Subscribe
Share
Share
Embed
In this episode of The Highlights, show host Thiago Tarraf Varella discusses his research on pre linguistic vocal learning in marmosets, and what this can tell us about human vocal development.
Brains, Black Holes, and Beyond (B Cubed) is a collaborative project between The Daily Princetonian and Princeton Insights. The show releases 3 episodes monthly: one longer episode as part of the Insights partnership, and two shorter episodes independently created by the 'Prince.' This show is produced by Senna Aldoubosh '25 under the 147th Board of the 'Prince.' Insights producers are Crystal Lee, Addie Minerva, and Thiago Tarraf Varella. This show is a reimagined version of the show formerly produced as Princeton Insights: The Highlights under the 145th Board of the 'Prince.'
Please direct pitches and questions to podcast@dailyprincetonian.com, and any corrections to corrections@dailyprincetonian.com.
SA: Hi everyone. My name is Senna and I'm an assistant podcast editor at The Prince. You're listening to the eighth episode of Insights: The Highlights. The Highlights is a collaboration between Princeton Insights and The Daily Princetonian. Insights is a newsletter written by Princeton undergrads, grad students, and postdocs, covering the most exciting and groundbreaking research being conducted here at Princeton, in the form of fun, short, easy to read reviews. We cover a range of topics including Psychology, Neuroscience, Biology, Computer Science, and Physics, just to name a few. Make sure you check out their website at insights.princeton.edu. I am joined today here with my co-host from the Insights podcast team, Sophia Villacorta. Say hi, Sophia.
SV: Hi everyone, really excited to be here.
SA: If you've been listening to insights for a while, you might notice that our usual host Thiago is not hosting today.
SV: This is because it's a special episode. Our guest is Thiago himself. If you're new to the podcast, Thiago Tarraf Varella is a PhD student in the Department of Psychology here at Princeton. Recently, his research paper “Cooperative care and the evolution of the pre linguistic vocal learning” has been included in the insights newsletter. His paper was reviewed by Sarah McFann, a PhD student and the Chemical and Biological Engineering Department at Princeton, in her review titled, “What other species can teach us about how infants learn to speak.” Welcome, Thiago.
T: Hey, thank you. It's super exciting to be here. I mean, I've been here for the other episodes, but normally on the other side, so I think it's gonna be a very good experience so that I'll be better when hosting the next episodes.
SA: That's awesome. So let's jump right into it.
SV: So I'm curious to know, what made you interested in specifically studying language development, at least for this paper?
T:I think that the problem, the specific problem of language development kind of came to me instead of me going to the problem. I found this lab based on how they were applying dynamical systems to behavior. And coincidentally, the behavior that they were studying was vocal development, which is actually an important distinction. We are not studying language development, but vocal, like it's prelinguistic, it’s pre language vocal development.
SA: Oh, that's awesome to hear. Um, could you please give us like a quick summary of like, what your research was about and what you were hoping to find in doing your research?
T: Yeah, I think that in very broad terms, I am trying to find similarities in how, so how behavioral changes happen. And that can be in different timescales. So for example, how evolutionarily, we started behaving the way we do, and how developmentally, we started behaving the way we do, and also in learning in real time. And this paper that you mentioned in the intro is actually an interesting example of an intersection between evolution and development, because we're basically studying how development influenced evolution. And I think that that's sort of what I'm trying to find out with my broader PhD program.
SV: Oh, yeah, that's, that's really interesting. I'm curious to know a bit more about how you went about working on your research and the collaborations that you made to make your paper possible.
T: I think that it was mostly via my lab. So when I came here, I started interacting with people from my lab, like that are postdocs, that are grad students. And this was the research that we were discussing, mostly. This paper is a review paper, it's not a it's not an experiment that I conducted. So, having access to all of this other research made by the people in my lab, put me too, in contact with all of these other papers that I mentioned, to be able to write this review. And, like make this connections. It was mostly my advisor, actually, it's a process like how I started with an idea and then he complements the idea.
SA: Yeah, that's really cool. Um, so like transitioning into the content of the paper specifically. I noticed when you decided to analyze the marmoset monkeys, you focused on three variables: altriciality, cooperative breeding, and vocal learning. Could you define these terms and their relation to your research, and why you chose to specifically focus on them when you were looking at vocal learning?
T: Let's start with defining what altriciality, corporate breeding and social learning and reinforcement learning is. So altriciality is the opposite of precociality. And precociality is when you're born and you're ready, while an altricial animal is when you're born, but are very dependent on your parents or other community members to survive, to move, to eat. Yeah, so this is altriciality. Cooperative breeding is when the infant is raised by other members of the family, and not just the mother. And finally, reinforcement learning. I think that initially, reinforcement learning is a computer science idea in which you try to teach an agent to learn something by giving rewards, which is the reinforcement. So whenever the agent does something correct, it'll get a positive reinforcement. And if it does something wrong, they either don't get any sort of feedback, or they get a negative reward. And somehow you make the agent learn through this process. We wanted to understand why humans learn how to make better vocalizations. And this has been a long term, long time debate in Psychology and Developmental Psychology. Initially, people thought that humans will learn through the process called maturation, which is basically just natural development of the body. But now we know that it's actually, there's a lot of feedback involved. So, the human babies will learn much faster and much better if there is consistent feedback from the parents during the process of the vocal learning. So, if the parents look at the baby and give enough attention to them when they vocalize something that is correct, like “mama,” usually when a baby says “mama” for the first time, it's like a whole party. [laugh] So this is one example of something that changed on the way that we understand vocal learning in humans. And this is essentially a reinforcement learning idea. So the fact that the baby said a word correctly, like “mama,” is giving the baby reward, But about the altriciality and the cooperative breeding, it was just an hypothesis. So we came up with this hypothesis, because the other nonhuman primates, like the squirrel monkey, and etc, they do not have, they are not as altricial as humans are. So for example, humans will increase the size of the brain like four times as they develop, from when they're born to an adult, right? While most of these other primates will only increase their brains by about less than twice the final size. So that shows how much less development, brain development there is after birth. And this is also an interesting idea for vocal development. Because, if you have this sort of feedback, while the brain is developing, this makes it much easier to learn. Because the connections that will form and the, the whole brain plasticity, which is the process of creating new connections in the brain, it's like very active, in, during this time, where the brain is increasing and being developed. So, that's where the altriciality came from. The idea for using altriciality in the process of vocal learning. And the cooperative breeding was one way that we found to join the two, the two issues of altriciality and reinforcement learning. Because when you have a cooperative breeding system, you do not have your mother giving you attention all the time. So you need to somehow elicit attention. And the best way to elicit attention is learned. So that's the main idea. So you will learn how to vocalize in a way that you can elicit attention from your caregivers that might be worried about other things because they’re not, their sole responsibility is not taking care of you. They have other responsibilities as well.
SV: Oh, yeah, that was a really great explanation of those terms. Could you kind of go over what makes marmosets so similar to humans or why you chose them for this experiment and any other possible similarities in relation to vocalization?
T: You asked about why marmosets. Sure. So I just explained on the previous answer about the hypothesis that we had that the, because of the cooperative breeding and the altriciality that kind of facilitated reinforcement learning so, how can I test this hypothesis of whether this is actually true? One idea is to find this process of vocal learning happening in another species. And fortunately, marmosets are one of the few mammals that do exhibit cooperative breeding. And another very interesting thing about marmosets is that this is not similar to humans, but they're always giving birth to twins, and sometimes triplets. And because of this fact, this kind of makes the marmosets being born more altricial than other primates, because, well, I guess there are mainly two reasons. First, the size, like the mother needs to carry the babies inside, their womb. So the bigger they are, the harder it is to deliver the babies. So if they are twins, you'll need to deliver them earlier, before they grow too much. And that is one factor related to being able to show. So marmosets are crucial and exhibit corporate breeding. So they are a great system for testing this hypothesis. And it's also interesting that they are not closely related to humans. So I believe that the marmosets' phylogeny, separated from, phylogeny is the evolution. So the marmoset evolution, separated from humans evolution, evolution about 40 million years ago, I think, while other monkeys separated from humans 25 million years ago. So the fact that humans, humans and marmosets exhibit, for example, cooperative breeding, and these other monkeys do not exhibit cooperative breeding, makes these traits, the corporative breeding, are one example of convergent evolution, which is one two species developed the same trait independently. And it's a very useful thing to observe. Because you can kind of study what are the, what are the incentive incentives? What are the constraints in the environment that led to the evolution of this specific trait. And because of all of these similarities, we decided that it will be an interesting example to study the vocal learning process, not in humans. Well, to test our hypothesis about the vocal learning, the evolution of vocal learning in humans using the marmosets.
SV: So I know that, speaking of vocal learning, I know that you mentioned that you brought an audio of the marmosets.
T: Yeah. So this is one example of immature vocalizations. So like almost, I think it was one day after they were born.
[immature marmoset vocalizations]
T: You can see that it sounds a little bit like a cry sometimes. So here's one example of the mature vocalization that it made, I think it's actually even higher pitched.
[mature marmoset vocalization]
T: The mature vocalizations are a lot more tonal, while the immature vocalizations have, they're not as like, they're not just one tone, but rather, it's either like very dissonant, or sometimes they have some, some vocalizations that sound like birds. This just shows how much the vocalization changes from when they're born until like, I think that this last example that I showed you is about two months in.
SA:Oh, wow.
T: Yeah.
SA: Oh, that's really interesting. So kind of shifting to the results of the paper. Could you tell us why the results are important and then like possible, like future shifts in your research you want to like take from this specific paper.
T: The main result, essentially is, so we wanted to test the hypothesis of whether humans, whether altriciality and corporate breeding are related to the vocal learning process. So we found that marmosets exhibit otter chirality and corporate breeding so the big question is, do they also exhibit vocal learning? As you saw in the vocalizations, their vocalization changed a lot. But this is not necessarily a matter of learning. It could be just a maturational process. But there are other papers that actually show that this is in fact, a learning process. This can have many broad implications within my own research project. I'm just trying to understand behavior in general and formulating these relationships between evolution and development. So this is one example of a behavioral trait, vocalization, that evolved to be more complex due to the way that it is being developed. So altriciality and cooperative breeding are things that are related both to genetics and their environment, like how they are raised. And both are very much relevant to the development of the individual. But this also has some other broad implications. For example, when we try to understand autism, one way that we characterize early on that one human might be part of the autism spectrum, is basically they start vocalizing and showing mature vocalizations a lot later. And in some cases, not even at all. So studying this process of how the vocalizations develop, could also help with understanding the onset of autism. And also, there are some disorders as well like some speech related disorders that could be related. Having a more, better understanding of why the vocalization develops at all is helpful later on to understand why some individuals are not developing their vocalizations as another typical individual would develop. And in terms of next steps, I am currently studying more about the development, so there is a paper that should be out anytime soon, hopefully, that is how both the energetic costs of vocalizing and the amount of information that is transmitted in a vocalization. These are two factors that are balancing themselves during the process of vocal learning, and vocal development in general. So basically, the idea is that when you're young, you try to optimize the vocals, the amount of energy that you're using when you're vocalizing especially in marmosets that vocalize very loudly and vocalize a lot and they are very small. The energy that these vocalizations require is proportionally big. So most of the vocalizations will be very energetically efficient. While when they grow up, this energetic constraint is not as strong as it was. And then they are able to vocalize, to elicit vocalizations that have more information that contains more information and are more able to transmit whatever type of information that they need to transmit, like their age, their gender, their health status, that sort of thing.
SA: Gotcha. Well, this was really fun talking about marmosets and vocal learning and everything. And thank you so much for joining us today.
T: Yeah, yeah. Thank you. It was, it's always very nice to talk about your research. And it was very fun being here.
SA: And thank you to Sophia for co hosting this episode.
SV: Yeah, thank you for having me.
T: Nice. Thank you both.
SA: This episode of highlights was written by Senna Aldoubosh and Sofia Villacorta. It was produced under the 146th Managing Board of the Daily Princetonian. For more podcasts and other digital media of the Prince visit the Daily Princetonian website. Many thanks to Thiago Tarraf Varella for speaking with us today. To read more about his work, check out the Princeton podcast Insights article covering his research which can be found in the description of this episode. Thank you for listening
Transcribed by https://otter.ai