For compressible mixing layers, the most widely used turbulence models, without further modification, cannot predict key compressibility effects. The present study examines both the capability of compressibility correction methods implemented in two-equation models and the improvements proposed by RNG methods. The turbulence models examined are the Chien low-Reynolds number k-e model and the Yakhot RNG k-e model. A Reynolds averaged, full Navier-Stokes flow solver called NPARC is used to examine the effectiveness of the corrections in a commercially available flow code. Although the investigation concludes that the RNG model produces results which are in conformity with experimental investigations, both models still fail to show the desired agreement with experiment that the compressibility corrections are claimed to produce.