认知与发展心理学研究室知识库

词汇的意义包括字面意义和非字面意义，其中隐喻意义是非字面意义的重要组成部分，在词汇意义学习研究中，大部分研究关注字面意义的学习以及特征表征的建立，也积累了一些实验证据，那么在隐喻意义学习中，学习者能否习得词汇的隐喻意义以及建立特征表征?本研究想借鉴字面意义学习的范式一一语境学习范式，考察上述问题。
首先，研究一考察被试能否习得词汇的隐喻意义，包括两个实验。实验1在学习阶段创设了支持字面意义的语境和支持隐喻意义的语境，分别学习词汇的字面意义和隐喻意义。在测验阶段，要求被试完成词汇判断任务，新词充当启动词，目标词包括三种类型:对应概念(反映了字面意义学习)、本体词(反映了隐喻意义学习)和无关词。结果发现在词汇判断任务中，并不存在新词对本体词的启动效应，也就是没有习得词汇的隐喻意义。我们认为，可能任务设置导致的不同认知资源需求程度影响了对隐喻意义学习效果的探测，因此实验2进一步使用语义相关判断任务，考察被试能否习得词汇的隐喻意义。这一任务需要被试有意识的加工词汇的语义信息，从而可以保证被试加工到词汇的语义层面。结果发现，在语义相关判断任务中，新词确实可以启动本体词，习得词汇的隐喻意义。总之，研究一说明被试能够习得词汇的隐喻意义，隐喻意义的学习比字面意义的学习需要付出更大的认知努力。
其次，隐喻是建立在共享相似特征基础上的，那么隐喻意义学习是否可以拓展到语义特征层面，建立特征表征?研究三进一步考察这一问题。隐喻包括本体和喻体，喻体传达了词汇的隐喻意义。喻体的特征可以分为与隐喻意义相关的特征和与隐喻意义无关、仅与其字面意义相关的特征。关于隐喻中的特征加工，研究者形成了多种模型，其中具有代表性的是特征比较模型和特征赋予模型。在词汇隐喻意义学习中，能否建立特征表征?这两类特征表征的建立是否存在差异?符合哪一模型的观点?该研究进一步使用语境学习范式考察上述问题，在学习阶段创设了支持隐喻意义的语境，测验阶段的任务与实验二相同，目标词包括四种类型:对应概念、隐喻相关词(与隐喻意义相关的喻体特征)、喻体相关词(与字面意义相关的喻体特征)和无关词。结果发现，隐喻意义学习中可以启动两类相关词，建立特征表征;并且隐喻相关词和喻体相关词诱发的脑电成分不存在差异，符合特征比较模型的观点。
总之，本论文通过两个研究，考察语境学习中词汇隐喻意义学习及其特征表征的建立过程。结果发现，在词汇意义学习过程中，被试确实可以习得词汇的隐喻意义，隐喻意义比字面意义的学习需要付出更大的认知努力。隐喻意义学习中可以建立特征表征，并且两类喻体特征表征的建立不存在差异，符合特征比较模型的观点。

Lexical meaning contains literal meaning and non-literal meaning, and the metaphorical meaning is an important part of non-literal meaning. Previous lexical meaning learning studies mainly focus on the literal meaning and the establishment of feature representation. Little is known about the metaphorical meaning learning. The current study adopted the contextual learning paradigm, which was widely used in literal meaning studies, to explore whether participants were able to get the metaphorical meaning of the novel word, and whether the metaphorical meaning could establish feature representation.
First, Experiment 1 and Experiment 2 were designed to explore the first question.In Experiment 1，the novel words were embedded in two types of learning contexts: one supported the metaphorical meaning (metaphorical meaning learning condition: ML condition), and the other supported the literal meaning (literal meaning learning condition: LL condition). The learning effect was assessed via a lexical decision task with event-related potentials (ERPs) being recorded. In this task, the novel words served as primes, the literal corresponding concepts of the novel words (CC targets),the topic words (TO targets) and unrelated words (NR targets) served as target words.The results showed that the novel words could prime could prime the CC targets, butthey could not prime the TO targets, indicating that participants could not construct the relation between the novel words and topic words. We thought that the task might affect the demand for cognitive effort. In Experiment 2, a semantic-relatedness judgment task was adopted to explore the novel words' metaphorical meaning learning. The results found that novel words could prime the topic words, indicating that learners could get the metaphorical meaning of newly learned words. In a word, the results of two experiments indicated that participants could get the metaphorical meaning of novel words, and the detection of the learning effect was affected by the task settings. The metaphorical meaning learning was more demanding than literal meaning learning.
Second, the formation of metaphor is based on similarities of properties between the topic and the vehicle. In metaphorical meaning learning, can participants establish the feature representation? The related feature of vehicle can be divided into two types:the first one is related to the meaning of whole metaphor, while the second one is only related to the literal meaning of the vehicle. In the psycholinguistic literature on metaphor, two models have been proposed to describe the different features processing in metaphor: property comparison model and property attribution model. Whether the novel word can establish feature representation in metaphorical meaning learning? Do the two properties have different activation mode? Which model can explain the feature representation in metaphorical meaning learning? Experiment 3 was designed to explore these questions. In Experiment 3, we adopted the contextual learning paradigm which was used in Experiment 2. In the learning phase, participants were asked to guess the novel word's meaning from discourses which supported the novel words'metaphorical meaning. The learning effect was assessed via a semantic-relatedness judgment task. In this task, the novel words served as primes, the literal corresponding concepts of the novel words (CC targets), the metaphor-related feature words (MR targets), the vehicle-related feature words (VR targets) and unrelated words (NR targets) served as target words. ERP responses to the targets were recorded. The ERP results showed that participants could establish the feature representation in metaphorical meaning leaning. Furthermore, there was no difference between these two related words (MR targets and VR targets), which was consistent with the property comparison model.
In conclusion, we explored the lexical metaphorical meaning learning and the establishment of feature representation in contextual learning. There were two main findings: the first was that learners could get the metaphorical meaning of novel words. The second was that learners could establish feature representation in metaphorical meaning leaning, which was consistent with the property comparison model.