450 



MEMOIRS NATIONAL ACADEMY OF SCIENCES. 



Table 118. — Correlation of intelligence ratings with score. 



[Vol. XV, 



Section 2. — Examination "a" and intelligence ratings. 



The data secured from examination of the groups described in table 80, p. 426, as alpha before 

 a, and a before alpha has been utilized in an attempt to discover what effect, if any, the revision 

 of the early examination a into the later alpha had upon the efficiency of the examination as in 

 instrument for the measurement of intelligence. 



Tables 119 and 120 give the results as far as total weighted scores (tests weighted as during 

 the examination in the first four army cantonments) are concerned, for a as a first and as a 

 second examination, the other examination given at the same sitting being alpha. Tables 84 

 and 121 give the results of the alpha examination of these same groups. It is immediately 

 obvious that the correlation of scores in both examinations, as either first or second examinations, 

 with ratings, are so nearly the same that there has been no very great change either for better 

 or for worse in the revision. To discover more exactly the effect of dropping tests 2 (memory 

 span for digits) and 10 (number comparison) table 122 was formed. The resulting correlation 

 coefficient is very slightly lower than for all 10 tests. Although unweighted scores have been 

 used in table 122 the result is almost exactly comparable with the result of table 119, owing to 

 the very high, practically perfect correlation between weighted and unweighted scores. But 

 the slight decrease in correlation when tests 2 and 10 are omitted may be entirely due to the 

 increase in zero scores — i. e., in the number of unmeasured cases, and may therefore not indicate 

 the loss of any qualitively essential or important feature of the examination. 



Table 119. — Correlation 0/ examination "a" scores with officers' ratings — "a" before alpha (Camp Meade). 



