Questionnaire Matched guise test Foreign researcher results

MGT, but not for the questionnaire. Only in the education level variable was there a match-up between the questionnaire and MGT, whereby significant differences in scores were noted in both. Yet in the education level variable the directionality of attitudes towards Indonesian was reversed in the MGT from the questionnaire. In many of the cases, though, individual questions in the questionnaire specifically dealing with language attitudes did result in more similar findings with the matched guise test.

4.5.5 Summary of matched guise test results

Overall, a significant difference between the Indonesian and Jambi Malay recordings in the MGT was not found in this study. Two of the speakers had very slightly higher scores for Jambi Malay versus Indonesian, and the remaining two speakers had higher scores for Indonesian than Jambi Malay. Where Indonesian had a higher score, there was a larger gap between SI and JM, and for one of the speakers it was statistically significantly larger. However, in that statistically significant case, the results may be skewed since the speaker’ voice was often mistaken for an ex-governor’s wife. Thus, the difference between attitudes towards JM and SI was so scant a distinction can hardly be made. But the scales were slightly tipped towards more positive attitudes towards Indonesian than JM. The differences found between the MGT and the questionnaire were also discussed, and it was shown how in every variable studied, MGT results varied from questionnaire results. This could be due to a problem inherent in the MGT which skewed results, or simply that reported language attitudes may not always be the same as actual attitudes observed or tested for indirectly. However, in many of the individual questionnaire questions which dealt with language attitudes, findings were more in line with MGT results. This indicates that language attitudes do not necessarily match up with language usage, which seems to be true for Jambi Malay people. This phenomenon was observed by Southworth in India 1985:227–228; he discussed the tendency for people to regard one language as “best”, but “[i]f we look at the actual behavior 10 of people in contemporary stratified societies, we see a different picture. Whatever people’s expressed or unexpressed views toward ‘correct’ language may be, when they use language they tend to use a variety which is appropriate to the context.” 4.6 Differences among researchers 4.6.1 Introduction In order to test for validity of the data, a test was performed on the average scores according to researcher. There was a total of seven researchers who administered both the questionnaire and the MGT, three of which were native to the communities under study referred to hereafter as local, L in the tables and figures, and three of which were residents of Jambi City referred to as non-local, NL.

4.6.2 Questionnaire

In figure 4.25 the mean questionnaire score for the local researchers was 1.443, and the mean score for the non-local researchers was 1.433—a minute difference. In fact, table 4.66 shows that the difference between local and non-local researchers’ scores was not statistically significant t statistic .45, p. 6551. 10 The emphases are Southworth’s. Figure 4.25 Average questionnaire scores by localnon-local researcher Table 4.66 Differences in average questionnaire scores by researcher test: independent samples t- test

4.6.3 Matched guise test

In figure 4.26, it is shown that in the MGT, like in the questionnaire, there was little difference between local and non-local researchers, though the local researchers again had a slightly higher average score. The difference between local and non-local researchers for the MGT was also not statistically significant, as seen in table 4.67 t statistic 1.25, p .2112. Figure 4.26 Average MGT scores by localnon-local researcher Table 4.67 Difference in MGT scores by researcher test: independent samples t-test

4.6.4 Foreign researcher results

In section 3.4 it was mentioned that during the pilot test it became clear that the foreign researcher’s presence disturbed the research process and might have lead to skewed results. This theory was tested, and the results will be given here. Figure 4.27 shows the average questionnaire scores by researcher. Note that there was a substantial amount of variation, with a Mudung Laut local A having the highest average score, and a foreign researcher K having the lowest score. The variation was statistically significant, as seen in table 4.68. An ANOVA test was performed on the following pairs: local researchers versus pilot test researchers P, and non-local researchers versus pilot test researchers. In both of those cases, the difference was statistically significant, though the composite F statistic was 2.97, p .0530. Figure 4.27 Average questionnaire scores by researcher including pilot test researchers Table 4.68 Questionnaire score by researcher including pilot test researchers test: one-way between subjects ANOVA In looking at the mean MGT scores, there was much less variation among researchers than in the questionnaire see figure 4.28. Here, the highest scores were found with the two researchers native to Mudung Darat S and T, while the lowest score was found with D, a pilot test researcher. But like the questionnaire, the variation between researchers in the pilot test and both local and non-local researchers was statistically significant see table 4.69. Figure 4.28 Total MGT averages by researcher including pilot test researchers Table 4.69 MGT scores by researcher including pilot test researchers test: one-way between subjects ANOVA

4.6.5 Summary of differences between researchers