YUZO HOSHINü 
263 
expected on the more-thaii-three-factor hypothesis, as, on the four-factor hy- 
pothesis, there is expected only one early constant individual among 254. In 
Table 9, there are three early constant and one late constant families (exclud- 
ing the pseudo-early and-late) among loi families (less than 254/2), and in 
Table 10, 2 early constant and 3 late constant families among 54 families (far 
less than 254/4). And moreover, by examining the variation types shown in 
Tables 9, 10 and 1 1, one will readily be convinced of the fact that the varia- 
tion types are much too simple for those of F3 progenies whose parents had 
more than three alleromorphic pairs. 
If assume the presence of three factors, there are two ways of inter- 
pretation ; (1) supposing that all three factors have the same hereditary effect, 
and (2) supposing that each of the three factors has a different hereditary 
effect. On the former assumption, there must be 4 kinds of constant families 
as shown in the following zygotic formula: — 
l[i aabbcc H 
I AAbbcc [I AABBcc 
I aaBBcG HI i aaBBCC IV {i AABBCC 
I aabbCC (i AAbbCC 
If we assume gametic purity, even on the above hypothesis, we should expect 
only 4 distinct constants and can not interpret the occurrence of the pseudo- 
early and pseudo-late constants in the F3 raisings and of those constant fami- 
lies whose variation means are located in various places between the variation 
means of the parent varieties in raisings. If gametic contamination be 
granted, our two-factor hypothesis explains the facts obtained by our experi- 
ments far better than the three-factor hypothesis. 
If we attribute a different hereditary effect to each of the three-factors, 
there must be 8 different constant families in Fg raising. This conflicts 
with the fact that we could group the intermediate constant families into only 
two, the early intermediate and late intermediate, in Tables 9 and 10, especialy 
in the former. And the variation types of the variable families in F^ and F^ 
raisings are too simple to be interpreted on this hypothesis. If we deduce, 
