The variability of response to selection. I. Interline and intraline variability in a population of Drosophila affinis selected for body weight.

M O S T expectations relating to breeding systems have been derived under the assumption of an infinite or very large population and are based almost entirely upon deterministic theories in which random variation is discounted ( KEMPTHORNE 1957). However, the advance of a finite genetic population through time is subject to variation through random sampling as a consequence of Mendelian segregation and selection. Therefore, quantitative estimates of the magnitude of the random variation inherent in finite breeding systems are necessary in order to determine the reliability of the existing expectations. Results from selection experiments involving both large animals and laboratory organisms have demonstrated that selection is a powerful tool for changing the genetic composition of a population. The expected rate of change per generation may be estimated from population genetic theory if the selection differential and the heritability of the characteristic under consideration are known. Actual responses, however, may be quite variable when only a small effective number of parents is used to perpetuate selected lines if random sampling is an important source of variation in a selection program. The variability of response to selection has apparently not been thoroughly studied and the various experiments designed to test this variability are not in complete agreement (CLAYTON, MORRIS and ROBERTSON 1957; MARIEN 1958; RASMUSON 1955; ROBERTSON 1955). Quantitative estimates of the variability between lines selected alike are needed to determine if the Variability in response to selection is an important source of variation in a selection program. The primary objective of the present study is to obtain the interline and intraline variance components from replicated lines of Drosophila ufinis subjected to two intensities of selection for body weight in each of two directions. With such information, it shall be possible to construct intervals to attach to various means in order to indicate the magnitude of the variability of response to selection.