Last August, as it does every summer, the College Board released its national SAT report, showing that the average score on the college entrance exam had climbed slightly, as had the scores of minority students. Many scholars used the results to assess the "achievement gap" that continues to separate black and Hispanic students from their Asian and non-Hispanic white peers.

But buried in the data was a fact overlooked by researchers and journalists: A record portion of the test-takers, 25 percent, had declined to disclose their ethnicity. Now, an independent analysis of years of SAT data suggests that growing American disdain for racial categories may cast doubt on how accurately the SAT can measure the achievement gap.

Dale Whittington, director of research and evaluation for the Shaker Heights, Ohio, school district, who investigated the discrepancy, said that for two decades the number of students who refused to list their race has been large enough to raise questions about using the SAT to research minority achievement. But the tripling of that number in just seven years "makes use of SAT data for such purposes irresponsible," she said.

According to Whittington, students not reporting their ethnicity have become "the largest minority group taking the SAT."

With so many students concealing their race on SAT applications, the College Board report of average scores for each ethnic group is open to question. The combined math and verbal score in 2003 for blacks (857) and Hispanics (about 905) cannot be reliably compared with the score for whites (1063) and Asians (1083) on the 1600-point test. Whittington and College Board officials say they cannot determine how big the discrepancy for each ethnic group may be.

Nevertheless, experts said, there is no question that a significant achievement gap exists among ethnic groups. A wide variety of standardized tests shows blacks and Hispanics trailing whites and Asians in educational proficiency.

Whittington's findings will appear in an upcoming issue of the online journal Education Policy Analysis Archives. Several testing experts endorsed the credibility of her findings, which drew information directly from the College Board's annual data charts.

Officials of the New York-based College Board, which owns the SAT, acknowledged the sharp growth in what they call "non-responders." More than 355,000 of the 1.4 million high school seniors who took the test in 2003 failed to provide their ethnicity in registering for the SAT. Similarly large numbers did not report their parents' education or income, the College Board said.

Colleges and universities have already expressed concern about the number of students declining to disclose their ethnicity on SAT applications. The large number of non-responders on the SAT makes it more difficult to determine the impact of recent efforts to change admissions processes to conform with the 2003 Supreme Court decision on affirmative action, they said.

College Board officials said they have since made it much more difficult to skip the ethnicity, income and parental education questions in the online SAT application, and are seeing a sharp drop in the number of non-responders this year.

Chiara Coletti, College Board vice president for communications and public affairs, said her office will in the future point out the non-response rate in its news releases and accompanying charts and graphs, rather than leave it to researchers such as Whittington to unearth.

But Whittington suggests in her paper, which can be found online at epaa.asu.edu, that even a sharply reduced percentage of non-responders will still leave ethnic analysis of SAT results in doubt. She cites the work of Howard Wainer, currently a professor of statistics at the University of Pennsylvania's Wharton School, with SAT data from 1981 to 1985.

Wainer found that 12 percent to 14 percent of SAT applicants failed to report their ethnicity. Unlike the most recent group of non-responders, the 1980s students who did not report ethnicity scored significantly below the national average. But there were still enough of them, Wainer said, to lead him to question the analytical usefulness of reported minority achievement rates.

"The nonresponse to ethnic identifiers is sufficient to introduce noise of a greater magnitude than the changes being interpreted as real," he said in a 1988 article in the journal American Psychology.

Whittington said the results of her study "not only confirm Wainer's results but also show that the noise generated by recent groups of non-respondents has increased from overwhelming to deafening."

The SAT ethnic statistics are troubling to school districts such as Shaker Heights and Virginia's Arlington County, which are part of the Minority Student Achievement Network that has been trying to raise achievement for African American and Hispanic students. Arlington schools planning and evaluation director Kathleen Wills said that the SAT non-response rate was so high last year that she and Superintendent Robert G. Smith discussed not distributing the ethnic results, "but we believe that could be interpreted as an attempt to withhold information."

Whittington and College Board experts said the 2003 non-responder group had more males than females, even though most SAT takers are female, and had an average SAT verbal score of 510, slightly above the national average of 507.

That composition differs from the makeup of non-responders in Whittington's school district, which she was able to determine by matching scores with other records on Shaker Heights High School seniors. The Shaker Heights non-responders were majority female, 82 percent white and 18 percent African American, while the portion of total students taking the SAT at Shaker Heights High was 59 percent non-Hispanic white and 36 percent African American.

Whittington said some students, particularly non-Hispanic whites, may have felt their ethnicity would not give them an advantage in college admissions. She said some students oppose racial categorizing in general and some African American students may have been influenced by the research of psychologist Claude Steele, who believes that black students do worse on tests when they reveal their race.

Washington area students who did not answer some personal questions on the SAT registration form gave various reasons for doing so. Julie Rager, a white senior at Mount Vernon High School in Fairfax County, said she skipped the family income and parental education questions on the SAT form because she did not know her family's income and did not think either question had any bearing on her scores.

Heather Morgan, another white senior at Mount Vernon, said she answered the race question but not the income and parental education questions because she thought they violated her privacy and did not "have anything to do with the outcome of the SAT."

But College Board officials say they think the principal reason for students skipping the questions was a poorly designed online application that made it easy for teenagers to avoid questions not directly related to signing up for the SAT. Brian O'Reilly, executive director of SAT information services, said when the percentage of non-responders hit the 25 percent mark last year, the College Board changed the form and the non-responses dropped rapidly.

Since May 22, O'Reilly said, applicants cannot complete their SAT registration online without addressing the ethnicity, family income and parental education questions. In each case, they can click the "I choose not to respond" option, but the number doing so is much less than 25 percent of applicants, he said. Complete figures will not be available until May.