The conceptual issues are briefly noted with respect to the distinctions between classification and diagnosis; the question of whether mental disorders can be considered to be ‘diseases’; and whether descriptive psychiatry is outmoded. The criteria for diagnosis are reviewed, with the conclusion that, at present, there are far too many diagnoses, and a ridiculously high rate of supposed comorbidity. It is concluded that a separate grouping of disorders with an onset specific to childhood should be deleted, the various specific disorders being placed in appropriate places, and the addition for all diagnoses of the ways in which manifestations vary by age. A new group should be formed of disorders that are known to occur but for which further testing for validity is needed. The overall number of diagnoses should be drastically reduced. Categorical and dimensional approaches to diagnosis should be combined. The requirement of impairment should be removed from all diagnoses. Research and clinical classifications should be kept separate. Finally, there is a need to develop a primary care classification for causes of referral to both medical and non-medical primary care.