The latest fad in American education is the small high school. Spurred on by grants of $1 billion from the Bill and Melinda Gates Foundation, dozens of cities are downsizing their high schools. It is time to step back and ask whether high schools can be too small to be effective.

Last spring, Bill Gates told the nation's governors that today's high school is obsolete and that the United States risks losing its competitive edge in global markets to nations with better schools. Thomas Friedman, in his best-selling book "The World Is Flat," warns that the United States is failing to prepare enough students for careers in mathematics, science, technology and engineering.

The nation's business leaders, governors and educators apparently agree, because Gates's remedy -- small high schools -- has become the remedy of choice in many cities and communities.

No one seems to have asked, however, whether a high school can be too small and whether tiny high schools will fix the dual problems of low performance by our best students and low graduation rates overall. Most new high schools enroll fewer than 500 students, and advocates for small schools think they should be even smaller, perhaps around 300, so that the school has the feel of a community or a big family.

But a high school can be too small to provide a solid curriculum and to offer advanced courses in mathematics and science and foreign languages. Our nation has had many decades of experience with small schools in rural areas, which were indeed like a community or family, and they were seldom exemplars of rigorous academic preparation. In many cases, they did not offer even the courses in calculus, trigonometry or physics that students need to prepare for college study.

The reason the issue of school size dominates educational discussion (aside from the Gates family's billion dollars) is that too many super-size high schools, especially in urban districts, are demonstrably unsuccessful. These schools, which enroll thousands of students, have been handicapped by the common practice of social promotion, which sends barely literate teenagers to high school despite their lack of basic skills. Lost in the milieu of a giant shopping mall in which no one knows them or guides them, many students become alienated, become discipline problems and drop out.

Clearly it makes sense to refashion the high school to meet the demands of the 21st century. But does it make sense to replace today's educational behemoths with high schools for 200 or 300 students? Such schools may be appropriate for youngsters in need of intense remediation, but they will be too small to provide highly qualified teachers of mathematics and science or the other academic courses needed by able students.

What is the ideal size of a high school, a size that maximizes personal relationships between students and adults and that produces high academic achievement? The research on this crucial issue is surprisingly thin. Most of the "research" on small schools consists of testimonials by advocates for small schools.

The only rigorous study was conducted by Valerie E. Lee of the University of Michigan and Julia B. Smith of Western Michigan University. Lee and Smith analyzed federal data for nearly 10,000 students in 789 public and private high schools of varying size. They sampled the performance of these students in mathematics and reading as they progressed from eighth to 12th grade. Lee and Smith concluded that the ideal size for a high school is 600 to 900 students. Size matters, they said, because it affects social relations within the school and the ability to mount a reasonable curriculum. Schools that are too large lack any sense of community and cannot shape student behavior; schools that are too small cannot offer a solid curriculum.

In their study, low-income students made the greatest academic gains in schools of 600 to 900 students. Indeed, the performance of low-income students was worst in schools with more than 2,100 students. Size did not make as much difference for students from advantaged backgrounds, but even their performance peaked in the schools that enrolled between 600 and 900. Academic gains for both low-income and high-income students declined as enrollment fell below 600, and declined again in schools that enrolled fewer than 300.

Small schools of fewer than 300 may be appropriate for some students, especially those who have been educationally unsuccessful, but they are not the right size for most students. If we move too far in that direction, we may have the paradoxical outcome of higher graduation rates and persistent mediocre achievement.

The writer is research professor of education at New York University and a member of the Koret Task Force at the Hoover Institution. She is also a senior fellow at the Brookings Institution.