Lombard speech
Let’s do a little test…
First, let’s take care of your audio settings:
- put on your headphones/ear buds/speakers
- turn your volume way down
- play this sound…
SOME WHITE NOISE…
- …and adjust your volume until it’s at a loud but still comfortable level.
OK, now we’ll do a short reading test:
- play the video below
- you’ll see a counter counting down from 3…
- …and then it will present a simple sentence on screen
- your task is simply to read out the sentence aloud.
Ready? Go!
Why thank you! OK, now let’s do this again. Make sure to keep wearing your headphones, play the next video, and read out the sentence aloud.
Surprise!
What’s going on?
Perhaps you noticed your voice sounding somewhat different the second time, when there was loud babble from 16 other talkers playing in your ears, compared to the first time (in quiet). This phenomenon is called Lombard speech (or: Lombard effect; Lombard reflex). It’s the type of speech people produce when speaking in noise.
But perhaps you didn’t quite hear yourself all too well because it’s hard to listen to your own voice when there’s other sounds around. So here’s two clips from a male speaker of British English (and a rather posh one, if I may say so…) giving you some really useful dietary advice. The first is from when he was speaking in quiet: this is called ‘plain speech’. The second clip is a recording of the same sentence but this time the talker heard loud noise over headphones, speaking in noise: ‘Lombard speech’.
-
PLAIN SPEECH
-
LOMBARD SPEECH
retrieved from the Acted clear speech corpus
What is Lombard speech?
In the clips above, you can clearly hear the difference between ‘plain speech’ and ‘Lombard speech’. And this speaker is not the odd-one-out: many vocal learning species, like dolphins, seals, and birds adjust their vocalizations when encountering noise.
In humans, Lombard speech sounds louder, higher pitched, is a little slower, with more pronounced higher frequencies, and clearer vowels. In our own research, we demonstrated that Lombard speech is also more rhythmic, having a stronger ‘beat’ to it compared to plain speech (see refs below).
Lombard speech rulez
Speech perception studies have demonstrated that these ‘vocal adjustments’ people make when speaking in noise actually have a purpose: they make you more intelligible! When you take the plain and Lombard clips above, scale their intensities to be exactly the same, and then mix them with loud babble, this is what you get:
-
PLAIN SPEECH in BABBLE (SNR = -6 dB)
-
LOMBARD SPEECH in BABBLE (SNR = -6 dB)
You may experience that it’s easier to pick out the male target talker from the babble in the second (Lombard) clip than in the first (plain) clip. Apparently, ‘speaking up’ actually helps!
Why is this important?
Lombard speech is more intelligible in noise than plain speech. This means that speech researchers can ‘borrow’ acoustic aspects of Lombard speech to boost speech intelligibility, for instance in hearing aids. So next time you wanna make sure your message comes across in that busy bar, you’d better boost your F0, raise your spectral tilt, and increase your vowel dispersion; got it?!
Relevant papers
(2018). Talkers produce more pronounced amplitude modulations when speaking in noise. The Journal of the Acoustical Society of America, 143(2), EL121-EL126, doi:10.1121/1.5024404.
(2020). Enhanced amplitude modulations contribute to the Lombard intelligibility benefit: Evidence from the Nijmegen Corpus of Lombard Speech. The Journal of the Acoustical Society of America, 147(2), 721-730, doi:10.1121/10.0000646.