A new method, called a DVR-DVI method, has recently been developed for predicting flowering stage of soybean plants, which has proved to be more practical and accurate than the traditional thermal unit method in many cases. Under the DVR-DVI method, emergence and flowering stage are represented by numerical values of 0 and 1.0, respectively. Such values are termed DVI, through which a model of the growth of soybean plants is constructed. Increment of DVI on each day is defined as DVR, which is calculated by a model, taking into account weather factors of each day. Such a model is the basis for predicting flowering date and indicating varietal differences of the related characteristics. Models for several varieties of soybean, which reflected their respective DVR for daily mean air-temperature and daylength, were developed on the basis of results of a field experiment with some daylength treatments. These models clearly explain complex responses of soybean to weather factors and varietal differences in earliness. They also show that dependence of DVR on temperature changes in accordance with daylength.