Wired Magazine:
Richard Nixon's Last Secret

Richard Nixon took the secret to his grave. Thirty years after the Watergate break-in, there's a ra ce to unerase - and it's hitting high gear.

By Tom McNichol

Paul Ginsberg owns more than 100 tape recorders, most of them crammed into his basement audio lab, a low-ceilinged lair in Spring Valley, New York, that he likes to call the Tape Cave. There's a clunky Revox reel-to-reel from the early '70s that looks like it's straight out of the Partridge Family's garage; 13 Tascam cassette decks modified to copy tapes at various speeds; dozens of microcassette recorders scattered about like matchbooks; and a Nagra SNS, the subminiature reel-to-reel that went up in smoke at the beginning of every episode of Mission: Impossible. But it's another reel-to-reel that holds a special place in Ginsberg's collection. "This is the baby," he says, watching the tape revolve. "It's a Sony TC-800B, the same kind used by the Nixon White House to record the Watergate tapes."

As it happens, Ginsberg has dusted off his old Sony just in time, because the Nixon White House is suddenly hotter than ever. Capitalizing on the 30th anniversary of the Watergate break-in, former Nixon counselor John Dean has stirred up a media frenzy with a book that claims to unmask the case-breaking source Deep Throat. Meanwhile, Ginsberg is among a handful of forensic audio experts vying to crack the infamous 181é2-minute gap in a White House tape recorded three days after the burglary.

Photo by Martin Esseveld "If I manage to do it," says forensic audio guru Ginsberg, "my basement will become a national historic spot."

The gap appears in a recording made on June 20, 1972, as Nixon discussed the Watergate break-in for the first time with his chief of staff, H. R. Haldeman. When the White House revealed that part of the conversation had been "accidentally" erased, a majority of Americans came to believe that the man who insisted "I am not a crook" was that and more. Nixon resigned in disgrace in 1974, but in the years since Watergate, the gap has remained a tantalizing mystery.

Watergate scholars can't help but wonder: What could be incriminating enough to make this the only Nixon tape erased, when so many hours of blatant cover-ups, dirty tricks, expletives, and Jew-baiting were left untouched? "It's the big unknown," says David Coleman, assistant professor at the Miller Center of Public Affairs at the University of Virginia, who's studying Nixon's taped conversations for the university's Presidential Recordings Project. "It could contain something innocuous. Or it could be another smoking gun."

Tape 342, as it's known by archivists, was last tested in 1974 by a panel of audio experts, who concluded that the erasures were done in separate segments. Whoever erased the tape pressed Record, stopped the tape, and hit Record again, between five and nine times - hardly an accidental erasure. But the panel was unable to retrieve any of the lost conversation. "The experts concluded that it was a deliberate erasure," emails Bob Woodward, who along with colleague Carl Bernstein did much of the groundbreaking reporting on Watergate for The Washington Post. "The public got the message: The tapes had very damaging information. The exact content of the 18 1/2-minute gap would be interesting."

Photo by Martin Esseveld Inside Ginsberg's Tape Cave: multispeed reel and cassette recorders, signal processing units, parametric equalizers, a dynamic noise eliminator, bandpass filters, and much more.

Last summer, the National Archives and Records Administration decided it was time to take another look, in the hope that advances in digital technology would be able to restore the conversation. NARA invited audio experts to demonstrate how they might retrieve intelligible speech. The competition - a battle of the bands among a dozen or fewer audio experts - is expected to last more than a year. Before any of them get their hands on the original tape, they must show they can retrieve sound from test erasures without doing damage. Those who succeed will get a crack at the real thing.

The recovery effort promises to be as tricky as Nixon himself. It's like solving a Zen riddle: What is the sound of no one talking? And NARA isn't paying any of the audio experts, though the attention they'll draw may be its own reward. Ginsberg is well aware of the prestige that would come with a successful decoding of Tape 342. "If I do manage to do it," he says, punching Stop on the TC-800B, "I guess my basement will become a national historic spot."

An inveterate tinkerer who counts Thomas Edison as one of his heroes, the 56-year-old Ginsberg has built or modified much of the equipment in the Tape Cave. Since 1974, he has run a business called Professional Audio Laboratories out of his home. Most of his work involves enhancing and authenticating surveillance tapes and other audiovisual evidence presented in criminal and civil proceedings. He's been involved in more than 1,600 court cases, including the Waco trial. He enhanced surveillance tapes recorded inside the Branch Davidian compound to show that the fire was set from within - not by the surrounding government troops.

For the Nixon case, Ginsberg is hoping that his TC-800B will give him a leg up on his competition. By carefully measuring the class characteristics of the Sony - wow and flutter (the percentage of low- and high-frequency speed fluctuations), head alignment, and track configuration - Ginsberg hopes to establish a technical baseline against which to compare subsequent testing. Being a lifelong audio packrat just might pay off. "This is one of those times," he says, "when it helps to have a crazy old piece of equipment."

Stroll into the National Archives in College Park, Maryland, and ask to check out Tape 342, and the archivists will look at you as if you've asked to wipe your feet on the Declaration of Independence. Tape 342 is treated like a priceless heirloom, locked in a vault kept at precisely 65 degrees Fahrenheit and 40 percent relative humidity. The tape has been played just half a dozen times in the last three decades, and only then to make copies.

Qualified researchers do have access to a copy. "I've listened to this thing more times than I care to admit," says Michael Hamilton, supervisory audiovisual specialist for the archives' Nixon Presidential Materials staff. "I've listened to it very close and very loud. I can't detect any conversation. But I haven't played with it electronically."

The gap occurs during a conversation between Nixon and Haldeman three days after the burglary of Democratic National Committee headquarters. Just before the erasure, there's the distinct ticking of a clock in the background - marking out the seconds of a doomed presidency. The overall sound quality is terrible, even by the crappy technical standards of the Watergate tapes. The tiny lavalier microphones hidden throughout the Oval Office and Executive Office Building to record conversations were cheap and poorly placed. The recordings were made on thin 0.5-mm tape running at the unusual speed of 15é16 inch per second - half the speed of a cassette recorder. The slow tape speed degrades the recording's already poor signal-to-noise ratio. At the point of the first erasure, the muffled conversation is suddenly replaced by a buzzing noise, presumably the sound of a 60-cycle hum leaking from the power grid as interpreted by a high-gain microphone input circuit. Throughout the gap, the buzz occasionally drops in volume, but never is there any discernible speech.

When the public learned in 1973 that the tape had been tampered with, Nixon's personal secretary, Rose Mary Woods, stepped forward with a convoluted story about how she could have been responsible. Woods said she was transcribing the tape when she took a phone call and left her foot on a pedal that may have caused the erasure. A widely circulated photo of Woods re-creating her improbable lean across her desk led many to believe that she was stretching more than just her torso. But even Woods insisted that she was responsible for no more than five minutes of the erasure.

The fact that the tape contained as many as nine separate erasures contradicts any notion that it was caused by an accidental press of the Record button. The culprit was either very anxious to protect the president or was a mechanical klutz. Both descriptions, Watergate scholars have noted, fit Richard Nixon. The 37th president was laughably inept when it came to technology. Haldeman recounts in his now out-of-print book, The Ends of Power, that Nixon struggled with the most basic functions of cassette recorders. The Army Signal Corps supplied Nixon with the simplest recorder available so that the president could dictate memos in the evening. But even then, the various buttons had to be marked so Nixon could use the machine without mixing things up. Put a man like that in front of a reel-to-reel, and it's easy to see how a simple erasure could turn into a clumsy mess.

More puzzling is why Nixon chose to erase this segment and no other. One theory is that he sat down sometime in 1973 with the intention of erasing all incriminating tapes, so that the Watergate special prosecutor couldn't get his hands on them. Nixon began with Tape 342, the first recorded conversation after the break-in, but quickly became overwhelmed. There were more than 2,800 hours of tapes, and the way he handled a recorder, it would have taken him a lifetime to erase everything. So the June 20 conversation was erased simply because it was the first Watergate tape chronologically.

An opposing theory is that the lost conversation is the one Nixon couldn't bear anyone to hear. Even Haldeman's somewhat sanitized version supports this. As he reconstructs it in The Ends of Power, a worried Nixon spends part of the meeting setting the Watergate cover-up in motion. During the discussion, Nixon is particularly concerned that the trail will lead to White House special counsel Chuck Colson. Linking Colson to the break-in would tie the burglary to the White House. "I know one thing," Haldeman recalls Nixon saying. "I can't stand an FBI investigation of Colson."

"Inadequate as [Haldeman's] version might be, it fits with the overall evidence in the case," notes Bob Woodward. "There was an effort to cover up and conceal."

But what Nixon said may pale in significance to the way he said it. Perhaps there's a fragment of conversation, even an aside, that's so outrageous that Nixon had to get rid of it. It could be something more incriminating than the so-called smoking gun conversation held three days later, in which Nixon instructs Haldeman to tell FBI investigators, "Don't go any further into this case, period." Or a comment more thuggish than telling John Dean that it wouldn't be a problem coming up with hush money for the Watergate burglars: "You could get a million dollars. And you could get it in cash." Something even more vile than Nixon ordering Haldeman, "Bob, please get me the names of the Jews, you know, the big Jewish contributors of the Democrats. All right. Could we please investigate some of the cocksuckers?"

Audiotape consists of a thin plastic base coated with ferric oxide, made up of microscopic magnetic particles. As the tape passes over the record head, the ferric oxide particles are reoriented by the magnetic field from the head. After recording, the tape is covered with patches of magnetization of various depths and directions, which are translated into sound. When a tape is erased, the erase head scatters the particles. "It's not a kitchen floor, where if you strip away all the tiles, you can see the original floor," says Ginsberg. "It's more like ripping up all the tiles and randomly rearranging them."

Although Tape 342 was recorded on a Sony TC-800B, it was erased on a Uher 5000 reel-to-reel. If the Uher performed a perfect erasure, the contest to retrieve the lost conversation is over before it begins. But any number of factors - a crack in the erase head, a dust mote on the tape or heads, slack in the tape at startup, head misalignment - can result in only a partial erasure. The fact that Tape 342 was recorded on one machine and erased on another increases the chances of an incomplete erasure. A decade ago, this wouldn't have mattered. But today's digital tools can help make something out of what sounds like nothing.

Removing unwanted sound with analog tools is akin to using a butter knife to cut out a bruise in an apple. The knife may remove the bruise, but it'll also carve out some of the good part. Digital audio tools are more like a laser scalpel, allowing the user to make hundreds, even thousands of fine cuts, shaving off specks of unwanted tones and noise so that all that remains is speech. Ginsberg has put this technique to use already, in a case that offers a glimpse at how he plans to attack Tape 342.

In 1997, a large New York brokerage house came to Ginsberg with a taped conversation between two brokers, a crucial piece of evidence in a multimillion-dollar lawsuit. The conversation was taped on a Dictaphone logging machine, which records 60 channels across a 1-inch-wide tape - "a lot of tomatoes to squeeze into one can," as Ginsberg puts it. The conversation in question was on channel 19, but a malfunctioning amplifier rendered the speech inaudible. To improve the signal-to-noise ratio, Ginsberg contacted Dictaphone and had them build a special tape head that played only channel 19. Then he modified the machine to run at four times normal speed, further boosting the signal without increasing the noise. "At first I didn't think there was a conversation there," recalls Ginsberg. "But then I could hear a very, very, very faint muttering."

Next, Ginsberg turned to an array of digital audio enhancement tools. He loaded the audio onto the PCAP II - a sophisticated audio-filtering workstation used by the FBI to clean up undercover surveillance tapes - which features a toolbox of digital signal processors. There's a high-pass filter to eliminate low-frequency "rumble," and a limiter for reducing gain on sudden, loud input signals. Notch filters can be set to reduce the signal level of an unwanted frequency. Comb filters attack harmonically structured noises, such as the 60-cycle hum found on the Nixon tape. An inverse comb filter enhances useful harmonically structured signals, such as vowel sounds in speech, allowing the user to tease voices out of background noise.

There's also a tone remover that eliminates constant frequencies. Human speech, which is highly intonated and constantly changing, is not affected. Up to five onboard equalizers can be adjusted to boost frequencies in the human voice range while filtering out extraneous noise. A 460-line spectrum analyzer allows the user to see the audio being processed as a waveform. An array of digital adaptive filters, like the PCAP, use multiple microprocessors to continuously analyze the audio signal and automatically make adjustments when they detect changes in the noise. Moreover, they can operate in the speech frequency bandwidth to reduce noise and not affect the voice.

After running the brokerage tape through the PCAP, Ginsberg was able to recover about 90 percent of the conversation, which helped exonerate his client. "Even I was impressed," Ginsberg says. "This conversation came from nowhere."

Of course, success wasn't entirely in the tools. Whether through heredity or training, Ginsberg is literally wired for sound, blessed with exceptional hearing. His family is always complaining that volume on the TV is too low, while he sits happily in his chair wondering what the fuss is about. He drives a Lexus GS 300 for its acoustics - he says the car has a great sound system and gives a quiet ride.

Photo by Martin Esseveld "If we get a couple of words, we'll be ec static," says Reames, who designed undercover recorders used worldwide.

This will be Ginsberg's second brush with Tape 342. While he was employed by Federal Scientific almost 30 years ago, a team of government officials approached a division of the company with Tape 342 in hand, wanting to retrieve speech from the gap. Ginsberg didn't work directly on the project, but his company was responsible for the first of many failed attempts to recover the 181é2 minutes. This time, Ginsberg wants to get it right - but not for the political or historical implications. He can't even remember if he voted for Nixon in 1972. "For me," he says, "it's all about the technical challenge."

If Paul Ginsberg is the Tinkerer, James Reames is the Spy. Reames, another entrant in the Nixon Gap Challenge, worked for the FBI for 32 years, at one point running the bureau's audiotape laboratory. He helped design the Nagra JBR subminiature recorder, which for years was the standard undercover recorder used by government agents worldwide. After leaving the bureau, Reames started his own company, JBR Technology, which does forensic audio work.

"I know most of the experts who studied the 181é2-minute gap in 1974, and they are goooooooood," Reames says, in a soothing Southern drawl that sounds like Andy Griffith extolling the virtues of a Ritz cracker. "We'll have to get lucky to retrieve any of the conversation. But this is one of those once-in-a-lifetime cases."

Like Ginsberg, Reames has a Sony TC-800B. He plans to apply some of the same techniques, with a few added twists. He'll play Tape 342 on a modified Nagra reel-to-reel fitted with a small video camera and two high-resolution mirrors mounted on either side of the tape path. A video record will protect Reames if there's any question about damage. It could also reveal oily smudges or even fingerprints. Reames could get lucky and find Nixon's big, klutzy thumbprint. "We'd take any of his 10 fingers," he says, smiling.

Photo by Martin Esseveld A modified Nagara reel-to-reel, owned by audio expert James Reames: Variable speeds allow searching for different sound frequencies, while hi-res mirrors detect tape damage.

He also plans to study the tape with magneto-resistive read/write heads, used in computer hard drives. The heads can sense small changes in magnetic direction, producing a detailed scan that can be used to map a digital image of the sound. If there's enough of a bitstream, the digital images can be translated into intelligible voices. Reames will use as many as 36 MR heads to scan the width of Tape 342, looking for a tiny "sliver track" of recoverable speech.

To find snatches of speech in the buzz and background noise, Reames has made an array of filters that will look for speech tones. If he finds signals that are aperiodic, it's a good indication that a word or a sound may have occurred in that frequency. If Reames manages to extract, say, the word yes, he can measure its frequency characteristics and search for the same properties elsewhere. In this way, he may be able to gradually piece together the message. "If we get a couple of words, we'll be ecstatic," Reames says. "If we get 30 words, we'll be even happier. And if we're lucky enough to recover the entire gap, we'll probably have a heart attack."

Reames' approach is being taken a step further by researchers working at two of the Commerce Department's Boulder, Colorado, laboratories. Scientists there have developed a technique known as second harmonic magneto-resistive microscopy (SH-MRM) for forensic authentication analysis. It involves passing a tape hundreds or thousands of times over a high-resolution magneto-resistive head. The sensors in the head move back and forth over the tape without touching it, gradually building up, line by line, a topographic map of the tape's magnetic field at millions of points. The mapped data can then be plugged into an imaging program. If enough data is recovered, investigators can begin to rebuild the original signal.

The SH-MRM technique is exacting; two inches of cassette tape can take an hour to scan. "If you're passing a half-inch-wide tape over a 1-micron-wide head," says David Pappas of the National Institute of Standards and Technology, "you have to scan it about 12,000 times to realize the full resolution."

One way to speed up the scan is to construct a system with an array of sensors - as many as 400 - to read many tracks simultaneously. Even with a single sensor, SH-MRM has been able to recover sound from an otherwise unplayable segment of tape rescued from a flight data recorder. The FBI is also using SH-MRM to analyze the thousands of cassettes it receives each year as evidence.

Pappas considered applying SH-MRM to the Nixon tape gap, but after September 11, the NIST lab has been focused almost exclusively on homeland security projects. "If somebody picks up our technology and applies it to the Nixon tape, more power to them," says Pappas. "But right now we have our hands full."

The contest to retrieve the 18 1/2-minute gap promises to be a long-running battle; at the National Archives, time tends to be measured in years, not months. If one or more audio experts is able to recover sound from the test tape without damaging it (and archivists are visibly nervous about harming the original), they will be issued a standard request for proposal. Once the bureaucratic formalities are satisfied, qualified audio experts would get their hands on the original Tape 342, possibly not until early next year.

Ginsberg is already immersed in the test recording sent to him by NARA. It's a fair approximation of Tape 342, a segment of speech and tones recorded on a Sony TC-800B and erased on an Uher 5000. Ginsberg has started tinkering, modifying his Dictaphone player to accept quarter-inch tape. After running the tape through the system, Ginsberg identified six discrete channels across the width of the tape that appear to contain signals - a good start, but still a long way to go. "It's analogous to an astronomer knowing where to point the telescope when he's looking for a celestial body," Ginsberg says.

Ginsberg is searching for a hidden world not in the infinite expanse of space, but among the infinitesimal ferric oxide particles scattered across a thin piece of plastic. Many hours of arduous investigation lie ahead, and Tape 342 has so far guarded its secrets well. But it would only be fitting if technology were responsible for revealing more about Richard Nixon than he intended.