The most popular model of microwave backscatter from rough water surfaces at mid-incidence angles (20¿<&thgr;i<70¿) is composite surface theory. This theory holds that the backscattered return is directly proportional to the spectral density of centimetric, Bragg-resonant water waves which are tilted and advected by longer waves. A stringent test of this theory is to measure, independently and from the same surface area, the normalized microwave cross section (&sgr;0) and the Bragg wave spectral density, and compare them using the theory. In this paper, we use a calibrated optical slope imaging system in a wind-wave tank to measure the two-dimensional wavenumber spectrum of short waves. From these spectra, we calculate both the pure Bragg scattering &sgr;0 which neglects longwave effects and the more complex composite surface &sgr;0. The results are compared with &sgr;0 obtained from backscatter measurements at X band (10 GHz) and Ka band (35 GHz) made between 28¿ and 68¿ incidence angle. We find that composite surface theory generally shows better agreement with experiment at both frequencies than pure Bragg scattering theory. The agreement seems best for friction velocities above 40 cm s-1. For all friction velocities up to 70 cm s-1, however, composite surface theory somewhat underpredicts the actual &sgr;0 in a majority of the cases. This is especially true for horizontal polarization at large incidence angles. We conclude that while composite surface theory accounts for much of the backscatter at both frequencies in the incidence angle range we examined, the discrepancy between the predicted and measured cross sections is sufficiently large that contributions from other scattering processes cannot be ruled out. ¿ American Geophysical Union 1995