l6 JOURNAL OF CONCHOLOGY, VOT-. l6, NO. I, JANUARY, I919. 



walls, each about 20 yai-ds long, H much shaded, the other two open. 

 All the loci would be fairly called sheltered, except B, which was 

 pretty fully exposed to the west, and C and D which are draughty 

 loci on higher ground towards the pass leading westwards from the 

 Glaslyn Valley. A, B, F, H, I, and J one would call loci of highly 

 similar characteristics ;• E and G similarly go together. 



The required data are set out in table I. The shells being 

 measured to o'l m.m., it is impossible to give the full details for 

 altitude, and these have accordingly been set out in groups covering 

 o'5 m.m. ; this is the less objectionable as in practice it is convenient 

 to make the necessary calculations from such grouped data rather 

 than from the individual measurements^ ; the diameter is much less 

 variable and can be given in full. The other data are the maximum 

 and minimum (to show the extreme range), the mean, the standard 

 deviation and the number of specimens measured. 



The longest shells are H, the shortest E. The difference between 

 E/H is I '302 m.m., between any other pair less until it dwindles to 

 o'ooy m.m. for A/I. It would almost stand to reason — or common 

 sense — that this latter difference can mean very little and that the 

 former may mean much. But whether it really means anything and 

 what we are to think of all the intermediate values cannot be deter- 

 mined by inspection and a hope of revelation ; the answers can be 

 obtained only by proceeding along some such method of evaluation 

 as has been outlined above. 



Applying this method to the figures in table I we get the results 

 shown in tables II to V. Table II gives the absolute difference in 

 millinietres between each possible pair, table III the standard error 

 of the differences in each case, and table IV the figures obtained 

 by dividing each difference by its standard error ; in each table the 

 right upper half refers to altitudes, the left lower half to diameters. 

 The results are summed up in table V, which shows the comparisons 

 which give significantly different results for altitude and diameter 

 respectively. 



There are 45 possible pairs for comparison from the ten loci. In 

 altitude 30 show differences which are real according to the rule, in 

 diameter 19 ; in 12 cases the samples differ from one another in 

 both diameter and altitude, while in 7 they differ in diameter but 

 not in altitude, and in 18 in altitude but not in diameter. In all, 

 therefore, there are 37 differences out of a possible 45. C, E, G and 

 H each differ from all the other samples in altitude or diameter or 

 both, and F alone shows fewer differences than not; A, B, D, I and 

 J differ from the majority of other samples, but each is not signifi- 



{ See the methods given in Yule, op, cit., pp. 112, 141. 



