Includes bibliographies. ; The dissertation concerns the procedure of teacher supervision as presently conducted at the University of Cape Town's School of Education, and suggests ways in which this can be improved. It is argued that teacher education should be situated in the paradigm of Action Research, and within that, in the approach known as Participant Observation. The language of the classroom is chosen as the means whereby classroom practices can be investigated. A detailed account is provided of a full sequence of clinical supervision, and use is made of complete lesson and subsequent interview transcriptions in this account. The methodological techniques of participant observation, interviews, 'triangulation' procedures and fieldnotes to supplement the analysis of transcripts and interviews, are used. In addition, three different ways of analysing classroom language are described, viz . discourse analysis, insightful observation and coding schemes, and their appropriacy for teacher supervision purposes assessed. Finally, teacher conclusions are drawn regarding an effective model for supervision. It is suggested that FIAC (Flanders' Interaction Analysis Categories) be used in combination with lesson transcriptions, and stress is laid on the need for teacher trainees and trainers to observe at all times the criteria for participant observation. Suggestions are put forward concerning possible implementation of recommendations made, and the need for future research in this area is emphasised.