論文

'Word learning in Japanese children: Implications for theories of lexical development.　Psychology Development, University of British Colombiaにて'

Syntax has been noted to play an important role in word learning in English; it distinguishes the fundamental conceptual difference between individuals (coded as proper nouns), nonindividuals (coded as mass nouns), and classes of individuals (coded as count nouns). The Japanese language does not have grammatical markers flagging the distinctions between count nouns and mass nouns, between proper nouns and common nouns, or between singulars and plurals. How Japanese 2- and 4-yr-olds assign meaning to novel nouns associated with familiar and unfamiliar animals and inanimate objects was studied in this paper. When a novel label was given to an unfamiliar object, children assumed it to be a name for a basic-level object category whether the referent was an animal or an inanimate object. If the named object already had an established name, and if the object was an inanimate object, the children mapped the noun to a subordinate category. When the named object was an animal, however, they tended to interpret the label as a proper name. These results demonstrate that in the absence of useful information from syntax, 2-yr-old Japanese children are able to fast map a noun to its meaning by coordinating word-learning biases and other available sources of information. !

幼児の語彙獲得に対して帰納論理プログラミング(ILP)に基づく立ち上げモデルを提案している．主たる特徴は以下の4点である．第一に，いかなる後天的な知識にも依存しない語彙学習立ち上げ過程の位置付け．第二に，この過程について近年認知科学によって見出だされた学習バイアスの導入．第三に，互いに依存し合う語内包学習と概念階層学習の共発展メカニズムの検討．第四に，ヒトの脳との機能的類似性の検討，である．

人間の知の性質を理解するうえで言語の存在が思考にどのような影響を及ぼすかの問題について考察した。言語が思考に大きく影響を与え，言語に依存しない生得的制約と相互作用をしながら思考を形成していく観点ではサピア・ワーフ仮説が支持される。また思考が異なる言語の話者の間で異なるか否かという狭義の伝統的な観点では，答は思考のどの側面を問題にするか，どの認知領域を対象にし，その認知領域において環境自身の構造がどれだけ直接的に人間の認知に訴えてくるかに依存することも考えられる。今後の研究の方向性としては人間の生得的・普遍的認知機構が環境の呈する知覚情報等とどのように相互作用するのかを探っていくことである。

'Temporal language is often couched in spatial metaphors. English has been claimed to have two space a time metaphoric systems: the ego-moving metaphor, wherein the observer's context progresses along the time-line towards the future, and the time-moving metaphor, wherein time is conceived of as a river or conveyor belt on which events are moving from the future to the past. In three experiments, we investigated the psychological status of these metaphors by asking subjects to carry out temporal inferences stated in terms of spatial metaphors. In Experiment 1, we found that subjects were slowed in their processing when the assertions shifted from one spatial metaphoric system to the other. In Experiment 2, we determined that this cost of shifting could not be attributed to local lexical factors. In Experiment 3, we again found this metaphor consistency effect in a naturalistic version of the study in which we asked commonsense time questions of passengers at an airport.'

'Whether there is an advantage for noun learning over verb learning has been much debated. To examine this issue, we tested whether it is easier for Japanese children to learn the meaning of a novel verb than of a novel noun. Being an argument-dropping as well as SOV language, Japanese has properties that are said to be advantageous for verb learning. A novel noun or verb was introduced while children were seeing a video. How Japanese children map and generalize the word was then tested. The Japanese 3-year-olds mapped a novel noun to a novel object in a dynamic scene, and generalized it to the same object even though it appeared in a different action. In contrast, although the 3-year-olds knew that verbs refer to category of actions, their generalization of verbs was overly conservative, not yet understanding that a verb should be generalized by the sameness of the action per se, and the object in the action can be changed. These findings suggest learning of verb meanings is more difficult than noun meanings.'

Researchers have agreed that children learn words in a principled (e.g., Carey, 1982; Markman, 1989). Several important word learning principles, including the whole object assumption, the taxonomic assumption, the shape bias, the mutual exclusivity assumption, have been suggested (Markman, 1989, 1992; Landau, Smith, & Jones, 1998; Imai, Gentner, & Uchida, 1994). However, there has been an unsettled debate in the literature as to the exact nature of these principles, particularly with respect to their origin. In this paper, I discuss the origin of the whole object assumption and taxonomic assumption, especially in light of innateness and domain-specificity. Reviewing the literature of children’s early word production and comprehension and that of crosslinguistic comparison of early vocabulary development, I suggest that the whole-object and taxonomic assumptions are likely to be universally applied independent of the structure of linguistic input but that these principles are not necessarily applied at the onset of word learning. I argue that it is not likely that the whole-object and taxonomic assumptions are innately specified as abstract rules. I argue that these word-learning principles are formed from a combination of domain general, basic cognitive facilities (such as detecting similarity among entities and putting similar things together to form categories, and extracting and generating abstract rules from a small amount of experience) and domain specific, pre-linguistically available ontological knowledge about the world. Finally, I evaluate the necessity of the whole-object assumption and the taxonomic assumption in the face of the position that social-pragmatic force is sufficient to guide word learning. Keywords: word learning principle (語意学習原理), whole object assumption (事物全体原理), taxonomic assumption(事物カテゴリー原理), innateness(生得性), domain-specificity (領域固有性)

Investigated whether learning the distinction betweensubstance names and object names is conceptually or　linguistically　driven, by repeating Soja et al.'s study with English-and Japanese-speaking children. (Japanese does not make the count-massgrammatical distinction proposed to contribute to learning the distinction.) Found evidence for universal use of ontological knowledge in early word learning, with some cross-linguistic differences.

【学会誌特集号編集】認知科学4-1特集「言語獲得」を企画・編集

This research has examined whether there is an asymmetry in children’s and adults’ notion of “likeness” in making inferences about the extension of a novel word and in making inferences about novel properties of objects. Recent research studying children’s word learning has suggested that children of preschool age rely on shape similarity rather than taxonomic relations when the two aspects are separated (e.g., Baldwin, 1992; Imai, Gentner & Uchida, 1994). In contrast, a body of work by Gelman has shown persuasively that even 2-year-olds project a property of an object onto other objects sharing the same label in spite of misleading perceptual similarity. This discrepancy between the results from the word extension studies and those from Gelman’s induction studies suggest that children use different criteria in determining “likeness” an across two situations (cf. Gelman, Coleman & MaCoby, 1986; Gelman & Medin, 1993). To examine this, four studies were conducted with Japanese 5-year-olds and adults on the item sets in which conceptual similarity and perceptual similarity conflicted . Children as well as adults shifted relative weight placed between shape similarity and conceptual similarity across the word learning task and the property inference task, relying on shape similarity more heavily in the former task. I argue that this asymmetry reflects children’s appreciation that shape has strong predictive power for inferring category-membership but not for inferring shared properties. Furthermore, the results suggested that children know how the relative weight between shape and conceptual similarity should be shifted across natural kinds and artifacts. I conclude that children’s word meaning and concept learning is theory-driven and principled not because they can bypass perceptual similarity but because they know, just like adults, when they should and should not rely on perceptual similarity.

This research examined the acquisition of human/animal numerical classifier categories in preschool children. In Experiment 1, using an error detection paradigm, we examined the developmental change of the knowledge 150 children possessed about classifiers. Overall, even the youngest group of children (early 4-years-olds) were aware of the grammatical function of classifiers, but they did not know the semantic criteria for each classifier category until around their late 5 years of age. Experiment 2 examined 60 4- and 5-years-olds who had not shown the sensitivity for the semantic rule dividing the hiki (small animal) category and the tou (large animal) category. The results showed that the 5-years-olds only, could spontaneously generate the correct rules and apply it to new items only through hearing adults’ use of classifiers. Based on such results, the interplay between the linguistic input, grammar and extra-linguistic cognitive categories in acquisition of classifiers was discussed. Key words: numerical classifier, preschool children, error detection paradigm, semantic rule, interplay between the linguistic input and cognitive categories.

Psychological studies of superordinates have generally treated them as equivalent. However, many languages distinguish mass superordinates (e.g., clothing) from count superordinates (e.g., vehicle). In the present paper, experimental evidence is presented which suggests that the two types of superordinates are conceptually distinct as well. One study showed that the members of mass superordinates more often co-occur. A second study showed that people more often interact with the members of mass superordinates in temporal proximity whereas people primarily interact with single members of count superordinates on a specific occasion. Also, properties that characterize an individual are more salient aspect of count superordinates. These findings imply that mass superordinates refer to unindividuates groups of objects, united by spatial and functional contiguity. Two other studies supported this hypothesis by showing that the class inclusion relation between a single object and a category is stronger for count superordinates. Taken together, the findings suggest that mass superordinates are not true taxonomic categories. We relate the findings to previous views of superordinates and to the count/mass distinction in general.

A study with three- and five-year olds contrasted two important proposals regarding children's assumptions about word meanings: the taxonomic assumption proposal and the shape bias proposal. Results suggest that perceptual similarity, particularly shape similarity, is very important in early word meaning but that children gradually shift their attention to deeper properties.