The therapeutic process is characterized by a paradox. On the one hand, the therapist confronts the patient, who has been rendered incompetent by his neurosis, in the role of teacher and provider of potentially valid linguistic interpretations. On the other hand, this difference in power must be cancelled at the end of treatment and the patient must be set free. The proposed solution of this problem is that psychoanalysis, as 'research project', is obliged to subject the rules and conditions it creates for the attainment of its goals to its own investigatory methods. The truth which is at stake in the therapeutic process is not 'ontological' but relative, resting on interpersonal consensus.