CORRELATION: RATIONAL AND EMPIRICAL 37 



Cuvier, 1 "where our theoretical knowledge of the relations 

 of forms would not suffice, if it were not filled out by 

 observation," that is to say, there are many cases of correla- 

 tion not yet explicable in terms of function. From a hoof 

 you can deduce the main characters of herbivores (with a 

 certain amount of assistance from your empirical knowledge 

 of herbivores), but could you from a cloven hoof deduce that 

 the animal is a ruminant, unless you had observed the con- 

 stancy of relation, not directly explicable in terms of function, 

 between cloven hoofs and chewing the cud? Or could you 

 deduce from the existence of frontal horns that the animal 

 ruminates ? " Nevertheless, since these relations are con- 

 stant, they must necessarily have a sufficient cause ; but 

 as we are ignorant of this cause, observation must supple- 

 ment theory ; observation establishes empirical laws which 

 become almost as certain as the rational laws, when they 

 are based upon a sufficient number of observations. . . . 

 But that there exist all the same hidden reasons for all these 

 relations is partly revealed by observation itself, independently 

 of general philosophy." That is to say, even correlations 

 for which no explanation in terms of function can be sup- 

 plied are probably in reality functional correlations. This 

 may, in some cases, be inferred from the graded corre- 

 spondence of two sets of organs. For example, ungulates 

 which do not ruminate, and have not a cloven hoof, have a 

 more perfect dentition and more bones in the foot than the 

 true cloven-hoofed ruminants. There is a correlation between 

 the state of development of the teeth and of the foot. This 

 correlation is a graded one, for camels, which have a more 

 perfect dentition than other ruminants, have also a bone more 

 in their tarsus. It seems probable, therefore, that there is 

 some reason, that is, some explanation in terms of function, 

 for this case of correlation. 



Nevertheless, the fact remains that many correlations are 

 not explicable in terms of function, and the substitution of 

 correlation as an empirical principle for correlation as a 

 rational principle marks for Cuvier a step away from his 

 functional comparative anatomy towards a pure morphology. 

 It is significant that in later times the term correlation 

 1 OssemensfossileS) i., p. 60. - Loc. tit., i., p. 63. 



