On Mendel's Lazus. 
197 
direction as their parent. Quite generally, the statistician speaks 
of a character as inherited whenever the number or “ constant ” B 
is greater than zero ; if it does not differ sensibly from zero the 
character is held to be noil-heritable, quite apart from the question 
whether the mean is more or less constant from one generation to 
the next, a consideration which does not affect the conception of 
individual heredity. 
It is important that the biologist should realise this distinction 
between individual heredity and race-heredity, or as I should prefer 
to term it, constancy of type ; for, although the two phenomena 
must be in some way related through the processes of reproduction 
and growth, approximate constancy of type is not only logically, but, 
I believe, actually quite compatible with very slight individual 
inheritance, the type being maintained the same by the constancy 
of external conditions and the action of selection. Moreover, and 
this is the most important point, the maxim “ like begets like ” 
does not hold in the same rigid sense for the individual and the 
race. One generation of a race is (approximately speaking) the 
same as the preceding ; apart from such changes as are hardly 
revealed except by measurement, the mean of the offspring is that 
of the parents. It would seem natural perhaps to assume that the 
same law holds for individual types within the race—why should 
an isolated group of individuals behave differently from the race as a 
whole? Yet the assumption would be false; the offspring of any 
abnormal individual, any individual differing from the mean of the 
race, are always, on the average, more mediocre than himself. In 
the terms of our “estimating equation” ( 1 ), this means that the 
. 
constant B is always less than unity—always a fraction, If the 
offspring simply centred round the parental type we should have 
Y = X 
always. For stature in man the value of B is about 03 to 05; for 
a certain character in Daphnia Warren found B = 0-6, for other 
characters in an Aphis, 0*54, 05, and 0-36, the reproduction being 
parthenogenetic in both these cases ; for vegetative reproduction in 
Lemna minor I find, so far as the results are reduced, B= 0 25 to 
0’60, roughly speaking. This phenomenon of the relapse of the 
offspring from the parental type towards mediocrity is termed 
regression. Regression and not constancy of type is for the statis¬ 
tician. the fundamental phenomenon of heredity and the prime 
fact to be explained by any physical theory. The absolute lack 
of any mention of the subject in most biological theories makes 
them seem, to him, in so far, curiously unreal. 
