Areality and Genealogy in Linguistic Typology : A Graphical Modeling Approach to Pacific Rim

As demonstrated by Ono (2020), previous “interdisciplinary” studies have misunderstood “missing values” in linguistic typology and imputed linguistically meaningless values on the missing values in the data. The main objectives of this paper are to propose exploratory statistical methods that enable researchers to approach areality and genealogy on languages by linguistic typology without imputing missing values. We will propose new statistical analyses combining correspondence analysis (Benzécri et coll. 1973), correlation analysis (Tsubaki and Tsubaki 1997), graphical modeling (Lauritzen 1996), and clustering techniques that are based on three linguistic requirements in linguistic typology: “correction on frequencies,” “profile view,” and “language as variable.” The proposed methods resulted in not only outperforming previous studies on areal and genetic grouping in Africa, Eurasia, Europe or Indo-European family, and Austronesian and Austro-Asiatic but also detecting Circum-Pacific structure (Nichols 1994; Bickel and Nichols 2006) in languages. From the viewpoint of linguistics, our results showed that the word-order features have been masking areal and genetic groupings in linguistic typology, and word-order features related to Adjectives and Negation are needed in further investigations of linguistics. Since areal linguistics has mainly been focusing on some features or values that are not found in other areas or not necessarily included in WALS, our results opened up possibility that linguistic typology contributes to area linguistics by the common linguistic typological features or values. Furthermore, the Circum-Pacific structure detected in this paper includes some Pacific Rim (Bickel and Nichols 2006). However, our classification result did not distinguish the coastal part of languages between the other parts in North and South America, the former of which strictly consist of Pacific Rim. Thus, improving the quality of the database in linguistic typology and developing statistical methodologies consistent with substantive viewpoints are still promising in addressing areality and genealogy, including Pacific Rim and North Pacific Rim (Miyaoka 1992) that cannot be traced back by another linguistic realm, including lexicostatistics.