Measurements of energy transfer rates in monochromatically excited iodine are used to test various theories of vibrational relaxation. The efficiency of vibrational energy transfer is observed to be greatest when the mean duration of the collision is equal to the period of vibration. An explanation for this is given in terms of a time-dependent perturbation theory due to SCHWINGER. Most of the energy transfer models developed for ultrasonic data (viz. : complete LANDAU-TELLER, COTTRELLREAM, SCHWARTZ-SLAWSKY-HERZFELD, RAPP et al.) reproduce this behavior, but the simple LANDAU-TELLER and MILLIKAN-WHITE theories do not.The magnitude of the transition probabilities derived from these models is in only fair agreement with experiment. The problem lies in that calculated values of P01 approach unity as ΔE approaches кT, and P(n, n ± 1) is customarily taken as (n + 1 /2 ± 1 /2) P01 . In order to conserve probability, a strong-coupling model will be required, such as has been proposed by RAPP and SHARP or SHULER and ZWANZIG.Inclusion of an attractive term in the intermolecular potential contributes a factor of two to three in collision efficiency, but does not alter the dependence on collision times. Three methods of carrying out the BOLTZMANN averaging are compared, namely, steepest-descents approximation, numerical averaging of transition probabilities, and numerical averaging of the rale expressions. These three methods yield essentially consistent results. The observed ratio of Δn = 1 to Δn = 2 efficiencies is given to good accuracy by the square of the exact matrix elements of the intermolecular potential.Several calculations of pure rotational and rotation-vibration energy transfer are also reported.