10 April 2014
With so many responding to media invitations to form subjective opinions as to whether Oscar Pistorius’ emotion is genuine, are we missing factual errors in the reporting of what he is actually saying? Could scientific analysis help here?
Pistorius’ testimony is reported everywhere as including the words ‘She wasn’t breathing’, but close analysis reveals a different interpretation. Listen to the audio and view a spectrogram.
Here’s a clip from Pistorius’ testimony
To hear it in context, go to http://www.bbc.com/news/world-africa-26939443
Now here’s the section of interest.
Careful listening makes it fairly clear he is not actually saying ‘She wasn’t breathing’ (which in any case could be seen as a strange comment from someone who had testified in his bail application that, when he found her, ‘Reeva was slumped over but alive’ – and, as pointed out by Ros Godlovitch Chappell, still relies strongly on his assertion that the victim was still alive when he found her).
I think he is saying ‘She was everything’. That interpretation is supported by acoustic analysis. Here is a spectrogram showing very clear acoustic evidence of the /v/ in ‘everything’ (in the section highlighted in green). This is quite different from how a spectrogram would look if he had said ‘wasn’t breathing’.
You might like to listen again and follow both phrases to see which seems more accurate.
She wasn’t breathing — She was everything
Is this the only error we are missing in these transcripts?
It seems likely that careful checking might reveal other errors in the transcripts being circulated by the media – and perhaps used by the judge in the case to help remember the testimony.
Do errors like this matter?
Arguably, the difference between these two phrases is a minor issue in the context of this trial overall – though equally arguably the correct version could give important evidence as to Pistorius’ state of mind and a useful check for consistency with other parts of his testimony.
The real significance however of this error, in terms of forensic transcription, is the insight it can give into how easy it is to make an error in transcription, and how unlikely it is that the error will be detected simply via people listening to the recording.