VIII HEREDITY 129 



another by fluctuation. Part of this blending, moreover, 

 is not fluctuation in the ordinary sense, but rather an 

 autogenous fluctuation,^ provoked by correlation with 

 other characters. Thus the modes of seed -weight in an 

 F2 are not only subject to a P.E. of 12 per cent., but before 

 this allowance can be made they have to be corrected for 

 correlation with diameter of the boll. 



A simple way of applying such correction is to dissect 

 the frequency polygon of the family, isolating the wide- 

 boll and the narrow-boll forms, for example, and plotting 

 their seed- weights separately {e.g., Fig. 69). Unfortunately 

 this method reduces the size of the groups under examina- 

 tion, which has never been excessive to begin with, so 

 that the precision gained in one direction is lost in 

 another. 



Again, since correlation exists between certain 

 characters, it might be thought that the slide-rule would 

 give the necessary correction. The difficulty here is to 

 find out the value and nature of the correlation. The 

 simplest Mendelian combination, AB x ah, where the 

 characters are simply linear measurements, gives a correla- 

 tion diagram in Fg, which consists of four groups, only 

 separable with difiiculty even when correlation is perfect 

 and linear, which — worked out in the conventional way — 

 gives a value for r from 0"6 downwards (Fig. 53). There 

 does not seem to be any method at present extant by 

 which a quantitative separation of these groups can be 

 made.* 



The difficulties enumerated, both of experiment and of 

 computation, have prevented the author from making any 

 exhaustive statement which can be considered as honest 

 from the scientific viewpoint. The results are in the 

 main indicative, and suggestive, often strongly so, but in 

 all cases open to criticism. 



* W.L.B. 21, and reply by Craig, J. I., Gairo Sci. Jour., Auj^ust, 1910. 



K 



