PSYCHOLOGICAL EXAMINING IN THE UNITED STATES AKMY. 



Part III.— MEASUREMENTS OF INTELLIGENCE IN THE UNITED STATES 



ARMY. 



CHAPTER 1. 



SAMPLING OF INTELLIGENCE RECORDS AND MODE OF ANALYSIS. 



Section 1. — The principal sample. 



In view of the fact that nearly two million psychological record cards were returned from 

 the camps to the Office of the Surgeon General it became necessary, for the statistical summary 

 of the results of examining, to select a considerable sample for analytical treatment. A sorting 

 of the entire group of cards would have been impossible with the facilities available and would, 

 moreover, have added little to the value of the results. The entire group is, after all, by no 

 means a random group, but depends upon such chance conditions as the particular camps in 

 which examining was progressing, the draft quotas sent to those camps, and many other minor 

 considerations that affected the numbers of examinations given to different classes of recruits. 



The problem of selecting a sample is obviously twofold. In the first place it is necessary 

 to obtain as far as possible a sample that is truly representative of the group under considera- 

 tion — in this case the drafted recruits of the United States Army. When the total group is 

 affected by such chance conditions as we have mentioned the sample must be chosen so as, if 

 possible, to be more representative than the total set of records. 



The second phase of the problem of sampling is the determination of the number of cases 

 in the sample. With respect to the reliability of results based upon a particular sample a law 

 of diminishing returns applies, for with equal added increments less and less increase of reliabdity 

 is acquired. Mathematically, the precision of data increases only as the square root of the 

 number of cases. There is no way of settling a priori upon the total number of cases for any 

 given sample. The greater the number of cases the more precise the data and the smaller the 

 differences that can subsequently be made out from it, but gross differences can be reliably 

 based upon a very few cases only. Since the comparisons can be made only subsequently to 

 the selecting of the sample, the safe rule is to choose as large a representative group as is prac- 

 ticable under the circumstances and to trust that it will prove itself a reliable basis for such 

 conclusions as may later be indicated. Thus it was planned to select a sample of approximately 

 100,000 records as representing white recruits in the draft Army and additional samples for 

 negroes, officers, and enlisted men of stages of military training more advanced than that of 

 the recruit. 



The ideal sample of data is the "random" sample. The selection of a "random" sample, 

 however, requires, paradoxical as it may seem, great pains and much planning. The psycho- 

 logical record cards received in Washington from the camps came in various sorts of files. Some 

 cards were arranged apparently without rule by companies; others were alphabetized by com- 

 panies; others were in large alphabetical files for total draft quotas, or even, in some cases, for 

 the total period of examining in the camp. Very often the individual examinations were sepa- 

 rated from the other examinations; sometimes even they were kept in separate files. Occa- 

 sionally alpha and beta examinations were separate. There was varied practice as to the 

 inclusion of negro recruits in the s.ame files with the white. The cards unfortunately were none 

 of them dated, so that it was not always possible in the case of separation of files to find the set 

 121435°— 21 36 553 



