LD calculated using narrow, phonetic transcriptions

to fit the Sui was not difficult. All changes that we made are given in table 7.2. We retained the secondary tone split markings in the Kam data see chapter 4, table 4.1 because these extra tones, where they occur, have very different phonetic realisations to their regular non-aspirated counterpart tones. Table 7.2. Adjustment of Kam transcriptions in preparation for Sui-Kam LD analysis Original transcriptions Replaced by ɑ a aC aːC aV aːV ɐC aC əu, ɐu au əi, ɐi ai ɯi ui ʊ u j ɪ j ə ɛ e

7.4 Results of Sui dialect comparison

7.4.1 LD calculated using narrow, phonetic transcriptions

Table 7.3 shows Levenshtein distances calculated using the narrow, phonetic transcriptions. The smallest phonetic distance was between SD and ZH both Central, at 0.030. The largest distance was between JR Southern and JL Pandong, at 0.162. 6 Distances of under 0.075 are highlighted in grey. At this threshold four clusters emerge: 1 Central and Western; 2 Eastern; 3 Southern; and 4 Yang’an. The two Pandong lects are outliers. Clusters as revealed by applying the WPGMA Weighted Pair Group Method with Arithmetic Mean clustering algorithm to these data are shown in map 7.1 and figure 7.7 . Prokić and Nerbonne 2009 urge caution when applying clustering algorithms. In their application of LD to Bulgarian dialects they found that their results varied significantly depending on the clustering algorithm used, partly because the clusters revealed were sometimes “mere artefacts of algorithms” rather than “natural groups present in the data set.” In order to offset this risk, we used various different algorithms including fuzzy clustering Nerbonne et al., 2008 to find the most stable clusters. The results were identical in all cases. Each cluster is represented using a different colour. Light blue corresponds to Pandong dialect, dark blue to Yang’an dialect. Light green corresponds to Southern Sui and dark green corresponds to other non-Southern Sandong varieties. 6 These distances are significantly smaller than for other dialect studies such as Yang 2009; Castro et al., 2010; and Jackson et al., 2012. This is largely due to the fact that we compared only cognates, whereas these other studies include non-cognates in their comparisons. Distances calculated using all of our Sui data without distinguishing cognates were much higher, generally in the range of 0.1 to 0.3. Map 7.1. Sui cluster map using narrow transcriptions and WPGMA Table 7.3. Levenshtein distances for narrow transcriptions distances under 0.075 shaded in grey Sandong Yang’an Pandong Central Western Eastern Southern SD ZH AT TP TZ DJ SJ RL JQ SY JR SW BL TN JL PD 0.030 0.074 0.067 0.072 0.069 0.036 0.060 0.055 0.060 0.066 0.058 0.069 0.070 0.071 0.092 0.077 0.08 0.103 0.102 0.102 0.059 0.081 0.085 0.098 0.099 0.113 0.055 0.039 0.068 0.07 0.104 0.108 0.089 0.095 0.100 0.103 0.073 0.071 0.108 0.111 0.095 0.106 0.104 0.110 0.047 0.085 0.085 0.121 0.129 0.108 0.118 0.113 0.119 0.051 0.054 0.095 0.096 0.093 0.096 0.116 0.085 0.100 0.099 0.054 0.070 0.074 0.122 0.119 0.134 0.132 0.12 0.121 0.117 0.105 0.134 0.143 0.148 0.145 0.132 0.120 0.118 0.117 0.127 0.112 0.123 0.108 0.141 0.143 0.154 0.126 0.063 0.129 0.136 0.119 0.126 0.123 0.114 0.104 0.107 0.156 0.155 0.162 0.161 0.153 0.151 0.141 0.145 0.121 0.117 0.144 0.111 0.102 0.092 0.149 0.150 0.150 0.132 0.150 0.125 0.098 Figure 7.7. Sui cluster dendrogram using narrow transcriptions and WPGMA. In order to check the validity of these clusters we produced a multidimensional scaling MDS plot of the data. This is shown in figure 7.8. Relative phonetic distances between Sui varieties are plotted onto a two-dimensional plane. Separate Pandong and Yang’an clusters are unambiguous. The Sandong varieties, however, appear to be in more of a “dialect continuum”, ranging from RL closest in pronunciation to Pandong to JR the most distant from all other varieties. SW Southern appears almost midway between the Central Sui dialects SD, ZH and the other Southern Sui dialects JQ, SY, JR. TZ, which belongs to Western Sui by other analyses, appears to be closer to Central SD, ZH. Figure 7.8. MDS plot of LDs based on narrow transcriptions.

7.4.2 LD calculated using broad, phonemicised transcriptions