Observations from therapeutic practice and a series of empirical findings, for example, those on discontinuous change in psychotherapeutic processes, suggest modelling the therapeutic process as a self‐organizing system with stable and critical instable phases and abrupt transitions. Here, a concept of psychotherapeutic change is presented that applies self‐organization theory to psychodynamic principles. The authors explain the observations and considerations that form the basis of the concept and present some connections with existing findings and concepts. On the basis of this model, they generated two hypotheses regarding the co‐occurrence of instability and discontinuous change and the degree of synchrony between the therapist and patient. A study design to test these hypotheses was developed and applied to a single case (psychodynamic therapy). After each session, patient and therapist rated their interaction. A measure of instability was calculated across the resulting time series. Sequences of destabilization were observed. On the basis of points of extreme instability, the process was divided into phases. Local instability maxima were accompanied by significant discontinuous change. Destabilization was highly synchronous in therapist and patient ratings. The authors discussed the concept and the methodological procedure. The approach enables the operationalization of crises and to empirically assess the significance of critical phases and developments within the therapeutic relationship. Copyright © 2013 John Wiley & Sons, Ltd. Key Practitioner Message We present a concept of change that applies self‐organization theory to psychodynamic therapy. We empirically tested the hypotheses formulated in the concept based on an extract of 125 long‐term psychodynamic therapy sessions. We continuously monitored the therapeutic interaction and calculated a measure of the instability of the assessments. We identified several sequences of stable and unstable episodes. Episodes of high instability were accompanied by discontinuous change. On the basis of these episodes of high instability, we divided the process into four phases. Mean values of variables differed across these phases. Destabilization proved to be highly synchronous in therapist and patient ratings. The approach allows to empirically assess critical phases and developments within the therapeutic relationship.