• No results found

Chapter 7 Quantitative results

7.2 Inferential statistics

The selection of the Mann-Whitney test is based on two principles which Gries (2014) suggested: the type of study being conducted and the type of variables involved. The quantitative strand of the current research aims at comparing two data sets, which are Appraisal options used in texts by L1 Chinese writers and those in texts by L1 English writers. The two sets are

176 independent in that there is no relationship between the observations of the use of Appraisal options in each of the L1C and L1E sub-corpora or between the two

sub-corpora. Therefore, the current research suits what Biber and Jones (2009) call a type of study which is designed to describe the “differences among texts and text varieties” (p. 1290). Each text is treated as “an observation,” and the “rates of occurrence” of linguistic features are interval variables that can be subjected to inferential statistics (p. 1290).

As the present corpus has a relatively small number of texts, a

non-parametric test is appropriate in that it does not assume a normal distribution of data (Field, 2009; Oakes, 1998). The Mann-Whitney U test, as the

non-parametric test for two independent samples, therefore is chosen to examine whether the observed differences in the relative frequencies (per 1,000 words, see 7.1.2) of Appraisal options used by L1 Chinese and L1 English writers are statistically significant. The null hypothesis predicts that there is no statistically significant difference in the use of Appraisal options by these two groups of writers, and a two-tailed Mann-Whitney test (p<0.05) was run on all data sets to examine the hypothesis.

177 Table 7.1 The Mann-Whitney Test Result of Appraisal

Ranks

L1 N Mean Rank Sum of Ranks

Normalized frequency of appraisal L1C 6 6.17 37.00 L1E 6 6.83 41.00 Total 12 Test Statisticsa Normalized frequency of appraisal Mann-Whitney U 16.000 Wilcoxon W 37.000 Z -.320

Exact Sig. (2-tailed) .818

a. Grouping Variable: First language

Overall, the null hypothesis was confirmed with a 95% degree of certainty. As can be seen from Table 7.1, the test result showed that the

normalized frequency of Appraisal in the L1C sub-corpus (Mdn = 59.30) did not

differ significantly from that in the L1E sub-corpus (Mdn = 62.48), U =16.00, p =

0.818, r = -0.09. However, this result should be interpreted as indicating that no statistically significant difference was found in the current corpus rather than that the two groups of writers represent a similar population in terms of using

Appraisal. This needs to be examined in further research by including more samples from each group. This result is also at odds with many previous studies which observed differences in using interpersonal language by L1Chinese and L1English writers who were at a lower education level (e.g. Chen, 2010; Hinkel, 1997, 2003; Hyland & Milton, 1997). It therefore can be assumed that as

178 language proficiency increases and experience in academic education

accumulates L1 Chinese writers, at least those in this study, have taken good command of making interpersonal meanings in thesis writing. This positive result also indicates that perhaps at this higher level of education the first language would not be a constraint on using devices for making interpersonal meanings.

Similarly, the L1 Chinese writers did not differ significantly from the L1 English writers in the use of the two main categories of Appraisal – Attitude and Graduation. Table 7.2 presents the inferential results about the comparisons of the normalized frequencies of Attitude and Graduation between the L1C and L1E

sub-corpora, respectively.

Table 7.2 The Mann-Whitney Test Results of Attitude and Graduation

Ranks

L1 N Mean Rank Sum of Ranks

Normalized frequency of attitude L1C 6 5.33 32.00 L1E 6 7.67 46.00 Total 12 Normalized frequency of graduation L1C 6 6.17 37.00 L1E 6 6.83 41.00 Total 12 Test Statisticsa Normalized frequency of attitude Normalized frequency of graduation Mann-Whitney U 11.000 16.000 Wilcoxon W 32.000 37.000 Z -1.121 -.320

Exact Sig. (2-tailed) .310 .818

179 Table 7.2 shows that the normalized frequencies of Attitude and

Graduation in the L1C sub-corpus (Mdn = 2.20 and 12.37, respectively) did not

differ significantly from those in the L1E sub-corpus (Mdn = 4.43 and 12.35,

respectively), U =11.00 and 16, p = 0.310 and 0.818, r = -0.32 and -0.09, respectively.The statistical test also confirmed that there is no significant difference in the use of Bare Assertions and Heterogloss by the L1 Chinese writers and the L1 English writers , U =10.00 and 17, p = 0.240 and 0.937, r = -0.37 and -0.05, respectively (see Table 7.3 below). As reviewed in 3.2, Swain’s (2010) research showed that non-native undergraduates generally experienced more difficulty with effective deployment of Engagement resources. However, the present finding indicated that doctoral students at least those in this study seemed to show similar rhetorical behavior in using Engagement particularly in terms of the relative frequency. Furthermore, the similar use of Heterogloss seems to indicate both groups of writers’ awareness of engaging with the reader and construing dialogic divergences for their texts.

Table 7.3 The Mann-Whitney Test Results of Bare Assertion and Heterogloss

Ranks

L1 N Mean Rank Sum of Ranks

Normalized frequency of bare assertion L1C 6 7.83 47.00 L1E 6 5.17 31.00 Total 12 Normalized frequency of heterogloss L1C 6 6.33 38.00 L1E 6 6.67 40.00 Total 12

180 Test Statisticsa Normalized frequency of bare assertion Normalized frequency of heterogloss Mann-Whitney U 10.000 17.000 Wilcoxon W 31.000 38.000 Z -1.281 -.160

Exact Sig. (2-tailed) .240 .937

a. Grouping Variable: First language

One more interesting finding to note is the Mann-Whitney test result of the use of Pronounce across the two sub-corpora, as shown in Table 7.4. Table 7.4 The Mann-Whitney Test Results of Pronounce

Ranks

L1 N Mean Rank Sum of Ranks

Normalized frequency of pronounce L1C 6 8.33 50.00 L1E 6 4.67 28.00 Total 12 Test Statisticsa Normalized frequency of pronounce Mann-Whitney U 7.000 Wilcoxon W 28.000 Z -1.761

Exact Sig. (2-tailed) .093

a. Grouping Variable: First language

Although the difference in the relative frequency of Pronounce between the L1C and L1E sub-corpus is non-significant, the probability (p = 0.093) is the least among all the Mann-Whitney results of other Appraisal categories and is closest to the critical p-value 0.05. This result seems to be understandable considering that Pronounce functions to make propositions with high authorial

181 emphasis and that L1 Chinese writers tended to use intensified assertions as reported by some previous studies (e.g. Chen, 2010; Hyland & Milton, 1997). Pronounce was also reported to be accounting for the lower score of

undergraduate argumentative essays in Wu’s (2007) study as frequent use of Pronounce appeared to result in a sense of highly insisting on the validity of one’s proposition and even some unwarranted assertiveness. Therefore, it would be useful to test if the difference becomes significant when more samples are included for comparison in future research.

182