Cognitive therapy is a relatively new approach to the treatment of pathological gambling. Theoretically, there are strong grounds for believing that cognitive treatments will be effective in helping individuals cut back and stop excessive levels of gambling. However, there is evidence that cognitive therapy for pathological gambling is being confused with cognitive-behaviour therapy. In this paper, the distinction between treatments that are cognitive and those that are cognitive-behavioural is highlighted. Such a distinction has strong implications for the manualisation of therapy. Additionally, a range of problems that confront the evaluation of all therapies for pathological gambling is considered. Spontaneous recovery without therapeutic intervention has been documented in both field studies of both problem and non-problem players and controlled trials of cognitive therapy compared to a waiting list control group. The implications of the phenomenon of spontaneous recovery for the evaluation of cognitive therapy are described. Other problems common to all evaluations of psychotherapies are considered in relation to gambling and recommendations made for outcome study designs.