
At the time, the Gulf of Tonkin incidents of August were not quite so controversial. According to the Johnson administration, the issue of the attacks was pretty much cut and dried. As the administration explained, our ships had been in international waters -- anywhere from fifty to eighty miles from the DRV coastline by some calculations, during the alleged second attack -- and were attacked twice, even though they were innocent of any bellicose gestures directed at North Vietnam. Secretary of Defense Robert McNamara had assured the Senate that there had been no connection between what the U.S. Navy was doing and any aggressive operations by the South Vietnamese.1
However, within the government, the events of 4 August were never that clear. Even as the last flare fizzled in the dark waters of the South China Sea on that August night, there were conflicting narratives and interpretations of what had happened. James Stockdale, then a navy pilot at the scene, who had "the best seat in the house from which to detect boats," saw nothing. "No boats," he would later write, "no boat wakes, no ricochets off boats, no boat impacts, no torpedo wakes -- nothing but black sea and American firepower."2 The commander of the Maddox task force, Captain John J. Herrick, was not entirely certain what had transpired. (Captain Herrick actually was the commander of the destroyer division to which the Maddox belonged. For this mission, he was aboard as the on-site commander.) Hours after the incident, he would radio the Commander-in-Chief, Pacific (CINCPAC) telling them that he was doubtful of many aspects of the "attack."
It would be years before any evidence that an attack had not happened finally emerged in the public domain, and even then, most reluctantly. Yet, remarkably, some of the major participants in the events still maintained that the Gulf of Tonkin incident had occurred just as it had been originally reported. Secretary of Defense Robert McNamara, in his memoirs In Retrospect, considered the overall evidence for an attack still convincing.3 The U.S. Navy's history of the Vietnam conflict, written by Edward J. Marolda and Oscar P. Fitzgerald (hereafter referred to as the "Marolda-Fitzgerald history"), reported that the evidence for the second attack, especially from intelligence, including a small amount of SIGINT, was considered conclusive.4
The public literature on the Gulf of Tonkin for years has been overwhelmingly skeptical about the 4 August battle. Articles that appeared in magazines within a few years illustrated the general inconsistency in the descriptions of the incident of 4 August by simply using the conflicting testimony from the officers and crews of both ships. The first major critical volume was Joseph Goulden's Truth Is the First Casualty, published in 1969. The most complete work to date is Edwin Moise's Tonkin Gulf and the Escalation of the Vietnam War. Moise's work has the dual advantage of using some Vietnamese sources, as well as small portions of a few SIGINT reports released to the author under a Freedom of Information Act request. Yet, even what few scraps he received from NSA were enough to raise serious questions about the validity of the SIGINT reports cited by the administration which related to the 4 August incident.5
The issue of whether the available SIGINT "proved" that there had been a second attack has been argued for years. In 1968, Robert McNamara testified before Senator William Fullbright's Foreign Relations Committee's hearings on the Gulf of Tonkin that the supporting signals intelligence was "unimpeachable." On the other hand, in 1972 the deputy director of NSA, Louis Tordella, was quoted as saying that the 4 August intercepts pertained to the 2 August attacks. In a 1975 article in the NSA Magazine Cryptolog, the Gulf of Tonkin incident was retold, but the SIGINT for the night of August 4 was not mentioned, except for the "military operations" intercept, and even then without comment.6 The Navy's history of the Vietnam War would misconstrue the SIGINT (disguised as unsourced "intelligence") associating portions of two critical intercepts and implying a connection in the evidence where none could be established.7
Except for the sizable collection of SIGINT material within NSA, and a much smaller amount from the archives of the Naval Security Group (which essentially duplicates portions of the NSA holdings), almost all relevant material relating to the Gulf of Tonkin incidents has been released. Although the questions about what happened in the Gulf of Tonkin on the night of 4 August have been fairly well answered by the evidence from all of the other sources -- radar, sonar, eyewitness, and archival -- the SIGINT version needs to be told. This is because of the critical role that SIGINT played in defining the second attack in the minds of Johnson administration officials. Without the signals intelligence information, the administration had only the confused and conflicting testimony and evidence of the men and equipment involved in the incident. It is difficult to image the 5 August retaliatory air strikes against North Vietnamese naval bases and installations being ordered without the SIGINT "evidence."8 Therefore, it is necessary to recount in some detail what signals intelligence reported.
For the first time ever, what will be presented in the following narrative is the complete SIGINT version of what happened in the Gulf of Tonkin between 2 and 4 August 1964. Until now, the NSA has officially maintained that the second incident of 4 August occurred. This position was established in the initial SIGINT reports of 4 August and sustained through a series of summary reports issued shortly after the crisis. In October 1964, a classified chronology of events for 2 to 4 August in the Gulf of Tonkin was published by NSA which furthered the contention that the second attack had occurred.
In maintaining the official version of the attack, the NSA made use of surprisingly few published SIGINT reports -- fifteen in all. The research behind the new version which follows i based on the discovery of an enormous amount of never-before-used SIGINT material. This included 122 relevant SIGINT products, along with watch center notes, oral history interviews, and messages among the various SIGINT and military command centers involved in the Gulf of Tonkin incidents. Naturally, this flood of new information changed dramatically the story of that night
of 4/5 August. The most important element is that it is now known what the North Vietnamese Navy was doing that night. And with this information a nearly complete story finally can be told.
Two startling findings emerged from the new research. First, it is not simply that there is a different story as to what happened; it is that no attack happened that night. Through a compound of analytic errors and an unwillingness to consider contrary evidence, American SIGINT elements in the region and at NSA HQs reported Hanoi's plans to attack the two ships of the Desoto patrol. Further analytic errors and an obscuring of other information led to publication of more "evidence." In truth, Hanoi's navy was engaged in nothing that night but the salvage of two of the boats damaged on 2 August.
The second finding pertains to the handling of the SIGINT material related to the Gulf of Tonkin by individuals at NSA. Beginning with the period of the crisis in early August, into the days of the immediate aftermath, and continuing into October 1964, SIGINT information was presented in such a manner as to preclude responsible decisionmakers in the Johnson administration from having the complete and objective narrative of events of 4 August 1964. Instead, only SIGINT that supported the claim that the communists had attacked the two destroyers was given to administration officials.
This mishandling of the SIGINT was not done in a manner that can be construed as conspiratorial, that is, with manufactured evidence and collusion at all levels. Rather, the objective of these individuals was to support the Navy's claim that the Desoto patrol had been deliberately attacked by the North Vietnamese. Yet, in order to substantiate that claim, all of the relevant SIGINT could not be provided to the White House and the Defense and intelligence officials. The conclusion that would be drawn from a review of all SIGINT evidence would have been that the North Vietnamese not only did not attack, but were uncertain as to the location of the ships.
Instead, three things occurred with the SIGINT. First of all, the overwhelming portion of the SIGINT relevant to 4 August was kept out of the post-attack summary reports and the final report written in October 1964. The withheld information constituted nearly 90 percent of all available SIGINT. This information revealed the actual activities of the North Vietnamese on the night of 4 August that included salvage operations of the two torpedo boats damaged on 2 August, and coastal patrols by a small number of DRV craft. As will be demonstrated later in this chapter, the handful of SIGINT reports which suggested that an attack had occurred contained severe analytic errors, unexplained translation changes, and the conjunction of two unrelated messages into one translation. This latter product would become the Johnson administration's main proof of the 4 August attack.
Second, there were instances in which specious supporting SIGINT evidence was inserted into NSA summary reports issued shortly after the Gulf of Tonkin incidents. This SIGINT was not manufactured. Instead, it consisted of fragments of legitimate intercept lifted out of its context and inserted into the summary reports to support the contention of a premeditated North Vietnamese attack on 4 August. The sources of these fragments were not even referenced in the summaries. It took extensive research before the original reports containing these items could be identified.
Finally, there is the unexplained disappearance of vital decrypted Vietnamese text of the translation that was the basis of the administration's most important evidence -- the so-called Vietnamese after-action report of late 4 August. The loss of the text is important because the SIGINT record shows that there were critical differences in the English translations of it issued both by the navy intercept site in the Philippines and
NSA. Without the individual texts (there were two of them), it is difficult to determine why there are critical differences in the translations and more importantly, to understand why two separate North Vietnamese messages were combined into one translation by NSA.
Before a discussion can begin, it is necessary to understand how the Gulf of Tonkin incidents came to happen, the way they did, and what their significance was for the Johnson administration. To do that, we need to consider the Desoto mission that the Maddox was conducting at the time, as well as the Defense Department's OPLAN-34A missions against the Democratic Republic of Vietnam (DRV). It was the convergence of the two that embroiled that ship in the crisis in the Tonkin Gulf.
Physically, Desoto mission destroyers were unique in their configuration -- a small van lashed to the ship which housed intercept positions for voice and manual Morse communications. There also was a position which intercepted noncommunications emissions such as radars, referred to as electronic intelligence or ELINT. Finally, a communications position, which allowed the detachment to send and receive messages from the other monitoring stations in the area, as well as other SIGINT organizations and commands, via the Criticomm communications system, was located in the hut. The hut was manned in shifts from a complement of twelve to eighteen officers and men from the Navy's cryptologic element, known as the Naval Security Group (NSG). However, contrary to some assertions, the Desoto missions were not the functional or operational equivalent of the ubiquitous Soviet electronic collection trawlers.9 The Desoto missions primarily served the mission needs of local commanders, although the received technical support in the way of technical working aids and intercept data from NSA.
The Desoto patrols had a two-part missions: to collect intelligence in support of the embarked commander and higher level authorities and to assert freedom of navigation in international waters. The early Desoto missions in the waters [***** ***** ***** redacted text ***** ***** *****] had been tracked by the coastal radar surveillance networks belonging to the naval forces [***** redacted text *****] While an occasional communist patrol ship would come out and shadow the U.S. patrol, little else happened.
However, when the Desoto patrol first was proposed for the waters in Southeast Asia, its
mission was expanded. First of all, the commander, Seventh Fleet, wanted a patrol to move in closer than the original twenty-mile limit -- as close as twelve miles. Additionally, the Desoto mission was expanded to include a broader collection of "all-source intelligence," namely, photographic, hydrographic, and meteorological information.10
In mid-January 1964, COMUSMACV requested that the Desoto patrol scheduled for February (USS Radford, DD-446) be designed to provide the forthcoming OPLAN-34A program with critical intelligence regarding North Vietnam's ability to resist its projected commando operations. However, in this case, the Radford's mission was canceled so as to not interfere with OPLAN-34A missions planned for the first two weeks of February.11
This is an important point, although a subtle one, for understanding the events of 2 to 4 August. Inasmuch as there was an interworking between the two programs, and this remained a point of contention in later congressional hearings, as well as a source for speculation by the press, the Desoto mission remained merely one of collection of intelligence which could be of use to the OPLAN-34A planners and commanders back in Danang and the Pentagon. There was no direct operational connection between the two programs. They were managed under separate officers and were not known to coordinate mission planning, except for warnings to the Desoto patrol to stay clear of 34A operational areas. At least that was the understanding back in Washington.12
In early July, General Westmoreland requested more intelligence on Hanoi's forces which were capable of defending against an expanded OPLAN-34A program. Specifically, Westmoreland required intelligence on the DRV's defenses in those areas targeted for July operation -- Hon Me, Hon Nieu, and Hon Matt Islands, as well as the area around the port of Vinh Son, south of the islands. In response, Admiral Sharp, CINCPAC, issued a new directive for a Desoto patrol whose purpose was "determining DRV coastal patrol activity."13

Gulf of Tonkin region of interest to OPLAN-34A and Desoto missions' alphabetic points denote Desoto mission start and stop positions.(Courtesy of Naval Historical Center)
That the two missions might run up physically against one another was a consideration at both MACV in Saigon and CINCPAC (and CINCPACFLT) in Honolulu. But Westmoreland assured the navy commanders that as long as the Desoto patrol stayed within its schedule and area of operations, there would be no problem. Westmoreland added that all the Studies and Observations Group (SOG), which ran the OPLAN-34A missions, needed in the way of an
alert, was thirty-six hours' notice of any change. They could then adjust any planned 34A operation. The navy accepted these reassurances from MACV.14
The first Desoto mission in the Tonkin Gulf region ran from February to March 1964. The USS Craig (DD-885) sailed near Hainan Island towards the Vietnamese coast and then turned back north towards Macao and Taiwan. [***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****] The North Vietnamese *** tracked the Craig as it swung south of Hainan Island, but had made no reaction even though they knew that it was a U.S. warship. It was uncertain to the Americans what the Vietnamese precisely know of the Craig or its mission.[ ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****]
During this mission, there was a Naval Security Group DSU aboard whose task was to provide tactical intelligence to the Craig's commander, as well as intercept unique communications and electronic intelligence in reaction to the vessel's presence. The Craig also received support from the SIGINT facilities in the region: the navy and air force COMINT sites in the Philippines [***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** *****] No Vietnam-based sites were involved since the area of Craig's mission barely touched on the DRV's territorial waters, and then only briefly, although it was suspected that the North Vietnamese navy at least once did report the Craig's position.17
However, there were two critical differences between the Craig's Desoto mission and that of the Maddox which followed it in late July and August: The Maddox would sail along the entire DRV coastline, while, at the same time, OPLAN-34A maritime missions against North Vietnamese coastal installations were being carried out. By July, the North Vietnamese were reacting aggressively to these raids, pursuing and attacking the seaborne commando units.
In mid-July 1964, the JCS approved another Desoto mission, which would concentrate on collecting intelligence on North Vietnam's coastal defense posture. The USS Maddox, under the command of Captain John Herrick, loaded up its intercept van in the [***** redacted text *****] The sixteen members of the DSU boarded, and the ship departed for the Gulf of Tonkin. The Maddox had received no additional instructions to its standard collection mission and apparently was not aware of specific OPLAN 34A missions in the area.18 However, the Maddox was not on a purely passive mission. U.S. intercept sites in the area were alerted to the real reason for the Desoto missions, which was to stimulate and record (my italics) North Vietnamese [***** redacted text *****] reactions in support of the U.S. SIGINT effort.19
CINCPAC's orders to Herrick were equally explicit and ambitious: locate and identify all coastal radar transmitters, note all navigational aids along the DRV's coastline, and monitor the Vietnamese junk fleet for a possible connection to DRV/Viet Cong maritime supply and infiltration routes.20 Whether these missions could be completed was questionable: the DSU was limited by its few positions and equipment in collecting such a large amount of communications. The Maddox had been ordered by CINCPAC to stay eight nautical miles from the North Vietnamese coastline, but only four miles from
any of its islands.21 It would be attacks on these islands, especially Hon Me, by South Vietnamese commandos, along with the proximity of the Maddox, that would set off the confrontation.


Captain John J. Herrick (left), the on-site task force commander,
and Commander Herbert Ogler, commanding officer of the Maddox
[***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****]
[***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****]
[***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****]
[***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****]
At the beginning of 1964, the Department of Defense, which had started its own program, assumed control of all of these covert missions. It merged its own project [***** ***** ***** redacted text ***** *****] and organized all of the new missions under OPLAN 34A-64, [***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****] OPLAN 34A originally was planned to last twelve months and was to be a program of selective intrusions and attacks of graduated intensity. The purpose of these actions was to "convince the Democratic Republic of Vietnam leadership that their continued direction and support of insurgent activities in the RVN and Laos should cease."24
The OPLAN reflected the current American strategy of escalation of the war through graduated response. The U.S. established four levels of actions; each proceeding one was a qualitative and quantitative increase in the sensitivity of target selection and the intensity of the application of force. It began with harassment attacks and operations, whose cumulative effect, though labeled "unspectacular," was to make Hanoi aware of them to the extent it would allocate forces to counter them.25 If this approach failed, then the next level -- tagged as attritional -- was to attack important military and civil installations whose loss could cause "temporary immobilization of important resources" which, in turn, might create or increase opposition amongst the North Vietnamese population to the government in Hanoi. The third level, termed punitive by the 34A planners, was meant to cause damage, displacement, or destruction of those facilities or installations considered critical to the DRV economy, industry, or security. To protect itself from further attacks would mean that the DRV would have to redeploy resources originally meant to support the war in the south to the needs of internal security. The planners admitted that the operations at this level would involve large enough forces that they would be necessarily overt. But the planners felt that these attacks could be attributable to the South Vietnamese.26
The final step of the plan was the initiation of an aerial bombing campaign designed to damage the DRV's capacity to support the southern insurrection or cripple its economy to such an extent that it would realize the extent of its losses was not worth the support of the war in the South. At this point, the planners in Washington believed that Hanoi's reaction to the attacks would be based on two factors: its willingness to accept critical damage to its own economy by continuing supporting the war in the South, and the possible support of the People's Republic of China. The plan did suggest that the communists would choose to continue to support the southern front, and it left open the possibility of further operations to offset the anticipated Chinese aid.27
The major operational components of OPLAN 34A were airborne operations that inserted intelligence and commando teams into North Vietnam, and maritime operations (MAROPS) which consisted of hit-and-run raids on coastal installations and facilities. These latter missions were known under the operational title Timberwork. The teams were made up of mostly South Vietnamese Special Forces, known as Luc Luong Dac Biet or Biet Kich, with some foreign mercenaries (mostly Chinese and Koreans) to crew the attack craft. The American involvement, though extensive in the planning, training, and logistics portions, was minimized to achieve the usual "nonattribution" status in case the raids were publicized by the North. No Americans were allowed to participate in the actual raids.
Despite all the planning, there was little confidence in the effectiveness of the OPLAN 34A operations. CIA chief John McCone suggested that they "will not seriously affect the DRV or
Norwegian-built 'Nasty' fast patrol boat, the primary platform for maritime operations under OPLAN 34A
cause them to change their policies."28 Defense Secretary McNamara, when he returned from an inspection trip to South Vietnam in March 1964, described OPLAN 34A as "a program so limited that it is unlikely to have any significant effect." The operations were described by other officials as "pinpricks" and "pretty small potatoes."29
The Johnson administration was dissatisfied with the initial results of OPLAN 34A and sought a stronger approach. By June 1964, a new OPLAN, designated 37-64, had been developed jointly by the National Security Council, the JCS, and MACV. This new OPLAN called for a three-pronged approach to "eliminate to negligible proportions DRV support of VC insurgency in the Republic of Vietnam." Three military options were put forward: ground action in Cambodia and Laos to eliminate VC sanctuaries and supply points, increased levels of 34A attacks on Hanoi's coastal installations, and South Vietnamese and United States bombing of ninety-eight "preselected" target in North Vietnam.30
If the commando raids had been such failures, why did they continue to be staged? The truth is, Washington was anxious to support the shaky regime of General Khanh, who had succeeded to the presidency of South Vietnam after Diem's assassination. Until a better plan, such as 37-64 could be implemented, then doing "something," even as ineffective as the raids, was the course Washington chose to follow. In spite of Hanoi's gains for the first six months of 1964, if America's determination to succeed could be communicated to Khanh, then the South Vietnamese might be reassured of the prospects for victory.31 This was Washington's policy: to prop up Saigon. Yet, this was a structure built on unsupported assertions.
The reality for Washington was that the increased tempo of maritime commando raids had only raised Hanoi's determination to meet them head on. Through June and July 1964, NSA and the navy monitoring site in the Philippines reported that the conflict along the coast of North Vietnam was heating up. Communications about small boat actions, commando landings, and high-speed chases out at sea were intercepted and reported back to Washington. What the reports showed was a North Vietnamese navy emboldened to more aggressive reactions to incursions by the commandos from the south. For example, on 28 July, after an attack on the island of Hon Gio, DRV Swatow-class patrol boats pursued the enemy for forty-five nautical miles before giving up the chase.32 Earlier, on 30 June, another patrol boat had taken potshots at two jet aircraft flying along the coast and claimed a hit.33
By early June, Hanoi's stepped-up defensive posture had registered in its radio traffic. On 8 June, NSA reported that the level of North Vietnamese tactical radio communications had increased almost fourfold during the early part of June from the previous period in May, probably in reaction to attacks along its coast. It also reported that DRV naval patrols now seemed to cover its entire coastline.34 Clearly, Hanoi was determined to defend itself resolutely. Whether or not the Vietnamese believed that the Americans were preparing for a larger war was not important. What was critical was that the situation along North Vietnam's territorial waters had reached a near boil.35
The SIGINT support to OPLAN 34A started at almost the same time as the operations began. Codenamed Kit Kat, the effort required that the then current ceiling of 660 cryptologic personnel in South Vietnam had to be raised. In February 1964, an increase of 130 personnel for Kit Kat was approved by CINCPAC.36 The ASA moved personnel from the Philippines to Phu Bai, and the Naval Security Group added coverage of North Vietnamese naval communications to its mission at San Miguel in the Philippines. The Air Force Security Service units at Monkey Mountain near Danang increased their coverage of the communications of DRV navy and coastal surveillance posts. A small special SIGINT unit at Tan Son Nhut Airbase, known as the Special Support Group (SSG), was formed in late February to coordinate Kit Kat support between the intercept sites and the Studies and Observations Group.
A few last notes before we review the attacks. It will be necessary to limit the discussion to the role SIGINT played during the incident. Other evidential sources, such as that from the American ships' own radar, sonar, and visual sightings, will be mentioned in passing simply because they are part of the story and cannot be altogether ignored. However, the brunt of the following discussion will center on the SIGINT evidence because of its critical role in convincing the Johnson administration that the attack actually occurred.
Besides the NSG detachment aboard the Maddox (USN-467N), other SIGINT elements that were involved in the events of the next three days included a Marine SIGINT detachment (USN-414T), collocated with the Army Security Agency intercept site at Phu Bai (USM-626J), and the NSG site at San Miguel, Philippines (USN-27), which also had a Marine SIGINT contingent, but the latter was not designated separately as was the Marine group at Phu Bai. It would be the intercept and reporting by the Marine unit at Phu Bai and the navy site in the Philippines which would prove critical to the events in the Gulf of Tonkin.
[***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****]
[***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****
***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****]
A large number of the reports by the various field sites and NSA were issued contemporaneously with the events themselves. A few of these would be cited in the various after-action analyses and postmortems that attended the Gulf of Tonkin. However, many more field translations and reports based on the intercept during the period of the incidents would be issued as late as two to four days after the crisis. The reason for the apparent delay was that the request from NSA for ALL intercepts came only on 7 August.35
Because of the nature and enormous amount of the SIGINT evidence used here for the very first time in discussing the Gulf of Tonkin crisis, we will need to present it in a format which will highlight that information. Rather than try to retell the story all at once and incorporate the new evidence into the narrative, which could be overwhelming, especially to those readers not intimately familiar with the events of 4 August, a different track will be used. We will break down the events into their separate days. First, we will review the details of the known engagement of the afternoon of 2 August. While there is no controversy surrounding this fight -- at least there is no question that it occurred -- there is an important point to draw from it: that is, the North Vietnamese communications profile during a naval combat engagement was revealed. For ease of reference, we shall refer to this communications profile as the "command and control communications and intelligence" system or C3I. This is a functional description used widely in the intelligence and defense communities to describe the process whereby the individual elements of intelligence (information/intelligence), command and control (interaction by command authorities), and communications (communications links among all operating elements and units) are combined in military operations.
After looking at the "uneventful" day of 3 August, we will consider the "official" version of the engagement of 4 August. Although, as we progress through the narrative, we will consider the problems with the various other pieces of evidence which support the contention that an attack occurred, the emphasis will be on the SIGINT "clinchers," that is, those reports that convinced the Johnson administration that an attack had occurred. These items will be presented when and how they appeared to the participants.
Finally, we will go back over the clinching SIGINT "evidence" of 4 August and illustrate what problems exist with the individual pieces. In this section, the entire scenario of what was reported and, more importantly what was not reported, will be considered. We will review closely the technical problems with the two critical SIGINT reports which prop up all of the other evidence of an attack by the North Vietnamese. In this approach we will consider how the product was developed and the serious problems in translation, composition, and reporting of the information.
One last item. For purposes of clarity, all time references will be marked either Zulu (time ("Z", or Greenwich Mean time) or Golf ("G," or Zulu +7), which is the time zone for the Gulf of Tonkin. While the actual time of the incidents was in local, or Golf time, SIGINT reports were issued in Zulu time. This is done because of the worldwide nature of SIGINT reporting. The use of Zulu time allows for a consistent and universal benchmark for analysts and recipients of the intelligence. To further confuse the issue, the U.S. Navy used Hotel time (Zulu +8) in all of its messages, which is carried over into its history of the Vietnam War. Then there are the events in Washington, D.C., and NSA HQ, Fort Meade, MD, which are in the Eastern time zone, or
Romeo ("R", or Zulu-5 hours). The latter times will be notated "EST" for Eastern Standard Time. All times will be given in the military twenty-four-hour clock. So, all "P.M." times after 1200 hours can be determined by subtracting 1200 from the time: e.g., 1700 hours equals 5:00 P.M. Also, it must be remembered that events in the Gulf of Tonkin occurred west of the international date line, so that certain events in the region were occurring the next day in terms of Washington's time. For example, if something happened at 1500 hours Zulu, it is reflected as 200 hours Golf, 2300 hours Hotel and 1000 hours Romeo of the same day. However, t two-hour advance in Zulu time, that is, 1700 hours on 4 August, means 0000 hours Golf and 0100 hours Hotel time on 5 August, while Washington will be 1200 hours on 4 August. For ease of reference, the reader can observe that there is a twelve-hour difference between Washington and the Gulf of Tonkin.
On the morning of 1 August, the ASA site at Phu, Republic of Vietnam, monitored a DRV patrol boat, T-146, a Swatow- class patrol craft communicating tracking data on the Maddox to another Swatow. At the time, between 0700G to 0730G (0030Z), the Maddox was located nine miles southeast of Hon Me Island moving northeasterly. The Swatow- class patrol craft was one of a group supplied by the People's Republic of China. It was a fairly large patrol craft displacing sixty-seven tons. It had a top speed of forty-four knots and a cruising speed of twenty knots. It was armed with two 37-millimeter (mm) antiaircraft (AA) gun mounts, two 20=-mm AA mounts, and carried up to eight depth charges. This armament limited the Swatow's role to countering other small vessels. The Swatow carried the Skin Head surface search radar. The Swatows often worked in tandem with P-r torpedo boats, acting as communications relays between North Vietnamese naval command centers and the P-4s, whose long-distance communications capability was limited. This was a role that the Swatows filled all during the next few days' action.40

The T-146 patrol craft also ordered the other craft to turn on its "equipment," which probably referred to its Skin Head radar. However, the Maddox did not intercept any emissions from the Swatow's radar. The North Vietnamese boats referred to the track as the "enemy"; the equation of the term to the Maddox was made by Phu.41
Shortly after 2300G (1600Z), on 1 August, the naval intercept site in the Philippines reported that the DRV naval base at Ben Thuy (18°39'N, 105°42'E) had informed an unidentified entity, possibly the T-146 patrol boat, that it had been "DECIDED TO FIGHT THE ENEMY TONIGHT [1 Group unreadable] WHEN YOU RECEIVE DIRECTING ORDERS." The base also queried the boat if it had received the "enemy's" position change from another naval entity, possibly an authority on Hon Matt Island (18°48'N, 105°56'E).42 The Maddox was informed of this intercept. A half hour after receiving the most recent report, Captain Herrick informed Seventh Fleet and CINCPAC that he had terminated the mission because of indications of an imminent attack and was now heading east out of the patrol area at ten knots. These indications of an attack were from Vietnamese communications intercepted by the two field sites, as well as the NSG detachment aboard the Maddox. Throughout the rest of the day, these stations would monitor the North Vietnamese ships-to-ship and ship-to-shore manual Morse and voice communications nets. They intercepted the all-important vectoring information, the orders from shore commands, and all the tactical communications. However, the DRV boats made no hostile moves against the Maddox that day.
Throughout the night of 1/2 August, according to the intercepted communist messages, the North Vietnamese continued to track the destroyer as it remained east of Hon Me Island, some twenty-five miles offshore. Still, nothing had happened that night, and so the Maddox returned to its patrol line off the DRV coast on 2 August.
During the early morning, the Maddox, which was heading along the northern track of its patrol area, was notified of further North Vietnamese tracking of its movements. The North Vietnamese naval motor torpedo boat squadron stationed at Port Wallut command was receiving the tracking. A coastal surveillance radar station on Hon Me may have been ordered to begin tracking the destroyer "continuously." (It is possible that this station had been inactive during the previous day so as to deny any information on its operation parameters from the American monitoring effort.)43
More ominously for the Maddox, the communists also had ordered P-4 patrol torpedo boats (MTB) and Swatow-class patrol boats to begin concentrating near Hon Me Island later in the morning.44 These patrol torpedo boats had been supplied by the Soviet Union. The P-4 boat displaced twenty-five tons. Its top speed was fifty knots; its cruising speed was thirty knots. It had two twin 12.7-mm machine-gun mounts and two eighteen-inch torpedo tubes. The P-4 boat also carried a Skin Head surface search radar. The reporting from the American intercept sites construed the Vietnamese boat concentration near Hon Me as a prelude to an attack on the Maddox.45
NSA feared that an attack on the Maddox was in the offing. At 1002G (0302Z), on 2 August, NSA sent an urgent message to a number of commands and sites in the region warning of a possible attack. Included in this message was CINCPACFLT, MACV, and the Commander, 7th Fleet. Ironically, the Maddox was not on distribution for this message; the DSU would have received the message, but it was not addressed either. The gist of the message was simple: repeated attacks by "enemy vessels" on Hon Me Island had led Hanoi to make preparations to repel any further assaults. NSA added that

"...THE INDICATED SENSITIVITY ON PART OF DRV AS WELL AS THEIR INDICATED PREPARATION TO COUNTER, POS[SI]BLE THE DRV REACTION TO PATROL MIGHT BE MORE SEVERE THAN WOULD BE OTHERWISE BE ANTICIPATED." The problem with the Maddox not receiving these critical warnings would not be resolved until after the first attack.46
Shortly before noon, at 1144G (0444Z), the Marine SIGINT group attached to the ASA site at Phu Bai, RVN, intercepted a message from the T-142 Swatow-class patrol boat to the DRV naval base at Port Wallut which stated that "[WE] HAVE RECEIVED THE ORDERS. [T]146 AND [T]142 DID USE [1 Group unreadable] HIGH SPEED TO GET TOGETHER [PARALLEL] WITH ENEMY FOLLOWING LAUNCHED TORPEDOES."47 The Phu station issued a Critic, short for a critical message, that alerted all relevant commands, and the Maddox, of the planned attack. In the same report, the Phu site added that four boats, T-142, T-146, T-166, and T-135, had been engaged in tracking and following an "enemy" which "is probably the current mission." The final paragraph of the message added that the DRV naval facility in Port Wallut was acting as the shore-based "coordinator/director" for the surveillance of the probable vessel.48
About a half hour later, at 1218G (0518Z), another Marine SIGINT detachment stationed with the navy monitoring station at San Miguel, Philippines, intercepted the same message. This later intercept is not unusual; it meant that the Vietnamese were retransmitting the message to ensure its reception. However, this intercept was reported in a different manner than Phu Bai's version. The second version was reported as a translation instead of a report. In essence, this meant that the actual intercept was reported, and not a restatement of its contents. Therefore, some interesting items of intelligence, which were missing from the first report, were included.
First of all, the second version contained what is known as the "file time" of the DRV message, that is, the time when the message was entered into a log prior to its transmission by the Vietnamese radio operator on the T-142. In this case, a file time of 1113G was noted. This time reference tells us that there was a half-hour delay between the receipt of the message from the originator and the initial transmittal of the "attack" message (1144G/0444Z), as well as an hour's difference in the second intercept (1218G/0518Z). The differences are interesting for two reasons. First of all, if the intercept times from both American sites reflect the beginning of the actual intercept of the Vietnamese transmissions, then the half-hour difference suggests that the "attack" message was sent more than once. Why more
than once? It is possible that Port Wallut had not received the first transmission from T-142, although the reports from both Marine sites imply that the message was received each time. Secondly, the lag between the file time and the actual transmission time by the Vietnamese, if figured from the American time of intercept, suggests that the Vietnamese were having difficulties in transmitting messages in a timely manner. This delay, as we shall see, becomes an important element in determining the DRV intentions.
At about this time, the three torpedo boats had arrived at Hon Me Island. The Maddox, which was steaming on a northeast heading away from the island, had observed visually the arrival of the three boats. Shortly afterwards, the two Swatows were seen by the Maddox in the area of Hon Me. The five North Vietnamese boats now were concentrated at the island.
The "attack" message was followed up by another message, this time from Port Wallut to T-146, which was intercepted at 1306G (0612Z) by the Marines in the Philippines. The message instructed T-146 (and probably T-142) to "LEAVE 135 AND TURN BACK TO [THE PATH] OF THE ENEMY." The " 35" that T-146 was told to leave turns out not to have been an individual boat, as earlier reported by the Marines, but the squadron designator for the three P-4 torpedo boats which would take par tin the upcoming attack. These three boats made up the Section 3 of Squadron 135.
These five boats, which included the P-4 boats, T-333, T-336, and T-339, departed Hon Me Island at about 1300G, quite possibly on their way to seek out the Maddox.49 Within the next hour a set of apparently conflicting orders was sent to the Vietnamese boats. At 1409G (0709Z), Port Wallut notified both Swatow craft that the "enemy" was a large ship bearing 125 degrees (from My Duc?) at a distance of nineteen miles at a speed of eleven knots on a heading of twenty-seven degrees. This put the target on a north-by-northeast heading, which matched that of the Maddox. The same message also included a garbled phrase to "THEN DETERMINE," but it is unclear what this phrase meant.50 However, according to Edwin Moise, the North Vietnamese said that Section 3 received its order to attack the destroyer at 1350G.51 Since the file time of the message from Port Wallut was 1400G, this may have been the "attack" message.
However, there is a complicating factor. At 1403G (7030Z), just six minutes earlier, the site at San Miguel had copied a message from Haiphong to the two Swatow patrol boats which told them to "ORDER 135 NOT TO MAKE WAR BY DAY." Furthermore, the message added that all of the boats were ordered to head towards shore (though an intercept of the same transmission by the Marines at Phu ordered the boats first to pretend to head towards shore), and then return to Hon Me Island.52 Although this message was sent shortly after 1400G (0700Z), it contains a file time of 1203G (0503Z). This means that this message, which constitutes an order to recall the boats, was originated some two hours before the order to attack was transmitted! A second intercept of the same message added that T-146 was supposed to order the recall of Squadron 135's torpedo boats.53 According to Moise, the North Vietnamese claimed that a recall order was sent after the attack message, but T-146 never relayed it in time.54
This conflict in orders by command elements from Port Wallut and Haiphong indicates that there was a loss of control of the situation. It further suggests that the DRV naval authority in Haiphong had never wanted the attack to occur, at least not during the day, when conditions were not favorable for surprising the Maddox. Since the boats continued their attack on the destroyer, it appears that the recall order was ignored. The deciding factor for the Vietnamese boat commander may have been the much earlier file time of the recall order; the attack message with the more current file time
probably superseded everything else in his decision.
At around 1400G, the Maddox's radar detected the approach from the southwest of the three P-4 torpedo boats. Forewarned by the SIGINT of the Vietnamese intentions to attack, the Maddox then started turning eastward, then to the southeast and increased its speed from eleven to twenty-five knots. The North Vietnamese boats initially may have missed the turn to the southeast by the Maddox. They probably had been visually tracking the American vessel.55 There is no SIGINT evidence that their Skin Head radars were active, though the Vietnamese claimed the boats used it. Pictures from the action appear to show the radar masts still upright and not lowered in a combat position. By the time the Vietnamese did react to the Maddox's change in course, they found themselves in an unfavorable attack position. They were chasing the Maddox from its rear starboard, that is, from the northwest, which meant it would take some time, even with a near twenty-knot advantage in speed, to achieve an optimal firing position for their torpedo run -- perhaps as long as thirty minutes before they could execute a turn on an attack heading. By 1430G, Commander Ogier ordered the Maddox to go to general quarters.
At about 1440G (0740Z) the Maddox sent a flash precedence message to various commands in the Pacific that she was being approached by high-speed craft with the intention of attacking with torpedoes. Herrick announced that he would fire if necessary in self-defense.56 He also requested air cover from the carrier Ticonderoga, which was then 280 miles to the southeast. Four F-8E Crusaders from the carrier, already aloft, were vectored to the Maddox. The destroyer Turner Joy (DD-951) was ordered to make best speed to the Maddox.
For the next twenty minutes, the chase continued. The Vietnamese boats inexorably closed the gap between the themselves and the destroyer. At 1500G, Captain Herrick ordered Ogier's gun crews to open fire if the boats approached within ten thousand yards. At about 1505G, the Maddox fired three rounds to warn off the communist boats. This initial action was never reported by the Johnson administration, which insisted that the Vietnamese boats fired first.

A few minutes later the Maddox resumed fire. Through the shellfire, the DRV boats bore in on the Maddox. But their attacks were ineffective. Within fifteen minutes of Maddox's first salvo, jets from the carrier Ticonderoga had arrived and attacked the Vietnamese boats, leaving one dead in the water and the other two damaged. As for the Maddox, she was unscathed except for a single bullet hole from a Vietnamese machine gun round.

2 August naval action. Note the use of Hotel time (Z+8/G+1). (Courtesy of Naval Historical Center)
There would continue to be confusion over losses for some time. The DRV claimed that two aircraft had been shot down. In reality, one of the navy's jets had sustained wing damage during its maneuvering for the attack and was escorted out of the area by another jet. Both aircraft departed the area under full power, the black exhausts trailing from their engines probably appeared as battle damage to the Vietnamese sailors.57 The damaged navy jet would be forced to land at Danang.

As for the attacking communist vessels, eventually all three struggled back to their bases. The one craft, T-339, thought to be dead in the water and claimed to have been sunk by the Americans, and, incidentally, initially reported sunk by the Vietnamese as well, actually restarted its engines and managed to limp back to shore. On board were four dead and six wounded Vietnamese sailors out of a crew complement of twelve. However, the other Vietnamese boats were unaware of what had happened and reported T-339 as sunk, and would continue to do so for days afterwards.58
At 1630G (2330Z), the Vietnamese patrol boat, T-142, received orders to concentrate back at a location north of Hon Me Island, and to make contact with another possible Swatow-class patrol boat, T-165. T-146 also received orders from Haiphong to send two boats out and help the P-4s of Squadron 135 to return.59 Two days later, on the afternoon of 4 August, T-146 would report to Haiphong the damage to the boats during the attack. T-333 had been hit three times and suffered scattered damage to its water pipes and lifeboat. Its auxiliary engine had been hit and oil pressure was low, suggesting a leak. Still, the boat was assessed as being "lightly damaged." On the other hand, T-336 was described as being "heavily damaged with many holes." Its fuel oil was contaminated, possibly by sea water, and the barrel of one of its deck guns was ruined.60 The boat's crew had suffered at least two wounded as well. The status of both boats and T-333's crew is important to remember when we look at the events of the later evening of 4 August.
In Washington, the reaction to the attack was relatively subdued. Since no Americans had been hurt, President Johnson wanted the event downplayed while a stern note of protest was sent to the North Vietnamese. (Ironically, this message was the first diplomatic note ever sent to North Vietnam by the United States.) The president had said that we would not "run away"; yet we were not going to "be provocative." However, Hanoi was to be informed in no unambiguous terms that any more unprovoked actions would entail "grave consequences."61
The lack of any reprisal was surprising, especially since freedom of navigation was one of the official reasons for the Desoto missions. However, it is likely that there were mitigating factors which caused Washington to pause. Secretary of Defense McNamara was incorrect to claim that the Vietnamese had fired first.62 At the same time, the Johnson administration had seriously miscalculated the reaction by Hanoi to the OPLAN 34A missions. It had never considered that the communists might correlate the attacks with the presence of the American destroyer.63 NSA, monitoring the increasing aggressiveness in DRV naval communications, had seen the possibility and had warned everyone, except the Maddox.
Furthermore, Washington, through the intercept of the DRV's naval communications, had seen the confused set of orders sent to the boats, which suggested that Hanoi had lost control of the situation. McNamara would state, "We believed it possible that it had resulted from a miscalculation or an impulsive act of a local commander.64 It seemed that everyone was trying to defuse the crisis.
DIRNSA, concerned about possible aggress reactions [***** redacted text *****] ordered all the sites in the region to maintain "extreme collection, processing, and reporting vigilance on part of all with reporting accomplished IAW [in accordance with] established procedures and at precedence appropriate to activity, especially in regards to [*** redacted text ***] reaction.65 A SIGINT Readiness Level Bravo Lantern was declared. Under this readiness level, eight field sites were tasked by NSA to monitor for any North Vietnamese [*** redacted text ***] reaction to the patrols. The brunt of the intercept and reporting was handed to the navy at San Miguel and the Army and Marine missions at Phu Bai.66
The Pentagon was not going to wait around for another incident to happen, either. Plans were put into motion to augment U.S. forces in the region, including deployment of United States Air Force combat aircraft to the Philippines and the dispatch of the carrier Constellation to join the Ticonderoga. A second destroyer, the Turner Joy, already had been dispatched to rendezvous with the Maddox. CINCPAC ordered both ships back to the patrol area, seeing it "in our interest that we assert right of freedom of the seas." CINCPACFLT issued new rules of engagement for the next three days which
allowed both ships to approach the North Vietnamese coast as close as eight nautical miles and four miles from its islands. The two destroyers were ordered to arrive at their daylight patrol point about one hour before dawn. One hour before sunset they were ordered to retire east out to sea during the night.67

If the Pentagon brass was anxious to insert its ships into harm's way, Captain Herrick was more cautious. In an after-action report transmitted that evening, which reviewed the attack and the successful American defense, he added a warning: the "DRV HAS C[AS]T DOWN THE GAUNTLET AND NO[W] CONSIDERS ITSELF AT WAR WITH WITH US." He added a concern that the DRV's torpedo boats, especially at night, could hid and then approach the destroyers with little warning.68 He stated that the Maddox and the Turner Joy, with their five-inch guns and top speed of thirty-three knots, were inadequately armed for defense against such boats. He suggested that the Desoto patrols would be safe only with a cruiser and continuous air cover. One last item was reported by Captain Herrick: the Maddox's long-range, air search radar (AN/SPS-40) was inoperative, and the fire control radar (AN/SPG-53) belonging to the USS Turner Joy, which had just arrived to reinforce him was out of action indefinitely.69
At the close of 2 August, the North Vietnamese boats were hiding in coastal waters caring for their casualties and waiting for orders as to what to do next. The Maddox was joined by the Turner Joy out at sea, and both were being replenished with ammunition and supplies while under way. The had been ordered to return to the coast at daybreak.
The SIGINT community could be proud of its efforts during the day. The field sites and NSA had intercepted, processed, and reported North Vietnamese naval communications in such a rapid and clear way that everyone in the Pacific command was aware of the approaching attack. It also had provided the information to Washington that suggested that Hanoi's grip on events was less certain than was expected. At the same time, by monitoring the DRV's naval communications, the cryptologists had developed a picture of the command and control elements prior to an attack: extensive tracking by coastal observation posts; the identification of a target and the communication of an attack command; and the use, if limited, of radars in locating the target. The Maddox had never been explicitly named as the target of the attack; in fact, there was just the notation of an "enemy"; however, the analysts at Phu Bai, San Miguel, and inside the Desoto hut had correlated the North Vietnamese tracking with the American ship. The Maddox had been fixed in the minds of the American cryptologists as an "enemy vessel" to the North Vietnamese; they would be on the lookout for possible new attacks. The question was, though,
was Hanoi spoiling for another round with the U.S. Navy?
At the same time, the State Department publicized the note it had sent Hanoi protesting the attacks. It concluded with the words "The United States Government expects that the authorities of the regime in North Vietnam will be under no misapprehension as to the grave consequences which would inevitably result from any further unprovoked military action against the United States forces."70
Despite the increased North Vietnamese vigilance and the observed sensitivity to American and South Vietnamese naval activity in Hanoi's territorial waters, COMUSMACV went ahead with an OPLAN 34A mission scheduled for the night of 304 August. In accordance with an earlier agreement, the Maddox and Turner Joy were advised to avoid sailing in the area bounded b y the 17th and 18th parallels. A 34A mission against the radar site at Vinh Son (17°57'N, 106°,30'E), which involved a four-boat task group, set sail at 1510G (0810Z) on 3 August. At midnight it shelled the radar station. One of the boats broke off and attacked a nearby security post and was pursued for a short distance by a North Vietnamese patrol craft.
By mid-morning of 3 August the two destroyers were heading to their patrol station, which was about 100 miles northwest of the new 34A mission area. They expected to be on station by early afternoon. However, this location kept them in the area of the island of Hon Me, which was the focus of DRV naval activity during the ensuing day and night.
Meanwhile, the North Vietnamese were concerned with the salvage of their damaged boats. Just past midnight on 3 August, T-142 and T-146 were in the area of Hon Me Island trying to contact another Swatow, T-165, as well as find the missing boats from Squadron 135. At 0300G (2000Z), T-142 sent an after-action report to the T-146 (for relay to Port Wallut), which highlighted the previous afternoon's combat. It included a chronology of the various actions the squadron's boats carried out from 0935G to 1625G when they attacked the Maddox.71
Even by mid-afternoon of 3 August, naval headquarters in Haiphong still did not know where the torpedo boats were and demanded that the Swatows inform it when they knew their situation.72 However, the SIGINT site at Phu Bai misconstrued this search and salvage activity as a prelude to a potentially dangerous concentration of enemy boats. It issued a Critic at 1656G (0956Z), which placed six DRV patrol and torpedo boats near Hon Me Island.73 However, the report was wrong in that it identified the squadron reference "135" as a boat, as well as locating the two torpedo boats, which, at the time, were still missing. The ominous concentration of boats simply was not occurring. However, this incident revealed how tense the situation had become. It also illustrated a precedent by the field site at Phu Bai for misinterpreting Hanoi's intentions.
Almost as soon as the two destroyers arrived on station south of Hon Me Island in early afternoon, they were shadowed by a DRV patrol boat which tracked them using its Skin Head radar.74 The tracking continued through the afternoon into early evening. The Haiphong naval authority and the Swatow boats near Hon Me
exchanged position information on the two destroyers as they moved from the north to south and back north on their patrol.75 At one point, another Swatow, T-379, erroneously identified as an SO-1 class subchaser, was ordered to go out and observe "different targets," which probably referred to the American ships.76 The North Vietnamese also detected aircraft in the area of the Desoto patrol, though it is unclear from their report whose aircraft these were. However, the commander, 7th Fleet, had ordered a continuous combat air patrol accompanying the two destroyers. The navy jets flew their cover to the east of the Desoto position so as to avoid infringing on DRV air space.77
By early evening, Haiphong ordered T-142 to track the Desoto patrol. T-379, which earlier had been instructed to observe the Desoto patrol., had sailed to Hon Ngu Island (18°48'N, 105°47'E). It had arrived at 2250G (1550Z) and reported that the situation at sea was "peaceful."78
T-142 took up a position to the north of the two destroyers and stayed with them, reporting the location of the American ships to Haiphong either directly to naval HQs or relaying reports through T-146. Both U.S. ships reported being followed from the north at a distance of thirty-eight miles by a DRV patrol craft using its Skin Head radar. By this time, 2252G (1552Z), the Desoto patrol was heading southeast out of the patrol area as had been instructed earlier.79 Tracking of the destroyers ended soon after when they were out of range.
Meanwhile, the main concern of the DRV navy was the recovery operation for the boats damaged during the 2 August attack. Late in the night of 3 August, Haiphong informed T-142 that the salvage tug Bach Dang would soon leave Haiphong (it was not clear from the intercept if the time of departure was 0100G, 4 August/1800Z, 3 August) and head towards Hon Me Island to tow T-333 and T-336 back to Haiphong or Port Wallut, which was their unit's base.80 It was expected that the tug would arrive at about noon on 4 August. Meanwhile, T-146 was ordered to stay with the two damaged boats from Squadron 135 and report their position and status.
So ended 3 August. That evening's 34A raid on Vinh Son was protested by Hanoi. In its complaint, it accused the two destroyers of participating in the raid. Although the DRV's own tracking of the two ships had ceased some hours before, and they could not be certain of where the American ships were, the Vietnamese had inferred anyway that the Desoto ships were involved. It may not have been the right conclusion, but the Vietnamese believed it. Washington still did not think that Hanoi would act.
The North Vietnamese had been tracking the Americans. Haiphong informed T-142 at 1610G (0910Z) that they had located the destroyers near 19°36'N and 106°19'E traveling on a southwest heading. However, this last position of the two ships had been acquired by the North Vietnamese some two and one-half hours earlier at 1345G (0645Z).81 At approximately 1600G (0900Z), following his operational directive from
CINCPAC to be clear of the patrol area by dark, Herrick ordered the patrol to head due east.
At 1115Z (1815G), the naval SIGINT detachment aboard the Maddox received a Critic from the Marine SIGINT unit collocated with the ASA at Phu Bai, which stated, "POSS DRV NAVAL OPERATIONS PLANNED AGAINST THE DESOTO PATROL TONITE 04 AUG. AMPLIFYING DATA FOL."82 Twenty-five minutes later, Phu Bai issued a follow-up report at 1140Z (1840G) which reported, "IMMINENT PLANS OF DRV NAVAL ACTION POSSIBLY AGAINST DESOTO MISSION."83 The report went on to add that three DRV boats, T-142, T-146, and T-333 had been ordered at 0927Z (1627G), the time the message was intercepted by Phu Bai, to "make ready for military operations the night of 4 August." Although the report did not specify the nature of the military operations, the Marines appear to have concluded that it was an attack against the Desoto. The NSG detachment informed Herrick. Within an hour, at 1240Z, he informed CINCPAC and other commands that he had received " INFO INDICATING ATTACK BY PGN P-4 IMMINENT. MY POSITION 19-10N 107-00E. PROCEEDING SOUTHEAST."84 At this point, the two ships were about eighty to eighty-five nautical miles from the nearest DRV coastline and began to head southeast at twenty knots.
A short time later, just after 1300Z (2000G), the Desoto vessels acquired their first radar contacts. The Maddox reported that it had detected "two skunks" (surface contacts) and three "bogies" (air contacts) on its radars. The surface contacts were about forty to forty-five miles to the northeast of the two destroyers, putting them about 100-110 miles away from the Vietnamese coast at sea, but very close to Hainan Island.85 (The appearance of aircraft returns (bogies) on the destroyer's radar has generally gone unremarked upon by various commentators. Herrick speculated that these were terrain returns. Whatever the case, these false "bogies" suggest Maddox's air surveillance radar was still malfunctioning.) The Ticonderoga ordered the four jets on CAP to cover the two ships. It scrambled four more A1H Skyraiders. Within an hour, the aircraft were overhead.
At about 2045G (1345Z), Herrick reported he had lost the original surface contacts: they had never closed to less than twenty-seven miles from his own ships. At 2108G (1408Z), Maddox detected another return -- first identified as one boat, later thought to be several boats in a tight formation -- this time only fifteen miles away to the southwest, moving towards the destroyers at thirty knots. Nine minutes later, naval A-4 Skyhawks flying air cover were vectored towards the supposed boats. Although the pilots could see the wakes of the destroyers clearly, they could see no boats at the point the radar indicated. At 2131G (1431Z), this radar return disappeared.86
Then at 2134G (1434Z) came the most important radar contact of the entire incident. What appeared to be a single boat suddenly appeared on the Maddox's radar screen east of the two destroyers at 9,800 yards and closing at nearly 40 knots. The Turner Joy detected another object approaching, but on a different heading, distance, and speed. According to Marolda and Fitzgerald, the navy claimed that this was the same return as the Maddox's.87 At 2137G (1437Z) at a distance of 6,200 yards from the Desoto vessels, the return tracked by the Maddox appeared to make a sharp turn to the south. This maneuver was interpreted by the Maddox combat information center as a turn after a torpedo run. If this was a torpedo launch, then it was an extraordinarily desperate one. Hanoi's tactical specifications for its P-4s called for torpedo launches at ranges under 1,000 yards. At over 6,00 yards, it was unlikely a torpedo launched at a moving target could hit anything.88 The sonar operator aboard the Maddox detected a noise spike on his equipment, but did not report it as a torpedo. This conclusion was reached on the CIC. However, the Turner Joy never detected any torpedoes on its
sonar. Nor did it detect any torpedoes at all on its sonar that night.89
At 2140G (1440Z), Herrick informed CINPACFLT that he had commenced firing on the attacking PT boat. The Turner Joy had begun firing at its return shortly before this. Both destroyers had a difficult time holding a radar lock on their targets. Within five minutes, the return on Maddox's radar, which was moving away from the destroyers, disappeared from its screen at a distance of about 9,000 yards. The one that the Turner Joy was tracking kept approaching, and at a distance of about 4,000 yards, it disappeared as well.90
For the next fifteen minutes all surface contacts were gone from the radars of the two destroyers. Then, at 2201G (1501Z), more contacts were detected coming from the west. Now the thickest part of the naval action commenced. The two destroyers gyrated wildly in the dark waters of the Gulf of Tonkin, the Turner Joy firing over 300 rounds madly at swarms of attacking North Vietnamese boats -- maybe as many as thirteen -- and dodging over two dozen torpedoes. Another twenty-four star shells had been fired to illuminate the area and four or five depth charges had been dropped to ward off the pursuing boats and the torpedoes. The Maddox vectored overhead aircraft to the surface contacts, but time and again the aircraft reached the designated point, dropped flares, and reported they could not find any boats. By the time the attack was considered over at 2335G (1635Z), Herrick reported two enemy patrol boats sunk and another damaged. (The count of the damaged boats varied; Herrick believed that the DRV boats sank one of their own accidentally. It is not understood how he arrived at this conclusion, except as a misinterpretation of the radar data which itself was of dubious quality.)

It should be mentioned again that the radar returns from both ships were not continuous trackings. Rather, they were mostly flashing returns, that is, they appeared on the scope, held for a few sweeps of the radar, then disappeared. Other targets would suddenly appear a few miles from the destroyers, hold for a while and then disappear. They came from all directions. As each return was logged, it was assigned a target designator, a single letter. One officer from the Turner Joy described the confusion of proliferating targets this way: "We were getting blotches on the radar screen -- nothing real firm, so we were whacking away at general areas with proximity fuzes, hoping to get something."91 A target would apparently be hit and then disappear as if it had completely and instantaneously incinerated in an explosion -- contrary to what had happened two days earlier when the North Vietnamese PT boats would take several hits but remain afloat afterwards. The Maddox's main gun director
maintained that the ship was never able to acquire any of the targets during the battle; he figured he was shooting at the high swells brought on by the storms.92 Ironically, during all of this latter action, the Maddox never fired a round; its radar never acquired another target after the initial one detected two hours earlier.93
The sonar returns of the supposed torpedo attacks were later determined to be a result of the high-speed maneuvering by both U.S. ships. As we saw above, the first "evidence" of a torpedo launch by the enemy boats came from radar. When one of the radar tracks turned away to the south from a westerly heading, this was interpreted by the Americans as a torpedo launch. The sonar rooms in both destroyers were then alerted to a possible torpedo attack. Four crewmen aboard the Turner Joy thought they saw a "white streak" in the water as the ship turned.94 Both vessels had then gone into wild evasive maneuvers to avoid the torpedoes that were thought to have been launched against them. It was this high-speed gyrating by the American warships through the waters that created all of the additional sonar reports of more torpedoes. Every time one of the destroyers changed course, the sonar reported the distinctive high-speed sounds of torpedoes. Eventually, Herrick and the other officers realized what was happening: the rudders of the two ships had caused the high-speed returns when they reflected the turbulence of the ships' own propellers.95
Within an h our of the end of the attack, Herrick relayed his doubts about the attack in an after-action report. After reviewing the number of contacts and possible sinkings, he stated, "ENTIRE ACTION LEAVES MANY DOUBTS EXCEPT FOR APPARENT ATTEMPTED AMBUSH AT BEGINNING."96 Herrick then suggested in the morning that there be a thorough air reconnaissance of the area for wreckage. In a follow-up message, Herrick added that the Maddox had "NEVER POSITIVELY IDENTIFIED A BOAT AS SUCH."97
Herricks' doubts did not sit well with Washington. Since the first Critic warning of the attack, which had arrived at 0740 EST, Washington had been following the action in the Gulf of Tonkin. At 0925 EST, Secretary McNamara had called the president with the news of the imminent attack. At 1000 EST the flash message from the destroyers that they were under attack reached the Pentagon. Within three hours after the attack ended, 1400 EST, President Johnson had already approved a retaliatory strike against North Vietnamese naval bases to be carried out at 1900 EST, 4 August (0700G, 5 August.)
Precisely why President Johnson ordered a retaliatory strike so quickly is not totally clear, especially when there was conflicting evidence as to whether it had actually occurred. Johnson was in the midst of a presidential campaign and his opponent, Republican senator Barry Goldwater from Arizona, a noted hawk, would have gained in the race if Johnson had hesitated or refused to retaliate. Johnson, even in his pose as a moderate relative to Goldwater, could hardly appear weak before a public audience demanding a counterstrike.98 It also has been suggested that when Johnson first learned of the possible attack, that is, the first Critic issued by Phu Bai, he decided to use the warning as an excuse to get Congress to pass what was soon to be known as the Gulf of Tonkin Resolution.99
Whatever the president's own rationale for ordering the air strike, he required immediate verification of the North Vietnamese attack because of the doubts that started to be openly expressed within the administration. At around 1400 EST, Admiral Ulysses S. Sharp, CINCPACFLT, called the Pentagon with the news that "a review of the action makes many reported contacts and torpedoes fired 'appear doubtful'" because of freak weather, over-eager sonar operators,
and the absence of visual sightings.100 McNamara called Sharp, who added that there was "a little doubt on just what exactly went on."101 Messages buzzed back and forth between Washington and the Pacific, demanding information and then getting contradictory evidence of the attack. The Desoto mission reported that except for possibly the first torpedo report at 2159G (1459Z), all others were caused by reflections off the two destroyers' screws.102 At the same time, Herrick reported that the air cover from the two carriers was unable to locate the targets because of poor weather. Yet the carrier Ticonderoga transmitted its own evaluation in which the pilots had 'REPORT[ED] NO VISUAL SIGHTINGS OF ANY VESSELS OR WAKES OTHER THAN TURNER JOY AND M[ADDOX]. WAKES FROM TURNER JOY AND M[ADDOX] VISIBLE FROM 2-3000 YARDS."103 Crews from the two destroyers reported seeing nothing for certain. One sailor thought he had seen flashes of gunfire, but wasn't sure.
Then, like a classic deus ex machina, along came a second SIGINT report that seemed to clinch the case for an attack. This report was a translation issued by NSA on the 4th of August at 1933Z (1433 EST in Washington) and was leaped upon by administration officials, especially the secretary of defense, Robert McNamara, as direct evidence of the attack. What this translation appeared to be was a sort of North Vietnamese after-action report. An unidentified North Vietnamese naval authority had been intercepted reported that the DRV had "SHOT DOWN TWO PLANES IN THE BATTLE AREA," and that "WE HAD SACRIFICED TWO SHIPS AND ALL THE REST ARE OKAY." It also added that "THE ENEMY SHIP COULD ALSO HAVE BEEN DAMAGED."104
At 1640 EST, Admiral Sharp again called McNamara with more information on the attack. Just before 1700 EST, McNamara and the JCS met to evaluate the evidence on the attack. They concluded that it had occurred and that five factors were critical: "(1) The Turner Joy was illuminated [by a searchlight] when fired on by automatic weapons; (2) One of the destroyers observed cockpit [bridge] lights [of one of the DRV patrol boats]; (3) A PGM 142 had shot at two U.S. aircraft (from COMINT); (4) A North Vietnamese announcement that two of its boats were 'sacrificed' (from COMINT); (5) Admiral Sharp's determination that there was indeed an attack."105
Of the five pieces of "evidence," two were from the same NSA product issued that afternoon (EST). If the two pieces of visual evidence -- the searchlight and cockpit light reports -- were contentious, the SIGINT, was in the minds of the secretary of defense, the JCS, and the president, the "smoking gun" evidence needed to justify the air strikes on North Vietnam.106

(0000Z) on 5 August, CINCPAC received the order to execute the retaliatory raid, codenamed Pierce Arrow. At 1030G (0330Z), naval strike aircraft from Ticonderoga were launched. By early afternoon they hit several targets in the DRV, including almost all of its naval installations.
However, there were many problems specific to the SIGINT information which emerged almost as soon as it was being reported. In this section we will reconsider what happened that night using all of the relevant SIGINT. We will begin with the initial order to the Vietnamese boats ordering them to make ready for military operations.
What made this intercept a Critic was the interpretation put to it by the Marine SIGINT site at Phu Bai, which stated that this was an "OPERATION PLANNED AGAINST THE DESOTO PATROL."108 The follow-up report from Phu Bai amplified the original Critic and maintained, as well, that the attack was against the Desoto mission.109 When one considers the events of 2 August, this interpretation was n to totally unfounded; one could see a reference to a military operation being directed against the American warships. However, the text of the intercept never mentioned a target or any objective of the military operation, or even the nature of the operation. As we shall see soon, not everyone who saw this intercept jumped to the same conclusion that an attack against the American ships was being planned.
Another problem is that the decrypted Vietnamese phrased for military operations, hanh quan, has an alternate meaning of "forced or long march or movement," which, in a nautical context, could refer to a voyage by both T-146 and T-333. As it turns out, this is the activity that the intercept was actually alluding to.
For at 1440Z, almost at the precise moment: that Herrick ordered his two destroyers to open fire on the approaching radar returns, the Phu Bai intercept site issued a spot report which
![]() Phu Bai Critic alerting Desoto patrol to possible attack. |
![]() Translation of the intercepted original of the "attack" order. |
stated that both DRV torpedo craft, T-336 and T-333, the latter of which earlier had been reported ready to attack the Desoto patrol, were, in fact, being readied to be towed to either Haiphong or Port Wallut. This second report carried two salient points: First, at 1946G (1246Z), Swatow T0142 reported to Haiphong that the tug Bach Dang was unable to return to port. T-142 also included the statement that if the ship [Bach Dang] "MET THE DESOTO MISSION, IT WAS TO [A]VOID THEM."110 Besides being a warning about the Desoto ships, the message also implied that the North Vietnamese thought that the destroyers were close enough to shore to be a threat to DRV vessels, whereas, at this time, the American ships were far out at sea. In all probability, the North Vietnamese had lost track of the American destroyers (an issue which we will discuss further on in this narrative).
The second point of the Phu Bai report was that at 2031G (1331Z) T-142 had informed an authority in Port Wallut that the tug was towing the two craft from Squadron 135. The analysts at Phu Bai added this comment to the end of their report which read, "WITH THE MTB 336 ADDED TO ITS STRING, IT AP[PE]ARS THAT T333 WILL NOT PARTICIPATE IN ANY MILITARY OPERATIONS." So, the boats originally reported being ready to attack the Desoto patrol, were incapable of even moving on their own!
In fact, this attempted salvage of the two damaged torpedo boats would occupy the efforts of Hanoi's sailors for much of the night of 4/5 August. The Vietnamese would try various methods of getting the two damaged P-4s to a port for repairs. During the 2300G hour, T-146 was ordered by Haiphong to escort the Bach Dang as it returned to base. When that was completed,
T-146 was ordered to Bay Chay, a point near Haiphong harbor.111 Shortly afterwards, T-142 informed Haiphong that the very busy T-146 was now to tow T-336 back, but since the latter boat was short of fuel, the T-333, which was short of oil but under tow from the Bach Dang, could transfer one to five tons of its fuel to its sister vessel.112 At 1830Z on 4 August (0130G on 5 August), the navy monitoring site at San Miguel intercepted T-142's report to Haiphong that T-146 had completed its preparations for the two torpedo boats by 0100G 5 August (1800Z 4 August).113 So, in reality, none of the boats named in the original attack Critic in fact participated in anything but salvage efforts.
Remember, Captain Herrick did not know that the original Critic was really an interpretation, and that there was no explicit reference to an attack on his ships. He accepted the Critic's contents as intercept of actual Vietnamese plans to attack his ships when he informed the Ticonderoga task group commander of his decision to leave the area. He added his own twist to the report to include specifically the unsupported amplification mentioning the involvement of North Vietnamese P-4 torpedo boats when only one was mentioned as a potential participant in the unidentified operations, and then only if it could be refueled.114
The possibility that, even if the interpretation was incorrect, the Marine Critic was justified in light of the events from two days earlier, does not stand up when we consider that another site, the navy intercept station at San Miguel, Philippines, had translated the same "operations order," but reported it in a much different fashion. The navy translated the same intercept and then reported it at a Priority precedence, two levels below a Critic (or one level above Routine). The navy analysts titled the report "REPLENISHMENT OF DRV NAVAL VESSEL." The San Miguel report translated the critical sentence as "T146 SUPPLY FUEL FOR THE 333 IN ORDER TO GIVE ORDERS TO PUT INTO OPERATION ((2 GR G)) WITH T146."115
The difference (and correctness/incorrectness) between the translations is not important as much as the fact that San Miguel viewed the information as nothing more than the refueling of the damaged torpedo boats. This was in line with an earlier intercept of a query from Haiphong to T-142 asking if T-333 had been refueled yet.116 Unfortunately, because the San Miguel version was a lower precedence, it was released much later. In fact, it came out at 1838Z (0038G), some two hours after the destroyers had stopped shooting.
The quandary created by the reports about the salvage operations is this: If the original suspect vessels, the two Swatow-class patrol and two damaged P-4 torpedo boats, were not participating in the anticipated "attack" against the Desoto patrol, then who exactly was going to attack? No other messages had been intercepted which suggested that any other DRV boats were handed the mission of attacking the American destroyers. In fact, there was no intercept at all which hinted at an attack; nothing at all like what had been intercepted on 2 August. So, if the original culprits were involved in salvage operations, then just what was going on in the Gulf of Tonkin?
For NSA and the rest of the SIGINT participants, the second Phu Bai report should have acted as a brake to any further reporting about an attack. It directly contradicted the interpretation -- remember, it was an interpretation only -- contained in the initial Critic which claimed an attack was being prepared. At this point, all the SIGINT community could accurately state was that there was no signals intelligence reflecting a planned or ongoing attack against the Desoto mission.
Except this is not what happened. The second Phu Bai report was not used to report
what was going on in the Gulf of Tonkin. Instead, the problem posed by the second Phu Bai report was handled in a curious manner. Late on 4 August, Washington (050130Z August 1964), NSA issued a Gulf of Tonkin situation report which covered the events of 4 to 5 August. At the end of the report, NSA added these interesting sentences: "ALTHOUGH INITIAL MESSAGES INDICATED THAT THE T142, T146, AND T333 WOULD BE INVOLVED IN THE ATTACK . . . SUBSEQUENT MESSAGES [not further identified in the report -- a curious lapse by NSA which we will address in detail later] SUGGEST THAT NONE OF THESE [BOATS] WAS INVOLVED. REPORTS FROM THE MADDOX THAT IT WAS UNDER ATTACK SOME SEVENTY NAUTICAL MILES NORTHEAST OF THE NAVAL BASE AT QUANG KHE SUGGEST THAT NAVAL UNITS SUBORDINATE TO THE SOUTHERN FLEET COMMAND . . . WERE INVOLVED. . . .117
However, the effort to find "culprits" only compounded the errors: the only boats known to be stationed permanently at Quang Khe were Swatow- class patrol boats which did not carry torpedoes.118 All P-4 torpedo boats staged from Port Wallut far northwest of the action. Accusing the Swatow craft of participating in the attack was no "solution"; in fact, it only added to the confusion. In reality, though, this statement by NSA was a vain attempt to cover the problem of the contradictory report from Phu Bai. It was nothing but speculation -- ignorant speculation at that. Furthermore, this summary report still did not address the issue of the total lack of intercept of any North Vietnamese attack command and control communication.
Fingering the Swatows as the culprits only made the "attack" scenario more improbable for another reason. The distance from Quang Khe naval base (17°46'N, 106°29'E) to the reported first radar plot by the Maddox, forty to forty-five nautical miles northeast of its position, is about 120 nautical miles. However, this distance should not be construed as a "straight line" dash from Quang Khe. Because the DRV boats were "detected" coming from the east, they would have had to travel in a long arc northward and then southeast around the American destroyers which were speeding to the southeast. Also, remember that the Maddox and Turner Joy did not "detect" these boats until they approached from the east, so the route to the north of the American destroyers had to be at a distance sufficient to avoid discovery by radar. This lengthens to a distance of around 180 nautical miles. Since the "attack order" was issued at 1115Z and the initial radar plot was at 1336Z (and we are presuming that the postulated boats left at the exact time of the first intercept, or were soon under way at the time), then the boats would have had to have been traveling at a speed of nearly seventy miles per hour (about 110 kph) to have been where the Maddox first detected them -- at a rate some 58 percent higher than the Swatow's known top speed!
The only other base from which the "attack" could have been staged was Port Wallut, which was the base for the P-4 Squadron 135. The distance from Port Wallut (21°13'N, 107°34'E) to the initial point of detection by the Desoto radars is about 140 nautical miles. However, the same problem exists here as for Quang Khe, though not quite as extreme, for the P-4s. The scenario presumes that they would have been moving at a little less than seventy miles per hour, or a good 40 percent higher than the boat's listed maximum speed.
Another possibility to consider when looking at the "attack message" is that there was some other activity to which the "military operations" (if that is the interpretation one could have) might have referred. In fact, there was something else going on that night of 4/5 August which is seldom mentioned in the public record: a maritime OPLAN-34A mission was, in fact, moving northward along the DRV coastline at the time when the American destroyers were shooting away at those radar returns. The Marolda and Fitzgerald history of the U.S. Navy in Vietnam
fails to mention the ongoing 34A mission. Official Washington as well never mentioned this 34A mission. In classified hearings in February 1968, Secretary of Defense McNamara never mentioned this mission, claiming that the last one prior to the 4 August attack occurred on the night of 3-4 August. Obviously, if the 34A mission of the night of 4-5 August were known at the time, it would have undercut Washington's claim that nothing else was happening that night which might have provoked Hanoi.
This 34A mission had been scheduled back at the end of July by COMUSMACV, which then had informed Washington of the missions planned for all of August. This particular foray's main objective was the shelling of the island of Hon Matt. It is not certain when this mission left Danang, though it was normal for the boats to depart in the late afternoon to take advantage of darkness by the time they reached the DRV coastline. So a departure time between 1500G and 1600G (0900Z) would not be too far off.
At 2316G (1616Z) the Marine mission at Phu Bai intercepted a message from the DRV naval HQ in Haiphong to T-142 that six enemy raiding vessels had been located somewhere south of Thanh Hoa (20°00'N, 105°30'E). (The actual position is confusing due to a garble in the text transmitted from Phu Bai. Neither the time of the enemy boats' position nor their course is clear.)119 This intercept occurred only a few minutes before the JCS approved an urgent recall order from CINCPACFLT for the 34A mission to be discontinued and return to Danang immediately.120 It is possible that the Kit Kat support element may have passed this intelligence to the MACV/SOG, which in turn began the recall.
In light of what finally transpired with T-142 and the two P-4 torpedo boats, it seems that they were not part of a defensive plan against the raiders. That this Swatow received the message about the raiders does not seem odd in light of the fact that T-142 seems to have served as some sort of radio relay for other boats or as a communications guard vessel for all DRV naval operations: a majority of intercepted messages during the period seem to have been sent to or through T-142. From other intercepts, we know that at least another Swatow, T-379, was near Hon Matt; two others, T-130 and T-132, were near Hon Me Island; and T-165 had deployed, as well. If the DRV was planning to attack the 34A raiders on 4 August, these craft would have been the logical ones to use because of their substantial deck gun armament. However, no other communications activity related to any other Swatow patrol craft was intercepted that night. So it remains uncertain what, if anything, Hanoi was planning to do to fend off the 34A mission of 4 August.
During the entire day of 4 August, most of the communications intercepted from either DRV naval command entities in Port Wallut or Haiphong either were directed to the craft involved in the salvage and recovery of the two Squadron 135 torpedo boats, or else were
relays of tracking reports of the Desoto patrol, and these latter messages were exchanged with T-142, which was involved in the ongoing recovery operations. The only other messages which were intercepted contained orders for other Swatow-class patrol boats to move to positions along the coast: T-130 and T-132 were ordered to Hon Me Island, while T-165 was ordered to leave Haiphong at 1448G (0748Z) and move to the entrance of an unspecified bay.121
During the 2 August attack, there were elements of high-level control from the naval commands at Port Wallut and Haiphong, both of whom sent orders and tracking reports to the attacking boats. The Swatows, principally T-142, acted as a communications relay between the torpedo boats and the onshore commands. The messages were transmitted using high frequency manual morse communications which were intercepted throughout the day, even during the fighting. Finally, there were sporadic boat-to-boat VHF, tactical voice communications which the intercept positions aboard the Maddox's hut could intercept, at least until the destroyer activated its fire control radars, which interfered with the navy's monitoring.
However, not one of these elements was detected during the night of 4 August. Trying to find more evidence of the purported attack, NSA had queried the NSG detachment aboard the Maddox on 6 August to supply urgently all intercept that "PROVIDES PROOF OF DRV ATTACK ON FOUR AUGUST UPON U.S. NAVAL VESSELS."122 Within five hours came the disheartening reply from the DSU. There was no manual morse intercept to prove the DRV attack of 4 August. Furthermore, voice intercept was nil, except for signal checks between two unidentified stations.123
The tracking messages locating the Desoto patrol ships had been intercepted by the Americans early in the day of 4 August. However, the last credible position of the American ships was passed at 1610G (0910Z) from Haiphong to T-142. The position, 19°36'N, 106°19'E, was fairly close to the Desoto patrols position at the time. This was just about two hours before Herrick ordered his ships to head east in reaction to the Phu Bai Critic.124 However, it should be pointed out that this position report was sent to the T-142, which was involved in the salvage of the two torpedo boats. There is no evidence that the T-142 relayed it to any other boat or command.
One more position report on the Desoto patrol was sent from Port Wallut to a probable vessel at 2246G (1546Z), which was about an hour after the supposed engagement had begun. This position report might seem as related to the action, except for two problems. First of all, the report located the American ships thirty-five nautical miles east of Hon Matt Island, which places the destroyers some eighty nautical miles northwest of where they actually were at the time! In addition, the report does not carry the time associated with the Americans' position. (The reported location suggests, however, at least from the track the patrol took that night, that this position report was about four to five hours old.) So, this information could hardly be used by any North Vietnamese boats intending to attack the Americans. Secondly, the message includes an order (or advisory) to the recipient to maintain a continuous communications watch with an unidentified entity, as well as to "go close to shore."125 This latter command seems to be hardly intended for boats looking to attack the American ships; rather it appears suited for the boats involved in the salvage operations or the other patrol boats spread out along the DRV coast.
The issue of DRV tracking of the Desoto patrol is important. For in September 1964 NSA would release a report on Vietnamese coastal radar operations during the period. In this report, NSA would contend that active tracking by the coastal observation posts equipped with coastal surveillance radars would indicate hostile
intention by Hanoi. The report pointed out that there was no tracking of the Craig earlier in March.126 This was not quite true: the DRV was aware of the location of the destroyer, but its time off the Vietnamese coast was quite short so the tracking was spotty.
The same report also pointed out that the Maddox was under "constant" radar surveillance before it came under attack on 2 August. However, the report then ducks the issue of the observed sporadic tracking by the North Vietnamese on 4 August with the claim that "The evidence is still inconclusive in light of the virtual absence of trackings on 3-4 August before the second attack."127 The evidence would never be found. The final report from the DSU aboard the Maddox showed only occasional coastal tracking from shore stations and North Vietnamese boats on 4 August. And it had ended by mid-afternoon.128
Finally, the Americans detected no Skin Head emissions during the "attack" on 4 August. Keep in mind that during 3 August the DRV boats that shadowed the patrol used the Skin Head surface search radars, and that these emissions were detected by the ELINT position in the intercept hut aboard the Maddox. These signals were also intercepted during the morning and early afternoon of 4 August.129
While it is true that no North Vietnamese radar emissions were detected during the 2 August attack on the Maddox, it must be remembered that this attack occurred in the daytime under nearly ideal conditions.130 Yet, the DRV boats had initial difficulty visually locating and then following the Maddox. What we are confronted with in the second "attack" is the proposition that the North Vietnamese boats themselves, which the Turner Joy and Maddox detected using only their radars, could find the Americans so far out at sea (over 100 nautical miles), in heavy swells (three to six feet), at night, with a low cloud cover, without using their radars. Even if the North Vietnamese had the equipment to receive the American radar pulses, this information would have given them only a crude bearing on which to track. They could not determine distance, speed, or anything else with which to plot any sort of torpedo attack.131 Besides, that, how could they even begin to track the American ships when the latest valid position was almost five hours old!
In the Sherlock Holmes story "Silver Blaze," the great Victorian detective and his assistant, Dr. Watson, are confronted with the paradox of a crime which cannot be proved to have happened. In the story there is this exchange:
Is there any point to which you wish to draw my attention?To the curious incident of the dog in the night-time.
The dog did nothing in the night-time.
That was the curious incident, remarked Sherlock Homes.132
And so it is with the 4 August incident: there were no DRV naval communications or radar emissions which were normally associated with a naval engagement. Just two days prior, the Americans had an opportunity to observe Vietnamese naval communications during the attack on the Maddox. Among other things, they had seen that the Vietnamese had difficulties in setting up and maintaining control of an attack, as the incident with the conflicting orders illustrated. And so there should have been a generous amount of intercept of any communications which would have supported the claims of the two American destroyers.
Yet, nothing as much as a single bark was intercepted. As Holmes would come to conclude that no crime was committed, so we must conclude that, since U.S. SIGINT never intercepted anything associated with an attack, none ever occurred. And the contention that all possible communications and emissions reflecting an
attack might have gone unheard can be dismissed. As Gerrel Moore, the officer-in-charge of the DSU on board the Maddox, observed: "I can't believe that somebody wouldn't have picked up something."133 Finally, a review of the DSU intercept log for 4 August showed no variation in Vietnamese communications procedures which could suggest that any change or changes, such as new operating frequencies, callsigns, or procedures, were implemented just for the "attack" that could elude American intercept.134

The first difficulty with the intercept is that it does not resemble an after-action report of the type which had been intercepted early on 3 August by the Marine element at Phu Bai. That intercept, sent by T-142 to T-146 and the Port Wallut HQ of Squadron 135, contained a chronology of events beginning at 0925G on 2 August when T-146 met the three boats from Squadron 135 and guided them to Hon Me Island. The report noted that the attack against the Americans began at 1525G, and that by 1625G, all the boats had received the orders to break off the attack.135
In the 4 August translation, there is no chronology associated with the supposed downing of the aircraft. There is no mention of any participating boats or units, except to mention that two were "SACRIFICED . . . AND ALL THE REST ARE OKAY." The only sense of when anything happened comes with the beginning phrase, "AFTER THE 135 HAD ALREADY STARTED TO REPORT TO YOU." In fact, the entire report seems incoherent, not the type one expected to see sent by an officer on the scene, as had been intercepted on 3 August. It rambles, mixing morale boosting statements with seemingly repetitious references to planes being shot down and then seeing them "sink."
Secondly, there is a problem with the translation of a critical passage: "WE SACRIFICED TWO SHIPS AND ALL THE REST ARE
OKAY." Unfortunately, the original, decrypted Vietnamese language version of the message cannot be located in the NSA Archives. Also, a possible original translation of the entire message (or part of it), numbers "T162-64" and issued by the navy site at San Miguel, cannot be found in the NSA Archives files of that site's 1964 translations. Without either document, we are left with the conjecture of what Vietnamese words were seen by the navy analysts and linguists at San Miguel and their counterparts at NSA.
However, from the existing records, what we do know is that the translation finally issued by NSA was not what was initially reported by San Miguel. At 15550Z (2250G) on 4 August, when the American destroyers were shooting away at those radar returns, San Miguel intercepted a message which it identified as being sent from T-142 to an unidentified entity at My Duc (19°52'N, 105°57'E). In total, the report, numbered "R38," read:
WE SHOT AT TWO ENEMY AIRPLANES AND AT LEAST ONE WAS DAMAGED. WE SACRIFICED TWO COMRADES BUT ALL ARE BRAVE AND RECOGNIZE OUR OBLIGATION.136
How the translation changed from "comrades" in the San Miguel version to "boats" in the NSA version is unknown. Edwin Moise, in his study of the Tonkin Gulf, suggests that a Vietnamese sentence to the effect of losing two comrades could hardly be construed to mean two ships "HAI DONG CHI HY SINH" or "HAI DONG CHI BI HY SINH" are possible Vietnamese phrases which could be translated to "sacrificing two comrades."137 The Vietnamese word for boat, "TAU," had been seen in earlier intercepted messages. This would be consistent, since Hanoi's messages usually shorted the word to just the letter "T" from where the same letter designators for Hanoi's boats comes from, such as "T-142," "T-146," etc.
A possible argument that there was a garble in the encryption of the message which could have led to confusion does not hold. [***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** redacted text ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** ***** *****