2QQ PLANT-BREEDING 



or less frequent occurrence of the observed combinations. 

 On the other hand the conception does not exclude cases of 

 complete mutual dependency of characters; it is even prob- 

 able that it embraces many of them, but that our knowledge 

 is still too incomplete to allow us to draw a distinction be- 

 tween these and the ordinary cases. The other extreme in 

 the long list of possible combinations is evidently given by 

 casual and fortuitous coincidences, accidentally repeated 

 in some group of observations. A continued investigation 

 would show these to have no real value, and so they may 

 here be left out of consideration. 



After our hasty survey of some of the most interesting 

 facts, we may now proceed to inquire into their causes. 

 These may be brought under two heads. The causes may 

 be internal or external. In other words the correlation may 

 depend on some inner connection of the qualities, or simply 

 on corresponding changes induced by environmental in- 

 fluences. Evidently, the internal correlations are the most 

 interesting, and moreover those which may be the most 

 completely relied upon. Outer life-conditions, working in 

 the same direction on different characters, on the other hand, 

 are more easily understood and more directly accessible for 

 experimental study. On this account a rational treatment 

 will have to begin with the latter, and discuss the former 

 or internal causes of correlation only after the field has been 

 cleared as much as possible of its foreign elements. 



For this reason I have chosen for this chapter the study 

 of the influence of external life-conditions on the phenomena 

 of correlation. 



Correlated variability is quite an ordinary feature in all 

 plant life. It may be seen almost everywhere. As soon as 

 a plant deviates from its type, it will be disposed to do so in 

 more than one character. This rule holds good for rare and 

 casual abnormalities, as well as for the more normal, so- 



