Detailed studies have been carried out for fetch and duration limited wave growth and fully developed sea state, but much less is known about changing wind conditions. The subject of this study is the response of waves to a change in wind speed. The main point of interest is the rate of response at the different frequencies. Two third generation wave prediction models were used, WAVEWATCH (Tol man , 1991) and the model of Resio and Perrie (1991). In third generation models all relevant processes of generation and dissipation of energy are represented explicitly by the three source terms: input of energy from the wind, nonlinear interactions between spectral components and dissipation of wave energy by various mechanisms. The main difference between the two used models is the way they calculate the nonlinear interactions. The operational model WAVEWATCH uses an approximation; the academic model of Resio and Perrie gives a full solution of the nonlinear souree term. To reduce the computation time required to calculate all wave-wave interactions, this model is restricted to either duration or fetch limited wave growth in deep water. With the model of Resio and Perrie a series of calculations was performed for sudden increases and decreases in wind speed. A simple equation which describes the model behaviour was formulated. An important coefficient of this equation is the adjustment rate immediately after the change in wind speed. Another meaningful parameter of the model behaviour is the time needed to reach a new equilibrium. Both parameters show that the rate of response increases with increasing frequency. The adjustment to a new equilibrium is faster for larger increases in wind speed, but slower for larger decreases. Field measurements of wind speed, direction and wave height were carried out on Lake George near Canberra (Australia). Analysis of the wind data resulted in five cases with a clear change in wind speed and a sufficiently constant wind direction. These cases were analysed with Fast Fourier Transforms. The measurements were compared with the results of WAVEW ATCH and the model of Resio and Perrie. The measurements were do ne in a water depth of about 2 meters; transitional water depth for the considered situations. Because in the model of Resio and Perrie bottom influence cannot be taken into account, the comparison with the measurements is only meaningful in situations with relatively short waves, which are not influenced by the limited depth of the lake. In these situations the results are reasonable. The results of WAVEWATCH do not agree very well with the measurements, e.g. the significant wave heights are too small. The parameterizations of the source terms need to be adjusted to the circumstances at the lake. The bottom roughness is probably much smaller than the 'standard' roughness for oceans.