The present study proposes a new method for analyzing classroom discussion process. This method is called Inter-utterance Quotation Network Analysis (IQNA) that was developed with the concept of inter-textuality and network analysis. Using IQNA, researchers can obtain visualized representations of discussion process and indices of qualities of discussion. Although IQNA proposes several indices for quality of discussion and explorative analysis devices, the present study focuses on finding an index of how much a certain classroom discussion is actively managed by students and examining validity of it. Establishing such an index is very important task for educational research, since student’s active engagement in a classroom discussion is crucial for their knowledge construction. A reading comprehension unit for sixth- graders in a Japanese elementary school was examined with IQNA. As a result, it is found that “teacher’s frequency rate of being quoted by students” is most appropriate as an index of student-centered classroom discussion.

For the last decade, educational researchers have enthusiastically scrutinized classroom discourse. One of the reasons is that, at least, the researchers have recognized that learning is not just absorbing information, but reconstructing prior knowledge with new information. Discourse processes should be analyzed because learning process can be observed in verbal language communication in classroom. When they examine discourse process, qualitative analysis is frequently employed. Qualitative analysis enables us to access to ecologically valid and embodied findings in a natural context for learning. However, qualitative analysis has a few methodological limitations. It is difficult to establish reliability of a finding with qualitative approach. Accordingly, comparison and accumulation of the findings are also relatively difficult in general. More practically, it is time-consuming for fully understanding transcribed discourse not only for researchers but also for readers of published papers. Those characteristics may sometimes keep away some researchers and readers from the research on discourse process.

                To compensate those limitations of discourse analysis, this study proposes a new method for analyzing discussion process. This method is called Inter-utterance Quotation Network Analysis (IQNA) that was developed with the concept of inter-textuality (Kristeva, 1966) and network analysis. Using IQNA, researchers can obtain visualized representations of discussion processes (see Figure, though this function is not dealt here) and indices for qualities of discussion. A theoretical assumption of IQNA is that classroom discourse can be understood in terms of quotation relationships of words because student’s new understanding and interpretation consist of quotation from others’ utterance to some extent. In IQNA, “quotation” is operationalized as using the word previously uttered by a different person from the original speaker. Following this operationalization, quotation relationship can be computationally processed using transcription and measured quantitatively. Although IQNA proposes several indices for quality of discussion and explorative analysis devices, the present study focuses on finding an index of how much a certain classroom discussion is actively managed by students and examining validity of it. Establishing such an index is very important task for educational research, since student’s active engagement in a classroom discussion is crucial for their knowledge construction.

Method

Data Collection: Target classes were 15 lessons of a reading comprehension unit in an elementary school in Fukuoka city, Japan. The class consisted of 39 sixth-graders (21 girls and 18 boys). All students were native Japanese speakers. The teacher was Mr. Y who has been worked for over 20 years and is well known in the school district as an expert for student-centered communicative teaching. All lessons were videotaped and transcribed. Speaker’s name and the serial number of conversational turn were also identified.

Lesson Plan and Activities: A learning goal in the unit was to understand 3 fictional stories in depth and discuss on 3 stories so that they can construct their own interpretation of a main concept of the stories. One story of them is broadly included in Japanese textbooks for sixth-graders. Mr. Y designed a lesson plan for the unit in 3 steps: (i) Oral and silent reading, (ii) Writing on their first impression of the readings, (iii) Small group / whole class discussion. Two styles of discussion were introduced in the unit. One is teacher-initiated discussion and the other is student-initiated discussion. Of 15 lessons, 9 correspond to the former, 4 correspond to the latter, and remains correspond to neither. This characteristic of the classroom was used as a variable in later analysis.

Analysis

The transcription was analyzed with IQNA method. The procedure is as shown blow. (1) Prepared transcriptions of all classes. (2)Transcriptions were decomposed into morpheme. (3) Deleted irrelevant words referring the concordance. (4) Generated “word by context matrix” for each class. (5) Calculated “quotation relationships” among words. (6) Drew graphs using the quotation information. In 2nd, 3rd, and 4th processes, 2 freely distributed applications (ChaSen and KH Coder) were employed. In 5th and 6th process, the original programs that the author developed with Microsoft VBA on Microsoft Excel were employed.

Result

To explore an index of student-centered class, 9 candidate variables were prepared as blow. “Centralization” shown in (e) and (f) is a popular index used in network analysis. This index indicates how much a certain network is controlled by the most influential member.

(a) Total utterance frequency in whole class.

(b) Students’ utterance frequency.

(c) Teachers’ utterance frequency.

(d) Teacher’s rate of utterance frequency.

(e) Degrees of centralization for the frequency of quoting others’ words.

(f) Degrees of centralization for the frequency of being quoted by others.

(g) Total frequency of quotation in whole class.

(h) Teacher’s frequency rate of quoting students’ words.

(i) Teacher’s frequency rate of being quoted by students.

Among these candidates, variable (i) seemed the best index which would predict which class has more student-centered structure. To ascertain this finding, t-test was used to examine group difference in mean variable (i) between student-initiated classes and teacher-initiated classes. In student-initiated 4 classes (M = .04, SD = .02), teacher’s words were significantly less quoted by students than in teacher-initiated 9 classes (M = .14, SD = .01) (t (11) =3.93, p < 0.01, two-tailed).

Discussion

It is found that “teacher’s frequency rate of being quoted by students” is most appropriate as an index of student-centered classroom discussion. Of course, further validation is needed. A contribution of IQNA is that it would give us quantifiable measures for quality of discussion through computer processing. This enables researchers to compare different discussions and examine the factors which may affect quality of discussion in more empirical and explicit manner than qualitative discourse analysis.

                 A limitation of the index proposed is applicability to other genres of discourse. The classes analyzed here are open-ended creative discussions. It may be difficult to apply the same index to other sort of discussion such as discussion on mathematical problem solving in which the repertory of word used in classroom is fairly restricted. Therefore it is necessary to examine to what extent the index can apply.