CORRELATED VARIATIONS. 73 



correlation is obviously less close than between arm and 

 arm, or leg and leg. Still again, personal experience 

 teaches us that there is correlation between even the 

 length of the face and that of the limbs. Tall men as 

 a rule have longer faces than short men; or, a more 

 striking instance, greyhounds have long heads and long 

 legs on the one hand, as compared with bull-dogs with 

 short heads and short legs on the other. Between the 

 length of face and length of limb, however, it is clear 

 that there is a less degree of correlation than between 

 length of arm and of leg, and between certain other 

 organs of the body the connection must be less intimate 

 still. It follows, therefore, that between the various 

 parts there may exist all degrees of correlation, stretch- 

 ing from an almost perfect degree of resemblance down 

 to an absolute lack of it. We must also recognise the 

 existence of negative correlation, in which the variation 

 of one part in one direction is accompanied by a greater 

 or less degree of variation of another part in the oppo- 

 site direction. Here again we may experience all de- 

 grees of negative correlation, just as of positive. In- 

 stances of negative correlation are much less frequent 

 than those of positive, and the only one known to me in 

 the case of man is that recently discovered by Professor 

 Pearson.* It was found that between stature and 

 head index the correlation is distinctly negative, or 

 that brachycephalic or relatively broad-headed persons 

 are slightly shorter than dolichocephalic or narrow- 

 headed. 



From what has been said it is clear that a bald state- 

 ment that in such and such a case one part or organ is 

 *Proc. Roy. Soc., Ixvi. p. 23, 1900. 



