Increasing sophistication in the neurosciences has re-stimulated expectations that the underlying biological mechanisms involved in the psychoses might finally be clarified. These expectations however coexist with a sense of mounting frustration over the continuing failure of such progress to occur. One important factor in this failure is an enduring reliance upon obsolete nosological concepts and tools. The persistence of concepts such as schizophrenia and manic-depressive illness, which are substantially based upon unsustainable assumptions concerning the relationship between pathophysiology, symptoms and outcome, interferes with the appropriate deployment of new technologies and the interpretation of the resulting data. As many have suspected for some time, we are in a state of chronic paradigm failure, yet the lack of a superior alternative makes it difficult to dispense with traditional concepts. Following a review of the principal nosological paradigms, it is argued that the psychopathology of psychosis is more fluid than acknowledged by the dominant neo-Kraepelinian paradigm in particular. An alternative, which may be termed the loose linkage model, is proposed, which though it falls short of full paradigm status, does represent more accurately the true state of affairs in psychotic disorder, with fewer historical preconceptions. Its importance lies particularly in its significance for research strategy, and its capacity to minimise early iatrogenic sequelae of the diagnostic process and permit a more honest approach to clinical care.