394 MEMOIRS NATIONAL ACADEMY OF SCIENCES. [vol.xv, 



but this is a matter of less importance, since relatively few cases are graded on Stanford and 

 grades are ordinarily assigned mainly in the lower ranges. 



Considering the dividing line between D and E with a view to its efficiency in recall of 

 cases for individual examination we note, by reference to table 54, that the same difficulties 

 appear in establishing this dividing line as we found with reference to present weighted score 

 (see pp. 386ff), viz, that we must either recall too few whom we want or too many whom 

 we do not want. 



The dividing line of 20, which was adopted, recalls the following numbers as distributed 

 by Stanford mental age: 



Below 8 years, 4 cases (100 per cent of all cases below 8 years, M. A.). 



8 to 9 years, 11 cases (50 per cent of all cases S to 9 years, M. A.). 



9 to 10 years, 11 cases (17.7 per cent of all cases 9 to 10 years, M. A.). 

 Above 10 years, 15 cases (2.7 per cent of all cases over 10 years, M. A.). 

 Total number recalled, 41 cases (6.3 per cent of total English-speaking group). 



Comparison with the figures given on page 387 for the present dividing line of 40 (weighted 

 score) shows that while the suggested dividing line of 20 (suggested raw score) recalls a few 

 more cases, the increase is mainly due to the addition of cases who tested low on Stanford 

 and whom it was desirable to recall. 



Any reduction below the point of 20 seemed unsafe, since it would mean missing too 

 many cases in the dubious range who should be seen, if possible, before being passed upon. 



Section 6.— Summary. 



The following modifications, all of which are incorporated in the final edition of the 

 Examiners' Guide, were made in beta after its first introduction into the scheme of camp 

 examining: 



1 . Slight changes in procedure recommended by the camps were introduced. 



2. Time limits on five of the beta tests were reduced as a result of inspection of distribu- 

 tions of men making low scores in alpha. 



3. Test 8 (spot pattern) was dropped on the basis of evidence to the effect that it could 

 be dropped without appreciable lowering of correlation between beta and Stanford-Binet. 



4. Raw scores were adopted, since their use meant both a saving of time for the scoring 

 units and a decrease in probability of errors and since their use involved no appreciable reduc- 

 tion in correlations with either the Stanford-Binet scale or alpha. 



5. The numerical value of test 1 (the maze) in the total was reduced by giving one-half 

 point instead of one point for each half maze correct. This change was due to the evidence 

 from correlations to the effect that this was one of the least satisfactory tests of beta; also 

 because inspection of tables showed that many low-grade men as measured by other stand- 

 ards made very good scores in the maze, thereby raising their total scores unduly and con- 

 tributing to the difficulty of establishing a satisfactory lower dividing line. 



6. New standards for assigning grades were established which increased the consistency 

 of grades as assigned on beta and on alpha. 



The mam conclusions based on this more extensive trial of beta were as follows: 



1. Beta evidently constitutes an important addition to the series of examinations for 

 recruits for the following reasons: (a) It helps to compensate for the fact that alpha becomes 

 rapidly too severe at the lower end of the scale, since beta makes discriminations below the 

 zero point of alpha. (&) It affords a possible method of examining men who are illiterate in 

 EnglisTi, whether foreign or native born. For this purpose it is at least far more satisfactory 

 than alpha. 



2. Beta is a distinctly easier type of examination than either alpha or the Stanford scale. 

 For the total English-speaking group of 653 cases the mode falls in the upper part of the scale 

 and quite near to the upper limits of beta. In other words, there is no trailing off, such as alpha 

 shows, at the upper end. This results in making beta less discriminating than alpha for the 

 upper portion of the scale. 



