On the perennial argument about grading “on the curve” in college courses

Abstract

If letter grades assigned in a course are determined by the percent of objective test items answered correctly, then under various assumptions about the mean and standard deviation of normally distributed test scores, the distribution of the percent of students receiving each letter grade is highly anomalous. Conversely, if one insists that letter grades be “reasonably” distributed and at the same time determined by a typical percentage scheme, then allowed test parameters are severely constrained and not the ones which psychometricians consider most defensible. One recourse available to an instructor who does not wish to grade “on the curve” is to adjust the test parameters to dubious values by manipulation of item difficulty so that a desired grade distribution based on percentages is achieved. This informal adjustment procedure constitutes an indirect and inefficient method of employing test norms — in effect, grading “on the curve.”