The Copyright Nobody Renewed

The register

Every field on those cards is a fact somebody wrote down. oreg is the original registration number and its date; the R number is the renewal registration and the day the renewal was recorded; the two-letter code is the statutory ground on which the filer claimed the right to renew. That code is not decoration. The 1909 Act, now at 17 U.S.C. 304(a)(1)(C), hands the renewal term to a specific queue: the author if living, otherwise the widow, widower or children, otherwise the executors, otherwise the next of kin. The Great Gatsby was renewed in 1953 by C, a child. Their Eyes Were Watching God was renewed in 1965 by NK, next of kin. The register is a machine-readable record of who outlived whom.

Twenty-eight years, then a form

The Copyright Act of 1909 gave a published work an original term of twenty-eight years from the date copyright was secured. To get the second term you had to file. The Copyright Office states the consequence of not filing without any softening:

Copyright owners of these works had to file a renewal application before the end of the twenty-eighth year to extend copyright protection into the renewal term. If a renewal registration was not properly filed, copyright protection ended on the twenty-ninth anniversary date of the original term, and the work entered the public domain.

U.S. Copyright Office, Circular 6A, Renewal of Copyright, rev. 03/2021

That is the whole mechanism, and it is why this page exists. A formality-based term is a filter: it lets go of everything whose owner has stopped caring, and it keeps everything whose owner is still paying attention. The second instrument shows the filter running.

The twenty-eighth year, and the wall at 1964

Set the class filter to the book classes and read the lower band. Nearly every filing sits on two columns. The law did not merely permit renewal in the twenty-eighth year; it made any other year useless, and the filers behaved accordingly.

Now look at the upper band on the right. Registrations from 1963 carry thousands of renewal filings. Registrations from 1964 carry almost none. Nothing happened to authors in 1964. What happened is that in 1992, Congress reached back and made renewal automatic for everything copyrighted between 1 January 1964 and 31 December 1977, so the filings those years would have generated in 1991 and 1992 were never made. The wall in the chart is a law passed twenty-eight years after the fact, visible as an absence.

Seven per cent, or twenty-eight? Both, and it is the denominator

The figure everyone repeats is that only about seven per cent of copyrights were renewed. It is a real number from a real study: Barbara Ringer counted renewals against Class A registrations for fiscal 1932 in the Copyright Office's own renewal study. The trouble is what Class A contained. Move the blocks out of the denominator and watch the famous number dissolve.

The unattributed remainder is worth naming out loud. The three named components add to 50,944, and Ringer's Class A total for fiscal 1932 is 57,065. We could not attribute the difference from the re-analysis as published, so the instrument carries it as its own block and lets you switch it in or out. That is a gap in our reading, not a claim about the Copyright Office.

The second modern count comes from the other end. In 2019 the New York Public Library finished converting the book volumes of the Catalog of Copyright Entries for 1923 to 1964 and could, for the first time, count registrations and renewals from the same digitized source rather than from a sample. Their headline is 162,416 renewals against 642,206 registrations. But eleven of those forty-one years are not measured.

So the honest range for "how much of the American book output of the renewal era is free" is not a number, it is an argument with named participants. Ringer's seven per cent is true about a denominator that is mostly pamphlets and magazine pieces. The book-proper rates from the same registers run twenty-six to thirty-three per cent. NYPL's twenty-five per cent is measured for thirty years and assumed for eleven. And John Wilkin, one of the co-authors of the re-analysis, argued the year before that estimates of the public domain from this period have been inflated, comparing Ringer against HathiTrust's Copyright Review Management System determinations. That disagreement is live. This page does not settle it.

The term machine

Given the facts about a work, the American rules produce a date. Not a status, a date. Set the facts and the machine walks the chart and prints both the date and the path it took.

The objection: nobody wanted the unrenewed ones anyway

Two operable answers, both of which say that a missing renewal is not the end of the inquiry.

One. The story under the film

It's a Wonderful Life is the standard example of a lapse that made a classic. Republic Pictures did not file a renewal in 1974, television stations discovered a Christmas film they did not have to pay for, and two decades of free broadcast built the audience the film never had in 1946. Then in 1993 Republic reasserted control, and the thing it reasserted was not the film's own copyright. It was the copyright in Philip Van Doren Stern's short story, plus the separately registered score.

Search the dossier above for The greatest gift. The record is there: AA446178, registered 20 December 1943, renewed R505822 on 10 May 1971 by Stern himself, claimant code A. The film lapsed. The story did not. Under the reasoning of Stewart v. Abend, 495 U.S. 207 (1990), you cannot keep exploiting a derivative work through the underlying work's renewal term without the underlying rights.

And Abend itself has a record in this dataset too. Cornell Woolrich promised the motion picture rights in his story to the producers of what became Rear Window, and promised to renew and assign the renewal term as well. He died in 1968, before the renewal could be made. Look at the dossier card for It had to be murder: registration B528177 of 2 January 1942, renewed R475565 on 29 December 1969, four days before the twenty-eight years from that 2 January 1942 registration ran out, by Chase Manhattan Bank as executor, claimant code E. That single line, filed by a bank in the last week of the year, is what the Supreme Court's decision runs on: a Hitchcock film could not be shown without it.

The 1992 Act changed the arithmetic for everyone who came after. 17 U.S.C. 304(a)(4)(A) now says that if no renewal claim is registered, a derivative work prepared under a pre-expiration grant may continue to be used during the renewed term. Register the renewal and you keep the Abend leverage; skip it and you keep the automatic term but lose that particular lock. The term machine flags both branches.

The same lock works at a smaller scale, inside the covers of a single book, and it is one of the two caveats NYPL itself attaches to its unrenewed count: those books, it warns, might "include previously published material still under copyright, or [have been] published abroad 1st & meet certain other URAA conditions". A book whose own registration has no renewal against it can still carry material somebody else registered separately and did renew: a reprinted story, a set of plates, a translation, a preface commissioned from another writer. The containing work's lapse frees the containing work. It does not free the parts that arrived with their own copyrights, and this page's register search, which matches on the book's own registration number, cannot see them at all. The second caveat is the URAA branch, and that is the next section.

Two. Works can come back out of the public domain

In 1994 the United States implemented its Uruguay Round obligations, and section 104A restored American copyright in foreign works that had fallen into the American public domain for failing formalities, including failing to renew. The Copyright Office puts the effective date and the eligibility test plainly:

January 1, 1996, was the effective date of restoration of copyright for works from countries that were members of the WTO or the Berne Convention on that date. [...] The copyright in any work in which copyright is restored will last for the remainder of the term of copyright that the work would have enjoyed if the work had never entered the public domain in the United States.

U.S. Copyright Office, Circular 38B, Copyright Restoration Under the URAA

The Office's own worked example is the fourth preset on the term machine: a French short story published without notice in 1935 is treated as if it had been published with proper notice and properly renewed, so its restored copyright runs to the end of 2030. Press that preset and the machine reproduces the date. A search of the whole 445,386-row renewal dataset for The Hobbit returns nothing, in any spelling, in any class, in any year. That is not a finding that the book is free. Tolkien was a British national and the book was first published in London in 1937, which is the exact fact pattern section 104A was written for. Press the fifth preset and watch a work with no American renewal on record carry an American term into the 2030s.

Section 104A has no floor under the publication year. The only date it tests is whether the term had already expired at home, so restoration reached back past 1930 as easily as it reached 1937. The sixth preset is the case that proves it, and it is the case this page originally got wrong: a London book of 1930, notice, no US renewal, still protected at home on 1 January 1996. An earlier version of this machine treated everything published before 1931 as beyond restoration's reach, called that book free since its 1958 anniversary, and was out by sixty-eight years. It was in fact under a restored American copyright from 1 January 1996 until the end of 2025, and it reached the American public domain on 1 January 2026, thirty years later and by expiry rather than by lapse. Set the sixth preset, then set the source-country question to "already public domain there", and watch the same book fall back to its 1958 anniversary date.

The Supreme Court was asked whether Congress may do this and said yes. Golan v. Holder, 565 U.S. 302, was argued 5 October 2011 and decided 18 January 2012. Ginsburg wrote for the Court, joined by Roberts, Scalia, Kennedy, Thomas and Sotomayor; Breyer dissented, joined by Alito; Kagan took no part. The public domain, in American law, is not a one-way door.

The passport

One work, one set of facts, four legal systems, four different answers. The divergence is not exotic; it is the default. Set an author who died recently enough and a work old enough and you can build a book that is free in the United States and in France while remaining owned in Germany and in Mexico.

Set the author's death year to 1951 and then to 1952 and watch Mexico jump seventy-one years. That is the non-retroactivity rule biting: a term that had already run out before the 1982 extension stayed run out, and a term that had one year left in it was carried forward through every later extension. Public domain status is not a property of a work. It is a property of a work in a place at a time.

The unprovable negative

Here is the honest frontier, and it is not a rhetorical one. Everything above rests on searching a register for an entry and not finding it, and this page has already demonstrated on itself that a search can fail on a real record over a definite article. The volumes were scanned, run through OCR and keyed, and the maintainers say so themselves. Titles change between registration and renewal. Authors are indexed under a pseudonym one way and a legal name the other. Renewals were batched, so twenty-eight registrations can hide behind one entry. And the bundled slice here covers eighteen registration years of a forty-one year era.

So the question is not whether a search can prove a negative. It cannot. The question is what a failed search is worth, and that is a number you can compute, if you are willing to say out loud what you are assuming.

Move the coverage slider from 99 per cent down to 90 per cent and, at the NYPL headline base rate, the posterior falls from about 0.980 to about 0.953. That is the shape of the whole problem. A library deciding whether to put a 1938 novel online is not choosing between "public domain" and "in copyright". It is choosing a position on this curve and then deciding how much liability it will carry there. That is why probable-public-domain books stay dark, and why HathiTrust built a whole review programme, the Copyright Review Management System, to make the determination one book at a time with human reviewers rather than by rule.

How the paper became searchable

None of this was operable twenty years ago. The Catalog of Copyright Entries was dozens of fat volumes a year: the Copyright Office printed it from 1891 through 1978, then issued it on microfiche from 1979 to 1982, and after 1982 the only way to see it is the Office's own online catalogue (Circular 22). Project Gutenberg volunteers transcribed the renewal volumes for 1950 to 1977. Stanford Libraries built the first searchable renewal database in 2007 from Class A book records, which you can still query. NYPL then did the harder half, converting the book registration volumes for 1923 to 1964 to XML, and reparsed the renewals into the tab-delimited form this page uses, unrolling batch entries so that a single renewal covering twenty-eight registration numbers becomes twenty-eight matchable rows. That unrolling is the difference between one match and twenty-seven false negatives.

The digitization is not finished. The pre-1953 pamphlet and serial volumes, Group 2 and Part 1B, are not converted. The pre-1923 and post-1964 volumes are not converted. Non-book classes for 1950 to 1977 are not in this dataset at all, which is exactly why the silence about a 1946 motion picture in these files proves nothing about that film. Every one of those gaps is a reason the coverage slider in Instrument VI is a slider.