A new experimental college-admissions test appears to augment the SAT's ability to predict college students' first-year grade point averages, according to a report presented here on Saturday at the annual convention of the American Psychological Association. The researchers who developed the test, which is designed to measure intelligence more broadly than does the SAT, suggested that it would also reduce ethnic and racial disparities in admissions scores.

"Right now, admissions testing procedures are basically atheoretical," said Robert J. Sternberg, the lead investigator of the Rainbow Project, as the experimental effort is known. "There is no theoretical connection between what goes on in psychology and what goes on in admissions testing. ... We want to provide greater equity in admissions and reduce the amount of talent lost to society." Mr. Sternberg is also a professor of psychology at Yale University and the psychology association's president.

The Rainbow Project team includes five researchers at Yale and two at the University of Virginia. Its test is intended to be used in combination with, not as a replacement for, the traditional SAT-I test. (The SAT-I test measures verbal and mathematical ability, and will soon add a writing component. The newer SAT-II tests assess skills and knowledge in particular subject areas.)

The project is financed entirely by the College Board, the nonprofit organization that administers the SAT. The project's research is strictly experimental, however, and the College Board has made no pledge to adopt such a test.

Mr. Sternberg designed the test around his notion of "triarchic intelligence." In that model, the analytic intelligence measured by the SAT is complemented by measures of creative and practical intelligence. These three categories are not defined as completely distinct types of intelligence, but instead are situational. That is, creative intelligence is the ability to apply knowledge in novel situations, practical intelligence is the ability to apply knowledge in everyday situations, and analytic intelligence is the ability to apply knowledge abstractly.

Since 1985, Mr. Sternberg and his colleagues have conducted dozens of studies that, he says, demonstrate the internal validity and predictive power of the triarchic model.

The Rainbow Project's experimental test includes three components. First, the students take the Sternberg Triarchic Abilities Test, a multiple-choice test that measures all three types of intelligence in Mr. Sternberg's model.

Second, the students are given a number of "tacit-knowledge inventories" that are designed to measure practical intelligence. One item, for example, measures students' common-sense knowledge of how to succeed academically, by asking how consistently (on a scale of 1 to 7) they attend class, take comprehensive notes, and so on.

Finally, the students are given performance tests designed to measure creative intelligence. Some are asked to write two very short stories with titles selected from a short list: "A Fifth Chance," "It's Moving Backwards," and so on. These stories are assessed by six judges on measures of originality, complexity, emotional evocativeness, and descriptiveness. Other performance tasks include writing captions for cartoons and composing oral stories that must include objects in drawings presented to the test-taker.

At Saturday's panel, Mr. Sternberg presented data from the first phase of the experiment, in which the test was administered to 973 students at five community colleges and eight four-year colleges in the spring of 2001. Some of those 13 colleges are highly selective, and others have essentially open admissions. The students' SAT scores mirrored the general national score distribution of students who attend at least one year of college.

Using a variety of statistical techniques, Mr. Sternberg and his colleagues found that the Rainbow Project's test -- when combined with the students' SAT scores -- significantly boosts the ability to predict students' cumulative first-year grade point average. SAT scores alone account for 10.3 percent of the variance in the students' grades, Mr. Sternberg's report said, while all of the test scores taken in combination account for 27.3 percent of the variance.

The researchers also found much smaller racial and ethnic disparities in Rainbow Test scores than in the students' SAT scores. "We contrasted whites' and Asians' scores against the scores of African Americans, Hispanic Americans, and Native Americans," Mr. Sternberg said, "and found that ethnicity predicted very little about the scores in our Rainbow test."

The project is viewed skeptically by Robert A. Schaeffer, the public education director of the National Center for Fair and Open Testing, which campaigns against high-stakes ability tests. In a telephone interview, Mr. Schaeffer offered praise for Mr. Sternberg's past work, but said that from his organization's perspective, "All of this is unnecessary and a waste of research time and money, unless you come up with some extraordinarily strong predictor, which nobody has, and which personally I doubt you can, simply because of the difficulty of predicting an adolescent's behavior one or four years hence."

"The best predictor that anyone has found," Mr. Schaeffer continued, "is high-school record, expressed as grade or class rank. And that's not an answer the College Board likes, because it says you don't need their test."

A different kind of skepticism is offered by Stephan Thernstrom, a professor of history at Harvard University and the co-author of No Excuses: Closing the Racial Gap in Learning, which is scheduled to be published by Simon & Schuster in October. Mr. Thernstrom believes strongly in the utility of the traditional SAT. He said in a telephone interview, "My experience has been that by the time they enter college, there are huge differences among students in their basic skills, their literacy, and their numeracy. ... I think a great deal of this discussion fails to comprehend what elite schools require of their first-year students." He fears that Mr. Sternberg's project might dilute the SAT's predictive power.

A second phase of the Rainbow Project experiment will involve a much larger number of subjects. This new phase will track students' performance beyond the first year of college, and will look specifically at how they perform in especially difficult first-year classes, such as organic chemistry. Results from the second phase will be released in approximately three years.

The Rainbow Project is just one of several experiments currently being financed with an eye toward possibly augmenting the SAT, according to Amy Elizabeth Schmidt, the College Board's executive director for higher education research. Her organization is also supporting investigations into new methods of measuring students' motivation and study skills and of verifying students' accounts of their extracurricular activities.