Experimental data were reviewed critically for four types of intermolecular energy transfer in the gas phase:(1) exchange of vibrational energy, (2) interchange of vibrational and electronic energy, (3) exchange of electronic energy, and (4) interchange of electronic and translational energy. The rate of exchange of vibrational energy between NO and a number of diatomic molecules has been measured by fluorescence and by flash photolysis. The probability of energy transfer decreases rapidly with increasing discrepancy between the vibrational frequencies of the two colliding molecules, and the results are in satisfactory quantitative agreement with predictions of Schwartz, Slawsky, and Herzfeld theory. The spin-orbit relaxation of Hg(63P), Hg(63P1) + M → Hg(63P0) + M*, was investigated by flash spectroscopy, and was shown to occur with M = N2, CO, H2O, or D2O. There is no systematic variation of quenching cross section with the minimum energy which cannot be converted to vibration in the quenching molecule. It was suggested that if a substantial transfer of electronic energy occurs, e.g., Na(32P) + M→Na(32S)+M *, the yield of vibrational energy in the quenching molecule is generally small. Quenching occurs because of a strong interaction between Na(32P) and M which permits near crossing of potential curves in the collision complex. It was pointed out that the apparent electronic relaxation time of a metal atom in a shock-heated gas will be approximately equal to the vibrational relaxation time if quenching produces a finite yield into any one of the vibrationally excited levels of the quenching molecule. Exchange of electronic energy between atoms was reviewed and it was shown that to a first approximation transfer occurs to minimize the change in internal energy. If several opportunities occur with ΔE < ∼1000 cm−1, the course of the energy transfer cannot be predicted either from the magnitude of ΔE or from the optical rules. The quenching of an excited atom by a polyatomic molecule depends on the reactivity of the polyatomic molecule, which again suggests that strong interaction facilitates transitions between potential surfaces in the collision complex. Hg(63P0) is curiously stable to collisional deactivation. The probability of electronic−translational energy transfer was shown to decrease rapidly with increasing magnitude of the energy to be transferred. It was suggested that several types of energy transfer conform approximately to the law log P = A ΔE + B, where P is the probability of transfer per collision and ΔE is the energy to be converted to translation. For fixed masses and temperature, A is the same constant for several different types of energy transfer, and B can be neglected except for vibrational exchange.