Laboratory experiments were conducted to measure the joint distribution of surface elevation and slopes for nonlinear random waves. The slopes of random waves were measured by a small and precise accelerometer as well as the surface elevation. The correlation coefficients between the surface elevation and slopes were very small for nonlinear waves with the JONSWAP spectrum and Bretschneider-Mitsuyasu sprectrum. The skewness and kurtosis took different values from a Gaussian distribution. A method for estimating the joint distribution of surface elevation and slopes were developed by using the principle of maximum entropy. The estimated distributions were compared with the measured distributons, together with the Gram-Charlier series and two variate Gaussian distributions. The present method found to produce a similar distribution of the measured one.
