The Role of Assessment in Teaching

Top

Ways to Evaluate Student Learning
Once you have collected all the measures you intend to collect -- for example, test scores, quiz scores, homework assignments, special projects, and laboratory experiments -- you will have to give the numbers some sort of value (the essence of evaluation). As you probably know, this is most often done by using an A to F grading scale. Typically, a grade of A indicates superior performance; a B, above-average performance; a C, average performance; a D, below-average performance; and an F, failure. There are two general ways to approach this task. One approach involves comparisons among students. Such forms of evaluation are called norm-referenced since students are identified as average (or normal), above average, or below average. An alternative approach is called criterion-referenced because performance is interpreted in terms of defined criteria. Although both approaches can be used, we favor criterion-referenced grading for reasons we will mention shortly.
NORM-REFERENCED GRADING
A norm-referenced grading system assumes that classroom achievement will naturally vary among a group of heterogeneous students because of differences in such characteristics as prior knowledge, learning skills, motivation, and aptitude. Under ideal circumstances (hundreds of scores from a diverse group of students), this variation produces a bell-shaped, or "normal," distribution of scores that ranges from low to high, has few tied scores, and has only a very few low scores and only a very few high scores. For this reason, norm-referenced grading procedures are also referred to as "grading on the curve."
CRITERION-REFERENCED GRADING
A criterion-referenced grading system permits students to benefit from mistakes and to improve their level of understanding and performance. Furthermore, it establishes an individual (and sometimes cooperative) reward structure, which fosters motivation to learn to a greater extent than other systems.
Under a criterion-referenced system, grades are determined through comparison of the extent to which each student has attained a defined standard (or criterion) of achievement or performance. Whether the rest of the students in the class are successful or unsuccessful in meeting that criterion is irrelevant. Thus, any distribution of grades is possible. Every student may get an A or an F, or no student may receive these grades. For reasons we will discuss shortly, very low or failing grades tend to occur less frequently under a criterion-referenced system.
A common version of criterion-referenced grading assigns letter grades on the basis of the percentage of test items answered correctly. For example, you may decide to award an A to anyone who correctly answers at least 85 percent of a set of test questions, a B to anyone who correctly answers 75 to 84 percent, and so on down to the lowest grade. To use this type of grading system fairly, which means specifying realistic criterion levels, you would need to have some prior knowledge of the levels at which students typically perform. You would thus be using normative information to establish absolute or fixed standards of performance. However, although norm-referenced and criterion-referenced grading systems both spring from a normative database (that is, from comparisons among students), only the former system uses those comparisons to directly determine grades.
Criterion-referenced grading systems (and criterion-referenced tests) have become increasingly popular in recent years primarily because of three factors. First, educators and parents complained that norm-referenced tests and grading systems provided too little specific information about student strengths and weaknesses. Second, educators have come to believe that clearly stated, specific objectives constitute performance standards, or criteria, that are best assessed with criterion-referenced measures. Third, and perhaps most important, contemporary theories of school learning claim that most, if not all, students can master most school objectives under the right circumstances. If this assertion is even close to being true, then norm-referenced testing and grading procedures, which depend on variability in performance, will lose much of their appeal.