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



373 



The Pearson correlations of beta total with alpha total are: Total of 10 beta tests, 

 unweighted, r with alpha 0.795; total of eight beta tests weighted, r with alpha 0.793; total of 

 eight beta tests, unweighted, 0.790. The eight tests indicated were those included in the 

 revised scale. 



For the Camp Lee groups, in which all the alpha scores were below 100 (weighted score), 

 the correlation of alpha total with beta total was 0.65. 



The scatter with mental age for the Vineland group of 77 feeble-minded subjects is 

 shown in table 45. It will be seen from the table that the beta examination is difficult enough 

 to prevent feeble-minded subjects from making high scores. On the grade basis later adopted 

 only one of the 77 subjects of mental age 6 to 11 years made a grade of C, and only a small 

 proportion of the total number succeeded in making D. Certain of the tests proved especially 

 effective in keeping down the scores of the feeble-minded, notably test 12 (geometrical con- 

 struction), on which none of these subjects scored above 2. It was partly on this ground 

 that test 12 received a heavy weighting. 



Table 45. — Total score, weighted, eight tests. 



The distribution of scores on each of the 15 tests for each group (unselected, low, and 

 feeble-minded) was considered. Although each of the tests brought out differences between 

 the groups, some did so much more sharpie than others. This was an important factor in 

 the final selection of tests. 



Summarizing, the investigation brought out the following facts : 



1. That group intelligence tests for foreign and illiterate subjects are feasible, and that 

 the procedure employed in the trial series is fairly satisfactory. 



2. That the 15 tests included in the preliminary form of beta are not of equal validity or 

 practicability, but that all except three or four are usable with only slight changes. 



3. That the total score of eight of the tests correlates 0.79 with alpha total and makes a 

 fairly sharp distinction between groups known to differ in intelligence. 



Section 5. — Choice of tests. 



This was determined chiefly by the following considerations: (1) Correlation with 

 alpha and beta total scores; (2) clearness with which average, low, and institutional sub- 

 jects were differentiated; (3) ease of demonstration, as shown by low percentage of zero 

 scores; (4) economy of time in giving and scoring, and economy of space in record blank. 



The purely practical considerations were important. Tests 2, 7, 8, and 11 gave entirely 

 too many zero scores for inclusion in their present form. Test 1 5 was eliminated chiefly because 

 it requires about 50 per cent more time to give than most of the others and because it is difficult 

 to score. For the sake of convenience no two-page test was included; test 5 being reduced 

 to one page and retained, while tests 2, 8, 11, and 14, all two-page tests, were eliminated. 



The following are the important facts regarding the several tests, including reasons for 

 retention or rejection: 



Test 1. Maze. — Retained without change because it makes an excellent "first" test. It 

 can be successfully demonstrated; gives few zero scores. It correlates fairly well with total 

 scores of alpha and beta, but does not make a very sharp distinction between good and poor 

 groups. 



Test 2. Form recognition. — While it is possible that this test might have been improved 

 by easier demonstration samples and easier items at the beginning of the test, it had serious dis- 

 advantages. The demonstration was unsuccessful (19.3 per cent of zero scores at Camp Meade 

 and 31 per cent at Camp Lee). It showed the lowest correlations with intelligence of all the 

 tests tried, and is rather too easy, once the idea is caught. It was therefore dropped. 



