344 



MEMOIRS NATIONAL ACADEMY OF SCIENCES. 



[Vol. XV. 



entire regiment except for 45 men who did not make a single point on any test. The only 

 elimination from the other group was of those born in non-English speaking countries. Com- 

 parison of the score distributions for the two groups brings out the following: 



"Oral directions" was made harder, as was intended, and the distribution curve was greatly 

 improved. In "practical judgment" the proportion of zero scores was very greatly reduced 

 and the distribution improved. The synonym-antonym test appears not to have improved. The 

 introduction of five harder items at the end was of doubtful value, and the excessive number 

 of zero scores shows that the effort should have gone rather to improvement of instructions and 

 to the substitution of easier items at the beginning. The distribution of scores for "disarranged 

 sentences" was materially improved, but the number of zero scores is still too large. "Number 

 series" was made very much better. Its present form yields far fewer zero scores, and the form 

 of distribution curve is also improved. The attempt to simplify the instructions for the analogies 

 test was not particularly successful, as the distribution for the revised form is practically the 

 same as before. As was intended, the information test was made harder, but without increasing 

 materially the number of zero scores. Data presented elsewhere indicate that "number series" 

 and "practical judgment" were especially improved by the revision. 



Tables 25 and 26 give an opportunity to compare the merits of examination a and alpha 

 as indices of a pupil's probable success in high-school classes. The pupils of both groups were 

 enrolled in the first year of high school. The groups were tested under equally favorable con- 

 ditions and the higher correlation shown by table 26 seems to indicate a decided superiority of 

 alpha over examination a for this purpose. The correlation in table 26 (0.455) is equivalent to 

 a much higher correlation in a wider-range group. 



Table 25. — Correlation of raw total of examination (a) with average class mark of 494 California high-school pupils 



(r=0.343). 



Average class 

 mark. 



Total raw score, examination a. 



30-44 



45-59 



60-74 



120-134 



135-149 



150-164 



165-179 



195-209 



210 or 

 ove». 



Total. 



Total. 



76 



92 



105 



34 

 114 

 182 

 93 

 54 

 11 

 1 



Table 26. — Correlation of alpha score with average class mark of 480 California high-school pupils [not the same pupils 



as in Table 25) (r=0.4SS). 



The most important c riticisms of alpha are : (1) That it is less effective than examination a 

 in the lower ranges of intelligence, and (2) that it is made up of tests too highly intercorrelated. 

 The first criticism is answered by the fact that alpha was not intended to serve as a measure 

 of low-grade men, since it was planned that all who made low scores in alpha would be required 



