In this section, the ways of how data were analyzed are introduced. This section starts with the coding system of language teacher code-switching used in this study. Then, the procedure of data analysis is introduced.
The Coding System of Functions of the Teacher’s Code-switching The coding system used in this study was based on Moradkhani (2012), which was adapted because it was the most comprehensive coding system among the several that I had reviewed (Flyman-Mattsson & Burenhult, 1999; Rolin-Lonziti & Brownlie, 2002). The scheme includes six functions, including translation, metalinguistic uses, communicative uses, managing the class, building rapport with students, and
providing instructions, as shown in Table 2.
30
Table 2
The Coding System Used in the Study
Function Subcategory
Translation Translation of a word
Translation of an entire sentence Metalinguistic uses Contrast
Comment
Grammar explanation Highlighting
Communicative uses Clarification
Checking comprehension Directives
Marker
Managing the class Reprimand or disapprove Giving feedback
Reminder Building rapport with students Telling jokes
Revealing emotions Providing instructions Giving instructions
Giving prompts Pointer
Nominate
Using administrative vocabulary
Note. Adapted from “The typology of EFL teachers’ code-switching: A validation study” by S. Moradkhani, The Journal of Teaching Language Skills, 37(1), pp. 113-115.
The Procedure of Data Analysis
Merriam (2009) has stated that “Collection and analysis should be a
simultaneous process in qualitative research” (p. 169). Following her suggestion, I started to analyze the data while I was still collecting data. In this study, the data were analyzed in the following stages. First, I transcribed the recorded lessons verbatim (See Appendix C). As mentioned above, the classroom was observed over a whole semester, excluding the introduction of the course (Week 1), the Double
31
Tenth Day (Week 4), the midterm exam (Week 10), the dance contest (Week 13), and the New Year’s Day (Week 16), and the final exam (Week 18). In total, the class was observed for 12 weeks. The data collection schedule, the duration of the classes, and the time the teacher actually talked are shown in Table 3.
Table 3
Data Collection Schedule
Stage Week Date Total class time Actual teacher talk %
Beginning 2 103/09/26 1:28:32 0:55:39 63
3 103/10/03 1:22:20 0:55:45 68
5 103/10/17 1:20:58 0:46:55 58
6 103/10/24 1:22:50 0:51:10 62
Middle 7 103/10/31 1:20:48 0:55:58 69
8 103/11/07 1:21:45 0:55:48 68
9 103/11/14 1:25:54 0:53:12 62
11 103/11/28 1:14:04 0:49:04 66
12 103/12/05 1:03:08 0:26:14 42
End 14 103/12/19 1:17:29 0:52:53 68
15 103/12/26 1:20:16 0:55:00 69
17 104/01/09 1:07:00 0:46:00 69
As shown in Table 3, the teacher’s class was observed for 12 weeks, but for this study, I decided to analyze four classes in the beginning, middle, and near the end of the semester respectively (i.e. weeks 2, 3, 7, 8, 15, and 17) because there was the most maximum percentage of language the teacher spoke in these class periods. When transcribing the recorded lessons, I followed the conventions for transcriptions (See Appendix D) adapted from Yu (2007). In the transcripts, students’ conversation practice with their classmates, students’ grammar exercises, and the teacher’s roll-call were not taken into account because in these situation, the teacher did not speak English or Chinese for the teaching purpose.
32
Second, to analyze the functions of the teacher’s code-switching, I coded his utterances based on the coding system adapted from Moradkhani’s coding scheme.
Afterwards, I calculated the frequencies of different functions of the teacher’s code-switching to see when he frequently code-switched from English to Chinese.
Third, the data from the interview (See Append E) and the stimulated recall interview (See Append F) were also transcribed verbatim in Chinese to obtain the teacher’s teaching belief and his perceptions of code-switching in the classroom.
The interview data were analyzed through content analysis to generalize the themes.
It is noted that I translated some of the transcripts from videotaped recordings and interview into English to show the findings.
Trustworthiness
The issue of trustworthiness is crucial for qualitative researchers because it “lies at the heart of issues conventionally discussed as validity and reliability” (Seale, 1999, p. 467). In order to ensure the trustworthiness of the result of this study, three
techniques were adopted: prolonged engagement, persistent observation, and triangulation (Lincoln & Guba, 1985; Merriam, 2009).
First, prolonged engagement refers to the researchers spending adequate time in the field; it helps fulfill two purposes. First, it can prevent the data from being distorted due to the presence of the researcher and the video camera set up in the setting for data collection. With prolonged engagement, the participants involved will get accustomed to the presence of the researcher and the video camera. Thus, their behaviors in the field will be similar as usual (Wu, Sung, & Chien, 2010).
Second, it can also help the researchers to build trust with the people involved (Lincoln & Guba, 1985). For this study, I stayed in the classroom for an entire semester, observing the teacher and the whole class. By doing so, the people in the
33
classroom were familiar with my presence, and I was able to get reliable data.
Second, through persistent observation in the field, the researchers are more able
“to identify those characteristics and elements in the situation that are most relevant to the problem or issue” (Lincoln & Guba, 1985, p. 304). In this study, I observed the teacher and the whole class in almost entire semester, which allowed me to get general ideas and then focus on specific incidents where the teacher code-switched in the classroom. By doing so, more supporting data were obtained.
Third, to avoid misinterpreting the data, I also triangulated the research findings through several data collected in this study. Triangulation means that the researcher utilizes various types of data, methods, investigators, and theories to verify findings (Ary, Jacobs, & Sorenson, 2010; Shenton, 2004). In this study, after collecting the data, I checked whether the data from my observation, the interview, and the
stimulated recall interview were consistent. With the use of these multiple data sources, the internal validity was increased.
Besides the techniques mentioned above, the intercoder reliability was also considered to see how consistently the data were coded by two people. For this study, I (Coder 1) invited one female graduate student majoring in English, as Coder 2, to analyze a selected transcript. Before coding the data, I introduced the coding scheme to Coder 2, so that she was familiar with it. After that, we began coding the same sample transcript. We discussed the result of the coding until we two reached consensus. Once fully aware of the coding system, we then continued to code another transcript. The intercoder reliability was computed by the percentage agreement through SPSS 17.0. The results indicated that the intercoder reliability was at .77. The intercoder reliability was acceptable because the coefficient was from .61 to .8 (Ary, Jacobs, & Razavieh, 2009).
34