148 
On Homotjumsix in Eggs of the Common Tern 
Certain facts are at once obvious from this Table, others are obscured. In 
the first place length and breadth of the egg of the Common Tern have a rela- 
tively small relationship, while the relationship between the two girths is between 
TABLE II. 
Correlations of Characters iv t]ie same Egg. 
Cori'GlSition 
T? din n I'V Q 
Length and Breadth 
Longitudinal and Equatorial Oirths 
Length and Longitudinal Girth 
Breadth and Longitudinal Girth ... 
Lidex and Longitudinal Girth 
Inde.x and Length ... 
Index and Breadth ... 
L , B 
^ , Gi 
B , Gi 
BIL, Gi 
BIL, L 
BIL, B 
+ •2220+ -0374 
+ -5297 + ^0284 
+ •8804 ±-0088 
+ •5216+ •028(5 
- ^3832 + -0336 
-•7284+ •OISS 
+ •SOSS ± ^0294 
Mottling and Ground Cohiur 
Mottling and Index ... 
Mottling and Breadtli 
Ground Colour and Index 
Ground Colour and Breadth 
M, G 
M, BjL 
M, B 
C, BIL 
C , B 
+ -2260 (corrected d) 
— '1550 (corrected rj) 
— •1803 (corrected j;) 
•0000 (corrected 
— •1506 (corrected r)) 
More mottling, deeper groiuid colour 
Less mottling, higher index 
Less mottling, greater breadtli 
No relationship 
Fainter ground colour, greater breadth 
t 
tw^o and three times as great. This probably flows from the consideration that 
the correlation of (x, and Gj^ arises from B being a factor in both and only 
secondarily from the correlation between L and B. The correlation of the 
longitudinal girth with egg length is 60% higher than that of longitudinal 
girth with egg breadth; both these correlations are more substantial than that 
of the longitudinal girth, Gi, with the egg index, 5/X. The egg index correlated 
with length is large and negative, and with breadth considerable and positive, 
precisely the results we should anticipate would appear if the correlation were 
largely spurious*. 
In order to ascei'tain how far it was possible to predict the longitudinal girth 
from length and breadth, double (for L and B) and triple (for L, B and BjL) 
regression formulae were worked out. The following equations resulted : 
(i) Gi-Gi= 1-2701 {B-B) + 1-6415 (L - L), 
Gi = 1-2701 B + 1-G41.5 L + -8224, 
(ii) Gi-Gi = - 17-2930 (i? - 5) + 14-6374 (L-L) + -7636 (/ - 7), 
Gi=- 17-2930 5 + 14-6374 L + -7636 B/L - 52-7239. 
The first seventeen eggs were taken as a random set to test these results upon 
with the following values : 
• As a matter of fact the correlation of index and length for a constant breadth is - -997 and 
of index and breadth for a constant length is + -996 instead of unity. These values indicate how closely 
the linearity of regression holds in these quantitative measurements. 
or, 
and 
or. 
