Tuesday 15 June 2010


There’s one Zulu word we have all become familiar with over the last week or two: vuvuzela, the blaring plastic horn blown by South African football supporters.
Turn on your TV right now to whichever sports channel is showing the England's soccer game against the USA in the World Cup in South Africa. Turn the sound up. Why does it sound as if several dozen propeller-drived airplanes have started up their engines in the stadium? Has someone dropped one of the commentator's mikes into a huge beehive? No. It's just that South Africans love to bring annoying three-foot-long one-note plastic trumpets to every game and blow them continuously. (They all seem to be tuned roughly to A below middle C.) [Geoff Pullum in Language Log].

Harry Campbell draws my attention to the claim reported in the media.
A vuvuzela is tuned - to use the term loosely - at the B flat below middle C, and has a similar frequency to speech tones, says Trevor Cox, president of the Institute of Acoustics. This makes it particularly tricky for broadcasters to tune out, as to do so would dampen the commentators' voices - and not in a good way.
"I'm looking at its wave patterns and there are at least six very strong harmonics in there. It would sound really horrible to notch these out - if one coincides with the vowel sound e, you won't be able to hear the -es in the commentary. It would sound unnatural."

Er, well, yes. I’m not going to start messing around with spectrograms myself, but it is obvious that the vuvuzela racket covers quite a range of frequencies. It is not a ‘pure tone’ like a whistle, which would have a single frequency that one might filter out.
The fundamental note is reportedly “roughly [at] A below middle C” or “at the B flat below middle C” (accounts vary, as you see). Let’s say 230 Hz or thereabouts. That would imply harmonics (overtones) at integer multiples of that frequency: 460 Hz, 690 Hz, 920 Hz etc. This coincides with the range of vowel first formants (not just e, all vowels), which is approximately 250–1000 Hz. So if you filter out the vuvuzela noise, you’re going to lose information that could be important in keeping speech intelligible.
Nevertheless, here is a plugin that claims to do that. Feel free to try it.
I think a better way would be for broadcasters to separate the commentary feed from the crowd noise feed, and to suppress all or most of the latter. Except for people who like the noise of the vuvuzela.


  1. "I think a better way would be for broadcasters to separate the commentary feed from the crowd noise feed, and to suppress all or most of the latter."

    Now, now. Don't start using logic. That's hardly fair on the journos.

    As for the tuning I assumed that as a musician Pullum had judged it by ear, but perhaps I'm wrong.

  2. I'm surprised you don't have anything to say about the pronunciation of the word: FACE or DRESS?

    I should expect FACE, which is also Wikipedia's choice. However I usually hear DRESS from the match commentators on US television (who are mostly British).

    And count me as a vuvuzela fan, on television at least! They remind me of the local color provided by conch shells in cricket matches in the Caribbean.

  3. God forbid the commentators use mics that only pick up their voices, then process-out the horns. But what would be left? Commentary, white noise from the crows, and whistles. Sounds a little boring, I like the ominous swarm-of-killer-bees sound, it really gets your heart going as the tension mounts in a game.

  4. Today's Guardian has a video demonstration from a vuvuzela orchestra member.

  5. Surely the crowd noise and the commentators' voices will already be miked separately. And given that, for just this reason, commentators traditionally use lip mikes, the sort held close to the mouth, and sit in a (somewhat) soundproofed booth, I find it hard to understand what the problem is here. The noise would have to be VERY loud to be impossible to deal with. It must come down to an aesthetic judgment by the people mixing the sound -- like the question of how loud the applause and laughter should be in a comedy programme, or bursts of music on a specch channel like Radio 4, where many people agree with the character in Stan Freburg's skit on the banana boat song: "too loud man, too piercing".

    In any case, the Centre for Digital Music's devuvuzelator seems to refute the BBC's acoustics expert pretty effectively.

    As for the word itself, Samora Ntsebeza's pronunciation on that YouTube demonstration make a pretty clear case for DRESS rather than FACE.

  6. Yes. Zulu /e/ has two allophones: [e] if the following vowel is high (= /i/ or /u/), [ɛ] otherwise. [vuvuˈzɛːla], tones unknown.

  7. I can barely hear any difference between when the "devuvzelator" is on and when it's off (at least in the youtube video -- I haven't bothered to download the plugin myself) -- but I think maybe it's because I myself am developing a partial deafness to frequencies in a narrow band centred around about 230 Hz and one around 460 Hz... :-)

  8. Here's a music score to enable you to play along with the crowd.


Note: only a member of this blog may post a comment.