 |
|
 |
| Neural correlates of non-native accent perception |
| by Patricia Bestelmeyer, D. Robert Ladd and Pascal Belin |
| Further Information |
| SUMMARY |
This paper reports a preliminary study of what goes on in the brain when we listen to speech in different varieties of our own language. The study was a cooperative project between researchers at the University of Glasgow and the University of Edinburgh.
The paper was presented at Neuroscience 2010 in San Diego on 14 November. Because of press conference coverage, it has attracted a certain amount of interest. Since the paper presented at the conference goes into a lot of technical detail about both the brain anatomy and the experimental techniques we used, we have posted the following less technical description of our work for interested readers. |
|
| The "Temporal Voice Area" |
| Work by the Glasgow team (VNL laboratory) beginning in the early 2000s has located an area in the brain (the Temporal Voice Area or TVA) that seems to specialise in processing characteristics of individual voices. The TVA is close to some of the areas that are known to specialise in producing and understanding speech, and to some of the areas involved in hearing generally, but it is clearly separate from them. |
|
 |
| The location of the Temporal Voice Area. This shows the TVA in the right hemisphere of the brain, but there is a comparable area in the left as well, where many language-related areas are found. |
|
| A lot of the Glasgow team's work has concentrated on non-linguistic vocalisations - sounds like coughs, groans, screams and laughs - and shows that the TVA is especially involved in processing these sounds. The Glasgow team have also concentrated on analysing some of the purely physical features that make it possible to identify individual voices - things like overall pitch, and the resonances of the vocal tract. These things are clearly different from what is involved in understanding ordinary speech, and the TVA seems to be involved in paying attention to them. |
|
| Does the TVA pay attention to people's accent? |
| Unlike a lot of the Glasgow team's previous research, this study was designed to see what happens when language and voice are not so clearly separate, namely in an individual's accent. Accent is definitely a characteristic of individual voices, but to know what kind of accent somebody has, you have to know something about what they're saying. You can't identify accents in a language you don't know. We predicted that an individual's accent would be processed as if it were a feature of their voice, even though recognising the accent depends on understanding the speech at the same time. More specifically, we predicted that the TVA would be involved in recognising accents. |
|
| The Brain Imaging Study |
| To test our prediction, we scanned people as they listened to a long series of PIN numbers (like "two-four-five-eight"). The PIN numbers were spoken by nine people with one of three different accents - Scottish, General American, and Standard Southern British (sometimes referred to as Received Pronunciation or BBC English). We used a technique called the adaptation paradigm (see illustration below). Basically, this technique allows us to see which areas of the brain tune out when they encounter a familiar stimulus. When we find a brain area that tunes out like this in response to a particular stimulus, we interpret this to mean that that area is involved in processing that kind of stimulus. We expected that the TVA would tune out when it was confronted with repeated examples of the same accent, and perk up when it heard a different accent. |
|
|
|
| |
 |
| The adaptation paradigm as applied to our study. A series of stimuli (here shown as spectrographic representations of the PIN numbers) is presented to the listener in the scanner. If the accent changes from one stimulus to the next, we predicted that the TVA will remain active - "no adaptation". But if the accent is the same from one stimulus to the next, we predicted a slight drop in the activity of the TVA - "adaptation". The effects for any given pair of stimuli are very small, but over a long experiment with many repetitions the average difference between adaptation and no adaptation can be clearly detected. |
|
| What we found |
It turned out we were only partly right. Because the experiment was carried out at Glasgow University, most of the available volunteers were Scottish undergraduates. So to keep the group of experimental participants as uniform as possible, we only scanned people who had spent all their life in Scotland. This had an unexpected effect on the results.
When our volunteers heard repeated American or Southern British accents, their brains responded roughly in the way we had expected - the TVA seemed to tune out a bit. This suggests that the TVA is involved in processing people's accent, as we predicted. But when they heard repeated Scottish accents, the TVA actually perked up a lot.
Right now, that's as far as we've gone. We know that Scottish listeners respond differently to Scottish accents than to other native accents of English, but we don't really know why. The box below contains some ideas. Meanwhile, we're thinking of ways to explore this question further. But we hope that this kind of research may eventually give us new insights into the prejudice and snobbery that often accompany reactions to other people's accents.
|
|
| What's the explanation? |
One possible way of thinking about what we found is that, for Scottish listeners, the Scottish accent doesn't really count as an accent. Instead, it's part of the background expectations of Scottish listeners, and so listeners tune in to focus on the differences between different Scottish voices. With American and Southern British accents, on the other hand, it may be that Scottish listeners treat the accent itself as one of the identifying characteristics of the speaker's voice, just like pitch or resonance.
Researchers have noticed things like this before: that we respond differently to familiar things than to unfamiliar things. So one possibilty is that the brain just stores up examples of all the voices it hears and constructs a kind of prototype of what words are supposed to sound like. As long as a speaker fits the familiar prototype, the listener concentrates on other features to identify the individual voice. According to this view, the business of constructing prototypes could just be a fairly mechanical process that would be affected by whatever voices a person is exposed to.
But simple familiarity may be only part of the story. Our volunteer listeners were mostly university undergraduates, and if you think about it, you realise that they will have been exposed to lots of Southern British accents during their lifetimes - on radio and TV, in university lecture halls, in recorded messages in lifts and at call centres, and so on. So they'll be familiar with Southern British accents. Yet somehow, they've constructed a prototype that isn't based on all the voices they hear, but only the voices of wha's like us. In other words, the prototypes seem to be built up taking the listener's social identity - in this case, their Scottish identity - into account. Social scientists have thought a lot about the idea of identity, but it's hard to pin down, and hard to imagine how it might be involved in the mechanical construction of prototypes.
To see whether this is the right kind of explanation, we will need to do more detailed studies. For example, we might compare people's social attitudes with their brain responses. Somebody who feels strongly Scottish might have a stronger difference between their brain responses to Scottish accents and their brain responses to other accents, compared to somebody who is happy to consider themselves British or European. We hope to be able to carry out research like this soon. It will be a real opportunity for social science research to influence brain research and vice-versa. |
|
|
 |
 |
 |
|
|
We gratefully acknowledge
the support of : |
 |
 |
 |
 |
 |
 |
 |
 |
|
|
|
