北方言語研究 = Northern Language Studies;第7号

FONT SIZE:  S M L

計量多次元尺度構成法の有効性と限界 : アイヌ語を例として

小野, 洋平

Permalink : http://hdl.handle.net/2115/64514

Abstract

The main objective of this paper is to clarify possibilities and limitations on Multidimensional Scaling (MDS) in linguistics, taking the (dis-) similarity data in Hattori and Chiri (1960) as an example. The present author first shows the advantages of MDS, which can transform the dissimilarity data into multidimensional coordinate systems, and we can interpret those plots with the linguistic knowledge. However, MDS is an approach, which cannot be recommended from statistical point of view, in which the dissimilarity or distance matrix is calculated directly from the cognacy data. In this paper, the author then demonstrates that such method can omit the core information from the original data. Indeed, Ono (2015) found out the interesting insight on Ainu dialect with the statistically recommended approach using the original data of Hattori and Chiri (1960). However, the way the author, as statistician, does not think as desirable is one approach that can be easily used for non-specialist of statistics to test the hypothesis about the relationship of languages or dialects and it is useful for linguists to understand how MDS works in the frame of statistics because some previous lexicostatistical researches, which applied MDS to cognacy data, should be reconsidered from this point of view. Therefore, demonstrating the possibilities on MDS, the present author introduces (metric) Multidimensional Scaling, applies it to the (dis-) similarity data in Hattori and Chiri (1960) and then reconfirms statistically one of the characteristics in Ainu on dialects in and around Hiratori, which was indicated many time philologically but first verified statistically in Ono (2015). Through this introduction about possibilities and limitations on MDS, the author desires to promote the fruitful reexamination of linguistic researches using MDS especially in lexicostatistics, the better understanding on the application of statistics to linguistic data among linguists and complementary studies among linguists and statisticians.

FULL TEXT:PDF