360 THE AMERICAN NATURALIST [Vol. LI V 



20-21 0.2 —9.93 



'22- 23 0.0 —8.93 



24- 26 0.2 — 7.93 



27- 29 0.6 —6.93 



30- 32 1.8 —5.93 ■ 



36- 39 8.2 —3.93 



40- 43 9.6 —2.93 



44- 48 11.1 —1.93 



49- 53 11.9 —0.93 



54- 59 11.1 -f 0.07 



73- 80 5.9 +3.07 



81-88 5.1 + 4.07 



89- 97 5.5 + 5.07 



98-107 2.3 + 6.07 



108-118 1.4 +7.07 



119-131 0.0 +8.07 



132-145 0.2 +9.07 



The following figure is based on the same arrangement 



This is much closer to a normal distribution of frequencies than 

 in the ordinary method. It is correspondingly more reliable in 

 the determination of the various constants. 



If the biological assumption upon which this tabulation is 

 based is correct the classes are of equal value as far as the factors 

 affecting facet numbers are concerned though the facet ranges 

 are different. In following out this view the intervals on the 

 variation scale have been expressed in terms of class units, each 

 unit being equivalent to a factor which produces a change of ten 

 per cent, in facet number. Some arbitrary point, for instance 

 the mean of the unselected stock, may be taken as the point of 

 reference or zero and every facet value has a corresponding fac- 

 torial value on the scale. The variation constants may be ob- 



