Scientific Journal Publications: On the Role of Electronic Preprint Exchange in the Distribution of Scientific Literature

1. Method of research: I was employed as a programmer/analyst specialist at Cornell University Library from January 2001 until August 2003. For almost two years during this time, I was assigned technical administration for supporting daily submissions to arXiv. I acquired first-hand knowledge of the inner workings of the system as both a technical and a social object, and was involved in some of the institutional migration process from Los Alamos National Laboratories to Cornell University Library, including broad examination of system needs, planning new developments and addressing policy issues. The contents of this paper are based on participant observation (and comprehension; Collins, 1985: 171-72). The text draws from my knowledge and experience, and from my understanding of the burning issues debated by advocates of the arXiv model and related initiatives. I draw from personal/ professional communication that includes extensive unrecorded conversation in the workplace and off the record - in person, at meetings and over the phone. This communication is supplemented with notes and email exchanges to follow up on key topics. All statistical data that I have collected and processed are publicly accessible using Internet harvesting protocols and I contacted my correspondents regarding the use of spoken and written remarks and gave them copies of a draft to examine.

2. Arpanet was conceived and planned by Joseph Carl Robnett, ‘Lick’ Licklider and others, for the Information Processing Techniques Office (IPTO) of US defence research, of which Licklider was the first director. An interactive overview of the email and Internet and related histories is available at <http://livinginternet.com>.

3. A trend in the development of the service was to add specialties that cater to interests of HEP, although the growth of the system has stretched to other communities. Archives included as of October 2004 were:

4. See, for example, Bijker (1987) on the theory of invention and Hughes (1985, 1987) on ignoring disciplinary boundaries and on technology as problem-solving systems and means for reordering the world. See also Woolgar (1991) on configuring users and machines.

5. The notion of ‘core-group’ (a group of researchers and research units who actively and interactively engage in a scientific research area) was coined by Collins (1985; see discussion in Collins, 1999). The larger community that I refer to as ‘HEP’ comprises more than one core-group, and much of the literature that is of interest to that community is read across the boundaries of core-groups. This can surely raise a problem when preprints are inspected, but that problem is the same as with formally published works. In both cases uncertainty can arise, but I argue that the delay of formal publications gives arXiv salience for immediate research uses, and more immediate communication channels (consultancy for example) are used to settle uncertainties introduced with preprints.

6. For an insider’s account of the early history of technical implementations for arXiv and its patrons, see Ginsparg at <http://arxiv.org/blurb/blurb.ps.gz>.

7. The section number refers to the online version available at <http://arxiv.org/blurb/blurb.ps.gz>.

8. This section largely follows a historical overview by Heath O’Connell (2002), published with High Energy Physics Libraries Webzine and available at <http://library.cern.ch/ HEPLW/6/papers/3/>. It also draws from discussions with correspondents that helped clarify my own understanding of this and other historical records.

9. Ibid., at <http://library.cern.ch/HEPLW/6/papers/3/#SPIRES%20and%20Internet>.

10. See Donald E. Knuth’s web page at <http://www-cs-faculty.stanford.edu/ ~ knuth/>; also about TeX use at <http://www.tug.org/>.

11. The arXiv operation is all Linux-based using mainly the Perl programming language, and it has until very recently been exclusively Unix/Linux-oriented and expected its patrons to feed only source data from papers. Patrons have always been expected to be competent in scientific computing, which includes an understanding and being up-to-date about the operational details behind digital imaging and image conversion, writing and compiling TeX source, postscript interpretation of documents, the use of font files, document style files and other factors that are at play in the arXiv system.

12. See <http://citebase.eprints.org/cgi-bin/search>.

13. See their site at <www.openarchives.org>. This technology is referred to as aggregation services of digital libraries.

14. On a technical note, it is interesting that, although the development and maintenance of the auto-TeX compiler have contributed to the ‘open-source’ GhostScript developments in accordance with ‘open-source’ ideology, and that the design of arXiv has been all free-ware computing, the arXiv system source has never been made available to the public under any scheme such as the Open Source Initiative (see <www.opensource.org>). One reason for this is that it has never been sufficiently complete as a system to be one single software package, which is transferable (and manageable) as such, but is a compilation of many such packages. See also <http:/ /sourceforge.net> as one example of a large-scale Open Source development organization.

15. See Bijker (1987) and Pinch & Bijker (1984, 1987). But referring simply to a ‘technological system’ already suggests a system of social arrangements (MacKenzie & Wajcman, 1985; Woolgar, 1991).

16. Quantitative biology started posting on arXiv in September 2003. See also the ‘Math Front’ at UC Davis for an entry to all the math specialities that use arXiv at <http://front.math.ucdavis.edu/>.

17. See Ginsparg (2001). He offers a detailed discussion of these figures. They are used in a talk given in 2001 and are clearly ‘ball park’ figures. However, they measure against APS revenues as an example of a professional society disseminating output, against Elsevier as an example of a for-profit trade publisher, and so forth. An average figure was estimated of about US$50,000 per produced scientific paper in HEP. This cost would typically cover the expenses of research: namely, salaries, overhead and experimental equipment, but not the editorial, printing and distribution costs of the final paper. In other words, the output would still need to be published. Then a comparison was made between a pricey trade publisher in the field with US$10,000- 20,000 per paper in revenues, a professional society in physics estimated to generate about US$2000 per paper in revenues, and an operation that takes data feed from an existing print publisher and converts it to Web readable formats, operating at approximately US$100 per paper and more. It was also pointed out that managing the conventional editorial process could never cost less than about US$1000 for each peer-reviewed and published physics paper, given the way this process is generally orchestrated. However, a point was made in personal communication about this particular claim, to the effect that the figure reflects the cost of peer review in physics by the American Physical Society journals, and that the cost of the peer review process can vary greatly, depending on discipline and the type of process.

18. For example, Greg Kuperberg, professor of mathematics at UC Davis, argues that open reviews, such as Math Reviews and Zentralblatt, provide a much more helpful screening and evaluation of the mathematical literature than the anonymous refereeing system is capable of (Kuperberg, 2002).

19. See also Kuperberg (2002). Both Ginsparg and Kuperberg argue that there is a persistent and misleading idealization of the role of formal peer review.

20. Ginsparg (2003) asks if the peer review could be better ‘focused’, as he puts it.

21. The author probed specially for this view among arXiv programmers in a project meeting with them 12 February 2002.

22. See Traweek (1988), chapter 3, on this matter in the HEP community. She speaks at length about the sorting mechanisms and, interestingly, in her observation the young physicists are never made aware of this process. It is often only the ‘bright and blunt bastards’ that catch on to what is expected of them.

23. For an interesting input, see Bachrach et al. (1998).

24. To take some examples, the home of Information Research: An International Electronic Journal, offers a title page of well-known electronic journals that are peer reviewed but free of charge at <http://informationr.net/ir/titlepages.html>. The Public Library of Science (PloS) is another initiative that calls on scientists in biology and medical research to resist the privatization of scientific knowledge (see <www.plos.org>). Trade publishers have been able to establish a monopoly-like situation because university libraries and research institutions have no choice but to subscribe to academic journal publications in order to uphold proper status of their institutions. See also <http://pubmedcentral.org>, an archive of peer-reviewed life science journals that are free of charge. Stevan Harnad is one pioneer who should also be mentioned: he has written numerous papers on the subject (see, for example, Harnad, 1998, 2001). His contributions include among other things the eprints.org project. This project is dedicated to opening access, through so-called distributed author-institution self-archiving and is meant to operate alongside the formal peer review process and to take advantage of reference linking and aggregation services to maximize the distribution of works. He is also the founder of CogPrints.org.

25. Just to give one example, the engineering library at Cornell University cancelled 35 subscriptions for an annual charge of US$59,000 in autumn 2003. According to a statement from the director of collection management, electronic-only format at that point was priced at 115% of the cost of a paper subscription with the same trade publisher.

26. See Kling et al. (2001) for an interesting treatment of the E-Biomed experiment for medical research (a mix of refereed and unrefereed works), which later became a nonprofit scientific publisher of peer-reviewed works as the Public Library of Science (PloS) at <www.plos.org/>, and also PubMed Central, whose service offers free access to already existing life science journals at <www.pubmedcentral.org>.

27. See for example Knorr-Cetina (1999) for comparisons of the cultures of HEP and molecular biology; also Collins (1999) on expert readership, Hilgartner (1997) on the Sokal affair, and Kling et al. (2001) on the failure of the unrefereed exchange service of the E-Biomed proposal.

28. See for example the first three pages of <http://arxiv.org/pdf/hep-ex/0406040> (listing the BABAR Collaboration).

29. Knorr-Cetina (1999). Chapters 7 and 8 discuss the notion of erasing and restoring the individual as an epistemic subject.

30. Somewhat related to this is an interesting discussion about the debate between Thomas Hobbes and the experimentalists over the notion of ‘public’ in public witnessing in the early experimental life (Shapin & Schaffer, 1985). It is clear that high-level competence is mandatory in order to evaluate works, whether by direct inspection or aided by consultation from colleagues.

31. See Hughes (1985, 1987) for an interesting discussion about system builders, ignoring disciplinary boundaries. He also suggests that analytical categories (political, science, technology, economics) should be used sparingly.

32. See for example Bijker (1987), for a theory of invention; also Pinch & Bijker (1984, 1987) on the social construction of facts and artefacts, and MacKenzie & Wajcman (1985) for introductory essays.

33. MacKenzie & Wajcman (1985) talk about building on already existing solutions, and about gradual developments, new combinations and new configurations, for which existing technology is a precondition. Hughes (1987) suggests that systems with high momentum tend to exert soft determinism on other systems, groups and individuals.

34. Star & Griesemer (1989) discuss boundary objects, such as repositories and descriptive objects, and argue that they are plastic enough to adapt to various local needs, but robust enough to maintain common identities across sites. Seen in relationship to methods standardization, they point out that consensus is not a necessary condition for entities to cooperate, or for successful conduct.

35. See for example Woolgar (1991), MacKay & Gillespie (1992) and Pinch & Trocco (2002). See also Bohlin (2004) for a discussion of stabilization processes in the competition between new science communication regimes, as he describes them. Bohlin also points out how his use of the Social Construction of Technology (SCOT) model is atypical, precisely because so much is unsettled; but he importantly points out that it will not be some sort of a system superiority that determines the outcome.

36. A sizable group of authors have problems using TeX when it comes to uploading their source files into an automated TeX compiler at arXiv. Astrophysicists have openly complained about having to compress their figure plots and colour photos ‘more efficiently’ according to the arXiv technical administration. See also an interesting and related discussion in Pinch & Trocco (2002) about how users of music synthesizers mostly use pre-set sounds and do not take advantage of the possibility to develop new sounds with the instrument.

37. Greg Kuperberg, professor of mathematics at UC Davis, leads a group of moderators for mathematics specialities who have shaped this process to their own special needs. Kuperberg also was commissioned to assist with an ongoing research project on the development of the auto-TeX system and the TeX services.

38. It is also very interesting to look more closely at the involvement, over time, of the areas of physics that are remote from HEP, which already indicates a divorce of designers from patrons. Designers shift to operating as a group of socially and culturally distant software engineers in the frustrating situation of attempting to cater to needs of a largely unknown body of customers. It is a well-known situation that leads to all sorts of unpredictable and problematic results. See an interesting sociological perspective in Woolgar (1991).

39. Refer to the following translations of the abbreviations used for the archives in the system:

40. The so-called ‘Journal-Ref’ in arXiv is strictly used for existing journal citations when they become available and those references are either provided by services that automatically harvest them (SLAC) or they are provided manually by the authors themselves. Consequently, the column marked ‘%’ below ‘In print’ in the Appendix (Tables A and B) refers to the percentage of papers in the database that are already published.

41. From phone conversation and email exchange with library personnel at SLAC.

42. From phone conversation, office talk and email exchange with arXiv advocates and personnel.

43. Greg Kuperberg at UC Davis manages this work and he also maintains the UC Davis, Math Front to math papers in arXiv; see their site at <http://front.math.ucdavis.edu/>.

44. From phone conversation, office talk and email exchange with arXiv advocates and personnel.

45. Andrew Odlyzko framed a similar distinction in his paper ‘Peer Review and Non-Peer Review’ between significance and correctness, as he puts it (Odlyzko, 2001b).

46. See also Hagstrom (1965), chapters I and II.

47. See Knorr-Cetina (1981), Latour & Woolgar (1986 [1979]) for discussions of the scientific paper as an endproduct of research and a retrospective construction of that research.

48. An interesting borderline case here, as regards physics, would be the story of the Bogdanov brothers. Apparently they employed the correct language and earned a degree in physics, but left the physics community uncertain as to whether their work was a hoax. One measure is to say that they never worked with anyone and were never in collaboration with known bodies, which is usually questionable in the larger research community. A good entry to the Bogdanov case is a letter from John Baez to the Google newsgroup ‘sci.physics.research’ (23 October 2002), pointing out the possible hoax. It has links to their thesis, curriculum vita and four paper references: see <http://groups.google.com/groups?q = g: thl3105378894d&dq = &hl = en&lr = &ie = UTF-8&safe = off&selm = ap7tq6%24eme%241%40glue.ucr.edu>.

49. From email exchange with an arXiv moderator, June 2003.

50. This example is only one of many promotion mechanisms in academia, which apparently are quite similar from one institution to another, and on both sides of the Atlantic. Such similarity supports the main argument here, that academic status and prestige are obtained in an interdisciplinary social setting. However, there are outer social circles that differ importantly from those identified here. Two of my correspondents pointed out that the peer review process has direct use value, for example in mathematics and some areas of physics. Large numbers of scientists live on the outskirts of the practice, and therefore they rely on bibliographies, citation figures and citation references in their work: for example, teachers in universities and institutes of higher education who teach the science to a certain degree, but are themselves not contributors to the construction of scientific knowledge.

51. See for example Kuperberg (2002) and Guédon (2001).

52. See also Ginsparg (2003) where he talks about a ‘desired signal’ that individual researchers or groups wish to or are required to ‘transmit’.

53. This is of course a playful example, but the general idea is in accordance with the type of peer review that physicists and mathematicians, using arXiv, can and do expect from conventional publications. The content is usually only checked for relevance to ongoing research; that is to say, it is accepted if it is interesting and not obviously wrong.

54. Ginsparg (2003) actually discusses the possibility of gaining better focus to peer review process, and he asks if it is really the case that the ‘desired signal’ can only come from the traditional publication practices that cost US$1000 for each paper.

55. There is likely to be more to this situation. While the lack of motivation can be directly associated with career care-taking, two of my correspondents said they would like to see further studies on differences in rank or in age group, for example, given that younger persons might be more eager to supply an electronic system with a journal reference in hope of attracting more readers through the online system, while established scientists can rely on their reputation and other channels. In addition, there may be some relationship between the fact that the older generations used to have secretaries to typeset their papers and generally take care of things such as getting papers to publishers and polishing curricula vita, while the younger and computer-savvy generations have taken on these tasks themselves.

56. This valuable point was made clear to me by Neil Stephens during a session in which I introduced a draft of this paper to colleagues in the Knowledge, Expertise and Science (KES) group at Cardiff University.

57. See Guédon, 2001: chapter III.

58. See Odlyzko (2001a) for statistics on the use of scientific literature in different formats and discussion about slow and fast changes.

59. See <http://arxiv.org/Stats/au_all.html>.

60. The full set of all 659 tables (35 questions) is available at <www.diglib.org/pubs/scholinfo> (Friedlander, 2002).

61. See also Bohlin (2004) for discussion of this trend in journal policy.

62. A correspondent pointed out a single case in arXiv from November 1997 <http://arXiv.org/hep-th/9711200> that wasn’t formally published for roughly a year, but was the most cited paper in 1998. The author had 26 papers in arXiv dated before that one, and of these 25 published with elite publications, most of them already published at that point. Currently SLAC shows over 3250 citations of this one paper.

63. From communication with Harry Collins. This paper is by P. Astone, D. Babusci, M. Bassan, P. Bonifazi, P. Carelli et al., co-authored 2002, and is available at <http:/ /arxiv.org/gr-qc/0210053>.

64. This example refers to the case of the Bogdanov brothers: see note 48.