How to Read a Declassified Document

The document is a legal artefact

A released page usually gets read as history. Here's what they knew, here's what they said. It's also a legal object, and nearly every mark on it is a claim somebody had to make under a rule, with a procedure for contesting it.

Learning to read those marks is worth an afternoon, because they tell you things the text doesn't. Which sentences were the sensitive ones. Whether the classification was a judgement or an inheritance. Whether the gap in front of you is national security or somebody's name. And whether what you're looking at is a released record at all.

The specimen above is constructed, nothing on it is real, and the eight features worth knowing are numbered.

The banner is the least informative mark on the page

The big line top and bottom gives the classification of the page as a whole, which means it gives the highest marking anywhere on it. A page banner-marked SECRET can be four-fifths unclassified.

The marks carrying information are the small ones in brackets at the start of each paragraph. (U), (C), (S), (TS). Portion marks, required precisely so a reader can tell what the classification was for.

This is the single most useful habit to pick up. When a document gets cited as proof that something embarrassing was secret, check the portion marks on the embarrassing paragraphs. If the damaging material is marked (U), it wasn't classified to protect it. Sometimes the sensitive thing is the sourcing, who said it or which intercept it came from, and the substance was never secret at all.

A redaction is an argument, not a void

Every withholding has to cite one of nine exemptions, and the code is normally printed on or beside the bar. That code is the agency's legal position, and an appeal contests it specifically.

Two bars of the same width are not the same kind of absence. A (b)(1) withholding is classified national-security material and has a shelf life, so it can be released later as the material declassifies, which is why requesting the same document again years later is worth doing. A (b)(6) withholding protects personal privacy and generally doesn't lapse while the person is alive.

The one that draws the most criticism is (b)(5), covering internal memoranda, the deliberative process privilege. It lets advice be withheld without any harm being demonstrated, which is why people call it the withhold-it-because-you-want-to exemption. Two things have narrowed it. The FOIA Improvement Act of 2016 added a foreseeable-harm standard across the board: an agency may withhold only if it reasonably foresees that disclosure would harm an interest protected by an exemption. The same Act provided that the deliberative process privilege cannot be applied to records created 25 or more years before the request. So if you're looking at a (b)(5) bar on a document from the 1970s, that isn't a lawful withholding, and saying so in an appeal is free.

Exclusions are a different thing from exemptions

An exemption withholds a record that exists and says so. An exclusion lets an agency respond as though the record doesn't exist at all.

There are three and all are narrow. Ongoing criminal investigations where the subject is unaware and disclosure would interfere. Informant records filed under the informant's name, unless the informant's status has been officially confirmed. And certain FBI records on foreign intelligence, counterintelligence and international terrorism, while their existence remains classified.

The practical consequence is that "no records responsive to your request" doesn't always mean no records. Usually it does. But the statute permits a particular class of answer that is technically true and substantively false, and knowing the category exists is part of reading the reply.

"We can neither confirm nor deny"

The Glomar response is its own category and the name comes from a real case.

In the 1970s a journalist asked the CIA for records of its efforts to discourage reporting on the Hughes Glomar Explorer, a ship built under cover of deep-sea mining to raise a sunken Soviet submarine. The agency wouldn't say whether such records existed, on the argument that the answer itself would reveal something exempt. The D.C. Circuit allowed it in Phillippi v. CIA in 1976, and the response has carried the ship's name ever since.

It isn't a refusal to release records. It's a refusal to confirm the question. And it's appealable like any other denial, most often by showing the existence of the records has already been officially acknowledged, which turns out to be a common situation once an agency has made a press statement.

How to tell a release from a transcription

The last feature on the specimen is the dullest and the most useful.

A genuinely released page carries release numbering. A Bates number, a document identifier, a page count, stamped at review rather than at creation. It carries a release stamp with a date and an authority. It carries banners. A page circulating as an image with none of that is not a scan of a released record, whatever the text on it says.

This matters because the commonest way a false document enters circulation isn't a forgery. It's a transcription. Somebody types out what a document is said to say, the typed version spreads, and the original either doesn't exist or doesn't say that. Release numbering is how you get back to the artefact, and it's how anybody else checks you.

The dates are worth attention too. A document declassified in 1998 went through a different review standard than one declassified in 2015, and the same page released twice can carry different redactions each time. Comparing two releases of one document is among the most productive things a researcher can do, and it costs nothing but the second request.

Where this goes next

All of the above assumes you have a document. Getting one is the other half, and the archive's guide to filing a FOIA request covers where to send it, what it has to say, what it costs, how long it really takes, and what to do when the answer is no, including the administrative appeal, which is free and succeeds more often than people expect.

The archive's own declassified-document collection indexes material from the National Archives, GovInfo and the CIA's CREST release, and matches it against the files here. Those pages carry their release numbering. You can check them the way this page describes, which is the point of describing it.