“Phonetic change in an Antarctic winter”
The Journal of the Acoustical Society of America 146, 3327 (2019); https://doi.org/10.1121/1.5130709
Authors: Jonathan Harrington, Michele Gubian, Mary Stevens, and Florian Schiel
How does spoken accent develop? In this episode, we speak with Dr. Jonathan Harrington of the Institute of Phonetics and Speech Processing at Ludwig-Maximilians-University of Munich, about the answer to this question. Dr. Harrington studied phonetic changes to individuals’ speech after a prolonged stay in Antarctica. Learn about variables that make up a person’s accent, the kind of stimuli that cause shifts or changes in a person’s accents, as well as Dr. Harrington’s methods of research for studying accent.
Read more from The Journal of the Acoustical Society of America.
Learn more about Acoustical Society of America Publications.
Music Credit: Min 2019 by minwbu from Pixabay. https://pixabay.com/?utm_source=link-attribution&utm_medium=referral&utm_campaign=music&utm_content=1022
Malené Walters 0:06
Welcome to Across Acoustics, the official podcast of the Acoustical Society of America’s publications office. On this podcast, we will highlight office research from our four publications: The Journal of the Acoustical Society of America, also known as JASA, JASA Express letters, Proceedings of Meetings on Acoustics, also known as POMA, and Acoustics Today. I'm your host, Malené Walters, Publications Business Manager of the ASA. Today, our guest is Dr. Jonathan Harrington, Professor of Phonetics and Speech Processing at the Institute for Phonetics and Speech Processing, Ludwig-Maximilians-University of Munich, Germany. We will be discussing the paper phonetic change in an ad Arctic winter published in the November 2019 issue of the Journal of the Acoustical Society of America, co-authored by Dr. Harrington. Hello, Dr. Harrington, welcome to the podcast. How are you doing today?
Jonathan Harrington 1:05
Oh, thank you. Thank you very much.
Malené Walters 1:09
We start with our guests, giving our listeners a brief history of their background. Can you tell us a bit about yourself and your background?
Jonathan Harrington 1:16
Yeah, so as you said, I'm a professor of genetics at the University of Munich. I started out on this academic path as a PhD student of linguistics at the University of Cambridge in England. And I specialize there then in phonetics and speech science. Then I worked as a postdoc at Edinburgh University when they started up the Center for Speech Technology Research. That was in 1984, and we were working on automatic speech recognition. And then after that, in 1989, I went to Macquarie University, Sydney, Australia, and directed the Speech, Hearing, and Language Research Center, and was also involved in cognitive science as well there. Yeah, I was there until 2001. Then we left Australia, me and my family, took up a professorship at the University of Kiel, in North Germany from 2002 until six. And after that I have while I've been here in Munich since then, so that's a sort of brief summary of where I've been at least here.
Malené Walters 2:57
Wow, very good. You've traveled a lot for work.
Jonathan Harrington 3:02
We have, yes.
Malené Walters 3:04
Now let's jump into your research. I wanted to start with letting our listeners know that this research was funded by a European Research Council Advanced Grant, is that correct?
Jonathan Harrington 3:14
That's correct. Yes, that's right.
Malené Walters 3:16
Wonderful. And I know from researching that type of grant is very competitive and it expects support investigator driven research.
Jonathan Harrington 3:24
That's right. Yes, that's right. So it's basically five years worth of funding, basic research as well, which is actually very, very good, and it's generously funded. And anyone from around the world can apply for one of these things. So it just then you have to be based in Europe. That's all if you decide to take up the grant.
Malené Walters 3:50
I see. Very good, very good. Okay. Now, to move into your paper, you discuss an acoustic analysis made of the speech characteristics of individuals recorded before and during a prolonged stay in Anarctica. Could you explain why you and your co-authors began the study of these speech characteristics?
Jonathan Harrington 4:14
Yes. So it was actually something which we have been thinking about for some time even before we started the ERC project. And what we wanted to do with this is to work out the conditions under which spoken accent starts, how it actually gets going. So we know of course, that when there was migration to America with the Mayflower and settlement in Australia and New Zealand and so on, that after a while they develop their own characteristic way of speaking, i.e. their own spoken accent. And so that must have something to do with then this isolation. So that's the first thing. And interaction as well, because of course, with the Mayflower folks or even of course, in Australia as well, people had to cooperate with each other, and talk to each other. Because if they didn't, well, then they wouldn't survive. So the issue is, then how can you replicate isolation and interaction in modern times? And so that's then when, when we thought of Antarctica because they these recordings took place during an Antarctic winter. And while of course, there is still some internet contact with the rest of the world, although they said that, because of course, this is all via satellite phone that it is fairly sparing so, but the main thing is, then that you have a group of people isolated, because in the winter months there, there is no way in and no way out. So we’re actually stuck there for three, four or five months period, whether you like it or not. And so that's the first thing. And of course, then they have to cooperate with each other, and they have to interact with each other. Because, as in, you know, first settlement times. If you don't, then you'll die. So it's those two conditions, really, which are very important. And so now, of course, one thing is, is that the timescale of being in Antarctica is very, very short. I mean, it's only three, four or five, six months or something like that. And so, I mean, that, you know, doesn't compare with folks went in the Mayflower. I mean, you know, they were actually stuck there for the rest of their lives if they survived that. So. So what that means then, of course, is that the magnitude of x and change which we expect from Antarctica is very, very small in comparison, for example, with development of an American accent, which of course has evolved over hundreds of years. Yeah.
Malené Walters 8:11
I see. Interesting. Now, you touched on this a bit. But could you kind of go into detail of how an individual acquires specific accents, could you describe the variables that make up an individual's spoken accent?
Jonathan Harrington 8:30
So most of this it is true is set, of course, first of all, in childhood, so the accent with which you speak in your earliest years is your caregivers’. So it's, it'll be you know, that of your parents, that of your family, i.e. those to whom you are exposed most of the time. That of course, means then that when a child goes to school, that the accent will change. And if of course, then for example, the parents have moved in some new dialect area, then the child will speak in that dialect, and it will no longer speak with the accent of its parents. So for example, I mean, I mean, when we went to Australia, and when my daughter had just been born, and she went to an Australian school when she was four, and I speak with a British English accent, my wife with a more of an Irish accent, and she spoke with an Australian accent, which she still has. That is because in the daytime, she interacted with Australian children. So that's the first thing and then but accent though in adulthood is malleable. So that if you move region into different dialect region, then whether you like it or not, you will pick up some of the characteristics of that dialect. Okay. Now, I mean, the size of that shift in adults is smaller than it is for children who, you know, start out in school, but it is measurable. Yeah. Mostly comes down to who you talk to and how often. Yeah, so I mean, if, of course, you move to Canada, for example, and never ever speak to Canadians, well, then you're not going to acquire a Canadian accent.
Malené Walters 11:20
Ah, okay. That's interesting. Now, to go into your research a little more, your research measures changes to speech as a result of spending time in Antarctica. And you chose a small group of individuals who would be spending several months in an Antarctic winter, as you mentioned. And this group was part of the British Antarctic Survey. Could you describe the parameters and focus of the study?
Jonathan Harrington 11:48
What we did, first of all, is they have a training session, basically, the British Antarctic Survey has a training session, which is then where all of those who are due to go to Antarctica meet up for the first time. And it was in that training session then where I explained that I wanted to record their speech and asked for volunteers. And what we wanted to do and what was very important, indeed, is that we should record from them before they went to Antarctica and then in Antarctica itself, so that we could then measure any change then in Antarctica relative then to this baseline, as it were. Yeah. So that was the first thing and then it was also important that the Antartic Winterers did not know each other beforehand, which they didn't, so that was also fine. And we recorded from them then, either actually on that training day, or if that wasn't possible, then we went to their homes, and we recorded from them there. Now they, when they are in Antarctica, they are very, very busy with all sorts of things, with all sorts of scientific things, which are yeah. And I mean, I should add to that the Winterers weren't only scientists, they were support staff as well. So you have for example, medical staff, and you have engineers, you have plumber, for example. And so, you know, I mean, you know, basically so that they can survive, therefore that you know, time so… and their accents were very, very varied. So, one or two of them spoke a bit like I do with a British English accent, but then we had some from Scotland. We had one person from the United States, I think it was from the west coast. We had a first language German speaker and a first language Icelandic speaker as well. And that's all fine because of course, that is exactly what we wanted was this sort of mixture of lots of accent types because, of course, if accents had all been the same, well, then it will be far more difficult to expect and measure any action change. Now, as I said, they have limited time. So ideally, of course, we would have liked to have recorded from them in conversations, but that just wasn't possible. So we chose then sets of words, which they just read one at a time from a computer screen. And these words, contained vowels, where we thought, if there is going to be any change, then it might be here. So I should say that as far as that was concerned, it was a bit of a fishing expedition, which was the we just weren't sure whether we were going to find any change whatsoever. In fact, I was expecting to find no change. Yeah, yeah. Well, that's it really yeah. I mean, it's this is this is it you see with the, with European Research Council, once then that you do high-risk research? So this was an example of high-risk research? Well, we might well have found absolutely nothing. Yeah. Okay. So that's what they did. And they read these words, then, before they went to Antarctica, and then in Antarctica, every month or so, and the recording itself took only 15 minutes.
Jonathan Harrington 16:49
And we were looking then for changes to vowels. And we chose vowels as well, because vowels change very, very slowly, you can actually see you incremental changes, let's say from an oo to an oooo, which, you know, that whereas, whereas with consonants I mean, if an if an “s,” for example, if a “sir” changes into a “shir,” which you do have actually in North America, so that Columbus, Ohio, for example, rather than saying “street” you’ll hear “street” with a “sh.” But that change from “s” to “sh,” is often not as gradual, right? So it's a sort of jump from “es” to “esh.” And since, of course, we had only four or five months or so, and since we were expecting incremental changes, and that's one of the reasons then as well, why we focused on vowels.
Malené Walters 18:07
I see. And after the recordings were complete, how was the data analyzed, and how did you apply the agent-base computational model to the recordings?
Jonathan Harrington 18:21
Yeah. So the VAT. So, the analysis was an acoustic comparison of the vowels in Antarctica, relative then to the baseline and here then we measure the resonances, the resonance frequencies, basically, of the vowels which are also called formats. So the reason then why, when I say for example, “ee” and then “oo,” the reason then why you and I hear “ee,” rather than “oo” is because the formant frequencies are markedly different for an “ee” compared with an “oo.” So we were, what we wanted to see then if there was any evidence then of formant frequency changes. So that was that comparison. And what we found was the vowels which we were looking at. Most of all, were the “ee” vowel and happy and city, ie the one which is spelt with a Y, then the “oo” sounds in words like food and the “oh” sound in window. And what we found was that there was no change in the “ee,” that there was a very slight shift then in the “oo” sound so that so that the second formant frequency increased for the “oo” and for the “oh,” which means then that they were produced with a very, very slightly more forward tongue position than before Antarctica. So I'll exaggerate what that sounds like. So that if “oo” then has a low second form of frequency it sounds like “oo” and if it has a high second form and frequency it sounds like “ooo.” So from “oo” to “ooo” now, the shift though, from was nothing like as large as from “oo” to “ooo,” but it was in that direction. So, that was one of the changes anyway. And then we found as well that there was convergence, at least for that “oo” and “oh,” so that the what that means then is that the Winterers sounded just that little bit more like each other, I mean microscopically so, when they were in Antarctica than before Antarctica. So, that was that bit. Then for the agent-based model, this is a computational model, where we represent each of the Winterers with virtual agent, and we supply this agent then with a very basic vocabulary. So that in this case, then the vocabulary was those words, which they heard read out, we tell the agent, whether the word has an “ee” vowel or an “oh” vowel or an “oo” vowel, and most importantly then is that we give each of the agents the recorded speech data before they went to Antarctica, okay. So, then you have these agents there with this speech data vocabulary, which is very small, and then we get these agents to talk to each other. So that one of the agents transmits a word and the vowel and then recorded speech signal to an agent listener. And then there are various criteria as well for whether or not that listener memorizes what it has just heard. And we repeat this then, for example, I think this was 50,000 times. So, so, we you know, randomly selected then an agent has in this in this way. And so, then, what we did was we wanted see, first of all, then if, after then this computational interaction, there was any acoustic change. And most importantly, then, was this computational change in the same direction as change which we had found in Antarctica. And to a certain extent, it was so that what we found them was that the computational model predicted that there should be no change to “ee.” It exaggerated vastly the degree of change in “oo” in words like “food,” so output then of the computational model was a much more fronted “oo: than had actually taking place in Antarctica and, and it's got the shift then to “oh,” more or less right/ So what that means is then is that we found then that there had been some shift to Winterers’ speech in Antarctica. And to a certain extent, we could predict this with our computational model.
Malené Walters 25:24
Wow, that's amazing. Really, really, really amazing. That, you know, what the Winterers, the vowel change in the Winterers, kind of, you know, was, you know, what the model showed as well.
Jonathan Harrington 25:40
Well, yes, that's right. I mean, I was not expecting, first of all to find any change, and then was pleasantly surprised as well, when not only then, when we did find change, but that we could also model it computationally. But of course, I should say that this is only based on 11 speakers. Yeah. So I mean, it might well have been the case, if we had had a different set of speakers that, that it wouldn't have worked out. So yeah, I think that that actually, I mean, that is that is a limitation of this study which I would like to emphasize here that yes, it was a very nice finding, but it is only for 11 speakers.
Malené Walters 26:41
Yeah, that makes sense. And lastly, are there any other details you would like the listeners to know about the research?
Jonathan Harrington 26:52
Details itself? Well, I can say that we would like to replicate this, of course, with a larger group of speakers. But it's very difficult to do in that way. You know, I mean, to find your speakers, and of course, I mean, if this had been in the COVID times, then I mean, we just wouldn't have been able to do it at all. I guess that this has implications, for example, for I mean, if we should send astronauts on a, on a mission to Mars, where they're away from home and isolated for 18-- well actually, for you know, long, long period of time-- then the implications are that that they would actually start to develop a Martian accent. That needs some empirical investigation. But that is actually what the implication is anyway, the isolation with interaction leads then to accents like American English, New Zealand English, for example. And what we want to work out then is if you know something about the speakers that go into the mixture, i.e., for example, those on the Mayflower, or those who, you know, first settled in New Zealand, can you then predict the direction in which the accent will develop? And that's, that's what we have tried to do with this limited study.
Malené Walters 29:02
Wow, very interesting. I could just listen to talk about linguistics and vowels forever. This is really, really interesting.
Jonathan Harrington 29:11
It's a fascinating subject, isn't it? Yes, very good. I mean, and you know, that that there are studies as well which you know, show that accent development is microscopic. So there’s research for example, by Katie Drager from Hawaii. And she showed for example, that a group of schoolchildren who regularly meet for lunch every day over a long period of time, begin to develop characteristics which they have and which aren't shared then by other members of the same school. Yeah. Well, that's right, you see, and it's, of course, because we imitate each other, and we imitate each other as well in in a very, very subtle, fine way that we can't even hear. Right? But of course, then the issue is, if you can't hear it, or if you're not aware of it, how does this imitation come about? Because you'd have thought, you know, I mean, if I wish to imitate another person, that I you know, have to be able to identify aspects of that person's voice or accent and then copy those features, but that isn't though how this limitation works. I mean, even if we, for example, converse over, over, let's say, I mean, as we are doing now, over an hour or so, at the end of that conversation, I will in certain ways actually sound very, very, very slightly more like you and vice versa. And there are the experiments which actually, you know, show that, so, how does this happen? That is a fascinating question. And it is the most important thing that to, you know, explain then how new accents get started.
Malené Walters 32:02
That is, that's amazing. Yeah, honestly, amazing. I would, I would love to have a British accent. So, in a minute way, I love that.
Well, thank you so much, Dr. Harrington. This has been very enlightening. I thoroughly enjoyed learning more about your research, and I'm sure our listeners will as well.
Jonathan Harrington 32:29
Well, thank you so much for asking me along. And I've, I've really enjoyed talking to you.
Malené Walters 32:35
Same here. Thank you so much.
Thank you for tuning in to Across Acoustics. If you would like to hear more interviews from our authors about the research, please subscribe and find us on your preferred podcast platform.