Justice by Statistics


The Harvard Crimson assumes no responsibility for the sentiments expressed by correspondents, and reserves the right to exclude any communication whose publication may for any reason seem undesirable. Except by special arrangement, communications cannot be published anonymously.

To the Editor of the CRIMSON:

Perhaps even more farmous than "Listerine" is the well known "One out of every five" advertisement which shakes its bony finger at all who venture forth from the academic confines of the Yard; yets its boaring seems to have been quite missed in the editorial entitled "Murder by Statistics," which appeared in Wednesday's Crime. The thesis of this editorial, which has been several times sponsored by Boston and other papers upon hearing of the system of marking under which the percentage of men altaining the various grades is predetermined is that such predestination is both "unfair" and "unjust." They seem to view the "distribution curve" of marks as akln to the prophecies of the crystal gazer and the vagaries of the weather man.

Of course an out-of-hand determination that 10 1-2 men would get A's and the rest would flunk would be little better than assigning the specific marks to the individuals upon their enrollment in the course, but what many people seem to miss is the fact that the "distribution curve" system of marking, when properly applied, is a bet on the consistency of performance of a large group, the class, as against the consistency of performance of one man, the instructor. And that man's task is proverbially difficult, though it ranges in degree from a course in elementary mathematics to are advanced course in English composition (with CRIMSON editorials at some undetermined point in between).

In short, it appears that all three points in the editorial are ill taken. In order to secure consistancy of marks it is probably not necessary, perhaps even not best, to use the same system in courses of widely different degrees of objectivity of attainment, though it is surely possible. Any properly derived "distribution curve," such as those based on the wide experience of the many large courses here, appears far better calculated to give the students their due than the flat of any one person; and finally both empirically and theoretically. "C" will be found to fall remarkably near the "average," even for a weighted group such as is found in the bulk of college courses. And as for advanced courses and those of very small size. In the one case it is hardly vital to discuss "probation" (as does the editorial), and in the other just surprise yourself by counting the number of blue eyes in a chance gathering of 25 and compare the result with a similar count in some other chance group. Of course this doesn't mean that they will be found in the ratio of exacty "one in five," though it may help to console some who are on "the danger line." Respectfully, Philip Walker '25