106 VARIATION, DISTRIBUTION, AND EVOLUTION OF THE GENUS PARTULA. 
generally speaking, while the Astaman series from even higher levels shows a 
greater average length in comparison with the colonies from the neighboring areas 
of Torre and Garapan. Again we may compare an association from the low plain 
where the soil is alluvial with another from a higher station with exposed limestone 
rock, without discovering any consistent relation between an absolute or pro- 
portionate measure of the shells and such external conditions; a case in point is the 
comparison of the Astaman and the Garapan series. It is not necessary to mul- 
tiply illustrations, for the recorded figures clearly demonstrate the conclusion in 
question. 
It is instructive to compare the whole series of shells from Saipan with those 
from Guam in order to obtain an exact quantitative statement of their differences 
(table 53). While the Saipan collections were taken from 8 localities only, they 
are at least representative, even though a more extensive series would undoubtedly 
yield more or less divergent results. Taking the figures on their merits, the most 
striking differences are shown in the proportionate measures; the Saipan shells 
are collectively narrower, with relatively broader apertures, and with a noticeably 
higher proportionate value of the length of the aperture to that of the whole shell. 
TABLE 54.—Partula gibba, Saipan. Statistics of tooth development. 
Series. 
Puntan Flores 
Sadog Tase 
Puntan Muchut 
Fanaganam 
2 316 24 
[8722' ps Cty Ith. S pact.] (0.9 p. ct.] 

The statistics relating to the degree to which the tooth is developed are inter- 
esting (table 54). The feature in question is produced very weakly if at all; 87 
per cent in Saipan and about 84 per cent in Guam are devoid of any trace, approxi- 
mately 12 per cent in Saipan and 13 per cent in Guam display a small nodule, 
while only about 1 per cent in Saipan as compared with nearly 3 per cent in Guam 
possess a very small tooth. It is the Chalankiya collection which brings up the 
last figure for the Saipan series of gibba. 
Finally we come to the ultimate statistical analysis of the Saipan material, 
which is concerned with the quantitative characters of the several color-classes, 
colony by colony (tables 55 and 56). The fundamental question at issue is whether 
a given class, distinguished by its coloration, is also statistically differentiated 
from its fellows everywhere; the answer is negative. By way of illustration we 
may follow the bicolor and castanea classes through the several associations, 
omitting the small Torre series, contrasting them on the basis of the character of 
