Recorded Text Test RTT: Story question and answer method

Extra-linguistic factors such as attitudes towards speakers of another dialect can also play an important role in intelligibility Wolff 1959, Schüppert and Gooskens 2010. If people are motivated to understand another dialect, they may understand a great deal; if they are not motivated at all, they may understand very little. Levels of inter-dialect contact are very difficult to measure because of the complex social networks involved. Speaker attitudes and motivations are even more unquantifiable. Therefore most dialect intelligibility studies have sought to isolate and measure inherent intelligibility. The same is true for this study.

8.3 The history of intelligibility testing

Dialectologists have used various methods for testing inherent intelligibility with varying levels of success. No one method has emerged as the most effective and reliable Gooskens 2013. We describe a few of these here before explaining our own method of testing, the “sentence retelling method”, which we believe to be a significant improvement on other methods. Before we proceed, some words should be said about opinion testing. The simplest way of evaluating mutual intelligibility is to ask the speakers directly how well they understand the other dialects in question. This can be done in two ways: 1 ask speakers of dialect A how well they understand speakers of dialect B; or 2 play speakers of dialect A a recording of dialect B and ask them how much they understand. By its very nature method 1 restricts the researcher to subjects who have had contact with other dialects, thus producing a measure of perceived cross-dialect inherent and acquired intelligibility. Method 2 seems more promising. However, various studies have shown that reported perceived intelligibility varies greatly from speaker to speaker. Towards the end of their interesting study of intelligibility between Chinese dialects, Tang and van Heuven conclude: We feel, then, that the correlation between [opinion and functional] tests is not good enough to recommend the indiscriminate use of opinion tests as a substitute for functional test procedures. We advocate that, whenever the resources are available, mutual intelligibility should be tested functionally. Tang and van Heuven 2009:727 We now look at some methods of functional inter-dialect intelligibility tests which have been used in the past.

8.3.1 Recorded Text Test RTT: Story question and answer method

SIL International has more experience in intelligibility testing than any other linguistic or academic institution in the world. For many years, the most popular method used by SIL linguists has been the Recorded Text Test RTT. This was first developed by Bradley 1968 and promoted through Casad’s 1974 manual Dialect Intelligibility Testing. The test was developed after trying out various other methods, including translation and retelling methods, which are described in detail by Casad 1974:52- 86. The Recorded Text Test was introduced to China by Allen 2004 in his Bai dialect survey. Variations of it have been used by, among others: Chang et al., 2009 for Nisu; Hansen and Castro 2010 for Zhuang; and Shi 2010 for Kam. The basic method involves making recordings of free narrative texts in each dialect to be tested. The recordings are then played back to subjects in each test location and content questions are asked, in the subjects’ own dialect, to test comprehension of the text. In the final version of the test, each question is embedded in the recorded text after the part of the text to which it pertains. Advantages of this method include: 1 participants listen to natural texts with familiar content; 2 participants hear and respond to questions in their own dialect—no translation is involved during administration of the final test; and 3 since the test is based on only ten questions, it is extremely straightforward to administer and score. In our view, though, the drawbacks of this method outweigh the advantages. Firstly, the authors’ own experience has shown that it is extremely difficult to elicit non-repetitive narrative texts sufficiently abundant in content to produce meaningful measures of intelligibility. Secondly, comprehension scores resulting from this type of test cannot be meaningfully compared with each other since every story is different in content. Thirdly, participants are often uncomfortable with a “question and answer” type test, especially when administered by outsiders. Fourthly, only ten questions are asked about each text. The notion that correct responses to such a limited set of stimuli represent “full comprehension” of the text is dubious. Finally, preparation of this type of test is complex and cumbersome, involving translation of the questions into every dialect on which the text will be tested and insertion into every recording in the appropriate place.

8.3.2 Word recognition tests