Paper presented to the ICSU Press Workshop, Keble College, Oxford, UK, 31 March to 2 April 1998
The electronic submission of papers to learned journals is often presented as if it offered nothing but benefits, both in timing and cost. The comments made by Barbara Kirsop to this Workshop take very much this approach. In fact, there are significant costs and problems to be handled, if the benefits that are readily apparent on first consideration of e-publishing are to be achieved at an acceptable cost, and without loss of the rigorous academic scrutiny that traditional scientific publishing achieves.
Lindsay Ross International has published electronically scientific and other technical information at various levels since late 1995. Over that time the publishing involved has changed from simple HTML available to all Web users, to a mix of HTML for general reference, SGML for indexers, abstractors and searching applications, and PDF for authorised users only. This is similar to the pattern described by Boyce  for the American Astronomical Society in Paris in 1996.
The company has experienced problems that are not generally taken into account in estimates of both timing and costing of electronic publishing. While these have been exacerbated in some cases by the speed of the development programme at Lindsay Ross, they are of general interest because they seem to be underrated in the literature. (See, for example, the description of electronic publishing on the Web given by Wentk , and the comments of Dr Barbara Kirsop). Although the discussion that is given below reflects the experiences of a small company, many of the points made are applicable to publishers of any size.
Challenges to Production
In handling traditional hard-copy submitted papers, a journal will limit the range of languages that will be accepted. Although authors may not be able to write in the required natural language at 'native' levels, humans are able to tolerate a high proportion of grammatical and syntax errors and still accurately assess the underlying scientific merit of the material submitted. This allows the Editorial Office to allocate an accession number to each paper received, and pass it to an editor who will rapidly decide on suitable referees for any given paper, taking account both of linguistic and scientific expertise.
When papers are submitted in electronic form, this poses problems of data conversion from an extremely wide range of word-processing packages. Something like thirty packages are in everyday use and many more are available, and to refuse to accept software that is virtually out of date in a developed country would exclude many authors from the developing world. The editorial production office must therefore cope with the need to convert submissions from many formats and print them out, before accession numbers can be allocated, and appropriate referees can be selected.
In order to allow for the software available to referees, the editor is likely to have to supply electronic papers in recent, but not the latest version of 'standard' software such as Tex or Word. However, the editor will need to have the latest version of this software on at least some of the office computers, so that well-equipped authors and referees are catered for. This imposes both a time delay and a skill requirement on the Editorial staff, as well as the need for a wide range of conversion software and several versions of 'standard' software.
There is clearly a potential benefit in the more rapid transmission of papers for review to and from referees. However, the ability of authors to make virtually endless small changes to a submitted paper means that costs and timings can easily spiral. This can only be controlled by imposing a strict discipline on authors - in effect, forcing them to treat the process rather as if it were still hard-copy based. In addition, it is far easier to differentiate between early drafts of a paper in hard-copy form, and the later, corrected version, than is the case with electronic versions of a paper. This, therefore, demands precise physical and electronic storage and indexing of the versions of a paper, when it is held on a floppy disk or an equivalent storage medium.
Lustig  mentions the danger that peer review may be reduced in coverage, or even given up as a result of EP. Fortunately for standards, he believes this is not likely in the near future - but that was two years ago.
A major benefit of electronic submission is the potential to avoid the need to re-key or scan in printed material (both of these are slow and subject to error, and the resulting document may require editing before use). This benefit goes some way to offset the time required to convert data from the input format to its final form, but the skills demanded to deal with the electronic material are in great demand and are expensive.
The Pattern of Production
Perhaps the most widely perceived benefit from the electronic manufacture and delivery of learned journals is the reduction in the time taken by the total production cycle. In less than three years greater and greater electronic intervention has led to changes in procedures that are examined below.
A conventional production cycle is shown in Figure 1. The process changed very little between 1977 (which marked the demise of hot metal typesetting) and 1995 (when web technology was first used commercially). Using a Physics journal as an example of monthly publication, the cycle took 26 to 28 weeks from manuscript submission to publication. Of this, the peer review process took up 6-7 weeks, or about 25% of the total. The use of email to transmit papers between the editorial office and the referees has typically reduced this lead time by 40 to 50%, down to 3-4 weeks.
The existence of electronic files reduces both the time and the cost of typesetting, although these savings have to be offset against the need to provide SGML (Standard Generalised Mark-up Language) tagging during typesetting, for use by indexers and abstractors who are thus able to interrogate the electronic text. Web delivery of the journal, with its SGML content, is also a costly process, which is considered under 'Challenges in Viability' below.
The use of illustrative matter within journal articles, especially halftones or colour, further complicates the electronic delivery procedures, and while time benefits may still accrue, so too will costs. Inevitably there will be continuing pressure to 'evolve' e-journals away from their paper forbears as video clips, three dimensional illustrative matter and sound are incorporated into papers, and this will have a further impact on costs.
Even though the paper medium for delivery continues to predominate, opportunities for shortening the production cycle are still growing. An example is the ability to distribute a journal electronically, and print in the target countries, rather than to print in one place and distribute physically. This avoids costly intercontinental mailings and provides a speedy delivery cycle to the end user, which is even of benefit for journals with a small subscriber base. However, as costs are saved in one area (reduced mailing) they accrue in another, in that there will be several sets of start up printing costs, with the management time needed to control them. Figure 2 shows that real savings due to e-publishing continue to be elusive.
Figure 2a, b, c
The production process continues to incur new costs as the need for electronic archiving complements that of paper. This is further complicated in that no formally agreed archiving method exists for e-journals as yet, which makes the process of referencing more complex, and adds to the costs involved.
Challenges in the use of Journals
Even if a library has several copies of a journal, only a limited number of people can use them at the same time. Electronic versions of the same journal can clearly be used by many people at the same time, unless there is a restriction on the number of copies that can be in use at any given time. As the economic pressure on academic and learned institutions grows (not to mention the pressures on commercial publishers!), there is a strong temptation to limit the number of subscriptions to one per institution. e-publishing might appear to offer a most attractive alternative to institutions, by comparison with taking out multiple subscriptions to hard-copy journals, because it seems to offer such easy multiple access. Okerson  gave a very helpful discussion of the position the librarian is in during the Paris conference.
In the longer term, groups of institutions may wish to aggregate their library provision, so allowing them as a group to subscribe to more journals in total. From the publishers' point of view, this might lead to a disastrous reduction in subscription income, and thus cause the demise of many prestigious journals. The most likely way to avoid this is to switch either to a form of 'pay per view' or to some method of licensing that would involve paying a fixed fee for single accesses, and further fees to allow multiple access up to fixed thresholds. This would generate income from every use of a given paper or journal, or create a status quo in which multiple users paid for multiple accesses. Pullinger  has discussed this previously, and will no doubt do so at this Workshop. While this might not appear attractive to publishers at present, data on usage of journals is urgently needed to provide the basis for designing a satisfactory method of funding e-publishing in the future.
Challenges in Viability
In addition to the obvious cost of a server that has to be available at all times, and will therefore require both a 'shadow' server (that may be in use during working hours for other purposes) along with an uninterruptable power supply (UPS); acting as a Web provider implies other costs. First, the network will require rapid digital input/output, which happens to be expensive in the UK. Second, staff require training and experience in operating systems (Unix or NT 4 being the most popular), in languages (HTML, SGML, Perl, Java), in word-processing software (the packages are too numerous to list), and conversion tools, database and specialist software. This takes time, greatly enhances the value of the staff involved and so has salary implications, and leaves companies vulnerable to loss of key personnel.
If learned journals are published electronically, reductions in time from submission to publication are obviously achievable, and the scientific work concerned can be made extremely widely available, but the resultant cost is likely to be significant. All the functions that are involved in publishing a journal in hard-copy form still have to be carried out, and e-publishing requires additional hardware, software and liveware. The first two need to be constantly updated, while the last will require costly salary increases and, frequently, costlier replacements.
There seems to be a feeling in some quarters that e-publishing will bring in an era of cheap but high quality scientific publishing - a sort of 'perpetual motion' phenomenon. For the reasons we have set out above, this is a fallacy. Difficult choices will have to be made as to how e-publishing is to be financed. If scientific information is to be published electronically, while conforming to the rigorous standards of peer review that we take for granted today, then it can only be achieved - as ever in life, for there are no free lunches! - at a price, which the users will have to pay in one way or another.
1 Boyce, P.B., A successful electronic scholarly journal from a small society, in Proc. Electronic Publishing in Science, ICSU Press, Paris, 1996, pp 43-47.
2 Wentk, R., The Which Guide to the Internet, publ. By Which ? Ltd., 1997, p 62.
3 Lustig, H., Electronic publishing: the role of a large scientific society, in Proc. Electronic Publishing in Science, ICSU Press, Paris, 1996, pp 127-130.
4 Okerson, A., A librarian's view of some economic issues in electronic scientific publishing, in Proc. Electronic Publishing in Science, ICSU Press, Paris, 1996, pp 143-150.
5 Pullinger, D.J., Economics and organization of primary scientific publications, in Proc. Electronic Publishing in Science, ICSU Press, Paris, 1996, pp 151-161.