It is shown that the adaptive theory of the training of yeast cells and the Mendelian view of heredity are not incompatible with one another, and the chemical basis of these phenomena is discussed. The growth characteristics of three yeast strains in a galactose medium have been examined. One, aneurin-exacting, will not adapt to galactose under any circumstances; another, which is not exacting for any growth factors, is initially capable of poor growth only, but eventually trains to grow at an optimum rate; a third, which is biotin-exacting and pantothenate-stimulated, grows at the optimum rate immediately. A detailed examination of the second strain has shown that the adaptation is a response of the bulk of the population and not due to a selection of variant cells. Analysis of four-spored asci of the third strain has shown, however, that absolute ability to utilize galactose segregates as though controlled by a single gene. Nevertheless, those segregants which can metabolize galactose at all exhibit initially a varying degree of training towards this substrate, and, moreover, show the characteristic adaptive phenomena on subculture. The material basis of these effects is considered.