SummaryThe author analyses two experiments with alternative forms of training of teachers of modem languages, carried out in Greece in 1982 to 1985. Examining first the ideology of training, she first calls in question some basic assumptions concerning teacher models and the processes of training, which risk perpetuating traditional, teacher‐taught, attitudes. The concept of training is set against that of teacher autonomy; the rigidity suggested by ‘training’, it is suggested, can be counter‐productive. The approach adopted in the experiments described has been that of maximising autonomy.The first experiment extended over three years, commencing 1982. It aimed at reestablishing the theory‐practice link, in the hope of bringing into being a ‘teacher‐researcher’. The strategies adopted are described, with an emphasis on the multiplier effect of the first group on return to their schools, and with a subsequent extension to initial training as well as in‐service. Despite difficulties, positive results were achieved.The second experiment commenced in 1983 and drew upon the existing project. It aimed at the production of new curricula and evaluation methods, and upon new curriculum materials. Teachers were actively involved in the process, their involvement being essential in the eyes of the project team. The experience was highly innovatory, leading to a complete restructuring of class activities and to the emergence of new pupil‐teacher relationships, and of a new learning transaction.