Miscellanea 
expressly states that he is dealing with a stable population. Notwithstanding Galton's own state- 
ment as to his series holding for a stable population, Professor Castle does not hesitate to apply 
it to a cross which produces a totally different population I My types may alter and do alter in 
any way from one generation to a second, (ii) Galton supposed that we may apply his scries 
not only to the deviation of an individual from type, but to the " whole heritage," he divides the 
oflFspring up into groups of individuals following special ancestors. This is an extension I liave 
expressly and repeatedly disagreed with. I look upon such a distribution as not falling under 
the Law of Ancestral Heredity at all, but as part of the Theory of Reversion or of Alternative 
Inheritance. My agreement with Galton consists in tlie general concepti<nis (a) that the 
deviation of an individual from the type of his own generation depends ultimately on the 
deviations of his particular ancestors from their types, and (^>) that the proportions of such 
deviation from type contributed by each generation of ancestry diminish in a geometrical series. 
This series is not Galton's. It may be deduced from the ancestral correlations, and in the case 
of man, horse and dog the series are within the limits of errors of observation identical. It is 
these ancestral correlations (which have no relation whatever to type, and are only connected 
with the regression coefficients which give the proportions of the deviation by complex determi- 
nantal relations) which Professor Castle cites, and states do n(jt agree with the numbers of 
coloured mice that von Guaita found descended from albino-grey crosses in successive 
generations ! 
Now Professor Castle was perfectly free to ignore my work, or to confess frankly that he 
could not understand it, but he commits a grave breach of scientific decorum when in a paper, 
not taking hasty journalistic form but published by the American Academy, he states that " the 
test of von Guaita's mice is conclusively in favour of Mendel's Law and against the law of 
ancestral heredity," and yet shows that he has either not read, or not been capable of properly 
citing, the two papers in which the meaning of this law is discussed. Either Professor Castle is 
so ignorant that he does not know that a coefficient of correlation cannot be a group frequency ; 
or, he has directly misquoted my memoirs because any form of argument suffices for the audience 
he wishes to appeal to. It is a fundamental canon of scientific discussion that, if you wish to 
contradict a man's results, you should know what those results are and cite them correctly. It is 
a breach of scientific decorum to assert that a man's theory is opposed to certain facts, when you 
have demonstrably not the faintest notion of what that man's theory is, or how his results ai'e 
reached. 
I stated in my paj^er of 1903 on the Law of Ancestral Heredity that "as far as the 
available data at present go, inheritance coefficients for ascending ancestry are within the limits 
of observational error represented by a geometrical series and by the same series." Professoi- 
Castle remarks : 
" It should be observed that the ' available data ' upon which principally Pearson bases his 
conclusions consist of two cases of pigment inheritance, one in man, the other in the horse. 
A third well-known series of this sort has not been utili/.ed by Pearson, though our informati(jii 
about it is much more complete and precise than that about either of the other two. I refer to 
our statistics about colour inheritance in mice recorded by von Guaita*." 
Now there are two points to be considered here. Professor Castle states (i) that the 
information is far more complete and precise, and (ii) that it is available for discussion by aid 
of the Law of Ancestral Heredity. Any one who has compared the total number of original 
parents or of offspring in any generation in von Guaita's statistics will realise that they are 
absolutely incomparable with the numbers we possess for man, horse or hound. They are also 
statistically insufficient to give correlation coefficients significant having regard to their probable 
errors. Further, the data given, as in almost all Mendelian categories, are so wanting in 
* Castle: loc. cit., p. 224. 
