One of the possibilities to find out the stage of teachers' professional development is to investigate the perception of instructional events. These methods have been used by Berliner et al. (1988) and by Sato et al. (1993). In the research that was carried out in Tartu University, a methodology similar to Sato, et al (1993) was used. The subjects of the experiment were shown a videotaped lesson of Estonian language in grade seven and while watching, they were simultaneously asked to comment on the outstanding moments and events. The methodology of Sato et al (1993) was initially followed, according to which the comments of the research subjects were divided into idea units and thereafter the idea units about teaching and learning were analysed separately. The categorising method can be described as follows:The subjects' comments were divided into (idea)units. First, it was determined whether the unit was about teaching or learning and then, in turn, whether it was a fact, impression or interpretation. Besides that, the idea units emerging in the comments were categorised according to significance and relevance. Despite following the methodology, the categorising of idea units often appeared to be unreliable in Estonian circumstances.The modified methodology of analysis is based on Gagné's model of instructional events. In case of every comment it was decided which instructional event it was about and whether it was a fact or an interpretation. During the analysis difficulties appeared in differentiating between some instructional events, on the same time it became clear that Gagné's model does not comprise some teaching- and learning-related points, which caught the eye of the research subjects. Thus, two new categories were added which treat the concrete organisatory actions in class or the general organisation of study work. The development of the new methodology can be described by the following aspects: 1. The stimulating material - the videotape of the class - was divided into components according to Gagné's model of instructional events in order to decide more easily which stage of the class the subject's comment belonged to. 2. The comments, taperecorded during the viewing of the film, were transcribed and the received text was divided into segments according to speech pauses. The comments were equipped with a time scale so that while categorising them, it would be possible to compare the subject's text with the class recording, which was the basis for the experiment. In order to simplify the coding/categorising of the segments, a coding table was used together with texts describing the categories. It appeared, that one comment could sometimes reflect several instructional events.3. Gagné's model was replaced with a simplified version, where some instructional events were united (like checking what was studied, giving the feedback and assessing the study results were now viewed as "checking"), for the text analysis proved that those stages were often overlapping or very hard to distinguish in the subjects' comments.4. In order to simplify the data analysis the comments which described several stages of the class were divided into parts. 5. When difficulty arised in deciding whether a text segment can be classified as any of Gagné's stages of instructional events or as an organisatory action, the former classification was preferred.6. The analysis proved that comments can often be classified as inapropriate, irrelevant, and significant. It was decided that the relevance (significance) would only be determined about comments viewed as interpretations.7. In order to increase the reliability of categorising, a consensus-based deciding was applied.The paper introduces the elaborated method of content analysis in details and draws some conclusion on research findings.