While psychotherapists discuss amongst themselves and in supervision the struggles they experience face-to-face with their patients, empirical study of therapists' experience of psychotherapy has been scant and there has been little investigation of therapists' difficulties. We require tools for such investigation and in this initial report we describe the development of a taxonomy of situations that psychotherapists experience as difficult and indicate lines of research to which the taxonomy might be applied. We also suggest that our procedure might serve as a model of collective self-study for other groups of practitioner-researchers. Using accounts of our own difficulties as therapists, we constructed a consensual set of experimental categories that could be applied to describe them. These were designated as Therapist (T)-Incompetent, T-Damaging, T-Puzzled, T-Threatened, T-Out of Rapport, T's Personal Issues, Painful Reality/T's Ethical Dilemma (a combined category), T-Stuck, and T-Thwarted. We then employed a fresh set of 30 accounts, which were also drawn from our own experience but had not contributed to the development of the taxonomy, to study the taxonomy's reliability. Each of us rated the presence and predominance of the categories for each of the 30 accounts. Both presence and predominance of all nine categories in the taxonomy could be reliably identified, with about three raters required to ensure good reliability for all categories; alphas (for our group of seven raters) were in the range 0.88-0.95 for presence-absence and 0.83-0.94 for predominance. We conclude that the taxonomy is a serviceable tool for further investigation. We also show that we are discriminable from one another as therapists in terms of the kinds of difficulty we reported, differences amongst us conforming to what we know of one another's therapeutic practice. We note implications for the diagnostic use of the taxonomy in training and supervision.