The random walk models such as the Stochastic-Separated-Flow (SSF) model are commonly used in the simulation of two-phase turbulent flow. However, there exists an essential assumption of correlativeness between the gas and drops through the equation of motion in the random walk models. This work makes probabilistic investigations on the turbulent dispersion effects using the SSF model for two test problems. One is a drop-loading, planar mixing-layer flow; the other is a hollow-cone spray. It is found that the correlative assumption between the gas and drops embedded in the SSF model is only valid for the small drops under some special conditions. Thus, the SSF model is generally incapable of predicting the turbulent dispersion effects accurately.