The existiJ?g theory of qaantitative variation, in the form generally applied to animal breeding, is essentially descriptive in the sense that it describes the variation in a population in terms of certain statistical parameters. 1Prom these parameters, it is theoretically possible to predict how the mean of the ]?op~tlation will change as a resuR of continual selection. Bltt as we know that selection will change geese frequencies and that these control the values of the descriptive parameters,

more »

... this prediction of change in mean is of limited value because we cannot predict the chang e in gene frequency. The problem of how long the response will continue is therefore only to be solved by experiment. In an earlier paper (O]ay~on, Norris & A. l%obertson, 1956), we have described the effects of selection for abdominal bristles in a poptd~tioa of D~'oso2ohila ~neh~nopaste< We were there concerned with the early generations. The response of the population to different methods of selection (individual, half-sit and full-sit) was in reasonable agreement with predictions based on parameters derived from the base poptdation. One of the experimetats, that invobing selection of the extreme twenty individuals out of 1.00 of each sex in five replicates in each direction, was carried on for twelaty generations and in some of the lines for up to thirty-five, and it is these long-term effects of selection, pre-ctiet~bIe only in a very general sense, that we wish to discuss here. It was remarked in the paper on short-term response that the repeatability of the replicates in respon.se was i?ot high in the sense that the different replicates soon established a definite order between themselves, the differences in the high Ih~es being greater than wend be expected on any simple genetic sampling explazlation. This individualRy of the lines in respect to mea~ in the later generations spread to other aspects such as variation (Table 3, Fig. 4), ratio of scores in males and females (Fig. 3), and lethal genes present (Table 1). Axmther interesting feature was the lack of predictability of the behaviour of a given line in %tare generations. At generation 20, it w~s decided to discontinue two of the replicates in each set. This decision was taken on the basis of the mean res]?or~se in the line. In each set, the 'best' two and the 'worst', its terms of response, of the lines were kept. At the 8:3rd generation, when seleetiol~, filaally ceased, it was surprising that in both cUreetiolas, the line which had. only been retained because of its early slow response was now leading the field. The approach to final stability, which might on general gronnds Be expected to be a gradual deceleration of progress as the limit was reached, was in ma~y * ~[embcr of the Scientific Staff~ Agrimdtur~l Research Council, Gre~t I~ritah~.