Comparing frequency-based with sequential discourse analysis: convergence or divergence?
I. Karasavvidis, J.M. Pieters and T. Plomp
University of Twente, The Netherlands
Discourse analysis in psychology and education can be distinguished into two broad categories: qualitative and quantitative. In the case of the latter, discourse is firstly segmented and these segments are subsequently assigned to a number of codes. The frequencies of occurrence of these codes are used as scores for further statistical analyses. Exclusive consideration of frequencies is problematic on two counts. Firstly, an utterance always addresses and responds to previous utterances and, thus, is dialogically related to them. Secondly, the same utterance in two different contexts may have potentially different meanings, and, therefore, the local context has a bearing on the meaning of the utterance [4]. It is argued that sequential analysis of discourse provides a very promising alternative to analyzing classroom discourse because the dialogic relationship of each code with the others is maintained as well as the meaning of every code is preserved by considering specific codes preceding and following it. The importance of sequential analysis of discourse is illustrated by examining data from an empirical study. Twenty grade ten students were divided into two groups of ten and were tutored by their teacher in how to solve correlational problems. Students in one group used paper and pencil, while students in the other used a spreadsheet. All tutorials were videotaped, transcribed verbatim and subsequently coded by two trained judges using a coding scheme we developed for the purposes of the study. The main question we were trying to answer was whether the two groups differed as far as the teacher and student codes were concerned. The discourse data was subjected to a conventional frequency-based statistical analysis as well as to sequential analysis [1, 2, 3]. A number of statistically significant differences emerged when the groups were compared using the code frequencies as a criterion. In sharp contrast, a drastically different picture of the data surfaced with sequential analysis, some discourse patterns being either unique to each group or more likely in either of them. These divergent findings are discussed and some implications for discourse analysis in the educational and psychological scene are drawn.
Poster presented at Measuring Behavior '98, 2nd International Conference on Methods and Techniques in Behavioral Research, 18-21 August 1998, Groningen, The Netherlands