W. Rowan, K. M. Parker and J. Bell 
147 
rather unexpected result may, perhaps, be found in the suggestion that the total 
amount of pigment is the same in both eggs, but the mottling and ground colour 
will appear denser on the smaller surface of the smaller egg. The point deserves 
consideration on the basis of larger numbers and possibly better defined measures 
of pigmentation. 
(3) Means and Variability. 
Table I gives the means, standard deviations and coefficients of variation of 
the several characters studied. It will be seen that the tern's egg has for 
quantitative characters relatively small variation. The values of the coefficients 
TABLE I. 
Means and Variabilities {Absolute Measurements in Gentimetres). 
Character 
Mean 
Standard 
Deviation 
CoefHoient 
of 
Variation 
Length L 
Breadth B 
Girth Gi 
Girth G,, 
Index BIL 
Index of Ovality, 0* ... 
4-14+ -007 
2-98+ -004 
11-39+ -015 
9-59+ -014 
72-04+ -130 
56-35 + -171 
-180+ -005 
-099+ -010 
-37fi+ -010 
•347 + -010 
3- 449+ -096 
4- 334+ 121 
4-34+ -12 
3-33 +-09 
3-30+ -09 
3-62+ -10 
[4-79+ -13] 
[7-09+ -22] 
of variation are less than many of those which we find for the human skull 
(3 to 8), but greater than those we know for the wing of the wasp. It is very 
doubtful whether the coefficients of variation of the indices should be included 
in such considerations, for the object of the use of these coefficients is to get 
rid of absolute lengths, and this is already done in the case of indices ■[-. It is 
noteworthy that the length of the egg is only slightly more variable than the 
breadth and the breadth-girth is actually more variable than the length-girth. 
(4) Correlations. 
If we turn to the correlation of characters in the same egg, we note that 
while the ordinary product-moment correlation r has been calculated for all 
measurable pairs of characters, this is not possible for the ground colour or the 
mottling. Where mottling has been used with a quantitative character there ?; 
has been calculated and both corrections used. Where mottling has been con- 
sidered in conjunction with ground colour, there we have adopted mean square 
contnigency correcting for both number of cells and for class-index correlations. 
* 0={BIL)l\-2-{lilL)]. 
t For example, if we take 1/0 for our index of ovality its mean =176-32, the standard deviation 
= 11-24 and the coetiicient of variation =6-38. Is O or 1/0 the more variable? It does not seem that 
the coefficient of variation can help us in such a problem. 
19—2 
