378 JOHN C. PHILLIPS 
weights of each pure race as class means in a variation table, which 
shall contain a convenient number of intermediate classes, say 
seven. Tables may thus be constructed separately for each sex 
(tables 6 and 7). 
In both tables the variation of the pure races is seen to be about 
classes 2 and 10 as their respective modes. Indeed the tables 
were constructed with that end in view, the observed means of 
the weights being made the means for classes 2 and 10. Since 
further the number of intervening classes is the same in both 
tables, these tables may legitimately be combined, class by class, 
to obtain a table of weight distribution for both sexes, which 
however will be free from any serious error? due to the fact that 
the two sexes differ in size. Such a combination table is table 8. 
From an examination of these tables it is clearly seen that (1) 
the F, animals vary closely about the middle class, 6, exactly 
intermediate between the parent races. The extent of variation 
of the F, animals is small. (2) the F, animals vary about the 
same intermediate class, 6. The amplitude of variation of the F, 
animals is greater than that of the F; animals, but does not ex- 
tend beyond the nearer limit of the respective grand-parental 
races. In the 33 F, animals studied no variate occurs which is as 
small as the mean of the small race or as large as the mean of the 
large race. No case of complete segregation occurs. (3) The in- 
creased variability of F, as compared with F, may be regarded as 
due to partial segregation of genes having multiple representation 
in the gamete, or as due to modification of gametes in other ways 
as a result of their association in F, zygotes. The evidence at 
present available is insufficient to decide between these contrasted 
views. 
In conclusion, I wish to express special thanks to Professor 
Castle for valuable help in correlating the results of this exper- 
iment. 
2 This difficulty was met by Galton (’89, Natural inheritance) in a different 
way, which however would for our present purpose, it is thought, be less accurate 
as well as more laborious. Finding that the average male measurement was 
greater than the corresponding average female measurement he ascertained the 
ratio of the two to each other and then converted all female measurements into 
equivalent male measurements by multiplying by this ratio. 
