157 



Professor Thorndike's suggestions fall "into two divisions ac- 

 cording as one searches for means of measuring the specific 

 information, skill, interests, and habits added by courses in 

 science, or the more general changes in total mental make-up — 

 in, for instance, open-mindedness, accuracy, zest for verification 

 and the Hke. 



"The specific changes are, of course, the easier to measure. 

 Indeed, my first suggestion is that we make scientific use of the 

 measurements that we already make. For example, the regular 

 school examinations are, or should be, careful scientific measures 

 of important changes in our pupils. If we would test our classes 

 with the examinations set by other teachers, have the pupils' 

 work graded by other teachers, and print questions, work and 

 grades, we should be making a start toward a real measurement 

 of educational achievement. If examinations are worth giving 

 at all, they are worth giving, at least occasionally, in such a way 

 as to measure not only how well a pupil has satisfied some par- 

 ticular person, but also what he really is or knows or can do in 

 certain special fields. 



"We need thousands of significant questions, in each science, 

 thousands of 'originals' in physics, chemistry and biology Hke 

 the originals of geometry; and above all we need to have thou- 

 sands of classes tested by outside examiners; for if an examina- 

 tion, instead of being a hasty, subjective selection of questions, 

 graded still more personally (and alas, how hastily), were made 

 a serious educational measurement, the examination papers of a 

 year would alone give us a large start toward knowledge of 

 what science teaching actually does. 



"Knowledge may, however, be measured more conveniently 

 than by the examination of notebooks, essays, or replies to 

 questions of the ordinary sort. These have the merit of adequacy 

 and richness, but the defects of measuring too many things at 

 once and too indefinitely. Greater uniformity in the use of the 

 test, quickness in scoring it, and freedom from ambiguity in the 

 numerical value assigned can be secured by the exercise of 

 enough ingenuity. I will mention two tests as samples of the 

 many that are possible. The first is an adaptation of a test, 



