Scientific discourse is the evolutionary path for science. An autonomous system generates concepts and theories, which are tested and challenged by the scientific community in a discursive infrastructure that depends heavily on scientific journals. Those concepts and theories change and adapt, or they die off.

Today a new electronic environment is testing and challenging the scientific journal, which itself could evolve — or decline — depending on its response to that new condition. Electronic publishing has the potential to transform scientific communication. At its most basic level, the availability of scientific journals on the Internet allows scientists and professionals to access the most current information from their homes and offices. Yet unless those journals exploit the Internet's potential for two-way communication, enabling readers to interact and contribute, they will still be pale imitations of paper journals.

I propose a new model for publishing on the Internet, one that takes advantage of the two-way communication possibilities it affords. I call it "Interactive Publishing." Here is how it works:

• Instead of layout designers, we make use of an automated-formatting software that converts simple ASCII documents into HTML.
• The system supports graphical illustrations and automatically inserts them into the text. Hypertext is also inserted into the articles.
• Instead of a hidebound peer-review system, we use an interactive "vote," in which those with comments and suggestions post them along with the article.
• Instead of a lengthy discussion carried out over a period of months and years as letters are submitted to journals and await publication, we allow anyone to post letters, and allow authors to answer them immediately.

As a model, I have started the Electronic Journal of Cognitive and Brain Sciences (EJCBS), an e-journal that works without editors. As a graduate student in neuroscience, I find the current journal-publishing enterprise rather old-fashioned, and not at all evolutionary. A journal that supports immediate publication seems to be an obvious solution to the time frustrations of publishing in science. The system I designed to do this is simple: It creates a Web page for each document automatically.

The software engine behind the EJCBS is a compiler that automatically converts the text to HTML. I'm working on it in my spare time (when I'm not working on my Ph.D. thesis). The idea originated from the frustration that news groups and mailing lists do not support graphical communication. Although Web pages enable graphical illustrations, they are less interactive than a news group or a mailing list. It became clear that I needed special software that compiles documents and graphics to HTML. I explored the software market as well as Internet software resources, and realized that there is no application that directly supports what I want to do. So I wrote a simple program that converts a text file to an HTML file.

I wanted to create my own e-journal, but the administrative work involved required much more time than I could afford. The idea of automatic publishing occurred to me when I was listening Stevan Harnad's lecture on "Skywriting," a bright metaphor of the aspiration and failure of electronic publication. The main source of resistance to publishing in an e-journal is the general attachment to the printed paper. I realized that the fear of electronic publishing is partially derived from the immaterial nature of electronic media, for which we readers and authors must be compensated by some other features.

Those features can be speed of publication, low cost, interactivity, broad distribution, and innovative quality control. Those features are all integrated into the system I call "interactive publishing."

Interactive publishing, especially in the areas of science and business, allows for bidirectional and rapid communication of ideas, which could bring global changes in the way people understand political, business and scientific concepts. The impact of interactive publishing could be enormous. It redefines concepts of traditional publishing, such as editing, acceptance, reviews and comments, and archives.

Editing

Traditional journal editing and reviewing is a long and costly process, sometimes not finished until the paper is out of date. EJCBS uses autopublishing software. With that software a Web server can make the submitted information immediately available for review and criticism.

EJCBS authors format their own articles. If they know HTML, they can edit their documents according to their preferred style, save it in the local computer and submit the URL. The URL will be linked to the table of contents of the journal. If they don't know HTML, but they are using a standard word-processing program such as WordPerfect, they can use the built-in HTML converter that comes with the word-processing program.

But I don't expect most authors to create an HTML version of their papers, so I encourage them to submit their manuscript in ASCII, and I have written a program that automatically converts their plain-text articles to standard HTML. Figure insertions, if indicated, are managed by my software if authors send both the text and the image files to the EJCBS server by FTP. For figure insertion the text must include a simple control sequence in curly brackets such as {fig01.gif} where the image should be inserted. My software looks for the file with that name and inserts it. Similarly, an embedded link to a footnote or to a figure caption can be placed anywhere in the document. For example, {link01.html} will link the file "link01.html" to the original document. After compilation, an HTML page will be created with pictures and links to other pages. Using the curly-bracket characters, the authors can control the figure placement and the hypertext structure. It does not require more attention than following the format requirements of any printed journal. My instructions explain step-by-step how to create and send the ASCII format file to the Web server. (Editor's note: The conversion program is no longer available as of January 2001.)

After a successful submission, authors are shown immediate "proofs" of their submitted documents, and can check the look of their document and images, and check embedded links. The proof is not linked to the journal until the author corrects any errors and approves the final document. Then the corrected HTML document is added to the journal's table of contents and becomes immediately accessible on the World Wide Web.

All those editing procedures are interactive, and the changes take place immediately. No publication delay is involved; there is no cost for editing, printing and distribution. Replacing human editors with a computer autopublisher saves time and cost, but of course, there is still the big problem of maintaining the quality of the submitted articles.

Acceptance

Paper journals institute a comprehensive review from an editorial board of luminaries in the field who pass written comments to editors, to the reviewers, and to authors. Authors then rewrite and resubmit the paper.

EJCBS accepts articles for publication before review. To create and maintain standards, EJCBS uses a two-tier acceptance procedure that makes reviewing automatic and allows readers to control final acceptance.

Review Status: A newly submitted article is published in EJCBS in "review status." It is linked to an evaluation form that readers are encouraged to complete. The form includes only a few brief questions about essential qualities of the scientific publication such as "How significant is the problem discussed in the paper?" Readers click one of five buttons from "Not significant" to "Extremely significant." The buttons allow busy readers to review the article quickly and anonymously. The answers are transferred to a database that tallies results at the end of a month of review-status publication.

Archive Status: Articles that receive a score of 80 percent or higher, are transferred to an archive of accepted papers. Those papers that do not receive an 80 percent score are rejected, and the links are removed from the table of contents.

Reviews and Comments

One major weakness of paper publishing is the lack of discussion. Reviewers rarely see one another's work, and readers never see it. Paper journals could send the manuscript to selected peer reviewers and publish the original paper with the correspvonding commentary articles and author's response. This is currently a successful central part of the print journal Behavioral and Brain Sciences. Still, that is only one cycle of dialogue. The reviewers have no chance to reply to the author's responses. E-journals can easily handle multiple cycles of discussion. The author can decide how much response is necessary; simple comments can be addressed with simple answers, and more complex and relevant issues are likely to generate more debate. While the authors are not required to respond to every comment, adequate responses could influence the readers to evaluate the paper more positively. I believe that the authors and critics will be able to moderate the discussion appropriately.

When the target paper is accepted, some of those commentaries will also be published and archived in association, linked to the main paper. To decide which to publish, the commentary must also undergo the process of acceptance or rejection, and a shorter evaluation is attached to each commentary. The commentary evaluation has just one question with five choices. Only those commentaries that exceed the 80 percent threshold of acceptance will be published. That way the readers can directly control the contents of the journal. On the level of contents the journal is edited by the readers.

One inherent problem of such an automatic publishing system is the statistical nature of the control. The calculated average of the poll may represent the standard of the average reader and not the best experts. A very important paper may be rejected by the readers just because it deviates from the scientific paradigm. Just as in any other democratic system, voting does not always result in the best decision. Experts may be in the minority; dilettantes may hold sway. One way to overcome that problem is to give greater weight to the opinion of those readers who submit a commentary. That can counterbalance the votes of readers who evaluate without fully reading and understanding the paper. By changing the weight setting in the program, the system operator is able to attenuate or amplify the influence of different reader categories (readers, critics, and reviewers). Another control parameter is the threshold of acceptance. By changing the threshold of acceptance in the program, the system operator can influence the quality of the journal. Lowering the threshold allows papers to be accepted with lower average score, but it also increases the number of publications. Those parameters are flexible and we should experiment with them.

The importance of the reviews is not only to facilitate discussions. It also presents major benefits to the authors who submit their papers. They gain valuable feedback from the readers. Based on the opinions and commentaries or reviews, authors can improve their research, their paper, and the quality of the publication. The modified paper can later be submitted to a paper journal in an improved form with the mistakes and ambiguities identified and corrected.

A dilemma arises when paper journals refuse to consider papers that have been published elsewhere. Authors who seek the prestige of an established journal may not wish to publish first in an experimental electronic journal. But that problem is acute only during the transition period, until the e-journal becomes a generally accepted and referenced publication.

However, this legal ambiguity makes the start of an e-journal extremely difficult. In order to survive the transition period, EJCBS does not prohibit authors from publishing their modified papers in traditional journals. It seeks, instead, to serve as an experimental publication to help authors crystallize their final papers. If the article is published in paper in a better, cleaner version, authors can request that EJCBS delete the electronic version of the paper from the archive. But many times the printed paper is a compromised version of the original manuscript and authors may like to have their "director's cut" version available.

Archive

The "Los Alamos National Laboratory (LANL) Research Library" in physics is an example of an efficient electronic archive. The LANL Research Library collects experimental reports that may later be turned into research papers and submitted for publication elsewhere. The archive does not conflict with publication in a printed journal. Instead, it is an extremely useful tool for uncovering unexplored problems in physics. The LANL Research Library also shows that after reaching a critical mass the archive can become a primary tool of physics research. In this sense the LANL Research Library is an e-journal that is halfway between a paper journal and EJCBS, in that it is "preprint," but it doesn't go as far as EJCBS in promulgating a new form of peer review.

In the EJCBS all the accepted manuscripts are archived. I also encourage authors to send their URLs to our URL library. At the moment EJCBS has two separate URL libraries, the first a collection of other e-journals and electronic resources that I found useful, and the second, pointers to downloadable manuscripts or HTML documents submitted by the authors. The latter is author driven. My software lets any author submit a title and URL and automatically adds them to the library.

Natural Selection of Ideas

The unexploited potential of electronic publishing is to create an autonomous system that would be able to control itself based on rules. Those rules of submission, evaluation and acceptance, however, require an ethical group of users. Those who evaluate and criticize the papers must be honest and try to be unbiased. They must also take responsibility for their commentary. If the scientific community is up to that challenge, my system will evolve and be able to control itself like an autonomous organism.

An automatic publishing system could be thought of as an ecosystem, which provides an environment for the evolution of good ideas and scientific paradigms based on community consensus. That ecosystem would not only provide nutrition for the evolution of ideas, but in also providing a selection factor, it also eliminates the least viable ideas. The reader community decides whether a scientific finding has a logical flow and scientific importance. The reader community votes on the adequacy of the methods used to test the hypothesis. The reader community determines whether the conclusions are grounded. The archive can slowly evolve under the readers' opinions and control.

Please join this experiment by submitting a paper or commentary to EJCBS. In fact, I welcome your comments on this paper.

Zoltan Nadasdy is a Ph.D. candidate at Rutgers University, where he is studying the neural mechanism of learning and memory, continuing work done in 1983 at Eotvos Lorand University, Budapest. He received his MA degree in psychology from the Eotvos Lorand University in 1988. His dissertation research focuses on the "neural code" for memories. He has developed an on-line database, the "Basal Forebrain Cholinergic System."