THE DISTRIBUTION OF VARIANTS IN NATURE 99 



(/) Aubertin, Ellis and Robson (193 1 ) studied colonies 

 of Cochlicella acuta in W. Sussex in respect of three 

 main types of shell-colour. 



(i) 21 comparisons were made (I.e. p. 1042), and of 

 these 2 only showed equal distribution of the 

 types. In the rest no regularity of incidence 

 was found, but either one class or two 

 tended to preponderate at the expense of the 

 third. In each colony, however, all three 

 types are usually well represented, and in 63 

 cases there were only 7 instances of a colony 

 having less than 20 °/ of any one type. 



(ii) The various colonies differ significantly in 36 % 

 of the possible comparisons. The authors 

 say that on the whole (p. 1047) very little 

 relation exists between the distance separating 

 the colonies and the differences in shell-pattern. 

 But this is not quite true, as nearly all imme- 

 diately adjacent colonies tend to show very 

 little difference one from another. Neverthe- 

 less it is true that some adjacent colonies may 

 differ significantly and distant ones may be alike. 



From these summaries we may form the following conclu- 

 sions. 



(1) Populations of land snails tend to occur in colonies 

 having a different facies, the differences having little correlation 

 with differences of environment (Alkins, Crampton, Aubertin, 

 Ellis and Robson) except perhaps in size (Boycott). 



(2) Continuous populations (/) may be divisible into sub- 

 ordinate areas with a statistically different composition. 



(3) In two cases ( (b) and (/) ) these differences are main- 

 tained with a tolerable degree of uniformity over a limited 

 number of years (up to ten) . 



(4) While each colony tends to show a fairly wide range 

 of variation, certain classes of variants tend to preponderate 

 and often whole classes may be absent. One gets the impression 

 that colonies exhibit the results of obligatory selective mating. 



(5) That certain classes tend to occur in a high percentage 

 might suggest that selection may be at work ; but we think 

 that this is unlikely, as (e.g. in Rensch's observations) we find 



