The Process

Building an accurate and authentic transcript from a CVR, and then correctly interpreting the results, can be a painstakingly slow and tedious task. Specific protocols are followed in order to prevent dubious or unclear passages from being transcribed as definitive phrases. Dedicated audio equipment able to filter out extraneous noise and enhance other sounds must be employed in order to give reliable results where recordings give poorer-quality data.

Once transcription is complete words and phrases must be read, first in isolation, and then from within a broader context, in order to either support or contradict any initial interpretation. A common trap is to ignore this broader context and assume that the first interpretation, made without contextual reference, is the correct and authentic representation. It is almost inevitable that this approach will lead to an erroneous conclusion. In fact, it is normal when constructing a transcript to state at the beginning of the transcript itself: Never draw conclusions, nor ascribe meaning to a phrase, using the CVR as the sole source of information.

To illustrate this, imagine recording a meeting with five people conducting two simultaneous conversations. This scenario is not unlike the environmental situation on the flight deck of a large 1970s-era aircraft with flight engineers and perhaps navigators present as well as the pilots. Then imagine asking five untrained people working independently to play back the tape on their own personal equipment and transcribe their results. They will tend to write what they think they hear, or worse, what they want to hear, rather than what is actually said. Moreover it is likely there will be five widely differing results, any or all of which may be misleading. Without the services of an audio expert it is quite easy, for instance, to blend two simultaneous conversations into one and thus reach a completely erroneous result.

Despite the poor quality, the lay person could be persuaded by eloquent argument that any (or indeed all) of these “transcripts” are correct, even if they contradict one another. This is a good example of the Rashomon Effect 2, in which one person’s truth conflicts with another person’s equally convincing truth, and is a well-known phenomenon in psychology.

For this reason, properly trained CVR analysts will do only two things when constructing a transcript: they will attempt to determine exactly what words were spoken, and will attempt to attribute those words to a particular person. It is absolutely critical that transcribers do not attempt to determine why something was said, nor attempt to interpret or infer any meaning, during the transcription process. That will come later, when the transcribed words are analysed and interpreted alongside any corroborating information.

Good rules and protocols for CVR transcription and interpretation are essential for forming strong defences against flawed results. They facilitate the production of a true and authentic record, and promote an accurate understanding of the events in question. However even with these precautions in place, highly experienced air accident investigators are still not immune from hearing the things they want - or expect - to hear on a CVR. Nor are they immune from ascribing meaning to a phrase based only on supposition, or perhaps subtle pressures from vested interests.

They must therefore always remain vigilant, and flight recorder specialists are trained to understand this. In order to assist investigators further, it is axiomatic that transcribers and interpreters must be available that are in current flying practice with the same airline, and on the same type of aircraft, as was involved in the accident.

Moreover, they must not work alone or without the assistance of trained audio specialists. If there is significant disagreement among transcribers, or if none are able to determine what was said in relation to a particular passage, that particular passage must be deemed to be unintelligible, or not sufficiently intelligible to give a reliable result. In addition, anyone working on an investigation or a transcription should remain free of preconceived opinions or ideas as to possible causes of the accident. This will guard against the brain being fooled into hearing something that supports a pre-conceived theory, rather than what might actually have been said.

The initial CVR Group assembled by Mr R Chippindale travelled to Washington and underwent specialised training, adhered strictly to that training, and produced a single, handwritten transcript. For their efforts and professionalism they were roundly praised by NTSB and FBI experts.

The transcript actually published in the formal accident report was significantly different to the version produced by the CVR Group. Chippindale made 55 changes to the transcript without consultation with the CVR Group and in direct contravention of accepted protocols. Chippindale’s actions were inexplicable - ‘at best, an extraordinarily non-standard performance; at worst, a highly improper and prejudicial act.’

The complete failure of the accident-investigation system to identify and prevent unprofessional conduct like this was what Justice Mahon later described as “culpability of the organisational system” (Mahon, 1981).