8.6 Conclusion
When discussing comprehension percentages given by a question and answer RTT method, Casad 1974:46 suggested that 80 was a suitable threshold for indicating “sufficient comprehension” for use
of common materials. However, Casad also notes that not all RTTs or languages are equivalent and that the threshold could actually vary from 75 to 85 depending on the test. The methodology used for our
Sui survey was quite different from Casad’s question and answer method. It is more detailed and our scoring is much finer. Because of this, and from previous experience with this type of test, we feel that a
higher threshold of 85 is more realistic. It is certainly unlikely that varieties that show less than 80 intelligibility would be able to use common mother tongue materials.
Figure 8.6 shows the sixteen datapoints grouped into “intelligibility clusters”. Each percentage indicates the threshold at which all the locations within that cluster can understand the reference variety
which is underlined. For example, the twelve locations in the “GC Central” cluster can all understand over 85 of the GC lect. The five locations in the “SY Southern” cluster all scored over 75 mostly
closer to 80 on the SY test. All fourteen locations in both the GC Central and SY Southern groups scored over 70 on the GC test.
Figure 8.6. Intelligibility clusters indicated by RTTs. The results indicate that Central Sui is the best variety on which to base language development
efforts, since Central, Western, Eastern, Yang’an and JQ and SW both Southern can all understand over 85 of this variety. We judge that materials based on the Central Sui lect could be used by speakers
from all these other regions without too much difficulty. If language development efforts are to include Sui speakers of all dialects, our results suggest that
separate materials must be developed for Pandong speakers and Southern speakers in Libo county JR and SY, since these two groups of people understand far less than 85 indeed, less than 75 of the
Central Sui lect. As we noted earlier, it is possible that JR would be a better choice of reference lect for Southern Sui than SY. This requires further investigation.
8.7 Methodology critique
Overall, we are extremely satisfied with the RTT methodology used for this study and we feel that the results give a realistic picture of the overall interdialect intelligibility situation in the Sui area.
Perhaps the greatest weakness of our study is the fact that it sometimes gave us a better idea of “acquired intelligibility” than “inherent intelligibility”. Because of the small geographical area covered
by the survey, the greatest challenge was to weed out test participants who had had contact with other
varieties. Our potential pool of test subjects in each village was often limited because many villagers were either busy in the fields or had gone far away to do migrant work. In the Yang’an area it was
impossible to find test subjects who had had no exposure to Western or Central Sui varieties. And in Western and Central areas it was also difficult to find test subjects who had never heard the Yang’an
dialect before.
Despite this, inherent intelligibility between Yang’an and Central Sui can be guessed at by looking at the results of the Yang’an test in Eastern Sui areas. Eastern Sui is linguistically extremely close to
Central Sui see chapters 5, 6 and 7. Moreover, to our knowledge, Eastern Sui speakers have no contact with Yang’an Sui at all. Thus their extremely low scores on the Yang’an test on average around 50
indicate that inherent intelligibility between Central Sui and Yang’an Sui is probably also very low.
In terms of test design, we feel that one particular weakness was the length of our sentences. Most of our sentences consisted of at least four separate clauses. They were significantly longer than the
sentences used in a previous Lalo intelligibility survey in Yunnan Lam et al., 2010. Other aspects of these two surveys’ methodology were virtually identical. The Lalo survey’s sentences generally
comprised no more than three clauses each, and the raw home town test results were consistently higher— an average of around 98 for the Lalo survey as opposed to around 90 for the Sui survey.
Because our sentences were longer, the test subjects had to remember more details and were more likely to inconsistently leave details out during the retelling.
The quality of the recordings and the clarity of speech of each of the four speakers in the RTT recordings also surely had some influence on the results. For example, the SY Southern speaker spoke
faster and occasionally “swallowed” his words, making it harder for test subjects to pick out some details. Nevertheless, our scoring method took account of most such instances—if the home town test
subjects could not consistently retell certain elements, we did not count these elements in the final scoring.
For this particular study, we feel that the sentence L2 retelling or “sentence translation” method worked well because of the relatively high levels of proficiency in Chinese dialect Southwestern
Mandarin among our participants. In other surveys, a mother-tongue retelling method may work better, although the test design would have to be altered to prevent successful “mimicking” of words being
counted as successful comprehension.
180
9 Conclusions
Andy Castro
9.1 Visualisations