THE MANNER OF OCCURRENCE OF MUTATION 455 
the most extreme possible type, namely, white—the original 
white, the ‘new white’, and the deficiency ‘ultra-white’—while 
four were very nearly white—namely, ivory, tinge, buff and 
écru. Of the remaining three, cherry and blood are at least as 
far removed in the color scale (as determined by the thickness of 
solutions required to imitate them) from red as they are from 
white. Only coral remains as a variant of moderate degree, but 
even it is scarcely darker than the lightest eye variations 
hitherto observed in other loci than W. If anything, then, 
there is a piling up of the curve of variation near its extreme— 
a phenomenon diametrically opposite to the most fundamental - 
characteristic of all ‘probability curves.’ 
The relative deficiency of mutations of lesser magnitude is 
not due to their being less easy to observe, for most of the eye- 
color variations which occur in other loci, and which deviate 
from normal much less than does the average mutant of W, are 
so distinctly different from the red that their classification and 
separation from it in crosses is quite certain. There may of 
course be some variations so slight that they will often escape 
recognition, but all the variations observed in W are quite out- 
side this range, and could not fail to be detected in practically 
every instance in which they occurred. If, then, the variations 
of W form a probability curve, there should, within this range 
which is certain of detection, be a greater number of variants 
found at smaller deviation values from red than at greater, 
instead of the reverse grouping which obtains. Although ten 
might ordinarily be considered a small number for a sample, 
this discrepancy between the expectation based on ‘chance 
fluctuation’ and the actual finding is so marked as to afford 
definite contradiction of that idea. 
We must conclude that the variations in W are not distributed 
in magnitude like the deviations found in random sampling 
(Muller, 718). On the basis of this finding, furthermore, the hypoth- 
esis becomes improbable that very small deviations in the factor 
W—lying on or behind the borderline of visibility—are any more 
frequent—if as frequent—as the occasional visible mutations which 
have been detected. Finally, it is seen that selection with regard 
