An investigation into the underlying mechanisms that cause bottom reverberation fluctuation statistics to deviate from the typically assumed Rayleigh distribution reveals that sediment ``patchiness'' is a major contributor. The investigation involved processing both shallow water data and simulated data. The shallow-water data, collected during a SACLANT Centre experiment, include low-frequency measurements made over chalky sediments with varying degrees of roughness using different pulse types (cw and FM) and pulse lengths (2.2 and 8 s). The processing produced both Rayleigh and non-Rayleigh statistics. The results indicate that sediment type, pulse type, and pulse length are not a factor per se in the deviations of the statistics from Rayleigh. However, sediment ``patchiness,'' defined as an ensonified area containing patches which contribute largely different backscattering strengths within the area, does cause the statistics to deviate from Rayleigh. The simulated data reinforce these conclusions and provide insight into the spacial and temporal extent of ``patchiness'' required to cause a deviation in the statistics.