American schoolchildren have grown perceptibly worse over the past five years in their ability to solve relatively simple mathematical problems, according to a survey of mathematics education released today.

Fractions, percentages and word problems of all kinds -- even ones as simple as "What is the sum of 54 and 21?" -- defeated large numbers of the 9, 13, and 17-year-olds tested, the survey found.

Fewer than half the 13-year-olds and only 13 percent of the 9-year-olds (mostly fourth graders) could determine "what does 2/3 of 9 equal?" (6).

Slightly over a third of the 13-year-olds and a bit over half of the 17-year-olds could "express 9/100 as a percent" (9%).

Nearly a quarter of the 9-year-olds could not correctly answer "what is the sum of 54 and 21?" although if it were set up for them as 21 + 54, 90 percent could provide the solution (75).

Performance at all three age levels has declined between 1973, when the tests were last given, and 1978 when this set was run, with the falloff sharpest among the older children.

Educators affiliated with the survey pointed to the discrepancy between computational skills and problem-solving and linked it to the "back to basics" trend that has surfaced in education in the interval between the tests.

"I think we should be asking ourselves if we are creating a generation who can compute but who don't know when to compute," said Roy H. Forbes, director of the National Assessment of Educational Progress, which conducted the survey.

"What is missing is problem solving performance," said Dr. Shirley A. Hill of the National Council of Teachers of Mathematics. "Often [the pupils] give answers that are not even reasonable."

Hill, a member of a panel of educators who interpreted the test results, pointed to a question in which the children were asked to figure out how many cars would be required to transport a number of people. Many, she said, gave fractional answers.

Hill charged that there has been "an excessive narrowing" of mathematics teaching.Schools have "been responsive to public pressure" to emphasize computation to the point where the teaching of application has been slighted, she said.

The decline, however, parallels a long-term one that has been recorded in college board aptitude scores. That decline, which began in the mid-1960s, encompasses the old math, the new math, and the back-to-basics math, and Hill acknowledged that she might be overemphasizing the role of recent curricula.

"One doesn't ever really know" the causes, she said, noting that a study of college board scores several years ago cited "multiple factors" that could be responsible for the decline.

The National Assessment of Education Progress is a long-term, government-sponsored program to try to evaluate how well American children are doing in a variety of subjects. The NAEP runs tests in 10 fields, from reading and writing to science, music and social studies, and reports periodically on its findings. The math tests involved 71,000 children.

"The pattern found in mathematics -- larger declines for older children and more of a tendency for declines . . . on higher-order cognitive skills -- is very similar to that found by the National Assessment in reading and writing," Forbes said in a prepared statement.

Thus, he said, children are doing well in basic mechanics of a broad range of subjects but fall down when it comes to making use of them.

Forbes and Hill both expressed concern that the "minimum competency" tests now being given in a number of jurisdictions may be, in Hill's words, "coming to dominate the curriculum." She said she fears that many schools are gearing their teaching to these tests.

The one bright spot in the results was the performance of 9- and 13-year old blacks, who showed substantial improvement.

Hill said she believes that "compensatory programs in inner cities are beginning to have an effect."

However, the report noted, "by age 17 favorable trends for blacks were no longer evident."

The report divided its results by race, sex, region of the country, parental education and community size. Findings included that whites do better than Hispanics who do better than blacks, though the younger blacks have been closing the gap somewhat; boys and girls do about the same at 9 and 13, but boys do better at 17; the more education their parents have, the better the children do.

Also, suburbanites do better than big-city dwellers, and "advantaged urban" dwellers do better than rural residents who do better than "disadvantaged urban." Northeastern children do the best, children from the central part of the country next best, westerners next best and southeasterners worst, though the 9-year-olds in the southeast showed the only improvement besides the younger blacks.