INTRODUCTION 9 



with a specific name. The very idea of variation assumes 

 deviation from a norm which is invariably the character of a 

 group defined (whether as species, subspecies, or race) by taxo- 

 nomic procedure. To anticipate the discussion on the species 

 (Chapter III) we must point out that the latter is not a group 

 with standardised properties by which it can be invariably 

 recognised as such. It is an abstraction from a number of 

 individuals varying in such a way that any group or groups 

 defined must do some violence to the natural divergences that 

 certainly have always occurred in time and very frequently 

 occur in place. There are further difficulties to note which 

 arise from the actual imperfections of taxonomy. The vast 

 literature of taxonomy and the categorical nature of its 

 definitions obscure the incompleteness of our knowledge 

 in this branch of zoology. In certain limited groups in which 

 abundant series have been collected and studied critically the 

 status of the species at least rests on a solid foundation. In 

 many groups, however, particular species are known only from 

 a few individuals, sometimes of one sex only. Sometimes our 

 knowledge of the range of variation of a species depends on 

 whether two forms found in different areas are really identical 

 and no adequate comparison of them has ever been made. 

 Often purely nomenclatorial difficulties intervene, e.g. where 

 one species is known under more than one name in different 

 countries. All these difficulties are intensified when we are 

 dealing with the finer taxonomic units, such as very closely 

 allied species or geographical races. Many generalisations 

 about the variation of particular species are still rendered 

 dubious in this way, probably many more than is usually 

 supposed. The imperfections of taxonomy in this respect 

 are doubtless temporary, but they are at the present time a 

 great practical difficulty in the investigation of variation in 

 nature and not uncommonly they produce an element of 

 doubt in generalisations as to distribution and similar matters. 

 A species, like a molecule, is a statistical summary, and a 

 comparison of its properties with those of related forms can 

 most efficiently be made with the aid of statistical methods 

 involving tests of significance. When simple measurements, 

 such as those of size, are being made or when the material 

 studied consists of numerically small samples, these tests are 

 often indispensable, but in a broad survey like the present 



