This section will illustrate how the data are analyzed in the present study, including ways of identifying pragmatic transfer, in section 3.5.1, and how we performed quantitative analysis, in section 3.5.2, and qualitative analysis, in section 3.5.3.
3.5.1 Identifying Pragmatic Transfer
There are two major ways to identify pragmatic transfer (Kasper, 1992). One is the informal estimation of differences and similarities among the percentages of L1, L2, and IL. That is, similarities among L1, L2, and IL are categorized as positive transfer; differences between IL-L2 and L1-L2, and similarities between L1-IL are categorized as negative transfer (i.e., L1 influence). The other way to examine L1 transfer at the pragmatic level is to see if there are significant differences in a certain pragmatic feature between interlanguage and learners’ native language. More specifically, positive transfer (i.e., successful L1 transfer without leading to pragmatic failure) is operationally defined as the “lack of statistically significant differences in the frequencies of a pragmatic feature in L1, L2, and IL” (p. 223). That is, L1, L2, and IL do not differ significantly from one another. Negative transfer (i.e., the inappropriate L1 transfer to L2) is operationally defined as “statistically significant differences in the frequencies of a pragmatic feature between IL-L2 and L1-L2 and
lack of statistically significant differences between IL and L1” (p. 223). That is, EFL learners’ L2 performance is similar to L1 baseline group’s rather than L2 baseline group’s performance. Since the present study involves quantitative and qualitative analyses of correction, both of the two ways will be employed to examine interlanguage performance.
The pedagogical implication for examining pragmatic transfer is that if EFL learners exhibit negative transfer in correction, this implies that they do not fully acquire the target-like pragmatic knowledge. Then there is still space left for learners’
to make progress and for instructors to help learners acquire the pragmatic knowledge to overcome L1 transfer that may lead to pragmatic failure.
3.5.2 Quantitative Analysis
To examine the differences among and within groups in the quantitative analysis of perception of face-threat in SRQ and strategy use in DCT, the SPSS software is employed to carry out the statistical analysis. As for the perception data in SRQ, the analysis of variance (i.e., one-way ANOVA) is used to examine the differences of perception among groups. If significant differences are detected, post-hoc comparisons of Scheffé tests are employed to examine the differences between each pair of groups. In addition, to measure the influence of social factors, i.e., social status and social distance, on participants’ perception of face-threat, one-way ANOVA is used to investigate participants’ perception in different social status situations,22 and t-test is used to examine the effect of social distance on participants’ perception.
Regarding the use of correction strategies, two types of Chi-square tests are employed: tests of independence23 and tests of goodness of fit. Tests of independence
22 When there are significant differences, post-hoc comparisons of Scheffé tests are also executed to detect the differences between each pair of groups.
23 The results of tests of independence are the same as those of tests of homogeneity because they are
are used to examine the relationship between the variable of group and the variable of strategy use on the whole. When the result gets to the significance level, this shows that the variable of group has strong association with the variable of strategy use. That is, participant groups have preferences in the use of a certain strategy. As for tests of goodness of fit, they are used to investigate the differences of a certain strategy used by the four groups. If significant differences among the four groups are detected, it is not easy to examine the post-hoc comparisons, for the results can not be evaluated by the present SPSS software. Some of previous studies (e.g., H. Chen, 2006) used several pairs of Chi-square tests to examine the differences between each pair of groups. However, this approach may increase the likelihood of Type I error.24 A better way is to employ the formula of a posteriori comparisons in Lin (1992) because this procedure makes it possible that all the analyses share the same Type I error. The formula of Lin’s (1992) posteriori comparisons is shown in Figure 3.2.
'
Figure 3.2 A formula for posteriori comparisons (Lin, 1992, p. 291)
Note: ( ' )
j p j
p −
ψ = is the difference of percentage between two groups.
'
derived from the same formula (cf. Chang, Chang, & Lin, 2000; Lin, 1992). The difference between a test of independence and a test of homogeneity lies in their different foci: the former one aims to examine the relationship between two investigated variables, whereas the later one attempts to examine the differences of strategy use among groups.
24 Type I error refers to the fact that when actually there is no difference, the researcher thought there is a difference.
3.5.3 Qualitative Analysis
Qualitative analyses will provide much insight for the understanding of the speech act of correction as previous studies have indicated that different ethnic groups may employ the same set of strategies with different types of content and different linguistic forms (e.g., Blum-Kulka, 1983; Chen, 2007; H. Chen, 2006; Lin & Chen, 2006; M. Chen, 2006). Thus, the present study conducted qualitative analyses on participants’ production (e.g., linguistic forms, content of strategies, and softening devices) with a focus on the comparison between interlanguage and L2.
3.6 Summary
This chapter first provided the information about the four groups of participants in the present study. Then the rationale for employing the three research instruments, i.e., Discourse Completion Tasks, reasons of opting out, and Scaled Response Questionnaire, was discussed in section 3.2. Section 3.3 described the procedure of data collection. Based on the collected data, in section 3.4, we developed a data-driven coding scheme. The scheme includes the 13 strategies under the three main categories (i.e., direct strategies, indirect strategies, and opting out), 11 types of reasons of opting out, and softening devices (i.e., external modifications and internal modifications). All the collected data were coded by the researcher, and 20 percent of the data were randomly selected and coded by a second researcher for interrater reliability. Last, in section 3.5, we also illustrated our data analyses, including ways of identifying pragmatic transfer, quantitative analysis, and qualitative analysis.
CHAPTER FOUR
RESULTS AND DISCUSSION
This chapter provides the results and discussion of the present study.
Participants’ perception of the face-threat scaled from SRQ will be presented first in section 4.1. The perception usually influences participants’ production, i.e., the use of correction strategies, which will be analyzed quantitatively and qualitatively in section 4.2. When choosing to use verbal strategies to realize the correction, people sometimes use more than one strategy to ensure successful meaning transfer. Thus section 4.3 will present the different correction patterns used by the four groups with a focus on two-strategy patterns. In addition, the verbal strategies are often mitigated by external modifications. Thus section 4.4 is dedicated to the discussion of the use of external modifications and their relation to the head act strategies. Last, section 4.5 will have a general discussion on the findings of the present study.
4.1 Overall Results of Scaled Response Questionnaire (SRQ)
Scaled response questionnaire (SRQ) is used to examine participants’ perception of face-threat toward the speech act of correction in different scenarios. In SRQ, the subject’s perception with the least face-threat is represented by the number one and the perception with the greatest face-threat by the number five. The overall means of SRQ is shown in Table 4.1. The results from one-way ANOVA show that there were significant differences among the four groups (F = 99.59, p = .000***). Post-hoc analysis of Scheffé test indicated that the two EFL groups scaled the face-threat significantly higher than the NSC group ( p = .000***). However, the means of NSC does not differ significantly from NSE ( p = .305). That is, both Chinese and Americans had similar perception of face-threat toward the speech act of correction.
Table 4.1 Overall means of perception of face-threat from SRQ by the four groups NSC EFL-L EFL-H NSE
M SD M SD M SD M SD
F ( p) & post-hoc comparison
2.83 1.34 3.28 1.27 3.26 1.33 3.02 1.38
F = 99.59, p = .000***
NSC & EFL-L (.000***) NSC & EFL-H (.000***) NSC & NSE (.305) EFL-L & EFL-H (.998) EFL-L & NSE (.065) EFL-H & NSE (.102) Note. ***p <.001
Thus, both monolingual groups were predicated to have similar preference in the realization of correction strategies of similar directness level. On the other hand, learners scaled the face-threat significantly higher than the NSC group, and they were predicted to employ less direct strategies and use more modifications for mitigation than the NSC group. Furthermore, the means between EFL-L and EFL-H did not differ significantly from each other ( p = .998). This suggests that learners’
proficiency level did not have a great influence on learners’ perception. Since the two groups of EFL learners had similar perception toward correction, they were predicted to use correction strategies of similar directness level.
On the other hand, participants’ perception of face-threat can be influenced by the two contextual factors, social status and social distance. The influence of social status on participants’ perception will be discussed in section 4.1.1, and the influence of social distance will be discussed in section 4.1.2.
4.1.1 SRQ and Social Status
The result of means of perception of face-threat scaled from SRQ in different social status situations (H to L, Equal, L to H) is shown in Table 4.2. The four groups scaled the imposition most face-threatening in low to high status situations, then in high to low status situations, and least face-threatening in equal status situations. The four groups’ perception of greater face-threat in low to high and high to low status
Table 4.2 Means of perception of face-threat from SRQ in different social status
Overall .882 (.415) Overall 6.107 (.002**) Equal & L to H (.003**)
Overall 5.102 (.007**) Equal & L to H (.008**)
Note. *p < .05, **p < .01, ***p < .001
situations may indicate that they tended to be concerned more about the relationship with people of asymmetric status.
The four groups’ perception can be clearly demonstrated in Figure 4.1. The two EFL groups’ average means of perception are higher than the two baseline groups’ in the three different social status situations. Learners’ averagely perception of greater
Figure 4.1 Means of perception of face-threat from SRQ in different social status
situations by the four groups
face-threat could result from the fact that they were in lack of confidence and felt insecure in using an unfamiliar language (Edmondson & House, 1991; Lin, 2006a 2006b). Then they scaled the face-threat higher. This follows that they were predicted to use less direct strategies and more modifications for mitigation.
4.1.2 SRQ and Social Distance
The result of means of perception of face-threat scaled from SRQ in different social distance situations is shown in Table 4.3. The four groups tended to scale the face-threat higher in correcting an unfamiliar person than in correcting a familiar person. However, they demonstrated different sensitivities toward social distance. The two groups of learners perceived the imposition significantly more threatening in greater social distance situations than in less social distance situations (EFL-L: t = 3.559, p = .000***, and EFL-H: t = 3.182, p = .002**), whereas Americans did not ( t
= .535, p = .594). Learners’ perception in different social distance situations could be influenced by their L1 because the NSC group also scaled the offense significantly more threatening in greater social distance situations than in less social distance situations ( t = 2.208, p = .029*). This implies that social distance was an important variable for Chinese and learners’ perception in different social distance situations, but not for Americans.
Table 4.3 Means of perception of face-threat from SRQ in different social distance situations by the four groups
NSC EFL-L EFL-H NSE
Figure 4.2 Means of perception of face-threat from SRQ in different social distance situations by the four groups
Furthermore, the two EFL groups’ perception of face-threat was greater than that of the two baseline groups in both of the two different social distance situations. This finding, similar to that in section 4.1.1, exhibited learners’ insecure feeling and lack of confidence in the use of a foreign language. Thus, learners were predicted to use less direct strategies and more modifications to soften the imposition.
4.1.3 Summary
SRQ is used to elicit participants’ perception of face-threat toward correction.
The results of this section can be summarized as follows. In the overall results of SRQ, there were no significant differences between the native Chinese and Americans’
perception. However, the two EFL groups scaled the face-threat significantly greater than the NSC group. This reflects learners’ insecurity in learning English. Thus, learners were predicted to employ less direct strategies and to use more modifications for mitigation than the two baseline groups.
In addition, participants’ perception could be influenced by social status and 0
1 2 3 4
+D -D
NSCEFL-L
EFL-H
NSE
social distance. The four groups tended to scale the imposition most threatening in correcting people of higher social status and those of greater social distance. These findings on the two factors as important variables are in accordance with Brown and Levinson’s (1987, 1978) claim which suggests that to perform speech acts on people of greater social power and social distance would result in greater face-threat. Thus, participants were predicted to use less direct strategies when correcting people of higher social status and greater social distance.