What we can deduce from a leaked PDF

In 1979, Bob Wood­ward and Scott Armstrong published The Brethren, a chron­icle of the Supreme Court during the tumul­tuous and conse­quen­tial terms from 1969 to 1975. Including, of course, the delib­er­a­tions around Roe v. Wade. I’ve recom­mended the book before—it’s my favorite work of legal jour­nalism.

At the time, The Brethren was contro­ver­sial. Despite the Supreme Court’s long­standing policy of secrecy around internal delib­er­a­tions, it was apparent that sources within the court had spoken to Wood­ward and Armstrong off the record. After the death of Justice Potter Stewart in 1985, Wood­ward confirmed that Stewart had been one of his key sources.

Thus, the bad news for those who contend that the recent leak of a draft Supreme Court opinion is “unthink­able” or, in the words of Chief Justice John Roberts, a “singular and egre­gious breach”—the horse is long out of the barn. Indeed, with so many more ways to securely leak infor­ma­tion these days, the only surprise in recent years is that there haven’t been more.

Much as I enjoy Wood­ward’s writing, his sources are not neces­sarily well concealed. One just needs to ask: “which person in this story takes the fewest hits?” For instance, in Wood­ward’s earlier book about the Trump admin­is­tra­tion, Fear, this line of thinking led inex­orably to former White House economic adviser Gary Cohn.

Cohn publicly ques­tioned the accu­racy of the book. Tellingly, he didn’t specify any partic­ular fact it had gotten wrong. In general, when sources deny jour­nal­istic reporting, I trust the jour­nal­ists, because there are still serious legal conse­quences for news orga­ni­za­tions that publish false­hoods; mean­while, no conse­quences at all for sources who issue blanket denials.

(This dynamic isn’t limited to polit­ical reporting. In 2018, Bloomberg Busi­ness­week published a story called “The Big Hack” that was vigor­ously denied by Apple and Amazon. Based on these denials, certain tech blog­gers became convinced that the story was false. The fact that neither Apple nor Amazon sued Bloomberg for defama­tion—despite being extremely rich, finicky, and liti­gious—made nary a dent.)

To be fair, this exchange of favors is not unique to Wood­ward. Rather, it’s a long­standing feature—or bug, some might say—of Wash­ington polit­ical jour­nalism. Much of the oper­a­tion of govern­ment is committed to the public record. But much more is not. Thus, leaks become currency, traded constantly. Without them, there would be no national polit­ical news.

So when you hear the cater­wauling—“egad, the leakers!”—assume it refers to the leaks that the cater­wauler finds unflat­tering. Although disclosing actual clas­si­fied infor­ma­tion is a crime, much infor­ma­tion about the govern­ment doesn’t fall into that cate­gory. In partic­ular, it doesn’t appear that leaking a draft Supreme Court opinion breaks any law. So the hot-blooded idea that the leaker should be “pros­e­cuted” is misplaced.

Not every leak is published, however. Over time, one of the reci­p­rocal favors that Wash­ington jour­nal­ists have offered is to plug certain leaks rather than publi­cize them. For instance, during his first 10 years on the Supreme Court—including the time depicted in The Brethren—Justice William Rehn­quist became addicted to Placidyl, a powerful seda­tive. Never­the­less, this fact was not mentioned in Wood­ward’s book, nor much other jour­nalism of the time. As best I can tell, the Wash­ington Post didn’t explic­itly connect Rehn­quist to Placidyl until after he had completed a detox program in early 1982. (Current Chief Justice John Roberts clerked for Rehn­quist during the 1980–81 term.)

Bringing us to this week’s leak by Politico of a draft Supreme Court opinion in the case of Dobbs v. Jackson Women’s Health Orga­ni­za­tion. I don’t usually comment on current events. But the possi­bil­i­ties for typo­graphic foren­sics were too intriguing to ignore.

Consis­tent with the Wash­ington jour­nal­istic prin­ciple of leaks-for-favors, I infer that whoever leaked this draft must foresee a benefit from the leak—as usual, cui bono?

There­fore, I don’t think the source is someone who works at the Supreme Court, like a justice or a clerk. Justices under­stand that they don’t always end up in the majority. Clerks rely on these jobs as a calling card for the rest of their careers. To be exposed as a leaker would amount to setting that future career on fire. It’s not worth the risk.

Though I’m not going to delve into the substance of this draft opinion, I believe it’s much more likely that the leaker is someone who supports the opinion rather than an oppo­nent—

So what can we tell from the docu­ment itself?

For thor­ough­ness, I ran the PDF through some meta­data checkers to see if there were any inter­esting tidbits left behind. There weren’t. Though I didn’t expect to find any, based on the appear­ance of the docu­ment.

How was it created? Let’s go in steps:

  1. An orig­inal color PDF was created on a computer using the US Supreme Court’s usual type­set­ting soft­ware. (And what is that? Programmer Faiz Surani noticed (perhaps unin­ten­tional) refer­ences in the Supreme Court Style Guide to a tool called “Opin­ions 2003”, which he spec­u­lated is a custom version of Microsoft Word 2003 used by the clerks for drafting opin­ions. This sounds plau­sible. For the type­set­ting and layout, designer Dan Rhatigan noted that the Supreme Court once used (and likely still uses) an XML-based publishing system made by Miles 33, appar­ently called OASYS. I’ve seen theo­ries else­where that LaTeX is involved—this wouldn’t surprise me either, because to my eye, the line breaking in Supreme Court opin­ions resem­bles that produced by the LaTeX algo­rithm.)

  2. It seems that the PDF was created on a modern computer and not with a different device because of the use of Arial in the upper right corner of the first page.

  3. It seems that the PDF was created in color because the yellow high­light around “1st Draft” is a rectangle that perfectly fits the text. Thus, the box must’ve been present in the digital file, and not, say, drawn by hand with a high­lighting pen.

  4. It seems the PDF was printed and stapled because of the pres­ence of staple holes on the top left corner of each page. The opinion is 98 pages, so that must’ve been a pretty big staple.

  5. It seems that the printed PDF was unsta­pled and then rescanned. Why? The reso­lu­tion of the page itself is very coarse and uneven, which is a kind of typo­graphic degra­da­tion char­ac­ter­istic of sheet scan­ners. Further­more, the pages have been scanned at different angles, which indi­cates the use of a low-volume home-office device. A typical office scanner would have an auto­mated sheet feeder that would keep the sheets in a more uniform vertical orien­ta­tion.

  6. The text of the PDF is search­able because OCR was run on the PDF after it was created. Perhaps by the leaker, but more likely by the recip­ient, Politico.

It’s possible that Politico received the printed docu­ment and made their own scan. If that were the case, however, I’d expect them to have better quality scan­ning equip­ment and produce a nicer PDF.

But Politico has a strong incen­tive to protect their source. By making their own scan from a paper orig­inal, they wouldn’t open them­selves up to the disclo­sures of confi­den­tial infor­ma­tion that have tripped up others. (That said, printed docu­ments are not neces­sarily free of meta­data, as Reality Winner found out the hard way.)

Is it possible the docu­ment was scanned twice—once by the leaker, once by the publisher? I don’t think so. If it had been, I’d expect to see more pecu­liar pixel-level arti­facts and distor­tions.

So what does the state of the PDF tell us about the iden­tity of the leaker?

Of course, I’m prob­ably wrong. Best of luck to Chief Justice John Roberts on his inves­ti­ga­tion.

PS—should court delib­er­a­tions be confi­den­tial? Or should justices be required to post drafts of opin­ions to the judi­cial equiv­a­lent of GitHub?

On the one hand, we can see the virtue of a certain kind of delib­er­a­tive veil for written opin­ions: it allows the members of the Court to freely explore ideas and alter­na­tives, some of which may be strange or unrea­son­able, on the way to a final result.

On the other hand, oral argu­ments are out in the open. Partly because we the public are inter­ested in how the Supreme Court justices are framing the issues, and where they detect strengths and weak­nesses in the parties’ argu­ments.

Oral argu­ments happen in real time, however. The justices’ ques­tions are under­stood to be a tool for sharp­ening the presen­ta­tion, not for revealing their own posi­tions (though infer­ences are made anyhow).

Majority opin­ions, by contrast, repre­sent the offi­cial output of the Court, and thus neces­sarily should evolve more slowly. The delib­er­a­tive veil is a wise policy specif­i­cally because it promotes open-mind­ed­ness and flex­i­bility among the justices while allowing for a high stan­dard of output. (Most famously: the case where Chief Justice Roberts wrote both the majority and minority opin­ions.) Justices who had to put every draft in a public loca­tion would become more cautious.

Finally, as a typog­ra­pher, I think the Supreme Court’s apparent habit of type­set­ting draft opin­ions as if they were final is bad policy. I’ve cited the Supreme Court’s typog­raphy as the best in the country. I still think so. But part of the reason the Supreme Court’s typog­raphy works so well is because it visu­ally connects an opinion to a centuries-long tradi­tion of delib­er­a­tive thought. When that typog­raphy is applied to a draft opinion, it gives the ideas therein a grav­itas and authority they haven’t yet earned. Perhaps the people within the Supreme Court aren’t affected by such suppos­edly cosmetic consid­er­a­tions. But I’m certain that if this leaked PDF didn’t look like a finished Supreme Court opinion—say, more like this—the public recep­tion would’ve been quite different.

update, 82 days later

As for govern­ment using GitHub, US sena­tors Cynthia Lummis and Kirsten Gilli­brand recently put a draft of their proposed cryp­tocur­rency legis­la­tion on GitHub for public comment. This is the first time legis­la­tion has been presented in this manner, and it’s going the way you would expect, with sugges­tions like “add mario from mario 64”.

update, 198 days later

The New York Times reports that Rev. Rob Schenck claims he learned the outcome of a 2014 abor­tion-related opinion, Burwell v. Hobby Lobby, three weeks before it was released. That opinion was also written by Justice Alito:

In early June 2014, an Ohio couple who were Mr. Schenck’s star donors shared a meal with Justice Alito and his wife, Martha-Ann. A day later, Gayle Wright, one of the pair, contacted Mr. Schenck, according to an email reviewed by The Times. “Rob, if you want some inter­esting news please call. No emails,” she wrote.

Let’s compare that to my predic­tion for how the Dobbs leak was accom­plished:

I’d suppose it’s a friend, spouse, or family member of a Supreme Court justice who has consis­tently opposed Roe v. Wade, acting with some­thing between autonomy and plau­sible deni­a­bility.

For his part, Justice Alito issued a denial to the NY Times:

[The] alle­ga­tion that the Wrights were told the outcome of the deci­sion in the Hobby Lobby case, or the author­ship of the opinion of the Court, by me or my wife, is completely false.

Unfor­tu­nately, this is the kind of denial that raises more ques­tions than it answers due to the delib­er­ately narrow phrase “were told”. The denial would remain true even if, say, Ms. Alito had put a copy of the draft opinion on the table, allowed Ms. Wright to look it over, and then taken it back—no “telling”, just showing.

To be clear—if my hypoth­esis turns out to be correct, it will be one of the worst days ever for the Supreme Court and federal judi­ciary, and a terrible day for the United States. Though I followed the evidence where it led, I still would’ve preferred to be wrong.

update, 259 days later

The New York Times reports that the Supreme Court inter­viewed 97 employees and found no evidence that any had leaked the draft opinion. Though the article also notes that this inves­ti­ga­tion of “employees” did not include the justices or their spouses. Mission accom­plished? (The full Supreme Court report is here.)

update, 260 days later

In response to the welter of thinking-face emojis that were emitted in response to yesterday’s news that the inter­ro­gated “employees” did not include justices or their spouses, Supreme Court Marshal Gail Curley clar­i­fied that she “spoke with each of the Justices” but that none of the “cred­ible leads … impli­cated the Justices or their spouses” and thus she did not “ask the Justices to sign sworn affi­davits.” Makes perfect sense?

update, 262 days later

Another well-reported New York Times piece on the short­com­ings and snow­balling conse­quences of the Supreme Court’s leak inves­ti­ga­tion. This is fine?