CONTINUOUS VARIATION 



added to the non-heritable component which it shares with its 

 predecessors. 



Now any gene segregating in Fg will, by Mendel's principles, be 

 lioniozygoiis in half the F^ individuals and heterozygous in the other 

 half. If, therefore, we raise Fg's by inbreeding the F., individuals, half 

 will show segregation for each gene and half will not. Taking all 

 genes into account we shall expect that the average variance of all 

 the Fg's will lie between the variance of Fg on tlie one hand, and 

 the variances of parents and F^ on the other. For the average 

 heritable component of the F3 variance will be smaller than that of 

 Fo, but not, of course, so small as those of parents and F^ where 

 variation was solely, or nearly solely, non-heritable. But not all the 

 Fg's will be alike in cither variances or means. The variances will 

 reflect the different numbers of genes heterozygous in the different 

 Fg mothers. Furthermore, for all genes in wliich the mother was 

 homozygous, each F3 individual must carry the same allelomorphs. 

 The mean expression of the character in each F3 will therefore be 

 correlated with the expression in its Fg mother, just as Galton found 

 the heights of fathers and daughters to be correlated in man. 



So, even though we cannot follow segregation by ratios of types 

 in the mendelian fashion, we can still hope to detect its occurrence 

 by the biometrical properties of frequency distributions. And we 

 can make so many predictions about these properties, that if all of 

 them are borne out in experiment we can entertain no doubt that 

 mendelian segregation is occurring. 



A characteristic experiment of this kind is illustrated in Fig. 17. 

 It shows the inheritance of corolla length in a cross between two 

 lines o( Nicotiana longijiora as recorded by East (1915). The corolla 

 lengths are expressed in millimetres and the data are grouped for 

 convenience of presentation into classes each covering a range of 

 3 mm. and centred on 34, 37, 40, etc., mm. The frequencies arc 

 shown as percentages of the individuals of the various families which 

 fell into the different classes. The variances, measuring the spreads of 

 the distributions, are much the same for the two parents and the Fj, 

 whose mean is just about midway between those of the parents. 

 The F2 has a much larger variance, while those of the four Fa's 

 shown differ amongst themselves and lie between the values for Fg 

 on tlie one hand and parents and F, on the other. The arrows 



70 



