JOHN VORSTER HAD TAKEN the simple expedient of lying when confronted with allegations that South Africa was on the verge of a nuclear test. He followed that nuclear lie with another, claiming during an October 1977 interview that “I am not aware of any promise that I gave to President Carter”—and that he had only repeated his oft-made statement that South Africa was “only interested in peaceful development of nuclear facilities.” In response, the State Department released portions of an October 13 letter from Vorster to Carter, which included the pledge that South Africa did not have, and would not develop, nuclear explosives for any purpose.1 That letter also turned out to be a lie, for throughout 1977 and for the rest of the decade and beyond, South Africa would secretly continue along a path leading to its possession of a small nuclear arsenal.
In 1977 South Africa and Israel began carrying through on an agreement that had been reached in April 1976 when Vorster—who the British had jailed during World War II for his Nazi sympathies and refusal to serve in the military—visited Israel to meet with Israeli prime minister Yitzhak Rabin and other Israeli leaders. The two pariahs had agreed to trade thirty grams of yield-boosting tritium from Israel, code-named teeblare—Afrikaans for “tea leaves”—for fifty tons of uranium from South Africa, code-named mutton. The tritium would be flown to South Africa in 2.5-gram installments over eighteen months.2
During the same year that South Africa and Israel began implementing their secret nuclear trade agreement, South Africa’s minister of defense issued a white paper asserting that “we are today involved in a war, whether we wish to accept it or not.” That belief led to acceleration of the program and a Vorster meeting with his senior aides. He directed them to draft a memorandum describing the nation’s nuclear path, which he approved in April 1978. What Vorster sanctioned was a strategy that called for South Africa’s clandestine development of nuclear weapons, revealing that capability to the United States and other countries if South African territory was threatened, and then, if the secret disclosure had no effect, a public announcement and possibly a test. The program did not envision actual military use, for fear of retaliation.3
That September Vorster left office to become the nation’s president, and was replaced by defense minister Pieter Willem Botha, commonly known as “P.W.” and “the big crocodile.” Shortly after becoming prime minister, Botha established a cabinet committee to deal with the military uses of atomic devices. It would deliver its first recommendations the following summer, including construction of seven weapons. In the interim, cabinet members, at a meeting held on October 31, 1978, decided that the state-owned Armaments Corporation (Armscor), the Defence Force, and the Atomic Energy Board should begin work on a nuclear weapons program, a program that was immediately classified as top secret. Armscor would be responsible for designing and building the devices.4
The year was notable not only for the formation of policies and committees but also for the production of fissile material and construction of another nuclear device. In January the Valindaba Y-Plant produced its first batch of highly enriched uranium, although it was 80 percent U-235 rather than the 90 percent preferred by weapons designers. Production would continue for the rest of 1978 and well into the next year at the 80 percent level, until operations came to an abrupt halt in August 1979, when a major chemical reaction contaminated the plant and put it out of commission for over seven months. The year 1978 also saw the AEB complete construction of a second, smaller device, intended for an instrumented test, before bomb construction was assigned to Armscor.5
DURING 1978 and the first eight months of 1979, the U.S. intelligence community was, as Jimmy Carter had promised, monitoring South Africa’s nuclear activities and evaluating any data collected. Monitoring included use of the imagery and signals intelligence satellites operated by the National Reconnaissance Office, National Security Agency and CIA communications intercept operations, as well as attempts to recruit human sources with knowledge of South African nuclear activities. It also included the continued use of aircraft belonging to the defense attaché’s office. As a result, in April 1978 South Africa expelled three American diplomats who were alleged to have used the defense attaché’s plane to take aerial photographs of “strategic installations,” which apparently included the nuclear enrichment plant at Valindaba.6
Two months before the expulsions, the CIA’s Office of Economic Research prepared a paper for an interagency assessment of South Africa’s nuclear options, noting “there is no doubt that South Africa can afford to develop and test nuclear explosive if it chooses.” Articles that appeared in the agency’s Scientific Intelligence Weekly Review between December 1978 and the end of April 1979 discussed the capacity of the Valindaba plant, problems at the plant, and the presence of military personnel at the Pelindaba nuclear research center.7 Then, in late September 1979 the focus suddenly shifted from assessing South Africa’s progress toward developing an atomic bomb, to whether it had already built and tested one.
ABOUT 10:15 P.M. on September 21, at Patrick Air Force Base in Florida, technicians from the Air Force Technical Applications Center were conducting a routine readout of Vela 6911, which had been launched on May 23, 1969. That Vela satellite, along with the final pair launched in 1970, orbited the earth at a distance of seventy thousand miles—leaving the U.S. one short of the four needed to keep all of the earth under surveillance on a continuous basis. By 1979, Vela 6911 had substantially exceeded its projected lifetime, and if not for the controversy that followed the September 21 readout, might have been best known for its contribution to space science. For years it had provided data on Cygnus X-1, first believed to be an X-ray double star and then the subject of a bet between physicists Stephen Hawking and Kip Thorne as to whether it was really a black hole. (It is and Thorne won the bet.)8
Vela satellite bhangmeters had, over the years, detected a variety of light flashes, including lightning. Many were of no concern because their line of sight did not lead back to earth. Others did not have the optical signature of a nuclear detonation. In some of those cases, their origin was apparent. In others, the cause was a mystery, and the detected signal became part of the collection of Vela “zoo events.”9
During the readout of Vela 6911, AFTAC personnel watched as a stylus drew a figure representing the variations in light intensity, as monitored by the two satellite bhangmeters. There was no data from a third optical sensor, whose mission was to provide the geographic origin of any noticeable flash of light, because it was out of commission. Nor would there be any reading from the satellite’s electromagnetic pulse sensors, which were no longer functioning. But what the technicians saw was sufficient cause for concern. The stylus drew a figure with a double hump, indicating a brief intense flash of light, a dramatic decline in intensity, and then a second, longer-lasting flash. Such double flashes had always been associated with nuclear detonations, where the fireball’s surface is rapidly overtaken by the expanding hydrodynamic shock wave, which acts as an optical shutter and hides the small but extremely hot and bright early fireball behind an opaque ionized shock front which is comparatively quite dim. The initial flash normally lasts only a millisecond and emits about only 1 percent of the total thermal energy, although it is the point of maximum intensity. It appeared that some nation or nations, in some part of the world covered by Vela 6911, had detonated a nuclear device in the atmosphere.10
The area in view of the Vela bhangmeters was 3,000 miles in diameter, encompassing the southern tip of Africa, the Indian Ocean, the South Atlantic, and a bit of Antarctica. Examining the satellite readout indicated that the double flash was recorded at about 3:00 a.m. local time. Less than a hour after its technicians noted the signs of a double flash, AFTAC, based on a preliminary analysis, initiated a “pre-alert.” After further analyses, Alert 747 was declared at 3:30 a.m.11
Sometime that evening President Carter and other top government officials—including national security adviser Zbigniew Brzezinski, secretary of state Cyrus Vance, and secretary of defense Harold Brown—were informed of the possibility that a nuclear detonation had occurred. The next morning a committee of ten to twelve officials gathered to deal with the potential crisis. Gerald Funk, at the time the senior Africa specialist on the National Security Council staff, remembers Brzezinski calling to tell him to “get my toucus into work, that we had a little bit of a problem.” Funk also recalled that “we first convened a meeting in the Situation Room of the White House,” with Frank Press, the presidential science adviser “there and in charge.” During that and other meetings early in the crisis, Funk’s assumption was “that there had been in fact a legitimate sighting . . . that satellite had never failed to react positively, and had never given a false signal.”12
Others called to the meeting included Spurgeon Keeny, then the number-two man in the Arms Control and Disarmament Agency, who received a call from Harold Brown, an old friend from college, and Gerald G. Oplinger, Brzezinski’s aide for global issues. Oplinger recalled that “we went around and asked ‘Was it a test?’” The CIA and Defense Intelligence Agency representatives “said the odds were at least ninety percent that it had been a nuclear explosion,” according to Oplinger. Keeny was skeptical that a truly informed judgment could be reached so quickly.13 Everybody realized that they were just at the beginning of an investigation.
PRESIDENT CARTER and his chief policy advisers needed two questions answered: Were the indications of a nuclear test, which the specifics of the Vela signal suggested was in the 2- to 4-kiloton range, correct? And if there had been a test, who was responsible? The first question could be addressed in two ways: by looking for corroborating data and by exploring the possibility that the Vela had malfunctioned or detected an event whose optical signature cleverly duplicated that of a nuclear blast.
Verification that a test had taken place, as well as the identity of the culprit, could come from conventional intelligence methods—a spy at the heart of the guilty government, an intercept of communications between high-level officials or between the test site and officials in the culpable nation, or even overhead photographs that revealed test preparations. In addition, the CIA, AFTAC, and other intelligence agencies could search data gathered by other components of the Atomic Energy Detection System for signals confirming a detonation had taken place. Much of that search would have taken place anyway, as it had for acknowledged tests in order to determine their characteristics, but the uncertainty and the high stakes added to the urgency.
Part of that search would involve examining the data gathered by the other military satellite systems that carried sensors capable of detecting nuclear explosions. On September 22, 1979, the air force was operating three Defense Support Program satellites, equipped with both bhangmeters and infrared sensors, the latter capable of detecting the heat from a detonation. F-6, which hovered over the equator, 22,300 miles above Brazil, could “see” portions of the South Atlantic. The best view of the area of the suspect flash was from F-7, the Eastern Hemisphere satellite stationed over the Horn of Africa, whose sensors could view part of the South Atlantic, the northern portion of Antarctica, and all of the Indian Ocean.14
In addition, sensors were carried aboard two NRO Jumpseat spacecraft, which operated in highly elliptical orbit, and whose antennas intercepted a variety of communications, particularly from the northern reaches of the Soviet Union. Those spacecraft also carried an infrared sensor designated Heritage, which had the potential to detect the heat from a nuclear test.15
One Satellite Data System (SDS) spacecraft, which also operated in highly elliptical orbit, was equipped with nuclear detonation detectors. The first two SDS craft had been launched in 1976; their primary mission was to relay the electronic signals from the low-earth-orbiting Kennan (KH-11) imagery satellites back to the ground station at Fort Belvoir, Virginia, where they would be converted into images. Other missions that had been assigned to the spacecraft included serving as a communications relay between the remote tracking stations employed by the air force to exercise command and control of its satellites, as well as relaying communications to U.S. strategic bombers flying over the northern polar region. In 1974 the air force had decided to assign yet another mission to the SDS craft: carrying bhangmeters to augment polar coverage of nuclear detonations. The third SDS satellite, and the first to be equipped with the detonation detectors, was orbited in August 1978.16
Further, Defense Meteorological Satellite Program (DMSP) satellites might have detected a blast. In normal circumstances two of the satellites circled the planet in five-hundred-mile orbits. Their primary mission was to provide weather information—crucial to programming reconnaissance satellites as well as conducting military operations. Going along for the ride on the satellites launched in February 1976 and April 1978 was a gamma X-ray detector provided by AFTAC.17
The records of the other passive nuclear detonation systems operated by AFTAC could also be searched to see what they might have picked up on September 22. Those no longer included the acoustic sensors of the Dawn Star network, which had been shut down in 1975. But they did include the seismic sensors spread across the planet, as well as the seabed sensors employed for the Sound Surveillance System.18
Active measures could also be taken by AFTAC, the CIA, and other intelligence organizations. The Vela detection “set off one of the most extensive air sampling operations in recent years,” according to AFTAC historian Gerald Wright. The first planes arrived in the area on September 25, and began over a month of air-sampling operations. By the time they ended on October 29, five different WC-135B aircraft flew a total of twenty-two missions, which involved almost 222 hours of flying time. Other aircraft flew another three missions, spending eight hours aloft in search of debris. The aerial missions were flown to gather any debris emanating from four possible test locations. The Kalahari Desert, where two years earlier South Africa had been preparing for a test, was one location. Another was Prince Edward Island, which along with nearby Marion Island was a South African possession located between South Africa and Antarctica, far from shipping and commercial routes. In addition, two ocean locations could have been the source of acoustic and underwater signals that had been detected. Airflow from other parts of the general area of interest was also targeted. The CIA took a lower-tech approach, sending personnel into western Africa to gather tree leaves, which might be coated with radioactive residue from a blast.19
CORROBORATIVE DATA might also be found in the records of government agencies outside the intelligence and military communities, for the effects of nuclear detonation could have been detected by scientific instruments operated by other elements of the government and private scientific institutions, with missions far removed from nuclear detonation detection. Such instruments included earth resources satellites such as Landsat, civilian weather satellites like Nimbus and Tiros, and even the Arecibo Ionospheric Observatory in Puerto Rico. In addition, scientific institutions in the United States and abroad might have inadvertently collected data of significance. Over the next year, some of this data would be searched for and acquired by the United States, while other data would be volunteered once word of the possible blast became public.
In November corroborating data appeared to come from the Institute of Nuclear Science in Gracefield, New Zealand, located just north of Wellington, the nation’s capital. The institute had discovered an increase in radioactive fallout in rainwater samples collected between August 1 and October 28. The rainwater contained short-lived radioisotopes such as barium-140, praseodymium-143, and ytrrium-91, all fission products of a nuclear explosion. The institute’s director, B. J. O’Brien, noted that “we didn’t see much of an increase, just enough to suggest they came from a small nuclear test.”20
The increase in fallout was measured for radioisotopes with half-lives no longer than fifty-nine days, so that if they came from a nuclear detonation, it would have had to have been a recent one. The half-life of barium-140 is twelve days, while the half-lives of praseodymium-143 and yttrium-91 are thirteen and fifty-nine days, respectively. O’Brien observed, “What we see in our fallout here would be consistent with a nuclear explosion having a force equivalent to two to four kilotons” and “what we’ve seen couldn’t have come from an old test. . . . Whatever it is, it is a recent event.” A White House source was also impressed, telling the Washington Post that “radioactive fallout was the key missing element in what we thought originally was a clandestine nuclear test” and that “the fallout in New Zealand could well be that missing element.”21
But before the end of the month, the institute was backtracking, issuing a statement saying “there is no evidence of fresh radioactive fallout during the past three months.” After the initial announcement, New Zealand’s National Radiation Laboratory conducted its own analysis of the rainwater and found no evidence of fallout. The discrepancy was officially explained as resulting from the laboratory’s being equipped to detect only levels of radiation that would endanger health, so that slight fluctuations of radiation would escape notice. However, when scientists at the Gracefield institute tried, they were unable to replicate their initial findings. Subsequently, a U.S. government laboratory also tested the water and found no signs of radioactivity.22
Another possible confirmation came from scientists working at the ionospheric observatory at Arecibo, Puerto Rico, site of the world’s largest single radio telescope, one thousand feet in diameter. Established in 1963 as a result of Cornell University professor William Gordon’s quest to study the ionosphere, it was also used to collect the signals from Soviet radars after they bounced off the moon, but became best known for its role in searching for signals from any extraterresttial civilizations that might be trying to alert others of their existence.23
The scientists, Lewis Duncan and Richard Behnke, were using the observatory’s radio telescope to watch the upper atmosphere to gather baseline data in support of an experiment in which they planned to watch an Atlas Centaur rocket tear a hole in the ionosphere during its launch. Several hours after the apparent Vela detection, the telescope sensed a ripple moving through the ionosphere. The scientist who saw the ripple, which he called a pattern of “ducted ionospheric disturbances,” would later note that the time and direction of the ripple’s appearance was consistent with the Vela flash. A nuclear explosion can send a shower of electrons outward through the ionosphere in such a manner as to cause it to “bob up and down a little,” according to the scientist.24
GIVEN THE INITIAL IMPRESSION that Vela had witnessed a nuclear detonation and the importance of such an event, senior U.S. officials could not wait until all the evidence was in and analyzed before considering the possible implications. A discussion paper, prepared by the State Department for an October 23 meeting involving the secretaries of state, defense, and energy, the director of the ACDA, the chairman of the Joint Chiefs of Staff, director of central intelligence Stansfield Turner, and presidential science adviser Frank Press, laid out the issues.25
The paper noted that the intelligence community had “high confidence, after intense technical scrutiny of satellite data, that a low yield atmospheric nuclear explosion occurred in the early morning hours of September 22 somewhere in an area comprising the southern portions of the Indian and Atlantic Oceans, the southern portion of Africa, and a portion of the Antarctic land mass.” However, efforts to acquire radioactive debris “have been fruitless,” although debris “could have escaped our collection effort.” It reported that there was no corroborating seismic or hydroacoustic data, but “those systems’ existing capabilities to detect low yield nuclear events in the region of interest is poor.”26
The paper also noted that the Vela detection was not yet public knowledge, but that information “could leak at any time” and observers would assume that South Africa had tested a nuclear bomb. In that case, there were three concerns: One was that “the efficacy of U.S. intelligence systems generally and test ban monitoring capabilities specifically” would be questioned. In addition, public knowledge would impinge on U.S. global nonproliferation as well as African policy interests. The “nonproliferation stakes could be high,” the paper observed, if the Vela detection “caused a rupture in our nuclear negotiations with South Africa,” which included trying to persuade the Botha administration to sign the Nuclear Non-proliferation Treaty. On the other hand, failing to take action in response to the event could make it more difficult to deter proliferation elsewhere, including in Pakistan and India.27
It was also feared that public disclosure of the event would have a negative impact on efforts to achieve settlements in Rhodesia and Namibia. In Rhodesia, the paper noted, “disclosure of a possible South African nuclear capability might have some cautionary effect on the negotiating positions of the parties . . . but most likely would sharpen the lines already drawn.”28
One choice facing the administration was whether to confront South Africa, “the most likely responsible party.” There was a case for not taking the issue up with the Botha government. The evidence was not strong enough to permit an accusation, and South Africa was “likely to treat our raising of the subject as an accusation.” If guilty, the South Africans would deny involvement, and if not guilty they would “react violently and probably conclude that there is no further point in discussing broader nuclear issues” with the United States.29
On the other hand, not to go to the South African government would leave the United States “vulnerable,” particularly if the intelligence on the September 22 event became public, to charges that the Americans were unwilling to confront the likeliest perpetrator. On balance, it was concluded “there seems more to be gained than lost” by raising the issue with the South Africans.30
Only days after the paper was written, the feared leak occurred, making a meeting between the U.S. ambassador and South African officials inevitable. ABC News reporter John Scali, who had attained fame by serving as an intermediary between a Soviet intelligence officer and the U.S. government during the Cuban missile crisis, reported the detection on the evening of October 25. In response to Scali’s report, the State Department released a statement that evening, acknowledging that the “United States Government has an indication suggesting the possibility that a low-yield nuclear explosion occurred on September 22 in an area of the Indian Ocean and South Atlantic including portions of the Antarctic continent, and the southern part of Africa.” The statement also noted that “no corroborating evidence has been received to date.” In a press conference in Gainesville, Florida, the following day, secretary of state Cyrus Vance observed that “it is not clear that there has been a nuclear detonation” and “we don’t know that anything has happened in South Africa.”31
Over the next days, South African officials, acting as if they had been accused, and with stories on the incident appearing in the international and South African press, proclaimed ignorance, denied that their nation had anything to do with the explosion, and suggested alternatives. In response to the State Department announcement, foreign minister Roelof F. Botha proclaimed, “I know absolutely nothing about the matter.” J. Wynand de Villiers, the AEB’s chairman, labeled suggestions that South Africa might have detonated a nuclear device as “complete nonsense,” and went on to say, “If there was anything of the sort, my first reaction would be that some other power might have undertaken a test, but it was definitely not South Africa.”32
Vice Admiral J. C. Walters, the chief of South Africa’s navy, suggested that the cause might have been an accident aboard a Soviet nuclear submarine, adding that the presence of Soviet Echo II class submarines, each equipped with eight nuclear-armed cruise missiles, in the vicinity was common knowledge. He called the theory of a submarine accident a “real possibility.” It was not an assessment shared by the White House. “It was considered,” one official told the New York Times, “but we gave up on the idea very quickly.”33
WHILE SOUTH AFRICAN involvement in a covert nuclear test attracted the most attention, it was not the only possibility. The October 22 State Department discussion paper, while focusing on South Africa, observed that “we must consider the possibility that Israel could have detonated a device in this remote geographical area,” an action that, if verified, could have significant consequences for U.S. policy toward Israel, the Middle East peace process, and the incentives of Arab states to acquire nuclear weapons.34 And while the collectors in the U.S. intelligence community sought to obtain evidence of a test and identify a perpetrator, some analysts as well as knowledgeable observers tried to identify the risks and benefits of a covert test to a number of nations. One result was a December 1979 interagency intelligence assessment, The 22 September 1979 Event.
In the aftermath of the event the NSC had asked for an estimate, based on the assumption that a test had taken place, of which country or countries might have been responsible. Producing the estimate was the responsibility of the national intelligence officer for nuclear proliferation, who coordinated his effort with the intelligence community representatives of the Interagency Intelligence Working Group on Nuclear Proliferation.35
The report, completed in December, began by stating that the technical analyses suggested an atmospheric nuclear detonation had taken place near the earth’s surface, within a broad area that consisted primarily of oceans and was generally cloudy and with a yield equivalent to less than 3 kilotons.36
It then mentioned and dismissed a variety of possibilities. Pakistan and Taiwan “probably lacked sufficient fissile material for even a single nuclear explosive device.” Brazil, Argentina, and Iraq “almost certainly lacked the fissile material and non-fissile components required to fabricate and test nuclear explosive devices.” China and France had not signed the partial test ban treaty and were free to test in the atmosphere. They also lacked any “technical or political motivation” to conduct a clandestine test in the southern Indian or Atlantic Ocean.37
The Soviet Union was also mentioned and dismissed, although not without dissent, on the grounds that it would have to assume “inordinate political risks” in its relations with the United States to conduct a covert atmospheric test in violation of the partial test ban treaty. The DIA felt differently, arguing that if an atmospheric test were in the Soviet Union’s technical interest, “an anonymous test near an unwitting proxy state such as South Africa could have provided an attractive evasion method.” The Energy Department suggested that while the Soviets had no technical reason to conduct such a test, it could serve a political purpose—creating suspicion that South Africa was the guilty party, disrupting peace efforts, and polarizing moderate elements in southern Africa.38
The possibility of an accident—the “unintended firing and near-surface detonation of a nuclear weapon during military exercises”—was considered and judged unlikely. While an unintended explosion would have produced the double-flash signature, it would have been of lower yield than what was indicated by the Vela signals. In addition, an accident would have required multiple safety measures to have been neutralized. The memorandum also noted the absence of any known weapons carriers in the area on September 22, as well as the absence of any other signs that would be consistent with such an incident, including the disappearance of any nuclear weapons carriers or the presence of ongoing search-and-rescue operations.39
The report then, not dissuaded by official denials, turned to the two prime suspects: South Africa and Israel, either individually or jointly. The authors reviewed the history of South Africa’s nuclear activities, including the aborted test of 1977, and the role of Prime Minister P. W. Botha during his tenure as defense minister in expanding his government’s military capabilities. They commented that if Botha decided in favor of a test, an atmospheric test “over unfrequented international waters,” while entailing some risk of being discovered violating the partial test ban treaty, “would have offered a relatively quick, safe, and easy way” for South Africa’s weapons designers to test a device without leaving behind clear evidence. In contrast, an attempt to test underground would have been more likely to be discovered ahead of time, as in 1977, and would have left tangible indications of a detonation.40
In addition to examining South Africa’s motivation to test, the analysts looked at some circumstantial evidence. The Simonstown harbor and naval base had been declared off limits from September 17 to 23, while the Saldanha naval facility, whose tenants included a naval search-and-rescue unit, suddenly was placed on alert for September 21 to 23. Neither event, however, could be taken as a dramatic clue. The memorandum reported that while the Simonstown closure could have helped screen sensitive loading or unloading operations as well as ship movements, the U.S. defense attaché had been told by “several reliable sources” that harbor defense exercises had been conducted during the seven days beginning September 17. The defense attaché noted that the closure was “a regular practice linked to internal defense.” And while the alert at Saldanha appeared “unusual,” in that no explanation was given and no activity was observed around or in the port, the analysts were unable to state with any assurance that the alert was unique. In addition, Gen. Magnus Malan, chief of South Africa’s Defence Force, was reported to be in South America when the Vela detection occurred.41
Also noted were statements by Botha that hinted at South Africa’s possession of nuclear weapons. On September 25, three days after the double flash, he told a provincial congress of his ruling National Party that “South Africa’s enemies might find out we have military weapons they do not know about.” A month later, on October 24, during a dinner attended by past and present members of the AEB, he was reported to have paid tribute to South African nuclear scientists who had been engaged in “secret work of a strategic nature.” Their names, he said, could not be mentioned and they would never receive the recognition they deserved. Also of interest were the words and silence of foreign minister Roelof Botha, the day after the prime minister’s dinner remarks. He ridiculed speculation that the Vela had detected a South African test, but declined to say, when asked, that South Africa had not been responsible.42
Most skeptical of South Africa having conducted a test was the State Department’s Bureau of Intelligence and Research, which considered the evidence against South Africa “inconclusive.” While the INR believed that South Africa had an ongoing nuclear program, had probably acquired enough fissile material for a device, and might eventually take the risk of testing, the bureau also believed that the same factors that deterred South Africa in August 1977 were still effective in September 1979. The State Department’s intelligence organization did agree that if a test had taken place, South Africa was the most likely perpetrator.43
The possibility of a secret Israeli test also received significant attention. Several different circumstances could have pushed Israel to test in secret, and in the atmosphere, in violation of its commitment to the partial test ban agreement. The “Israelis might have conceivably foreseen [the need] . . . for . . . low yield nuclear weapons that could be used on the battlefield, or might have considered desirable a small tactical nuclear warhead for short range Lance surface-to-surface missiles,” the memorandum observed. In addition, Israeli strategists “might even have been interested in developing a fission trigger for thermonuclear weapons.” In absence of access to designs that had been tested, Israeli nuclear weapons designers would probably, it was asserted, want to test prototypes—and “a low yield nuclear test conducted at sea could have enabled them to make basic measurements of the device’s performance.”44
Political calculations, however, would mitigate against such a test. Israel’s leaders could not ignore a number of risks, including the adverse reaction from the United States, possible increased Soviet assistance to the Arab states, the likelihood of serious damage to the two-year-old peace treaty with Egypt, and the likely erosion of support among traditionally friendly West European states. In addition, the Department of Energy suggested that for Israel to explode a nuclear device in the vicinity of South Africa, and leave that country to take the blame, was not consistent with Israel’s policy or attitude toward its fellow pariah.45
THE POSSIBILITY of a joint Israeli–South African test, with South Africa testing Israeli designs in exchange for technical information, was also evaluated in the memorandum. According to its authors, both nations would have considered the trade-off between the benefits of cooperation and the security risks involved in a joint effort. Israel would have expected responsibility for a test to attributed to South Africa, and “would have calculated that South Africa, as a pariah state in need of reliable friends would have had every reason to preserve security and remain silent in the face of speculation about the Israeli role.” For such a test to be worth the risk for South Africa, Israel would have had to offer advanced weapons technology. South Africa would probably have had enough confidence in Israeli security, the DIA suggested, to consider a joint test.46
While the interagency memo had noted that as of September 1979 India probably had enough fissile material for a bomb, it did not really explore the possibility of Indian responsibility. Former Atomic Energy Commission intelligence officer Arnold Kramish did, in an article published in the summer of 1980. Kramish argued that the signals detected by Vela 6911 were most likely those from a nuclear blast and asked the same question the U.S. intelligence community had asked: who was most likely to be responsible?47
Kramish dismissed Israel, South Africa, and Pakistan as likely culprits. Israel would not want to risk disruption of the implementation of its peace treaty with Egypt and the wrath of the United States, while South Africa was concerned with the stability of the transition from white rule in Rhodesia-Zimbabwe. Pakistan’s possession of a bomb was, Kramish believed, questionable. While he had no direct or circumstantial evidence to offer, Kramish suggested that the momentum of Indian nuclear weapons development as well as concern over the threat from Pakistan may have pushed India in the direction of a test. Employing the Pokhran test site would have been “politically dangerous and unlikely to escape detection.” The presence of Indian navy ships or scientific vessels in the southern Indian Ocean would not be unusual, he wrote. The locale would also divert suspicion to South Africa.48
Sam Cohen, best known for his key role in development of the neutron bomb, had another candidate. Cohen, writing in early 1981, dismissed Israel and India as candidates, in part because they had signed the partial test ban treaty. They would not risk, he argued, the consequences of violating it. Cohen also suggested that the lack of radiation in the area after the Vela detection could be explained by a test of a neutron warhead. South Africa, never known to have tested any bombs at all, was unlikely, he argued, to start with a test of a sophisticated device.49
“But now take France,” he wrote. The French would have a much different risk equation, never having signed the partial test ban treaty. They had halted atmospheric tests in French Polynesia only because of political pressure. Since an atmospheric test would be the best way to measure the effects of a neutron bomb, which, Cohen noted, many observers believed France was developing, a covert test in an area where the risk of detection was low might be an attractive option.50
Cohen had two pieces of circumstantial evidence to offer. If Vela 6911 did register a test, it occurred within the three-thousand-mile-wide circle covering the Indian Ocean, South Atlantic, southern tip of Africa, and a small part of Antarctica. In that area lie the Kerguelen Islands, one of the areas once considered as a possible French nuclear test site, where the French had maintained a small scientific center with a staff of about one hundred since 1950. Cohen also noted that in June 1980, France had announced that it had tested a neutron bomb, “but not when and where.”51*
BY THE TIME the intelligence community’s analysis of who was likely to be responsible for a test was completed in December 1979, and as collection efforts continued, a number of panels or organizations tried to analyze the available data and to evaluate the possibility that the Vela signal was the result of a malfunction or natural phenomenon—efforts that would produce a series of reports over the next several years.
On September 25, just as the AFTAC aircraft had begun their air-sampling operations on the other side of the world, Richard Garwin, a University of Chicago–trained physicist and longtime adviser to the CIA and NRO on satellite reconnaissance issues, was asked to come to CIA headquarters for half a day. Harold Agnew, director of the Los Alamos National Laboratory, and Steve Lukasik, the former director of the Advanced Research Projects Agency, joined him to serve as three-man panel “to render some judgment as to whether this actually had been a nuclear test,” Garwin recalled.52
Garwin reported that “all we had were Vela reports” and “there were no other data . . . and there were little negative data because there had not been enough time to determine whether other detection systems had received similar signals or not.” He suggested waiting for more information to be gathered from debris collection and underwater sensors. But when pressed to give a best estimate because the “CIA had to tell the president something” before an item appeared in the newspapers, he came up with a 60 percent probability that there had been a nuclear test.53
The next month Garwin received another summons, this time asking him to serve on a panel established by presidential science adviser Frank Press to evaluate the incoming data on the September 22 event. Eight other scientists, with a variety of backgrounds, would join Garwin as members of the committee. They included William Donn of the Lamont-Doherty Geological Observatory, Ricardo Giacconni of the Harvard Smithsonian Center for Astrophysics, Richard Muller of the University of California at Berkeley, Wolfgang K. Panofsky of the Stanford Linear Accelerator Center, Allen Peterson of the Stanford Research Institute, and F. William Sarles of MIT’s Lincoln Laboratory. Also serving was a scientist with a long history of involvement in nuclear detection issues—Manhattan Project veteran Luis Alvarez, then at the University of California at Berkeley. Jack Ruina, an old friend of Press’s, a professor of electrical engineering and computer science at MIT, and former head of the Pentagon’s Advanced Research Projects Agency, was appointed chairman. Ruina described the group as a panel of people from across the political spectrum.54
Ruina and his colleagues were given a three-part mission: to review both classified and unclassified data that could help determine whether the Vela signal had been the result of a nuclear detonation, to consider the possibility that the signal was a “false alarm” resulting from a satellite malfunction, and to investigate whether the signal might have been of natural origin, possibly the result of two or more natural phenomena. The group would begin work on November 1, supported by the Office of Science and Technology Policy (OSTP), which Press headed.55
BEFORE THE END of that month several technical analyses were completed. Los Alamos scientist Guy Barasch examined the possibility that the flashes detected by Vela 6911 were the result of a natural phenomenon. Vela bhangmeters had been triggered hundreds of thousands of times by lightning, cosmic particles, and direct sunlight. He concluded that naturally occurring signals would not be confused with signals from a nuclear detonation, whose light signature is “unmistakable.” While pulsed light sources that matched either the intensity or the duration of a nuclear detonation occur in nature, or could be built, no known source matched both characteristics. In particular, Barasch dismissed the possibility that the Vela signal came from a lightning “superbolt,” lightning that is over a hundred times more intense than typical lightning and usually occurs over water when cold polar air moves in over warm, moist oceanic air. Barasch noted that “to achieve the pulse shape and peak-radiated power simulating a one-kiloton nuclear explosion, lightning would have to be both 400 times more energetic and 100 times longer in duration than ever observed for the superbolts.”56
Another two technical studies would follow in December and January. Possible Origins of Event 747 Optical Data, written by three employees of the Santa Barbara–based Mission Research Corporation—Dale Sappenfield, David Sowle, and Trella McCartor—appeared in December. The three authors examined the same possibility that the Press panel was asked to consider—that the Vela signal was of nonnuclear origin. They would find a possible explanation in the reflection of sunlight off some small irregularly shaped object, provided it passed in front of the Vela sensors with the proper trajectory. However, all such objects would be “highly contrived” and would have to be matched to a “restricted trajectory.” In addition, they were unaware of other occurrences where such objects created double flashes that were clearly not caused by nuclear detonations. As a result, it was hard, Sappenfield and his colleagues wrote, to believe that the first time such an object produced a double flash, it made one so similar to the signal created by a nuclear detonation.57
The trio also considered two factors that could raise doubts about whether Vela 6911 had actually detected a nuclear detonation. One discrepancy was “late first maximum”—that the maximum intensity associated with the first flash occurred later than expected relative to the maximum of the second flash. The second factor, which would play a significant role in the Press panel’s conclusions, was that the two bhangmeters did not yield equivalent or “parallel” readings for the maximum intensity of the second flash, as would be expected when the event they were sensing was many thousands of miles away.58
The authors did not find the first anomaly particularly troubling, believing that it was the result of “experimental conditions surrounding the nuclear explosion.” The different readings produced by the bhangmeters for the second flash were of far greater concern. The authors considered and rejected several potential explanations, including “smog, X-ray veil, surface effects, and bomb mass effects.” The malfunctioning of one of the bhangmeters was the only explanation they could find, the probability of which they considered “much higher than the probability of any non-nuclear explanation of either or both sensor signals.” A nuclear detonation was thus the only possible source of the Vela signal that did not seem “very improbable” to the trio.59
In January, two scientists from the Stanford Research Institute reported on their rush evaluation of the possibility that the double-flash signal resulted from a meteoroid. The only conclusions they were able to offer after their three-week study was that the scenario suggested by the Mission Research Corporation (MRC) authors and the one suggested by Gary H. Mauth of Sandia, which involved two meteoroids, were extraordinarily unlikely. Two meteoroids were likely to be responsible for a double flash like the one received by Vela only once in one billion years. However, meteoroid data from the Pioneer space probe indicated that the MRC and Sandia models did not exhaust “the possible means for producing optical signals from meteoroids.” However, given the limited time for the study, the authors were unable to reach any firm conclusion concerning the probability that the Vela signal was produced by a meteoroid encounter.60
THE REPORTS PRODUCED in late 1979 and early 1980 resulted from quick assessments, using whatever data was available, of the likelihood that the September 22 double flash represented a nuclear detonation. Authors and panels whose reports were published in the spring and early summer of 1980 had the advantage of more, although far from complete, data, as well as more time to analyze data on issues ranging from possible corroborative evidence to the feasibility of a sensor malfunction to a nonnuclear origin for the Vela signal.
Reports issued in May by Los Alamos and Sandia would not, specifically, attract public attention, although they reflected the opinion of those in the labs who worked on the issue, an opinion that would often be presented in the months ahead. The Los Alamos report was authored by Henry G. Horak, who had joined Los Alamos in the fall of 1967 after obtaining a doctorate in astronomy from the University of Chicago, developing an expertise in astrophysics, and teaching at the University of Kansas. In his paper, he noted that the bhangmeters on DSP Flights 6 and 7 did not trigger, and offered two explanations consistent with a nuclear test having occurred: the event did not take place within the satellites’ field of view, or the signal was weakened during transmission through clouds and, as a result, wasn’t strong enough to reach the necessary threshold of brightness for detection. It was also possible, he noted, that Vela 6911 had detected the blast through a break in the clouds, or through thin clouds.61
Horak confronted the issue of discrepant bhangmeter readings and suggested that the substantial difference in bhangmeter results for the second flash could have been caused by “optical background changes during the much longer second pulse.” The bottom line for Horak was that when one looked at all the data associated with the Vela signal—the double flash, the intensity of the flashes, and the time difference between the various portions of the signal—it was “strong evidence that a nuclear explosion actually produced Vela Alert 747.”62
The May 1980 Sandia report, authored by Gary Mauth, noted that none of the other satellites equipped with bhangmeters, including DSP and SDS satellites, had detected the double flash that the Vela did. He also disputed the notion of Sappenfield and his colleagues that the different bhangmeter readings could be explained by a malfunction, noting that laser calibration tests on Vela 6911, conducted on two occasions in November 1979, revealed no problems.63
Most importantly though, taking into account such factors as Vela 6911’s performance history and the very remote probability that the signal could be explained by encounters with meteoroids, Mauth came to the same conclusion as the Santa Barbara trio and Horak as to the likely cause of the Vela signal. It was reasonable to conclude, he argued, that the discrepancy in bhangmeter readings was the result of satellite motion enhancing the apparent strength of the signal recorded by the more sensitive bhangmeter during the second flash. As a result, the Vela signals were “fully consistent with those expected from a low-yield atmospheric [nuclear detonation].”64
BY THE TIME the Los Alamos and Sandia reports had been completed, Ruina and his colleagues on the Press panel finished their investigation, meeting for the third and final time in early April 1980. During meetings, the panel received briefings by AFTAC, the DIA, and other agencies. Not all those briefings were unanimously well received—Luis Alvarez recalled the DIA briefing as being one in which the group was shown “and quickly discarded confirming evidence from a wild assemblage of sensors.” Keeny recalls that Alvarez was “outraged by [the] DIA people,” believing they flagged possible “blips” in various sensor readings without considering whether such blips were part of the normal environment. One briefing was given to a select sub-group of the panel, and involved relevant human intelligence gathered by the CIA on topics such as the movements of South African and Israeli ships.65
In any case, the information received from those briefings, along with additional documentation, allowed Ruina and his associates to examine Vela performance, review the data that might corroborate a nuclear origin of the signal, and review the studies produced by Los Alamos, Sandia, and other agencies. They also commissioned and reviewed statistical studies of the light signals that had previously been detected by Vela satellites as well as computer models of natural phenomena that might have generated the double-flash signal.66
The Press panel’s written report, issued in late May, contained an assessment of the effort to obtain potential corroborating data, an analysis of the Vela signal and possible nonnuclear explanations, and the group’s conclusions.
The search for confirming data from other satellites carrying bhangmeters was unsuccessful, because “these other satellites were looking at different parts of the earth and due to weather conditions had very little coverage overlap with the . . . satellite that observed the light flash.” Nor were there any electromagnetic pulse or magnetic disturbance data that could be correlated with the September 22 signal. The debris collection effort, involving both aircraft and ground sampling, and hampered by the weather, also failed to produce positive results—a failure, the group observed, that did not provide conclusive evidence that no nuclear explosion had occurred.67
The group also did not put much weight in an acoustic signal received at a “distant recording site in the northern hemisphere at an appropriate time.” A second site in the Northern Hemisphere failed to detect an acoustic signal, as did sensors in Australia, which sound propagation models suggested would be the more likely recipient of such a signal—although AFTAC’s data indicated that no signal was likely to be received at any of those sites from a low-yield explosion. There was also, the panel concluded, a significant probability of a signal arriving “within the large time window allowed,” owing to the uncertainty about the location of the Vela signal.68
Ruina and his colleagues had also been briefed by the Naval Research Laboratory during their last meeting in April, on signals that were picked up at SOSUS sites. Signals a few decibels above background noise were detected at the several sites at times consistent with their direct arrival from a source near Prince Edward Island, far from shipping or commercial routes, and for rays reflected from the Antarctic ice shelf. The report noted that the data was “analyzed by a filtering procedure that was not normally employed with SOSUS data.” In any event, the presidential panel considered the NRL study too incomplete to serve as corroborating data because the data on the frequency and strength of background signals was insufficient to resolve “the ambiguity in signal identification and source locations.”69
The traveling ionospheric disturbance detected at Arecibo was not considered “useful evidence” at the time the panel concluded its deliberations. The Arecibo scientists who detected that ripple, Richard Behnke and Lewis Duncan, met with the panel in Washington, a meeting they said was “mass confusion . . . an exercise in distraction.” Members of the panel disputed their major findings—that the ripple came from a source at least 310 miles away, moving from south to east at a speed, between 600 and 750 meters per second, typical of ionospheric ripples. The group claimed that Behnke and Duncan had made major errors in their calculations, which resulted in fallacious findings concerning the direction and speed of the disturbance. Duncan would say that it surprised him that “people have tried as much as they have to discredit [their findings] and that critics had failed to appreciate the sophistication of the incoherent-scatter radar (the radio telescope) at Arecibo and the methods used.” Duncan, himself, was not fully convinced that Arecibo had detected a nuclear test.70
Panel member Richard Garwin agreed that the ionospheric ripple was the most plausible corroborating signal considered, and that its movement from south to north was “striking and unusual” since most disturbances moved in the opposite direction. However, Garwin added, the record of observations using the advanced radar at Arecibo was too small to allow many generalizations. In addition, he doubted that such a large disturbance could have been created by a low-yield detonation like the one Vela 6911 appeared to detect.71
In the absence of data confirming the detection, the panel placed great importance on its analysis of the Vela signal. The group’s report noted that it “has the right duration and the characteristic double-humped shape was recorded by both bhangmeters.” Three separate means of producing an estimate of yield, normally derived from the time of the maximum intensities of the first and second flashes, were also in rough agreement, which further supported the hypothesis that a nuclear blast had been detected.72
However, the panel argued, before accepting the proposition that the double flash represented a nuclear detonation, it was necessary to demonstrate that the signal had no additional characteristics that ruled out a nuclear origin. It was also important that no alternative explanation was more likely to be true—that is, another class of signals of which the September 22 signal was more likely to be a member.73
One potential problem for the scientists was that the total intensity of light detected by Vela 6911 was “considerably larger than expected for a hypothesized explosion with this measured yield.” The anomaly could be explained if the signal had been transmitted through clear skies—if the region where the light source originated had been essentially free of clouds. At the same time, the absence of nuclear debris could only be explained, they wrote, if there was heavy cloud cover and “local rainout.” Those seemingly contradictory requirements could be reconciled if the light were transmitted through a small local gap in the clouds, a possibility that Horak had noted in his paper.74
But, as with others who had examined the Vela detection, the most troubling issue for the nine members of the Press panel was the discrepancy in bhangmeter readings at the peak intensity for the second flash, which, they stated, had not happened with the twelve previous nuclear detonations detected by Vela 6911. As the panel explained, the readings for the bhangmeters would not be expected to be identical, since one was more sensitive than the other, but if on one occasion one bhangmeter recorded a 20 and the other a 10, then on other occasions when that first bhangmeter recorded a 20, the other should also record 10 or a nearby value, with some variation possible due to changes in the background during the recording of data. But in the case of the September 22 signal, the size of the discrepancy, the panel wrote, “assumes major significance,” and “throws doubt on its interpretation as a nuclear event.”75
The report went on to explain that “during the second hump, the ratio of bhangmeter signals is significantly different from what would be expected from a nuclear explosion near the surface of the earth. Such anomalous behavior was never observed in bhangmeter recordings of previous nuclear explosions. Thus, although the September 22 event displays many of the characteristics of nuclear signals, it departs in an essential feature.”76 That observation, was followed by another:
It is very difficult to account for such a departure if the source of the September 22 signal was at a great distance from the bhangmeters, i.e. on the surface of the earth. On the other hand, if the source of the September 22 signal were close to the satellite sensors, the relative intensity of the light incident on the two bhangmeters could be quite different from cases where the source is far away. That is, an object passing near the satellite might be more in the field of view of one sensor than the other, whereas at a distance the field of view of both sensors is essentially the same.77
The panel noted its consideration of a number of alternative explanations, including unusual astronomical events, ordinary lightning, superbolts of lightning, sunlight reflecting off other satellites, sunlight reflecting off meteoroids passing near the satellite, and sunlight reflecting off particles ejected from the collision of meteoroids upon impact with the spacecraft. The group dismissed the first five alternatives—meteoroids of sufficient size were considered too rare and traveled too rapidly through the field of view to generate the double-flash signal with the required time difference between the first and second flash, while other satellites were too far away to reflect enough light to trigger the Vela bhangmeters, for example.78
The sixth alternative was another matter: “a meteor impact with the Vela satellite appears to be the best candidate for a nonnuclear origin of the signal.” Such an impact, the scientists argued, could generate a large number of particles with greater mass than the meteoroid, particles that would move with a low speed relative to the satellites. They could well generate “the complicated time histories seen in the unexplained zoo events as well as in the September 22 event.”79
Panel member Luis Alvarez recalled that “some on the committee proposed that a micrometeorite might have struck the satellite and dislodged a piece of its skin. Reflecting sunlight into the optical system on one sensor but not into that of its neighbor, the debris might have caused the questionable event. We constructed a believable scenario based on the known frequency of such micrometeorite impacts that reproduced the observed light intensity and pattern.”80
The meteor impact theory became the panel’s best explanation of what happened on September 22, although it could not rule out the possibility that the signal had a nuclear origin. Ruina said that his group had begun its work assuming it would conclude that the Vela 6911 signal was the result of a nuclear detonation, but by their last meeting the consensus was that the difference between known blast signals and the double flash detected on September 22 was too great to accept the flash as proof of an explosion in the absence of corroborating data. One panel member summarized the shift in opinion in terms of odds, saying that “on the first day we were betting four-to-one that it was an explosion, and at the end we were betting four-to-one that it was not.”81*
THE PANEL’S CONCLUSION had been expected, and challenged by some, as far back as January. An article in the Washington Post reported that “scientists at [Los Alamos], in the Department of Energy and even a few technical people at the State Department question why there should be such equivocating” since “every sign they’ve seen identifies the event of September 22 as a nuclear explosion.” An anonymous State Department official told the Post that “the Vela satellite picked up a signature like this 41 times before. . . . In every one of those 41 instances, there was never any question about the fact that a nuclear test had taken place. Each of those 41 was undeniably a nuclear explosion. This one was too.” A Los Alamos scientist, noting that the second pulse of a nuclear detonation produces exactly ninety-nine times more light than the first pulse, stated, “This is what the Vela saw the night of September 22.” Subsequent to the report’s release, one expert at Sandia asserted that the White House’s reflected sunlight theory “strains credibility.”82
Not surprisingly, just as reports prior to the Press panel’s study concluded that the September 22 detection had resulted from a nuclear blast, so did subsequent reports. The DIA report, The South Atlantic Mystery Flash: Nuclear or Not?, was released to a select audience in late June 1980. It was written by Dr. John E. Mansfield, who held a doctorate from Harvard in physics and worked for 1965 Nobel Prize winner Julian Schwinger, and Lt. Col. Houston T. Hawkins, who had recently joined DIA, as head of the Nuclear Energy Division’s Nuclear Weapons Branch, at the time of the incident. The authors noted that much of the search for corroborating evidence failed–that while data “possibly related [to the double-flash] were . . . found in the records of a number of instruments . . . in each [case], the signal was very weak, embedded in noise, or of a phenomenon not well understood.” As a result, intelligence analysts and review panels tended to discount those signals as possible corroboration of a nuclear blast.83
Among the topics discussed by Mansfield and Hawkins were the absence of detected radioactivity, the traveling ionospheric disturbance detected at Arecibo, and the possibility that one or more micrometeoroids caused the signal. They noted that explosions near or at the ocean’s surface result in a large volume of sea water being vaporized or physically lofted into the radioactive cloud. The water then immediately begins to fall back to the ocean, so that a large portion of the debris from the explosion falls into the ocean just a few thousand yards from the point of detonation.84
Their treatment of the ionospheric disturbance that Behnke and Duncan detected included both a detailed description of the detector and a calculation of the probability of one or more northward-traveling ionospheric disturbances from random events passing the Vela circle at the Vela time as being not more than 0.02–making it very unlikely, although not impossible. In their discussion of micrometeoroids, they argued, based on the calculations performed by Stanford Research Institute scientists, that the odds of a single meteor observation with a peak intensity and duration, with no additional requirements such as rise time and double pulse, was one in one hundred billion.85
Although much of their reasoning, the data they employed, and even their conclusions remain officially classified to this day, within weeks after their report was completed, its conclusions were reported in the press. On July 15, 1980, both the New York Times and Washington Post carried stories asserting that the DIA report, although “hedged with uncertainties,” concluded that the signal detected by Vela 6911 came from a nuclear detonation in the South Atlantic. The stories also carried comments from anonymous White House and Defense Department officials disputing the conclusion reached by Mansfield and Hawkins. (Years later, Ruina would dismiss the study as being on “college freshman level . . . maybe college senior” from a technical stand-point,” while Richard Garwin recalled that it involved a statistical analysis that would have justified any hypothesis over time.)86
But the news coverage prompted the White House to do more than issue anonymous criticism. On the same day the stories ran in the Post and Times, the White House released an only-mildly redacted version of the Press panel’s report. The White House official said that it was purely a coincidence that the two papers came to light in the same week.87
Four days after Mansfield and Hawkins submitted their report, the NRL turned in a three-hundred-page study, Report of NRL Investigations Concerning the 22 September 1979 VELA Alert. The NRL’s involvement in the effort to determine the significance of the Vela detection began with a letter from John Marcum, OSTP’s senior adviser for technology and arms control, to Alan Berman, the director of research at the laboratory—an institution that began operations in 1923, and whose campus-like headquarters are located in an isolated corner of Washington, D.C. Berman had received his doctorate in physics in 1952 from a far older institution, Columbia University, where he was a contemporary of Frank Press’s. He had been named NRL’s director of research in 1967 and become accustomed to receiving requests from Press to look into some “strange problem.”88
Marcum asked that the laboratory undertake “an immediate study” of ionospheric data that might have a bearing on whether the bhangmeters on Vela 6911 had actually witnessed the light from a nuclear explosion. In addition, the NRL was to study any other signals that might be available through its own experiments or experiments undertaken by other observatories. The laboratory’s effort was dubbed Project Search.89 The major result of NRL’s work, its report, has never been released. However, the paper trail of memos and letters gives a partial view of the laboratory’s extensive research effort as well as some of the conclusions it reached.
On February 14, 1980, Berman, according to a sanitized memo, contacted a “foreign national” and asked him to contact several seismic observatories and find out if they had detected any unusual signals at the time of the Vela detection. The same individual was asked to determine whether an experiment in equatorial scintillations, which had a propagation path between Adelaide, Australia, and Papua New Guinea’s Manus island, or a particular high-frequency over-the-horizon radar had detected any unexpected signals. It would appear that the foreign national was Australian, and the radar was the Jindalee radar that began experimental operations in October 1976.90
The following day, a member of Berman’s staff, Frank Kelly, telephoned Dr. Robert A. Helliwell of Stanford University in search of data, information, and suggestions. Helliwell reported that he was supervising a project, which had stations in Antarctica and Canada, to make wideband synoptic recordings of ELF (extremely-low-frequency) and VLF (very-low-frequency) noise. The recorders were on for only a portion of each hour, but he promised to have his workers check to see if they were on at the time of the Vela sighting. He also provided leads about other American and foreign researchers who were monitoring ELF and VLF noise in the Antarctic, and suggested that information might be available from a French research station at Kerguelen Island. On February 25 Kelly called Dr. Nelson Spenser of the Goddard Space Flight Center because he had been told that Spenser’s Atmospheric Explorer Satellite might have detected ionospheric disturbances associated with the September 22 double flash. Spenser agreed to check to see if the satellite’s sensors had been turned on, and if so, if they had recorded any signal of interest. The same day Kelly was in touch with an individual connected with ship-based VLF studies, who also promised to find out if any relevant data had been gathered.91
John Goodman, whom Berman had released from all other duties the day after Marcum’s letter arrived so he could work on the Vela investigation, spent part of February 27 inquiring about the availability of infrared and other data from the Landsat earth resources satellite and the Tiros and Nimbus weather satellites. Goodman also penned a brief memo that day, recording that he had requested and was expecting copies of images taken by an air force DMSP spacecraft on September 22. Meanwhile, Berman spoke to an official from another organization, possibly AFTAC, who was responding to the research director’s question about signals that might have been detected by a particular seismic station. The results were negative, as they were with respect to another station that might have detected a signal.92
Goodman also informed Berman that day of a series of findings: examination of data from the navy’s Omega land-based VLF communications system as well as the service’s Transit navigation satellite system, which used very high and ultra-high frequencies, proved negative. In addition, Goodman’s memo raised questions about the Arecibo data. While it was apparent that Behnke and Duncan had detected an ionospheric disturbance, a “back of the envelope calculation” did not support the idea that it originated from an explosion in the South Atlantic, off South Africa. However, it was possible, Goodman reported, that the mathematical analysis being employed was not the proper one for a surface explosion.93
During the first days of March, John Goodman explored the possibility of obtaining data from a number of satellite systems. He had discovered the First GARP (Global Atmospheric Research Program) Global Experiment, a joint program of the World Meteorological Organization and the International Council of Scientific Unions. The program involved five geosynchronous and three polar orbiting weather satellites, as well as a number of ocean buoys that measured surface atmospheric pressure and sea surface temperature—sixteen of which were located at longitudes and latitudes monitored by Vela 6911. It was hoped that the weather satellite data would reveal the presence of any storms near South Africa and Arecibo.94
On March 5 Goodman received a call concerning some of the data obtained by the DMSP. One of the military weather satellites did “see an enhancement” of high-frequency noise at the approximate time of the double flash, near the conjectured location of the possible test. The noise, however, could have resulted from environmental conditions. At the same time, a more active cause, including lightning or a nuclear detonation, was possible. In a memo written that day, Goodman observed that the DMSP data “would appear, at present, to be quite interesting.” The NRL was also processing data concerning the electron content of the ionosphere at the time of the Vela signal in its search for corroborative data. In addition, there was hope that a military research satellite, launched in January 1979 and known as SCATHA (an acronym for Spacecraft Charging at High Altitudes), might have relevant data from a high-frequency experiment.95
NRL representatives continued to reach out to a number of American and foreign institutions that might have inadvertently gathered data relevant to the investigation. Frank Kelly was trying to obtain high- and low-frequency magnetic field data from a Japanese research station in Showa, Antarctica, and the Japanese Radio Research Laboratory was expected to provide data. VLF data from Kerguelen was also due to arrive at the NRL. The Smithsonian Astrophysical Observatory in Massachusetts had been asked to provide VLF records from sites in Latin America, the Middle East, Australia, and the United States. Even Air Traffic Control tapes were being searched for data. The remote possibility that neutrinos associated with the conjectured nuclear explosion had been detected by a massive neutrino detection system at the University of Pennsylvania led the NRL to ask the professor who ran the system to determine if any unusual background counts were observed. Consideration was also given to obtaining samples from American ichthyologists who were due to be aboard a French supply ship traveling to the Crozet Islands in the South Atlantic. The Americans would be collecting samples of the islands’ fish fauna for their studies, fauna that might contain radioactivity transported by the air or water from the Prince Edward Island area, about six hundred miles away.96
Three weeks before the NRL turned in its final report, Lothar H. Ruhnke, the head of the research department’s atmospheric physics branch, summarized some of the preliminary results and made some recommendations. He noted that an analysis was conducted to determine the probability of detecting radioactivity in the air and rainwater from a 2-kiloton explosion on September 22, 1979, at Marion Island. An important factor in the analysis was the major cyclone that approached the island on the day of the double flash and moved eastward. Based on satellite data and conventional weather maps, a trajectory analysis showed that a possible radioactive cloud was caught by the storm and stayed with it until at least September 28.97
Comparing the trajectories of the radioactive cloud, its horizontal expansion, and the paths of the AFTAC aerial sampling missions led to the conclusion that only one mission, flown on September 28 to 58° south latitude and 150° east longitude, could have intercepted the radioactive cloud. By that time, “fission products had decayed by rainout below the level of detectability with the exception of the gamma count.” Later, Ruhnke noted that it was necessary to conclude that if, under the prevailing meteorological conditions, there had been a 2-kiloton explosion on September 22, it could not be detected by airborne sampling methods more than five days later (that is, after September 26).98
Nor could such a blast be detected by the existing ground-based sampling network under average meteorological conditions. Ruhnke suggested that it might be possible to find evidence in snow samples from the glaciers of the Kerguelen Islands. Even one year after a detonation, the snow should still show considerable signs of radioactivity. He recommended an expedition to obtain radioactivity profiles of the snow on a glacier, down to a depth of one meter. His suggestion was passed on to Frank Press by Berman. But the French, according to Berman, “told us to go pound sand.”99
Such NRL collection and analysis efforts were undoubtedly described in its lengthy report. The most significant element of the report, which, in conjunction with the Vela signal, led Berman and his associates to believe a nuclear device had been detonated on September 22, was the analysis of the hydroacoustic signals detected—the same signals the White House panel had dismissed as inconclusive, but which Berman describes as being comparable to those received from French tests in the Pacific.100
The September 22 signals had been obtained from two sets of sensors: the SOSUS hydrophones, used by the navy to monitor Soviet submarine movements and by AFTAC to detect nuclear detonations, and the sensors of the Missile Impact Location System (MILS), which had been deployed throughout the South Atlantic to measure when and where test missiles fired from Cape Canaveral splashed down—information vital to evaluating missile accuracy. To determine if the signal was possibly the result of natural phenomena, the NRL searched the logs for “every minute of every day” for thirty days before the event and thirty days afterward and found none.101
The NRL scientists used the SOSUS/MILS data in conjunction with a working assumption: if the Vela 6911 signal had been from a nuclear test, then the most likely site was in the Prince Edward Island–Marion Island area. If one wanted to conduct a clandestine test, that area was a “splendid place to go,” Berman notes. The islands’ high mountains would be a good place to locate observation sensors. A barge carrying a nuclear device could be placed in the shallow water near the islands.102
That hypothesis was reinforced by calculations done by the Naval Observatory, which Berman had asked to determine the time of sunrise in the Prince Edward Island area—knowing that weapons designers preferred tests to take place shortly before then. Such timing allowed the blast to be seen and measured against a dark background, as well as permitting aircraft monitoring the blast effects to launch shortly after detonation. The answer that came back from the observatory placed sunrise ten minutes after the Vela detection.103
Analysts under Berman’s direction then developed a model of the hydro-acoustic signature that might be expected from a blast at Prince Edward Island. A massive acoustic signal would be expected to travel south, bounce off Antarctica, and travel north. Using the laboratory’s sophisticated computer capability, the analysts calculated the time when the original signal should have reached Antarctica, and then the speed at which it would have traveled north. It was then possible to compare the theoretical times with records of the SOSUS and MILS network. Berman recalls that a signal arrived “within a few seconds” of the time such a signal would have been expected to arrive at the SOSUS and MILS sensors in its path. It was “not positive proof,” Berman notes, but it was “interesting evidence.”104
After its report was completed, the NRL continued to receive new information and to analyze it. In late July, Berman wrote Press adviser John Marcum to report that Marion Island ionosonde records from late on the night of September 21, 1979, extending to October 1, 1979, had been received and subjected to a “very preliminary analysis.” That analysis showed an “effect of currently unexplained origin” between 2:45 and 3:00 a.m. on September 22. The “rather striking anomaly” was a “a major biteout or depletion of the ionospheric electron density” in the vicinity of Marion Island, while no similar effect appeared in the records for Johannesburg, Kerguelen, and Grahamstown.105
On September 25, 1980, Berman received a letter from Dr. L. Van Middlesworth, a professor in the physiology and biophysics department at the University of Tennessee’s college of medicine who for the previous twenty-five years had been examining samples of sheep thyroids from around the world. Van Middlesworth wrote that he had detected iodine-131 in the thyroids of sheep slaughtered in Melbourne, Australia, in November 1979. He also reported that subsequent to that time, no evidence of iodine-131 was ever detected in Australian sheep thyroids. In early November, Berman informed Marcum that based on their analysis of the data, the head of the radiation survivability and detection branch of the NRL and his staff “believe that Dr. Van Middlesworth’s data constitutes a positive case for the proposition that Australian sheep ingested the fission product [iodine-131] during the month of October 1979.” Further, Berman wrote that it was not “inconsistent with the observed evidence” that iodine isotope detected in the thyroids “could have been associated with a postulated nuclear detonation in the vicinity of Prince Edward and Marion Islands.”106
AT THE TIME of the release of the Press report, the NRL’s conclusion was not known outside of government circles, but at the end of August 1980 it was revealed in the pages of Science magazine. The disclosure came not from an anonymous source at the NRL, but from research director Alan Berman. Berman, who was irritated by the Press panel’s dismissal of his organization’s work as “incomplete” and “ambiguous,” pointed out that when the panel’s report was drafted in April, the NRL had not yet completed its work. He also noted that the NRL had assigned seventy-five people to work on the project for several months, whereas the White House panel “undertook no study of its own [but] listened to presentations. . . . They heard various opinions and came to their own.” The preponderance of evidence indicated that a nuclear blast had taken place, according to Berman. And, in response to a question from the Washington Star’s John Fialka, the NRL research director remarked that “a hydro-acoustic signal was detected which had characteristics similar to those received from known nuclear detonations.”107
Science also reported the criticism by a White House staff member of the navy lab’s analysis. The anonymous staffer called the issue of the hydro-acoustic signal a “dead horse.” He asserted that the NRL study was fraught with ambiguity because two signals had been detected: a weak one followed by a strong one.* If they came from the same source, he continued, the initial signal had to have arrived directly, while the second was a reflected signal. Most of the mathematical analysis was based on the second signal, and no amount of sophisticated mathematics, according to the staffer, could determine with confidence the origin of the reflected signal. In addition, the aide claimed that if the event had originated at Prince Edward Island, as the NRL study assumed, it should have been detected by another satellite, but was not.108
IN THE IMMEDIATE YEARS after the completion of the Press panel, DIA, and NRL reports, the national labs as well as private contractors would complete additional studies. None would provide conclusive proof that a test did, or did not, take place.
In August 1980, Sappenfield and his two colleagues at Mission Research Corporation completed an expanded version of their December 1979 study. They reached the same basic conclusion as they had nine months earlier—that the likelihood of a nuclear explosion was much higher than the likelihood of any nonnuclear explanation for the triggering of the bhangmeters. They concluded for the first time that the reading of the less sensitive bhangmeter was more likely to be correct, and that the data from that sensor indicated a surface burst. They also argued that there were some serious flaws with the Press panel’s meteoroid theory—that the natural speed of objects spun off the satellite by a meteoroid would be one-tenth the speed required to generate the double-flash signature, and that the objects would be going in the wrong direction.109
In April 1981 a trio of Los Alamos scientists noted that about two minutes after the Vela detection, a Tiros-N weather satellite detected an “electron precipitation event.” While the event was “unusually large,” they did not find it to be unique, and suggested it was probably due to natural causes. They also noted that a patch of auroral light suddenly appeared in the sky above Syowa Base in Antarctica a few seconds after the Vela event, which was consistent with, but not proof of, an electromagnetic pulse resulting from a surface nuclear burst.110
In addition, the authors noted that it was possible to study natural and weapons-related perturbations of the ionosphere by monitoring long-path VLF transmissions, and that the U.S. Navy operated eight Omega VLF transmitters for navigational purposes. Since a nuclear explosion might have affected transmissions in the region, they examined transmissions along two different paths that cross that region. Neither yielded evidence of any abnormal variations at the time of the double flash.111
A January 1982 report by Carl J. Rice of the Aerospace Corporation, whose mission was to provide technical support to NRO and air force space programs, focused on the infrared data obtained by the two DSP satellites (Flights 6 and 7) whose footprints overlapped that of Vela 6911. A pair of signals, with the appropriate differences in time and intensity corresponding to a double flash, would suggest that a nuclear test had occurred, as would a single, but sufficiently bright, pulse. One problem facing the author was the lack of precise information about where the “detonation” took place. The initial signal from a low-yield detonation would not necessarily stand out from the variety of other infrared signals that the DSP spacecraft would pick up during their scans of the South Atlantic region. An initial signal of sufficient intensity would be noticed, as would a lower-yield signal at a precise suspect location. In the end, no confirmation of a test could be found—while at least one signal merited special attention, it was insufficiently intense and “very unlikely to represent the nuclear event in question.”112
In May, four Los Alamos scientists completed their study of the Vela signal. They discounted a number of explanations for the discrepancies in the bhangmeter readings, including atmospheric absorption of the signal or cloud cover. They went on to present their own model for the September 22 event, a model that remains largely classified but may have relied on the effect of surface bursts on bhangmeter readings. However, it does seem likely that the authors shared the belief of many of their colleagues at Los Alamos and Sandia that the September 22 event was a nuclear test, as they wrote that “our model is consistent with the apparent absence of nuclear debris, the collection of which is required by some analysts for absolute confirmation of an atmospheric detonation.”113
WHILE THE OFFICIAL STUDIES of the Vela incident would decline over the years, there would be continued speculation and reporting in the media. One of the first allegations came on February 21, 1980, when CBS aired a story by correspondent Dan Raviv, who quoted from a book—None Will Survive Us: The Story of the Israeli Atom Bomb—written by Israeli journalists Eli Teicher and Ami Doron. Although a novel, it recounted Israel’s path toward nuclear weapons—apparently accurately enough that the Israeli censor refused to permit its publication. Raviv also interviewed the authors, who claimed the September 22 event was a nuclear test conducted by Israel with South African assistance. In response, a spokesman for Israeli defense minister Ezer Weizman quoted him as stating that “nothing like that took place.”114
A similar charge was made on December 21, 1980, when Israeli state television broadcast a British-made program on Israeli–South African nuclear cooperation, which alleged that the 1979 flash came from a test of a new naval nuclear warhead, developed jointly by South Africa and Israel.115
In his 1991 book, investigative journalist Seymour Hersh reported that Vela 6911 had detected, according to former Israeli government officials “whose information has been corroborated,” the test of a low-yield nuclear artillery shell and that the test was not the first, but the third Indian Ocean test. At least two ships from the Israeli navy had sailed to the site ahead of time, and a group of Israeli military officials and nuclear experts, as well as the South African navy, observed the test, according to Hersh. A similar accusation was made by Dieter Gerhardt, who was subsequently convicted of spying for the Soviet Union, but at the time was commander of the Simonstown naval base near Cape Town. He said that shortly before the Vela detection, a fleet of Israeli ships had made a port call at the naval base and that the flash was the result of a joint Israeli–South African test, Operation Phoenix. The test was supposed to have gone unobserved, according to Hersh’s source: “There was a storm and we figured it would block Vela, but there was a gap in the weather—a window.” It has also been suggested that Israel may have received information that the Vela bhangmeters were no longer working.116
In April 1997 an article in the Israeli Ha’aretz daily newspaper stated that South African deputy foreign minister Aziz Pahad confirmed that the double flash had resulted from a test, although he did not suggest that Israel was involved. In July the Los Alamos Laboratory News Bulletin ran an article based on the supposed revelation, with the headline “Blast from the Past: Lab Scientists Received Vindication.” The Albuquerque Journal also followed up on the apparent disclosure, reporting comments from David Simons, a Los Alamos physicist, who had joined the lab’s verification program in 1979, that similar discrepancies had been observed in Vela bhangmeter signals from earlier atmospheric tests.* It also reported the belief of one expert who helped design the satellite that the panel’s explanation of how a near-perfect image of a nuclear explosion could be caused by a meteoroid strained credulity. He added that when an explanation was needed, “the zoo animals come marching out of the woodwork.”117
Later that month Aviation Week & Space Technology ran a similar story, stating that “a South African government official has confirmed that his nation detonated a nuclear weapon in the atmosphere in September 1979.” But Pahad’s press secretary disputed the reports, stating that Pahad had merely noted the “strong rumor” that a test had occurred, and suggested the allegation should be investigated.118
MANY WHO directly confronted the issue at the time—whether as intelligence producers or consumers—believe that a test did take place and that Israel was probably responsible. Adm. Stansfield Turner, Carter’s director of central intelligence, believes the Vela detection of September 22 was of “a man-made phenomenon.” Several natural phenomena would have had to take place at the same time to simulate a nuclear blast, requiring a degree of coincidence Turner finds implausible. Leonard Spector, an expert on nuclear proliferation, who was at the time an aide to Senator John Glenn and received a number of intelligence briefings on the incident, recalls that he was “uncomfortable with the Ruina report.” His impression is that the accretion of evidence points to Israel.119
Another former Glenn aide who attended some of the same Vela incident briefings with Spector, as well as others that Spector did not have the clearances to attend, was his boss, Leonard Weiss. Weiss gave up his position as a professor of applied mathematics at the University of Maryland to join the senator’s staff, where he worked on nonproliferation issues for over two decades. When he first heard of the incident, he “jumped on it” and asked for a briefing. He recalls that his briefers were “too zealous in trying to tell us that nothing happened.” His initial conclusion was that a test had taken place, and he was prepared to say so in an interview with the CBS Evening News. Before he could, the White House warned Glenn that if such a claim was made, “all hell would break loose.” Glenn told his aide, as the television crew was setting up in Weiss’s office, that while he could go on the air, he could not make such an unequivocal claim.120
A few weeks after that interview Weiss finally received access to AFTAC’s data base of zoo events—access that he had sought, and originally been denied, given claims that the Vela signal might have been the result of a zoo event. Weiss recalls that when he examined the Vela zoo, he was “astounded”—it was clear to him that “this was no zoo animal.” Rather, it was the “classic wave produced by a nuclear explosion,” with its double hump. It was “unmistakable.” The inability to detect radiation associated with a test was not surprising, he remembers, because for some French tests, of which the United States had advanced knowledge, monitoring aircraft were not able to find the plume or found it late. He also recalls that there was “hardly anybody in the intelligence community who didn’t think it was a test.”121
That view was not shared by Press panel member Luis Alvarez. In his 1987 memoirs, he wrote, “I doubt that any responsible person now believes that a nuclear explosion occurred because no one has broken security, among South Africans or elsewhere.”* During the summer of 1994 many of Alvarez’s colleagues on the presidential panel met in San Diego at a meeting of the JASON group, which consists of scientists from outside the government who provide their expert advice to the Department of Defense. They received CIA briefings, largely on the lack of new evidence concerning the incident that had appeared in recent years, and, according to Jack Ruina, “found no reason to change their mind.”122
New information in the early 1990s suggested that South Africa was not responsible, at least not by itself, for any test that occurred on September 22, 1979.† As of 1995 the U.S. intelligence community had not reached an official consensus about the cause of the double flash. But according to one account, based on interviews with U.S. officials, “unofficially, the widespread view in the U.S. government was that the Vela satellite had detected the test of a low-yield nuclear bomb and that Israel alone was responsible.”123
Spurgeon Keeny, in early 2003, described himself as “uncertain” as to whether Vela 6911 detected a nuclear explosion on September 22, 1979. He noted that an Israeli technician studying at MIT, and “clearly an Israeli agent,” as well as a few Israeli officials hinted that Israel did conduct a test that day. However, he argued, if Israel had conducted a test, it would have involved a major expedition. The resulting spread of knowledge about the activity, along with the conflicts of personalities that make up Israeli politics, would have resulted in disclosure of the test, he believed. Ruina, in 2004, argued that if it was a nuclear explosion, it was clearly in the 2- to 3-kiloton range—in contrast to the standard yield of first tests, between 15 and 20 kilotons. That Israel might have detonated a low-yield neutron bomb is a hypothesis he considered “far fetched, then and now.”124
There remains a lack of hard evidence, and if Israel did indeed test a nuclear weapon that day, it is unlikely that former Israeli officials will come forth or that Israeli documents confirming Israeli involvement will emerge—at least any time soon.
___________
* Both pieces of evidence were weak. By 1979 there was a joint French-Soviet meteorological facility on Kerguelen, making it somewhat unlikely that the French would involve it in a covert nuclear test. In addition, a number of underground tests in French Polynesia during 1970s have been associated with the French neutron bomb program.
* Another panel, the CIA’s Nuclear Intelligence Panel (NIP), whose members included Harold Agnew, Louis Roddis Jr., and Edward Teller, reached a different conclusion. According to Donald Kerr, who chaired the panel’s study, and who had served in the Carter administration as acting director of defense programs at the Energy Department, and became the CIA’s deputy director for science and technology in 2001, “We had no doubt it was a bomb.” See Seymour Hersh, The Samson Option: Israel’s Nuclear Arsenal and American Foreign Policy (New York: Random House, 1991), pp. 280–281. The panel reviewed data from the event during a February 11–13, 1980, NIP meeting, according to a memorandum: Chairman, JAEIC, Memorandum for: Director of Central Intelligence, Deputy Director of Central Intelligence, Subject: Judgments of the DCI’s Nuclear Intelligence Panel on the 22 September 1979 Event, February 14, 1980. The panel’s report (Judgments of the DCI’s Nuclear Intelligence Panel on the 22 September 1979 Event, n.d..), attached to the memo, was released in 2004 in response to a Freedom of Information Act request—with all discussion of facts as well as the panel’s conclusions redacted.
* The aide was apparently referring to a weak signal that arrived before the signal that the NRL concluded had been reflected off Antarctica. Berman has noted the arrival of a very weak signal following the Vela detection. Interview with Alan Berman, Alexandria, Va., March 27, 2003.
* According to one individual involved in the Vela incident debate, the same bhangmeter discrepancy occurred on a previous Chinese test, possibly the atmospheric test of September 13, 1979. Richard Garwin recalls that the only significant discrepancy was between the September 22 event and all nuclear detonations detected by Vela 6911. Private information; e-mail from Richard Garwin, November 4, 2004.
* Given the decades-long veil of secrecy concerning many highly classified projects in the United States, including satellite reconnaissance projects, Alvarez was far from being on solid ground. Daniel Ellsberg has noted that “secrets that would be of the greatest importance to [the public, Congress, and the press] can be kept from them reliably for decades by the executive branch, even though they are known to thousands of insiders.” The veil of secrecy is far tighter in Israel, where military censors are able to review and withhold news considered damaging to national security. See Daniel Ellsberg, Secrets: A Memoir of the Pentagon Papers (New York: Viking, 2002), p. 43.
†See chapter 9.