364 
Note on R. O. Punnett's Memoir 
a lo7iff series of mammals. I told Mr Puniiett that this vahie alone seemed sufficient to show 
that the material was inadequate. He then added 136 individuals in 15 families. They gave : 
Series II. 
13G Individuals in 15 Families 
136 Individuals in 15 Families 
Character 
Parental Correlations 
Fraternal Correlations 
Kaw Value 
Corrected Value (a) 
Raw Value 
Corrected (a) 
Corrected (b) 
Posterior spine 
Anterior spine 
Wliole vertebrae 
Half vertebrae 
Total segments 
•096 + -057 
•357 + -050 
•139 + -057 
■232 + -055 
•379 ± -050 
•142 
■396 
•177 
•252 
•332 
•473 + ^015 
•269 + -018 
•469+ 015 
•238 + -019 
•477 ±^015 
•479 
•294 
•475 
•246 
•457 
•484 
•318 
•482 
•253 
•437 
]\Iean 
•241 
•260 
•385 
•390 
•395 
The results were now much lower for the parental correlations, and slightly lower for the 
fraternal correlations. The diflferences between the two sets are in a number of cases significant 
compared with the probable eri-or of the differences, and the only safe conclusion to be drawn is 
that the two samples from two different years are not random samples of the same population. 
Mr Puunett tells me that he cannot fix on any diflerence in the method of collecting or observing 
the two series. Judging by p. 321 one appears to have been procured at a rather earlier date than 
the other. Clubbing the two sets together we had the values for fraternal and parental correla- 
tion given in Mr Punnett's Tables 8 and 9. These arise from 224 individuals in 25 families for 
the parental, and 229 individuals in 27 families for fraternal values. The general result is, as we 
might expect, midway between the two. 
Can we venture to draw any definite conclusions from these divergent values, and, if so, what? 
Now, I think we may safely say that Mr Punnett has demonstrated the existence of inherit- 
ance in these meristic characters ; but that the paucity of material does not allow us to definitely 
fix its intensity. 
The values found are very similar in intensity to those first found for stature in man from 
Mr Galton's material, mothers and sons ^302, mothers and daughters ^284*. These were based 
on 200 families. Again from the same data : Brother and brother = ^391, sister and sister ^444, 
sister and brother •375. On Mr Galton's hypothesis the ancestral law gives for parental corre- 
lation •SOO, and for fraternal ^400 1. Or, we might even say that Mr Punnett's fish results agree 
on the average fairly well with those of Mr Galton for stature, and confirm his view of the 
ancestral law|. 
But this is, I take it, to miss the real point at issue. Such low values as Mr Punnett finds 
have been found before in man, in lepidoptera, and in molluscs, but in all these cases in short and 
doubtful series. They tell neither for nor against the ancestral law, but they are opposed to any 
hypothesis as to sensibly equal jjarental correlation in different species and for a variety of 
characters. That is a very vital point which no one can at present consider finally settled. 
* Phil Trans. A, Vol. 187, p. 270 and p. 281. 
t R. S. Proc. Vol. 02, pp. 397 and 410. 
X Similarly we have for a short series of 100 families, the inheritance of cephalic index, mothers and 
daughters •SOO and brothers and sisters -340, for example. R. S. Proc. Vol. 62, p. 415. 
