The feasibility of a forecasting scheme for predicting the onset of a major El Ni-o in the ocean is demonstrated using the linear numerical model of Busalacchi and O'Brien and the interannual components of the shipboard observed wind for 1961-82. The model upper layer thickness anomaly in the eastern Pacific, which was used as the predictand, was estimated after three months of steady wind integration. A lag time of three months is used to permit the propagation of a large El Niflo-type Kelvin wave across the Pacific Ocean. If the necessary wind changes required to generate a large El Niflo type Kelvin wave have already taken place in the western and/or central Pacific, El Niflo could be predicted one to three months in advancefrom knowledge of the wind field alone. Starting from November 1963, three-month steady-wind integrations were performed for the wind condition of each of the seven months (November to May), for each of the 15 years extending from 1964 to 1978. This period includes four El Ni-o years. The probability distribution function of the three-month running mean of the upper layer thickness anomaly in the eastern Pacific was estimated separately for the El Nifio and the non-El Nub groups using "bootstrap" estimates. The separation of the two probability distribution functions allows for the establishment of a criterion for forecasting El Nifio. An independent wind data set for the period 1979-82, which includes the onset of the 1982/83 El Nub, was used to test the feasibility of the forecasting scheme. If the null hypothesis is that a sample year is a non-El Niflo year and based on a forecast criterion of false positive error (false alarm rate) ~ 0.01, which corresponds to false negative error ~ 0.52-0.86 (which corresponds to a probability of detecting the occurrence of an El Niflo ~ 0.14-0.48), the 1982/83 El Niflo would be forecast to be underway following the analysis of the April 1982 surface wind field. Since the objective is to establish a forecasting scheme for predicting the onset of a major El Niflo, the forecast scheme is chosen to be one of low risk and low power in prediction performance.