no. 2.] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES ARMY. 543 



Table 199. — Comparison of results by first and second trials — Examination a. 



Statistical data. 



Group. 



Enlisted 

 men. 



Devens, 

 officers' 

 training 

 camp. 



Taylor, 

 officers' 

 training 

 camp. 



Devens, 

 officers. 



School 

 children. 



Number of cases 



Mean: 



First trial 



Second trial 



Mean gain 



Percentage of gains 



Percentage of losses 



Percentage same 



r (first and second trial) . 



190 



155.6 

 180.5 

 24.9 

 79.5 

 7.4 

 13.2 

 .97 



331 



283.5 



313.4 



29.9 



90.9 



8.5 



.9 



225 



278.8 



312.0 



33.4 



90.2 



4.9 



4.9 



.84 



202 



313.3 

 330.2 

 16.9 

 74.7 

 19.3 

 5.9 

 (') 



324 



112.6 



152.0 



39.4 



91.1 



2.7 



6.2 



.90 



1 Not computed. 



• All groups show a high percentage of cases who improved their scores on second exam- 

 ination, varying in the different groups from 75 per cent to 91 per cent. The mean gain in score 

 varies from 16.6 for the officer group to 39.4 for the school children. The relatively small 

 amount of gain in the officer group is largely attributable to the high initial scores of this group. 

 (It may be noted that 19.3 per cent of this group actually made lower scores than in their 

 first examinations.) We have already noted the circumstances with regard to examination 

 forms which made for somewhat exaggerated gains in case of the group of school children. 

 Owing to the uncertainties connected with this group we shall disregard it in the remaining 

 discussion. The correlation coefficients (Pearson product moment) are given where these 

 have been computed, though the significance of these is more directly related to the problem 

 of reliability than to that of practice effeots. 



Inspection of the scatter tables for the various groups shows that there is not an even 

 gain over the whole range. For the two officers' training camp groups correlations were com- 

 puted between score in the first examination and change in score on second examination. The 

 correlation coefficients were both negative: Devens, r= — 0.476; Taylor, r= — 0.38. In 

 other words, there is reduction in gain toward the upper part of the intelligence scale. As a 

 further check on the relationship of position on the scale to amount of gain we have computed 

 the mean gain for oases making given letter grades on their first examinations. These are given 

 in Table 200. (Where the number of oases in any letter grade was very small the figures have 

 been omitted and the number of cases indicated in parentheses.) 



Table 200. — Mean gain on second trial of groups classified by grade made on first. 



In the light of the facts whioh we have stated regarding variations in conditions it would 

 be unprofitable to discuss the detailed differences which appear between the different groups. 

 Several facts, however, stand out rather prominently: 



(1) The amount of gain does not decrease steadily toward the upper part of the scale as 

 the correlation coefficients of score in first examination with gain may have implied. Instead, 

 the gain is apparently least at the two extremes, but, aside from this, reaches its maximum 

 toward the upper end in either the C 4- or the B range. 



(2) There is a distinct drop in gain for the A cases in all groups. 



