Things have been rather quiet with the Blog recently. I’ve been in Dubai for a week looking at opportunities to take dBx Acoustics global, and then had my last workshops for my MBA at Manchester Business School. In between, I’ve been out on site at Strawberry Field (yes, in Liverpool!), playing “hunt the noise source” in the Midlands, and beginning the design work for my new project at Truro College.
In the midst of this, though, I’ve been avidly following the trial of Oscar Pistorius. The last post I did on the acoustics of the Oscar Pistorius trial got more hits in a day than I usually get in a month – thank you for that! Clearly I wasn’t the only person out there wondering about some of the evidence presented with regards to what people could and couldn’t hear, and waiting for the day when either the prosecution or the defence would bring out their acoustics expert to answer the questions once and for all.
One important update to the basic analysis I presented last time – it appears there was a window in the toilet, and it was open. However – it also seems that there were houses in between the witness house and the scene, so the end result balances out the same. I think it’s highly unlikely that these witnesses would have been woken by screams – and there certainly are houses much closer which would have heard more.
It appears that some tests were done, on a firing range, at night, to record the sound of a cricket bat against a door, and of a gun being fired. There’s some good and some bad in this science; if you’re going to measure sounds then it’s appropriate to do it in the absence of any other background noise, so at night is reasonable. However, this seems a little redundant if you’re not then going to do a frequency analysis of the sounds, or also measure and correct for any background noise that may be present.
Also, measuring at 90 and 180m seemed lacking in any real use; the inverse square law dictates how much the sound would decrease by over a doubling of distance. But once again, it seems nobody was measuring absolute noise levels. Would you need to? Not necessarily – it depends if you’re trying to demonstrate that the two sounds could be confused, or if you’re trying to demonstrate what could actually be heard over a distance.
There is a good video presented by Alexander Jason where he compares gunshots to cricket bat sounds over 180m and based on this, although I still think I can tell the difference between the two sounds, it’s plausible that if you didn’t know what you were listening for and you were awoken by them in the night – especially if you then were told next morning there had been a shooting – that you could confuse the two.
I’d also suggest that whilst there were good reasons for measuring on a firing range, this could in itself be misleading – depending on what the range is like, it could introduce reverberance (e.g. reflections from concrete bunker walls) which affect the sound – better to make such recordings in a free field free from as many reflections as can practically be achieved. Finally – playing the sounds back in the acoustic of the courtroom means that while sounds can be compared, they are not presented as a true reflection of what happened during the test – perhaps better to listen through good quality headphones?
As Alexander Jason points out in his notes, the point of the test he carried out was simply to see whether over 180m a gunshot and bang on the door could possibly sound the same. However, for the purposes of evidence, it is surprising that the acoustic analysis for the court wasn’t taken further.
As I noted in the previous blog, the only way to know for sure what was heard by neighbours would be to either replicate the sounds on site (not possible due to subsequent development), or to create a computer model. This is relatively straightforward, and can give both a quantified prediction of the sound levels received at neighbouring houses (useful in conjunction with a background noise survey to see what actually could have been audible), as well as an auralisation based on audio recordings to replicate (over headphones) what might have been heard.
Why not do this? It would be relatively quick and easy, if done by a qualified acoustician. It’s accurate, it’s based on the laws of physics, and the inputs to such a model can all be traced and verified. A large part of the prosecution’s case was based on the question of “who heard what” – but the defence has missed a great chance to explore and challenge this.
We have been told that when he screams, Oscar Pistorius sounds like a woman. As you’ll remember from the original blog on the subject, I’m unconvinced by this argument. Apparently they have done some ‘decibel tests’ – a phrase which already makes me wonder if these tests may have been carried out by a car mechanic, or perhaps a lifeguard – but as yet, no recordings or evidence have been presented in court. Ultimately, though, male and female vocal physiology are very different, and I’m not the only one who thinks this line of defence is very unlikely.
One of the most disturbing things about the way the acoustic evidence has been treated in the trial is the use of a non-expert expert, Roger Dixon. Its not clear that Mr Dixon has any specialist acoustics knowledge or training – in fact, he’s a geologist by profession, not a profession known for requiring a grasp of physics. This knowledge did lead to a slightly silly conversation on Twitter, but I think it makes an important point…
Putting aside that this “expert” was also expected to know about ballistics and blood spatter, and the more general question as to why the defence though this might be a good way to handle things, I do wonder why an acoustics expert couldn’t be found, even if they had to be flown in. We’re rare, but not THAT rare. And as experts go, we are good value for money. I think a good acoustic assessment with some proper testing and modelling, could make all the difference to one of the sides in this trial. Unfortunately, it doesn’t look like we’re going to see that happen.