The Rutgers Optimality Archive, ROA, has just passed 1,000 postings, including 123 dissertations. To mark the occasion, I offer you an exclusive interview with Alan Prince, who originally set it up back in 1993.
SNARL: Where did the idea for the Archive come from?
AP: Long before electronic texts, and even before easy xeroxing, the field was awash in blue mimeos and other low-grade copies of the latest papers. This defined the front lines, and no one was waiting for the judgment of gatekeepers. Nor was there much of a sense that academic careers were to be built by the major presses, or even perhaps that there were ‘careers’ in that sense. So the idea of free circulation — along with the view that you could judge things as well as anybody — has been out there for a while.
The immediate model, though, was the Neuroprose Archive created by Jordan Pollack (now at Brandeis, then at OSU) in the late 1980’s. This served the burgeoning connectionism community, and I came into contact with it through palling around connectionists as an offshoot of work I did with Steve Pinker (you heard about some of that last week in Alec Marantz’s talk). It was rather dazzling – you log in anonymously via ftp and have instant access to an entire, current literature.
Another model was the math-physics repository at arXiv.org (which now includes computer science and quantitative biology). This is huge beyond the dreams of linguistics — 500,000 papers, and quite fully integrated into the way those fields operate.
AP: Lack of oversight. For some reason I had fairly extensive privileges on the RuCCS server, and the Neuroprose model seemed easy to imitate. So I just made up a directory one day and got some people to send in papers. ROA-1 kicked things off with a nice short paper by Armin Mester and Jaye Padgett which uses generalized alignment, and in particular Kirchner’s flash about how to do directional footing, to get syllabic ‘directionality’ effects. At this point, it’s hard to reconstruct the details, but I remember writing and re-writing the instructions about how to use anonymous ftp, each time feeling that I’d reached a new height of clarity. I also cobbled together some kind of descriptive indexing system for the texts, which was based on the insight that a mark-up language used back-slashes and cryptic abbreviations. In retrospect, looking back from the link-clicking web world, the big easy, it’s hard to believe the lengths you had to go to just to up- and down-load stuff. We allowed every possible format – ranging from text to binary printer code.
One big break-out came when Mike Strickland, then a grad student in high-energy physics and the companion of a linguistics grad student (and a very good fiddle player, who played occasionally with the linguistics dept jam-band that Jaye Padgett dubbed ‘Eric and the Ominoes’, after the drummer) — when he offered to web-ize the interface. The then sysad of the RuCCS server wouldn’t let it run any scripts, so all the software ran on the hardware of a private company called webslingerZ, in North Carolina.
The modern era arrived when the interface was completely overhauled by Paul de Lacy, Graham Horwood, Eric Bakovic, and myself. de Lacy re-did the software from top to bottom — a very large project, it turned out, — and Horwood and I worked on design. Currently, de Lacy still watches out over the software — our most recent big move was to install full text search — and Bakovic still handles the day-to-day management, dealing with authors, problems, smoothly and endlessly. These are huge, very much behind-the-scenes contributions.
SNARL: Was there much controversy associated with the idea of free publication?
SNARL: Yes. Much push-back from journals and publishers, and so on?
AP: Remarkably little, and I think considerably less as the world has marched on. In the early ’90’s, the editor of a ‘prestigious, highly selective journal’ (as one writes in promotion narratives) actually blocked a colleague at RuCCS from publishing in the tech report series there. Out in the big world, on the Front to the math archive at http://front.math.ucdavis.edu/, there’s a list of more than 50 journals that accept submissions through the archive. You post and simultaneously indicate to the journal that you’re submitting. Linguistics hasn’t reached that level of integration yet, but I suppose it could happen.
AP: Interesting issue, but remember that there are strong natural forces that regulate nearly as well as reviewing. Post to ROA and more people will be able to see your work than if you put it any journal or book.
SNARL: Do you see no role for paper publication and reviewing, then?
AP: Let’s separate these issues. Paper publication is clearly moribund, a kind of animated corpse at this point, but reviewing lives on as useful. I suppose the question is, how useful, what’s the marginal utility of the extra investment of time and tsuris. There’s clearly some cross-over point, where the delay and the incomprehension rate outweighs whatever improvements ensue. Everybody has their stories — one colleague was chastised recently for using the logical symbol ‘horseshoe’, which it was felt would put too much strain on the delicate intellectual constitution of the readership, I kid you not. But we’ve all seen papers improve markedly as arguments tighten and errors disappear. (And we’ve all seen things get worse, with suppression of the interesting bits.) The posting process is really part of the reaction-and-review-and-rethink cycle that makes good work better.
SNARL: If paper publication is dead, or undead, as you claim, what’s with the shelves and shelves of books and journals and encyclopedias and handbooks that come out every year?
AP: A puzzle, indeed. The advantages of electronic presentation are just too enormous. The obvious attractions are universal access and searchability. Another major one is ease of correction and updating, which is now not even a consideration in journals and books. Others, just over the horizon, follow from the suspension of size and content limitations on documents and documentation with huge cheap storage available. How much data should you cite in support of an argument? All that you’ve got, at least by link: tapes, films, numbers. You used a program to calculate? — show us the code and let us run it. How much should you quote from a text you’re arguing with? I like to imagine that the standard of citation should be a link to the exact text, and that then the quality of argument would be forced up (and the degree of distortion, down).
This paradise of access has not yet arrived, but even in the context of the Optimality Archive you can see an immediate effect of sheer space: the range of things that count as papers is much wider. Journal space is held to be at a premium, but why? Many papers on ROA, which contain valuable results, could have no other home under current understanding of what publication is. So let’s change the definitions to accord with current realities.
All these considerations apply with full force to textbooks, as well. Peter Ladefoged, to cite one prominent case, was very good about getting a lot of the material associated with his excellent (and ultimately very expensive) phonetics books out there free on the web. There’s an example to follow — just drop the paper part, and the ‘for sale’ sign. Right now, we’re struggling with the paradox of marketing: the agency that exists to deliver goods for the benefit of its customers must also strenuously restrict their flow, for its own benefit.
SNARL: But aren’t you confounding ‘electronic’ and ‘free’?
AP: The two go so nicely together.
SNARL: We’re reaching the end of our space. May the next 1000 be as interesting as the first.
AP: And may the same be true for the issues of SNARL.