Deposit Institutionally, Harvest Centrally

University of Michigan’s digital repository now available through PubMed:

"Researchers who find articles by University of Michigan (UM) authors in PubMed can now directly -- and for free -- link to the full text using Deep Blue, UM’s digital repository, via PubMed’s LinkOut feature. Deep Blue is an online archive that preserves and provides access to UM intellectual and creative work. It is the first institutional repository to provide such links."

Congratulations to the University of Michigan and PubMed for adding this excellent and timely feature (to both PubMed and Michigan's Institutional Repository [IR], Deep Blue)! But why stop there?

The implications are obvious: Central Repositories [CRs] (like PubMed Central and Arxiv and CogPrints) should not be deposited in directly, because that merely complicates and competes with a systematic worldwide policy of depositing all institutional research output in each institution's own, OAI-compliant IR. Institutions are the primary research providers. They have the greatest stake in ensuring that all their own research output is maximally visible, accessible, and usable, thereby maximizing the institution's research impact. Institutions are also the best placed to showcase, monitor and reward the self-archiving of their own research output.

All institutions should mandate that all their research article output must be deposited in their own IR. Research funders (like NIH) should also mandate that all the research article output from the research they fund must be deposited in the fundee's own institution's IR.

Then CRs like PubMed Central as well as indexers like PubMed (or Thompson ISI or Scopus or Google Scholar) can either link to or harvest from the network of interoperable, OAI-compliant IRs.

In this natural way -- "deposit institutionally, harvest centrally" -- all of research output can be systematically made OA. Instead depositing willy-nilly in IRs or CRs will only create confusion and resistance on the part of researchers, who will understandably only wish to do the keystrokes once.

IR software can also help with automatic exports to other OAI-compliant sites where desired, as well as with version control.

Now that the NIH OA self-archiving mandate is imminent, it is all the more important to reformulate it in a way that will scale systematically to all research output worldwide, in all disciplines, rather than leaving it as one non-scaling special case for NIH-funded biomedical research.

And remember that the Web era means distributed content provision and central harvesting, Google-style. It is not, as in paper days, that all the content needs to go in one central physical space.

Swan, A., Needham, P., Probets, S., Muir, A., Oppenheim, C., O’Brien, A., Hardy, R., Rowland, F. and Brown, S. (2005) Developing a model for e-prints and open access journal content in UK further and higher education. Learned Publishing 18 (1). pp. 25-40.
Abstract: A study carried out for the UK Joint Information Systems Committee examined models for the provision of access to material in institutional and subject-based archives and in open access journals. Their relative merits were considered, addressing not only technical concerns but also how e-print provision (by authors) can be achieved – an essential factor for an effective e-print delivery service (for users). A "harvesting" model is recommended, where the metadata of articles deposited in distributed archives are harvested, stored and enhanced by a national service. This model has major advantages over the alternatives of a national centralized service or a completely decentralized one. Options for the implementation of a service based on the harvesting model are presented.

Stevan Harnad
American Scientist Open Access Forum