Week 9: Accent, fatigue, and the party line

Intelligibility is a property of the channel, not of the speaker — and the shared frequency is a safety device nobody designed

3 May 2027

Two ideas this week, and they pull against each other.

Intelligibility belongs to the channel#

The instinct is to locate intelligibility in the speaker: that person is hard to understand. It is more useful, and more accurate, to locate it in the whole path — speaker, encoding, medium, noise, listener, and the listener’s expectations.

The same speaker is perfectly clear to a colleague on the same sector and opaque to a crew arriving from another region. Nothing about the speaker changed. What changed was the shared context the listener brought.

Three things degrade the path, and none of them is accent:

• Band limiting. VHF voice is roughly 300 Hz to 3 kHz. That removes much of what distinguishes consonants, which is why the alphabet from week 2 is built out of vowel-and-stress contrasts.
• Fatigue. On both ends. A tired listener’s comprehension degrades before their speech does, so the person least able to understand you sounds fine.
• Expectation bias. The strongest of the three. You hear the clearance you expected, and you hear it clearly.

That last one is week 3’s hearback failure, generalised. It is not a lapse of attention. It is what perception does: it fills gaps with priors, and it does not tell you it has.

The party line#

Aviation VHF is a shared frequency. Everybody on it hears everybody else.

From a pure information-design standpoint this is a defect. It is inefficient, it makes simultaneous transmission possible, it means a busy sector wastes most of its bandwidth on messages irrelevant to most listeners.

And it is also, accidentally, one of the most effective safety devices in the system. Because everybody hears everything:

• You build a picture of the traffic around you that nobody transmitted to you.
• You catch a clearance issued to a callsign similar to yours.
• You notice the controller is saturated from the rate, not from any words.
• And you can hear that somebody else is about to do something wrong.

Nobody designed the party line for those. It falls out of using one radio channel, and it turns out to matter enormously. Hold onto that — week 10 is about a technology that removes the problem the party line causes and takes the party line with it.

Air Canada 759, 7 July 2017#

At just before midnight at San Francisco, Air Canada 759 was cleared to land on runway 28R and lined up instead with taxiway C, on which four loaded airliners were waiting for departure. It descended to about 100 ft and overflew the first of them, then went around at a minimum of about 60 ft over the second.

Nobody was hurt. This is the closest thing in the modern record to a second Tenerife.

The communications sequence is the reason it is in this course.

The crew, uneasy, asked the tower to confirm the runway was clear. The controller — correctly, and uselessly — replied that there was nobody on runway 28R but them:

ACA759   Tower, just want to confirm, this is Air Canada seven five nine,
         we see some lights on the runway there, can you confirm we're
         cleared to land?
TWR      There's no one on runway two eight right but you.

That is quoted from the record rather than offered as a model, and the notation says so: atc-verbatim, not atc. Neither transmission is standard phraseology — “just want to confirm”, “we see some lights on the runway there” — and neither is the point. Marking it as a model would be the mistake week 7 warns about in the other direction.

Every word of that is true. It is also an answer to a question the crew were not really asking. They were not on 28R.

What stopped it was the party line. A pilot waiting on the taxiway — not in the loop, not addressed, simply listening — transmitted that the aircraft was lined up on the taxiway. The tower then instructed the go-around.

The NTSB found the probable cause to be the crew’s misidentification of taxiway C as the runway, from a lack of awareness that the parallel runway was closed, with expectation bias and fatigue as contributing factors. Read the report for those findings. Read the transcript for this course’s finding: the loop between the two parties whose job it was closed correctly and produced the wrong answer, and a bystander broke it open.

The tension#

Week 7 wanted the frequency disciplined and standard. This week points out that the frequency’s value here came from somebody who was not part of the exchange at all, listening to traffic that was none of their business.

Both are true. A designed channel and an overheard one are doing different jobs, and a design that optimises one can quietly remove the other.