以其精细特征来表示这一观点纳入了语义记忆的组织中。我们要求 20 名正常志愿者通过完成简单的句子主干（例如，猫头鹰是__，有__，能__）来描述 64 个具体项目（来自生物和非生物类别）的属性。在稍后的日期，相同的参与者会对相同的概念进行原型和熟悉度的评估。生成的特征根据知识类型（感官、功能或百科全书）进行分类，并根据主导性（指定该属性的参与者数量）和独特性（被视为该特征特征的概念类别内的范例比例）进行量化。结果表明，评分的典型性与概念的熟悉度以及与同一类别内范例平均值（类别质心）的距离有关。特征数据库还用于复制、解决和扩展关于语义表示结构的各种先前观察结果。具体而言，我们的分析结果（1）解决了关于生物与非生物概念中感官与其他类型属性的相对比例的两个相互矛盾的说法；（2）提供了关于在不同领域区分概念与其类别坐标的特征类型的新信息；（3）证实了一些关于生物特征之间的相关性高于人工制品之间的相关性的先前说法。
