Two Boeing 747s collided on the runway at Los Rodeos Airport in Tenerife on 27 March 1977, and 583 people died, which remains the largest death toll in any civil aviation accident. The case is taught everywhere as the origin of standardized cockpit and tower phraseology, and that is fair as far as it goes, although the wording failure sat on top of a terrorist bombing, a mass diversion into an airport that could not hold the traffic, drifting fog, and a tower with no way to see either aircraft. This piece works through the radio mechanics and what they changed.

The bomb at Las Palmas and the diversion that filled Los Rodeos

The day began as a normal Sunday of charter and scheduled traffic into Gran Canaria Airport at Las Palmas, the main gateway for the Canary Islands. Early in the afternoon a small explosive device detonated inside the passenger terminal, injuring several people, and a phoned warning of a second device led the authorities to close the airport and evacuate while a search was conducted. Responsibility was claimed on behalf of a Canary Islands independence movement. No one died in the bombing, and the airport reopened later the same afternoon, by which time the diverted traffic had already been sent to other fields in the archipelago.

The largest share of it went to Los Rodeos, on the north side of Tenerife, an airport that sits at roughly 2,000 feet of elevation and that handled nothing like the volume Las Palmas did. Several diverted airliners arrived over a short period, among them KLM Flight 4805 from Amsterdam and Pan American Flight 1736, which had come from Los Angeles by way of New York and was carrying a large group of passengers bound for a cruise ship. The Pan Am crew asked to hold clear and remain aboard rather than land, since their passengers did not need to deplane, and were told to land because the ramp situation at Las Palmas was unresolved.

Los Rodeos has one runway with one parallel taxiway alongside it, and the diverted aircraft were parked on that taxiway because there was nowhere else to put them. That single decision, made under the pressure of an unplanned mass diversion, set up everything that followed, because it meant that when Las Palmas reopened and the aircraft began to leave, departing traffic had to taxi on the runway itself rather than beside it. Low cloud was drifting across the field in patches while this was happening, and visibility at the airport changed substantially from one minute to the next.

KLM elected to take on a large load of fuel at Los Rodeos rather than wait to refuel at Las Palmas. Published accounts of the investigation generally put that refueling at something over half an hour, and while it was underway the Pan Am aircraft could not get past the parked KLM to reach the runway, since the lateral clearance between the Pan Am wingtip and the KLM was measured on the ramp and found to be insufficient.

Backtracking a single runway in cloud

Departures were being made from runway 30. With the parallel taxiway occupied, the tower instructed aircraft to enter the runway and taxi down it in the opposite direction to the takeoff run, then turn 180 degrees at the far threshold and depart. This procedure, usually called backtracking or backtaxiing, is routine at small airports and entirely normal, and it depends completely on only one aircraft being on the runway at a time, or on everyone knowing exactly where everyone else is.

KLM was cleared to backtrack the full length of the runway and make the 180 degree turn at the end. Pan Am was then cleared onto the runway behind it, with instructions to leave the runway at the third exit and continue to the holding point on the taxiway. The exits connecting the runway to the parallel taxiway were angled back toward the terminal end of the field, which meant that an aircraft taxiing in the backtrack direction had to make a very sharp turn to use one. The investigation examined the geometry of the third exit and found that using it would have required a turn of roughly 148 degrees, which is not something a 747 performs as a single movement.

The cockpit voice recorder from the Pan Am aircraft captured the crew discussing which exit was meant and counting them off as they went, and the aircraft continued past the third toward the fourth, which lay at a much shallower angle and could actually be taken. Whether the controller’s count started at the first exit the aircraft would physically encounter is one of the points on which later analyses differ. The relevant fact for radio purposes is that the tower could not see either aircraft, the aircraft could not see each other, and Los Rodeos had no surface movement radar, so the controller’s entire picture of where two 747s were sitting on his runway came from what the crews told him on the air.

The runway centerline lighting was reported unserviceable at the time, which removed one of the cues a taxiing crew would ordinarily use in reduced visibility. Fog was thickening as the two aircraft moved, and the KLM crew, having completed the turn at the threshold, were lined up on runway 30 facing a Pan Am aircraft they had no way of seeing.

One AM channel, no ground radar

Civil aviation voice communication in the VHF band, roughly 118 to 137 MHz, uses amplitude modulation, and it did then as it does now. That choice has a consequence that matters enormously to this accident. An AM receiver has no capture effect, so when two stations transmit at the same time on the same frequency, the receiver does not lock onto the stronger one and discard the weaker. It reproduces both, along with a heterodyne whistle at the difference between the two carrier frequencies, and the result ranges from partially intelligible to a squeal that carries no information at all.

This is often described as an advantage of AM for aviation, and in a sense it is, because a pilot who hears a squeal knows something happened, whereas an FM system would have delivered one clean transmission and silently discarded the other. The same property makes a stuck microphone audible rather than invisible. What AM does not do is tell the transmitting stations that they doubled, so both pilots who keyed up at the same moment heard their own sidetone and had no cue whatever that another aircraft was on the air with them.

The controller at Los Rodeos was working two 747s on a runway he could not see, using a single simplex channel, with no ground radar display, no data link, no electronic strip system, and no automated conflict detection of any kind. His only instrument was the readback. Modern airports layer surface movement radar, multilateration, runway status lights and in some cases controller pilot data link on top of that voice channel, and every one of those layers exists because the voice channel by itself cannot confirm where an aircraft physically is.

It is worth being blunt about what the channel could and could not carry. A frequency is a shared medium, and in 1977 aviation had no mechanism for priority, preemption, or queuing on it, so an urgent transmission competed on equal terms with a routine one. The Pan Am crew’s statement that they were still on the runway was the most important sentence spoken that afternoon, and the medium had no way to give it precedence over anything else.

Route clearance, takeoff clearance, and the word in between

A point that gets lost in popular retellings deserves stating flatly, because it is the hinge of the whole case. Two different clearances were in play. The first is the air traffic control clearance for the route, which tells an aircraft where to go and at what altitude after departure and can be issued at any point on the ground. The second is the takeoff clearance, which authorizes the aircraft to begin its takeoff roll on a specific runway. The KLM aircraft received a route clearance that afternoon, and no takeoff clearance was ever issued to it.

Lined up on runway 30, the KLM first officer advised the tower that the aircraft was ready and was waiting for its ATC clearance. The controller issued the route clearance, which in the published transcript directs the aircraft to a beacon, gives a flight level and a heading, and includes the phrase describing a right turn after takeoff. The word takeoff therefore appeared inside a clearance that had nothing to do with authorizing a takeoff, in an instruction issued to an aircraft already in position on the runway.

The first officer read the clearance back, and at the end of the readback he appended a statement about the aircraft’s status. The published transcript renders it as a declaration that they were now at takeoff, and the precise wording has been argued over for decades, because the rendering varies slightly between the Spanish investigation’s transcript and other published versions, and because the speaker was a non-native English speaker using a construction that has no fixed meaning in either English or standard phraseology. Depending on how a listener parsed it, it announced either that the aircraft was in position waiting or that it was beginning to roll.

The captain had already released the brakes and begun the takeoff roll while that readback was being transmitted. The controller replied with a single word, rendered in the transcript as OK, which acknowledged nothing in particular and authorized nothing at all. After a pause of roughly two seconds he added an instruction to stand by for takeoff, saying he would call them, and that instruction is the one that would have stopped the roll had it been received in the cockpit.

The blocked transmission and what each cockpit heard

The Pan Am crew, still on the runway and listening to the same frequency, heard the KLM readback and understood exactly how dangerous it was. The first officer keyed up and transmitted that they were still taxiing down the runway, identifying the aircraft. That transmission overlapped the tower’s transmission almost exactly.

In the KLM cockpit, the result was the word OK followed by a heterodyne squeal lasting a few seconds, which is the audio signature of two AM carriers in the same receiver. The instruction to stand by for takeoff was inside that squeal, and so was the Pan Am warning. Neither transmission reached the KLM flight deck in an intelligible form, and either of them on its own would have been enough to stop the takeoff roll.

What followed is documented on the KLM cockpit voice recorder and is one of the reasons this accident became a foundational case in crew resource management. The controller asked Pan Am to report when the runway was clear, and Pan Am acknowledged that they would report when clear, which the KLM crew did hear. The KLM flight engineer then asked, in Dutch, whether the other aircraft was not clear yet. He was asked to repeat it, repeated it, and received an emphatic affirmative from the flight deck, after which the takeoff roll continued with no further challenge from anyone on board.

The KLM captain rotated early and dragged the tail on the runway, and the aircraft was airborne but low when it reached the Pan Am 747, which was turning left toward the fourth exit. The KLM’s landing gear and lower fuselage struck the Pan Am across the top of its fuselage behind the wing, tearing the upper deck open. The KLM climbed briefly, stalled, and came down on the runway a short distance beyond, sliding and burning. Both aircraft were destroyed by impact and fire at about 1706 local time.

The casualty figures and where the reports disagreed

All 248 people aboard the KLM aircraft, 234 passengers and 14 crew, were killed. The Pan Am aircraft carried 396 people, 380 passengers and 16 crew, of whom 335 died. Those numbers give the total of 583 that appears in the Spanish investigation and in every subsequent authoritative treatment, and unlike the casualty figures for many historical disasters, this one is not seriously contested. The survivor accounting is where published accounts vary slightly, because a number of people who escaped the Pan Am aircraft alive died of their injuries afterward, so counts taken at the scene and counts taken later are not the same figure. The commonly published number of survivors is 61.

The investigation was led by the Spanish civil aviation authority, with accredited participation from the Netherlands and the United States and from KLM, Pan Am, Boeing and the Air Line Pilots Association. The Spanish report identified the fundamental cause as the KLM captain taking off without clearance, and listed contributing factors including the use of non-standard phraseology by both the KLM crew and the controller, the simultaneous transmissions that blocked the warning, the weather, and the continued presence of the Pan Am aircraft on the runway.

The Dutch authorities did not sign up to that emphasis without reservation. Their commentary accepted that the KLM aircraft began its takeoff without a clearance and did not dispute that fact, while arguing that responsibility was more widely distributed than the Spanish report allowed, pointing to the controller’s non-standard wording and his use of an ambiguous acknowledgment, and to the Pan Am crew’s failure to leave the runway at the exit they had been given. Those two positions genuinely differ in their allocation of responsibility, and anyone writing about this accident should say so rather than presenting one of them as the settled verdict. KLM later accepted liability and paid compensation.

Separately, the Air Line Pilots Association published a human factors study of the accident, and a large body of analysis since has focused on the KLM captain’s position as the airline’s chief of flight training, the fact that he had checked out the first officer flying with him, and the pressure created by Dutch duty time rules that had recently become stricter. That work is analysis and interpretation of the recorded behavior rather than a formal finding of cause, and the formal findings remain those published in the Spanish report and in the Dutch commentary on it.

What actually changed in the phraseology

The reform that came out of Tenerife and the accidents around it is narrower and more specific than the folklore suggests, and its central rule is easy to state. The word takeoff is reserved for the issuing and the cancellation of a takeoff clearance, and it is not used at any other time. When a controller needs to talk about the departure in another context, the word is departure, so a route clearance now speaks of a right turn after departure. When an aircraft is being held in position on the runway, the phrase is line up and wait, which contains no form of the word takeoff at all.

The United States kept its own phrase, position and hold, for many years after ICAO had standardized on line up and wait, and the FAA harmonized with the international phrase in 2010, which is worth remembering the next time somebody tells you that a good idea reaches every operator quickly. Readback requirements were tightened so that clearances and instructions affecting runway occupancy are read back in full, and the controller is required to listen to the readback and correct it, which converts the readback from a courtesy into a verification step with a named owner.

Ambiguous acknowledgments were pushed out. Roger and OK confirm receipt of a transmission and authorize nothing, and neither is acceptable as a response to a clearance that requires a readback. Conditional instructions involving an active runway are restricted, so a controller does not say something that a crew can parse as a clearance contingent on an event the crew cannot see. The current rules live in ICAO’s air traffic management procedures document and in Annex 10 on aeronautical telecommunications, and in the United States in the FAA’s air traffic control handbook and the Aeronautical Information Manual, and anyone who needs the exact current wording should pull the current edition rather than trusting a summary, including this one.

Two other threads run from this accident. ICAO’s English language proficiency requirements for flight crews and controllers, phased in during the 2000s, addressed the problem that standard phraseology only helps when the plain-language fallback also works. And the accident became a keystone case in the development of crew resource management, together with the loss of a DC-8 at Portland in 1978, because a flight engineer raised the correct concern, was overruled, and did not raise it a second time.

What a public safety radio shop should take from it

The first transferable point concerns doubling. On an analog FM public safety channel, and on P25 and other digital systems, the capture effect means that when two units key up at once, the receiving end usually hears one of them cleanly and hears nothing at all from the other, which is worse than the aviation situation rather than better. The dispatcher hears a complete, intelligible transmission and has no indication that a second unit was talking. Aviation at least produced a squeal that told everyone something was wrong, whereas on your system a mayday transmitted simultaneously with a routine status update can vanish with no squeal, no error tone and no entry in the log to show it was ever made.

The second point concerns the dispatch console. On most systems the console transmits with enough signal strength, or with enough system priority on a trunked system, that a dispatcher keying over a field unit wins, and the field unit’s transmission is lost. That is normally the behavior you want, and it becomes dangerous when it happens during the four seconds a crew spends declaring an emergency. Find out what your console actually does in your own configuration, because the behavior varies with the console model, the system type and how the priority settings were left at installation.

The third point is acknowledgment discipline. A copy, a ten four, or an OK confirms that audio arrived and confirms nothing about understanding. Tenerife turns on a controller saying OK to a transmission whose meaning he had not resolved, and on a crew treating that OK as permission. If an instruction involves a crew entering a structure, changing tactics, going to a different channel, or leaving a position, the correct response contains the substance of the instruction rather than an acknowledgment token.

The fourth point is the flight engineer. He asked the right question, he was told he was wrong by the most senior training pilot in the airline, and he did not ask again. Fire and EMS have exactly the same gradient between a probationary firefighter and a battalion chief, and the answer aviation arrived at was to write challenge and response obligations into procedure, so that a crew member who has raised a concern once is required by the procedure to raise it again when it has not been resolved.

What to do at your agency

• Have your communications officer test doubling on your primary dispatch channel and on your fireground talkgroup this month, with two portables keyed simultaneously at different distances, and write down what the console operator actually hears so that your telecommunicators know the failure mode by experience rather than by theory.
• Ask your console vendor or system administrator, in writing, what your dispatch console does to a field unit’s transmission when the dispatcher keys up during it, and file the answer with your communications documentation.
• Have your training officer add one item to the next officers’ meeting already on the calendar: any instruction that moves a crew, changes tactics or changes channels is read back with its substance, and an acknowledgment token by itself is not an acceptable response.
• Have your quality assurance reviewer pull five recordings of working incidents from the last quarter and count how many critical instructions received a substantive readback rather than a copy, then bring the count to the next training committee meeting.
• Write one paragraph into your existing mayday or emergency traffic SOP stating that any member may transmit emergency traffic without permission, that a concern raised and overruled is to be raised again to the next level, and that the incident commander acknowledges by repeating the reported condition.
• Have your communications officer check your channel and talkgroup names for words that carry more than one meaning in your organization, and rename anything that a stressed member could hear two ways.

Takeaways

• The Tenerife collision of 27 March 1977 killed 583 people, comprising all 248 aboard the KLM 747 and 335 of the 396 aboard the Pan Am 747, and that total is consistent across the Spanish investigation and later authoritative sources.
• A terminal bombing at Las Palmas, a mass diversion into an airport with one runway and a taxiway full of parked aircraft, drifting fog and the absence of ground movement radar created the situation, so treating the accident as a pure radio misunderstanding misstates the record.
• No takeoff clearance was ever issued to the KLM aircraft, which held a route clearance and an instruction to stand by, and the widespread version in which the tower cleared the flight and the crew misheard is a myth.
• Aviation VHF uses amplitude modulation, which has no capture effect, so the tower’s stand by instruction and the Pan Am warning arrived in the KLM cockpit as a heterodyne squeal and neither one was received.
• The Spanish report identified the KLM captain’s takeoff without clearance as the fundamental cause with several contributing factors, while the Dutch commentary argued for a wider distribution of responsibility that included the controller’s phraseology, and those two positions differ rather than agreeing.
• The phraseology reform reserved the word takeoff for issuing and cancelling a takeoff clearance, replaced it with departure everywhere else, introduced line up and wait for runway holding, tightened readback requirements, and removed roger and OK as acceptable responses to a clearance.
• On FM and digital public safety systems the capture effect makes doubling worse than it was at Los Rodeos, because the dispatcher hears one clean transmission and gets no indication that a second unit was talking at the same moment.
• The KLM flight engineer asked the correct question, was overruled by the most senior training pilot at the airline, and did not ask again, which is the same authority gradient that exists between a probationary member and a chief officer on any fireground.