The Eras of Taylor Swift's Changing Dialect

Longitudinal studies of how an individual's accent changes over the course of their life are hard to come by. Fortunately, Taylor Swift's decade-plus career-- and the numerous interviews she's given over those years-- has opened a window into our understanding of how and why dialect changes may occur on an individual level. We talk to Miski Mohamed and Matthew Winn (University of Minnesota) about their work analyzing the shifts in Taylor Swift's speech over the years.

Associated paper:
- Miski Mohamed and Matthew B. Winn. "Acoustic analysis of Taylor Swift's dialect changes across different eras of her career." J. Acoust. Soc. Am. 158, 2278–2289 (2025). https://doi.org/10.1121/10.0039052

Read more from The Journal of the Acoustical Society of America (JASA).
Learn more about Acoustical Society of America Publications.

Music Credit: Min 2019 by minwbu from Pixabay. 

Transcript

ASA Publications (00:26)

Being a scientific podcast, it's not very often that we get to discuss pop culture, but today we do! Specifically, we'll be talking about the speech of Grammy-winning singer and songwriter Taylor Swift. With me are Miski Mohamed and Matthew Winn, who recently published the article, “Acoustic analysis of Taylor Swift's dialect changes across different eras of her career.” Thank you for taking the time to speak with me today. How are you?

Matthew Winn (00:49)

I'm doing great.

Miski Mohamed (00:50)

Doing good.

ASA Publications (00:52)

Good. So first, tell us a bit about your research backgrounds.

Matthew Winn (00:56)

I can start. My research is about speech communication and how it's impacted by hearing loss. So very different than studying Taylor Swift. This is kind of a one-off for us. I mainly work with adults who listen with cochlear implants. So they've undergone a transition of losing their hearing and then regaining it and relearning how to listen. And our lab focuses on understanding the things that make listening effortful for people in that situation and things that might impact their daily communication, even though it's not captured by their clinical evaluations.

Miski Mohamed (01:29)

I'm an AUD grad student and the research that I'm interested in is about speech perception and listening effort. So, besides looking at Taylor Swift's life and analyzing that acoustically, I've worked on research that involves listening effort, anticipation of speech, and how to understand how these situations impact hearing loss.

ASA Publications (01:49)

Interesting, interesting, okay. So, what are the scientific reasons for analyzing Taylor Swift's speech? Or perhaps more succinctly, why pick a big star besides name recognition?

Matthew Winn (02:00)

There are a lot of reasons why Taylor Swift is ideal for this kind of study. There is already some good science out there on how to compare dialects for groups of people. You can take measurements of people in New York, people in California, in Georgia and compare those groups, but it’s much more challenging to see how an individual’s speech changes over time because for one, most people don’t move around to different parts of the country, and even those who do, don’t have someone following them around with a microphone capturing everything they say. 

But Taylor Swift is in that position, exactly – her speech has changed over the years, and also her voice has been recorded so often, like in interviews and promotional events, that we can get a window into those changes when they happened, as they happened, in a way that is normally not possible.  

Miski Mohamed (02:53)

Yeah.

ASA Publications (02:53)

It’s hard to do longitudinal studies with speech.

Miski Mohamed (02:58)

Yeah. And it's just so rare to be able to observe how an individual's speech changes over the many years, but because so few people have recorded enough data to actually have that analysis, but Taylor Swift is obviously popular enough to have been interviewed and recorded like many times over the years, so we are able to see those changes in those recordings. 

What makes Taylor's voice so valuable to us scientifically is that she had many specific reasons for changing her voice over those years. She was either positioning herself in a different geographic place or with different social groups that have those distinct dialects. So we can see if her voice had actually changed in specific periods of her life. So we're able to actually match those periods to the different goals that she had with connecting with different groups. So we're able to see those subtle changes through a documented backlog of interviews that are accessible. 

And this type of analysis has actually been done before. There has been a study analyzed Queen Elizabeth's voice over the years. She's always recorded a Christmas podcast each year, and it turns out her speech actually changed over the years from a more upper class style into a more general public way of speaking. So this shows us that this kind of like long-term analysis is actually something that can be done.

ASA Publications (04:23)

Interesting, yeah.

Matthew Winn (04:24)

Yeah, and another thing that makes this interesting, I mean, apart from the obvious interest that a lot of people have in Taylor Swift, is that she just really brings us an opportunity to look at things that are just so hard to look at normally. And one of the things that we're trying to emphasize in this paper is that even though a lot of people think of dialects as belonging to geographic regions—you know, the north, the south, the west, you know, however you think about it—dialects also serve a social purpose. And when Taylor Swift was in Nashville, she was not only in the American South, she was also integrating into a country music community where the Southern accent is a key unifier just within that community. And so when she's in New York, her voice had a, she had a, not only changed the way she pronounced words, but also lowered her voice pitch. And so that also might signify something different, not just being part of a music community, but as someone who's emerging as a person who wanted to make important contributions to national discussions of social change and feminism and musicians’ rights. And so, all of these changes we can link to what she was doing at the time and who she was trying to connect with at the time. And the voice pitch thing turned out to be an analysis we did just near the end of the project. And even though she's capable of speaking with authority on all those issues, no matter how high or low her voice was, there’s a tendency for people to maybe take someone more seriously if they have a lower voice pitch, and she might have been taking advantage of that tendency to strengthen her message. So there's just so many different angles that you can think of these voice changes, in addition to whatever interest you might have in the singer herself.

ASA Publications (06:04)

Yeah, yeah. It's funny because we just actually did an episode on the Iconicity and Sound Symbolism issue of JASA, and voice pitch being seen as dominant or important came up in that discussion as well. So it's interesting to hear that happening with a specific person. Okay, so, but to get back into your article, generally speaking, what acoustic features are you looking at when you study dialect and what differences might you expect to see from people of different regions?

Miski Mohamed (06:34)

So the main thing that people analyze when they study dialects in English is the vowel articulation. Vowels carry a lot of cues that tell us where someone is actually from. So that's what separates a New York, Alabama, Chicago, or Minnesota accent, is the specific way that the tongue moves when you're actually articulating a vowel. So we can measure this by tracking the first two resonances of the vocal track as someone says a word, and actually plot these onto a graph. And then you can set up the graph in a way that can interpret the small differences in the trajectory of the tongue movement in one dialect versus another. So by actually plotting these points, you can see how one accent’s graph compared to another one will have slight changes in where the actual vowel can start, where the vowel can end, and the actual way that it's moving throughout with the tongue.

ASA Publications (07:32)

Ohhh… Okay, interesting, interesting. That makes those graphs in the articles make more sense when I see them. So looking back at previous research in this area, what do we know about how a person's dialect or accent might change over the course of their life?

Miski Mohamed (07:47)

So, for the most part, a person will acquire their dialect from their peers. We know that parents will play a role in this, but it's mostly your peer group that will shape the way that you speak. One of the other things that we know about dialects is that when they change, it's usually because the young women are leading the change by adopting new words, new pronunciations, and then everyone else usually catches on later.

ASA Publications (08:11)

Wait, so young women are the trendsetters in language acquisition or dialect acquisition? Oh, that's cool. Anyway… Okay. So are there differences in how a person who's been speaking a dialect their entire life would sound compared to someone who has moved to a region more recently?

Matthew Winn (08:28)

Yeah, for any dialect, there's a lot of potential features that one can express. And, you know, if you go to any particular region, if I go into Philadelphia and I have a checklist of all the things that makes up a Philadelphia accent, no person I encounter is going to check off every item on that list. Everyone kind of takes a selection of those features and it becomes their personal expression. And one of the key differences between speaking that dialect natively, you know, from childhood, versus acquiring it later in life is that you're much less likely to change some of those subtle things later in life. So in the case that you do adopt some of these features later in life, like Taylor Swift did, it's probably only the most salient things you're gonna pick up on, right? So like in Nashville, I'm gonna notice, because I'm not from there, I'm gonna notice when someone says the word “ride” as like “rod”. And like, gonna notice that because it's so different than how I say it. If they say the word “cute,” like “cyu-ute,” that's such a departure from my native dialect that I'm going to think of that and, you know, just dwell on that. But there are other really subtle things, specific word choices and some of the patterns that happen, like if this sound occurs before that sound, then the change doesn't happen. And that's one of my favorite examples of that is in the Philadelphia dialect of English, which I grew up around the OO and the O vowels are pronounced differently when the letter following it is an L. And we saw Taylor Swift maintain that in the recordings we had. So she grew up pretty close to Philadelphia and had some of those features of a Philadelphia native speaker, which was pronouncing those two vowels, OO and O, way far back in her mouth when the next sound is L, e more so than what you'd expect from someone who didn't come from Philadelphia. So that's probably not something that you'd pick up on unless you really grew up around it. So it's those patterns that stick with you as you move around if you're acquiring a dialect as an adult.

ASA Publications (10:24)

Okay, that's interesting. What would be an example of that? Like a word? Yeah.

Matthew Winn (10:29)

Oh, a word like school or cool, whereas in a lot of dialects that “ooo” might sound more like “eww,” but in Philadelphia it's much more like ooooo, like way farther back in the mouth.

ASA Publications (10:43)

Oh… Interesting. Okay. So let's get into your study and Taylor Swift. So did you focus on Taylor's singing, talking, or a little of both? And how did you select audio to analyze for your study?

Miski Mohamed (10:56)

So we did begin by trying to look at singing, but we ended up just focusing on her conversational speech. So we'll talk about singing a little bit later. So in this study, we actually just focused on her regular speaking voice whenever she was in interviews, just because singing is so hard to analyze when there's lots of background things going on, like guitars and drums. So we decided to just use recordings of interviews with basically no background noise. So the audio was actually from interviews that Taylor did while she was on tour promoting different specific albums throughout her career. And then the strategy that we used was to find interviews from those different time points. So when she was living in a specific city like Nashville, Philadelphia, or New York.

ASA Publications (11:45)

Okay, okay. That all makes sense. So how did you actually conduct the analysis of her speech?

Matthew Winn (11:50)

Well, we wanted to sample as many different vowels as we could because, although there are some vowels that really serve to be the markers of different dialects, you want to look at the whole space. But there are constraints to doing this kind of research that make it pretty difficult because some vowels are rarely ever spoken. So you don't have an equal distribution of like 10 EE vowels and 10 OO vowels and 10 AH vowels. Like the vowel in the word “boy,” that OI sound happens so rarely that you don't get enough samples to include it in the analysis. I mean, thankfully the key dialect markers are common enough that we captured plenty of those. But even then, there are some constraints. So when a vowel is followed by an N sound or an M sound or NG, we can't include that vowel because that's what we call a nasalized vowel because you start to pronounce that M sound even before you're done with the vowel and it complicates things. So there are all these historical things that people have documented that we're able to take advantage of to make sure the analysis was good and clean. And in these cases, as we're just listening to speech, all of these constraints don't really bother us. It doesn't get in the way of us understanding someone. But for a computer analyzing that sound, as we did, those little tiny things can be disruptive and throw off our whole analysis. So what we did was essentially capture as many different vowels as we could, dozens and dozens and dozens of vowels for every different time point in Taylor's career. And then Miski went ahead and did all the other analysis, which she can tell you about.

ASA Publications (13:27)

The brutal work of a grad student. 

Miski Mohamed (13:31)

Yeah, so then to actually make those measurements of the vowel, we use the software that's very familiar to people that do speech communication research called Praat. So we wrote a Praat script that just went and found every vowel that we labeled, and then it made measurements at 10 different time points throughout the vowel. So then at each of those time points, it was able to estimate the two formant frequencies that I was talking about, which are the key measurements of the vowels. And then there was a whole lot of data cleaning and statistical analysis and everything that you can find in the paper that I did spend hours and hours inspecting every single vowel to make sure it was tracked correctly.

Matthew Winn (14:15)

Yeah, can confirm.

ASA Publications (14:15)

Yeah, that sounds like a lot of work. So what changes did you see in Taylor's vowels over the course of her career?

Matthew Winn (14:21)

There are two really clear examples we can describe, and I mentioned them in passing before, but I can kind of get into more of the details. The first thing is that she adopted features of that southern accent specifically when she lived in Nashville. So the signature vowels in this dialect are that I vowel, like I said in the word “ride,” which becomes more similar to the word “rod.” It's not exactly like that, sort of in between, like “rah-ed.” And we call this monophthongization of I, and the reason that we use that word is we think of I as a diphthong. It has an onset and an offset. It starts with AH and ends with EE. And that AH-EE blend is what we expect for dialects outside of Nashville. But inside of Nashville, it sort of ends where it starts. It starts in AH and it just kind of ends there too. So that vowel becomes kind of shortened in its trajectory. 

And another thing in Nashville is this what we call vowel fronting. So this happens a lot for the OO. So if you have like a word like “who,” that would shift more toward like “wh-ew.” So that ooooo-eww, that if I slow it down, it's kind of sounds like I'm not happy to say it. But it's just one of those things that happens in English and in most dialects in English, this actually does happen. So this is not just a Southern thing. So if we have like the word like noon, your tongue also fronts. But what we found is that Taylor did this even in words where it's less common in other languages and exaggerated it even more where you would find it elsewhere. So it's just an exaggeration of that fronting, which is one of those signatures of Southern dialect.

ASA Publications (16:03)

Okay, okay. And then you also looked at Taylor's voice pitch, not just vowels. How did that change?

Miski Mohamed (16:08)

Right, so that was the second major thing that we measured. So we were able to see a lowering of Taylor's voice pitch when she moved to New York City. And this was in 2019. So this was a time in her career when she became more known for speaking up on issues in the world, like issues of like social change and feminism, as well as musicians rights. So it's possible that she was lowering her voice to ensure that she was being received as a voice of authority.

ASA Publications (16:38)

Okay, so can you tell us more about the kinds of messages and ideas that she was trying to elevate at that stage of her career?

Miski Mohamed (16:44)

Yeah, during that stage of her career, Taylor's voice took on a bigger role for her, both literally and figuratively. So she became more outspoken about being able to own her masters of her first few albums, and, like,  highlighting how young women are taken advantage of in all kinds of professional careers, regardless of how long they've been doing it or their success in it. And Taylor was also speaking up about different social issues due to the political climate as well as personally leading movements in Tennessee. So Taylor was more deliberately speaking up for herself and others. So lowering her pitch can be seen as just like heightening her position in society at that point.

ASA Publications (17:32)

To seem more dominant or more like, yeah, a position of authority. Okay.

So regarding pitch, the study happened to cover over a decade of Taylor's life. How much would this change in her pitch be related to aging?

Matthew Winn (17:46)

Yeah, this analysis did Taylor's life from age 19 to 30, and people's voices do change over those years. To be specific, the pitch dropped from having a median value of 197 hertz to a value of 182 hertz, that lower value being when she was living in New York, and that's a pretty noticeable difference. And we do have multiple ways to interpret this. We think that the interpretation based on her social and leadership goals makes a lot of sense because in previous analysis of women's voice pitch across the lifespan, this whole time frame that we measured has been treated as one time point. And that's to say, it's a distinct period that's different from childhood and it's different from old age, but we don't have strong reason to think that things just naturally change within this window on their own. Now there could be some research out there that we just didn't see. But from the research that we did look up and find, this is treated as one time frame. So when we're measuring these changes, we want to look for other explanations. And based on her visibility in the media, based on the topics that she was talking about and gaining leadership with, we're inclined to think that had an influence on these measurements.

ASA Publications (19:03)

Okay, okay, that makes sense. So she's just kind of generally considered a young woman for the entire time and not really a difference between childhood and young womanhood and younger and older.

Matthew Winn (19:16)

Right.

Yeah, I would just add that like for all these things, I think that it's probable that Miski and other people who are really into Taylor Swift, like you've probably noticed this over the years, and one of the things we're really happy to do in this study is to provide concrete measurements of these things, not only the voice pitch, which is pretty easy to measure, but those vowel articulations where you can measure something and really see how much it changed, which is why we thought it was a good fit for this journal because it's an acoustics journal.

ASA Publications (19:46)

Yeah, right, it's all these little acoustic features.

Matthew Winn (19:49)

Yeah.

So I mentioned this shortening of the “I” vowel to something more like “ah” – that’s how we heard Taylor say that vowel while she was singing country music in Nashville. But then in our recordings from the year 2012, when she was transitioning into pop music, the end point of that I vowel was extended back into what you might call its full range. And the way we know that is that you can measure the second resonance – or what we call the second formant – of her voice, and see that it rises up to about twenty three hundred Hertz. This is not her voice pitch, but it’s an overtone that stands out and tells us how far forward the tongue is in the mouth. Back in the country music era, this measurement came out to be much lower, around 1800 Hertz. And this is a difference that is very perceptible. You and I would just hear it as a marker of a Southern accent. And so the comparison of her pronunciation of that vowel during and after her country music period was something that was pretty easy to measure. And you can see in one of the figures in the paper, where we show how to track those differences that are very noticeable, and even some of the ones that are less noticeable.

ASA Publications (20:57)

So what are some of the reasons why these kinds of dialect change happen?

Matthew Winn (21:02)

Well, we were hoping that you would invite Taylor on the show and we could ask her herself. 

ASA Publications (21:06)

I mean, right?

Matthew Winn (21:08)

So without trying to get into her head too much, there's a few ideas that make sense, including geography and culture and aspirations, which we've talked about. So geographically, she was surrounded by people in these cities who spoke with the local dialect. You know, when she was living in Nashville, she wasn't just doing country music. She was surrounded by people who sounded like they were in Nashville.

And people normally assimilate to some degree adopting some of the local ways of speaking. But culturally, particularly in Nashville, her dialect reflects what we feel an authentic country singer should sound like. So the prototype of a country singer is someone who speaks with a Southern accent. So that was probably another motivation.

Miski Mohamed (21:47)

And then we can also think about her aspiration to be a voice of authority in her generation as we were talking before. So being someone that can stand up for others and talk about hypocrisy and double standards in society and how the music industry treats men and women, and being able to just point out the patterns that can lead to some people getting privilege and other people having it held back. And just being someone that can advocate for musicians who don't have as much power to speak up for themselves. So in that case, just lowering the pitch of your voice might have an impact of adding some extra strength to the message that you're trying to put through.

ASA Publications (22:30)

Yeah, yeah, that makes sense. So what were the biggest challenges in this project?

Miski Mohamed (22:34)

So obviously when like I was thinking about the project I could like hear the tiny changes myself being a fan of Taylor Swift but actually like putting in the analysis of the acoustics is like very tedious, so you have to just like listen to the recordings over and over again and just like triple check everything that we're measuring. And then also like being able to catch when the computer makes a mistake because there was a slight change that it didn't notice, and then being able to add that statistical analysis to actually measure the data.

Matthew Winn (23:08)

Yeah, I agree with all that. And I think that in addition to the technical process that Miski described, I think a big challenge was needing to engage with the whole area of scholarship that was outside of our normal lane of audiology and hearing. So we understand how hearing works and how speech communication works from the hearing side. So we have some confidence in measuring and expressing our knowledge of the basic speech acoustics, but when writing this paper, we also had the responsibility to contextualize this work in a long history of scholarship on the social use of language. And so it's important to try hard to get that right because sociolinguistics already is a field that's traditionally been underappreciated, and they bring so much to help us understand why these changes happen, and what are the right methods to capture those changes, and how to contextualize them in so many interesting ways.

ASA Publications (24:05)

Yeah, yeah, totally. So there any other things that got left out of this paper that you thought were interesting?

Miski Mohamed (24:11)

So, originally we did want to compare how duplicate recordings could be measured and see how consistent they were. So Taylor had re-recorded entire albums in her catalog, so there was a great opportunity there to be able to compare the same songs that were recorded in 2008 compared to 2021 and being able to compare the same songs that are sung in the studio versus in live recordings. But as it turns out, she just so consistent with those things that there wasn't a lot of differences to show for that.

ASA Publications (24:44)

Huh, that's really interesting. I guess it means too, that if you get to go see her show now, she will sound just the same as she did, you know, singing your favorites from 2008, today 2025. Do you have any closing thoughts?

Matthew Winn (24:59)

Yeah, I mentioned how, you know, how the real experts in this area are in sociolinguistics. So going into this whole process, I felt, and I think that Miski felt as well, that we have this relatively simple story to tell, like, okay, we have some changes we're hearing in Taylor Swift's vowels as she's living in different cities. But then we submitted this paper to JASA and the reviewers taught us some, some big lessons. They really challenged us to not just make some measurements and put it out there and walk away, but to embrace more of the knowledge that sociolinguistics brings to bear on this topic. And they challenged us to do better statistics and to dig deeper so that we can be much more confident in the conclusions that we were drawing. So I'm just really grateful for the time and effort that those reviewers put into motivating us and challenging us to do better. Some people complain a lot about the review process, and I have in the past complained about the review process, but in this case I think that they taught us so much and made this paper so much better than it was, and we won't ever find out who those reviewers are because of the blind process, but I hope they're listening to this and feel some sense of satisfaction out of this final product as well.

ASA Publications (26:12)

Yeah, thank you, reviewers!

Miski Mohamed (26:13)

And then one last thing, Taylor, give us a call, come to Minnesota and just hang out with us. And hopefully by the time this comes out, her new album is out too. So I hope everybody will listen to that.

ASA Publications (26:25)

Ha ha ha! Well, thank you for taking the time to speak with me, and I hope Taylor sees your article and you do get to hang out with her because that would be amazing.  I personally found it really interesting to hear what we can learn from studying how one person's speech can evolve over time, and hopefully your research will help us better understand how dialect can change on the individual level. that you again and I wish you the best of luck in your future research whether it's related to sociolinguistics or cochlear implants or whatever.

Matthew Winn (26:35)

Thank you.

Miski Mohamed (26:56)

Yeah, thank you.