Ground truth · the United States Reports, counted

Dissented, but Did Not State His Reasons

For a century and a half the Supreme Court published as one voice, and it was a habit, not an agreement. Across all 572 volumes of the United States Reports there are 26,745 decided cases; in the 1930s 11.7% of them carried a separate opinion, and by the 1950s 59.7% did. In 1936 fewer than one case in ten drew a separate opinion; ten years later more than half did, and it has never gone back.

Livingston, J., dissented from this opinion, but did not state his reasons. Higginson v. Mein, 8 U.S. (4 Cranch) 415, 421 (1808). The last line of the Court's opinion, verbatim from the scanned report

That sentence is the whole of Brockholst Livingston's disagreement. Not a paragraph, not a reason: a line the Reporter added after the judgment, the way you would note the weather. Five years later the same formula records Justice Johnson doing the same thing, word for word.

It reads today like a missing page. It was not. For most of the Court's history a Justice who lost had three things he could do, and writing an opinion about it was the rarest. He could say nothing at all. He could have the Reporter note that he disagreed, without saying why. Or he could write, and for a hundred and fifty years almost nobody did.

Then, in about ten years in the middle of the twentieth century, that stopped being true, and it has never gone back. The whole of it is countable, because the Court publishes its own record and that record is now free to read. What follows is all 26,745 decided cases in the 572 bound volumes of the United States Reports, one mark each.

The wall

One row per year, 1792 at the top to 2014 at the bottom. One mark per decided case, left to right in the order the Reports print them. A row's length is how much the Court decided that year.

The shape is the finding. For a hundred and fifty rows the wall is nearly empty; the ochre marks scattered through it are the cases where somebody's disagreement was noted in a line and never argued. Then the red arrives, not gradually, and after it the Court never again publishes a year that looks like its first century.

Ten years

Two crossings, with no model behind them. The last year in which fewer than one decided case in ten carried a separate opinion was 1936 (9.3% of 150 cases). The first year in which more than half of them did was 1946 (54.3% of 140). Ten years apart. Nothing before 1936 looks like anything after 1946.

Share of decided cases carrying at least one separate opinion, by decade
decade	cases	with a separate opinion
1790s	38	5.3%
1800s	150	4.0%
1810s	322	2.5%
1820s	350	5.1%
1830s	427	8.9%
1840s	337	9.8%
1850s	844	10.3%
1860s	867	7.8%
1870s	1,906	10.2%
1880s	2,324	4.3%
1890s	2,380	5.7%
1900s	1,883	8.3%
1910s	2,294	4.9%
1920s	1,900	7.1%
1930s	1,568	11.7%
1940s	1,417	40.8%
1950s	998	59.7%
1960s	1,264	65.1%
1970s	1,632	68.5%
1980s	1,640	74.0%
1990s	1,055	70.5%
2000s	797	68.3%
2010s	352	57.7%

Where you draw the line

Every count needs a denominator, and this one has a choice in it. The Reports print orders as well as opinions: 328,456 Supreme Court entries in these volumes, of which the overwhelming majority are one-line denials of certiorari. A case counts here as decided if its lead opinion carries an author line and the case runs to at least 400 words, which is roughly a page of the Reports.

That threshold is a judgement, so here is the whole family. Move it and watch what happens to the curve.

Minimum words for a case to count

each faint line is one threshold; the bold one is your selection the shaded strip is 2013 onward, where the bound volumes stop mid-Term

The level moves. The break does not. At every threshold from zero to 3,200 words the curve is flat and low for a century and a half and then climbs, in the same decade, to a plateau it has never left.

The longest silence

Run the 26,745 cases in report order and look for the longest stretch with no separate opinion in it at all, and no note of one either. It is 105 consecutive decided cases, from 8 U.S. 421 (Pollard v. Dwight, 1808-03-15) to 11 U.S. 366 (Lee v. Munroe, 1813-02-26): 1,809 days, near enough five years, in which the Court decided a hundred and five cases and not one member of it is recorded as disagreeing with a word of any of them.

Relaxing the rule to ignore the one-line notes, the longest run of cases with no written separate opinion is 175 cases (136 U.S. 211 to 139 U.S. 137, 287 days, in 1890–1891).

The check that had to come first

There is an obvious way for this whole finding to be false. These volumes are scans. If the software that split each case into opinions was worse at nineteenth-century typesetting than at modern typesetting, if it quietly glued old dissents onto the end of the majority instead of separating them, then the numbers above would look exactly like this and mean nothing.

So every case counted as unanimous was opened again and its full text searched for a separate-opinion heading the typing had missed: 19,648 cases read, a heading found in 1,549 of them (7.88%). The answer is the opposite of the worry. The typing misses more in the modern era than in the old one, because the modern era is where the one-line note survives, attached to summary dispositions. The audit is an upper bound (some hits are a lower court's dissenting judge named in the recital of proceedings, not a Justice), and even taken at face value it widens the gap rather than closing it.

Typed separate opinions, and the upper bound after adding every missed heading
decade	cases	as typed	upper bound
1790s	38	5.3%	5.3%
1800s	150	4%	4.7%
1810s	322	2.5%	3.1%
1820s	350	5.1%	5.7%
1830s	427	8.9%	12.6%
1840s	337	9.8%	12.5%
1850s	844	10.3%	17.8%
1860s	867	7.8%	9.5%
1870s	1,906	10.2%	15.3%
1880s	2,324	4.3%	6.5%
1890s	2,380	5.7%	14.3%
1900s	1,883	8.3%	19.1%
1910s	2,294	4.9%	11.6%
1920s	1,900	7.1%	14.7%
1930s	1,568	11.7%	17.2%
1940s	1,417	40.8%	53.1%
1950s	998	59.7%	69.4%
1960s	1,264	65.1%	72.3%
1970s	1,632	68.5%	71.9%
1980s	1,640	74%	75.2%
1990s	1,055	70.5%	71.6%
2000s	797	68.3%	70.9%
2010s	352	57.7%	66.8%

Checked against two counts made by hand

Two published series overlap this corpus, and neither was made from these scans. Both were coded by people reading books.

The first is as close to a control as this question has. In 2001 Lee Epstein, Jeffrey Segal and Harold Spaeth obtained Chief Justice Waite's private conference docket books, which record how each Justice actually voted in conference between 1874 and 1888, and set them beside the published Reports. Of 2,863 cases, 9% appear in the Reports with one or more dissenting votes. 40% were nonunanimous at conference.

Counting the same window here gives 3,319 decided cases, 6.2% with a written separate opinion and 9.8% once the one-line notes the audit recovered are added back. Their 9%, this page's 9.8%, from different books by different methods a quarter of a century apart.

Their other number is the one to keep, and it is the reason for this page's title. Four of every ten of those cases had a Justice on the losing side. One in ten said so in print.

The second series is Walker, Epstein and Dixon's, hand-coded Term by Term for 1935 to 1946, which is the window that contains the break: dissenting opinions per opinion of the Court. The same quantity computed here runs lower throughout, for reasons already on this page (their denominator excludes the per curiam dispositions this one keeps, and the typing misses the one-line notes). The shape is the same one. Pearson correlation between the two series across the twelve Terms: 0.8993.

Dissenting opinions per opinion of the Court, 1935 to 1946
Term	this page	Walker, Epstein & Dixon (1988)
1935	0.12	0.14
1936	0.07	0.11
1937	0.12	0.17
1938	0.13	0.25
1939	0.18	0.15
1940	0.16	0.16
1941	0.19	0.29
1942	0.20	0.43
1943	0.30	0.54
1944	0.41	0.51
1945	0.40	0.51
1946	0.48	0.57

A small correction, offered

John P. Kelsh counted the Reports by hand for exactly this purpose in 1999, and his article is the best thing written on the question. In a footnote he dates the form this page is named after:

I say “nearly” always because this period also saw the first dissent without opinion. See Herbert v. Wren, 11 U.S. (7 Cranch) 370, 382 (1813) (Johnson, J., dissenting). John P. Kelsh, The Opinion Delivery Practices of the United States Supreme Court 1790–1945, 77 Washington University Law Quarterly 137, 147 n.65 (1999)

The Reports appear to hold an earlier one. Higginson v. Mein, 8 U.S. (4 Cranch) 415, 421, decided 14 March 1808, ends with the identical formula five years before Herbert, and the dissenter is Livingston rather than Johnson. Both sentences were read here in two independent scans: the Caselaw Access Project's, and the Library of Congress's own PDF of the bound volume (tile.loc.gov/…/usrep008415.pdf and usrep011370.pdf), where they read, respectively, “LIVINGSTON, J. dissented from this opinion, but did not state his reasons.” and “JOHNSON, J. dissented from the opinion of the Court, but did not state his reasons.”

This is offered as a correction and not as a gotcha. Kelsh's own footnote, immediately before that sentence, observes that “Justice Livingston, for example, never once explained what motivated him to make his disagreement public”, so he plainly knew the man. He may also have meant something narrower by first than the words carry. Either way the earlier case is there, and anybody can open the same two PDFs.

What the audit found

The by-product is better than the check. Searching for missed headings turns up the vanished form itself: disagreement recorded as a fact rather than an argument. A few, verbatim from the scans, with the reporter's own spelling and the scanner's own errors left in.

…shall desire it. Livingston, J., dissented from this opinion, but did not state his reasons. Decree reversed.… 8 U.S. 415 · Higginson v. Mein · 1808-03-14

…herein’ contained. Johnson, J. dissented from the opinion of the Court, but did not state his reasons.… 11 U.S. 370 · Herbert v. Wren · 1813-02-26

…for uncertainty. Mr. Chief Justice Marshall. In this case I dissent from the opinion which kas been delivered on one point;. the validit… 15 U.S. 316 · Shipp v. Miller's heirs · 1817-03-05

…affirmed.' Mr. Chief Justice' Marshall, Mr. Justice Washington, and Mr. Justice Duvall, did not sit in this cause.’ -Mr. Justice S… 22 U.S. 581 · Renner v. President of the Bank of Columbia · 1824-03-05

…the United States, Mr. Justice Washington, Mr. Justice Thompson, and Mr. Justice Trimble, dissented; Mr. Chief Justice Marshall, Mr. Justi… 25 U.S. 213 · Ogden v. Saunders · 1827-02-19

…decisions. Justices Thompson and Baldwin, dissenting. This cause came on to be heard on the transcript of the record from the circ… 28 U.S. 222 · Patapsco Insurance v. Coulter · 1830-01

Who wrote separately

Author lines are the reporter's own text, run through a scanner, so they are matched against a hand-entered roster of 103 surnames; 94.1% of separate opinions matched a name and 653 did not. Surnames the Court used twice cannot be told apart here. Johnson, Rutledge, Marshall, Harlan, White, Jackson, Lamar, Chase, Roberts each name two different Justices, so those rows are the pair, not the person.

surname on the opinion	separate opinions	dissents	concurrences
Stevens	1,052	615	338
Douglas	767	546	172
Brennan	675	386	210
White (two Justices)	547	298	205
Harlan (two Justices)	524	296	171
Scalia	506	194	269
Blackmun	493	221	231
Marshall (two Justices)	450	330	85
Black	420	279	96
Rehnquist	405	287	85
Frankfurter	379	234	134
Powell	338	126	171
Stewart	316	176	129
Thomas	285	123	134
Breyer	256	139	89

How much gets written

The other half of the change is length. The number of separate opinions per decided case went from 0.133 in the 1930s to 1.265 in the 1980s. Over the same stretch the median opinion of the Court roughly doubled, from 2,045 words to 4,445. A separate opinion, when one was written, was always a real opinion: its median length barely moves across two centuries. What changed was how often somebody wrote one.

decade	cases	separate opinions per case	median words, opinion of the Court	median words, separate opinion
1790s	38	0.053	542	2,510
1800s	150	0.047	831	561
1810s	322	0.028	1,146	508
1820s	350	0.054	2,126	2,513
1830s	427	0.11	2,202	1,819
1840s	337	0.131	1,988	1,692
1850s	844	0.134	1,806	1,394
1860s	867	0.087	1,191	889
1870s	1,906	0.121	1,513	356
1880s	2,324	0.051	1,781	1,031
1890s	2,380	0.066	2,219	1,758
1900s	1,883	0.098	2,042	1,162
1910s	2,294	0.055	1,682	1,537
1920s	1,900	0.091	1,543	982
1930s	1,568	0.133	2,045	1,437
1940s	1,417	0.555	2,573	1,234
1950s	998	0.868	2,313	920
1960s	1,264	1.012	2,657	1,188
1970s	1,632	1.2	3,598	1,094
1980s	1,640	1.265	4,445	1,480
1990s	1,055	1.232	4,205	1,697
2000s	797	1.301	4,147	1,428
2010s	352	0.926	4,620	2,180

What this does and does not show

It shows what the Court published. It does not show what the Justices thought, and thanks to Waite's docket books we know roughly how far apart those two things were: in the 1870s and 1880s, four cases in ten were nonunanimous in the room and one in ten was nonunanimous in the book. A unanimous line of cases in 1890 is not evidence that nine men agreed. It is evidence that disagreement was not the kind of thing you put in print.

So the rise measured here is, at least in part, a rise in disclosure rather than in disagreement. That is not a weakness in the measurement; it is the thing being measured. What changed in the 1940s was what the record was allowed to contain.

Nor does it show a cause. The break sits in the years around the elevation of Harlan Fiske Stone to Chief Justice in July 1941, and the argument over whether that is why has been running for forty years. Walker, Epstein and Dixon made the case for Stone's leadership in 1988 after testing and rejecting four rivals: the Judiciary Act of 1925, caseload volume, the mix of cases, and the arrival of a bench appointed almost entirely by one President. Stacia Haynie answered in 1992 that the critical actor was Hughes, not Stone. Smyth and Narayan, estimating where the breaks are rather than assuming one, found three rather than one. Kelsh argues the spike of 1941 was a spark landing on kindling a century old. This page takes no side. It says only, and exactly, when the record changed.

The check, shown

Everything above is recomputed by research/the-vote-you-were-allowed-to-see/verify.mjs, which re-derives each number from the case records by a path independent of the script that produced them, and fails if any disagrees.

Source. All 572 volumes of the United States Reports from the Caselaw Access Project's static mirror, https://static.case.law/us/, fetched 2026-08-01. Judicial opinions of the United States are not under copyright. fetch.mjs re-fetches; distil.mjs reduces the volumes to one record per case; data/cases-min.jsonl.gz is that record set, committed, so the whole pipeline runs from this repository alone.
Universe. 329,995 cases in the reporter, of which 1,539 were decided by other courts entirely, because the early nominative volumes report Pennsylvania cases beside Supreme Court ones. Dropping them is the first thing the analysis does. 328,456 Supreme Court entries remain; 26,745 of those are decided cases under the rule above.
Separate opinion means an opinion after the first typed dissent, concurrence, concurrence-in-part, or plurality. Rehearings and clerical entries are excluded.
Coverage ends in 2014, with volume 572. The last two years are partial and the wall shows them as they are.
Scan quality. These are optical readings of printed books, and the oldest volumes read badly: the long s of eighteenth-century type comes back as an f. Quoted passages are reproduced exactly as the scan has them, errors included, rather than silently corrected.
The seriatim era is not resolved here. Before 1801 the Court used no settled form: counting the cases Dallas reported for 1790 to 1800, Kelsh finds 45 of 63 (71%) issued as a brief unattributed order “By the Court” and 15 of 63 (24%) delivered seriatim, one Justice at a time. In these scans the seriatim opinions are frequently merged into a single block with a garbled author line, so the first two decades of the wall are the least trustworthy part of it, and no claim here rests on them.
Before 1834 the Reports are a private venture. The Court did not require its opinions to be filed with the Clerk until March 1834, and the early Reporters printed what they chose: Warren estimated Dallas omitted at least a tenth of the cases decided in his sixteen active Terms. Any run of silence in the Dallas and Cranch volumes is a fact about what one man had printed, which is exactly why the two quoted sentences from that period were re-read in a second, independent scan.
Provenance is mixed in the early volumes. What was scanned is the bound United States Reports as Harvard held them, and for the nominative series that means a mixture of original printings and later full reprints (volume 11 is Cranch's own 1816 printing; volume 8 is a 1911 Banks Law Publishing reprint). That matters because Peters's Condensed Reports of 1830 deliberately pared away concurring and dissenting opinions, so a corpus built on condensations would understate exactly this page's subject. These are not condensations, and the two passages this page quotes were checked against the Library of Congress's separate scans.
Numbers. 19,648 cases re-read for the typing audit; 11,138 separate opinions counted in total; 10,485 of them matched to a surname.

Works cited

Caselaw Access Project, Harvard Law School Library Innovation Lab. https://static.case.law/us/. The bulk data carries no use restriction (the Ravel Law limitation expired in March 2024), and the opinions themselves have never been copyrightable: Wheaton v. Peters, 33 U.S. (8 Pet.) 591 (1834), the Supreme Court's own Reporter suing his successor, and Banks v. Manchester, 128 U.S. 244 (1888). CAP digitized books published through 2020; because the bound Reports run years behind the Term, the last United States Reports volume in it is 572 (October Term 2013).
Lee Epstein, Jeffrey A. Segal & Harold J. Spaeth, “The Norm of Consensus on the U.S. Supreme Court”, American Journal of Political Science 45(2) (2001), 362–377. Waite's conference docket books against the Reports, 1874–1888.
Thomas G. Walker, Lee Epstein & William J. Dixon, “On the Mysterious Demise of Consensual Norms in the United States Supreme Court”, Journal of Politics 50(2) (1988), 361–389. Table 1 at 364 is the series compared above.
John P. Kelsh, “The Opinion Delivery Practices of the United States Supreme Court 1790–1945”, 77 Washington University Law Quarterly 137 (1999). The hand count of the early Reports, and the footnote corrected above.
Stacia L. Haynie, “Leadership and Consensus on the U.S. Supreme Court”, Journal of Politics 54(4) (1992), 1158–1169; Russell Smyth & Paresh Kumar Narayan, “Multiple Regime Shifts in Concurring and Dissenting Opinions on the U.S. Supreme Court”, Journal of Empirical Legal Studies 3 (2006), 79–98. The two standing objections to a single 1941 break.
Craig Joyce, “The Rise of the Supreme Court Reporter”, 83 Michigan Law Review 1291 (1985), for the early Reporters and what they left out.
Library of Congress, United States Reports scans: tile.loc.gov/storage-services/service/ll/usrep/, used here as the second, independent reading of the two quoted 1808 and 1813 sentences.