As if the above were not a sufficient challenge to our psychopharmacological foundations, surprisingly little information has been collected systematically from routine clinical practice and we are forced to extrapolate from relatively small scale clinical trials where subject numbers are low and the population under study are unrepresentative of depressed patients as a whole. Where is our evidence base for the use of antidepressant drugs in those with coexisting psychiatric and physical disorders, in those with multiple previous episodes, or in those who have failed to respond to other treatments? It is estimated, and widely accepted, that between 30 and 40% of depressive episodes (in the ‘real world’) respond poorly to antidepressant medication (Thase and Rush, 2000). Despite this, a recent systematic review of the evidence base for the efficacy of pharmacological interventions for depression following a single antidepressant drug treatment failure found little evidence upon which to guide pharmacological management and commented that all published trials were simply too small to detect an important clinical response (Stimpson et al., 2002). With respect to the common and profoundly disabling clinical problem of chronic depression, where a depressive episode has lasted longer than 2 years, we are even less wellinformed. In the largest and best quality clinical study conducted with patients with chronic depression, Hirschfeld et al. (2002) randomised 681 patients with chronic, nonpsychotic depression to the antidepressant medication nefazodone, a cognitive behaviour-analysis system of psychotherapy (CBASP), or a combination of the two, and reported remission rates for study completers after 12 weeks of 22%, 24 and 42%, respectively. Clearly, these are highly encouraging data suggesting beneficial acute effects with the combination of appropriate psychotherapy with antidepressant drug treatment. However, a recent audit of attendees with chronic depression at our clinic found that none of our patients would have met entry criteria for this study, generally because of having been treated by three or more previous antidepressant drugs, or because of having received a course of ECT. As a consequence, we cannot be confident that the findings from Hirschfield and colleagues’ study are likely to generalise to our patient population. Also, as previously highlighted, a recent re-analysis of these data suggests that there are specific sub-groups of patients with chronic depression (those with histories of early trauma) who may only respond to psychological treatment methods. Hence, in the murky ‘real world’ of clinical practice: 1. Antidepressant drugs are likely to be a lot less effective than most of us believe. Indeed, we must face up to the possibility that they may exert no specific efficacy in a majority of patients with depression. More research is needed to produce good data on effect size in different groups of depressed patients. 2. Some individuals will respond dramatically to one pharmacological agent and not to another, even when these drugs are from the same pharmacological class. Given the complex biological systems in which these drugs must operate, perhaps we should not be surprised. 3. The clinical features of depression are substantially responsive to apparently non-specific ‘environmental’ and ‘psychosocial’ interventions, and to placebo. Similarly, structured, targeted psychological therapies can exert powerful ‘antidepressant activity’ that, at least in some studies, appears equivalent in potency to that of antidepressant medication. 4. Antidepressant actions are likely to be pharmacologically diverse. Sometimes entirely different modalities of treatment, e.g. ECT, neurosurgery, psychotherapy, light therapy, will elicit clinical responses when aggressive drug treatments have failed. 5. Antidepressant drugs have important therapeutic actions that may have little, or nothing, to do with specific antidepressant activity, hence their apparent efficacy in the management of anxiety, chronic pain, obsessive– compulsive disorder, and post-traumatic stress disorder. The functional changes that are brought about by these drugs, within different conditions, appear to be different, suggesting that the same drug may have different neurobiological effects depending on the baseline ‘brain state’ of the patient (Saxena et al., 2002, 2003) Similarly, the neurobiological effects that result from antidepressant drugs may also have different psychological and cognitive effects depending on the baseline state. Such questions reveal potential common pathologies in a variety of psychiatric disorders, which could be open to animal modelling, and which may help us understand the disorders at a more basic scientific level.