Central vs. Distributed Archives
harnad at ecs.soton.ac.uk
Mon Feb 24 12:46:44 EST 2003
On Mon, 24 Feb 2003, Hugo Fjelsted Alrøe wrote:
> [Thread: http://www.ecs.soton.ac.uk/~harnad/Hypermail/Amsci/0293.html]
> I have noticed that you lately recommend exclusively institutional eprint
> archives and not (inter)disciplinary archives.
> Why is that? What are the reasons for not recommending disciplinary
> archives? As you well know, the most successful archive we have seen
> (arxiv.org) is disciplinary, and there are a few others on the way.
Both institutional self-archiving and central self-archiving are
welcome and valuable contributions to open-access. Moreover, because of
OAI-compliance, they are all interoperable. So the short answer is that
it makes no difference. But there is a bit more:
Strategically, several years ago, I could see no reason why large
central archives like the Physics ArXiv should not subsume all of the
literature, in all disciplines. But gradally two problems become
apparent, along with their solutions:
Problem 1: ArXiv itself, though the biggest, is still growing too
slowly, even in Physics: It is growing linearly, which means it will
still be another decade before we arrive at a year when *all* of that
year's physics publications are self-archived.
Problem 2: The central-archiving of ArXiv was generalizing even more
slowly to other disciplines: CogPrints (at 5+ years), another central
archive, still only has about 1500 papers, compared to ArXiv's (at 11+
Solution 1: The Open Archives Initiative in 1999 provided an
interoperability protocol that effectively made all compliant archives
equivalent, whether they were central or institutional.
Solution 2: What is needed to accelerate self-archiving is an *incentive*,
and it is clear that that incentive is something that is shared by a
researcher and his own institution, not a researcher and his discipline
or a central archive.
The purpose of self-archiving is to maximize the visibility,
accessibility, usage and impact of one's research. In a word, to
maximize research impact. The benefits of research impact are shared by
researchers and their institutions. It is one of the main factors in
determining salaries, promotion tenure, research-funding, prizes and
prestige. These are all shared interests for researchers and their
institutions. They are behind the "publish or perish" injunction. This
means that the institution is not only a natural ally in self-archiving,
but it can even be the provider of the carrot and the stick, as an
extension of exactly the same considerations as those underlying
publish-or-perish: Maximize research impact.
It is for this reason that I think institutional self-archiving holds
greater promise for propelling open-access to critical mass than central
archiving -- or, as the effect is additive, I should really say: than
central archiving alone.
> If I am to guess, you might be thinking that authors can be pressured to
> place their papers in institutional archives by making it a condition in
> their employment contracts, or something similar. This pressure can also be
> applied in at least some kinds of disciplinary archives (such as
> http://orgprints.org), by way of making the condition in the research grant.
> And the motivation is straight forward: what the public pays for should be
> made publicly available.
I agree. And both of these pressures are welcome. But the institutional
self-archiving solution is more general, and pan-disciplinary. It is
easier to create and fill institutional archives (using local carrots and
sticks) than to create a central archive for each discipline and get all
researchers to fill it. Institutional self-archiving also benefits from
a wider institutional interest in making institutional digital output
and holdings (not just refereed research) openly accessible (though I
confess that this double mandate has been a 2-edged sword, also causing
confusion about what the target contents of institutional archives
should be, and thereby slowing rather than hastening the self-archiving
of refereed research output).
I would say that when an institution has adopted a policy of mandatory
self-archiving for all its researchers, it is easier and more general
to also provide the local archives to do it in, rather than to rely
on their being spawned and sustained by some external central entity
for each discipline. The policy is then also a uniform, self-conained and
self-sufficient one, whereas "self-archive somewhere" would have
been too vague and would not fit most disciplines yet (rather the way
"publish in an open-access journal" would be a premature injunction in
most disciplines and specialties today).
Last, there is a link between self-archiving and research assessment:
Institutions and their researchers are assessed in various ways on the
basis of their research output. It hence makes sense to have a uniform
way to assess them digitally. A standardized digital CV is one component
of this, and links to the full-texts of all refereed research is
another. Obviously the links could be to a local institutional archive
or a central one, and if either of these were already universal, either
would serve just as well. But I am betting that the institutional path
is the more widely applicable one, it is self-contained and does not
depend on developments elswhere, and it can be uniformly implemented for
> One possible benefit of (inter)disciplinary archives is that they can better
> support a kind of 'community feeling' (which a journal can also sometimes
> offer), and that this community feeling can help improve research
OAI-interoperability provides exactly the same community feeling, on a
distributed rather than a centralized basis. (Moreover, once a
discipline's total output is all self-archived, the "community" is the
same either way.) Journals offer another kind of community, but that
continues, as we are talking about the self-archiving of refereed
research: that means research that is appearing in a peer-reviewed
The Learned Society for a discipline continues to be a community, but
remember that there is no such Darwinian benefit as doing something "for
the good of the discipline" (any more than there is any such Darwinian
benefit as doing something for the good of the group or the species).
The evolutionary "unit of selection" in research is the researcher, or the
researcher's institution (a kind of kin-selection). Sometimes there are
also research collaborations distributed across institutions. But apart
from that, there is no shared interest with an entity corresponding to
the discipline itself. On the contrary, within a discipline, an element of
competition prevails: for grant funding, for priority, for recognition --
indeed, for research impact itself.
More information about the Jrnlnote