Inside the Georgia State Opinion


On Friday, the long-awaited decision in the Georgia State e-reserves case (a.k.a. Cambridge University Press v. Becker) dropped. By way of context, the case is a challenge by three academic publishers (Oxford University Press, Cambridge University Press, and Sage Publications) against Georgia State University’s e-reserves policy. The publishers sued in April 2008, in a lawsuit funded by the Association of American Publishers and the Copyright Clearance Center, claiming that the e-reserves policy went far beyond the bounds of fair use. Georgia State, as a state university, invoked the doctrine of sovereign immunity, the practical implication of which is that the publishers can only obtain injunctions against future infringements, not damages for past infringements. Since it also tightened up its e-reserves policy in December 2008, it also successfully argued to the court that only the uses made under the new policy should be relevant to any potential injunction.

There was a trial a year ago, and then long silence from the court. Now we know why it was taking so long: the opinion is 350 pages. That number is a little misleading, in that over two thirds of the opinion are dedicated to a highly methodical copyright ownership, infringement, and fair use analysis of seventy-four separate claims of infringement, using standard templates and highly repetitive language. Having now dug through the details, I’d like to offer a few observations.

First, over a third of the claims didn’t even make it to the fair use stage at the heart of the case. In many cases, the publishers were unable to prove to the court’s satisfaction that they owned the copyright in the portions of the books that were copied and uploaded. Sometimes they couldn’t produce a timely registration certificate and there were proof problems with originality; sometimes they couldn’t find a work-made-for-hire agreement or copyright assignment from the authors of individual chapters in edited volumes. The court was unsympathetic: no documented chain of title, no lawsuit. There’s a looming e-rights mess, loosely akin to the robosigning mess around ownership of securitized mortgages: in both cases, the putative owners don’t have all their papers in order. This opinion either recognizes or contributes to the mess, depending on your point of view.

Other claims dropped out before the fair use stage because they were uploaded to the e-reserves system but never downloaded by students. The court dismisses these from the lawsuit as de minimis, explaining that these uses by the University, while technical implicating the copyright owners’ exclusive rights, don’t affect the incentives for authors to create. This puts more teeth in the de minimis doctrine in copyright: it goes beyond the view that de minimis means “not substantially similar.” It also strengthens the argument that “internal use” copies never used to reach an to an audience that reads them for their content don’t infringe. Think, for example, of the HathiTrust’s archive of scans from Google Books.

(As an aside, the e-reserve logfiles played a key evidentiary role in the case. Specific users were never identified, but if a file had a total hit count of two, it’s unlikely that students actually read it. This stands in contrast to other cases, like American Geophysical, which was tried by sampling: the parties selected a single scientist at random, examined his files looking for photocopies, and treated him as representative of a cohort of 500. Here, the logs permitted an analysis of the copying done for numerous faculty members—presumably all those who assigned any excerpts from any of the plaintiffs’ books.)

When the court did reach fair use, it held across the board that two of the four factors favored Georgia State. The purpose of the use, while not transformative, was nonetheless for highly favored educational purposes by a nonprofit institution. And the nature of the works was consistently informational.

On the third factor, the amount copied, the court repudiated the Classroom Guidelines, calling them “not compatible with the language and intent of § 107.” It noted that the numerical limits in the Guidelines are so stringent that not one of the excerpts at issue in the case would fit within them. It was particularly uninterested in the Guidelines’ position that copying not “be repeated with respect to the same item by the same teacher from term to term,” which the court described as “an impractical, unnecessary limitation.”

Instead, the court fashioned its own quantitative test. For books of nine or fewer chapters, the court set a threshold of 10% of the total page count; for books of ten chapters or more, the threshold was a single complete chapter. (The chapter-based rule creates an odd incentive for publishers to create books with a surfeit of tiny chapters.) Copying of any amount under this threshold, the court held, would be treated as “decidedly small.” In practical terms, this ended up being a one-sided bright-line rule: copying of less than 10% or one chapter always ended in a fair use win for Georgia State.

Finally, the fourth factor, the effect on the market, favored the publishers whenever CCC was offering a digital license for copying the book in question, and favored Georgia State whenever there was “no evidence in the record to show that digital excerpts from this book were available for licensing” as of the date of infringement.” In practice, this was another one-sided bright-line rule: no digital license meant an instant win for Georgia State. The court repeatedly emphasized that students would not have bought the assigned books as a substitute for the excerpts posted on the e-reserve system.

This treatment of licensing is likely to have significant implications. On the one hand, it suggests that libraries may have a freer hand to make expanded uses of orphan works, since by definition, no one will be licensing them. And on the other, the court didn’t consider photocopying licenses to be a suitable substitute for digital licenses. This will put significant pressure on publishers to turn on digital licensing.

Only in seven instances did Georgia State use more than 10% or one chapter of a book that was available for digital licensing. When this happened, the court took a more detailed look at the specifics of the book’s licensing market and the portion copied. Generally, this turned on whether the book made significant revenues via licensing: if so, the use was unfair. (In one instance, the court did a “heart of the work” analysis under factor three to find no fair use because the professor had assigned chapters that “essentially sum up the ideas in the book.”)

Thus, the operational bottom line for universities is that it’s likely to be fair use to assign less than 10% of a book, to assign larger portions of a book that is not available for digital licensing, or to assign larger portions of a book that is available for digital licensing but doesn’t make significant revenues through licensing. This third prong is almost never going to be something that professors or librarians can evaluate, so in practice, I expect to see fair-use e-reserves codes that treat under 10% as presumptively okay, and amounts over 10% but less than some ill-defined maximum as presumptively okay if it has been confirmed that a license to make digital copies of excerpts from the book is not available.

The most interesting issue open in the case is the scope of any possible injunction. Given that Georgia State won on sixty-nine out of seventy-four litigated claims, while the publishers won on only five, I expect that the any injunction will need to be rather narrow. But given how amenable the court’s proposed limits are to bright-line treatment, it is likely that the publishers will push to write them in to the injunction.

My bottom line on the case is that it’s mostly a win for Georgia State and mostly a loss for the publishers. The big winner is CCC. It gains leverage against universities for coursepack and e-reserve copying with a bright-line rule, and it gains leverage against publishers who will be under much more pressure to participate in its full panoply of licenses.


I’d like to point out one other issue that got somewhat submerged in the publishers’ poor recordkeeping: The US offices of two of the three publishers are notorious for treating works as work-for-hire that are outside the statutory eligibility criteria in ยง 101’s badly-written definition. A book-length work by a single author is not in any of the categories of eligible works… and especially not when the work preexists any agreement to treat it as WFH.

The less said about the insistance of many law journals that articles published in those journals are WFH, even when not commissioned for that journal, the better!


Thanks for the very helpful analysis. I am curious about whether section 110 of the act might offer greater protection to faculty in cases where fair use may not be enough. It seems that faculty have a choice on whether to post online content through their own course management system, or hand it off to the library for inclusion in e-reserves. I know the Georgia State case did not address section 110, but I am grappling with the intersection of fair use and section 110.


I don’t understand how the judge can disavow the old guidelines(which I concur with), while creating a new one of her own. I feel a little sick over this ruling. GSU and education in general seems to have won, but I don’t think fair use, in all its subtlety, has really been worked out. You are astute in concluding that the real winner could be CCC. Aren’t institutions in for a tidal wave of new digital versions, which has the potential to trump fair use even further?


I agree with the comments on WFH and the expectation that digital licenses for textbooks will/should now be negotiated routinely. The court seems to have punished textbook publishers for not publishing digital editions. Publishers should be presumed to have the right to publish as their market demands, but the court (ironically, given ultra-conservative Georgia) has decided publishers’ business decisions be damned, digitize or die!

As for the determination that if no one reads an unpermitted copy there’s been no infringement, does that mean that GSU will be infringing as soon as two people access a book? In that case, all the unread unpermitted copies should be sequestered so that those two people can’t get at them.


I’m still troubled by her flat assumption that educational use with nothing more favors fair use. There’s ample precedent requiring the use to be transformative before it favors fair use.