1. Using the method given in the first paper in this series, estimates of p and a have been obtained for farm crops of winter beans grown in 1950 and 1951, p being the frequency of a dominant allele, in this case for black hilum, (1 — α) the proportion of successful fertilizations in which the pollen comes from the same plant as the ovule and α the proportion in which the source of pollen may be any plant in the crop. In the 1950 crop the estimates were in the 1951 crop they were The difference in the estimates of p is not surprising in view of the fact that the two crops were grown from different stocks of beans. The difference in the estimates of α is possibly due to the 1951 crop being much the denser.

2. In both crops, as in the case previously reported, selection of high-yielding parents produced a frequency distribution of genotypes showing no influence of inbreeding. The data from the 1951 crop failed to give a satisfactory fit with expectation. Both these facts are unexplained by the simple model of a Mendelian population varying at only one locus and with all genotypes of equal fitness. They can both be explained by amending the model so that the population is considered to vary at several loci and that heterozygotes have a slight advantage.

3. Theoretically, a bulk made up of the produce of parent plants selected for high yield should contain substantially more heterozygotes than a bulk of produce of unselected parents. This was confirmed by hilum colour counts. Estimates of yield show that the greater heterozygosity is not reflected in any increase in yield. A possible explanation is offered in terms of information theory.