The rise of the Internet in the 1990s helped spark a radical idea for turning primary science publishing on its head. If journals charged authors a fee to publish, instead of charging readers and libraries a fee to subscribe, said the advocates, published peer-reviewed papers could be provided free to anyone in the world.

This simple-sounding notion provoked visceral debate, resulting in extensive multisided arguments and antagonism among advocates of this and other forms of 'open access' and publishers, librarians and funders. Most of that rancour has now given way to greater pragmatism, dialogue and compromise. There is a broad appreciation that change is inevitable, but that constructive change takes time, thought and experimentation.

There is a growing demand among lawmakers and funders for greater public access to the literature.

Open-access pioneers such as the not-for-profit Public Library of Science (PLoS) and its commercial cousin BioMed Central have successfully shown that the author-pays model can be financially viable in the real world — something many had doubted. But the demonstration also offers a dose of reality. PLoS's goal when it launched in 2003 was to prove that high-impact journals could be paid for by author fees of just US$1,500 per paper. Yet author fees for its top journals have risen to $2,900 per paper, and the organization's finances are critically dependent on the high volume of papers published in its online journal PLoS ONE. This low-overhead journal, which charges $1,350 per paper, does not make editorial judgements about its papers' merits, it simply passes them through peer review to certify that they are technically sound.

The PLoS experience highlights the challenge in applying the author-pays model universally. Many journals, such as Science or Nature and its sibling journals, rely on their subscription fees to support the costs of high selectivity, added-value editorial content, such as reviews, and online enhancements. Such journals would need to charge fees several times that of PLoS ONE to cover their costs and support investment. So although author-pays could be a viable model for many lower-overhead journals, its broader uptake within the publishing industry will depend on the level of funds that research agencies are willing to make available for scientists to pay publishing fees.

One valuable and established intermediate model is the hybrid approach, in which subscription journals give authors the option to pay a fee to make their article freely available instantly. Nature Publishing Group will soon be launching its first Nature research journal of this sort, Nature Communications. Economists who have studied the science publishing industry argue that the sector will ultimately evolve into a mix of open-access, subscription and hybrid journals, rather than a monoculture.

In the meantime, there is a growing demand among lawmakers and funders for greater public access to the literature, in particular in fields where public interest is strong, such as biomedicine. This demand seems most likely to be met at least for the foreseeable future by a different model of openness, which was articulated in a 2007 bill requiring researchers at the US National Institutes of Health (NIH) to make authors' or publishers' versions of research papers publicly available in the PubMed Central repository within 12 months of publication. There is speculation that President Barack Obama might soon issue an executive order extending this requirement to all federal research agencies (see page 822). Legislation to that effect has also been introduced in the US Senate, and may soon be introduced in the House of Representatives.

Nature's publishers have consistently backed the NIH mandate, and support its extension to other agencies. But whatever form the extension takes, it should be flexible about the compulsory time limit within which papers must be deposited in archives following their publication in journals. The NIH initially insisted on a 6-month embargo interval, but agreed to extend this to 12 months after protests from some publishers. Governments must not impose a one-size-fits-all embargo interval. At a time when many academic libraries are facing deep budget cuts, they may be tempted to axe subscriptions to many journals on the grounds that all but the most recent content is freely available in archives such as PubMed Central. And that, in turn, could particularly hurt journals in disciplines such as the social sciences, in which researchers use older material far more frequently than do those in fast-moving fields such as molecular biology. Publishers must be able to negotiate embargo intervals that will fulfil their obligation to allow greater public access but not jeopardize their businesses. And publishers, in turn, need to recognize that science's social contract is evolving towards greater openness.