An embodied semantic mechanism for mimetic words in Japanese.

摘要

The assumption that the connection between a symbol and its meaning is arbitrary has been the foundation of modern linguistics, symbolic artificial intelligence, and most theories of knowledge representation. On the other hand, recent developments in cognitive science provide a new perspective: concepts and symbols emerge through the interaction between the environment and the organism, which suggests a non-arbitrary or iconic basis in the symbol-meaning connection. The contention of this thesis is that the meanings of mimetic words are also the product of interactive processes involving bodily sensations, articulatory sensations, and the linguistic sound system.;Mimetic words in Japanese are a word class similar to onomatopoeic words in that language sounds are used expressively, but they are more numerous and have much broader usages than onomatopoeic words in Indo-European languages. Similar word classes are found in many African and Asian languages. The first claim of this thesis is that Japanese mimetic words can be characterized as a word class that expresses human sensory experiences. Basic mimetic words use language sounds to imitate sensations including body movements, touch, vision, smell, taste, and sound. For example, the word pyonpyon expresses jumping and hopping, and sarasara conveys the smoothness on the skin.;In order to connect sound sequences and sensory experiences, the hypothesis of articulatory mediation is proposed. The claim is that articulatory sensations, associated with the production of sounds, connect sound sequences with a wide range of bodily sensations.;A questionnaire experiment showed that those who do not understand Japanese can choose the correct mimetic word given a situation and a choice of words, and that those who actually pronounced the words performed better in the task. A computer program was constructed to simulate the mechanism of articulatory mediation. The program was able to learn the correlations between the articulatory dynamics and the semantic scale values for the words, and it was also able to predict the meanings of new mimetic words.;These results establish the existence of an iconic, embodied connection between sound sequences and meaning at the basic level of mimetic words in Japanese.