Japanese Lexical Variation Explained by Spatial Contact Patterns

Abstract

In this paper, we analyse spatial variation in the Japanese dialectal lexicon by assembling a set of methodologies using theories in variationist linguistics and GIScience, and tools used in historical GIS. Based on historical dialect atlas data, we calculate a linguistic distance matrix across survey localities. The linguistic variation expressed through this distance is contrasted with several measurements, based on spatial distance, utilised to estimate language contact potential across Japan, historically and at present. Further, administrative boundaries are tested for their separation effect. Measuring aggregate associations within linguistic variation can contrast previous notions of dialect area formation by detecting continua. Depending on local geographies in spatial subsets, great circle distance, travel distance and travel times explain a similar proportion of the variance in linguistic distance despite the limitations of the latter two. While they explain the majority, two further measurements estimating contact have lower explanatory power: least cost paths, modelling contact before the industrial revolution, based on DEM and sea navigation, and a linguistic influence index based on settlement hierarchy. Historical domain boundaries and present day prefecture boundaries are found to have a statistically significant effect on dialectal variation. However, the interplay of boundaries and distance is yet to be identified. We claim that a similar methodology can address spatial variation in other digital humanities, given a similar spatial and attribute granularity.

Quantifying the spatial pattern of dialect words spreading from a central population

“…On the network of 46 prefectures, we regard Kyoto prefecture as the centre, P 0 , from which every linguistic variant derives. One typical method for modelling the linguistic diffusion on a network of cities is to use the gravity model, in which the extent of interaction between two cities is assumed to be proportional to the product of their population sizes and the inversed square of the distance in between [8,10]. However, since this assumption would always give a ii ¼ 1 in our model, we instead follow Burridge [11] to incorporate a modified gravity model, or the interaction density, w ij , which is defined as the time people in P i spend interacting with speakers in P j : Here, we adapt his eqn (2.3) to our model:…”

“…As with Newton's Law of gravity, (4.1) has a long algebraic tail. Geolinguistics has been adopting a variety of measures for geographical distance, such as Euclidean distance [9], great-circle distance (shortest distance on a sphere surface) [8], travel distance [8,9] and railway distance [20]. Here, we use the great-circle distance between prefectural government offices (buildings), summarized in [21].…”

“…One typical method for modelling the linguistic diffusion on a network of cities is to use the gravity model , in which the extent of interaction between two cities is assumed to be proportional to the product of their population sizes and the inversed square of the distance in between [ 8 , 10 ]. However, since this assumption would always give in our model, we instead follow Burridge [ 11 ] to incorporate a modified gravity model, or the interaction density , which is defined as the time people in spend interacting with speakers in Here, we adapt his eqn (2.3) to our model: where is the distance between and and denotes the population size of Constant represents the half-decay distance, that is, the distance at which the interaction density is halved relative to that within the same node, where words tend to spread farther when is larger.…”

“…cultural macroevolution; [7]). Similarly, on the level of dialect variation, much quantitative research has shown that the linguistic distance of each locality is to some extent explained by the geographical distance, although its correlation coefficient varies depending on how geographical distance is measured [8,9]. These studies give impression that the similarity of language reflects the phylogeny of human groups, but it is also often the case that the same dialect variant of a word is documented in phylogenetically distant local groups [10,11], which is likely due to diffusion of words between groups.…”

“…Effect of a barrier in the two-dimensional diffusion model. Populations form a 15 Â 15 square, with the central population being at(8,8) and a barrier at(11,8). (a) The mean word age and (b) the standard deviation in word age.…”