According to the findings, science and math achievement tests are the strongest predictors of college performance, while some of the language tests are the weakest.

(Editor's Note: This story is a follow-on to yesterday's piece, SAT Test Called Valid Predictor Of College Success, further refining the findings on these ubiquitous -- and always controversial -- exams.)

The ETS has offered SAT II subject tests (formerly called SAT achievement tests) for several decades, but few colleges and universities have required applicants to take them. As a result, research had focused primarily on the validity of their more widely used counterpart, the SAT I Reasoning Test.

However, a proposal by University of California president Richard Atkinson to replace the SAT I with an achievement test developed by the state spurred interest in more information about SAT II validity.

Atkinson has proposed that UC rely on a battery of SAT II tests until the state can develop its own test. Currently, UC requires applicants to take three SAT II exams: the writing test, one of the two math tests, and a third exam of their choice.

In response to the proposal, ETS researchers Leonard Ramist, Charles Lewis, and Carolyn McCauley-Jenkins undertook a detailed analysis of SAT I, SAT II, and high school grades as predictors of freshman GPA and performance in specific disciplines. UC also conducted its own validity study, based on data from its student body.

Of the two reports, the UC study has the advantage of being based on more recent data. The ETS analysis, however, is by far the more extensive and includes findings regarding the validity of individual tests that the UC report does not.

Of the SAT II tests for which data are provided, three are in the natural sciences and two in math. After adjustment for range restriction and other statistical artifacts, correlations between these five and freshman grade point average (FGPA) are striking.

All science tests show correlations exceeding .5 -- and chemistry and math II tests have the highest validities of the 14 subject tests studied.

Among the remaining SAT II subject tests, only English composition (now known as Writing) had a correlation with FGPA exceeding .5. Subject tests for German language and European history showed considerably less validity. The Spanish test proved the weakest predictor of all and was the only test where the correlation between scores and grades fell short of 0.2.

The analysis also found that while science and math tests are sound predictors of future performance in these disciplines, their validity is enhanced by combining scores with results from the SAT I. When results on both SAT II science tests and SAT I were considered together, predictive validity for the same subject taken in college was the highest achieved for any subject for which an SAT II is available. Nonetheless, the incremental validity of adding results from SAT II science or math tests to the SAT I alone was modest -- increasing predictive validity by .05-08.

The report also considers performance of different demographic groups, breaking down results by sex, ethnicity and native language. Test-takers were also categorized by an academic composite based on grades and test scores.

Minority students, those whose best language is not English, and students in the low academic composite generally performed better on the SAT II science tests than would have been expected based on their SAT I scores alone.

By contrast, students in the high academic composite performed less well on the chemistry and physics tests than predicted based on their SAT I scores.

The most dramatic difference between expected and actual performance was seen among the small number of females who took the SAT II physics test. In addition, black students and those whose best language is not English performed better than predicted on SAT II math tests. This suggests that questions on the SAT I math sections are more dependent on English skills than those on the SAT II.

A key limitation of the report is the absence of data for a broader range of academic outcomes. In a previous work on the predictive value of admissions criteria, author Ramist and colleague Nancy Burton reviewed evidence that a combination of SAT score and high school grades predicts multiple measures of success. In addition to GPA, these include college or departmental honors, acceptance to graduate or professional school, and completion of a graduate or professional degree. Corresponding data for the SAT II remain to be presented.

ETS cautions that some changes have been made to SAT II tests since data for the report were collected; whether these might impact the results is unknown. Also, as a group, SAT II test-takers are somewhat stronger academically than the general college-bound population, since most schools requiring the SAT II are highly selective ones. Whether the findings reported here would change if SAT II test-takers become more representative of the college-bound population remains to be determined.

The new report, titled Using Achievement Tests/SAT II: Subject Tests to Demonstrate Achievement and Predict College Grades; Sex, Language, Ethnic, and Parental Education Groups, is available free of charge on-line in PDF format at this URL.

(Editor's Note: This story is a virtual synopsis of an article by Patricia Hausman, Ph.D., in Science Insights, Vol 6, No 5, a publication of the National Association of Scholars. For the complete article, with tables, see this URL.)

Related website:

National Association of Scholars


[Contact: Patricia Hausman Ph.D.]

06-Feb-2002