ETS Plans "New" GRE

University Testing

The Graduate Record Exam (GRE), already evolving from pencil-and-paper to computer-based administration (see Examiner, Winter 1994-95), is scheduled for more changes in the next few years, according to its manufacturer, the Educational Testing Service (ETS).


Plans Address Content and Format

Two new General Test sections Writing and Mathematical Reasoning will be added. The three sections currently used Verbal, Analytic, Quantitative will be revised, and constructed response items will be integrated into the currently all multiple-choice test.


ETS calls these planned revisions an attempt to create some customization in the GRE; currently, over 400,000 students per year take the exact same test for entry into a wide range of graduate study fields. An ETS report, Toward a Description of Successful Graduate Students, concedes that the limitations of graduate school admissions tests in the face of the complexity of the graduate education process have long been recognized and that critical skills associated with scholarly and professional competence . . . are not currently measured by graduate admissions tests.


Starting in 1999, two separate GRE General Test packages will be available, each designed, according to ETS, to measure skills needed for different fields. One, for humanities applicants, will contain the revised Verbal, Analytical and Quantitative sections plus a new 45-minute Writing section. The second package, for math/science students, will also have the revised Verbal and Analytic and the Writing test, but a new Mathematical Reasoning section will replace the Quantitative component.


The revisions consist primarily of changes in the number of questions targeting particular skill areas. The new Quantitative section, for instance, will contain fewer Computation items and more Data Interpretation. The Verbal modifications will entail adding more paragraphs to Reading Comprehension question sets (a similar change was made to the SAT in 1994 but problems such as gender bias persist with that test).


ETS's addition of constructed response items to the GRE may reflect its concern for public relations and sales rather than educational integrity. The company receives intense criticism from education experts and consumers for its overuse of the multiple-choice testing format. Research shows that multiple-choice items favor males (see Examiner, Winter 1993-94), and they are widely believed to undermine educational quality by promoting rote instruction and superficial learning. However, sample GRE constructed response items available from ETS are essentially multiple-choice items with the answer options removed.


Bias & Misuse to Remain?

Currently, bias in the GRE may prevent many students from gaining admission to graduate schools and qualifying for scholarship funds, despite their abilities as proven in college. The GRE scores of whites, as opposed to under-represented minorities, and males, as opposed to females, cluster at the top of the score scale (see table).


Research from the National Science Foundation, the University of Florida, and Texas A&M University shows the GRE underpredicts the success of minority students. And an ETS study concluded the GRE particularly under-predicts for women over 25, who represent more than half of female test-takers. There is no evidence that ETS's new GRE alleviates the test's biases.


GRE misuse is another problem not addressed in the plans for change. ETS s Guide to the Use of the Graduate Record Examinations Program states a cutoff score based only on GRE scores should never be used as the sole criterion for denial of admission . . . Using GRE scores as the sole criterion may have a discriminatory impact and is therefore never appropriate. But only 10% of schools said they adhere to these guidelines, according to an ETS survey, The Role of GRE General and Subject Test Scores In Graduate Program Admission. Twenty-seven percent of schools that require the GRE reported that they use a cutoff, and 10% that recommend the test use a cutoff. Fifty percent of graduate departments requiring GREs reported using General Test scores to determine fellowship awards, and 46% of departments that only recommend applicants take the test use it for that purpose.


The GRE's standard error of measurement (SEM) which determines the point spread that must separate two individuals scores in order to say with reasonable confidence that they are truly different ranged from 96 to 118 points (depending on the section) in 1994-95. So while it cannot be said with reasonable confidence that 420 and 530 represent different levels of whatever is measured by the GRE Analytic section (SEM 118), a 420 scorer would nonetheless be shut out from a school or scholarship with a 500 cutoff.


Because the GRE is biased against women and under-represented minorities, lacks predictive validity and is frequently misused, graduate schools should base admissions on applicants college records and eliminate their use of the test new or not.