Perceptual Representations in Conceptual Tasks

概念任务中的感知表征

• 批准号: 9421326
• 负责人: Lawrence Barsalou
• 金额: $ 15.25万
• 依托单位: University of Chicago
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1995
• 资助国家: 美国
• 起止时间: 1995-02-01 至 1998-01-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9421326&HistoricalAwards=false
• 关键词: Perceptual; Representations; Conceptual; Tasks

The aim of the proposed work is to determine whether perceptual representations enter significantly into conceptual processing. Subjects will perform conceptual tasks that have traditionally been explained with propositional, non-perceptual theories of knowledge, e.g., feature listing (what features are typically true of a chair) and property verification (e.g., does a chair have a seat). Of interest will be whether subjects use perceptual representations in these tasks, even though the task contexts do not require or demand that they do. The use of perceptual representations will be detected in two ways: instructional equivalence and perceptual work. Instructional equivalence occurs when subjects who receive no imagery instructions (`neutral subjects`) perform similarly to subjects explicitly asked to use imagery (`imagery subjects`). Perceptual work occurs when performance of both neutral and imagery subjects is affected by perceptual variables. Initial experiments on feature listing and property verification provide strong evidence for both instructional equivalence and perceptual work, indicating that neutral subjects use perceptual representations. Further studies will replicate and extend these findings. For feature listing, various experiments examine the roles of perceptual representations in conceptual combination, event concepts, abstract concepts, gesture, and production modes. For property verification, various experiments use distractor relatedness, preparation time, part expectancy, and perspective to examine the role of perceptual representations. In an additional project on modality-specific interference, several experiments test the prediction that perceptual representations of conceptual knowledge should produce specific patterns of attentional interference. Should these projects provide evidence that perceptual representations play central roles in conceptual processing, theories of cognition and intelligence would require significant reformulation. Most existing theories assume that intelligent behavior reflects the processing of an abstract, amodal, and arbitrary language of thought, much like the symbolic computer languages used widely in electronic computation. According to existing theories, perceptual experience is transduced into a set of conceptual symbols that bear no perceptual resemblance to their referential origins. An alternative, and much less explored view, is that conceptual symbols are fragments of perceptual states, extracted directly from these states via selective attention. The resulting conceptual representations are not static holistic 'pictures in the head,' but instead are systems of perceptual symbols that combine productively to produce infinite conceptual structures. Additionally, these perceptual symbols are analytic, schematic, multimodal, and dynamic, offering a rich expressive system for representing a wide variety of concepts, not just those for concrete objects. Intuitively, such perceptual symbol systems resemble modern multimedia computing much more than traditional non-analogue computing. Evidence continuing to indicate the presence of a perceptual symbolic system in humans, has strong implications for understanding and applying intelligence as well as theories of the brain's operation and visions of artificial intelligence. Rather than viewing intelligence as the manipulation of arbitrary amodal symbols, it would come to be viewed as arising from a productive system of perceptual symbols.

这项拟议工作的目的是确定知觉表征是否显著地进入概念加工。受试者将执行传统上用命题、非知觉知识理论解释的概念性任务，例如特征列表(椅子的哪些特征通常是正确的)和属性验证(例如，椅子是否有座位)。令人感兴趣的将是受试者是否在这些任务中使用知觉表征，即使任务语境并不要求或要求他们这样做。知觉表征的使用将通过两种方式来检测：教学对等和知觉工作。当没有接受表象指令的受试者(“中性受试者”)的表现与明确要求使用表象的受试者(“表象受试者”)的表现相似时，教学等价性就发生了。当中性受试者和想象受试者的表现都受到知觉变量的影响时，知觉工作就会发生。特征列举和属性验证的初步实验为教学等价性和知觉工作提供了强有力的证据，表明中性被试使用知觉表征。进一步的研究将重复和扩展这些发现。对于特征列表，各种实验考察了知觉表征在概念组合、事件概念、抽象概念、手势和生产模式中的作用。对于属性验证，各种实验使用分心物关联性、准备时间、部分期望和视角来检验知觉表征的作用。在另一个关于特定通道干扰的项目中，几个实验测试了概念知识的知觉表征应该产生特定注意干扰模式的预测。如果这些项目提供证据表明知觉表征在概念加工中发挥核心作用，那么认知和智力理论将需要重大的重新表述。大多数现有的理论认为，智能行为反映了一种抽象的、多变的和任意的思维语言的处理，就像电子计算中广泛使用的符号计算机语言一样。根据现有的理论，知觉经验被转化为一组概念符号，这些符号在知觉上与其参照来源没有任何相似之处。另一种较少探讨的观点是，概念符号是感知状态的片段，通过选择性注意直接从这些状态中提取出来。由此产生的概念表征不是静态的整体“头脑中的图画”，而是感知符号系统，它们有效地结合在一起，产生无限的概念结构。此外，这些感知符号是分析性的、示意性的、多通道的和动态的，为表达各种概念提供了丰富的表达系统，而不仅仅是具体对象的概念。直观地说，这种感知符号系统比传统的非模拟计算更像现代多媒体计算。继续表明人类存在知觉符号系统的证据，对理解和应用智能以及大脑运行理论和人工智能的愿景具有重要意义。与其将智力视为对任意非标准符号的操纵，不如将其视为产生于感知符号的生产性系统。