AbstractA thermally generated reactive intermediate is generally presumed to consist of a population with a distribution of lifetimes characterized by a single, well‐defined parameter, τ. The decay of an instantaneously generated population, whose conversion to products occurred by parallel unimolecular steps, would be expected to be defined by a simple, single‐exponential expression: A(t)=A0 exp(−t/τ), where A(t) is the population at time t, and A0 is the population at t=0. In this model, the lifetime is also related to the first‐order rate constants, ki, for the unimolecular product‐forming steps; specifically τ = (Σiki)−1. The ratio of the products is equal to the ratio of the rate constants for their formation. Given an accurate potential energy surface (PES) for the reaction, one would expect to be able to compute the values of the ki under the specified reaction conditions, and hence both the lifetime of the intermediate and the ratio of products. This simple picture, which has formed the basis for most mechanistic investigation of reactive intermediates, and most kinetic modeling of reactions in which they are involved, has been challenged in recent years by molecular dynamics (MD) simulations, which have predicted that a variety of thermally generated intermediates should exhibit bimodal or even multimodal lifetime distributions. In several cases the shortest lifetime populations have been found to lead to products in a ratio quite different from that for the longer‐lived populations, and also quite unlike that expected from RRKM or transition state theory calculations based on the same PES used for the MD simulations. This commentary reviews the origins of this behavior and discusses the challenges of designing experiments to test the predictions. Copyright © 2003 John Wiley & Sons, Ltd.