Online Self-Archiving: Distinguishing the Optimal from the Optional

From: Stevan Harnad <>
Date: Tue, 11 May 1999 13:40:11 +0100

This is a reply to Arthur Smith of APS, but first an introduction


The 2/3-year long American Scientist September-Forum Discussion so far
has helped to crystallize a number of features of the online archiving

Before replying to Arthur Smith, below, I will summarize some critical
distinctions that some contributors are failing to make. I think that
when these are made, broad lines of agreement will be seen to emerge.

I think we can all agree on what is optimal now: that the refereed
journal literature should be available online on every
scholar/scientist's desktop for free. (Ask yourself, component by
component, what it would mean to deny this? It is BETTER that it not be
available online? It is BETTER that it not be available free? Save
thoughts about why you may think it might be impossible for later; we
are at the moment asking whether it would be optimal.)

> From the optimality of a free online refereed journal literature, it
follows, on an individual author basis, that publicly self-archiving
one's papers would free the literature. If every author did that today,
the optimal would be here tomorrow.

Having agreed on what is optimal, the rest is about how to get there
from here (and whether it is possible at all). Some steps towards the
optimal have face-validity: they are in and of themselves part of the
solution. Self-archiving authors' own refereed papers now is such a step.

But other steps are optional: (a) Individual authors can decide case by
case whether or not to self-archive unrefereed papers too. (b) Readers
can decide whether they are satisfied with the author's self-archived
versions, or they still want to keep using the paper or online
page-image versions too. (c) Institutions can decide whether or not to
continue purchasing the toll-based S/L/P versions of papers even after
they are also available free. (d) Publishers can decide whether to keep
producing and selling the S/L/P versions to readers' institutions as
their product, or to instead sell quality control and certification to
authors' institutions as their service.

Still other steps are provisional, and conditional on the outcome of
the above options (author self-archiving may be supplemented by official
publisher overlays on the archive; institutional archiving may be
subsumed by a global virtual archive; indispensable commercial
"add-ons" might be created that everyone wants to pay for). Many
contingencies are possible. No one knows exactly what shape the online
literature will take eventually.

For my own part, I am prepared to make some predictions, but they are
only predictions:

(1) Readers will self-archive all their papers online (with the
occasional exception of those unpublished papers that some may
sometimes wish to withhold for various reasons till they are accepted
for publication, if they ever are; then they will be self-archived

(2) Readers will overwhelmingly prefer to use the free online

(3) The S/L/P market will accordingly shrink radically.

(4) Publishers will prefer to scale down to providing quality control
only, funded by author-institution-end page charges.

For prediction (1) (author self-archiving) to fail, readers would have
to fail to understand the relationship between self-archiving and the
free online journal literature they want, or they would have to fail to
want it. I doubt the latter (but if it were so, there would be nothing
to be done); as to the former, I and others will work indefatigably to
prevent anyone from failing to understand it.

The overwhelming empirical evidence in support of prediction (1) from
the Physics Archive is that authors will indeed understand and
self-archive, once the archive is made available to them.
See: <>

For prediction (2) (reader preference for the free archive) to fail,
authors would have to understand that univeral self-archiving provides
the free archive, and go ahead and do it, but readers would have to
fail to use it.

Again, the empirical evidence from physics is that readers will become
quickly and overwhelmingly addicted to using the free archive.
See: <>

For prediction (3) (S/L/P cancellation) to fail, libraries would have
to continue to want to pay for S/L/P despite (2).

So far, there has been no sign of significant S/L/P cancellations
because of the Physics Archive. This may mean cancellations will never
come, and that this prediction is wrong, or that cancellations will
come more slowly than the change in useage patterns, and may possibly
be awaiting a generalization of the effect from Physics to the rest of
the disciplines.

Note that (1) and (2) are in NO WAY conditional on the outcome of
prediction (3). The optimal resource would be there, and in full use,
whether or not (3) happens.

For prediction (4) (transition to page-charges) to fail, prediction (3)
would have to fail. Again, (1) and (2) in no way depend on the outcome,
whichever it proves to be.

Now back to quote/comment mode:

Arthur Smith <apsmith_at_APS.ORG>, Mon, 10 May 1999 distinguishes 3
alternatives, whereas in reality there are only two. What he describes
below as the "Harnadian" (II) vs. the "Ginspargian" (III) system are in
reality IDENTICAL! It is only his interpretation that is different.

System I, the traditional system, is one in which the journal
literature is available only via S/L/P tolls. Systems II and III
are identical. They simply make the same literature available online
for free. The latter represents the optimum, as described above. The rest
is optional, provisional or conditional. And no one knows what the
eventual outcome will be.

> Let's consider 3 systems of payment for the peer review,
> selection, and presentation work that is what journals do:
> (I) The traditional system, where the author pays (let's say)
> nothing, transfers full copyright to the journal, and the reader pays
> all costs which can be artificially inflated because the journal
> has monopoly control over the article.
> (II) The "Harnadian" system, where the author or author's institution
> pays all costs, and readers have free access to all the literature.

No, the "Harnadian" system, is simply authors self-archiving to create
a free online archive. The page-charge model is a conditional
prediction on which the self-archiving initiative in no way depends. If
I am wrong about the market, then page-charges will not be necessary,
and a parallel S/L/P edition of the literature will continue to exist.


> (III) The "Ginspargian" system, which is the current situation in
> physics: the author pays nothing or a nominal page charge, transfers
> partial copyright to the journal, and the reader pays all journal costs
> which CANNOT be artificially inflated because the journal no longer has
> monopoly control: the author has also (at least optionally) posted the
> article to a free preprint archive. (Apologies to Paul Ginsparg if he
> doesn't actually believe in this system.)

No, the "Ginspargian" system is likewise simply authors self-archiving
to create a free online archive. Paul and I have differed in our
emphasis on local vs. global archiving, and on the archiving of
preprints vs. reprints, but the "system" is exactly the same.

(The real difference, of course, and historians will take due note of
this, is that whereas I simply preached and polemicized, Paul actually
DID it, creating a self-archiving resource that was astoundingly
successful, both providing and proving the way for us all. That is why I
always characterize myself as merely playing John the Baptist to
Ginsparg's Messiah.)

> The Harnadian system (II) solves the problem from the "author service"
> perspective, that readers run into barriers trying to read an article.
> This can happen in (I) and (III) where they, or their institution,
> cannot afford either a subscription or the specific one-time
> pay-per-view cost

Incorrect. Once every physicist is putting every published paper in the
Physics archive, there is no barrier with III.

> (note that "barrier" is different from "totally
> inaccessible" - everything is available through interlibrary loans, or
> through the Copyright Clearance Center, although these generally
> constitute time and effort barriers, as well as monetary barriers).

Note that the L and the P in S/L/P stand for Site-License and
Pay-Per-View, and that, exactly like Subscription (S), they constitute
a financial firewall separating the literature from its readership.
There is only paid access. (The fact that our institutions "subsidize"
our access through S/L/P should not be allowed to obscure the fact that
access is completely conditional on paying out the cash, and on the
reader-end, whether individually or collectively.)

> However, note that under the Ginspargian system (III) readers can also
> get fee-less access to every article via eprint archives.

Indeed, and this is why II and III are identical. The difference you
adduce is just in long-term predictions, on which the self-archiving
principle in no way depends.

> And in reality, researchers wanting to read an article can generally
> get a free copy out of the author one way or another under any system
> (how many of us have fulfilled requests from developing countries, or
> even smaller colleges in the U.S., for reprints of our work?)

True, but what is your point? The online archive simply allows us to
provide free reprints to anyone and everyone forever. Our history of
taking pain to provide free copies the old, hard, expensive way should
be taken as further evidence for the optimality of free

> But there is a serious problem with the Harnadian system from the
> "reader service" perspective: some authors or their institutions, and
> therefore some critically important work, may not be able to afford the
> monetary or other barriers associated with an author-pays system.
> Did Einstein have an "institution" or the means to pay substantial
> author charges in 1905?

Please! Are you saying that Einstein today would not have had access to
the Web, to self-archive in LANL?

I don't think this is an argument against page-charges either (how much
will it really cost? how many noninstitutional authors are there? will
it not be cheap enough to make the obvious solution a slush fund for
disenfranchised authors?).

But this is a false issue: It quarrels with prediction (4), above,
which is in turn conditional on prediction (3). These may or may not
turn out to be correct, but self-archiving should nevertheless proceed
apace, as its optimality in no way depends on the outcome. In is an end
in itself.

> Now it has been proposed that authors or institutions who cannot pay
> under the Harnadian system could receive subsidies of some sort.
> Fine - but many readers who cannot pay under systems I or III
> are also eligible for various subsidies, developing country journal
> discounts, free back-copies of journals, etc. Why would the subsidies
> be more effective under the Harnadian system?

Again mixing the optimal with the optional/conditional. Never mind.
Let's self-archive and let the market decide the rest. With a free
archive, there can't be any real losers (among authors, readers, and
their institutions).

> Both Systems II and III of course allow authors to freely post their
> work on preprint servers. The problem from the "reader service"
> perspective under the Harnadian system is not access - it's that the
> journals (or whatever replaces them) are not serving the
> reader-oriented role of providing a prominent, official, authoritative
> place for important work. The enormous mass of "raw" author-provided
> material is what makes the journals worth money to readers, under the
> Ginspargian system.

This is only if we blur the distinction between self-archived (i)
refereed papers, (ii) unrefereed papers, and (iii) the rest of the
stuff on the Web. For the time being, clear self-tagging will be
enough to sort these; later on we can worry about making the
classification more formal and reliable. The simple truth latent in
this is that (i) corresponds closely enough to the current refereed
journal literature -- minus some page-based frills. The approximation
can and will be tightened as useage grows.

And if the approximation is never close enough to put an end to all
demand for the enhanced S/L/P version, and hence scaling down to page
charges never becomes necessary, that is perfectly fine with me! My
mission is not to downsize journal publishers but to free the refereed
literature! If one can happen without the other, so be it!

> Also note that access is improving even under the traditional
> publication system, system I. Whether prices are improving is another
> matter, but electronic delivery on a pay-per-view basis should present
> a lower barrier for readers than a trip to the library, search of the
> stacks, and session at the photocopy machine, or especially relative to
> a $20-$30 document delivery fee and a two-week wait.
> And I think we'll see pay-per-view prices come down as electronic
> payment systems improve. At least where competition is available
> (the Ginspargian system).

Fine. I'm still betting that free will beat fee. But if some will
always prefer what fees can by, there is nothing whatever wrong with

> Now, let's turn to the question of how a journal gets paid.
> I have argued earlier that, at least under systems II and III
> the total global cost is not going to be much different. Costs
> certainly should drop getting out of system I, and I think opening
> up preprint archives to other disciplines is a great idea. But
> it's already there in physics.
> Anyway, Harnad does not agree on system II/III costs being equal:
> sh> But that is not the only issue: As long as one is creating a product
> sh> with more features wrapped into it than necessary, the costs will be
> sh> higher than necessary. Quality control is the only essential service
> sh> that learned journal publishers will perform in the online era.
> sh> That is an author-end service.
> But a "quality-controlled" literature serves both authors and readers.
> And why should we deny those who read journals the opportunity
> to request enhanced services? As long as there is competition
> for access to specific articles (the Ginspargian system) let
> the market decide what is "necessary"!

You are absolutely right. I never contested this. (I simply predict that
there will be no market left once the free version is available. And I
don't mind in the least turning out to be wrong in that prediction. The
mission is to free the literature. The economic future-casting is just
amateurish speculation on my part.)

> And on payment, a practical problem crops up - how do you enforce the
> Harnadian system? With I and III it's easy: you don't allow access to
> the journal to readers who don't pay. With author-pays, especially if
> submission charges are not allowed, the journal has already sunk its
> costs in reviewing the article. How do you not reveal the fact that the
> article was accepted while expecting the author to send the publication
> charge? How do you prevent an author from not paying but then posting
> "Accepted for publication in the Journal of ..." along with the article
> on a preprint server?

Arthur, are you serious? Do you think you are dealing with the bootleg
video market here? I edit a journal with an impact factor of 15. If I
and my referees invest our time and effort into refereeing a paper,
recommending revision, refereeing the revision, etc. etc. and then
finally accept the paper, and the author withdraws it after all that
and posts it as "Accepted for publication in...", do you have any
doubts about what the outcome would be? The author would be blackballed
from any further submissions to the journal, and everyone in the field
would know of it, including the referees (who refereed it for free, and
will never waste their efforts on work by that author again).

Please, keep the counterexamples realistic for this specific
constituency and service, otherwise you cannot hope to have them taken
seriously. (A delinquent author is not like a delinquent subscriber!
This is reader-end thinking again...)

> How do you deal with authors and institutions who
> are clearly well-funded but claim poverty (as many do now on page
> charges - I wouldn't want to name a certain major institution near Lake
> Michigan would I? But maybe they've reformed since a decade ago when I
> first heard complaints.)

I think the costs of quality control will be so low, and institutions
will be so much better off without their S/L/P expenses (and
publication pressure and benefits will continue to be so high, if
another incentive was wanted) that most will have no trouble or
hesitation about covering up-front costs. But the self-archiving
initiative in no way depends on my oracular powers in this regard.

> However, there is one area where the Harnadian system improves things -
> institutional payments (if they can be gotten out of the institutions)
> are fairer than under any current reader-pays system. By basing it on
> number of published articles, the payment becomes more closely
> proportional to actual research activity. Some such restructuring of
> payment based on institutional size or activity makes a lot of sense,
> but the Harnadian system is not necessarily the fairest. For example,
> it penalizes research institutions (particularly national laboratories)
> to the benefit of teaching-oriented institutions which may still make
> heavy use of current and archival journals in their instruction. The
> Harnadian system also makes the payment schedule very uneven for a
> small institution with widely variable numbers of publications each
> year. That was my thinking when I got into the following earlier
> discussion:

All depends on how much the no-frills service of quality control will
really cost. Let's just concentrate on freeing the literature for now,

> > > A start towards this would be simply to change our pricing structure
> > > to scale prices in some way according to the number of authors or
> > > researchers at an institution - but we really don't want to
> > > go to direct usage based pricing, as that (institutionalized
> > > pay-per-view in a sense) discourages use, which is the opposite of what
> > > we all want. So some kind of "number of authors" based price makes a
> > > lot of sense.
> >
> > This has become a Mobius strip! What on earth does "pay per view,"
> > which is a reader-end concept (the "P" in S/L/P), have to do with an
> > author-end service?
> I was simply thinking of how we could perhaps move to the fairer cost
> structure of the Harnadian system whether or not we actually make the
> journals free to all readers. Since the institution pays (not the
> author) either way according to Harnad, it really shouldn't matter how
> the journal figures out the charges, should it? As long as we have the
> competition of the Ginspargian system, the charge is not going to be
> excessive.

I can't really see why quality-control costs should be scaled to
institutions, with lower costs per paper to more research-active
institutions. Let's cross that bridge when we know we need to, and know
what the real residual costs then will be. For now, let's just

> > This is all reader-end, access-toll-based thinking: The tolls need to be
> > paid, or the papers are simply not accessible.
> No - they are accessible, even under the traditional system I, and much
> more so under the Ginspargian system III. Just because they aren't
> accessible freely from journal-controlled systems doesn't mean they are
> not accessible.

I completely agree that once the entire journal corpus is available
free, it will be available free.

> > [... on costs being artificially high ...]
> > I have no paranoia whatsoever regarding APS, for example. I
> > simply believe that you need to be protected from yourselves: As long as
> > you persist in thinking in terms of providing a reader-end product,
> > your "g" will be needlessly (even if unwittingly) inflated by expenses
> > from inessentials wrapped into that product.
> Who will police costs under the Harnadian system? Or will institutions
> only allow their researchers to submit articles to journals with the
> lowest bid price? If prestige and reputation determines where articles
> get sent, then what is to keep the "best" journals from having
> artificially high author charges, and if it is not prestige and
> reputation, then on what will this competition between journals be
> based?

It will be based on the quality of each journal's peer review,
authorship, contents, impact factor. Journals will attract both their
authorship and their referee/editorship by their quality, just as they
do now. I doubt that the better journals will want or need to charge
more, but I would really rather not second-guess these contingencies.
They will sort themselves out. It is self-archiving that we need now.

> As long as the journal monopoly is gone (the Ginspargian system, and of
> course that was Harnad's original subversive proposal) there should be
> very few "needless inessentials" that journals waste money on.

If journals can scale down S/L/P so that it can survive in co-existence
with a free archive, without the need to switch cost-recovery to
page-charges, I will still be at peace in my grave. The free literature
is fee enough for me.

Stevan Harnad
Professor of Cognitive Science
Department of Electronics and phone: +44 1703 592-582
Computer Science fax: +44 1703 592-865
University of Southampton
Highfield, Southampton
Received on Wed Feb 10 1999 - 19:17:43 GMT

This archive was generated by hypermail 2.3.0 : Fri Dec 10 2010 - 19:45:32 GMT