The influence of quasi-realistic high-frequency wind variability for the probability distribution of surface waves and for the skill of the predictions with the wave model WAM is investigated. So far, the sensitivity of the surface wave field to high-frequency wind variability has been demonstrated only for idealized wave conditions or particular events using rather simplified high-frequency wind forcings. In this study the problem is addressed more generally by estimating the high-frequency wind variability (for periods shorter than 6 hours) and by assessing its relevance for a particular North Atlantic winter. Wind observations with 20 min time resolution are used to build a statistical wind generator that reconstructs (in a statistical sense) that fraction of the wind variability that is missing in the analyzed model wind fields. These quasi-realistic wind fluctuations, superimposed on analyzed 6 hourly wind fields, produced an increase of the air-sea momentum flux and resulted in a moderate but systematic increase of the average wave heights and in their short-term variability. While the results are qualitatively consistent with the findings of earlier studies under simplified conditions, they differ quantitatively. The change of the distribution of the wave data is found to be less pronounced for young sea states in the storm track region than for older sea states in the low latitudes. The response yields everywhere a significant amplification of the spectral variance for periods below 12 hours and an increase of the probability of extreme wave heights. Thus the increased air-sea momentum flux is seen to be effective for predictions of the probability distribution of wave data and may also influence predictions of the ocean circulation. ¿ 2000 American Geophysical Union