Collision kernels describing elastic collision effects in saturation spectroscopy are completely calculated for the metastable 3 P 0 state of neon, starting from true experiment-deduced potentials. Collision processes with and without metastability exchange are respectively considered. A direct observation of these two kernels is provided by a two-laser time-resolved saturated absorption experiment. They are in good agreement with theory.PACS numbers: 34.40.+n, 32.70.Jz Line shapes in saturation spectroscopy are strongly affected by collisions. As in linear spectroscopy they are sensitive to quenching collisions and to dephasing collisions (broadening and shift), but in addition, as was pointed out early on, they are sensitive to velocity changing collisions (VCC). Many theoretical and experimental works have been devoted to this subject.The quantity of interest is the collision kernel W(v'-*v) which describes the probability per unit time that a collision will change the projection of the atomic velocity on the laser beam axis from v' to v. Many theoretical papers deal with the formal determination of the kernel from scattering amplitudes (sometimes from the interatomic potentials) and with the evaluation of the line shape from the kernel. 1 " 5 Nevertheless in all detailed calculations of line shapes phenomenological potentials (hard sphere, 3 R" 6 , 5 Lennard-Jones) or phenomenological kernels (Keilson and Storer 6 ) are used. In this paper, we present the first attempt to link all the steps from a realistic potential to a direct experimental observation of the kernel. The metastable 3 P 0 level of neon (ls 3 in Paschen notation) has been considered, since the Ne*-Ne potentials had been calculated and the differential cross sections measured.Experimentally the situation is comparable: Many experiments have been performed, but they have been interpreted with phenomenological kernels (strong or weak collisions 6 ). Furthermore in several cases the observed background (strong collisions) or the broadening of saturation peaks (weak collisions) is actually produced by several collisions for each atom and does not clearly represent the kernel.In our experiments we used a transient method previously described, 7 which has been improved to allow the recording of the saturation line shape after a very short time delay following the switching on of the saturating beam. Thus, the observed signal is the result of only one collision per atom. Another novel feature of this work lies in the calculation of a metastability exchange kernel and in its experimental observation in a mixture of two isotopes.Because of its short lifetime, the upper level of the transition (2p 2 for the 613.3-nm line or 2/> 5 for the 626.6-nm line) provides a negligible contribution to the saturation signal. On the other hand, when a collision changes the velocity of an atom that is in a coherent superposition of both levels, this coherence is destroyed, since very different interatomic potentials are involved 8 (except for very small scattering angl...