It becomes more apparent, as we attempt to offer a statement of failures as taken from the various reports, that they are not truly comparable. The bases of such percentages are not at all uniform. The basis used most frequently is the number enrolled at the end of the period rather than the total number enrolled for any class, for which the school has had to provide, and which should most reasonably form the basis of the percentage of failure. Furthermore, the failures for pupils who drop out are not usually counted. Yet, in most of the reports, the situation is not clearly indicated for either of the facts referred to. Still more difficult is the task of securing a general statement of failures by subjects, since the percentages are most frequently reported separately for each class, in each subject, and for different buildings, but with the number of pupils stated for neither the failures nor the enrollment. The St. Paul report[8] is an exception in this regard. To present the full situation it is indeed necessary to know the failures for particular teachers, subjects, and buildings, but it is also frequently necessary to be able to make a comparison of results for different systems. Consequently, in order to use the varied reports for the attempted comparison above, the plan was pursued of averaging the percentages as stated for the different classes, semesters, and years of a subject, in each school separately, and then selecting the median school thus determined as the one best representing the city or the system. This method was employed to modify the reports, and to secure the percentages as stated above for Denver, Paterson, and Butte. Any plan of averaging the percentages for the four years of English, or similarly for any other subject, may actually tend to misstate the facts, when the percentages or the numbers represented are not very nearly equal. But, in an incidental way, the difficulty serves to emphasize the inadequacy and the incomparability in the reporting of failures as found in the various studies, as well as to warn us of the hopelessness of reaching any conclusions apart from a knowledge of the procedure employed in securing the data. The basis is also provided for some interesting comparisons by isolating from the general distribution of failures by school subjects (p. 19) the same facts for the failing graduates. That gives the following distribution. THE FAILURES BY SCHOOL SUBJECTS FOR GRADUATES ONLY Total Math. Eng. Latin Ger. Fr. Hist. Sci. Bus. Span. or Subj's. Greek 5803 B. 660 403 521 241 191 180 251 91 7 6334 G. 782 347 673 257 240 410 394 162 12 Per Cent of Totals 24.8 12.9 20.5 8.5 7.4 10.1 11. 4.3 .3 SIMILAR PERCENTAGES FOR THE NON-GRADUATES As above 23.6 18.3 17.7 10.1 5.3 8.4 10. 6.3 .1