In proposing a major new initiative in mathematics (see page 931), Rita Colwell, director of the National Science Foundation (NSF), is recognizing problems that have been apparent to US mathematicians for some time. If enacted, its positive ramifications will extend beyond university mathematics departments.

The United States enjoys global leadership in many branches of mathematics, but this conceals some festering problems. First are weaknesses in its educational system. According to the Third International Mathematics and Science Study, US students do quite well in mathematics at age 9, rather badly at 13 and abysmally at 17. US universities also do poorly at recruiting American students into mathematics, especially at the graduate level. Twenty years ago, three-quarters of US mathematics PhDs were Americans: today, more than half are foreign-born.

For those who persist into academia, grant support is paltry by US standards. And agencies that once gave generous support for university mathematics, such as the departments of energy and defence, now leave the NSF to foot two-thirds of the bill.

Yet mathematicians are much in demand. Their research has been applied in search engines and other Internet tools, and underpins progress in other disciplines. Thanks to the rise of genomics and structural biology, and the realization that modelling can yield valuable insights into biological systems, this is now as true in the life sciences as in the physical ones.

Colwell is aware of these challenges. In an ideal world, she says, the NSF's $125 million mathematics budget would double in two years, and double again thereafter. Progress towards these goals would allow bigger grants, better linkages with other disciplines, new research institutes and improved ties between universities and high schools.

If this wins the backing of Congress and the next US administration, the whole of science —and society at large — stands to benefit. The initiative would concentrate on aspects of mathematics — such as the study of dynamic systems and the modelling of uncertainty —that will help tackle problems that affect us all.