Abstract
Molecular beam experiments provide fascinating data on how atoms move in the course of chemical reactions. In order to theoretically reproduce these data at relatively low computational cost and to interpret them, nuclei are often treated as classical particles, even though we have known for about a century that, in the range of energies usually available to chemical systems, their motion is best described by the laws of quantum mechanics. Nevertheless, over the last decade this approximation has been shown to work unexpectedly well, provided that a few constraints are introduced into the calculations in order to take into account the quantisation of product internal motions. Why this is so and the nature of the previous constraints are the central issues of this review article.