The world of science has accelerated. With the advent of online publication over the past 10 years, it no longer needs to take months or years for an accepted paper to become available to journal subscribers, and the term 'monthly journal' is losing its meaning. Articles are published online weeks to months before publication in print, with benefits all round: authors can make their peer-reviewed results available to the scientific community quickly, readers can keep abreast of the latest developments and publishers can provide a continuous stream of content in an increasingly competitive market.

But the downside of early online publishing is a confusing array of publicly available article types, awaiting print publication in various stages of editorial preparation. Adding to the confusion, interactive journals such as Atmospheric Chemistry and Physics place papers online first for peer review, and then in their final form. As the focus of scientific journals is moving from print to electronic publication, each publisher makes their own decision regarding the balance of speed versus the completeness of published work. But when papers go online before they are in final form, uncertainity arises regarding the canonical publication date.

Publisher's policies regarding the accessibility of online articles are equally piecemeal. Science Express — where Science papers are posted online up to six weeks ahead of publication in print — is available to site licence subscribers only as a premium add-on. And when journals of the American Geophysical Union publish 'in press' papers before their print version, only the titles of these papers are available to non-subscribers. On publication in print, abstracts are also free to access.

Nature Geoscience papers are published online in their final, definitive form — fully proofread and formatted — and the date of online publication is the date of record. However, we consider papers elsewhere as published as soon as the scientific content is fully available online, with a Digital Object Identifier (DOI name). That is, we are happy to highlight 'in press' articles, whatever format they are in. We also count them as part of the body of existing literature when assessing the advance of a submitted paper over existing knowledge.

Given the way the publishing industry is moving, it seems unlikely that print publication dates will play a role in the long run. And as the demand for print subscriptions wanes, unified payment models for accessing papers online and in print are likely to evolve. What needs to be decided is how much a preliminary paper published online should be allowed to change before it constitutes a new paper.