624 



MEMOIRS NATIONAL ACADEMY OF SCIENCES. 



[Vol. XV, 



ciently approximate to the normal distribution (allowance being made of course for the proper 

 distribution of inadequately measured or unmeasured cases) to permit us to treat their scales 

 as linear and the regressions of the corresponding variables upon each other as linear. To be 

 in accord with our assumption we must admit that the other three variables, beta tests 1 2, 

 and 3, do not fit well into a linear system. 



' ' I ' I I I i 



8 



•o 



We shall now consider somewhat more specifically the histograms of the various tests. 

 It will be seen from mere inspection that none of the distributions of "class marks" except 

 alpha 1 and 7 and beta 1 and 6 could be fitted satisfactorily by any type of Pearson frequency 



