Objectives. We used a new instrument, the Agnew Relationship Measure (ARM), to examine the alliance's dimensionality and to construct scales for use in future studies. Design. We studied the alliance as reported on the ARM by both clients and therapists in the Second Sheffield Psychotherapy Project, a randomized comparison of two contrasting time‐limited psychotherapies for depression. Methods. Clients (N = 95) and therapists (N = 5) completed parallel forms of the (ARM) after every session (N = 1120). Results. Five scales were constructed, based on results of simultaneous components analyses and considerations of conceptual coherence and comparability across client and therapist perspectives. Conclusions. Bond, partnership, and confidence overlapped statistically (consistent with previous studies), but there are conceptual and empirical reasons for retaining the distinctions. Openness represented a relatively independent dimension. The client initiative scale had low internal consistency but the items may have value for future investigations.