No. 567] EFFECT OF DISTRIBUTION ON SPECIATION 131 



It should be remarked that a unit of area in this con- 

 nection should be considered a distributional unit, not a 

 geographical unit. In other words, while the addition of 

 one hundred square miles might or might not involve a 

 change in the life of a region, the addition of a new ' 'life 

 zone," " fauna," or association" (see p. 155) would 

 inevitably involve a biotic change, and therefore the addi- 

 tion of one or several of any of these distributional areas 

 should be considered as an addition of a unit, comparable 

 to another unit of similar kind. 



Two possible ways of testing this hypothesis present 

 themselves. We may compare the faunas of distribu- 

 tional areas of dissimilar size, or we may compare the 

 specific and generic differentiation found within families 

 occupying areas of different extent. The former method 

 we should expect to work out with a fair degree of 

 accuracy, but the latter involves so many modifying cir- 

 cumstances that even if sufficient data were at hand, it 

 would be difficult to prove anything by it. In the first 

 place there is the difficulty of comparing, in a distribu- 

 tional sense, the areas occupied by different families, 

 since, as pointed out above, the geographic areas do not 

 necessarily coincide at all with distributional areas; in 

 the second place, while it is justifiable to compare the 

 speciation of a family in one region with the speciation of 

 the same family in another region, it is of doubtful value 

 to compare the speciation of one family with that of 

 another in the same or different regions, unless the other 

 factors controlling their speciation be comparable or 

 nearly so. In view of this there are few families which 

 could be advantageously compared with each other as to 

 speciation in relation to extent of distribution, yet in the 

 families which do seem to lend themselves to such a com- 

 parison, the evidence all points towards the correctness 

 of the law here proposed. 



The bats seem as favorable for such an interfamily 

 comparison as any group of mammals that could be 

 selected, and the table (Table I) of their distribution by 



