Fig. 5: Modelling of the systematic distances (sys_distm) in the Shanghai dataset.
a1–a7 Multi-dimensional scaling (MDS) plots of the mean sys_dist for each dialectal variety spoken in Shanghai, with arrows connecting each sub-dialect’s old version to its corresponding new version and statistical centroids annotated at the top-left of each plot. b1–b7 Triangles representing the relationship between one variety’s mean sys_dist to the statistical centroid (the horizontal coordinate) and its mean neighbuorhood sys_dist (the vertical coordinate, n_neigh = 3, taking three closest neighbours into consideration); Points representing the other varieties’ mean sys_dist (the vertical coordinates) to this variety. c1–c7 Points representing the relationship between one variety’s mean sys_dist to the statistical centroid (the horizontal coordinate) and its mean neighbourhood density ( = 1 – mean neighbourhood sys_dist, the vertical coordinate), again with arrows connecting each sub-dialect’s old version to its corresponding new version. d1–d7 Paired box plots for the range, quartiles, and median between old (o, in the left) and new (n, on the right) varieties’ mean sys_distm to the Shiqu variety, with results of paired t-tests and Wilcox-tests annotated at the top of each plot. These plots are vertically arranged according to the phonemic units: syllables (Syl: a1, b1, c1, d1), segmental combinations (Seg: a2, b2, c2, d2), onset consonants (Ons: a3, b3, c3, d3), rhymes (Rhy: a4, b4, c4, d4), vowels (Vow: a5, b5, c5, d5), final consonants (Fin: a6, b6, c6, d6), and tone classes (Ton: a7, b7, c7, d7). The labels and points are colour-coded consistently.
