In order to avoid these difficulties the method of selection according to progeny has been devised. The essence of this method is to select for stock, not the best individual plant, but the plant whose progeny yields the greatest weight of seed per unit area. This method was applied with great industry and some success in the Minnesota wheat breeding experiments of Willett Hays. Large numbers of promising plants were collected from a plot of the best variety in that district. The seed from each plant was rubbed out and sown separately. One hundred seeds from each plant were sown on small separate plots which were carefully marked out and labelled. Every possible precaution was taken to make all the little plots uniform in every way. By harvesting each plot separately, and weighing the grain it produced, it was possible to find out which of the original plants had given the largest yield. This process was repeated by sowing again on separate plots a hundred seeds from each individual plant from the best plot, and again weighing the produce of each plot. After several repetitions it was stated that new strains were obtained which yielded considerably greater crops than the variety from which they were originally selected. These results were published in 1895, but no definite statements have since appeared as to the success ultimately attained. This method of selection is undoubtedly more likely to give successful results than the method which depends on the selection of plants for their apparent good qualities; but it has several weak points. In the first place it is almost impossible to make the soil of a large number of plots so uniform that variation in yield due to varying soil conditions will not mask the variations due to the different cropping power of the seed of the separate plants. Many experimenters are still at work with a view to overcome this difficulty. Secondly, plant breeders are by no means agreed on the exact theoretical meaning of improvement by selection. The balance of evidence at the present time seems to tend towards the general adoption of what is known as the pure-line theory. According to this theory, which was first enunciated by Johannsen of Copenhagen as the outcome of a lengthy series of experiments with beans, the general population of plants, in say a field of wheat of one of the standard varieties giving an average yield of say 40 bushels per acre, consists of a very large number of races each varying in yielding capacity from say 30 to 50 bushels per acre. These races can be separated by collecting a very large number of separate plants, sowing say 100 seeds from each on a separate plot, and weighing the produce separately. The crop on each plot, being the produce of a separate plant, will be a distinct race, or pure line as it is called, and each pure line will possess a definite yielding power of its own. If this is so the difficulty of soil variation can be