BOYCOTT: LOCAL VARIATION OF CT.AtTSILIA BIDENTATA. I i;; 



therefore, two are significantly different,' the rest negative. 

 Similarly ft from locus H was collected in six batches — 



Number. Mean length. Standard deviation. 



1 ... 54 ••• ii'5o4 ••• 0-591 



2 ... 59 ... ii'566 ... o-8i2 



3 ••• 42 ... ii'3S2 ... 0-502 



4 ••• 59 ••• 11483 ••• 0-631 



5 ••• 39 •■• ii"35i ••• 0-613 



6 ... 53 ... ii'58i ... 0-603 



The evident difference is 5/6 = 0-230 m.m., which is only 1-7 

 times its standard error ; in none of the fifteen possible comparisons 

 is there any significant difference. 



It is evident, therefore, that samples of snails collected from the 

 same locus will nearly always work out as " not different " by this 

 method of examination. I have dwelt at some length on the evidence 

 for this, as it is evidently a fundamental point on which the validity 

 of this method of enquiry largely depends. That samples collected 

 in different loci frequently give significant differences, the sequel 

 shows fully enough. 



With these preliminary considerations, we may proceed to the 

 examination of our immediate problem. Measured collections are 

 available from ten different loci in the neighbourhood of Portmadoc ; 

 do the shells from these different places differ in size ? 



The topographical relations of the loci are shown on the sketch 

 map, which indicates also the 500 feet contour. Their individual 

 characteristics cannot be satisfactorily described in detail, but the 

 leading features of similarity and difference may be shortly summar- 

 ised. All were stone walls of the local slaty rock, and all were at a 

 low altitude, E to J practically at sea-level, A and B about 100 feet, 

 C and D about 200 feet. A : about 90 yards long, both sides^ of a 

 narrow rather shaded lane with metalled surface. B : roadside wall, 

 open, about 50 yards. C : the abutment of an old tramroad bridge, 

 about ten yards, open. D : similar, rather more shaded. E : a ruined 

 house, deeply shaded. F : roadside wall, moderately shaded, 40 yards. 

 G : an old shed in a wood, deeply shaded. H, I, and J : roadside 



1 The eight samples were collected successively along about 80 yards of wall. There is 

 every possibility that the area I have regarded as locus A is not homogeneous, and indeed if one 

 accepts the statistical test in its entirety the results prove this rather than an occasional failure of 

 the test. Theoretically, however, the test is relative rather than absolute, and, pending further 

 more systematic observations made with the point specially in view, I should regard the meaning 

 of these exceptions as doubtful. Testing by another method by which each sample is compared 

 to the seven others taken together, 2 alone is significantly diflferent. 



2 Samples taken separately on the two sides gave the answer " not different," but it was 

 certainly an error of judgment to take both sides as forming one locus; a metalled road cannot 

 be very pei-meable to iJzV/fw^ii/a. 



