Page:Finch Group report.pdf/86

 uses that can be made of it. Those restrictions serve to limit—as they are intended to do—the usefulness of what is made available to readers via repositories. Hence as we noted in Section 4, evidence as to any potential impact on the viability of journals arising from the access provided via repositories under current restrictions is as yet not clear; journal publishing has continued to grow in recent years.

7.65. As to what version of a paper that can be deposited, relatively few subscription-based publishers allow the version of record—that is, the version finally published, with the functionality associated with links and semantic mark-up—to be deposited and made accessible. Those few that do—as, for example with the British Medical Journal—allow such deposit only after an embargo period. Most other publishers allow either the submitted or the accepted (after peer review) manuscript to be deposited; and policies vary as to which of those two it should be. Some of the major publishers co-operate with the NIH by depositing in PubMedCentral versions of the articles they publish, but with a disclaimer making clear that what is accessible there is not the version of record, which remains accessible only from the publisher’s site. Highlighting the status of different versions of the article in this way is now complemented by the CrossMark service which puts a kitemark on the version of record in its most up-to-date form.

7.66. Funders have in general sought embargo periods of twelve months, and publishers of subscription-based journals are very concerned at any moves to reduce that period, believing that it would lead to a loss of subscriptions that would put the viability of their journals at risk. The concerns focus on the half-life of journals in terms of downloads: the length of time it takes the articles in each volume to reach half the number of downloads they will reach in total. Some major publishers have supplied us with figures which indicate half-lives varying from two-and-a half years in fast-moving fields such as computer science to eight years in mathematics.

7.67. Publishers have also noted that the availability of articles via the large subject-based repositories such as ArXiv and PubMedCentral tends to reduce the number of downloads from publishers’ own platforms. That tends to increase the cost-per-download ratio for universities and others who pay for subscriptions for licensed access to the relevant journals via the publisher’s platform; and since that ratio is being used increasingly when universities review the journals to which they subscribe, some publishers are nervous about loss of subscriptions. On the other hand, evidence from the PEER project suggests that providing access to articles via repositories with high-quality metadata may lead to a marginal increase in