This paper investigates wrapped and unwrapped phase differences generated by a non-Gaussian scattering model: the two-dimensional random walk. Mean square values for these quantities are obtained for one and two scatterers, as well as the large scatterer limit when the field constitutes a circular complex Gaussian process. Numerical simulation is used to investigate the phase under more general fluctuation conditions, and reveals that the wrapped phase difference correlation converges rapidly to that result predicted for a Gaussian speckle field. Analytical results for the unwrapped phase indicate that this quantity transitions from a stationary process for one and two scatterers to a non-stationary process in the large scatterer limit. The nature of this transition is examined using numerical simulation for arbitrary scatterer number. Phase correlations are of consequence in various phase sensitive detection systems, and this paper examines both Gaussian and non-Gaussian fields.