LEARNING TO SPY; ANNALS OF INTELLIGENCE

ELSA WALSH. The New Yorker

Philip D. Zelikow is a professor of history at the University of Virginia and an expert on the predicaments of government. He worked on the National Security Council for the first President Bush and wrote a book on German reunification with his former colleague Condoleezza Rice. Recently, he took on a more demanding assignment: he was appointed the executive director of the 9/11 Commission, and oversaw its staff of eighty-two.

The commission's report, which was released this past summer, tried to explain why the United States was unable to protect itself from the attacks of September 11, 2001, and it made recommendations on how to better safeguard the nation. A particularly delicate task was analyzing why the Federal Bureau of Investigation, despite its storied history in law enforcement, failed as a domestic intelligence agency. "There were some things about what the F.B.I. had become that were just really indefensible," Zelikow told me in a recent interview. The question facing the commission was simple and far-reaching: Could the F.B.I. be reformed? Or, as many commission members had argued at the outset, should a domestic intelligence agency modelled on the British M.I.5 be established, limiting the bureau's responsibilities to fighting crime?

The F.B.I. director, Robert S. Mueller III, had started work a week before September 11th (he replaced Louis Freeh, who had resigned in June, 2001), and he quickly realized that the bureau had a major problem. "He understood that the F.B.I. did not fundamentally have the mind-set, institutional culture, or organization to collect and disseminate intelligence," Zelikow said. Zelikow was not referring to the bureau's publicized missteps: from revelations about sloppy practices at its forensic laboratory in the nineteen-nineties to the latest embarrassment--a report by the Justice Department's inspector general, released in late September, which concluded that tens of thousands of hours of taped material related to terror investigations have not been translated and, in some cases, have been erased. Rather, Zelikow said, Mueller--and, subsequently, the commission--was alarmed by the bureau's basic failure to provide usable intelligence.

"They literally didn't write intelligence reports," Zelikow said. Instead, agents would describe their interviews in a report called a "302," and put it in a case file. These reports sometimes were lost because the bureau's computers were several generations out of date and there was no reliable cross-indexing system. The other issue, Zelikow said, is that F.B.I. agents "don't ask questions the way an intelligence agent would ask questions. The agent is typically interested in the facts of an event. An intelligence agent is really interested in a person's whole world."

The F.B.I., which began in 1908 as a national detective force of thirty-four, now employs about twelve thousand agents. The commission, in its report, pointed out that the bureau "has long favored its criminal justice mission over its national security mission." F.B.I. agents, trained to gather evidence in criminal cases, tend to think like policemen--operating after the fact. Intelligence, almost by definition, deals with ambiguity. "F.B.I. people generally think that the intelligence people are creative thinkers who play fast and loose with facts . . . willing to be freewheeling and speculative without being unduly burdened by rigorous attention to evidence," Zelikow said. Jamie Gorelick, a commission member and a deputy attorney general in the Clinton Administration, told me that trying to get the F.B.I. fully integrated into the national-security apparatus "was probably the greatest source of frustration for me during my entire tenure at the Justice Department." The evolution of the bureau into a counter-terrorism agency that continues to investigate major crimes, such as kidnapping or public corruption, has become Mueller's biggest challenge.

In private meetings with Mueller, the commissioners had been brutal, but they were impressed by his willingness to revamp the bureau. Mueller, Zelikow said, gave the commission "unprecedented access to bureau employees and bureau files." The commission's recommendation that the bureau be left intact, the report said, depended on the F.B.I.'s ability to make "an all-out effort to institutionalize change." The conventional wisdom, Zelikow said, is that "the F.B.I. was the poster child for the broken agency," even though the C.I.A.'s performance in assessing and coordinating intelligence before September 11th was far worse. Mueller and the bureau were undoubtedly helped by personal contrasts between Mueller and the C.I.A.'s director, George Tenet, who was less willing to acknowledge management failures. In Tenet's testimony, Zelikow said, there were "a variety of important issues on which there was little or no recollection." Zelikow also said, "We didn't believe him anymore." (Tenet, who announced his resignation in June, said he was "outraged" by that characterization of his testimony, adding, "I told the truth about everything I've done." Several commission members have defended him; former Senator Bob Kerrey called Zelikow's criticism "completely unfair.")

As a signal of the F.B.I.'s commitment to major change, Mueller, in May of 2003, hired a former Russian linguist named Maureen A. Baginski to run the bureau's new Office of Intelligence. The job is something of a work in progress, but if the bureau takes the next step and establishes a separate Directorate of Intelligence--the "service within a service" which Mueller has asked for and the commission has endorsed--Baginski would head it. In the event that Congress creates a Cabinet-level "intelligence czar," Baginski may report simultaneously to Mueller and to the intelligence chief.

Baginski had spent the previous twenty-four years at the National Security Agency, the nation's largest and most secretive intelligence service, which focusses on intercepting intelligence from abroad. In her last post at the N.S.A., she was the head of Signals Intelligence, or sigint, which oversees the interception of electronic transmissions--the heart of the N.S.A. operation in Fort Meade, Maryland. "She became a personal symbol of Mueller's reinvention of the F.B.I.," Zelikow said.

Mueller had tried a number of fixes in the year and a half before he hired Baginski, and perhaps his most fundamental reform was to centralize counter-terrorism investigations at F.B.I. headquarters. Investigations had formerly been run from the bureau's fifty-six field offices, where the Special Agents in Charge often set their own agendas. Commission members were startled to learn that neither Mueller's immediate predecessor, acting director Thomas Pickard, nor his top deputies had been aware of the detention of Zacarias Moussaoui, the suspected "twentieth hijacker," by the bureau's Minneapolis office on August 15, 2001, although Pickard had called every Special Agent in Charge that summer warning of new threats.

Some reforms had the paradoxical effect of producing more intelligence than the F.B.I. could absorb. Mueller had tripled the number of intelligence analysts assigned to counter-terrorism, from a hundred and fifty-nine to four hundred and seventy-five. (Today, there are nearly fifteen hundred intelligence analysts, the majority of them assigned to counter-terrorism.) In the past, analysts were often promoted from support-staff positions; to improve their calibre, Mueller created the College of Analytical Studies, a seven-week training program modelled in part on a C.I.A. course, at the bureau's academy, in Quantico, Virginia.

Passage of the U.S.A. Patriot Act, in October of 2001, greatly expanded the surveillance powers of law-enforcement and intelligence agencies, and helped to tear down the "wall," shorthand for restrictions that kept criminal and intelligence investigators from sharing information. As much a "cultural edifice," as Zelikow put it, as a legal barrier, the wall took shape in the nineteen-seventies, after the exposure of F.B.I. abuses of power, such as the notorious harassment of Martin Luther King, Jr. Its elimination so increased the flow of information that one senior F.B.I. official likened it to "trying to get a drink of water from a fire hydrant." As a result, Mueller began assigning many more agents to the counter-terrorism center. Larry Mefford, who until last October was the bureau's head of counter-terrorism and counter-intelligence, said that the level of stress became intense. "We had a couple of heart attacks of employees," he said. "I had one employee commit suicide. His wife said that he was very concerned about missing something." Mefford, who is now in charge of global security for Wynn Resorts, added that the task was overwhelming: "We had to reengineer the F.B.I. We had to fight the war"--on terror. "And we had to justify everything we were doing."

The Defense Department had started its own ad-hoc domestic-intelligence operation shortly after September 11th. At first, this was simply an attempt to collect information about the security of military facilities, but the operation expanded, and the Pentagon often passed on tips--sometimes picked up overseas--to state and local officials. "It created a huge problem for the F.B.I., because the military unilaterally would share raw intelligence that was totally uncorroborated or, worse yet, absolutely wrong," Mefford said. "It could be everything from you name it--e-mail threats that came in, telephonic threats. Or it could be stuff brought from overseas that the military got totally out of context--that somebody in a bar in Hamburg, Germany, talked about blowing up a dam near Las Vegas. And, boom--it was in enforcement circles, and before we could get a handle on it to see if it was credible or not we had everyone going crazy."

I met with Maureen Baginski, who is forty-nine, several times during the past year, and one Saturday in June I visited her at the house that she and her second husband, a microbiologist and Russian linguist, had built four years ago. The house, about fifty miles from Washington, sits on twelve acres that the couple have been clearing and turning into a series of gardens overlooking a creek. A collection of girl action figures given to her by her staff--Xena the Warrior Princess and Wonder Woman among them--are perched on a bookshelf in her home office. For Baginski, the house is a place to escape and relax; on weekends, she sometimes cuts back brush with a chainsaw, a gift from her husband. On an ordinary workday, she leaves home at 4 a.m., and usually gets back at around 8:30 p.m.

Sitting at her kitchen table, wearing black jeans and a lavender T-shirt, Baginski told me that in her first few weeks on the job she had an "absolute crisis of confidence." She had been startled by how different the bureau seemed from the N.S.A. She thought of the F.B.I.'s Office of Intelligence as a "start-up company," and at one point said, "There was nothing there."

After a month of initial panic, Baginski said that she called in representatives from every section and told them that she wanted to capitalize on the F.B.I.'s investigative strengths--its attention to solid evidence, or what she called "the pedigree of the source," and the willingness of agents and analysts to acknowledge when they needed more information. "There's no inherent shame here in not knowing something. That just means go and get more so you can build. But in the intelligence community they are a little less comfortable telling you what they don't know."

After ten weeks, Baginski and her staff began to try to reinvent the bureau's intelligence program. "I draw a very strong parallel to my challenge at N.S.A.," she said, adding, "I was telling somebody the other day, 'I guess I'm incapable of new ideas, because I actually do realize I'm doing exactly the same thing at the F.B.I. that I did there.' "

Baginski started at the N.S.A. in 1979, as a Russian-language instructor. She had studied Spanish and Russian at the State University of New York at Albany, not far from where she grew up, the fifth of six children: her father had run the telecommunications center at suny-Albany after a career as an actor and a high-school English teacher. In 1976, Baginski spent a year in Moscow as an exchange student. The harsh conditions of Soviet life made her appreciate America, she said, and she returned determined to work in public service. After a short stint teaching, she was hired by the National Security Agency. She rose quickly to the agency's most senior positions, and, in 1997, Lieutenant General Kenneth Minihan, the N.S.A.'s director at the time, asked her to become his executive assistant, a job akin to chief of staff; in October of 2000, Lieutenant General Michael Hayden, who succeeded Minihan, gave her the job of heading Signals Intelligence.

Like the F.B.I., the N.S.A. had resisted change, in part from institutional inertia but also from a lack of resources; during the nineties, after the collapse of the Soviet Union, Congress had steadily cut the N.S.A.'s budget by about a third. The agency was, and to some extent remains, obsessed with classification and compartmentalization; words like "warlord" were frequently used to describe the territorial proclivities of its managers, who functioned with much the same autonomy as Special Agents in Charge do within the F.B.I. The technology used to collect data, for example, was often seen as being more important than the data--and the overriding impulse was to hold on to information and not share it. Even such essential functions as collection and analysis were kept separate. "You have to understand the culture here," Lieutenant General Hayden told me. "Security is very important--watertight door, so to speak--so if you spring a leak in one place you don't sink the whole ship."

Hayden wanted Baginski to remake and reorganize the N.S.A., and Baginski took on the assignment with great energy. She pushed her technical team to find new ways to monitor the Internet, something that the N.S.A. had resisted. She also helped to make "geolocation"--the precise, geographic location of a target's position--a deadly tool. She urged representatives from the collection and analytic units to be more responsive to other agencies, such as the C.I.A. and the Defense Intelligence Agency, which over the years had complained about the N.S.A.'s inattention to their requests. All this, in hindsight, seems like common sense, but at the time it was the equivalent of an organizational shakeup. "She was trying to break people out of 'This is what I do and this is only what I do, and I'm really good at it and leave me alone,' " Hayden said.

The N.S.A., and its signals-intelligence operation, had been built on a near-legendary ability to eavesdrop on the world, but by the turn of the twenty-first century sigint was functioning as if the Soviet Union still existed and the information age did not. America's enemies no longer had a fixed address, and they had learned to use new technologies and encryption techniques. Yet, according to Michael Wertheimer, who held the most senior technical position at the N.S.A. at the time and was part of a team that Baginski had asked to assess sigint capabilities, there were employees in the field who were still focussing on ancient technologies, such as what was called "manual Morse code"--people speaking dots and dashes out loud. "About seventy per cent of our resources were going toward traditional industrial-age stuff"--put up an antenna, aim it at the sky, and vacuum up everything you could, Wertheimer said--in inverse proportion to where his team had concluded that important information would be found. "I wanted them to stop the vacuum cleaner, to get at the secrets worth knowing," Baginski said. "You just wind up drowning in data, and you're not necessarily any smarter." She called it "hunt, not gather," a phrase that later became associated with everything she tried to change.

Baginski and Hayden were still determined to fix the way that information was shared inside the N.S.A., and, in early 2001, she announced a test run of what was called Operation sigint. Among other goals, the experiment focussed on the Internet, and it aimed to bring intelligence gathered from many sources into one accessible place. Baginski and her technical team chose five strategic and tactical areas, including access to a Chinese communications system and attempts to learn whether Russian organized crime was the source of weapons of mass destruction in Iran. "They weren't spring-training games, because they were against real targets," Hayden said.

Throughout Operation sigint, which involved about twenty-five volunteers from the agency, there was strong internal opposition. Senior supervisors in particular felt that Baginski was usurping their authority by organizing the team around "hot issues," which meant that some employees could circumvent their chain of command. The team had its most difficult time with the China analysts. When the team discovered that a communications system targeted by the N.S.A. was about to be phased out, the China analysts admitted that, although they knew about it, they had no alternative source of information. The Operation sigint team found a new source. "The country knew a lot more at the end of the six-month experiment than they would have had we not done it," Baginski said.

In the Russian project, the group pulled together information from across intelligence agencies that, in separate pieces, had not been useful: passport numbers, aliases, frequent-flyer numbers, credit-card and cell-phone numbers. In particular, the team was told to look for e-mails, even though other analysts did not believe that their targets used e-mail. In the course of the experiment, the group learned the truth about a connection between the Russians and Iranian W.M.D.s. "I can't say true or false, but I can tell you they got the answer," Wertheimer said.

Senior managers also complained that Baginski could be too abstract in stating her goals. "What came so easily to her, what was so self-evident that it didn't need to be said, just didn't work for other people," Wertheimer said. This gave her a reputation as a micromanager, because, inevitably, she'd end up walking people through each step. Others thought that she took failure personally. Wertheimer told me, "I have sat in the room when her displeasure was absolutely written all over her face. As sweetly as she said it, people left grumbling, 'God damn it, we didn't do it right again. What is it she wants?' "

During the summer of 2001, the intelligence community had been picking up hints of an impending attack. (On August 6, 2001, the Presidential Daily Brief was entitled "Bin Laden Determined to Strike in U.S.") The UBL (Usama bin Laden) unit at the N.S.A., a cadre of about ten language analysts, was aware of a high level of "chatter," but none of it was specific. After the attacks, Baginski called together her team and the discussion veered toward data collection: Which phone numbers should be listened to? Which antennas should point in which direction? "The immediate reaction was, Let's backpedal," Wertheimer recalled. "She said, 'No, that will not happen. We have a plan' "--referring to the strategy tested by Operation sigint.

After a few false starts during the course of September 11th, Baginski took over, with an approach that will be familiar to any reader of police procedurals: on a large piece of paper, she wrote the initials "UBL" and drew a box around them; then she asked her team to come up with any plausible connections, social and otherwise. As Baginski and her allies saw it, groups like Al Qaeda have to use banks, computers, and cell phones. "Their job is to blend in, so they aren't quite the cryptographic challenge" that, for example, the Chinese Embassy is, Wertheimer said. That means they're as susceptible as the average person to computer viruses and identity theft--and to being spied on.

Many analysts were sure that Al Qaeda didn't use the Internet--a theory based, apparently, on the fact that they had seen no evidence of it. "We said, 'You are now to presume [they] use e-mail. Find it,' " Wertheimer said. They started looking for clues in old intercepts. There were unhelpful detours (including a request for all Arabic-language e-mails), and a moment of near panic when an associate of the bin Laden family made a call to the East Coast and mentioned a "string of pearls," which was considered a code for nuclear weapons. (It turned out that the pearls were intended as a wedding gift.) But analysts began to build a system that could deliver quicker, more reliable intelligence.

When the Bush Administration began the bombing campaign in Afghanistan, in October of 2001, Hayden, Defense Secretary Donald Rumsfeld, and Tenet, among others, had lists of so-called high-value targets. "Geolocation became a massive issue for us that it never was before," Wertheimer said. This was the first war in which the N.S.A. could locate a communications device within yards; in practical terms, it meant that the N.S.A. staff was going to provide information that would be used to kill someone. The C.I.A., which was launching missiles from unmanned Predator aircraft, hit several targets during the Afghan campaign. Baginski could see fear on the faces of the civilian analysts. "It's a very hard thing to do when you're thousands of miles away in Fort Meade, Maryland, and you're going, 'Oh, my God, I wonder if I'm wrong. I wonder if I've made a mistake, wonder if this is an innocent person,' " Baginski recalled. "I had long talks with General Hayden about that--that these guys are not in uniform. These guys are civilians. We are going to have to walk them through this and let them know that right behind them is me--that no one is ever going to blame these guys."

Baginski's father had been a bombardier at age twenty-three during the Second World War and had been deeply affected by the experience. "He had dropped bombs on humans, and I think that's a pretty heavy burden for a very deeply religious man to bear," she said. She told the analysts, "I know what we're asking you to do is hard, and we have absolute confidence in you and you must know that we will support you." Eventually, Baginski created a separate geolocation unit of willing participants. "We have to advance the tradecraft," she said. "It was new--very new." There were some terrible mistakes, including the December 29, 2001, bombing after a wedding party, which left dozens of Afghan civilians dead. When that happened, Baginski tried to reassure the unit. "Just stay focussed," she recalled saying. Geolocation helped to find bin Laden's military planner, Mohammed Atef, who was killed in a bombing raid in November, 2001, and, four months later, Abu Zubaydah, who was the first top Al Qaeda leader to be captured. There have been other successes, although Osama bin Laden and his second-in-command, Ayman al-Zawahari, are still at large.

In the late spring of 2002, Hayden told Baginski to start coming up with a plan for Iraq. She was going to be his J-3--chief of operations, he said. It "tells everybody in the extended enterprise exactly what I think of her," Hayden told me. But, for the first time in two decades at the agency, there was a serious debate about going to war. Baginski had questions about the justifications, and she worried about the strains on her staff. She nevertheless told them that war was likely "and if anyone had a lack of comfort with that, then they needed to make a different decision about their workplace." To me, she said, "When I talked to the workforce, I was talking to myself at the same time."

The Iraq war was a chance for Baginski and others to create for the military giant chat rooms, known as "zirconchat," which provided intelligence about the battlefield almost as fast as it could be collected. The system would permit a properly equipped soldier in a Humvee to get a live feed on the most up-to-date intelligence and have instant contact with a large group; as many as two thousand people could be in touch at any given time. This gave American soldiers an enormous advantage as they went into combat. "We had more situational awareness about where the Iraqis were than they had about themselves," Lieutenant General Hayden said.

By then, however, Baginski was seeking a change for herself. She sensed that Hayden wanted to slow the pace of institutional reform, and that her efforts would become increasingly less welcome. "It is not necessarily a bad decision," she said. "It just wasn't what I thought was right." In January of 2003, she told Hayden that she wanted to leave after the military phase of the war was complete--perhaps for private industry or for teaching. On March 19, 2003, the night the Iraq war began, Baginski recalled that she walked through her building at the N.S.A. campus in Fort Meade. Computers were on, and everything seemed to be working. That night, she said, "I knew it was O.K. to leave. They had taken it beyond me."

Two months earlier, Robert Mueller spent a half day at the N.S.A. "He talked about his analytical challenge, about trying to change the culture of the bureau," Hayden recalled. "Did I have someone to recommend to him? And it wasn't long in the conversation before the name of Maureen came up." Baginski had her first meeting with Mueller in March of 2003; she started work two months later, at a salary of a hundred and forty-five thousand dollars.

Everybody, Baginski said, tends to make intelligence "sound so hard and mysterious, and it's really not. You need something, you go get raw material, and you add value to it. You put out a product and you keep adjusting, based on the feedback that you get. That really is all it is." Yet intelligence also relies upon effective tools, and until this past spring, when the F.B.I. finished installing nearly thirty thousand new desktop computers, high-speed networks, and a counter-terrorist database, agents often had to use nineteen-eighties-era technology with dial-up modems; it took eleven keystrokes to complete a search. When the final elements of the F.B.I.'s new computer system are fully installed--sometime in 2005, after more than a year's delay and at a cost of nearly six hundred million dollars--whatever is known about a case should be easily available to analysts.

One day, I asked Baginski what would happen now if someone like Zacarias Moussaoui were to be picked up, or if an agent in the field wrote a report like the famous "Phoenix memo," from July 10, 2001, warning of an effort to enroll Al Qaeda operatives in flight schools in Arizona. The difference, she said, is that every field office has a field-intelligence person. "And that field-intelligence person has the responsibility for insuring that the sum total of investigative product that the F.B.I. has is reviewed for intelligence value and disseminated. The other big difference is that at headquarters we have what we call 'operation specialists,' and their job is to then connect the dots across from Seattle and New York and Jacksonville." All threats, she said, were immediately shared with relevant internal law-enforcement authorities. She was sure that a Phoenix-like memo would be seen and acted upon.

Paul McCabe, a Special Agent who also serves as a spokesman for the bureau's Minneapolis field office, where Moussaoui was first detained, talked about changes he had experienced. He mentioned a recent tip that, in the past, would have gone into a "zero file"--ignored--"because it didn't have any criminal implications." With the sort of cross-checking that is now available to analysts, he said, the information had thrown up enough "red flags" to send it on to the Joint Terrorism Task Force in Minneapolis, where it is still under investigation.

Every day, the intelligence community gets about thirty-five intelligence reports from the F.B.I., up from four a day when Baginski started and compared with a couple of hundred from the C.I.A. But the quality of the bureau's threat-assessment and intelligence reports is still inconsistent. In late May, during a Senate Judiciary Committee hearing, Senator Dianne Feinstein, of California, complained that an April, 2004, intelligence report that she had requested from the L.A. field office was little more than a compilation of older data and "random comments on ongoing investigations." Baginski agreed with Feinstein. "It's uneven," she said of the threat assessments. John Brennan, a C.I.A. veteran and the head of the government's recently created Terrorist Threat Integration Center, sees improvement. He pointed to the involvement of the bureau and Baginski in raising the threat level during the 2003 Christmas holidays, when several international flights to the United States were grounded because of suspicious names on the passenger manifests. Brennan said that the bureau's response showed that the F.B.I. was on the right course. But it was also a reminder that too much information can be overwhelming.

In early summer, I asked Baginski if there have been any significant penetrations of Al Qaeda. "I'm confident that we are doing an incredible job getting into their information space," she said. "I really am. But I'd have to leave the penetration thing out of it, because it is very classified stuff." She went on, "Like any sober individual, you can be very happy with what you know, and then what keeps you up at night, of course, is what you don't know. So that's what my job is--to keep pushing what you don't know."

A critic of the bureau, John MacGaffin, who had been the C.I.A.'s No. 2 person for clandestine operations until 1993, and, until 1998, a senior consultant at the F.B.I, was pessimistic. MacGaffin, who had met with the 9/11 Commission, believed that the key problem was not a failure to connect the dots but a failure to find enough of them. "The thing that is not better--that is not more intense and is the sine qua non to doing this right--is spies," MacGaffin said. "People in the innermost councils of those who would do us grievous harm. And we do not have those in any adequate--even approaching adequate--number. And that's the only way you're going to beat this." He called this a failing of both the F.B.I. and the C.I.A.

When I repeated this analysis to Baginski, she said, "He's wrong. He's wrong. He's wrong," although she would not say in what way. Baginski, who has met with MacGaffin, added, "I honestly work in a different F.B.I. than he did."

MacGaffin shrugged off Baginski's response. The F.B.I., he added, not only failed to understand the reach of Al Qaeda's presence in the United States but also didn't know if there were sleeper cells here. Philip Zelikow, the 9/11 Commission's director, doubted that domestic sleeper cells pose a great threat. "The United States is a more hostile environment in which to operate than it was," he said. "It's more like operating behind enemy lines in enemy territory."

Mueller, in Zelikow's view, had addressed the most "indefensible" areas that had plagued the bureau but noted that some technology problems, which were never fixed under Freeh, Mueller's predecessor, still aren't solved. Mueller has sought advice from computer entrepreneurs like Larry Ellison, the head of Oracle, and, in June of 2002, he brought in a former I.B.M. executive, Wilson Lowery, to oversee the upgrading of the bureau's system. (Lowery had reportedly likened his challenge to "teeing off two hundred yards behind Tiger Woods.")

Baginski, Zelikow said, is trying to train F.B.I. agents to be better at collecting intelligence; agents can no longer become assistant Special Agents in Charge without intelligence training. But, Zelikow added, as "with a lot of these things, the implementation is ragged." Last summer and in the fall of 2003, the commission staff visited five of the bureau's field offices and found gaps between good intentions and reality; for instance, some analysts reported that they were still being called in to perform the duties of support staff.

To help make its case for survival, Zelikow said, the bureau provided the commission with case studies of post-9/11 investigations. Based on information obtained from Khalid Sheikh Mohammad, a key plotter of the 9/11 attacks, Iyman Faris, an Ohio truck driver, was secretly taken into custody in March, 2003, and accused of being part of a plot to bring down the Brooklyn Bridge by severing the suspension cables. Although the bridge operation was not considered feasible, there were concerns about a second wave of attacks. "Traditionally, the primary focus would be, Can we put this person in jail for what he did?" Kevin Brock, who was then the Special Agent in Charge of the Cincinnati office, told me. In this case, agents wanted Faris's help.

Attempts to gain cooperation, Brock said, are always "a delicate and fascinating dance"; agents told Faris that he faced a prison sentence of life without parole and, according to his lawyer, possible incarceration in the government's detention camp at Guantanamo Bay. Alternatively, they said, if he cooperated his sentence might be reduced, and he and his family might even be placed in the government's witness-protection program. Faris quickly agreed to cooperate and was moved to a safe house. A month later, he secretly pleaded guilty to providing material support to Al Qaeda and, over the next couple of months, was debriefed by both the bureau and the C.I.A. Unlike sessions conducted before September 11th, experts on Al Qaeda were monitoring the interviews. "They were driving the questions," Brock said.

Faris provided information on how Al Qaeda communicated and used the Internet, and agreed to make contact with other Al Qaeda associates; one of those associates was charged earlier this year with plotting to blow up shopping malls in Columbus, Ohio. In June, Faris moved to withdraw his guilty plea, saying that he had tried to deceive the F.B.I. because he wanted to gather material for a book. The court rejected this claim, and Faris was given a prison sentence of twenty years.

Zelikow, along with Baginski and other bureau officials, wouldn't talk about other specific cases. When I asked Zelikow if the F.B.I. had thwarted any attacks, he replied, vaguely, that the bureau has had some success, but said, "The more you know, the more it makes you modest about how much we really understand. And there are things about the 9/11 plot we still don't fully understand." Zelikow was struck by a singular difference between the attacks on New York and Washington and those which have occurred since. "The operatives were trained in country X and are deployed and act months, much later, in a country ten thousand miles away," Zelikow said of September 11th. "That's an international operation, and a fairly complex one on a micro scale. The singular feature of all the major post-9/11 attacks is that they are not really that international. They are regional and local, using assets that are more or less in place in the area. That's a striking thing."

Zelikow said that all the evidence he has seen suggests that Al Qaeda would have liked to launch another operation in the United States by now. The fact that it has not occurred reaffirms his belief that the collective efforts of the intelligence community have been able to degrade Al Qaeda's ability to conduct complex international operations. He went on to say that the absence of an attack before the Presidential election could be seen as "a testament to incapacity." But he still believed that the United States would be the target of an attack sometime in the next decade. Kevin Brock, who became Baginski's deputy earlier this year, said, "What concerns me is that most successful terrorist operations are born from very simple things to execute. So we also don't want to get too distracted by trying to overanalyze and try to guess at every exotic threat scenario that ever pops up."

In early October, a longtime senior official who has access to sensitive intelligence told me that Al Qaeda may be planning a "Tet-like offensive" of twenty to thirty simultaneous attacks across the United States. Brock said, "We all share in the worry that Al Qaeda wants to do us harm. Nobody disputes that. I can assure you, in my twenty years' experience I have never seen us turn over more stones than we're turning over right now."

John Pistole, who, in early October, became Mueller's deputy, has a gloomy view of what lies ahead. Pistole, who is forty-eight, is a lawyer and the son of a minister, and a twenty-one-year F.B.I. veteran. The intelligence function, he said, "is not a finished product here. Clearly, the roads have been laid, we're building it." But, even with the improvement, he told me, "every day that goes by brings us a day closer to the next attack." Earlier intelligence reporting indicated that "certain attacks are being held in abeyance in order to do a more spectacular attack," and the F.B.I., along with the C.I.A. and other intelligence agencies, was using all information, including interrogations of detainees and cooperating sources, to come up with possible scenarios. The C.I.A. regularly briefed the F.B.I. on its "red cells" exercises, in which it brought together diverse groups and asked them how they would launch an attack that would cause maximum economic and psychological damage. The groups included thriller writers and people whose backgrounds mirrored those of the nineteen hijackers. Some red cells lasted a year; one looked at how Al Qaeda could generate a disaster like a forest fire or a power-outage emergency. "If it's something we've never thought of, and is clearly a top priority, we investigate," Pistole said.

In June, Pistole told the 9/11 Commission, "I think we've probably prevented a few aviation attacks against both the East and West Coasts," but he cautioned that there were "operatives involved in those plots that we still cannot account for." He told me that there were six cases abroad where the F.B.I. had obtained information so specific that a foreign service was "able to go out and locate and arrest people in the process of buying materials." The F.B.I., he went on, had vastly improved its ability to track money contemporaneously, to the point where it could see where money was being spent at that moment, or tell a foreign service where "somebody using this code is going to come in and pick up some money to launch a terrorist act."

One F.B.I. counter-terrorism official told me that the bureau had recently sent informants to the Middle East, and that the C.I.A. had brought some of its sources to the United States in a joint effort with the F.B.I. to penetrate radical Islamic circles. There have also, apparently, been attempts by terror groups to recruit non-Arab-looking individuals and American citizens. I was told by two sources that the F.B.I. had interviewed a blond Arab-American whom the Saudis had arrested last summer; he had confessed to plotting to assassinate George W. Bush and was still in Saudi custody. Other intelligence agencies have been picking up apparent references to assassination plots against Bush and people close to him. Baginski, meanwhile, continues to make repairs in what has been a troubled institution--for instance, those thousands of untranslated and erased terror-related tapes and documents. Mueller said that the translation division is being transferred to the planned Directorate of Intelligence--in other words, the problem is being handed over to Baginski.

But the F.B.I. remains relatively small--the New York Police Department, with nearly forty thousand officers, is more than three times as large--and its resources are finite. At one point, Baginski told me that there would almost certainly be another terrorist attack in the United States. We were in her office, just down the hall from Mueller's. I had brought lunch, and Baginski had barely eaten a bite. "But when you ask me based on what evidence, it's nonspecific, just as it was, quite frankly, in the run-up to September 11th," she said.