490 MEMOIRS NATIONAL ACADEMY OF SCIENCES. [Tol. >v, 



Referring to table 144 and figure 2 we note first the variation in percentages eliminated 

 as illiterate in the different camps, ranging from 7.4 per cent at Camp Taylor to 20.4 per cent 

 at Camp Devens. This precipitates at once the question whether we can take these data at 

 their face value and conclude that Devens is actually characterized by a peculiarly large per- 

 centage of illiterates (including the foreigners illiterate in English), and whether the other 

 camps are characterized by the relative per cents indicated. 



Two fines of evidence bear upon this point. The first of these is drawn from the known 

 facts regarding procedure, as given on pages 473 ff. chapter 13. It has been shown there 

 that the standard for segregation recommended by Camp Devens and tried out for at least 

 part of the time was higher than that used consistently by any of the other camps. Camp 

 Lee, on the other hand, set up very nearly a minimal requirement. When, therefore, we find 

 Camp Devens credited with 20.4 per cent illiterates as compared with 17.7 per cent for Lee, it 

 seems probable that this is more significant of differences in standards of segregation than of 

 differences in numbers of "illiterates." 



We note also, as shown clearly by figure 2, that, whereas Camp Lee has both a large pro- 

 portion of illiterates and a large proportion of D and E men left over after elimination of the 

 so-called illiterates, Camp Devens has a large proportion of illiterates, but a very small pro- 

 portion of D and E oases as compared not only with Camp Lee but with Camps Dix and Taylor 

 as well. This fact suggests that the high literacy standard of Camp Devens operated to exclude 

 considerable numbers of the men of inferior intelligence. This probability reduces the signifi- 

 cance of the data concerning relative numbers of men making low scores in the various groups. 



If the group actually excluded from examination is to be treated as an integral part of the 

 whole, certain assumptions regarding the mentality of these oases become necessary. The 

 assumption which seems best justified is that such cases would mass toward the lower portion 

 of the intelligence scale. Results obtained later by the use of examination beta, designed for 

 illiterates show that this is the case to such an extent that the number of men eliminated 

 making scores above the average is small in comparison with the total and that the number 

 making scores above the upper quartile is negligible. 



One conclusion following upon this assumption is that the significance of percentages 

 making given scores for intelligence increases toward the upper portion of the intelligence 

 scale. Stated in other terms we can compare the groups with respect to the proportions of 

 superior men with a reasonable degree of accuracy, whereas we can not, from the present data, 

 compare them with reference to their proportions of inferior men. Accordingly we have shown 

 in table 144 the percentages of men making A and B grades, and would call especial attention 

 to the relative lengths of the bars representing these two grades in figure 2. From this point 

 of view the order of the camps from best to poorest is Dix, Devens, Taylor, Lee, although the 

 difference between Dix and Devens (1.2 per cent) is too small to emphasize. 



Comparison of groups in terms of some measure of central tendency seems also desirable. 

 We have chosen the median for this purpose since it was possible by this means to compare 

 total groups, assuming that the cases excluded for illiteracy would fall in the lower half of the 

 group. Medians computed thus are shown in table 144. Here, again, the order of excellence 

 of the camps is Dix, Devens, Taylor, and Lee, showing in this case a marked superiority for 

 Dix and only a slight difference between Devens and Taylor. The upper quartiles, which are 

 less affected than the medians by the uncertainty regarding the ''illiterates," show the same 

 order, with Devens approximating more closely to Dix and more widely separated from Taylor. 

 Lower quartiles were not computed, since the assumption that the excluded groups would fall 

 in the lowest 25 per cent seemed quite unjustified. Instead of offering the quartile deviation 

 (semi -interquartile range) as a measure of dispersion we therefore show the range from the 

 median to the upper quartile. If we could assume the distribution to be symmetrical this 

 measure would be identical with the quartile deviation. Since we have no data in support of 

 this assumption we offer this as indicating merely the range of variation for that portion of the 

 scale which we are justified in considering. The measure indicates a slightly wider scatter of 

 cases toward the upper end of the scale in the case of Lee and Devens than in the case of Dix 

 and Taylor. 



