For the last hour or so, I’ve been on wild goose chase looking for a judicial opinion. More accurately, looking for a commissioner’s report from the Court of Claims – essentially a trial opinion – for the case Williams & Wilkins v. United States.

I started with publicly available sources, still my habit despite currently having Lexis access. I couldn’t find a trial opinion with an internet search. On Court Listener, I came to the conclusion that Court Listener, at least, didn’t have it. Then I logged into Lexis to Shepardize the appellate opinion – Shepard’s showed no history below.

Next I did the thing that I’m trying to remember to do these days in lieu of banging my head against a task – I asked ChatGPT. (No, I haven’t switched to Claude yet. And yes, my success rate with this technique is frustratingly low, perhaps because I’m using the free version and not prompting well.) An interesting conversation ensued, which I wanted to include here but failed to export properly. (I wasn’t logged in during the conversation – when I logged in, it disappeared. I had printed to PDF already, but lots of pages were blank.)

From ChatGPT, I learned that the commissioner’s reports from that period generally were not published. It offered this charming suggestion (which I have verbatim because it was so terrible that I had to share with friends):

If you want

I can try to reconstruct the trial judge’s full reasoning (liability analysis, fair use rejection, policy arguments) from the available fragments and secondary sources—that’s often enough for serious research even without the full text.

Let’s not dwell on what kind of serious research such an item would be “enough” for.

I asked for help finding the official document. It told me to try a law library or Lexis or Westlaw. Since I was doing this during the daytime in India, when U.S. law libraries are not staffing chat reference, I asked about Lexis. It offered several strategies, none of which seemed obviously wrong, but none of them worked. Such is research – sometimes good search strategies yield nothing.

Then I thought I should give Lexis’s built-in AI, Protégé, a chance at the question. I started with the same prompt that I had provided to ChatGPT:

I’m researching a legal case, Williams & Wilkins v. United States. I have found two opinions. One from the Court of Claims with this citation: 203 Ct. Cl. 74 * | 487 F.2d 1345. Another from the U.S. Supreme Court with this citation: 420 U.S. 376. I thought there was also an opinion from the trial judge. Can you help me find it?

This conversation I do have a record of, because there isn’t a free or non-authenticated option with Protégé, and it saves all your conversations by default. Of course, since it’s Lexis, the conversation is subject to nigh-incomprehensible terms and conditions. Luckily, I have a lot of experience reading terms and conditions! Unfortunately, I work for a library that is also a Lexis customer (under whose license I am using Lexis), and I really don’t want to misinterpret the terms publicly. On the one hand, the terms are clear that Lexis isn’t claiming copyright in the AI-generated text (only any previously-copyrighted text in it, such as Lexis headnotes). On the other hand, it seems like the conversation might qualify as “Materials,” which I’m not supposed to put in a “database” and therefore am reluctant to post a full copy of here. I will, however, quote some “insubstantial” portions.

Protégé gave me a summary of the case (no info was new to me) and concluded with this: “You may need to consult the case record or filings from the Court of Claims for the original trial-level opinion.”

Naturally, I asked if that was something I could do on Lexis+. The following answer was a little interesting – it cited U.S. Code and the court’s local rules (current ones for the successor court, so not relevant for a 1970s proceeding) before drawing this conclusion: “These provisions suggest that official records, including trial-level opinions, are maintained and may be accessible through proper channels.” (Records available for a court case? I’m shocked!) Then it gave an oddly distant and imprecise answer to my actual question: “While Lexis+ typically provides access to appellate decisions, including those of the U.S. Supreme Court and the Court of Federal Claims, trial-level opinions may not always be included unless specifically published or made available by the court.”

Perhaps Protégé is more useful if you’re asking legal questions, but not being able to get answers about Lexis from Lexis’s own AI tool is a major limitation. As noted above, ChatGPT actually worked better for Lexis than Protégé did – at least it provided plausible search strategies.

At this point I went back to ChatGPT, said that I couldn’t find the document on Lexis, and asked if it might be available, digitized, from the federal government. ChatGPT did its ChatGPT thing – it told me how smart I was to think of that question. Then it pointed me to the National Archives. It told me to look at Record Group 123. It got some of the nitty-gritty of the finding aid wrong, e.g., saying that the series “123.2 Case Files for General Jurisdiction Cases 1855-1966” actually went to 1970. But, I am certain that the document I want is at the National Archives, and that sure seems like the right record group. The trouble is that it’s big, and some of it seems not to be described yet. There’s still a lead I could follow here, but at this point I changed course.

In the midst of the National Archives search, I realized something obvious. I was interested in this document because Peter Hirtle had mentioned processing the National Library of Medicine’s records of the case while working at NLM. (NLM was the defendant, basically.) Maybe there’s a copy in the records Peter processed. The great things about this approach are that the collection is fully processed, and the finding aid is digitized fairly well. The downside is that I’m less certain that a copy of the document will be there, as compared to in the records at the National Archives. If I had to guess, I think the document I want is Report to Commissioner, 2/16/72, in box 5, folder 7. I just wish the preposition were “of” – then I would be more certain. [Spoiler: Turns out that is the date of the document I want, so I’m pretty sure it’s the right thing.]

The NLM records became another abandoned lead, but the benefit was that I had turned my brain back on. I had stopped thinking about where the document should be (Court Listener, Lexis, the National Archives) and started thinking about ways I might find it. Recently, I read Betsy Bernfeld’s article “Free to Photocopy?: A Legislative History of Section 108, the Library Photocopying Provision of the Copyright Act of 1976,” published in 2006 in Legal Reference Services Quarterly. Bernfeld talks about all stages of Williams & Wilkins. Where did she read it? 

When discussing and quoting the trial-level decision, Bernfeld cites the Register of Copyright’s 1983 report on library photocopying and 108. (This report was required by a now-repealed section of Section 108 that required that the Register of Copyrights submit a report to Congress every five years “setting forth the extent to which [section 108] has achieved the intended statutory balancing of the rights of creators, and the needs of users.” I had a moment of hope! That report is on my desktop! Could the Copyright Office have included the trial-level decision as an appendix?!

Alas, no. The Register’s Report contains long quotations, enabling Bernfeld to quote the trial-level decision directly. But the full text is not there. What was the Register’s source? Here’s the citation:

Williams & Wilkins Co. v. United States, 172 U.S.P.Q. 670 (Comm’r Ct. Cl. 1972), rev’d 487 F.2d 1345, 1363 (Ct. Cl. 1973), aff’d by an equally divided court, 420 U.S. 376 (1975).

U.S.P.Q. – the United States Patent Quarterly! This is a legal reporter, but a bit of an odd one (at least by today’s practice). It publishes cases from a variety of courts, including some things not published elsewhere. Forgetting a crucial detail, I plugged the U.S.P.Q. citation into Lexis. It redirected me to a totally different case, In re Willis, from the Court of Customs and Patent Appeals. But that case was also from 1972. What was going on? Then I remembered that Lexis doesn’t include U.S.P.Q. It has a lot of the cases because they appear in other reporters. But it probably doesn’t have all of them, and I already strongly suspected that it didn’t have this one. And, I don’t have any direct digital access to U.S.P.Q. – it’s a Bloomberg product.

I went back to Google Search, still my most-used legal research tool (though it’s not what it used to be), and tossed in the citation from the Register’s report. The first hit was an article in Vanderbilt Law’s repository, by L. Ray Patterson. I opened it and searched for “U.S.P.Q.” There are two hits. The first cites Williams & Wilkins as appearing at 172 U.S.P.Q. 672, different from the Register’s report. But this had come up in my search results because there’s another citation, referring to 172 U.S.P.Q. 670. 

At that point, I was sure I would find the document, so I started writing up this blogpost. While the initial searching took an hour, I’ve easily spent two writing this. As I was typing that I don’t have Bloomberg access, I thought – maybe I should double-check. Turns out that I do.

I can now confirm that the citation in the Register’s Report was correct. It is Williams & Wilkins Co. v. United States, 172 U.S.P.Q. 670 (Ct. Cl. 1972), and if you have Bloomberg access, you can get the whole thing there. The Patterson citation was a red herring. I’m still not sure why Lexis resolved to what it did.

Once you have the proper citation, and indeed, the full text, it is much easier to find the document. So, for the benefit of anyone else looking for the trial opinion in Williams & Wilkins Co. v. United States, 172 U.S.P.Q. 670 (Comm’r Ct. Cl. 1972), rev’d 487 F.2d 1345, 1363 (Ct. Cl. 1973), aff’d by an equally divided court, 420 U.S. 376 (1975), here are a couple places you can find it for free on the web (if you find more please drop me a line – I would love to add to this list):

• In ERIC, as ERIC Number ED058921 – This copy looks very authentic, which is to say it appears to have been printed for distribution by the Court of Claims in 1972. It’s pretty hard to read, but the OCR seems decent.
• On Ronald B. Standler’s personal website – This is a transcription of the opinion, with commentary (that I don’t agree with). It appears to have been prepared with WordPerfect, so OCR isn’t an issue – the searchable text is perfect. It includes the pagination from U.S.P.Q., by permission.

A postscript

A few times this year, people have asked me what I’m doing on sabbatical, how it’s going, what it’s like. This post certainly isn’t the full story, but it’s a good example of why sabbatical has been special. What got me started on this was closely related to my sabbatical project – developing new teaching materials for CopyrightX: Libraries. I was captioning the recording of a special event yesterday with Peter Hirtle, in preparation for posting it publicly. Then I thought, wouldn’t it be great to share a Zotero folder with every document mentioned in this event? Were I not on sabbatical, that thought would have gone no further. But I started the folder, and it needed this case. Then I thought I should write a blog post. Writing the post helped me find the file. 

Was any of this necessary? No. But it was valuable. I tried new things. This was my first time asking a legal research question to ChatGPT (normally I just ask about OpenRefine), my first time using Protégé (3/10, may not use again), and the first time in a long while that I’d consulted a finding aid for a purpose other than analyzing the copyright issues involved in digitizing a collection. I saw research resources that I hadn’t used before, and I saw them with the eyes of a patron. And I blogged about it.