Editor’s note: Barton Gellman delivered the following lecture at Princeton’s Woodrow Wilson School in October 2003, shortly after his reporting on intelligence and the war in Iraq. Gellman’s lecture takes on a renewed relevance in the wake of his recent reporting on U.S. intelligence-gathering operations. Part 1 of Gellman’s lecture is also available.

If you came to the first of these lectures, you know that I spend much of my professional life in pursuit of secrets. I was thinking as I wrote that line that it would scan better if I said “other people’s secrets,” but that would not be true. The government has no private claim to secrecy; its secrets are held on our behalf.

When the government stamps something classified, my editors and I take that seriously. We treat it as a yellow light, but not a red light. We draw lines, but we draw them ourselves. Tonight I’ll describe how we have done that in actual cases.

For those who missed the chance last month to ask me, “Who elected you?”,  I’ll just repeat that I claim no special authority as a member of the Fourth Estate. My reasons for thinking that what I do is honorable, and socially valuable even, apply equally to any citizen with a weblog. I can describe those reasons in 191 words:

National security secrecy presents a conflict of core values – self-government and self-defense. If we don’t know what our government is doing, we can’t hold it accountable. If we do know, our enemies know too. That can be dangerous. That’s our predicament.

Wartime heightens the case for secrecy because the value of security is at its peak. But secrecy is never more damaging to self-government than in wartime, because making war is the very paradigm of a political choice.

No individual, and no institution, can be trusted to draw the line for us when these two interests collide. That includes the people with the classified stamps. Newspapers cannot appoint themselves as arbiters of national security. Political leaders, on the other hand, cannot be allowed to decide for us what we need to know about their performance.

In practice today, secrets are kept or broken by a process of competition. Governments try to keep them; journalists try to find them out. When The Washington Post turns up something sensitive, we consult with authorities before publishing. Often we agree to hold something back. This system works fairly well, and I see no acceptable alternative.


Tonight I’ll talk in more practical terms. First I’ll tell you how I find out things I’m not supposed to know. Okay, don’t take that too literally. You’ve heard the term “sources and methods.” It comes from the National Security Act of 1947. Intelligence agencies won’t talk about them. I can be about 50 percent more forthcoming. (I’m thinking of putting that on my business card: “50 percent more forthcoming than the CIA.”) I can’t be too precise about my sources, but I can talk pretty freely about methods.

Washington is consumed at the moment by a scandal over “leaks.” I’ll reserve the Wilson case – no relation to Woodrow, by the way – for Q&A. But I do want to talk about leaks, because I really dislike the metaphor. And I think the metaphor matters, because it colors a lot of commentary on reporters and sources.

Our everyday discourse is full of dead metaphors. You know what “scuttlebutt” means in its figurative sense, but could you find one on the deck of a sailing ship? A “leak,” on the other hand, is something we feel we understand.

Leaks come from plumbing. Nixon had leaks during Watergate. He sent Plumbers.

So you picture a big tub of secrets. The seal is good and watertight, but someone sneaks in and turns the tap. And now look, there’s a puddle all over the floor.

Who made that mess? If you think in terms of leaks, the answer is clear. It was the Leaker who turned the faucet, and in that sense the Leaker is the true author of the story. The reporter – shall we say Leakee? – is a conduit. No one puts it quite that baldly, but that is the necessary premise of a favorite Washington parlor game, Name That Leaker. Maybe you’ve played it. You read a story between the lines. What agenda is being advanced? Cui bono? Who, by inference, must be the Leaker? If you like, you can throw in a further deduction about the politics of the reporter.

This is a bad model of a complicated transaction, and it leads to bad surmise.

I’ve already mentioned the Conduit Fallacy. Good reporters – and by this I mean most reporters at serious news organizations – do not transmit a source’s claims unexamined. One difference between reporters and readers is that we do know who we’re talking to. This is not something I boast about; I’m as unhappy as anyone about the ubiquity of unnamed sources in the published product. My point is simply that we understand where our sources fit into the organizations and events we write about. We usually know their motives and biases, and, armed with that knowledge, we test their assertions against other evidence. PR professionals make a living pretending otherwise, but it just isn’t so – or hardly ever is so – that the story you read is the mere instrument of a source.

Some of you might think me naïve. I’m describing a professional ideal, but sources can undermine that by flattering a reporter, cutting deals for access, playing to the fear of being scooped or the hope for glory. What that misses is that reporters do the same – we make good use of vanity, grudges, office politics, and the wishes of our sources to promote an agenda. All that is part of the real world of newsgathering. So too, and just as common, is an idealistic drive of reporters and officials to do their jobs well and according to norms. In the end I don’t care all that much why someone talks to me, as long as I understand the motive and can verify what I learn.

There is another fallacy we could call Drop-a-Dime. The expression comes from ancient times, when it cost ten cents to use a payphone. The intimation here is that a source picks up a phone and says, “Wait ‘til you hear this one.” Now, undeniably, that sometimes happens. It may have happened in the Wilson case. But it is very far from typical of our daily work. I have long since given up on finding plain brown envelopes in my mailbox. In 15 years at The Washington Post, I can count on maybe two hands the number of good stories I’ve learned from someone who made contact out of the blue. And most of those calls resulted from a story I had already published. Unearthing information is a little bit like borrowing money: you need some to get some. By far the most common way we learn new things is the iterative work of beat reporters. I don’t mean this in some grand investigative sense. It’s routine. It can be as simple as a political reporter’s instinct, from bits and pieces collected over weeks, that Bob Graham is preparing to drop out of the presidential race. Ask the right person at the right time, and you may well get an answer. This is what the folks in Langley, Virginia, call “assembling the mosaic.”

Another false premise, an important one, could be called the Big Picture Fallacy. There are an awful lot of readers – including some who make their living on cable TV – who interpret every news story as an episode in a central running debate. It doesn’t have to be only one debate, though it often boils down, in the mail I get, to “Bush: Statesman or Menace?” With just a few more topics you can cover a lot of ground on the average front page. For example, “Iraq War: Good or Bad?” “Economy: Whose Fault?”

If everything boils down to one big question, you can divine the source and his or her agenda by the impact you think the story has on your cause. That is a cartoon. Sometimes the story is actually about the story.

And in a given story, it’s surprisingly hard to guess what interest a source is serving. An anonymous campaign official who seems to hurt one candidate may be working, as you might think obvious, for the other. But maybe not. Maybe the source has committed a “gaffe,” as Michael Kinsley famously defined the term in 1988: an inadvertent blurting of the truth. Maybe the source cares more about his relationship with the reporter than with his candidate. Maybe the source thinks his candidate is a fool. Maybe – this does tend to be related – the source is angry that his boss didn’t listen to sage advice. Maybe he’s not thinking about who should win the race, but something that matters more to him that day – whether the campaign gives him the money he wants for TV ads or get-out-the-vote, or which of several conflicting interest groups to cultivate, or who gets to be in charge of the campaign. You can learn about someone’s motive if you know who the source is, but it’s a mistake to think you can guess the source and motive if you know neither.


So how do I find out government secrets? I could dignify the process as “reporting by hypothesis,” but I like the military term better. The term is “SWAG.” It’s an acronym – what else?– and it stands for “scientific wild-ass guess.” I do a lot of guessing. And no, I don’t put guesses in the paper. I use them to guide my reporting. The metaphor I like best is pulling on threads.

A lot of what I learn is off the page. I’m unlikely to put it in the paper, not because it’s a secret but because it doesn’t fit the conventions of a news story. It has to do with organization charts, the language and background of specialists, the trajectory of careers, the detailed process of bureaucratic decisionmaking, the equipment and methods of people who do things I write about, the history of an issue, the relationships and rivalries of individuals. I go to conferences and collect collect obscure reference works. The Army field manual on helicopter maintenance helped us follow both Iraq wars, for instance.

This kind of background helps me guess, from little clues and emanations, what might be happening now. It also helps me figure out who would know it if my guess is right, and who might have a reason to talk.

Occasionally the background becomes part of the story, necessary to explain what is happening or why. In my first newspaper job, covering the D.C. Superior Court, I got a scoop about a notorious murder trial from two words spoken cryptically in open court. The judge called for legal briefs on a “bifurcated trial,” and then recessed. I ran back to my office and looked that up. In that context it seemed to mean the defendant was going to admit to the shootings but plead insanity. I used that knowledge to press for confirmation, and got it.

What we call “enterprise reporting” almost never comes from one or two sources who spill the whole thing. It comes from an accumulation of small facts that lead eventually to the big fact in our lead.

A lot of my guesses are wrong. If I spent my time trying to prove them, rather than explore them, I would be breaching a basic obligation to readers. I abandon far more guesses than I confirm.

In many cases, the ultimate story I’m reporting is classified, but there are unclassified signs around the edges. In the early 1990s, I suspected that the Clinton administration was going to shift course and intervene in the Balkans. I asked myself, which combatant command would take the lead? Had it formed a joint task force of the kind that runs modern military operations? Had the task force stood up a 24-hour operations center? Had the units cancelled leaves and called back personnel? All those things were discoverable, unadvertised but unclassified, and they gave me strong reason to think I was on the right track.

Sometimes silence is meaningful. You know that scene in the war movie where the guy who’s about to die says, “It’s awfully quiet out there”? When organizations stop answering questions they usually answer, and when old acquaintances stop returning calls, it tells me something. When I ask something sensitive, the way my interlocutor does not answer hints at whether I’ve touched a nerve.

When I think I’m onto something, I make a list of everyone who might know whether it’s true. Sometimes I know the person, sometimes I know the name, and sometimes I know only the job title. I’m a collector of phone books and lists. I ask myself, who cares least about protecting some small part of this secret? An Energy or Defense or Justice Department official may not know or care that a given detail is diplomatically sensitive. A career official at State may not care about the political implications for the White House. Once I have all I can get from those who have the least stake in a subject, I begin to ask questions of those who have the most.

Hardly anyone in government is comfortable about explicitly crossing the line into classified material. Sometimes a person will rationalize it with the notion that he is saying something I already seem to know. Sometimes the person thinks the subject is of overriding importance.

Last year I wrote a story about al Qaeda’s efforts to acquire biological weapons. White House and CIA spokesmen told me categorially that there was no U.S. intelligence assessment that al Qaeda was looking for smallpox. I had heard otherwise. I went back to my original source, who had carefully skirted the line on classified material. The source got angry. I said, the only way to resolve this is for you to read me the language of the document. And that’s what the source did. It was a Powerpoint briefing for White House senior officials. It said: “Al Qaeda is interested in acquiring biological weapons, to include smallpox.” I went back to the spokesmen. Oh, they said, you mean that assessment.


I promised to talk about how we decide what to publish and what to hold back.

It’s easiest to talk in detail about older cases, so I’ll begin there.

In the 1991 Persian Gulf War, I was skeptical of Gen. Schwarzkopf’s briefings on bombing progress. From one week to the next he increased his estimate of the daily damage to Iraqi tanks by a factor of ten. How could that be? Most of the tanks were buried, hard to find or hit. I learned that pilots had found a new way to use the FLIR, or forward looking infrared sensor. Usually they search for hot spots. The pilots found, though, that armor sheds heat at a different rate than desert sand. If they looked soon after sundown, they could find tanks by aiming at cold spots.

The Washington Post did not seriously consider publishing that story. We did not bother to consult with the government. We just sat on it. I was sorry to give up a scoop, but this was obviously a technique to which Iraq could take countermeasures. Publication would do concrete harm to the war effort, and it served no grand public policy interest to disclose it. I do so now because it has since come into the public record.

A more recent example, and more complicated. Last year my colleague Sue Schmidt and I learned that President Bush had deployed what we called, in shorthand, a “shadow government” of senior officials into underground bunkers far from Washington. There had been contingency plans for this all through the Cold War, called COG/COOP – continuity of government, continuity of operations plans. Bush was the first president to activate them.

This was a watershed. For the first time a president was saying, because of al Qaeda, that he could not be sure that Washington would be here tomorrow. It spoke volumes about the new insecurity of a post 9/11 world. And the whole thing was very highly classified, top secret codeworded information.

When the government learned I was asking questions, White House chief of staff Andrew Card called the executive editor of The Washington Post. He said he couldn’t believe the Post would publish such a thing, and if we seriously contemplated doing so he wanted an opportunity to be heard. Len Downie, the editor, called me. I phoned Andrew Card’s office the next day and said, here’s your opportunity to be heard. His deputy invited me to pay a visit.

I asked what, exactly, the government sought to protect. He said everything. I said I didn’t think that would fly, and I had the impression he did not expect it to. We talked some more. He cared most that we not disclose the sites of the bunkers, the names of those deployed, and the mechanics of the deployment. I told him, honestly, that I thought he had good reasons for concern, and I thought I would agree, but I wasn’t completely convinced. Details are vital in a story like this. Extraordinary claims require extraordinary evidence. If we are going to break something big, we need to show readers we know it’s true. These bunkers had been designed to withstand a hydrogen bomb. What exactly, I asked, did the White House think Osama bin Laden could do to them?

On the other hand, I knew already from my executive editor that there was no way we would publish the details. But it was the uncertainty on that point that got me into the chief of staff’s office, and I maintained it in part to continue the conversation. I said, I’m sure you don’t think I came here only to take things out of the story. What can you tell me? I learned a few things, including the numbers of those deployed. Once he was sure we would run the story, he gave me an on the record quote. Only for that reason can I tell you I was talking to Joe Hagen. The on the record quote served his purpose, but it also made our story much more credible.

A related case. In December I learned that the Energy Department’s national labs had undertaken a crash deployment of a prototype system to detect nuclear materials entering the nation’s capital. It was a distributed network of sensors called, aptly, Ring Around Washington. It didn’t work. Again, very highly classified. The story I was writing, a long one, asked the question: Are we safer after 14 months of war with al Qaeda than we were on 9/11? Ring Around Washington was highly relevant. I consulted with high ranking officials I can’t name. They wished we would not mention the Ring at all. What they really cared about, though, was that we not describe exactly why the system failed – how it could be defeated. I proposed a very general way to describe the flaws, and after a while we came to a formula we all could live with.

For the same story, I learned something I can’t describe here at all. It was a really nice find, something with considerable news value in the context of that story. All I can say is that it had to do with a technique al Qaeda was using to elude its U.S. pursuers, and it spoke to the cleverness and innovation of our enemy. I spoke directly and at length with the head of one of the 15 U.S. intelligence agencies, and we worked out a form of words I could use. But then Condi Rice’s office called. She asked for a conference call with me and with Len Downie, the editor. She said she could not live with the compromise, and she said a little bit about why. I tried to find another form of words. She asked us to drop it entirely. She very kindly said she thought I’d have an awfully good story anyway. And we dropped it.

The most complicated example.

It had to do with Unscom, the U.N. arms inspectors in Iraq through the 1990s. In August of 1998, I learned – and this was the fruit of a SWAG, a guess – that the U.S. government was quietly urging Unscom to back off. I described a phone call in which Madeleine Albright persuaded Richard Butler, Unscom’s executive chairman, to rescind his order for a surprise visit to the headquarters of Saddam Hussein’s special security organization. Washington was even then professing support for anywhere, any time inspections, and threatening the use of military force to compel them, but it had lost backing for that position in the U.N. Security Council. Albright tried to have it both ways, and I showed that.

Governments find it useful, often in good causes, to say conflicting things in different forums. I am in the information arbitrage business. I don’t collaborate in that effort. We believe in my business that the truth, an accurate depiction of the world as it is, has elemental value. We will not conspire to hold it back in support of some particular diplomatic result. Unscom was dying. Saying so may or may not have sped the death, but staying silent would not have saved it. We would probably not have stayed quiet regardless.

As I traced the death throes of Unscom, I discovered its extraordinary development into the first – and probably last – U.N. intelligence agency. It was actually improvising high-technology spy tools against Iraq. The first time I wrote about that, authorities told me I would put the lives of inspectors and clandestine operatives at risk if I included details. We compromised on the following language: “inspectors deliberately triggered Iraq’s defenses against a surprise search and used a new synthesis of intelligence techniques to look and listen as the Baghdad government moved contraband from the site.” A bunch of mumbo jumbo, and deliberately so.

I knew a great deal about the operation, and I sat on it for months. But Kofi Annan’s office started hearing rumors, and Annan assigned a competent investigator to learn more. Anything that smacked of espionage against a member state represented a huge threat to the U.N. system, as he saw it. In January of 1999, I told my sources that the story was beginning to seep out. Le Monde, al Hayat and the Boston Globe were pursuing it. On January 6, with notice to authorities, I wrote some of what I knew: Unscom had used evesdropping equipment, carried by inspectors, to monitor communications that Iraq knew were safe from satellites. I knew the type of equipment, the identities of the inspectors, even the radio frequencies. I pursued those details to be sure my sources knew what they were talking about. We never considered publishing them.

A few months later, I discovered the most stunning aspect of the story. There had been yet another level of espionage. The U.S. government planted listening devices in Unscom equipment to spy on Iraq in ways that Unscom itself did not know about – and that had nothing to do with Unscom’s mission. All those years, Unscom said Iraq was hiding weapons, and Baghdad said Unscom was a nest of spies. It turned out that both sides, more or less, were right. The CIA told me that there were clandestine operatives still in Iraq, and asked for time to get them out if we planned to publish. We waited. Then we published. That was a hard decision – it is possible that we stopped a productive intelligence operation – but I think it was the right one on balance. I’m prepared to defend it if you like, and I’ll turn this over now for questions.

Thank you.