bagley] EDUCATIONAL RESEARCH 175- 



process continually produces a concomitant variation of the con- 

 duct-outcome, we may be fairly confident that the teaching stands 

 to the conduct-outcome as cause to effect, — although through 

 what series of intervening links we may be quite unable to say. 



Under the conditions of present-day school organization, 

 there are two methods of comparing variations in methods of 

 teaching or organization of materials with variations in conduct- 

 outcome. One is the statistical method which is illustrated by 

 Stone's investigations referred to above. This method involves 

 a comparison of a large number of schools operated under dif- 

 ferent conditions, subjecting the pupils of all the schools to the 

 same test, and then determining whether there is any "correla- 

 tion" between the differences in results and the differences in 

 organization of subject-matter, method of teaching, or other 

 factor in which the several schools may differ. This method is 

 .important in that it deals with large numbers of pupils among 

 whom unforeseen individual differences, or differences not in- 

 cluded among the correlated factors, may be assumed to be 

 equally distributed; in other words, these unrecognized differ- 

 ences offset one another, and may consequently be neglected. 



The second method of investigation is that of "parallel 

 groups". In this method, two classes, approximately equal in 

 number, age, home environment, and capacity, are subjected to 

 different methods or different types of subject-matter and the 

 effect of these different treatments is compared by subjecting 

 both classes or groups to the same test. This method might be 

 perhaps characterized as yielding results less general in their ap- 

 plication than the statistical method, but on the other hand, it 

 permits a comparison of the results of disputed methods under 

 conditions where a test of the former type would be impractica- 

 ble, and the results obtained, if-sufficiently suggestive, would war- 

 rant the repetition of the experiment until a sufficient number 

 of cases had been accumulated to permit the drawing of general 

 conclusions. Such experiments, carefully planned to account 

 for all known influencing factors, and repeated to discover 

 whether any factors not foreseen were really operating, would 

 ultimately yield very valuable results. 



The net outcome of educational investigation which follows 

 the general lines here indicated would be a gradual accumula- 

 tion of standards to which the progress of pupils could there- 

 after be referred, and definite conclusions drawn as to the 



