GENETICS: W. E. CASTLE 
503 
ments made under varied conditions. On this general ground the 
full data would seem to be preferable to the selected data. But 
let us, for the sake of argument, accept the smaller amount of selected 
data. What are the observational facts? In a total of 1218 flies, 
the gene yellow was observed to separate from the other two genes in 15 
cases, and bifid was observed to separate from the other two genes in 43 
cases, but white was not observed to separate from the other two genes. 
The 15 cases, in which yellow separates, make up 1 .2% of the total, 
which is put down as the cross-over value between yellow and white, but 
in reality it is the cross-over value between yellow on one hand and 
the group white-bifid on the other hand. Likewise the 43 cases, which 
form 3.5% of the total, are put down as the cross-over value between 
white and bifid, but in reality they represent the cross-overs between 
bifid on one hand and the group yellow-white on the other. It is 
not certain that white with yellow attached would show the same 
percentage of cross-overs as without yellow attached. Indeed, in 
their doctrine of interference, the authors have long held that it would 
not. According to that doctrine, in case a break occurs between 
yellow and white, there is no chance whatever that a break will occur 
simultaneously in the near-by region between white and bifid, and 
vice versa, if a break occurs between bifid and white, there is no chance 
that a break will occur also between yellow and white. Accordingly 
in calculating the percentage of cross-overs for yellow- white, 15 should 
be deducted from the total number of cases, and in calculating the 
percentage of cross-overs for white-bifid, 43 should be deducted from 
the total, if the principle of interference is valid. Such correction 
would increase slightly the cross-over values for yellow-white and 
yellow-bifid, whose sum would now exceed the cross-over value given 
for yellow-bifid, 4.7, and would again put the three genes out of line. 
It appears, then, that the authors have failed in two different 
attempts to establish their linear theory in the case of the three genes 
yellow, white and. bifid. Morgan and Bridges in their equation, 
'about r -f 5 equals 'about 6,' seriously distort facts which should 
be very exactly stated if the linear hypothesis is to have a basis in 
fact as well as in fancy. The real relations, as given in their Table 
65, are 1.1 + 5.3 > 5.5. The new equation, which they now offer 
as a substitute for the old one, is 1.2 -f 3.5 = 4.7, which looks all 
right. The tw^o halves of the equation balance, but the balance is 
a forced one. To obtain it two very questionable things are done. 
First, nearly 99% of the authoritative observations are suppressed; 
