A rough surface is assumed to be composed of a slowly varying large‐scale roughness and a rapidly varying small‐scale roughness, and it is assumed that these are correlated. The exponential forms of both the autocorrelations of these surfaces and their crosscorrelation were used. This paper shows that the return from such a composite surface for a plane‐wave incidence can have either one of the following terms contribute a major portion of the total return power: namely, the large‐scale roughness term, small‐scale roughness term, and the crosscorrelation term. It is therefore concluded that contribution due to each term must be evaluated before neglecting its effect on the total return power. These results are equally applicable to acoustic and radar cases.