Rulings by the U.S. Supreme Court often come with great anticipation and attention, even true drama. Anxious crowds gather outside the court at dawn. Opinions first go out on paper to the waiting hands of television news interns, who sprint the documents to correspondents to be immediately deciphered on the air. Justices later announce their decisions in open court, and occasionally read aloud the opinions.

But when the court fixes mistakes in its opinions, it does so very quietly. No press releases. No public reading of corrections. For most of the court’s history, the justices have only signaled their fixes and edits by adding the word “modified” in small type to newly issued print and digital versions of the opinions.

The changes thus have proved hard to find — not just for the general public, but for lawyers and judges and scholars of the law.

Gabe Roth, executive director of Fix the Court, an advocacy group pushing for judicial transparency, thinks that’s a problem.

“The court does what it can to obscure its mistakes and to obscure some of the finer points of what they do,” Roth said.

Most changes in opinions are minor — fixing a fact or two, not altering a constitutional determination. But some of the changes made have been substantial.

One of the court’s most infamous rulings — Scott v. Sanford, which in 1857 held that African Americans whose ancestors arrived as slaves could not be U.S. citizens or sue for their freedom — was heavily altered shortly after its release. Chief Justice Roger Taney added more than a dozen pages to the Scott majority opinion, largely to defend against arguments another justice made in a furious dissent. Taney did not show the rest of the court his rewrite before publishing the final opinion.

In 2014, Richard Lazarus, a Harvard University law professor, exposed the court’s methods for handling mistakes in a law review article. Lazarus dove into archival files, the working papers of retired justices, and proofs of opinions to unearth the backstories for several altered opinions, including the Scott ruling. He argued the Supreme Court should better notify the public of its revisions since the opinions are widely referenced and quoted.

“Serious practical problems arise when the version of the court’s opinion upon which lower courts, other branches of government, and scholars and teachers rely can change, without notice, as many as five years after initial publication,” Lazarus wrote.

After the article, the court took a step toward greater transparency. Since 2015, the court’s website has flagged opinions that are modified after their initial release. It now also posts digital copies of the decisions that highlight edits.

However, this reform applies only to recent decisions. Finding earlier changes can be challenging, and details about the revisions — how the errors happened or were uncovered, for instance — are rarely in the documents.

This year, ProPublica fact-checked a sampling of the Supreme Court’s majority opinions from 2011 through 2015, and found several errors — including a couple of glaring inaccuracies in big decisions. A ruling that struck down part of the Voting Rights Act featured incorrect voter registration data. Another that helped reshape criminal sentencing relied on flawed studies claiming that judges were punishing defendants erratically. The bad information came from legal filings, from government records and from the justices’ own independent research.

To date, the court has not corrected those errors. ProPublica sent the court questions about its practices for revising opinions, but the court did not respond. Chief Justice John Roberts and the other justices have repeatedly declined to comment when ProPublica provided them with our findings. “As a matter of policy the court does not comment on its opinions, which speak for themselves,” Kathleen Arberg, spokeswoman for the court, wrote in a September email denying interview requests.

There will be little, or no, public notice if justices later fix the errant opinions. And there is slim chance the court will ever explain its revisions.

Federal laws dictate what happens to presidents’ and lawmakers’ work documents, which detail what government officials have done and why. Those records often become the public’s property.

But no statute or regulation governs justices’ internal records. Further, the court does not even have in-house policies for its files, Jill Lepore, a New Yorker magazine staff writer, reported in 2014. The justices themselves have complete control.

“They can shred them; they can burn them; they can use them as placemats,” Lepore wrote. “Texts vanish; e-mails are deleted. The court has no policies or guidelines for secretaries and clerks about what to keep and what to throw away.”

The court also has not permitted much access to the minimal information it does release about changes in opinions.

In the course of his research, Lazarus sought copies of the “change pages” that mark revisions in rulings for publishers. He obtained one year’s worth of the pages from an online service before the Reporter of Decisions of the Supreme Court, which disseminates opinions, directed the service to stop. The Reporter didn't respond to Lazarus’ direct requests for change pages until after his law review article published.

“My view is that they impeded my efforts,” Lazarus told ProPublica. “The Reporter’s view is that they were agnostic.”

Justices have occasionally been more open about fixing factual errors when the mistakes became embarrassing. But they never offer a full accounting of how the mistakes occurred, how they decided to correct them and whether they verified the rest of their facts.

In 2010, the court ruled in Graham v. Florida that juvenile crimes other than homicide cannot bring life sentences in prison without parole under the Eighth Amendment, which protects against cruel and unusual punishment. Justice Anthony Kennedy undertook his own research in the course of writing the majority opinion.

A study of life sentences for juvenile offenders by Florida State University professors became central to the court’s decision. It determined that such penalties were very rare. Only 109 juveniles were serving life sentences for crimes less than homicide nationwide, the study found, of which 77 were in Florida.

The study included data from 47 states. Lawyers for the state of Florida argued the research was not comprehensive nor independently vetted, and should not be trusted.

Kennedy sought to fill in the missing data. The Supreme Court’s library sent letters to officials in the states that did not provide information to the Florida State professors and to the U.S. Bureau of Prisons, asking for a count of juvenile life sentences for non-homicides. All provided information to Kennedy, and the justice added the new numbers to the earlier study’s tally. (He also counted an Oklahoma case based solely on a newspaper report.)

In the opinion, Kennedy wrote “there are 129 juvenile non-homicide offenders serving life without parole sentences.” A few days after the ruling, the Solicitor General’s Office, lawyers representing federal agencies, notified the court that its number was inaccurate. The prisons bureau had sent Kennedy a list of six inmates serving life inmates for juvenile crimes, but each of those convictions had involved homicide.

The justice used data he collected without double-checking it. Kennedy edited the opinion to subtract the federal inmates from his count and include a footnote acknowledging the error.