136 Francis B. Sumner 
that the weight has increased during this interval 48.4 per cent 
in the cold room (1. e., lighter) lot; 45.8 per cent in the warm room 
(heavier) lot. The gain in tail length for the cold room lot has 
been 25.7 per cent, as compared with an increase of only 14.2 
per cent on the part of the warm room individuals. In respect 
to each character, therefore, but more especially in respect to the 
tail, there has been a very obvious tendency toward a diminution 
of the differences between the two contrasted sets. 
Referring to variability, a comparison of the standard deviations 
for tail length shows that there has been a slight reduction in the 
absolute and a considerable reduction in the relative variability in 
both the warm room and cold room lots. The standard devia- 
tions for weight have undergone an increase in each lot, though 
the relative variability has remained practically unchanged. 
Frequency polygons (Figs. 11 and 12) have been plotted for 
tail length and foot length in this group of females. The polygons 
for the former overlap to a very slight extent; those for the latter 
are sufficiently distinct to admit of no doubt as to their significance. 
The males used for breeding were kept under the extreme tem- 
perature conditions until April 1, at which time they were 4to 5 
months old. They were then transferred to the “intermediate 
room” along with the females, and were killed upon reaching 
the age of 7 months. 
The measurements for this group are given in Table 1o. 
Comparing the present figures with those for the same mice 
at the age of 24 months, we find that the lighter warm room lot 
has gained to the extent of 32.7 per cent of its former weight, 
while the heavier cold room lot has gained only 25.1 per cent, 
again showing a tendency toward equalization. he same fact 
is evident in the growth of the tail. This has amounted to 17.9 per 
cent in the case of the cold room lot; 11.5 per cent in the case of 
the warm room lot. 
Levelling Down of Early Differences. 
Reference has more than once been made, in discussing partic- 
ular sets of measurements, to a tendency for these experimentally 
produced differences to diminish with growth. A table has been 
