The K-distribution models the statistics of the backscatter from a coherent imaging process. It has successfully modelled radar backscatter from land and the sea surface, but it does not appear to have been applied to sonar images of a range of seabed sediment types. This paper shows that the K-distribution provided a good fit to sidescan sonar images of regions of large particles such as stones and gravel, but did not fit fine sediments such as mud. The generalised K-distribution provided a good fit to all the sediment types considered and should eventually produce excellent region classification and target detection algorithms. A number of physical mechanisms that may be responsible for the Ricean speckle component of the generalised K-distribution have been put forward.