A critique is offered of the paper on non-interpretive mechanisms in psychoanalytic therapy by the Process of Change Study Group (Stern et al., 1998a). The attention paid to procedural learning and to the need to attend to implicit relational knowledge is welcomed but a number of the group's assumptions are questioned. In particular, it is argued that the differences between mother-infant and therapist-patient relationships are inadequately considered, that the polarisation of non-symbolic and verbal memory is unhelpful and that the separation of the therapy relationship into three strands is problematic. The assumption that pursuing the goal of intersubjective understanding is in itself therapeutic is questioned. It is suggested that the recognition of the importance of implicit relational knowledge calls for a wider reappraisal of important aspects of theory and practice and that direct means of making such knowledge explicit should be considered.