Theoretical assumptions about how psychotropic drugs 'work' are rarely discussed explicitly. In a 'disease-centred model,' drugs are believed to work by acting on a disease process. In contrast, in a 'drug-centred model,' the characteristic physiological, behavioural and subjective effects of drugs are used to define drug action. The therapeutic value of a drug stems from the usefulness of these effects in clinical situations. The disease-centred model appears dominant but has weaknesses: (1) it cannot logically justify the use of drugs since major pathophysiological hypotheses were derived from selectively observed actions of drugs; (2) comparisons between drugs believed to have specific effects in certain conditions and drugs thought to have non-specific effects fail to support it; (3) outcome measures for various disorders include items responsive to non-specific drug effects; (4) studies with healthy volunteers describe characteristic drug-induced states independently of a psychiatric diagnosis; (5) animal tests show effects with agents not usually thought of as specific treatments for the conditions modelled by tests. This article offers suggestions to develop a drug-centred model and discusses its potential impact on clinical practice.