Abstract

It is argued that the conventional formulation of quantum mechanics is inadequate: the usual interpretation of the mathematical formalism in terms of the results of measurements cannot be applied to situations in which discontinuous transitions (''quantum jumps'') are observed as they happen, since nothing that can be called a measurement happens at the moment of observation. Attempts to force such observations into the standard mould lead to absurd results: ''a watched pot never boils''. Experiments show both that this result is correct when the experiment does indeed consist of a series of measurements, and that it is not when the experiment consists of a period of observation: quantum jumps do happen. The possibilities for improving the formulation by incorporating transitions in the basic postulates are reviewed, and a satisfactory postulate is obtained by modifying a suggestion of Bell's. This requires a distinction between the external description of the whole of a physical system and internal descriptions which are themselves physical events in the system. It is shown that this gives correct results for simple unstable systems and for the quantum-jump experiments.