Seventy-two names are on the list of new members of the US National Academy of Sciences, elected on 1 May. Nine stand out: Tania Baker, Ursula Bellugi, Karen Cook, Mary Estes, Pamela Fraker, Angela Gronenborn, Helen Hobbs, Laura Kiessling and Eve Marder.

Two years ago, the academy elected 19 women to its ranks; this year, the number is less than half of that. Over the years there have been a plethora of programmes designed to introduce women into science, and more sporadic efforts to keep them in the career pipeline while they bear and raise children. Yet women have still not come remotely close to closing the gender gap at the senior level.

Of course, some women do reach the scientific élite: at its meeting last week, the academy awarded its highest honour, the Public Welfare Medal, to biologist Maxine Singer. But Singer has little female company at the top of the scientific hierarchy.

Roughly 10% of members of the science academy are women. This is up from just 6% in 2000, but is still a disappointing number. Even as the percentage of women rises in many research fields, women still find it harder to join the scientific élite — even in the United States, where they have had a firmer foothold for longer than elsewhere.

As US science's most exclusive club, the academy is fully aware that its membership is dominated by white males of a certain age, and has made attempts to address the fact. New members are nominated and elected by other members — which is, of course, a recipe for perpetuating such bias. But some of the academy's 31 discipline-based sections have adopted proactive schemes for identifying promising female candidates, and a set of nominating groups established in 2003 helps more women and younger candidates enter the mix of potential members. The Royal Society in London has also made efforts in the past five years to increase the number of women in its ranks.

Such measures are to be applauded, but they don't seem to be working as well as they might. Options for a more direct assault on the issue are problematic, however. Setting up any kind of quota system, for example, would trigger a cascade of difficulties, starting with the possibly diminished status of women elected as part of a quota.

A slightly higher cap on the number of members admitted each year might better reflect the growing size and academic diversity of the scientific community, and open up the pipeline a little for deserving candidates of both sexes. But it would do little to address the gender imbalance.

Perhaps the best thing the academy can do is find ways to get suitably qualified women on the ballot in each of its sections. Such a change may, for instance, require a stipulation that nominees from diverse backgrounds will at least appear on the ballot.

Academy members at all levels should also take a more prominent and public role in promoting initiatives that will secure fair treatment for women scientists. All too often, discussions about advancing women or minorities in science spring from the same people — usually the women or minorities themselves. Some leaders do get involved, but it is up to them all to recognize that broadening diversity is more than just a feelgood effort, something to chalk up as a good deed done in the name of equality and then be forgotten.

A new initiative in US physics is to be applauded for taking steps in this direction. On 6–8 May, the chairs of 50 physics departments, plus 15 senior managers from national laboratories, met in College Park, Maryland, to discuss how to double the number of women in physics by 2022. The fact that a number of high-level researchers attended is cause for optimism. It remains to be seen how this effort will develop over time, but other fields would do well to consider similar moves.

Women in the United States have been told for decades that they need to enter science at the bottom in order to make their way to the top. But this situation has been going on for too long. Those already in the scientific élite must take it upon themselves to bring about genuine gender equity.