CORRELATION 169 



the Salmon and Lebistes) external factors (? salinity and 

 temperature) can alter the average numbers of vertebrae and 

 fin rays. 



It is probable that the groups of characters employed in 

 diagnosing species are not usually held together by a corre- 

 lation of this sort. How far the coincident effects of several 

 separate factors or the multiple effects of single factors of this 

 order may have been influential in evolution must be left for 

 a later discussion (p. 172). Theoretically at least groups of 

 correlated specific characters might arise as the direct effect 

 of environmental causes or from simultaneous selective 

 processes. The value of this suggestion depends on the 

 evolutionary importance we attach to these processes. There 

 is, however, some evidence of a convincing nature that charac- 

 ters of the same kind as distinguish taxonomic species are 

 altered in association as the result either of single environ- 

 mental factors or of several such factors acting concurrently. 

 Thus it is known that in the Baltic Macoma baltica and Mya 

 arenaria are both smaller (Brandt, 1897) and have thinner 

 shells than usual (Mobius, 1873). Bateson (1889) found that 

 the proportions and shape of Cardium edule are modified in the 

 brackish water of the Sea of Aral. Sumner (191 5) experi- 

 mentally induced lengthening of tail and foot in white mice by 

 high temperature, and such differences are known to differen- 

 tiate the wild races of rodents. Perhaps we should draw 

 attention to Sumner's point (1932, p. 53) that, though in some 

 of his experimental cases we might expect ' parallel modifica- 

 tion by the environment, the latter cannot account for 

 correlations which increase in segregating generations of 

 hybrids.' 



There is a theoretical possibility that all the characters of 

 a species may be produced by several coincident selective 

 processes or by a single selective process affecting several 

 characters. The wing-pattern of a mimetic butterfly would be 

 an example of the latter. The pattern is composed of several 

 elements, all of which are associated in the mimetic effect and, 

 on the selection hypothesis, must have been produced co- 

 incidentally by a single selective process. As an example of 

 the modification of several quite distinct structures in rela- 

 tion to a special mode of life we may cite Hora's (1930, 

 passim) demonstration that in torrent-dwelling species several 



