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



nearly 2,000 tables. By considerable pruning and in particular by giving up the separation 

 by camps this number was finally reduced to 255 tables. After the sorting had begun it was 

 necessary to reduce the number of tables still further to 180 in order to keep the work within 

 reasonable time. Even with this reduction six Hollerith sorting machines and four Hollerith 

 tabulators were kept busy for the greater part of three weeks in making the sortings. 



Section 4. — Presentation of data. 



A word upon the history of the preparation of Part III will make understandable 

 both certain omissions from the part and also some failures of coordination between nat- 

 urally related chapters. Work upon the main statistical analysis of psychological examining 

 was begun in December, 1918, under the direction of Maj. Yerkes and the immediate supervision 

 of Capt. Boring. Plans for Hollerith analysis and the tentative lists of sortings referred to 

 above were made out and finally approved. Considerable difficulty was experienced, however, 

 in obtaining authority for clerical help to undertake the coding. 



Finally, after numerous delays, six clerks were secured early in January, 1919, and the 

 work of the selection of cards and of coding progressed. Capt. Boring and an assistant made 

 the selections from the files and counted the cards. Incidentally, at the same time they 

 unpacked the boxes from the camps and put these cards in files. The six clerks continued the 

 coding until March 15, when the coding was completed. During this period occasional samples 

 of the coders' work were checked. It was obviously impossible to check the entire task with 

 such insufficient clerical help; complete checking would have required a month to six weeks 

 longer. The transfer of the coded information from the psychological record cards to code 

 slips was made unexpectedly necessary by the system in use where the Hollerith sorting was 

 done. It had not been foreseen from the first that this sorting could not be done by the 

 statistical division of the Surgeon General's Office, and, when it was found necessary later to 

 accept the courtesy of the Actuary of the Bureau of War Risk Insurance, it proved necessary 

 to have the records in simpler form for the use of his punch operators. 



The transfer of all this information to code slips and its entire checking was accomplished 

 by about 150 clerks in 10 days' time. Thereafter the cards went to the punch operators and 

 early ^n April sorting on the Hollerith machines was begun. For this purpose the Actuary 

 of the Bureau of War Risk Insurance made available six sorting machines and four tabulating 

 machines, together with their operators. An officer of the Section of Psychology was placed 

 in charge. The tables were completed by the beginning of May, after approximately three 

 weeks of sorting. 



Had the preparation of Part III been delayed until the completion of these tables no 

 work for the part could have been undertaken until May. Since -the discontinuance of both 

 commissioned and clerical personnel was promised for July 1 such a procedure would have 

 been disastrous to the completion of Part III. The difficulty was avoided, though only in a 

 small part, by deciding on the program for Part III and having authors begin their work 

 where possible prior to the completion of the Hollerith sorting. 



It was foreseen that a scheme for the combination of examinations alpha and beta in 

 particular and of the individual examinations with alpha and beta would be necessary in any 

 such detailed comparison of groups as this part contemplated. Considerable doubt occurred 

 in the early stages of the work as to the proper mode of combination; for example, the question 

 arose whether regressions of beta and individual examinations upon alpha should be used or 

 regressions upon beta, or whether the examinations should be made equivalent by equating 

 scores corresponding to equal percentile points, or whether some other procedure should be 

 adopted. The fact that the distributions for these examinations are presumably not normal 

 distributions, that correlations between them are not linear, and that in general the examina- 

 tion scales are not made up throughout of equivalent units, cast doubt upon the validity of 

 any very accurate application of the ordinary theory of probabilities and of correlation to 

 these data. Accordingly Mr. Brown and Lieut. May prepared early in February a memorandum 

 presenting a plan of combination of examinations alpha and beta — the plan which is worked 



