Sonar and radar measurements both indicate that there are fundamental problems with the present wind–wave models of the sea surface at higher wave numbers: Backscattering strengths at high frequencies and small grazing angles are much greater than predicted 0(10–20 dB). In addition, the Doppler spectra show no evidence of the dispersion expected in gravity‐capillary régimes. The usual quasilinear wave theories do not account for such effects. Accordingly, a new qualitative four‐stage model for wave growth is proposed: (1) initiation of high‐frequency surface ripples (catspaws) by the wind; (2) nonlinear formation of steep wave fronts; (3) generation of low‐frequency waves by surface instability and wave–wind interaction; and (4) continued generation of unstable surface‐disturbances which chaotically degenerate into ensembles of solitons, undergoing continuous creation and absorption, with the incoming energy balanced by dissipation. A preliminary theory of the resulting soliton‐surface layer is discussed by D. Middleton [ComputationalAcoustics(North Holland, New York, 1988)]. Here, a broad range of experimental evidence is presented, in support of the proposed surface mechanism developed by Middleton. Thus, with assumptions not contrary to any of the known properties of the sea surface, quantitative scattering predictions of the model are seen to agree well with experiment, in particular with respect to backscattering strength, Doppler shift, and (point intensity) spectral levels, produced in known ‘‘bubble‐free’’ régimes. These results provide extensive inferential support for the sea‐surface and scattering models proposed here. However, further research is needed for explicit verification by direct observation and testing against more refined measurements of wave statistics and hydrodynamic theory.