Chapter 159Dialect Classification and Dialectometry

159.1 Regional divisions and documentary samples

The broad division into Hokkaido, Sakhalin, and Kuril varieties is useful geographically, but grammatical and lexical boundaries require comparison of particular records. Within Hokkaido, southwestern and northeastern groupings recur in the literature, alongside distinctive features of Sōya and variation within each group Nakagawa & Fukazawa (2022: §§1–2).

Three influential datasets must be distinguished. Hattori and Chiri's 1960 comparison sampled nineteen localities: thirteen in Hokkaido and six in Sakhalin, with separate Kuril material from Torii. Hattori's 1964 dictionary directly investigated nine varieties, adding Torii's Kuril records as a tenth. Its Hokkaido varieties were Yakumo, Horobetsu, Saru at Hiratori, Asahikawa, Obihiro, Bihoro, Nayoro, and Sōya; the Sakhalin variety was Raichishka Nakagawa & Fukazawa (2022: §2.1); Hattori (1964: 37–38). Asai's later comparison selected 110 items, while Hattori and Chiri's 1960 calculations used 91 selected items. Their coverage and selection are consequently different Nakagawa & Fukazawa (2022: §2.2; §3, opening).

159.2 What a lexical similarity percentage measures

Lexical comparison can count forms as corresponding despite regular sound differences. Nakagawa and Fukazawa illustrate this with the ‘tooth’ forms mimak, nimak, imak, imah. In their explanation of Hattori and Chiri's method, a correspondence counts as one match. Where multiple variants yield both a match and a non-match, the comparison contributes half a match. Uncertain or missing entries are excluded from the denominator Nakagawa & Fukazawa (2022: §2.1).

Their hypothetical calculation has 60 matches, 30 partial matches, and 10 non-matches: (60 + 15) / 100 = 75%. This illustrates the method; it is not a new measurement of two dialects. Decisions about variant selection, correspondence, and missing data affect the result. A percentage measures similarity under those decisions and does not supply a date of separation Nakagawa & Fukazawa (2022: §2.1).

The 1960 sample also omits Shizunai, Mitsuishi, and Urakawa. Descriptions of a boundary near Shizunai and Niikappu therefore draw on a broader body of evidence than that nineteen-locality table Nakagawa & Fukazawa (2022: §2.3).

159.3 Reanalysis and dialectometry

Ono reexamines the position of Northern Kuril vocabulary. His analysis finds its strongest Hokkaido relationship with northeastern varieties, including Asahikawa and Nayoro, followed by southwestern varieties. He also compares artificial modifications of similarity matrices to examine the behavior of different classification methods. Those modified matrices are methodological demonstrations Ono (2020: 234–235; 242–246, §2.3).

Fukazawa and Ono use correspondence analysis to reconsider dialect boundaries. They normalize some accentual, glottal, and phonetic variation while retaining contrasts such as car/cara. Their treatment of multiple variants lets a locality share lexical features with more than one group. The resulting overlap represents relationships in the analyzed data; it is not a genealogical tree with a date on each branch Fukazawa & Ono (2024: 161–171, §§4–5).

Kasuga's conference abstract extends classification to 75 localities using existing lexical records, regularity features, geographic information, and established labels. It reports leave-one-out prediction of broad labels for 73 of 75 localities, with a lower subgroup score of 0.653. These results evaluate recovery of the classification labels used in the study. They are neither a new 75-locality field survey nor independent proof of common ancestry for each predicted group Kasuga (2026: 92–93, poster P-11).

159.4 Classification and grammatical description

A regional label helps locate an example, but its grammatical form still needs a source and a speaker or locality where available. Accent, initial consonants, interrogative vocabulary, and kinship terms have partly overlapping distributions. These are compared in Chapter 160 (Hokkaido Dialect Variation in Sound and Grammar) and Chapter 161 (Lexical Geography and Survey History). The distinctions also matter when comparing Hokkaido with Chapter 162 (Sakhalin and Northern Kuril Comparisons): a regional summary cannot replace the documented paradigm of a particular variety.