Surface and profile measurements and their resulting statistics, based on samples of up to half a million heights, are compared and their interrelationship examined for several common engineering surfaces. The measurements are employed to check the applicability of the spectral moment approach to random surface specification. This technique relates many important geometrical properties of a surface to those of its constituent profiles. A relationship is found to exist between the sampling interval and the spatial size of features accommodated by this form of approach. This explains, for example, why 4 and 8 nearest neighbor summit-density analyses based on the same square grid sampling interval reveal very different results. Having established this basic relationship, good agreement is found between theory and measurement over a large range of sampling intervals. In particular, summit densities and distributions are shown to agree well with theory even for non-Gaussian height distributions. It is shown how the isotropic analysis can be extended to cover directionally anisotropic structures such as ground surfaces by defining equivalent movements based on two profiles at right angles. Here again measurements are in good agreement with theory.