On Sun, 20 May 2007, John Smith wrote:

> I wonder if we can come at this discussion concerning the impact
> of OA on citation counts from another angle?
> 
> Assuming we have a traditional academic article of interest to only a
> few specialists there is a simple upper bound to the number of citations
> it will have no matter how accessible it is.

That is certainly true. It is also true that 10% of articles receive 90%
of the citations. OA will not change that ratio, it will simply allow
the usage and citations of those articles that were not used and cited
because they could not be accessed to rise to what they would have been
if they could have been used and cited.

> Also, the majority of
> specialist academics work in educational institutions where they have
> access to a wide range of paid for sources for their subject.

OA is not for those articles and those users that already have paid
access; it is for those that do not.

No institution can afford paid access to all or most of the 2.5 million
articles published yearly in the world's 24,000 peer-reviewed journals,
and most institutions can only afford access to a small fraction of them.

OA is hence for that large fraction (the complement of the small fraction)
of those articles that most users and most institutions cannot access. The
10% of that fraction that merit 90% of the citations today will benefit
from OA the most, and in proportion to their merit. That increase in
citations also corresponds to an increase in scholarly and scientific
productivity and progress for everyone.

> Therefore any additional citations must mainly come from academics in
> smaller institutions that do not provide access to all relevant titles for
> their subject and/or institutions in the poorer countries of the world.

It is correct that the additional citations will come from academics at
the institutions that cannot afford paid access to the journals in which
the cited articles appeared. It might be the case that the access denial
is concentrated in the smaller institutions and the poorer countries,
but no one knows to what extent that is true, and one can also ask
whether it is relevant. For the OA problem is not just an access problem
but an impact problem. And the research output of even the richest
institutions is losing a large fraction of its potential research impact
because it is inaccessible to the fraction to whom it is inaccessible, 
whether or not that missing fraction is mainly from the smaller, poorer 
institutions.

> Should it not be possible therefore to examine the citers to these
> OA articles where increased citation is claimed and show they include
> academics in smaller institutions or from poorer parts of the world?

Yes, it is possible, and it would be a good idea to test the demography
of access denial and OA impact gain. But, again, one wonders: Why would
one assign this question of demographic detail a high priority at this
time, when the access and impact loss have already been shown to be
highly probable, when the remedy (mandated OA self-archiving) is at hand
and already overdue, and when most of the skepticism about the details
of the OA impact advantage comes from those who have a vested interest
in delaying or deterring OA self-archiving mandates from being adopted?

(It is also true that a portion of the OA impact advantage is a
competitive advantage that will disappear once all articles are OA.
Again, one is inclined to reply: So what?)

This is not just an academic exercise but a call to action to remedy a
remediable practical problem afflicting research and researchers.

> However, even if this were done and positive results found there is
> still another possible explanation. Items published in both paid for and
> free form are indexed in additional indexing services including free
> services like OAIster and CiteSeer. So it may be that it is not the
> availability per se that increases citation but the findability? Those
> who would have had access anyway have an improved chance of finding the
> article. Do we have proof that the additional citers accessed the OA
> version (assuming there is both an OA and paid for version)?

Increased visibility and improved searching are always welcome, but
that is not the OA problem. OAIster's usefulness is limited by the
fact that it only contains the c. 15% of the literature that is being
self-archived spontaneously (i.e., unmandated) today. Citeseer is a
better niche search engine because computer scientists self-archive a
much higher proportion of their research. But the obvious benchmark today
is Google Scholar, which is increasingly covering all cited articles,
whether OA or non-OA. It is in vain that Google Scholar enhances the
visibility of non-OA articles for those would-be users to whom they are
not accessible. Those users could already have accessed the metadata of
those articles from online indices such as Web of Science or PubMed,
only to reach a toll-access barrier when it came to accessing the
inaccessible full-text corresponding to the visible metadata.

> It is possible that my queries above have already been answered. If
> so a reference to the work will suffice as a response.
> 
> I am a supporter of OA but also concerned that it is not falsely
> praised. If it is praised for some advantage and that advantage turns
> out not to be there it will weaken the position of OA proponents.

Accessibility is a necessary (but not a sufficient) condition for usage
and impact. There is no risk that maximising accessibility will fail to
maximise usage and impact. The only barrier between us and 100% OA is
a few keystrokes.

It is appalling that we continue to dither about this; it is analogous to
dithering about putting on (or requiring) seat-belts until we have
made sure that the beneficiaries are not just the small and the poor,
and that seat-belts do not simply make drivers more safety-conscious.

> Even if the apparent citation advantage of OA turns out to be false
> it does not weaken the real advantages of OA. We should not be drawn
> into a time and effort wasting defence of it while there is other work
> to be done to promote OA.

The real advantage of Open Access is Access. The advantage of Access
is Usage and Impact (of which citations are one indicator). The Craig et
al. study has not shown that the OA Impact Advantage is not real. It
has simply pointed out that correlation does not entail causation. Duly
noted. I agree that no time or effort should be spent now trying to
demonstrate causation. The time and effort should be used to provide OA.

Stevan Harnad