The thermal activation of the capture coefficientcis very often treated in a semiclassical way, resulting from a simple energy barrier. Experimentally, this activation energyEbis said to be the slope of ln[c(T)] vs 1/TkT. Unfortunately, many experimental evidences deny this analysis and suggest that a single energy barrier is not enough to reproduce the physical behavior of a point defect. Here we propose a more complete approach to the problem of nonradiative carrier capture assisted by phonons, which is based on a less restrictive hypothesis and leads to a compact formulation of the capture coefficient. This formula reproduces the previous asymptotic forms at high temperature in the strong coupling limit, but remains valid over the whole temperature range and for any strength of the coupling between the lattice and the defect. This point is illustrated for typical cases and the accuracy of the new formula is shown in the case of the so‐called B level in GaAs, whose capture coefficient does exhibit a nonexponential behavior with temperature. Our wish is to put the theoretical point of view previously developed by Passler under a form allowing more extensive use in order to extract physical parameters from the experimentalc(T) curves with better accuracy.