Clinicians faced with the problem of finding the most effective treatment of individual patients often express disappointment and frustration when turning to the research literature for guidance. Treatments and measures reported are often standardised, may be carried out for a fixed period irrespective of response, and use atypical - even ‘analogue’ populations, while variables of most clinical interest - such as individual differences - are symbolically relegated to the error term of the ANOVA. One reaction to this is to dismiss the experimental method as irrelevant to clinical problems, usually in favour of intuition in one guise or another, as a more satisfying and creative activity. It is unfortunately true that rigorous experiment is no guarantee of rapid success in generating valid theory or successful application in the clinic - however it is the only available route to cumulative progress. To utilise it most effectively requires both awareness of relevant clinical questions by the research worker, and willingness to utilise experimental or ‘quasi-experimental’ (Campbell & Stanley, 1966) methods on the part of the clinician, rather than to retreat to the non-experimental and untestable position of the pre-scientific psychotherapists.