no. 3.] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES ARMY. 661 



The distribution of letter grades, as is shown in table 178, is affected by these differences 

 in difficulty between forms. 



It is obvious that some of these differences are large enough to be "significant" in the mathe- 

 matical sense not only, but also to be of importance in comparing the intelligence of groups. 

 Form 5 is easier by about 6 points than forms 6, 8, and 9 ; form 6 is more difficult relatively at the 

 lower end and form 9 at the upper end of the distribution, etc. A difference of 6 points between 

 groups is great enough to appear significant even when the number of cases is relatively small 

 and could lead to a "significant" but factitious intelligence difference between the two groups 

 if one had been examined by form 5 and the other by form 6. All forms were sent in equal 

 numbers to each camp, so that any large group selected at random is likely to include all forms 

 in approximately equal proportions. (See tables 174 and 175.) 



Figures 8 and 9 represent graphically these same data. Figure 8 compares the extreme 

 forms, 5 and 6, and shows how small their difference is in relation to the whole range of alpha 

 scores. Figure 9 is drawn in the same way as figures illustrating differences between organiza- 

 tions, between ranks, etc., and shows that some of these are not much greater than difference of 

 form alone could introduce. 



20% 



\Q%- 



D- 



C- 



—r- 

 C 



i 



c+ 



e> 



— r— 

 A 



Fig. 8. Comparison of alpha forms. Percentage distribution of letter grades on form 5 (6,648 cases) and form 6 (5,803 cases), taken from the 

 principal Hollerith sample, Group I (white draft). These forms are the extremes; form 5 is the easiest and form 6 the most difficult of the 

 five. All other pairs show less difference. 



Previous study of statistical reports on form differences sent in from several camps had 

 shown substantially the same relation to exist between forms. These figures are here presented 

 as corroborative evidence. 



