202 Experimental Zoology 
conditions, be awakened to activity, we seem to get further insight 
into the conditions. That alternate or contrasted characters 
may exist in a single germ is well recognized. A further exam- 
ple will show the probability of this explanation. If individ- 
uals of two races, having different ‘‘modes,”’ are crossed, the 
offspring will often show a tendency to give a curve with two 
summits — one for each parent mode. The characters of both 
parents are present in the offspring and sometimes one, some- 
times the other, dominates. These, however, are special cases 
and are by no means to be taken to illustrate the usual effects 
produced by crossing two races. 
Selection of Fluctuating Differences 
We come now to the question of inheritance in regard to fluc- 
tuating variation. Ordinarily, succeeding generations show 
the same variations. This applies when the pairing of the indi- 
vidual is left to chance alone, and when the external conditions 
are similar. But the outcome is different if individuals show- 
ing the same kind of fluctuation are picked out or selected and 
inbred. The most striking results are produced, of course, when 
individuals showing one extreme or the other are utilized. Sup- 
pose, for example, we select two individuals, standing near the 
same outer limit of the curve —two individuals, let us say, 
showing the fluctuation in question developed almost to the 
greatest degree ever found. If the offspring of this pair be 
measured, it will be found, in most cases, that the average has 
been raised in the direction selected. The curve has moved 
toward the variation selected in the parents. In other words, 
and this is the important point, the average of the individuals 
is now higher than before, i.e. there are now more individuals 
than in the preceding generation with the character in question 
better developed. By selecting again individuals from the same 
extreme, the average can again be transferred in the same direc- 
tion. As the process is continued, however, the average gain 
soon decreases, and finally stops without ever having transgressed 
the outer limit of fluctuation found in the first instance, provided 
