W. F. R. Weldon 
305 
consistently greater than those of the corresponding arrays from adult shells. In 
all cases the values obtained cross and re-cross each other in such a way that we 
cannot infer greater variability in either series than in the other, and wc have 
therefore no proof that any selective elimination of young occurs. 
TABLE IV. 
Standard Deviations of Arrays of Peripheral Radii, and of Perpendiculars on 
the Columella, corresponding to fixed type of one other Dimension. The 
Determinations made luithin the limits of Columellar Radius-length indi- 
cating half a Revolution of the Shell-spiral. 
Limits of 
(7, .si I 
- r CP 
ffp\ll-f-pp 
- r'^pp 
Columellar 
Eadius-lengtb 
Young 
Adult 
Young 
Adult 
Young 
Adult 
Young 
Adult 
2-00— 2-49 
0-0525 
0-0456 
0-1125 
0-1039 
0-0347 
0-0338 
0-0328 
0-0342 
2 -.50— 3-09 
0-0467 
0-0509 
0-1110 
0-1165 
0-0372 
0-0381 
0-0358 
0-0367 
3-10— 3-79 
0-0692 
0-0614 
0-1453 
0-1474 
0-0436 
0-0468 
0-0426 
0-0493 
3-80— 4-54 
0-0592 
0-0621 
0-1675 
0-1535 
0-0567 
0-0574 
0-0554 
0-0557 
4-55— 5-49 
0-0876 
0-0712 
0-1953 
0-2151 
0-0624 
0-0658 
0-0622 
0-0648 
5-50— 6-54 
0-0997 
0-0933 
0-2312 
0-2257 
0-0787 
00773 
0-0751 
0-0757 
6-55— 7-64 
0-0918 
0-1174 
0-2589 
0-2334 
0-0748 
0-0916 
0-0728 
0-0750 
To test the possibility of selective elimination still further, the standard devia- 
tion of an array was determined in every group of measures, for fixed type of two 
dimensions ; the S.D. of peripheral radius-lengths, for fixed type of columellar 
lengths and perpendiculars, being equal to 
^ V 1 - r-cp 
a similar expression giving the standard deviation of perpendiculars, for fixed types 
of columellar and peripheral radii. The results, given in Table V., show no indica- 
tion that the variability of adult arrays is less than that of the corresponding arrays 
of young individuals, so that here again the attempt to demonstrate a selective 
elimination during growth is unsuccessful. 
The method employed in determining the variability of the arrays depends of 
course for its validity on the linear character of the regression in each case. It 
does not seem worth while to publish the whole series of 21 correlation tables, from 
which the results were obtained, but the diagram Fig. 2, chosen at random from 
the series, gives a fair idea of the regression, and in this diagram, at least, there is 
no doubt of its linear character. 
Biometrika iii 39 
