Subscribe
Share
Share
Embed
The Sci-Files is hosted by Mari Dowling and Dimitri Joseph. Together they highlight the importance of science, especially student research at Michigan State University.
Welcome to the Sci Files, an impact 89 FM series that explores student research here at Michigan State University. We're your cohosts, Marty Dowling
Dimitri Joseph:And Dimitri Joseph.
Mari Dowling:Today, we have Yuting Gu with us to discuss her research in linguistics. Hi, Yuting. Thank you so much for joining us today. Could you please introduce yourself and your research?
Yunting Gu:Thank you so much for having me. My name is Yunting. I'm a 5th year PhD student in linguistics, and I mainly work on the sound aspect of language.
Mari Dowling:By sound aspect of language, do you mean how people are producing language?
Yunting Gu:So there are thousands of languages of the world, and there are some common features shared by those languages. And as linguists, we work on the language of human beings in general, and we want to find some ways to explain some common patterns.
Dimitri Joseph:That sounds like a very interesting topic, Yunting. What are the major components of a language? Just so that us and our listeners can get a background of of what happens in linguistic studies.
Yunting Gu:In the field of linguistics, we study many aspects of languages, and we there is a sound aspect. There's a sentence structure that's formerly called syntax, and also we analyze the word structure and that's called morphology, and also we analyze the social meaning of language. Right now, the field has more integration of other areas. We also analyze the psychological side of linguistics. And also probably you already have heard of CAT GPT.
Yunting Gu:There are also some linguists that work with computer scientists that help advance the models in natural language processing and large language models.
Dimitri Joseph:Cool. This field sounds very broad, and it seems like it can have a lot of applications from the computational world to the physical world, where we're communicating communicating with another, and how the impacts that it has on someone's psychology, etcetera. But just to take a step back and go back to what you said, your focus is on the sound in linguistics. What are the major sounds that are made in languages across the world?
Yunting Gu:Thank you for your question, and most people know that sounds can be categorized into consonant sounds and vowel sounds. And actually there is another way of categorizing sounds. And previously, I encountered a song that has an interesting title and lyrics, and I want to play it to you. To thank you this year.
Dimitri Joseph:That was a very lovely sound. I didn't get all the all of the lyrics, but it provided a certain tone that made me feel that put me in a certain mindset without me even knowing what exactly was being said.
Yunting Gu:Okay. So the title of this song is called sonority scales, and the first few lines of the lyrics is that sonority scales aren't enough to scream up your name to thank you, my love. And it goes on so now these scales won't reach this tune and every new word won't be enough to thank you this year, to thank you this life. Something like that. So do you guys know what does 'sonority scales' mean?
Yunting Gu:No. That's actually a linguistic term. Sonority basically means they inherit loudness of sound. And in linguistics, besides categorizing sounds into consonant sounds and vowel sounds, we also have a sonority scale that includes all the possible sounds of human languages, and that has many more levels or categories than just consonants or sounds. So it basically refers to the inherent loudness.
Yunting Gu:If we go back to the sound, it basically means synergy scales which encompass all the possible sounds in the world aren't enough to say thank you. So that's a basic metaphor that's used here, and I think it's interesting because that involves linguistic terminology.
Dimitri Joseph:Yeah. Yeah. Very, very poetic, and it it brings to mind a few different limitations about language.
Mari Dowling:I think that's a really, really artistic way of expressing thanks in a song. Could you tell us a little bit more about what you mean by inherent loudness in a sound before we get into how that
Yunting Gu:relates to your research? That's a great question. So just to give you a few examples, vowels are the highest or loudest in terms of this inherent loudness feature. And some consonants such as p, t, those have the lowest degree of this inherent loudness. And there are other sounds that are in the middle such as the l sound or the r sound, and they're kind of in the middle.
Mari Dowling:Do you have a a way of explaining what the difference is between, like, what defines a consonant and a vowel?
Yunting Gu:Actually, it's interesting that we never talk about this question in introduction to linguistic class, and we do have a way of introducing the sound aspect of languages. There is IPA, International Phonetic Alphabet, And behind this chart, basically, there is a logic is that there is this kind of a group of symbols that can denote all the possible sounds that a human make independent of the language that they speak.
Dimitri Joseph:Could you give us an example of a of a vowel sound and then give us an example of a consonant sound based off of that phonetic alphabet?
Yunting Gu:Yes. And if you are a English speaker, you can basically understand the vowels as the vowel symbols or vowel letters, and the consonant sounds are basically the consonant symbols if you are an English speaker. So maybe a or I, those are vowels, and or, those are consonants.
Mari Dowling:Given that there's this phonetic alphabet that describes these various sounds, I feel like that implies that there's some sort of quantifiable finite number of sounds that people can make. But is that the case, or do you think that there's more variation in the types of sounds that people can make that aren't necessarily covered by these phonetic alphabets?
Yunting Gu:That's a great question. And at some level, the answer is yes. At some level, the answer is no. So, basically, when we have a letter, there's an abstraction. So your pronunciation of MSU may be different from my pronunciation of it, but we have no problem understanding each other.
Yunting Gu:Because in our mind, there is this abstract, basically box of MSU pronunciations. And there are variations of it, but there is also some common things that make us understand each other. That's also the case for the phonetic alphabet I mentioned. Those alphabets also represent basically a box of sounds, and they cannot capture like all the realization of variations, but it can capture all the possible categories of those sounds.
Mari Dowling:So what are some of the categorizations of the sounds that exist? Is that, like, the vowels and the consonants and the snoriety? Or
Yunting Gu:Yeah. For instance, like, those are called stops. And, f, v, or s, and, those are called fricatives. And so those are different terms for that. And also the 'l' like sound or the 'r' like sounds, those are called liquids.
Yunting Gu:Within the category of vowels, there are also different types of vowels, such as high vowel, which is 'e', and then we have low vowels, such as a. And I also mentioned that sonority scales is relevant to all the sounds of human languages, and the vowels are the most sonorous. They have the highest degree of inherent loudness. And the stops, they have the lowest degree of inherent loudness. So fricative is louder than stops, and liquids are also louder than fricatives but have lower level of inherent loudness compared to vowels.
Mari Dowling:Okay. So it seems like all these different sounds that make up a language or your words fall into these various categories and have various aspects of them, not just it's a son or a sound or it's it's just that there's there's multiple aspects that make up a sound. And within a language, you consider these sounds to be part of multiple categories.
Dimitri Joseph:And now that we have this, this foundational understanding of sonority scales, what your focus is in with the linguistics. Could you give us more of an explanation of what exactly you study with sonority scales?
Yunting Gu:In my recent research, I track the lips and tongue movements of English speaker when they speak, and I found that for a syllable, the larger the inherent loudness difference between adjacent sounds, the larger the timing difference between the pronunciation of the two adjacent sounds. And that can be used to explain several common features of languages of the world.
Dimitri Joseph:This type of study seems like it requires very fine data points. How how do how do you go about collecting these measurements of lip and tongue motion and the measurements of sound?
Yunting Gu:Quick question. For my research, I first test this hypothesis using existing data from a corpus. The data was collected using x-ray microbeams, and the corpus is from the nineties. And, basically, they are x-ray scans, and there are some gold pellets on the lips and tongues of the speakers which track the movement of the tongue and lip while they speak. Because of the advancement of this field, we use electromagnetic articulography, which is EMA, and that have minimal harm to human beings.
Yunting Gu:We also have sensors on the tongue and lip of people, and there is a electromagnetic field that's next to the head of the speaker. And together with the system, we'll also record the speaker. And using this system, we are able to collect the data, and then we need to analyze that.
Dimitri Joseph:This is very scientific. It's this isn't just something you're doing by observation or to the common air. Based off of the study design or the brief study design that you just described, it seems like there can be a potential for, for bias where the findings that you have may be largely dependent off of the samples, the people, and their upbringing, and how they're used to enunciating or pronouncing specific words. From this data set, was there a variety of different backgrounds to kind of balance it and so that you can have a a more broad finding?
Yunting Gu:Yeah. The this question makes sense to me. And the current dataset I've already analyzed is from Wisconsin X-ray MicroPine database. So it largely have samples which are participants that are undergraduate students in Wisconsin. And just as you said, people from different areas do have different dialects, and there is variation in terms of the pronunciation.
Yunting Gu:And this kind of bias will always be there because right now we are collecting data at the MSU campus, and people might also say that it may not be representative of people from other areas of other age groups. But that's the starting point, and we always need some kind of starting point. And once we see some regularities, the next step is to replicate the study from people who study other varieties of the language.
Mari Dowling:I was wondering if you could explain a little bit more about what you're measuring in your study
Yunting Gu:for each syllable, I mentioned this concept of sonority or like inherent loudness. There is different inherent loudness for different combination of sounds. So that's one set of numbers or one set of data. And then another aspect of my finding is the timing difference between the two sounds. I found this positive correlation between the two measures.
Yunting Gu:So the larger the loudness difference between the two sounds, the larger the timing difference of the pronunciation of the two sounds.
Dimitri Joseph:Just so that I understand accurately, if there's a big difference between syllables, you know, within a word or within a a sentence, then we take longer to produce that sound?
Yunting Gu:That's exactly correct.
Dimitri Joseph:And so what's the implication of this?
Yunting Gu:The implication is that it can explain several cross linguistic patterns in terms of the sound of languages. So across different languages, even though there are different characteristics in terms of the organization of syllables, there are some common features across different languages of the world. So for instance, at the beginning of syllable, combinations such as pl is much more common than combinations such as pt, which is much more common than combinations such as l p, because we want rise sonority at the beginning of a syllable. And through the years, there wasn't a satisfactory explanation for this kind of observation. And the finding of my research could be relevant to that since it suggests that this kind of right singularity or this kind of larger loudness difference is related to larger timing a And the next step is to explain why.
Yunting Gu:It could be that that's relevant to some perceptual system or, like, the hearing of the speakers, or it may be it's just easier to pronounce things that way so that why that's more frequent.
Mari Dowling:So what I'm understanding is that these languages that might seem very different don't necessarily have that great of a difference in terms of their sound production and that there are these universals that tend to exist. And I know you said you're still looking at why, but do you have any ideas as to how that might have developed given that these different languages might be across the world or have never interacted with each other?
Yunting Gu:I don't have a answer that's backed up by scientific evidence, but my intuition is that that's part of the evolution. And recently, there is actually a paper published that's the first research that I know of that actually analyzed this rice sonority thing with mice. And it turns out that mice also have some sensitivity to this. So it's possible that this kind of universal it's part of human evolution, and also that's something inherent to the nature, but we're not sure yet.
Dimitri Joseph:Is there anything else that's connected to your finding about this pattern of the difference in sonority and the time relationship?
Yunting Gu:So right now, I'm evaluating it in English, and the next step is to collect some data on Chinese Mandarin and also exploring the pattern there. And the idea is that it can be replicated in other languages as well. Another potential implication or impact of this kind of research is relevant to fake speech because right now we can use computer to generate human sounds, and there was a research that they kind of use the understanding of speaking to detect whether a certain recording is generated by computer or real human. So, basically, they reconstruct the vocal track behavior of a certain recording, and they compare this reconstructed vocal track to human anatomy. And if that's not a realistic human anatomy, and then this sound recording is considered as a fake recording.
Dimitri Joseph:What are you expecting to see or expecting to find when you compare the Chinese Mandarin data versus the English data?
Yunting Gu:For the Mandarin data, the expectation is that the finding still holds true. So there is still this positive correlation, and that way we can generalize the finding from English to another language, and if we have these 5 or 10 languages as samples, they all seem to obey the pattern, maybe we can generalize the findings to all the languages. So, similar thing larger long distance difference relates to larger timing difference. That's the expected finding for the Mandarin experiment.
Mari Dowling:Given your findings on some universals between different languages in the sound production and spacing of these sounds, do you think there's ever going to be a way where we can come up with a, quote, unquote, universal language?
Yunting Gu:I think that's an interesting question, and there is a constructed language called Esperanto. And the idea is that it's the language of the world, but not everyone speak Esperanto. So I think that there will not be a one universal language of human beings because there is this critical period that one needs to learn their native language, and I don't think we can force infants to learn a specific universal language. And one message of your question is actually that languages can change, and that's definitely true, And there are languages that are dying. And also even if we look at one human being, the way they talk when they are older may be different from their speech when they are younger.
Yunting Gu:So to wrap up, there will be some common patterns shared by human languages, and there are many aspects of language that are believed to be innate to human beings, but there will not be this one language that reflect all the features.
Dimitri Joseph:It's not something that you mentioned, but I'm thinking that, I guess, it's natural to linguistic studies that this can be applied to people that have speech impediments or that struggle with communicating in a streamlined fashion or or communicating effectively because I guess with the realization that there are changes in volume or changes in in loudness or sonority, then once you realize that, then a person may be able to recognize the points in their speech where maybe they should slow down because that is what is natural to speech across all languages. And with that realization, that will help guide the way that they communicate.
Yunting Gu:Yes. That's a great point, is that our study on people without speech disorder can inform speech pathology, inform the studies that can help people recover from speech disorder. So that's the potential of implication of this kind of research.
Dimitri Joseph:Thank you, Yuting, for giving us this thorough explanation about linguistics and the patterns that are within our sound and just the the nuances to how we communicate.
Yunting Gu:Thank you.