24 
PEOFESPOE K. PEAESON OX THE TXFLUEXCE OF XATUEAL 
way influences the change in tiie correlation between organs, if the distribution he 
normal,'"' the change depends only on the stringency of the selection. Breeders who 
select by the size of an organ only are in that case very likely to reduce the variability 
of the organ in the selected group l)y far more than 50 per cent. Accordingly, it is 
not to he wondered at if they, to a great extent, destroy the correlation between the 
selected organ and other organs. Tliis destruction would appear as a want of relation¬ 
ship l)etween parts, possihlv as a want of flxity in type. 
By means of ecpiation (lii.) can easily be found from ;’,2 and f) = Sj/ctj. All we need 
is a table of trigonometrical functions. We observe that is always of the same 
sign and less tlian For many biological purposes the following graphical con¬ 
struction gives quite sufficiently accurate residts. Let CAB be a quadrant, say of 
10 centims. radius, and take tl)e ])oint P on this quadrant distant PN = Ihrj.i 
from AB. 
Take QN = 
V 
PN, and let AQ meet the quadrant in B, then BM tlie 
distance of B fi'om AB = lOiq^, and consequently determines iq^. If the figure he 
drawn on decimal paper tlie determination of iq.i is peculiarly easy. 
Graphical method of finding correlation between organs A and B after selection has acted on A. 
ri.> = Ri 2 = BM = ‘44, as before. See p. 23. 
* 
It will do so if the di.stribntion be skew, see ‘Phil, Trans.,’ A, vol. 191, p. 231. 
