We present a modified Brownian motion model for random matrices where the eigenvalues (or levels) of a random matrix evolve in time in such a way that they never cross each others path. Also, owing to the exact integrability of the level dynamics, we incorporate long-time recurrences into the random walk problem underlying the Brownian motion. From this model, we derive the Coulomb interaction between the two eigenvalues. We further show that the Coulomb gas analogy fails if the confining potential, $V(E)$ is a transcendental function such that there exist orthogonal polynomials with weighting function, $exp [-beta E]$, where $beta $ is a symmetry parameter.