All recently completed controlled two-year studies on intermittent, early neuroleptic intervention treatment have failed to compare favourably with studies on maintenance treatment concerning relapse prevention. The reason for this failure is still unclear. Therefore the implicit, but as yet unproven, hypothesis that a relapse can be predicted from prodromal symptoms was tested from the perspective of our German multicentre study. Results demonstrate that this is not the case. Possible reasons for and clinical implications of this negative finding are discussed.