420 



MEMOIRS NATIONAL ACADEMY OF SCIENCES. 



[Vol. XV, 



Assuming that we are dealing with an unselected group, table 79 shows that some limit 

 between double and single time would probably be more suitable. Arranging the tests here 

 in the order of "per cent attempting all items in double time" and arranging from lowest to 

 highest per cents we have 2, 4, 7, 6, 8, 3, 5. 



We can get before us a picture of the relative effects of double time on the separate tests 

 by collecting the various rankings of tests made in the preceding paragraphs. 



1. Tests in order of agreement of distributions for double time and single time, 2, 4, 6, 7, 

 8, 5, 3. 



2. Tests in order of degree of correlation between scores in single time and double time, 

 6, 4, 2, 7, 8, 5, 3. 



3. Tests in order of improvement of double time over single time as measured by officers' 

 ratings, 3, 4, 8, 5, 6, 7, 2. 



4. Tests in order of improvement of double time over single time as measured by correla- 

 tion with total score (single time), 5, 3, 6, 7, 8, 4, 2. 



5. Tests in order of improvement of double time over single time as measured by corre- 

 lation with total score of remaining six tests, 5, 3, 6, 7, 8, 4, 2. 



6. Tests in order of the degree in which they can afford increase of time without an injus- 

 tice to quick men, 5, 3, 8, 6, 7, 4, 2. 



Similar results were obtained in a second experiment with double time on alpha. The 

 data for this experiment are not given in detail because they add little that is new to the fore- 

 going conclusions. The conditions were somewhat different: The men were allowed to finish 

 the alpha examination in the usual manner; then, when the last test was completed, they were 

 instructed to go back to test 2, to draw a line across the page at the point which they had 

 reached when time was called and to begin there and work on down the page. After they had 

 continued with test 2 an additional period equal to the first, the same procedure was followed 

 successively with the remaining tests. Instructions were given not to make any corrections 

 or changes in the part of the test done during the first period. There were 155 men given 

 alpha at Camp MacArthur under these conditions. 



The correlation between total score, single time, and total score, double time, in this 

 experiment, is 0.967, as against 0.965 in the preceding experiment. The correlation of total 

 score, double time, with officers' ratings, is 0.559; of total score, single time, with officers' 

 ratings, 0.526 — a difference of 0.033 as against 0.04 in the preceding experiment. 



The correlations between double time and single time for the various tests are as follows: 



Tests 



Correlations. 



3 

 0.832 



i 

 0.960 



0.937 



The various tests may be ranked as follows: 



1. Tests in order of agreement of distributions for double time and single time, 2, 4, 6, 5, 

 7, 8, 3. 



2. Tests in order of degree of correlation between scores in single time and double time, 

 4, 2, 6, 5, 7, 8, 3. 



In addition to the foregoing data there are results for giving beta with double time to 60 

 men at Camp MacArthur. The number of cases is too small to justify detailed treatment. 



The correlation between officers ratings and beta total score, single time, is 0.460; with 

 beta, double time, 0.370. The correlation between beta total score, double time, and total 

 score, single time, is 0.950. This seems to indicate that the scale as a whole is affected very 

 little by doubling the time and that the effect is injurious rather than helpful. 



Considering the tests separately we get the following correlations: 



