SummaryThe limitations of current diagnostic categories are well recognised but their rationale, advantages and utility are often ignored. The scientific support for a ‘continuum of psychosis' is limited, and the examination of whether categories, a continuum or more than one continua, and alternatives such as subtypes or hybrid models, best account for the distributions of symptoms in populations has simply not been done. There is a lack of discussion, let alone consensus, about the critical aspects of psychosis to measure, the best ways to quantify those and how these would be applied in clinical practice. Systematic studies are needed to evaluate which of a range of plausible approaches to the classification of psychosis is most useful before change could be justified.