An earlier paper' on this subject summarized results of studies on the 1954 crop in 10 Southern States. At that time three approaches to forecasting yield from plant observations were considered : (1) the multiple regression approach, (2) a "probability of survival" model, and and (3) an empirical approach. This paper describes a forecasting procedure which developed from these earlier studies as more data became available and as an appropriate way of making use of the fruiting habits of the cotton plant in a forecasting formula was better understood. The most noteworthy advance was the discovery of a simple device for estimating fruiting rate from a maturity classification of plants on August 1. This is of major importance for the August forecast because all of the fruit contributing to final yield is not yet formed by that date. After September 1, additional fruiting is no longer an important factor for the Southern region as a whole. This new approach is being used experimentally in 1956. Another paper in this issue by Jack Fleischer reports on additional work being conducted on the "probability of survival" model.