Magda Colberg grew up in Puerto Rico, tutored by a cousin of King George VI, studying logic from age 14, becoming fluent in three languages and earning a doctorate in philosophy.

She's the bureaucrat charged with developing the first nationwide federal entrance examination offered in nine years. Her background is no more unusual than her job: writing an employment test that will select the best workers -- and at the same time produce a civil service that is representative of the diversity of the country.

A perfect nondiscriminatory employment test has been an elusive goal for decades in both the business community and the government. The Labor Department last week announced its intention to abandon the test it has used to refer millions of American workers for private jobs. The Office of Personnel Management is going in the opposite direction -- reintroducing nationwide tests after nearly a decade without them.

With federal judges staring down from the bench and the civil rights community watching from the front row, OPM has developed the exams to select about 5,000 federal employees a year for entry-level jobs on the government's management track.

Many people were surprised that the Reagan and Bush administrations would develop something so controversial and specialized in-house. This is, after all, the same government that contracted out the preparation of congressional testimony for the secretary of the Department of Energy.

But it turns out that one of the most out-of-the-way corridors of OPM harbors three dozen people, more than half with PhDs, who are acknowledged experts in the field of testing. Indeed, the government pioneered some areas of employment testing, developing tests that were once copied all over the world.

For nearly 10 years this group -- twice as large before a reduction-in-force at the beginning of the Reagan administration -- sat on the sidelines while the government used what former OPM director Donald Devine called "informal quotas" to fill its jobs.

When OPM officials were finally given the go-ahead to work on new nationwide tests, Colberg recalls being part of a larger group of research and testing experts "closeted in a conference room for a month," going over every possibility.

They decided to make half the examination a "biodata test" -- which is rather like a written personal interview. "Biodata" tests are multiple choice questionnaires about experience, skills and achievements in school, employment and other activities. Blacks do about as well as whites on these tests -- the government calls them Individual Achievement Records -- because they give applicants an opportunity to show how they have played the hand life has dealt them.

"Biodata" tests are quite common in industry, where the range of questions is broader than that used by the government. For example, one "biodata" question -- about the applicant's mother's educational level -- has been used by some companies and is "extremely valid" in predicting success on the job, according to researchers. In general, the higher the mother's educational level, the more likely an applicant is to be a successful employee, studies have shown.

This is the kind of question the government won't ask -- because it in effect discriminates against applicants whose mothers had little education.

The government's Individual Achievement Record has stripped out all questions that are beyond the applicant's control, according to OPM.

The other half of the new examination is a written test of language, reasoning and quantitative abilities.

Enter Colberg, 53, who wrote the guide to the development of a relatively unusual type of question based on logic. It is a formula that can be used to write questions with a wide range of difficulty using words and concepts drawn from the everyday world of the federal worker.

She was influenced by research that suggested minorities did far better on tests using familiar subject matter instead of questions drawn from more esoteric fields. The idea was to test reasoning ability without stacking the deck against applicants from disadvantaged backgrounds.

The new government test differs from the Scholastic Aptitude Tests familiar to college-bound high school students in that the SAT requires a greater "knowledge information base," according to Colberg, as well as more math (trigonometry and geometry).

In writing the new tests, the government has tried to make them directly relevant to skills employees need to perform the work they are applying for. Gone is the general knowledge section of the old PACE (Professional and Administrative Career Examination) test that proved so difficult for blacks and Hispanics.

"The knowledge was supposed to be accessible to everybody," Colberg said, but it turned out that "it discriminated extraneously against people not from the mainstream culture."

While the current tests are far from easy, they are bound to attract critics who argue that the government needs a civil service that is at home in the world of art and culture -- not just in federal offices and laboratories.

By contrast, Richard Seymour, the attorney who sued the government over the PACE, said he is withholding judgment on the new tests until he sees who the government actually hires. "Our approach is to look at the numbers," he said. Referring to a consent decree signed in 1981, he said, "The decree requires the government to do everything it can to eliminate needless adverse impact" on minorities.

"In some instances they cannot eliminate adverse impact, because the test measures real job-related differences," Seymour said. But much "needless adverse impact" can be eliminated by making questions less abstract and more concrete, he said.

"In PACE there were some questions where the inferences were not watertight and the correct answers had to be done by consensus of psychologists," Colberg said. In the new tests, "all questions have only one incontrovertibly correct answer," she said, although there are plenty of illogical biases that are sometimes confused with the correct answer.

One of the obvious concerns is that the new questions are too narrow and will simply select candidates who have studied logic and fail everybody else -- but Colberg says this is not so.

"We gave the test to middle-class students in three highly developed countries -- in France, Spain and the U.S.A. In France and Spain more than 60 percent of those who took the test had taken logic; in America, the figure was 3 percent. The Americans performed better than the French and almost as well as the Spanish," she said.

Most of the testing community is withholding judgment on the new exams until results and validation studies have been published, which will take more than a year.

Colbert denied that the government has used widespread tricks of the trade to eliminate adverse impact -- making the questions easier or reducing the passing score so that a larger proportion of those who take the test will pass.

"We are not going to make the grading easier on this test," Colberg said. "Generally we will never have to go to the bottom of the register {list} to hire people. But in some geographic areas where the labor market is poor we may have to hire {them}. We are not going to place the cut score so low that everybody will pass because then the predictive value of the test is lost."