CHAPTER NINE

ELECTRONIC PUBLISHING

Scholarly activity ordinarily culminates in publication (literally, the act of making public), the communication to colleagues and students of results, observations, and interpretations emerging from one's research. It is at this point that libraries have traditionally entered the process; they have collected and classified printed products published by academic and commercial publishers that serve as vehicles of scholarly communication. But libraries are, of course, indispensable to scholarship from its very inception. They give scholars access to past work and thus make future contributions possible. Those contributions represent not only a culmination, in that they communicate results of work brought to the point where publication is deemed warranted, but also a beginning, in that they furnish material that supports new scholarly ventures and stimulates new analyses.

PRIMARY AND SECONDARY TEXTS: THE GENBANK EXAMPLE

One characteristic of electronic information technologies is that they establish a different relationship between interpretive works and the underlying data or primary texts on which they are based. Consider the difference between the more traditional approach to assembling and publishing data and the electronic procedures of GenBank, the national repository for nucleic acid sequence data.[1] In the past scholars and scientists gathered data in an attempt to test a particular hypothesis. When sufficiently important results were obtained, a paper describing them would be published and the authors would typically include some of the data substantiating their conclusions. After electronic databases such as GenBank were established, the data contained in such articles would then be extracted and stored electronically. There were limitations inherent in such a model, however: the period between the submission of the manuscript of an article and its publication was inordinately long; and as the volume of data increased, publishers were understandably reluctant to include anything more than excerpts, especially given that sequence data in printed form was of limited usefulness.

These features of the current model argued for different procedures altogether. It was proposed that the full range of substantiating data be made available electronically at the same time that a paper presenting conclusions based on them appeared in print. The scientists responsible for maintaining GenBank have secured the cooperation of journal editors so that submission of the data to GenBank would be a necessary condition to publication of a paper based on them. Authors are required, further, to submit data in "machine-parsable" form. The direction of flow of the data between published articles and the databank has therefore reversed. Authors are now able simply to excerpt or cite data from the database. Readers are able to retrieve the underlying base data at the same time that they receive the articles themselves.

Such a model might apply to humanistic scholarship and scholarly communication as well. Many humanists would be as interested as scientists in having ready access to the full range of primary material underlying scholarly arguments (in this instance, of course, the material is ordinarily different in kind, usually text rather than data). In historical disciplines, for example, one can distinguish between the sources---contemporary chronicles and narrative accounts, letters, diaries, works of art, literature, and music, debit-credit registers, data on demographic trends, government statutes, and so on---and analyses that make use of such material and attempt to package and interpret it in particular ways. It is in the comprehensive assembling of the primary material that electronic information technologies are especially flexible and powerful tools, in part because "the facts are never all in," as Harlan Cleveland has said, and one therefore wants to be able to assemble material in a way that is appropriate to the dynamic quality of scholarship, and in part because the new technologies permit one to search the assembled primary material with ease and reorder and reassemble it in ways appropriate to one's purposes.

In the humanities, as in the sciences, many publishers are increasingly unwilling to print lengthy original source material because it is so costly to do so. "[P]rinting costs," wrote the late Eric Cochrane of the University of Chicago, "have all but extinguished the four-century-old tradition of European text editing."[2]

These developments highlight the virtues and limitations of each medium and the appropriateness of particular media to particular material. Electronic media are better suited to storing underlying raw material, whether data or texts. Print, in contrast, is better suited to presenting works of synthesis and interpretation that make use of the raw material. As Henry Riecken has suggested, "It may be no more sensible to publish a commentary on the Miller's Tale in electronic format than to embalm in a printed work the data from the Current Population Survey.... The text of the Federalist Papers was put into machine-readable form in order to carry out an analysis that resolved questions of disputed authorship of some of the papers; but the new format did not replace the bound volume for readers who want to absorb the thoughts and reflect on the aspirations of this stately document."[3]

The example of GenBank suggests that, at least in some scientific disciplines, printed journal articles need not contain the documentation for the argument they advance. A more appropriate means to handle documentation might be through reference to the contents of the dynamic electronic database, thus consigning the type of material to the medium that handles it best. One advantage of this approach is that it may lead to shorter, less expensive print products that contain little or no documentation. Storing the raw material in electronic form, conversely, would in principle permit individual readers to manipulate it in ways that print clearly does not. The GenBank model of publication also avoids what might be characterized as the compromises and half measures typical of the traditional vehicles of scholarly communication---printed articles that contain the scholarly argument in the main text with brief excerpts from the underlying documentation in footnotes, tables, and appendixes.

But the GenBank example assumes that secondary works will appear in print. The new technologies will also permit a fully integrated process in which works of synthesis and interpretation are also published electronically and linked to underlying base data or texts through flexible, sophisticated hypertext and windowing functions,[4] which offer considerable advantages over print in their ability to connect primary and secondary material.[5]

To be sure, various kinds of intellectual material (we purposely avoid the word "information") have distinctive characteristics and purposes. But that need not imply that literary or philosophical writing or scholarly works of analysis and interpretation should be exempt from electronic publication in the first instance. It is rather that the arguments for electronic publication of such works are not the same as for substantiating data or documentation. Print continues to be the preferred medium for certain purposes: the printed page is still more readable than a computer screen; a book is more transportable than a portable computer; a subtle argument that unfolds over many "pages" is often more accessible in print than on a computer screen.[6] These are arguments for a print option at some point in the process of scholarly communication because of the advantages that print affords; they are not necessarily arguments for the production of texts, whatever the genre or intellectual purpose, in printed form in the first instance. Even in the case of a commentary on the Miller's Tale, there are advantages to having the text stored electronically, in that a reader can instantly locate a passage in the commentary of particular interest using a search-engine. Electronic media may also be more appropriate media of distribution, with respect to economic cost-effectiveness, in very specialized fields where monographs can, at best, be expected to sell just a few hundred copies.

The availability of text in electronic form has important potential advantages, moreover, for purposes of resource sharing among institutions, in that texts stored electronically are almost instantaneously transmittable. One of the main impediments to resource sharing---the cost and perceived inefficiency of traditional interlibrary loan services---is therefore resolvable, at least in principle. Current interlibrary services, which have been described as "a bulky process in which a dozen people [labor to bring] forth a mouse,"[7] could be completely transformed. A reader at a university library in California might be able to review, on a read-only basis, the first few panels of a text maintained electronically at a library in Massachusetts and decide whether to request that the text be downloaded and printed. A transaction that may take weeks under present circumstances could take seconds, and the reader could choose not to request the text at all after having reviewed a small portion of it.

Electronic storage of text also permits the reader or end-user to tailor universal information to individual needs. McGraw-Hill, Inc., has developed a program that permits faculty members at the University of California at San Diego to design their own textbooks;[8] they can search an online catalog from a computer in the university's bookstore for materials in McGraw-Hill's Primis database, which contains the full texts of books, journal articles, and so on. The on-site publishing center compiles the materials, creates a title page, and adds an index, table of contents, and page numbers; the entire process can be completed in 48 hours. McGraw-Hill has been granted permission from copyright holders to reprint all materials in the database. Because the texts are created and "published" on demand, faculty members need not anticipate class enrollments weeks or months in advance to order books. More important, faculty members can design texts more precisely suited to their pedagogical purposes.[9] The economic implication of such a capability, as one economist observed, is that "[l]argely because of the changes in costs due to electronics, the size of the minimum press run for a title has become smaller. Indeed, McGraw-Hill publishes textbooks on demand from a database of articles or chapters.... The minimum press run is approaching one. The number of titles produced grows in part as titles for narrower audiences become economic."[10]

Electronic publication, finally, has the virtue that it is the only existing medium appropriate for publishing material of certain types, as Jerome Yavarkovsky, director of the New York State Library, has observed:

In some instances, research results are not published by conventional, printed means because the results can't be printed and still be meaningful. This is true, for example, when the results are three dimensional, graphic, moving simulations, or animations, or when the outputs are dynamic visual representations of variable processes or theoretical constructs. Traditional, printed publication is completely inadequate for disseminating research of this kind. Yet, this research should be included in what we refer to as "the literature."[11]

Because few electronic journals presently exist, no one can confidently predict what new paradigms of scholarly communication in the electronic age will eventually emerge. Our objective in this section is simply to offer a few suggestions as to how scholarly publication might work.[12] We would do well to heed the challenge issued by Ann Okerson, director of the Office of Scientific and Academic Publishing at the Association of Research Libraries:

It is critical that in starting virtually "from scratch" with a brand new "making public" vehicle, we are unfettered by old modes of viewing and doing publishing: by existing notions of publishing offices; marginal cost structure of publishing; the idea of "circulation;" indexing and abstracting; "monographs" and "serials;" advertising; ownership; possibly even profits. We have the opportunity to begin with a blank page---even that notion needs a new metaphor.[13]

ELECTRONIC PRODUCTION AND DISTRIBUTION

Assuming that the principal current document formats---monographs and periodicals---will continue to exist in an electronic environment, at least for a time, the new technologies will no doubt first be applied to the tasks of producing and distributing traditional kinds of documents.[14] Let us begin with production.

When a scholar finishes the manuscript of a journal article and sends it to a publisher, it is ordinarily subjected to editorial and peer review. Some observers have suggested that the traditional editorial and peer review processes might be circumvented altogether because electronic information technologies so facilitate communication among scholars. In Henry Riecken's words, "[a] jaundiced view might hold that desktop publishing suffers from the same blight as computer network `bulletin boards'---anyone with the equipment can `publish' whatever he wants, of whatever merit or interest, and the potential audience is left with the task of sifting through the midden." In our view, however, as in Riecken's, the peer review process is so fundamental to scholarly practice that it will continue to be critically important, whatever the medium of distribution.[15] However, the new technologies may well expand the review process and change its character, with preliminary versions of manuscripts being made available on a network for comment by interested readers. The author would then prepare a final version, incorporating any suggestions that seem to have particular merit.[16] Even if this expanded process does not develop, the traditional one is greatly expedited in an electronic environment. Editors can send manuscripts to reviewers and receive responses much more quickly, thus reducing the period between the time of the manuscript's submission and editorial decisions.[17]

The manuscript then goes into production, and here electronic technologies have already been effectively employed. If the author has submitted his or her text in electronic form, the cost of original typesetting is considerably lower, since the compositor's task is simply to edit the electronic version of the text, and original keying of characters is kept to a minimum.[18] Both author and editor can have changes made readily, using global search and change functions.

Whether the new technologies will result in other significant production cost savings is difficult to determine at this point. Many of the principal costs have less to do with the medium than with the nature of the activity. For example, many of the costs associated with the highly labor-intensive editorial and peer review processes will remain, while some costs associated with the actual production of the text---typesetting, in particular---can almost certainly be reduced.[19]

Once author and editor have settled upon a final version of the text, it is distributed. Traditionally, distribution has consisted of shipping or mailing printed and bound copies of the first copy to individuals, vendors, and libraries. In an electronic environment there might be a variety of distribution media.[20] Publishers might continue to make hard copy available on demand from their own printers to clients who do not yet have the means to make use of electronic versions of texts (or prefer paper). Alternatively, they might issue titles on CD-ROM or floppy disc; these have an advantage over printed volumes in that they miniaturize the text, so that the considerable space problems libraries face might be addressed by this kind of distribution.[21] Such products might either be sold to the user, whether individual or institutional, or leased. In either case the option to download and print the text or any portion thereof might be controlled by the sales or rental agreement. If the CD-ROM or floppy disc is sold or leased to an institution, the publisher might also specify in the agreement that the text is not to be mounted on a local-area network, so that more than one reader can have access to it simultaneously.[22] In any of these cases the cost of printing, if it is permitted, could be the responsibility of the individual user, just as now the individual ordinarily incurs the cost of photocopying printed material in a library's collection.[23]

The option that appears to be of unusual interest is that texts will be mounted on a local-area or wide-area network and clients provided with direct access to them online. Under such circumstances one can envision a variety of retrieval architectures. Individual institutions---colleges and universities---might choose to maintain local electronic repositories of frequently used titles they had purchased outright from the publishers; in that case the cost of the telecommunications connection with the publishers would be eliminated.[24] Alternatively, some publishers might choose not to sell titles but rather maintain them at central sites themselves and charge users on a fee-per-use basis. Yet a third kind of architecture may eventually evolve in which libraries at different institutions collaborate in collecting electronic materials; under those circumstances the consortium will own a full complement of resources but will distribute them among the different members of the consortium. Malcolm Getz has described an architecture of this type as "decentralized ... with multiple, autonomous nodes subject to a standard protocol that allows participants to search multiple sites easily."[25]

To what extent any of these methods of distribution will result in cost savings is difficult to determine now with any precision. Any of them could have significant positive implications for libraries' space problems, as we have suggested, although there will continue to be substantial costs associated with storage, especially if texts are maintained in online databases. William Y. Arms suggests that "[t]oday, storing a document on a computer is more expensive than on paper, but prices are falling rapidly. By the end of the decade, online computing will be much cheaper than storing books on library shelves."[26]

There are other costs entailed in distribution. Let us consider, for example, a journal published by a university press that is currently mailed to subscribers; if the texts of the articles are instead mounted at the university's BITNET node and subscribers elsewhere retrieve them from the network, the cost of mailing, currently recovered in the subscription price, is eliminated, as are the costs of printing and binding on the publisher's part. On the other hand, both the university where the journal is produced and the university where it is received incur costs associated with the operation of the computing and telecommunications facilities supporting the electronic distribution; such costs are ordinarily hidden, at least from the individual subscriber, yet they nonetheless exist.[27] How the costs for various phases of the process of scholarly communication in a fully electronic environment will compare to those incurred now is clearly a subject for considerable further investigation, as is the associated question of who will absorb which costs.

Now let us consider some of the possibilities inherent in these various options. First, individual readers could tailor the information to particular needs. The reader might be able to excerpt and print a single chapter (or page) of a monograph or a single article of a journal issue.[28] Indeed, as several observers have suggested, the very concept of an issue of a journal is challenged by electronic information technologies.[29] Articles are currently collected together and an issue published when there is a sufficient number of articles available to make up an issue. In an electronic environment one can envision a situation where a single article will be mounted on a network as soon as it clears editorial review; there will be no need to wait for other articles to reach the same point in the process. It is in this sense that some of the distinctions entailed by the technology of print invite reconsideration, as Ann Okerson has suggested.[30]

Precisely because electronic technologies permit individual information needs to be met, different subscription and pricing schemes and practices will almost surely develop, whether for individuals or institutions. For example, publishers or vendors would presumably be able to charge on the basis of the amount of material retrieved: the entire text of a monograph or any part thereof, an individual journal article, a page or even a few paragraphs of material downloaded selectively. Case Western Reserve University and IBM are attempting to monitor this process within the university through a software that functions as a kind of royalties manager. On a larger scale CitaDel, the new citation and document-delivery service of the Research Libraries Group, allows for online ordering, delivering, and billing.[31]

Publishers, moreover, may choose to offer various services of other kinds. Although many subscribers will wish to continue to receive routinely all the articles published in a particular journal, it will be possible for publishers to facilitate selective acquisition as well.[32] A comprehensive, fully integrated service could provide tables of contents, lists of article titles, or abstracts of periodical articles or monographs, as INFO-SOUTH now begins to do for Latin American studies. Prospective buyers could skim portions of a full text on a read-only, not-keep basis.[33] Readers could order individual items of interest by means of a simple command; the document could then be delivered by any number of methods.[34]

Some redefinition of the traditional roles of publisher, vendor, and library and of the relationships among them is almost inevitable. Because individual end-users will be able to purchase information tailored specifically to their needs, they will perhaps be more inclined to interact directly with publishers or vendors---that is, readers may not need to rely as much on the library. (One makes certain assumptions about the pricing by publishers and vendors of such services).[35]

Libraries, for their part, may develop a different approach to acquiring materials, more appropriate to the characteristics of electronic technologies. Such an approach has been termed the "just-in-time" model, as contrasted with the "just-in-case" model currently governing libraries' acquisition practices.[36] Because libraries will ultimately have ready access to electronic versions of texts stored elsewhere, they will be able to base acquisition practices on immediate reader needs rather than on a priori assumptions about what those needs might be. Instead of subscribing to a certain number of journals, for example, libraries may instead negotiate contracts for a certain number of discrete articles from a national database of articles, acquired in response to readers' requests.[37]

It also seems clear that there may be some blurring in the distinctions among the historical roles of publishers as producers, vendors as intermediaries, and librarians as archivists. Scholarly publishers have traditionally allowed titles to go out of print because of libraries' willingness to serve as "a sort of secondary distribution system and warehouse," to quote Henry Riecken.[38] The costs associated with archiving materials have been the responsibility of the library. In an electronic environment, where texts can be miniaturized and storage costs greatly reduced, publishers and vendors may be more inclined to maintain archives of texts for many years.

It may also be that traditional models of resource sharing among libraries will be affected. Libraries have historically turned to other libraries for materials not owned locally; if such materials continue to be available from publishers or vendors, libraries might be inclined to acquire them directly, if resources permit and the demand justifies acquisition. Under any of these circumstances, the rapidly changing state of the technology is likely to be an important dynamic. As Jerome Yavarkovsky suggested, "A necessary feature of online accessibility and network publishing is a commitment to archival storage. Once an article is accepted and published online, the publisher, or some other agency, must keep that publication available in perpetuity, in the same way that traditional publications are held permanently by research libraries. This archival responsibility might even be one of the roles possible for research libraries in the future."[39]

CHANGES IN SCHOLARLY ETHOS AND COMMUNICATION

Electronic information technologies may, some observers think, bring about far more significant changes. Indeed, some envision fundamental changes in scholarly ethos and practices of scholarly communication as a result of their introduction. Apart from the expanded processes of peer review described earlier, the new technologies could permit immediate and public response from readers to electronic publications and allow them access to the reactions of prior readers.[40] Further, since their invention centuries ago, footnotes and bibliographic references have connected texts to prior pertinent writings and documented the arguments authors wish to advance. As such, they serve as print equivalents of hypertext, as we have suggested.[41] But the analogy is far from perfect. Footnotes and bibliographic references are always selected samples of the full range of material that might be related to the text at issue. Electronic technologies such as hypertext permit access to a much wider range of relevant materials---both primary and secondary, substantiating and contradictory---than those authors have elected to call to readers' attention. In this new environment readers need not rely on authors' selective presentations of related materials and can, if they choose, seek additional references for themselves using electronic media.

Some have suggested that the technologies may permit the emergence of new institutions or infrastructures supporting scholarly communication. As Patricia Battin has observed, "The advent of electronic capabilities provides the university with the potential for becoming the primary publisher in the scholarly communication process. At the present time, we are in the untenable position of generating knowledge, giving it away to the commercial publisher, and then buying it back for our scholars at increasingly prohibitive prices. The electronic revolution provides the potential for developing university controlled publishing enterprises through scholarly networks supported either by individual institutions or consortia."[42]

How might such a system function? One proponent envisions two broad types of campus network nodes, editorial and distributing. The first of these would offer editorial software for creating and reviewing papers. As demand dictated, it might also store publications held by an editorial node at another campus, a redundancy that would reduce telecommunications costs and network traffic. The second type of node would exist solely to distribute publications by making them accessible online and would resemble traditional libraries by serving as repositories of publications created and issued elsewhere. The distinction expressed in this scenario is familiar and approximates the distinction between the collecting or archiving function currently performed by libraries and the production function performed by university presses. In the new environment librarians might continue to be responsible for decisions about which materials are made directly available to local readers. Access to materials stored elsewhere, however, might well be more directly controlled by the individual reader than is the case under the terms of current interlibrary loan services.[43]

A variety of pricing and compensation schemes, similar to those anticipated earlier, might emerge: a licensing fee paid to a particular publisher might permit a local node to provide access to all the publisher's works; subscription fees might be based on anticipated use of subsets of a node's publications, with users---institutions or individuals---paying for access to parts of the node's holdings; alternatively, access might be on a fee-per-use basis. For individual institutions there will continue to be important and difficult questions to resolve concerning the extent to which the costs of these services are assumed by the institution or passed off to the individual user.[44]

One can envision how the system might operate. Individual scholars interested in accessing particular titles at their own workstations might first determine whether the titles were among those purchased by their institutions and loaded at the local campus network node. If the texts were available locally, the individual readers might be able to access them on a read-only basis, just as now they access printed materials in the university library on a read-only basis. If they chose to download and print portions of the text retrieved, they might be expected to assume some of the cost of doing so, just as now they assume the cost of photocopying printed material. If the text were not available locally, they might be able to scan the first few panels of text on a read-only, not-keep basis by means of a telecommunications connection, either with the publisher's (editorial) node or with the distributing node at another campus within the consortium (in this instance, the local library or its descendant institution might continue to act as mediator in the process). If they decided, once again, to retrieve and print all or part of the text, they might be charged according to whether the title was acquired directly from the publisher or from a fellow member of the consortium.

What any of these scenarios might permit is an assertion (or reassertion) of the university's direct role in scholarly communication, as Battin suggested.[45] The commercial publishers entered the market because they offered economies of scale and technical expertise to professional societies interested in publishing conference proceedings and professional papers. To a considerable extent these new technologies may eventually obviate the need to rely so much on the commercial publishers for their expertise, especially since many of those involved in editorial work for learned journals, even those published by commercial publishers, are academics with university appointments; copyright, if it remains in force in anything like its present form, can be held either by the author or the sponsoring university.[46]

ISSUES TO BE RESOLVED: ADULTERATION OF TEXT, FUNCTIONING OF THE NETWORK

Clearly, a number of issues must be successfully addressed and resolved by all the relevant parties before electronic technologies, whatever their virtues, will be seen as preferable to print as a medium of distribution. There is, first, the question of ensuring the authenticity and integrity of the text. "Electronic text," Riecken notes, "is far more susceptible to distortion, adulteration, and mischievous or criminal alteration than printed pages are."[47] Text on CD-ROMs (read-only memory discs) cannot be changed and thereby prevent alterations of the sort that concern Riecken.

The very characteristic of electronic text that permits easy adulteration is in another way a source of concern to publishers, and in what follows here we are anticipating some of the substance of the later discussion on copyright. A technology (photocopying) already exists that provides prospective buyers of a printed media product with an alternative to buying it (although a less satisfactory one). Anyone who would otherwise subscribe to a journal who instead photocopies selected articles of interest takes advantage of an existing technology to avoid subscribing, and the publisher is denied revenue. Publishers are understandably concerned that making texts of scholarly material available electronically will greatly facilitate this kind of circumvention.[48] Publishers of computer software programs, for example, are aware that there are a great many pirate copies of their products.[49] Their response is to charge substantially for every copy legitimately sold to offset the economic effects of illegal copying, just as University Microfilms, Inc. does for photocopies of journal articles.

It will be important, moreover, to refine and extend electronic access functions approximating the kind of intellectual activity occurring when a scholar browses in a library or skims the pages of a learned journal. When the latest issue of a core journal in a particular discipline is published, many scholars at least skim the opening pages of articles outside their own subspecialty to have some sense of what colleagues are writing about generally in the discipline. If one had to retrieve the text of the latest issue from a network, would one be more inclined to retrieve and print only the articles in one's own subspecialty? Would one's own interests become ever narrower as a result? As suggested, a related issue concerns the ability to browse in a library. Some interesting discoveries are made serendipitously; will scholars be able to make such discoveries as easily in an electronic environment?[50]

Illustrative material---black-and-white or color reproductions of works of art or other images, graphs, and charts---presents particular problems, which may explain why until very recently there was not a single peer-reviewed electronic journal in the sciences, where scholars make extensive use of such material.[51] It is certainly technologically possible to offer such material and with fully satisfactory results; the issue is rather that to do so assumes the availability of specific equipment and software.[52] There are not yet universally accepted conventions for representing images digitally, and the inconsistent methods of representing and resolving such information are an added complexity.[53]

Finally, there are many questions to be resolved concerning the functioning of the network and the principles governing scholarly publication therein, some of which have been discussed by Ann Okerson.[54] Among them are issues of availability, affordability, and friendly access; the consequences of a shift from the current subsidization of the network to its eventual commercialization;[55] intellectual standards (the integrity of texts and privacy); underlying cost structure (cost recovery, methods of collecting and distributing revenues, and so on); ownership and copyright practices; and academic culture (incentives or disincentives associated with publishing one's work on the network instead of by traditional means). In the culture of print, the fundamental issues of this type were addressed and resolved many years ago, and paradigms governing scholarly practices and the functioning of the entire system, with all its complexities, have been widely accepted. The new technologies provoke fundamental reconsideration.

The various scenarios sketched here suggest both the potential and flexibility of electronic technologies and also some of their current limitations. The discussion to this point has been relatively abstract; one means of achieving greater specificity is by surveying some of the characteristics of the relatively few electronic journals currently existing[56] and some of the issues libraries encounter in collecting them.[57]

One electronic journal attracting particular attention is The Online Journal of Current Clinical Trials, edited by Edward J. Huth, M.D., former editor of the Annals of Internal Medicine.[58] For our purposes The Online Journal is an especially serviceable example since it typifies some of the issues presented by materials of this sort. It is believed to be the first peer-reviewed electronic journal containing illustrative material.[59] It will, moreover, be available only to subscribers who meet particular hardware and system requirements, at least initially, though eventually subscribers will be able to retrieve it using other systems as well. The contents will be machine-searchable by key words, subject, author, and title; in this instance the distinctive advantages of electronic text are thus exploited. They are further exploited in that texts will be corrected after their initial publication and marked to indicate that they have been so revised. Moreover, readers will be able to customize their subscription, specifying topics of particular interest on which they especially wish to receive articles and to some extent controlling format specifications.

The TULIP project of Elsevier Science Publishers mentioned earlier makes bitmapped images of articles in some of its printed materials-science journals available on the Internet.[60] Elsevier will be responsible for mounting the journals on the network; each of the clients---colleges and universities---will be responsible for determining how it will retrieve them. Materials science journals were selected precisely because they contain graphs, illustrations, and other such matter. Unlike completely electronic texts, which are transmitted in alphanumeric form, the bitmapped images made available through TULIP cannot be altered or customized by users nor are they searchable. In these respects TULIP's capabilities differ markedly from those provided by The Online Journal of Current Clinical Trials. Elsevier does plan, however, to supplement the bitmapped image of a text with searchable bibliographic information: author, title, and so on.

These initiatives present challenges to library staff members and others involved in providing information services to the members of an academic community. Many of the challenges have been discussed in a valuable study by members of the library staff at M.I.T.[61] Among them are issues of selection (including the matter of simply being aware of the existence of particular electronic journals), acquisition (knowing the correct subscription information and successfully placing orders and retrieving and downloading the texts), cataloging (providing readers with adequate access information, including information about the medium, how one subscribes, how one specifies the "location" of electronic materials within the "collection"), access and retrieval (including "adding value," such as the possibility of doing keyword or simple string searches of the ASCII text), indexing of contents, archiving, and so on.

Various initiatives at a number of institutions are likely to yield lessons of value to many others. At Carnegie Mellon University, the Mercury Electronic Library now stores information resources of various types, including the page images of the journals included in Elsevier's TULIP program. These can be read online over the campus network. The key feature of CMU's practical success is the early adoption of a distributed approach to computing---that is, acceptance of diverse types of workstations and formats for the storage of information, as well as of the concept of dispersed storage of information.[62] At Stevens Institute of Technology, students and faculty will be able to retrieve the texts of journals published by Engineering Information, Inc., from the campus network.[63] At Cornell University, readers located at their own workstations will be able to retrieve bibliographic information and abstracts of articles published in twenty of the American Chemical Society's principal journals since 1980, as well as the full, machine-searchable ASCII texts and images of illustrative material: line drawings of apparatus, plots of spectograms, chemical structures, photographs.[64]

More important than the differences among these projects is their common assumption that the roles of the component institutions in the process of scholarly communication are changing and in ways that can be understood only by practical attempts to discover what is effective.