Reader Comments

Post a new comment on this article

let's study the real history and data of publishing and reading

Posted by tmccormick on 31 Oct 2013 at 09:10 GMT

"this network model with a simplified equation that gives the proportion (or probability) of re-use:"

That's an interesting abstraction, but seems to me rather in the sociophysics realm, i.e. trying to model a human activity (here, reading) from first principles as if it's something inanimate like particle motion, and putting aside actual human history and study of the phenomenon. I'd be inclined to pay more attention to the decades or centuries of evidence and thinking we have about how publishing, reading, and knowledge dissemination work.

For example, if one is to talk of "diffusion" models, we have 50 years of studies to look at following Everett Rogers' coining of the concept with his "Diffusion of Innovations" in 1962. ( These show, for example, that diffusion and interest is not at all evenly distributed (as Neylon's '1 in X are interested' model might suggest), and proceeds mainly through certain key channels of peer influence.

Turning more specifically to research dissemination, we have decades of data and experience from e.g. arXiv and other repositories, CrossRef, link-resolver hubs, etc., and controlled studies done or being done on specific effects of different types of open access provision. (see especially the in-progress Sloan Foundation-funded study, “The Impact of Free On-line Repositories on the Diffusion of Scholarly Ideas” led Heekyung Kim and Erik Brynjolfsson at MIT / NBER).

At Stanford HighWire Press, my former employer, we had nearly 20 years of usage data from millions of researchers, plus many long-term ethnographic investigations of actual research behavior. It was empirically found, for example, that the further researchers are from their core expertise area, the more they rely on "brands" and long-established quality signals such as prestige journals. This presents a major challenge to our hopes, for example, for altmetrics, and it gives us a data-based picture far different than what a "Bayesian" or "percolation" models might suggest.

In short, when there is so much evidence, prior study, and existing arguments out there, I am skeptical of approaching open-access questions with an ex-nihilo sociophysics-style model. Publishing and reading are human phenomena with long histories; we should study and understand them as such.

Tim McCormick
Palo Alto, California

No competing interests declared.