The entire set of Seasat A satellite scatterometer (SASS) wind speed observations, Us, colocated with the buoy measurements of the wind speed Ub and wave height H1/3 is analyzed. The ''error'' Ub-Us is found to be influenced by the degree of wind-wave coupling. This coupling is quantified employing the ratio of the wave-to-wind energy densities: X~&rgr;wg⟨&zgr;2⟩/&rgr;aUb2. For the special case of a fetch-limited wave growth, X is shown to coincide with the wind fetch. It is found that when the coupling is weak, i.e., at large X, the SASS tends to overestimate the wind speed, and vice versa. The magnitude of the trend is evaluated roughly as 0.5 m/s per 100 km of X. The increased radar backscatter at large X is explained by invoking the concept (due to V. Zakharov and his collaborators) of a Kolmogorov equilibrium range appearing in wave spectra of sufficiently developed seas when the wind input is concentrated at high frequencies. In this case, the surface density of steep wavelet occurrence would be at its highest owing to a pronounced cascade pattern in the surface geometry. The fractal dimension DH of such an idealized surface is estimated to be about 2.333. Further, it is suggested that DH for a general case is a function of sea maturity. Finally, it is concluded that both the probability and the surface density of steep wavelet events are increasing functions of X. A major implication with respect to the electromagnetic scattering is that the so-called spike component of the backscattering coefficient, formed from the individual radar returns caused by the steep and/or breaking wavelets, is cotrolled primarily by the large-scale features of surface geometry, hence by such nonlocal factors of the wave development as the wind fetch. ¿ American Geophysical Union 1988