An experimental research on ship capsize was conducted in the experiment tank with wind tunnel at the Institute of Industrial Science, University of Tokyo, using a simple cylindrical model. Critical (minimum) wind velocity for capsizing was determined for a given period and height of regular waves. Two wave periods were chosen so as to be in near synchronism with the natural rolling and heaving periods of the model ship. The experiments revealed that : 1) for a given wave height, the critical wind velocity was somewhat lower in case of the capsize in short waves than in long waves, but the difference was rather small, 2) the ship was capsized by steady wind force (without waves) when her angle of heel had surpassed over about two-thirds of the angle corresponding to the maximum stability moment, 3) the model could, not be capsized (at least in this experiment) when the mean angle of heel was kept nearly null as the result of balancing of two heeling moments due to wind and waves. The authors can not explain these results theoretically at present but a tentative hypothesis to predict the critical wind velocity or wave height is presented. Some examples of theoretical calculation of nonsymmetric, nonlinear rolling of ships are added for a future development of this research.