It has been shown recently that molecular gas samples excited with coherent light can display a variety of transient phenomena, similar to those found in nuclear magnetic resonance. This article elucidates how these coherence effects can be used to isolate or unfold molecular collision mechanisms that normally remain hidden within the optical line shape. Elastic collisions, for example, are easily detected here in two-pulse photon echo experiments for a C13H3F vibration-rotation transition. The echo-decay function which provides a signature for the velocity-changing collision diffusion mechanism, is not just a simple exponential in time but exhibits an exp(-Kt3) contribution for short times and an exp(-Γt) decay for long times. This behavior, which is unknown heretofore, contrasts with spin echoes in molecular liquids where Brownian motion leads only to the cubic decay law. An exp(-Kt3) behavior can be understood in terms of a solution of the Fokker-Planck equation which describes the effect of Brownian motion on echo decay. Such a treatment is valid for small phase excursions; in the case of a gas, this implies the Doppler phase factor kΔuτ1, where k→ is the propagation vector of light, Δu is a characteristic velocity jump for a binary collision, and τ is the echo-pulse delay time. When kΔuτ1 as in the long-time regime, the Fokker-Planck solution fails. We, therefore, present a new solution to the Boltzmann transport equation using a weak collision model and find agreement with the entire echo time dependence observed. The echo measurements indicate very small changes in longitudinal velocity per C13H3F-C13H3F collision, i.e., Δu=85 cm/sec, thereby justifying the weak-collision model. The total elastic collision cross section is 430 2. It follows that elastic collisions lead to velocity thermalization in a time of ∼5 sec when the C13H3F pressure is 1 mTorr. A comparison is also made of the C13H3F dephasing time τ2 in a coherent Raman beat decay, which is independent of velocity-changing collisions, with the longitudinal decay time T1. Here, T1 represents the molecule-optical interaction time, due largely to jumps in molecular rotation (J) and orientation (M) state, and is obtained from a delayed nutation measurement. The fact that the pressure dependent part of τ2=T1 shows that T1 is also independent of velocity diffusion. Futhermore, when a hole is burned in the Doppler distribution, population recovery must be due to inelastic rather than elastic collisions. Optical Carr-Purcell echoes, multiple pulse echoes, provide a direct measure of the C13H3F transverse dephasing time T2 without the effect of elastic collisions while being sensitive to "phase interrupting collisions." Again, we find that T2=T1 so that phase interruptions are negligible. Had such a process dominated the two-pulse echo, an exp(-tT2) damping would have been noticed with no exp(-Kt3) contribution. Thus, the present study covers several new aspects of molecular collisions. It represents the first detailed examination of velocity-changing collisions by coherence methods and without the complication of Doppler broadening. © 1975 The American Physical Society.