454 TRANSMISSION 



presence of sunlight and the fixing of carbon ; and by such 

 other relations as are involved in direct causation. 



On the other hand, such a relation as deafness among teleg- 

 raphers or blindness among civil engineers or locomotive drivers 

 is unknown, because the conditions are such that the characters 

 in question are mutually exclusive. 



In general, however, correlation falls somewhere between i 

 and unity, and on one side or the other of the zero point ; that is, 

 a degree of relationship exists which is neither absolute, denot- 

 ing direct causation, nor negative, signifying mutual exclusion. 

 For example, a high degree of correlation exists between length 

 of cob and weight of ear. It does not amount to unity, however, 

 for the circumference also contributes to weight. 



Most results in living organisms are the effect of mixed 

 causes, and for this reason correlations are more complicated 

 than may at first appear. For example, many, if not most, good 

 cows have a capacious " barrel" and a roomy udder, and men 

 have been led to assume a perfect correlation between these 

 special characters and milk production; whereas the truth is 

 that the correlation, though high, is something less than unity, 

 because good cows are known with small barrels and with incon- 

 spicuous udders. Here is a real need for accurate methods of 

 determining what degree of correlation actually exists. The 

 average man asks whether or not two characters are correlated, 

 and expects a positive answer Yes or No ; whereas the question 

 should be, To what extent do the two characters appear together? 

 expecting for an answer a fraction lying somewhere between 

 zero and unity, say perhaps 40 to 60 per cent, as in correlation 

 of length to weight of ear. 



The student must distinguish clearly between correlation and 

 mere association. For example, we might ask the question whether 

 black pigs are more subject to cholera than are pigs of other 

 colors. The first step would be to establish a ratio between the 

 number of diseased pigs and pigs in general. This ratio would 

 now express the chances that a particular pig, irrespective of 

 color, will be afflicted with this disease, that is, by that opera- 

 tion of independent probability which we call chance. If now 

 we find upon inquiry that under the same conditions the ratio 



