K. Pearson 
363 
Note on Mr Punnett's Section on the Inheritance of Aleristic Characters. 
By Karl Pearson. 
I should like to add a few words (as he himself has suggested) as to the results of this portion 
of Mr Punnett's investigation. I feel the more anxious to do so because he has broken entirely 
new ground, which we can only hope will attract the attention it deserves, and so bring new 
workers to this field. If so, it is desirable to state the conditions under which it is possible 
to obtain final results. The father in the case chosen by Mr Punnett is quite unknown, hence 
it is absolutely needful, in order to i-each accurate values of the correlation, that we should have 
enough mothers of each class to insure that the average father of that class is practically the 
average father of the population. When Mr Punnett first sent me material for reduction 
it consisted of 88 individuals forming 10 families for maternal correlation, and 93 individuals in 
12 ftimilies for fraternal correlation. Allowing for four to five classes in the tables, it was 
obvious that the extreme arrays of offspring would be due to one, or at most two mothers. It was 
not likely, therefore, that the mean father of these groups would be the mean father of the 
general population. However, we ventured to reduce the data with the following results : — 
Correkitions. 
Series I. 
88 Individuals in 10 Families 
93 Individuals in 12 Families 
Character 
Parental Correlations 
Fraternal Correlations 
Raw Value 
Corrected Value (a) 
Raw Value 
Corrected (a) 
Corrected (b) 
Posterior spine 
Anterior spine 
Whole vertebrae 
Half vertebrae 
Total segments 
■440+ -058 
■404+ -060 
•369 + -062 
•225 + ■068 
•534 ± -051 
•534 
•444 
•467 
•281 
•706 
•150+^023 
•488 + -018 
•297 + -022 
•152±^023 
•294 ±^022 
•246 
•509 
•.364 
•177 
•504 
•342 
•530 
•431 
•203 
•714 
Mean 
•396 
•486 
•276 
•360 
•444 
(a) Corrected for selection of mothers out of adult $ population. 
(b) Corrected for selection of fathers also, on assumption that they were equally selected 
with mothers, although unknown. 
Corrections made in the manner indicated P/iil. Trans. Vol. 200, A, pp. 39—45. 
Now it will be seen at once that the average corrected parental correlation, i.e. ^49, is in essential 
agreement — not with the ancestral law which fixes no value for the correlation — but with the 
values we have found for man, horse, and dog. On the other hand the fraternal correlation is 
very low, ^44 or even 36, a value lower than we have yet found for any physical character in 
46—2 
