The frequency distribution of tornadoes in the United States is discussed briefly and an empirical method of forecasting these and allied severe local storms is described. It is shown that the formation of such storms can be forecast by evaluating certain empirical criteria related to the stability of the air column, the horizontal and vertical distributions of the moisture content and the wind. A typical tornado development is discussed from a synoptic point of view. An unusual situation and the verification of 35 experimental forecasts are briefly described.