November 18, 2006

Linguistics in the service of Plane English

No, it's not a typo. This is about the English used by pilots and control  towers -- "Plane English." Geoff Pullum's recent post Linguistics in the service of astrophysics prompted me to describe one way that linguistic service also extends to the field of product liability. Illustrating this is a 1987 Chicago case about a private Lear Jet that crashed seven years earlier in New Orleans, killing the pilot and his four passengers. It's a story about Plane English.

The lengths to which insurance companies will go in order to avoid payment seem to have no end. In this case, the company wasn't satisfied that pilot error caused the crash. Instead, it tried to put the blame on the Garrett Corporation, manufacturer of the airplane's engine. If it wasn't the pilot's fault, what else could have caused him to veer off course as he tried to land? The insurance company came up with the idea that it was a toxic gas called trimethylol propane phosphate (TMPP) that leaked from the engine into the cockpit, disorienting the pilot and impairing his sense of judgment.

One problem with this theory was that TOXLINE and MEDLINE searches showed that little was known about the effects of TMPP on any large, living being, including humans. The only research available at that time was on small animals, such as rabbits, mice, and rats. There was nothing showing that it affects the cerebellum, motor pathways to the brain stem, basal ganglia, or the descending pyramid of the cerebral cortex.

Another problem was that no evidence of TMPP was found at the crash scene so the insurance company had only a theory to work with (sound familiar?). But it reasoned that TMPP is a class of bicyclophosphates that are GABA inhibitors and since GABA inhibition affects speech in diseases like Huntington's Disease, the insurance company theorized that it had the same effect on the pilot's behavior. If they could prove their case, it could shift the blame for the accident from human error to a product liability case against a successful and presumably prosperous company.

Garrett's defense attorneys were then faced with the unusual task of combating a theory rather than physical evidence,which is the usual basis for such lawsuits. Enter the service of linguistics. The only accident evidence available was the air-to-ground communications of the pilot between Milwaukee and New Orleans. The pilot's voice sounded okay to the defense lawyers but to verify their suspicions, they called me to analyze the intermittent tape recorded communications from the time the plane taxied down the runway in Milwaukee, as it passed over the control towers in Chicago, Kansas City, and Memphis, and as it approached its final destination. The defense had to combat one theory -- that there was TMPP in the cabin -- with another theory -- that if such happened, the pilot's speech would give evidence of it.

I analyzed the pilot's syntax, word frequency, speech acts, pause fillers, and other evidence of cooperative conversation, theorizing that if he was being overcome with TMPP, these features would be likely to show it. There is pretty good evidence that this happens when people ingest large amounts of alcohol or drugs. But who knows whether the same thing obtains when TMPP enters the system? Neither side had real proof to back up the theories proposed.

In order to determine if there were aberrations in the pilot's syntax, I first needed to study a number of other pilot-to-tower communications to find out what the normal syntax patterns of Plane English, including variability within the optional and obligatory sytactic slots. I found the following: first comes an optional acknowledgment ("okay," "Roger," etc.), followed by an optional self-identification ("Mitsubishi seven two seven," "Mike Alfa," "Six Golf Hotel," etc.), followed by an optional early closing ("out," "okay," etc.), followed by an obligatory subject ("we," "five thousand feet," etc.), followed by an obligatory predicate ("climbing to five thousand," "ready to go," etc.), and, finally,a second closing slot (used when a subject and predicate occur). I found no aberrations from this formula in any of the pilot's communications over the three tower-contact segments of the flight beginning in Milwaukee, going over the Midwest, and ending in New Orleans. His syntax didn't appear to be confused.

The pilot didn't show any loss of language ability in his use of compound sentences either. They remained constant from the beginning of the flight to the end. Nor did he start using shorter sentences.  His words per utterance remained fairly constant throughout the three segments of the flight, averaging 9, 8.27, and 7.75, respectively.

I also examined the pilot's speech acts. He reported facts, such as his location, altitude, and flight course without confusion or hesitation. Failure to do so could have been interpreted by ground control as erratic behavior but no such complaints occurred. He replied to all of ground control's questions, acknowledged all instructions, repeated all information accurately, and even thanked the control tower once. His most telling speech act, which came as the pilot was trying to land, was to correct the tower's error when it misidentified him (I'll come back to this later).

And what about pause fillers,  those "uh," "um," "er" features that most of us use in daily conversation? Did the pilot use them excessively, perhaps suggesting that he was getting sluggish or beginning to be overcome by toxic fumes? There are two types of pause fillers. One is used when speakers are trying to get the attention of a listener or holding onto their turns of talk. The other type occurs when speakers are uncertain or forgetful about how to say something. The latter type might indicate a decrease in cognitive ability. The pilot used five of these but, interestingly, they all came when he was still on the ground in Milwaukee as he prepared to taxi down the runway. His "hey, listen to me" attention grabbing pause fillers all came at the end of the flight, as the pilot was struggling to set the plane down in a torrential rainstorm.

Now for the most obvious feature. Did the pilot begin to slur his speech, especially his fricative, affricate, and interdental sounds, at some time in the flight? If alcohol or drugs can cause this, couldn't TMPP do the same?  If it could, it didn't.  The pilot had to pronounce words like "Mitsubishi," "seven," "that's," "Kansas," "thousand," "the," "taxi," "its," and "Memphis" throughout his flight. No diminution of speech ability is noticeable here.

Finally, I looked at the pilot's conversational cooperation (relevance, informativeness, sincerity, and clarity) to see if it got worse as the flight progressed. Perhaps it would take an air-traffic specialist to testify about whether any of the pilot's communications were less informative than they should have been, but the taped evidence  shows no discomfort or complains by ground control to any of the pilot's statements during any segment of the flight. Throughout, ground personnel treated the pilot's reports of his readiness, distination, movement, and flight level as though they were relevant, informative, sincere, and clear.

Okay, so why did the plane crash? Oddly enough, this wasn't the focus of the trial. It was only to show that TMPP did or did not affect the pilot's judgment. Other aspects were not considered relevant but I'll go there anyway, because this is Language Log, not a trial, and I can do whatever I want here.

At the very end of the flight, the pilot radioed the local control tower in New Orleans that he was on his approach to land.  Just before this, the radio picked up another aircraft, Six Golf Hotel, whose pilot requested permission to abort his landing because of the heavy rainstorm. After this the transmission went:

Pilot: We're on the approach.

Tower: Six Golf Hotel, you're on the approach now?

Pilot: No, Mitsubishi  seven  two seven Mike Alfa. We're on the approach.

Perhaps realizing the mistake, the tower then gave Mike Alfa a weather report and asked him to report Alger when he passed it. Alger is a specified checkpoint in the landing approach. It's difficult to know what happened next but Mike Alfa's response was, "Okay, Mike Alfa," and we never heard from him again. He had already passed Alger by that time and the tape gives no indication that he had ever reported it.

Fighting the elements, being misidentified by the tower, and having already passed the Alger checkpoint, the pilot was pretty busy trying to figure out what to do. As it turns out, he was far off course and crashed on the shore of Lake Ponchartrain. It would appear that the pilot was, indeed, confused and disoriented, but there seems to be no language evidence  suggesting that this condition was caused by ingesting TMPP. And if the pilot was confused, ground control seemed to be just as confused. Note that it was the pilot who corrected the tower's misidentification, hardly evidence of cognitive impairment.

The insurance company did not prevail in this Plane English  trial.

> Posted by Roger Shuy at November 18, 2006 02:17 PM