In the Berlin psychotherapy study data were collected not only from the pat
ients but also from the therapists. Thus, the influence of personality vari
ables of the therapists on the multidimensionally registered therapy succes
s could be studied. To this end a nonlinear k-sets canonical analysis was a
pplied resulting in a sample-specific optimal scaling. The relationship pat
tern of the two sets of variables found was satisfactory as regards the int
ernal criteria of the programme as well as its clinical meaningfulness. The
results can be used in similar studies for specifying therapist-related hy
potheses.