Episode 80 - Matthew Hunt on voice pitch

There's a quick contact warning for this episode, we will be briefly discussing issues connected with domestic violence. Welcome to Lexus, the podcast all about language and linguistics. I'm Lisa Casey and Jackie Glancy. I'm Radarana and I'm Dan Clayton. So on this episode of Lexus, we're really pleased, first of all, to welcome guest presenter Ben Thandon, here's head of English language A level at Peter Simmons College in Winchester. So welcome Ben and we're also really pleased to welcome Dr Matthew Hunt, who's a lecture of social linguistics at the University of Southampton. So welcome Matthew, thanks for joining us. I don't, thanks for having me. And you know each other through kind of work that you've done at the Sixth Form College in Winchester, don't you? So you've kind of met before and you've done stuff with students, Matthew, aren't you? Yeah, I've been lucky enough to come along to Peter Simmons a couple of times in the last few years to talk to A level students about a range of different topics. I think last time I went, I talked about swearing, which is one of the things I've spent far too much time thinking about and trying to write about, as well as a great topic to engage students about. I think the time before that I came and talked about. There was a Scottish guy on the traitors a couple of years ago who said, "I amnt." And I said, "There's one other topic I really love talking about because it's very engaging and they're all aware of the traitors at the very least. So if I had a good time at Peter Simmons and a couple of other student colleges in the area over the last couple of years." So we're going to talk about a couple of papers that you've been working on. One is out now and I think the other one is forthcoming and you've been working quite a lot on voice pitch, haven't you? And a whole range of different ways of which people respond to pitch. So can you explain what linguists mean by voice pitch and how it varies from person to person and how we measure it? Yeah, absolutely. So when linguists talk about voice pitch, we're talking about how high or how low a voice sounds. But technically speaking what we're really talking about is how fast the vocal folds vibrate when we speak. So faster vibration means higher pitch, slower vibration means lower pitch. So it's not about how loud someone is or how expressive they are. It's about the frequency of the sound that they're producing. And pitch varies a lot between individuals, both for biological and social reasons. So biologically pitch is influenced by the size and thickness of your vocal folds, the length of your vocal tract. I like to use sort of musical instrument analogy. When I talk about that, you might think about the size of a tuba versus the size of a quarter and the kind of differences in pitch that those instruments display and also changes in puberty. So in general adult men on average have lower pitched voices than adult women due to changes that occur to their vocal tracks and their vocal folds during puberty. That's an average difference though as I'll come on to in a minute. So there's biological factors that play a role in pitch, but also social factors as well. So there's kind of gendered norms and expectations that I'll get into when I talk about my papers. That explain differences between the average pitch of men and women over and above biology. There's differences in accent, the particular speech community you come from, the culture you come from and even actually the language you speak will affect what is considered to be typical pitch, the professional role that you have. So your pitch might change depending on the job that you do, the kinds of audiences you talk to, and even moment to moment factors like emotion and confidence and stress. So those emotions like fear, one of the ones I'm going to talk about later, often lead vocal folds to tighten and there to be a smaller space which leads to a potentially higher pitch. We usually measure pitch with a value that we call fundamental frequency, which you'll see often written as F0. It's measured in hertz, which just means vibrations per second. So we might say that an adult female speaking voice might average at around 200 hertz. I had an male speaking voice my average around 120, 130 hertz, but these are averages. So you'll find plenty of women with voices much lower than that average or much closer to the male average and you'll find plenty of men with averages much higher. So there's huge amounts of overlap. And in terms of measurement, we take audio recordings of people in an ideal world. You would measure them in a soundproof booth with lots of fancy recording equipment. So we analyze their voices using speech analysis software, things like Pratt, PR, AT, which is the most common one. And that allows us to look at a person's average pitch, how much their pitch rises and falls, and how their pitch might change across different situations. Yeah. So things like a sort of vocal fry, a creaky voice, kind of come on to pitch as well or is that something different? They are subtly different. You might find they somewhat co-vary to an extent. So vocal fry might be associated with a particular type of pitch as well. And they also occur in the vocal folds, but they are distinct things. We would talk about them as different in social linguistic terms as different social linguistic variables that might cluster together in the speech of certain people. So someone who uses vocal fry, I think might also have a slightly lower pitch, but they are distinct things. Yeah. And you mentioned about age. Obviously puberty and the influence of hormones on vocal folds, isn't it? Yeah. So right? Vocal folds. I mean, the vocal track is the longer what we talk about, and I start to finish, and the folds are within the allowance. And so, presumably things like your age, as you get older, you make that maybe affected as well. And I don't know. The sub things to do, are kind of illness and how the actual physical tissue is affected might have an influence on pitch. Yeah, I could totally see that. I think if you're, as you get older, maybe when you're vocal folds relax, I could plausibly see that your voice gets lower. But I think the biggest governing factors are social ones, things like the gendering class. So certainly in women on average, middle class women tend to have a slightly lower pitch than working class women. But gender is the really big thing. And when I say gender, I don't just mean whatever gender you were assigned at birth, it's the kind of gender performance. Exactly, the gender performance, whatever degree to which you align yourself with a masculine or feminine norms, and everything else you do to do in that performance, the clothing you wear, the hairstyle you have, and all that other stuff. Pitch is just another part of that same performance that's same. We tend to talk about stylistic, brick-a-large, interlinguistic. So all the stuff that you put together, which is your stylistic performance, I suppose, voices just one of those things and pitch is one aspect of it. Yeah. No, that's really interesting where I'm putting it, isn't it? I've come across sort of brick-a-large in media theory as well. That idea of kind of assembling stuff from different places and putting it together into sort of an identity. So I guess it works in that way, does it? Yeah, absolutely. I mean, it's kind of acknowledging that none of these, none of the finds that we deploy of social meaning purposes happen in isolation. Yeah. So none of the linguistic variables that we use are interpreted by other people in and of themselves. So I'm interpreted in the context of the kind of person someone already thinks that I am, based on the other things that they have been able to see about me, based on how I'm dressed, based on maybe what gender they think I am, how tall I am, all this other stuff, plus all the other signs in my voice that I'm putting together with that first sign. So it's just about acknowledging, as I'm going to talk about in both of my papers, the importance of context when we're interpreting what people are doing with their voices. Sort of perceptions of pitch and the kind of judgments people make about it are really important. And as he said, particularly in relation to gender. So what sort of judgments are made? Or people make judgments about pitch in the kind of process of making judgments about people's voices more generally, which they do very fast and they often do unconsciously. So when people hear novel voices, they are comparing those voices to all the other voices that they've heard that they have in their memory banks and the kind of social information that they store in their memory banks that is with those voices, the kinds of people they've associated in the past with those kinds of voices. And pitch is one part of that. So pitch can influence how confident and authoritative people think you sound, how credible they think you sound and also your particular emotional states that is maybe more momentary. And in many contexts, lower pitch means authority and higher pitch means warmth, but maybe less seriousness as well. But the really important thing to get across is that these are social stereotypes. They're not intrinsic facts about voices. It's not the case that inherently someone with a lower pitch is more confident about whatever they're saying. This is just the interpretation that has been put on those voices by wider society. And because women tend to have higher pitch voices on average, these judgments tend to then become gendered. We see voices. We see voice pitch through a particularly gendered lens. And women often face what gets called a double bind. So if they display a pitch that people consider to be too high, then they're not taken seriously. But if they then modulate their pitch and make it too low, sometimes this can be seen as ill-inauthentic. So there's a kind of limit within which people expect women to behave and anything outside of that limit, they will get penalised for whether it's above or below it. So pitch is a particularly gendered feature, most linguistic features that social linguists look at will have some kind of gendered stratification. pictures. One of the ones people feel most confident about talking about as well in the general population because one of the most salient features as well. That's really interesting. I mean, it kind of like tallies with other stuff around social linguistics where you kind of hear linguists talk about how certain sounds don't necessarily index something necessarily, but they kind of, for people's perceptions, they will often kind of associate them with social class. So if you glotterize, that sound itself is not kind of upper class or lower class, but the kind of perceptions of it are aren't they? And equally, I guess we've kind of written text. We were talking to you a couple of American researchers fairly recently when they were looking at exclamation marks used in messaging. And that, they talked about the double bind there. That sort of sense in which, you know, women want to kind of appear warm, but at the same time don't want to appear like a pushover in sort of business situations. And the exclamation mark is kind of deployed to suggest kind of warmth and friendliness, but at the same time they've got to be conscious about not appearing kind of too friendly, otherwise they kind of judged really harshly. So seems to be kind of like across all sorts of language features. Absolutely. I mean, that's why we tend to talk about social meaning potential rather than social meanings. It kind of in and of themselves, we're always talking about the potential constellation of meanings that this particular feature might index. And often the indexation is indirect. So it's come via some other means for whatever mean for whatever reason, glottling is associated as a working class feature, but that the index is not a direct relationship between those things. It's kind of mediated by these other meanings associated with glottling toughness, other kind of attributes that then get associated with working class speech as well. And the same as true of gender and pit. Do you think some of those associations you're talking about? I mean, a lot of them coming from the media, do you think or is it more kind of lived experience of people that we're exposed to in our daily lives? Is it a bit of a mix of both or does one have a bigger role, do you think? It's probably a mix of both. I tend to think that presentations in the media are largely reflective of just how people tend to think and behave. And gender is one of those categories that is incredibly salient to people anyway. That said, I'm going to come on later to talk about the context of female football commentators and the way in which the media has played a big role in perpetuating some of the more negative stereotypes about female football commentators. And they definitely have a role to play. And they definitely are a kind of, I was about to call them a leading light, a negative leading light, whatever the, whatever a more negative version of that is in, in riling people up about language. But I think language is one of those things that everyone has an opinion of anyway as a starting point. And it's stuff that we acquire from when we're very, very young. So, you know, children are already acquiring subconsciously the gendered stratification of language. By the way, they see if they have a mummy and a daddy. By the way, their mummy and daddy talk to them. Right? So there's strong evidence that mummies and daddies will talk differently to boys and girls. And they'll use more or less standard features. So we're already acquiring that before we ever get anywhere near a newspaper or social media. But then the really negative stuff potentially is being stoked by those particularly strong voices. Yeah. Can we hear a little bit about your 2025 paper that does what does fear sound like? So that was all about domestic violence, wasn't it? And perceptions that might exist in a courtroom. Can you tell us a bit about that? Yeah, absolutely. So that's a project which that's the first paper on that project. And they've actually just got some grant money to do some more work on that in the next three years or so. And it's all about victimhood. So victimhood is in kind of the sociology literature is talked about as being a subjective phenomenon. So it's not the case that someone has a particular set of criteria and that kind of pushes them over the line. And then they are a victim. It's more that people give more or less legitimacy to people as victims, depending on how much they fulfill whatever stereotypes they have of victimhood. And one of the concepts that I take in this paper is the idea of the ideal victim. So particularly in domestic abuse context in a lot of people's eyes, consciously or unconsciously, the ideal victim is someone who is passive, someone who is blameless, and generally someone who is female as well. And by extension, the ideal offender in a lot of people's eyes tends to be male. That aligns with statistics as well because vast, vast majority of domestic abuse offenders are male and the vast majority of victims are female. That's not exclusive, but that's the general trend. So it's an area of life that is already bound up with lots of stereotypes. And I knew this partly from my own time. I spent a couple of years as a police officer in London dealing occasionally with domestic abuse contexts and seeing the ways in which victims will pitch their allegations and then seeing how other officers might talk about those allegations and give them varying levels of legitimacy. So I was interested to see what role voice might play in that dynamic. And I like, I wanted to focus on pitch because as I've sort of already said, pitch is capable of indexing a wide range of different stuff. It can index your level of fear. So as I mentioned earlier, when people are scared, their vocal tracks constrict, and their pitch tends to get higher. So in theory, the higher you make someone's voice, the more scared they should sound. But in reality, there's all this other social stuff that is complicating the message and complicating how people feel about pitch. So as well as varying the pitch in that particular study, we varied as well the sexual orientation of the person alleging the abuse. So the experiment was set up where I got, actually it was an undergraduate student who was a professional actress who did the voice for me. She was doing the the degree alongside her acting career. And she recorded a one-minute monologue making an allegation of domestic abuse, initially using male pronouns to describe the abuser and then second time around using female pronouns to describe the abuser. I took those recordings and bumped them up in pitch and also bumped them down in pitch. So that gave me six different versions of the same monologue where the only difference is pitch and the pronouns used to talk about the abuser. And then I asked people very simply how vulnerable, how scared, how rational, how in danger do you think this person is based just on listening to one of these monologues. And the findings suggest that pitch matters, but only when you are already fulfilling the stereotypes of the ideal victim. So in the context where the victim was talking about a male abuser, so in a heteronormative context where you have an ideal female victim in the eyes of wider society and an ideal male offender in the kind of domestic abuse scenario that most people imagine. If you lower the victim's voice, that makes her sound less scared and also less rational as well. So people think that she's being less rational in her allegation. If you flip it and then she has a female abuser, she's making an allegation against a female abuser, none of that matters. So you can change pitch and it no longer indexes fear, it no longer indexes rationality. And I make the argument in the paper that this is because once a listener hears that this person is female and has a female abuser, they already kind of rule them out as being that rational because female on female domestic abuse isn't generally taken as seriously. And that shows in other effects in the study where regardless of pitch, people thought that the victim of the female abuser was less in danger than the victim of the male abuser. So it's about this hierarchy of social category relevance, basically. But in particular context, in particular frames is how I put it in the, how I borrow it, moving government and I put it in my paper. Some social categories are more relevant than others. And for whatever reason in this, in the victimhood frame, what matters first is sexual orientation and then what comes after that is gender and gender performance, I suppose. And you might imagine that in other contexts, then through the way around, maybe those contexts in which class is the thing that matters most. But there's a kind of a hierarchy created where I've, I've, I've, I think I know what kind of person you are based on this bit of information. I then don't need to think about the other bits of information that might tell me about you. I mean, you mentioned some software earlier, but yeah, the sort of practicalities of an academic to kind of play around with pitch is that something that's quite easily done for an academic, because you mentioned in the methodology, you kind of, you know, the same bit of language, but altering it in those pitch related ways, how would you, how'd you go about doing that? It's actually remarkably easy. So to manipulate pitch, I used a program called Audacity, a UD, AAC, ITY, which people might be familiar with and Audacity is completely free. So students can use Audacity, students can use Pratt as well, they are both free open source software's and it was as simple as loading the voice and then manipulating based on a number of semi-tones. But being careful not to push things too far, so I meant there are probably more advanced software's out there that I'm not aware of, because I'm not so techy that would allow you to shift pitch and have it still sound like a human. That wasn't the, that's not the case in Audacity. If you go more than a couple of semi-tones up, it starts sounding like a robot. If you go too low, something very similar, but also we wanted, we also wanted the voice to be kind of categorically identifiable as as female, as actually, and actually as cis female, and if you push things too low, it makes people's assessments of the voice a bit, a bit murkier. Basically people are less sure. So I use Audacity, that's very simple to do, but you do have to be careful when manipulating voices digitally that they still sound real. So you have to do the work of norming them, asking people, does this sound like a real person? Could you imagine this voice actually existing as a sort of sense check before you start doing your experiment? Because in the other paper, I think maybe a similar sort of thing with that bit of software, but it's time male and female voices. So this was all about perception of common taste as well, isn't it? Could you sell us a little bit about that related to gender and pitch and perception? Yeah, absolutely. So this was, to be honest, another example of me trying to find an excuse to do research about something I was already interested in. I like football, I like watching men's and women's football, and myself and my two co-authors, Louis Strange, and Sophie Holmes, Elia, to Rolf-Hubble fans. and we thought, "Here's an excuse to spend some time "if you're talking about it and writing about it." So I think this is only a lot to people who are aware of that there is a big media storm or sort of lots of prominent figures in the media, particularly former male football players and current football pundits who get upset about women's voices when they are commentating on football. So the particular claim is that women's voices are too high to operate in that sphere, to commentate on men's football. There are other complaints made, but pitches, the kind of specific one that we honed in on. And we wanted to see whether or not that was a claim that held up in experimental conditions. So whether actually it really is true that when people listen to a voice, to a female voice, they care whether it's higher or lower. We had a sense that maybe it wouldn't be that straightforward and that the media representation of this wouldn't match up with real life and that to not to be the case. So we took this time a male speaker and a female speaker, we got them to re-record real football commentary. So they mirrored, it was a two or three minute video on Sky Sports YouTube channel. They did it line by line, matching the intonation, trying to match the accent and the two speakers themselves had very similar accents. And then we recorded that, put it back with the video as well. So it looks like they are commentating over real football. And then for the male voice, we raised the pitch to 70 tones and we made it higher. And for the female voice, we lowered it to 70 tones so we made it lower. And this and what we were interested in, so this is a very different context to the domestic abuse context and the domestic abuse context. What I think matters to people is rationality and fear and vulnerability. What matters here was a kind of combination of competence and likeability. What gets talked about in lots of sociological literature as status and solidarity. So status is how intelligent, unknowable and competent I think you are, status is how much I like you and how friendly and how cool I think you are. And there's slightly different per seps but they tend to be very gendered and they tend to show up in these kinds of tasks quite well. And what we found was it doesn't really matter what you do to the man's voice. Whether you make it higher or lower, people don't seem to change how they feel. But with the female voice, when you lower it, it makes no difference to how competent people think she sounds. So suddenly, if you lower her voice, people don't suddenly think she sounds more knowledgeable or more competent. But it does have a negative effect on how likable people think she sounds. So you take this woman's voice and you lower it to semi-tone and suddenly people think she sounds less friendly, less approachable. I think one of our things was less fashionable as well just because that's a word that tends to show up in the same per sep. So this is like really strong evidence of something I talked about earlier, which is this double bind. So you have prominent figures in the media saying, women's voices are too high. They are, you know, they are delegitimized in this particular space. If you sound too feminine, you will be dismissed. But our experiment shows that if you lower the pitch, they are dismissed for different reasons because they're being penalized for not sounding friendly enough. So again, you have an example of women being forced to operate within very set bounds out of which their behavior is dismissed and not seen as authentic. Was there a particular action you use for the data, presumably the same across the male and the female? Was there a particular action? Or would it maybe have had a different finding if it was a different action, do you think? Absolutely. I mean, so for that particular study, we used SSBE so they were standard Southern British English speakers. And we've actually just run another study where we compare SSBE accents with kind of estuary English, more SXC sounding, maybe more non-standard accents as well. So it's all the same manipulations, raising the male speakers voice in pitch and lowering the female speakers voice in pitch, this added complication of either they're speaking with an SSBE accent or they're speaking with an SX accent. And what we find kind of matches up, which is that again, when you change the male speakers voice, people don't really care. And this is a domain in which you find lots of non-standard accents. If you think about any time you watch football, you'll find all kinds of different regional accents operating across male and female speakers. So it's very people are very normalized to hearing actions in that context. So adjusting the male voice makes no difference at all, but again, adjusting the female voice makes a difference. So again, lowering her pitch makes people think she's less likable. If you give her an SX accent rather than SSBE accents, again, people are more negative. If you add them both together, it's even worse. So a low pitched SX accent is the voice that gets the most negative evaluations from our participants. Out of the range that you've looked at. Yeah, I'm just wondering, I mean, when you think about it, as you say, there's a whole range of non-standard accents across football aren't there. I mean, particularly I'm thinking like Lucy Ward, as a lead supporter, you know, she's X. She's X, she knows football very well. She worked in football, like many commentators have, many pundits have, but has a, you know, as a West Yorkshire accent, which is often seen as being more friendly in some, you know, some kind of polling, isn't it? Yeah, absolutely. I think, yeah, to to, for us to say that it's specifically about standard versus non-standard, does hide the fact that you're right. There's huge amounts of variation. I think the people we focused on for that discussion was Alex Scott, who has received comments in the media about her non-standard accent more generally and Karen Carney as well. Right. Isn't she from Birmingham? I think so, yeah. Littleins at least. Yeah, but I do wonder if particularly if you think about those cities where they have such a strong footballing identity, so that if it was a Scouse accent or a Mancunian accent or a Jordy accent, whether those might not operate in the same way, I think that's a nice idea. I think that's entirely plausible. And where you're right, there's these huge accent studies that show that Jordy accents are great example. People think Jordy accents sound very friendly and very approachable, and that's why people are building call centres and around that part of the country. Even when a Jordy is about to punch you in the face, they sound endearing. Exactly, exactly. So I think it's much more complicated than I've maybe portrayed it there. And also there's so much more going on than just pitch. All kinds of other features, gendered or otherwise, that might affect how people perceive those kinds of voices. That's really interesting because I mean, like you say, there's all those other factors on there. I mean, it's kind of how people look. It's the kind of back stories they might have, you know, it's who they might be related to, the teams they played for, all of that stuff kind of comes into play, doesn't it? But if it's just the voice, that's something a little bit more kind of isolated and I guess kind of pure, isn't it, the people are responding to. But it's still quite telling, isn't it, what you found. Absolutely. I think that's kind of the nature of experiments is it allows you to isolate these much smaller differences and say, well, you know, I kept everything else constant. I kept other aspects of their accent constant and I, they're reading literally the same words with the same intonation over the same video. The only thing I changed is pitch and people do seem to care about this woman's voice. It's also specific to football. So I feel I think to say is that football is this particularly kind of heteronaut of space, highly like masculine space. I think if you would just change the sport, I think it would make a huge difference. We talked about maybe doing the same experiment with tennis or gymnastics or something where representation, male and female representation is much on a much more equal footing. And at least anecdotally it feels like there's just as many kind of male and female commentators, certainly in gymnastics anywhere, don't know for sure about tennis. But where masculinity seems to play less of a big role in deciding who gets to do what. So I think shifting that, shifting the frame, which is again what I talk about both the papers, might make a difference there. That's really interesting because I think there was some recent work done by Selina Cuddy, University of York, on she was doing like a social-ingristic study of a women's football team. I think in South Yorkshire that she was looking at kind of how the women in that team, and I think there was a majority of that team were less being rather than straight. And there was a kind of performance of identity, a kind of football identity within that, that had some really interesting kind of threads to it. But as you say, it's kind of the research associations we football on there about kind of performance of identity within that. Absolutely. There's kind of what I talked about before. It's that toughness, which is kind of seems to be intrinsic to that kind of sport. And there's definitely a class element too. I mean, so the other sport I follow very closely is rugby, and hugely different in terms of how they are classed and the types of schools that we end up in those sports tend to have gone to. And as a result, in certainly in rugby, when you, when a working class player speaks to the media, Ellis Gennan is the most famous example. He has frequently picked up on his pronunciation, his dropping of G's, all those other stuff, which in a football context, people could not care less about. It's actually weirder to hear a male footballer wear the sort of posh SSPE accent I find. And that seems to be more striking, whereas the reverse is true in a sport like rugby. I guess rugby union, definitely. Rugby needs. I've got more associations with the North, isn't it? Yeah, that's true. I should be careful there. As a other Southern, there are very much a rugby union trope rather than a rugby league trope, for sure. I think football is an interesting case in showing you that what matters for class or certainly for class performance isn't just money. So you've got all of these professional footballers getting paid millions of pounds a year. But if you listen to them in interviews, none of them have changed their accents. They all still have them. But all of those, those of them who came from working class backgrounds originally still speak with exactly the same accent. So class is about so much more than just what a person earns, for sure. It's definitely more complicated. Yeah. So you mentioned about kind of applying these findings to areas beyond football, you know, other sports as well. Do you think it would be interesting kind of think about how they might apply to other areas outside of sport? So things like sort of voice overs, newsreaders, politics, courtroom testimony. I guess you're kind of touching on that with your domestic violence research. Do you think there'd be practical applications in which we might want to? encourage students to think about pitch in relation to those as well. I think definitely, I think some of these patterns travel and some of them don't and that's why context really really matters. So in those kind of authority heavy roles, things like newscasting, politics and courtrooms, I think certainly we're just talking about pitch, that's maybe where lower pitch tends to help. But only over up to a point, only over up to a point of still being perceived to be authentic and you have to remember people have these kind of, these anchors of what they imagined male speech to be or female speech to be or working to us, male speech to be whatever combination it might be. And you can only buy yourself so much advantage by changing a voice in those kinds of contexts. And again, that's particularly true because of the double-bind for women where sounding too authoritative can risk being judged as aggressive or cold, but at the same time if you are sounding too warm or emotional, you can risk maybe not being taken so seriously and in those authority heavy roles, that's definitely in highly gendered spaces. That's where the expectations are even strongest. So in a football context, in definitely in a courtroom testimony context way of the victim giving testimony where you got crimes that are themselves so gendered, that's where expectations are going to be strongest. And it's really listeners reacting to what they see as fitting with the role rather than someone's actual kind of vocal skill if you like and the same voice can be judged very differently across different settings. I think in terms of real world application, I talk about this a bit in my, what does fear sound like paper towards the end. When someone makes an allegation of domestic abuse, they'll be a written statement made if they're fit to do so, but they also kind of make a verbal statement and often nowadays that's recorded. So police officers are they wear body worn cameras, so whatever your first words to a police officer are, those are recorded and those can then be used as evidence. So in a particularly in context where a victim might not support prosecution, so maybe they've called the police or maybe someone else has called the police and they didn't want the police there, the police show up, see enough evidence to arrest someone, they can pursue it anywhere regardless of whether the victim does or doesn't want the person to be prosecuted. It's one of the few crimes for that's the case. What can then happen is if they don't support prosecution, they happen given their kind of written statement, the thing that gets shown in court is a video of them giving their allegation or at least saying what happened, whether they're making an allegation or not, and pictures one of the things everyone's going to hear. So the police officer has heard someone's voice, the CPS, who have given all the evidence, deciding whether to prosecute them, given that voice, there's a judge, there's a jury, there's all these other people who are at different points in the legal process are hearing a voice and making a judgment about how scared or how rational they think that person sounds and actually when police officers are in context where they don't support prosecution, they have to write afraid statements, afraid is the acronym for the statement that they have to collect and the F in afraid is fear. How scared do I think this person sounds and a police officer will just write, though, but the person seemed very scared and that statement again is used as evidence in victimless prosecutions, they can use what the police officer perceived in that context as fear or as rationality, as evidence to put in front of a jury, which is really, really serious, the fact that people come into the situation, police officers are just like every other person on the planet, no matter what training you put them through, they come with their own biases and their own ideologies, so they'll be seeing things through a particular lens and it's concerning, but that could then have a real world impact on the journey of victim goes through the court. I was just wondering when you were saying about that in terms of like training for police around those sort of things, are you aware of any ways in which police officers are a given specific training about the idea of sort of voice, voice pitch and what it might mean and overcoming biases? I mean we've seen some work being done in like the legal profession, for example, around accent bias, is there any equivalent maybe in the police? As far as I know, no, and that's possibly because at the moment accent is not a protected characteristic. So, so police are given the same kinds of unconscious bias training that you might get in any other profession, but they get that right at the start when they're in in in training school, once they're learning a bunch of other stuff, and again, it's anecdotal, but in my experience, in those particularly tense scenarios where someone's been working through the night for 10 hours, it's the last couple of hours of their shift, they turn up to a call and they've got to take a statement, they're tired, they've not had their food, you know, people revert back to their worst biases and no matter what training you give them, and I think, particularly people who work in those kind of stressful jobs, I think there's only so much you can do to train someone out of their biases. I mean certainly you can screen people, so there's an argument in the police at the moment about whether or not they are doing enough vetting in terms of getting people in the first place, so at the most people would strongly say that the answer is no, but the actual bigger question is whether it's really about vetting and more about the kind of culture that the police actually have. So just to link this back to both of the studies that I've talked about, as well as collecting this experimental data, we also ask people to respond to surveys that's given an indication of their just pre-existing ideologies around gender, so just how much people believe in traditional gender roles, and you kind of take all those scores, put them as a turn them into a composite, and you can look at how it functions in your model, and what we find is in general people with more conservative attitudes towards gender already are more susceptible to differences in pitch. They're more susceptible to being guided by higher or lower pitch or by the kind of male female abuse or dynamic in my study. So there's information you can try and collect about people, and actually if anything, that's the stuff that's more worth addressing, that could look kind of more fundamental belief systems that people have. You can tell someone trying not to focus too much on the picture, the voice that person had, but you know you tell them that and then three years later they're dealing, they're trying to deal with an incident, and again they're tired, they're not eating and so on, I think it's very difficult to address that. And as you say there will be some organisations and professions where there are kind of systemic issues around it aren't there as well, particularly those with the kind of power dynamic like the police or judiciary and things like that. Yeah absolutely, I mean to the next three years of this project, so I just got some funding to do more work and I'm very lucky that a local domestic abuse charity in Winchester stopped domestic abuse of agreed to work with me in helping me create new materials for experiments but also supplying participants for the experiments. So then this big project is going to be about expanding the base of that first paper looking at different accents, different regional accents, different racialized accents and classed accents, but comparing the general population with people who work in victim facing roles in victim support to see whether they are any better than the average person at filtering out those biases. So hopefully come back to me in three years and I'll have a bit more of an answer. Really interesting. Yeah. Just to go back to some of the methodologies for the experiments you spoke about, the matched guys technique was mentioned in a few of your papers wasn't it, which hopefully something a level students will be familiar with. Just remind us what that term refers to and the implications of it for your work. Absolutely. I mean the original matched guys was done using French English bilingual speakers so it was an experiment done to to work out whether or not people preferred hearing their own language and so they got the same speakers to record. I forget exactly what kind of text it was, but some kind of text, in English some kind of text in French and they generally found French speakers whether and even if they do speak English prefer hearing it in French and vice versa. So I mean matched guys is really just about having more than one voice where you've manipulated something whether that's digitally or whether you've just asked someone to do two different versions of the same speech in different accents and then seeing how people react. So I use like at scales so you ask people on a scale of 1 to 8 or 1 to 10 how competent, intelligent, articulate whatever someone sounds. Half of your population hear one voice, the other half hear the other voice and then you can directly compare the results and the idea is to keep everything else like it like in all experiments keep everything else as constant as possible. So keep them having reading the same text if you can have the same speaker even better. That's why people do digital manipulations often because you can have the same speaker and you can cut and paste different features into their voice so I have a paper where I cut and paste in or in into people's voices and lots of other papers that have done the same thing. You can do that in part very easily. So just keeping everything as constant as possible so you can say with a degree of confidence the difference in how people feel about these speakers is because of this accent difference rather than something else. I just think it a bit about the kind of investigation project students might do in their A levels as part of their their NEA. I mean you know they're keeping variables constant as often as something often as stressing to students. I mean you've given us some great ideas there of you know studies that could be done looking at kind of gender and perception. I mean is there anything else on the agenda you think that perhaps some A level students with an interest in gender perceptions of gender and language might be usefully exploring and realistically exploring. I think the big word there is realistic right so what can we ask A Level Students to do with the time that they have balancing or that you know there are other subjects and stuff like that and also with the resources that they have. I know that A Level Students don't have access to microphone and centrally and all that other stuff. I think analysing readily available discourses around gendered voices is a good place to start so it's like newspapers but not just the newspapers themselves the comments section that you find but beneath newspapers some of the less solubrious newspapers have very well populated comments at the end but there's also things like Reddit forums there's kind of online forums. I mean the first thing we did on this football project is we went to fan forums. We went to just particular clubs will have a forum where people can post whatever they want anonymously and the ones we found where lots of people can complaining about female football commentators and we have a paper hopefully coming out at some point next year about the language people use to talk about female voices in those contexts. But I think the stuff like that work 'cause experiments, you know, they're sexy, they're cool and they give you something quite controlled, but you have to find participants and you have to design things very carefully. If you can find things out there that are just ready made, or even do a survey using voices that are ready-based, you don't have to necessarily match things up. If you can just find two people that are roughly the same age, maybe the same gender, same kind of region, but they vary in one respect, maybe they're different in terms of class, or different in terms of sexuality, whatever it might be, and then you can just compare those two speakers that would be much more manageable, might be. - Yeah, 'cause often students are quite interdisciplinary, perhaps they're doing kind of NEAs related to media sources, you know, using the traits, using love Island or whatever, rather than perhaps, you know, finding real people in the real world, but you presumably, you know, favor more kind of the sort of reality than the via the media would you as a sociolinguist? - As best that I can, whether the resources I have, yes, very much so, I mean, I'd like to have best to answer this question, actually. So, for in my undergraduate classes, I teach a module on language variation and change, and for one of their assignments, they have to do a mini variationist project, so they have to identify a particular linguistic variable, you know, glottling in maybe a particular vowel, and they just have to compare two speakers for how they use that vowel, and that's an example of where they have to control, so they have to find two speakers who, again, are the same class, same region, same age, but maybe it's a male and maybe it's a female. And in that context, I ask them just to find stuff that already exists. Go to podcasts, go to reality TV shows, which offer this sort of middle ground, where they are sort of fiction and sort of nonfiction, right, so they're not scripted, but they're somewhat performative and the context have been created for them. - Scripted reality, I think, or dramaticity, I've heard it being called. - Yeah, exactly. - Scripted reality, which it sort of feels a bit like how I describe politics in some ways at the time. So those are nice examples that students can use without having to go through the rigmarole of ethics and finding real speakers. And what you're trying to do is demonstrate that you understand the investigative process, which I'm guessing is what really matters for students when it comes to getting good grades and so on. To me, that is very much sufficient. You're not trying to find original findings that you're going to go publish in a journal somewhere, which are trying to show that you understand how to build an argument, start to finish, and showing the methodologies that you've been taught and so on. So that would be my suggestion. - Yeah, that sounds really interesting. I mean, I think that is kind of what we're after for students, isn't it? Is to talk through their process, think about what they're going to do, gather useful data, analyze it, apply kind of ideas that they've done as part of a literature review, including, you know, reference, your two papers here, and any of the stuff in your bibliographies, well, that type them into sort of interesting dimensions. - Absolutely. So justifying your decisions, you know, using literature. One of the things I noticed even at university level is that students need to understand moving away from citing literature because you think you need a site literature because that's what you've been told to do. And it looks good towards using existing literature to justify what you're doing right now. So, you know, if you're using a particular method, let's say it's a match guys, you know, read a couple of studies that have done used similar match guys, techniques, cite them, and, you know, justify your decisions based on what they did. And that's kind of moving away from performative use of secondary literature towards integrated uses of secondary literature. - That's really, really good advice for a level, particularly in the world of chat GPT where, you know, increasingly you're seeing not just at a level, but everywhere, people doing literature reviews that they've got chat GPT to do for them. There's no learning, it's taken place. - And it's, you know, for the course weight, you get no credit for it and you've done for plagiarism. - Yeah, absolutely. - But it's one of those kind of things, isn't it? Where it's about, it's justifying, I've read this, these really interesting ideas, I'm going to apply these ideas to my own work, here are the kind of limitations, here's my rationale for doing it, here's what I'm going to make it work. That's what we really want, isn't it? - Yeah, absolutely. I mean, if it makes teachers feel better that Gen AI problem is not specific to students, I was reviewing PhD applications today and came to the reference list and found two papers that were alleged to have been written by people in my department and I know they worked well enough to know they definitely didn't write those papers. And I've even reviewed a journal article for a relatively well-known journal where one of the footnotes, if you take a citation from chat GPT and you put it as an HTML link, it has chat GPT at the end of that link and I was like, I know this is, you know, anyway, sorry, I wasn't trying to write that. - In too much detail, but it's really not, and unfortunately, it's not specific. - Yeah. - Very much not specific to students. And I guess we can put in the show notes, links to audacity and part, which you say, you know, freely available. I mean, I can picture examples where students at AI level could be using those to do like a mini project or an investigation. - Absolutely. I mean, for my undergraduate module, they can, sort of, for the variation and change stuff, if they want to look at vowels, they can use things like part and for all the plotting and graphs, just Excel. You know, I mean, I think Excel, you have to pay for Excel, but most schools have, you know, a Microsoft office. License, if you want to be really fancy, you can use things like, ah, but you need to know a bit of coding. Microsoft Excel, pivot tables and pivot charts are incredibly intuitive and easy to do to create things like box plots or bar graphs or whatever. You really don't need to be a computer whiz and there's plenty of support stuff out there on YouTube or wherever to tell you how to put those things together. So for any AI level student wanted to do something quantitative, I'd really recommend just like spend 10 minute Excel, looking at what it can do is a really powerful tool. No, great stuff. Fantastic. No, thanks very much, Matthew. It's really, really interesting and helpful. No problem at all. Thank you very much for having me. [MUSIC]