The purpose of the present study is to show the applicability and usefulness of the experimental method for investigating the semantic structures of polysemous words. This approach rests on a basic tenet that the semantic relatedness which native speakers perceive between the uses of a polysemous word should be relied on as fundamental data to clarify the polysemous structure of that word.
In this study, subjects estimated on a 7-point scale the semantic similarity of pairs of sentenses containing a polysemous verb simeru.
The results were analysed by a non-metrical multi-dimensional scaling and cluster analyses, and the whole structure of semantic relatedness for simeru was represented in terms of a configuration in the 3-dimen-sional semantic space.
In specifying the polysemous structure on the basis of the semantic space, it was argued, the most appropriate model should be one with chain structures with each sense connected on family resemblances.
According to this model each sense is characterized by semantic attributes which may have gradient values and overlap one another. As a result of the analysis on this model, simeru was found to comprise 8 elemental senses which in turn got together to form 5 major sense groups.