Variation as an Organic Function 175 
cases from the simplest unicellular plants and animals to the most 
complex associations of different individuals. 
As a practical criterion however, this standard is of very little 
value, as the way in which the surplus energy is expended varies 
enormously in different types. It is necessary therefore, to have 
recourse to more proximate tests and to confine comparisons within 
certain restricted limits in which the type is relatively constant. 
Particular groups of structures and functions that are physiologically 
important and have a long phylogenetic history are most convenient 
for this purpose. 
Now this test can be applied in two ways—by an indirect com¬ 
parison of general types in any presumptive line of phylogenetic 
descent, or by attempting to ascertain directly the effect of observed 
mutations on somatic function. So far as the first method is con¬ 
cerned, there is the obvious difficulty that we have nowhere any 
exact knowledge of the phylogeny of existing species, and so must 
be dependent for the most part upon the comparison of different 
collateral types. Paleontological data are of value only in comparing 
anatomical characters, the functions of which are known in living 
forms. Within the limits we have stated, there is, however, marked 
evidence of a progressive increase in physiological efficiency in the 
course of phylogenetic development, as may be illustrated by the 
evolution of: (1) the photo-assimilatory apparatus of green plants; 
(2) the vascular system of pteridophytes; (3) the circulatory and 
respiratory system of vertebrates; (4) the nervous system of mam¬ 
mals ; (5) the social organisation of hymenopterous insects. 
The progressive increase in the efficiency of these physiologically 
important characters implies, even more than their observed stability, 
that germinal variations affecting them are mutually compensative 
with respect to somatic function. This may arise as the result of 
large mutations affecting several characters being internally com¬ 
pensated, or as the result of parallel smaller mutations being gradu¬ 
ally compensated with respect to average somatic conditions. 
When however we proceed to examine the various kinds of dis¬ 
continuous variations of which we have actual experience, there is 
very little evidence that they generally have any favourable effect 
upon the physiological efficiency of the organism, or that they are 
in any way mutually compensative. On the other hand, the great 
majority of observed mutations, especially those which occur under 
experimental conditions, are confined to single physiologically unim¬ 
portant characters . Thus among plants we not infrequently observe 
