A class of retrospective measurements analogous to the “delayed choice experiments” of Wheeler and Greenberger is considered. A new argument shows that the reduction of the wave packet must affect the past states of the system. As a side product, our argument implies that the axiom about the reduction of the wave packet in relativistic space-time cannot be consistently introduced.

The statistical interpretation of the quantum mechanical wave packet contains a gap. The author outlines the problem without offering a solution.