BẢN CHÉP ÂM THANH 🔗
Speaker 1 0:00
We've got these changes over time that lead to differences in music across different cultural traditions of the world, and it's not just music that changes between those different traditions, it's actually perception and experience of music and the pleasure that we experience that varies across those different musical traditions.
PJ 0:22
Hello, and welcome to Chasing the Viaton. I'm your host, PJ Weary, and I'm here today with Dr. Marcus Pierce, reader in cognitive science at Queen Mary University of London, and honorary professor of neuroscience at Orhaus University, and we're talking about his book, Learning to Listen, Listening to Learn: Music perception and the psychology of enculturation. Dr. Pierce, wonderful to have you on today.
Speaker 1 0:46
Thank you, PJ. It's great to be here.
PJ 0:49
So, Dr. Pierce, tell me why this book.
Speaker 1 0:53
Well, music is all around us, it's universal across cultures, and it's actually uniquely human as well, and yet if we compare it to something like language, it's kind of hard to see why it's so universal. Why is it everywhere? Why do we find so much pleasure in listening to music, and that's really the question I think underlying the book. Why is it that music is so pleasurable,
PJ 1:22
but that's interesting. How important to you was that distinction between language and music, in the way that music is maybe more difficult to define or to track how it, how we learn it.
Speaker 1 1:38
So, I think as scientists we know we're attracted to mysteries, you know, we want to kind of figure out what's going on, and I think music is kind of a mystery, you know, it's everywhere, and yet it's, you know, it's kind of ephemeral too. So, so what I want to do in the book was to try to try to figure out what is it that motivates us to listen to music, we know where does that pleasure come from, and one of the, one of the starting points I think was to, if we're going to understand pleasure, then first we have to understand perception, how we perceive, how our brains make sense of musical sounds, because it's a really complex thing, there are lots of things going on all the same time, and the brain has to kind of figure all this stuff out, and it has to do it in real time, right? Yeah, so this is not an easy task for the brain, and, but rather than challenging, you know, we find it really pleasurable. So we need to figure out perception, and then the other sort of starting point for the book really was that existing theories of music perception are very much kind of rooted in sort of Western musicological tradition, and you know, I think that's understandable if you look at the context in which they were developed, and they've also made great contributions to our understanding, so, but I think it's what's missing is the diversity that we see across the world, I mean, if you just look at music, there are an enormous number of musical forms and variety across lots of different dimensions between musical cultures, and so I think we have to explain that diversity, and it's the diversity we can, we can, we can look at it in two ways, I think. So, one is just looking at across cultures today, as we witness the musical diversity, and the other thing is looking at changes over time, and even within the musical culture. If we take, you know, sort of European musical culture, which has quite a long history of recording, so certainly we have lots of records available, and we can see massive changes in in music over the past sort of several 100 years. So we've got these changes over time that lead to differences in music across different cultural traditions of the world, and what I think, in the what I try to do in the book, is address the point that it's not just music that changes between those different traditions, it's actually perception and experience of music, and the pleasure that we experience that varies across those different musical traditions, and excuse me,
PJ 4:19
I'm sorry, the pleasure changes,
Speaker 1 4:23
right?
PJ 4:24
That's interesting. Yeah,
Speaker 1 4:26
and I can, I mean, I can get into some specific examples of this. Yeah,
PJ 4:30
yeah,
Speaker 1 4:31
it's so well. For example, this is quite, quite a simple example is, is dissonance constants and dissonance, which, which we, when you put different pitches together, then some combinations sound pleasurable, and some don't, and this has been thought to be a kind of universal property of the auditory system. It's turned out over the past. Of decades we've got several results now showing that actually that's not the case, so it varies quite significantly if you look across different cultures. One example is the Chimani of Bolivia, who live in the Amazon rainforest, who've been studied by various people, and it turns out that their experience of consonance and distance is very different from from North American experience, for example, they showed shown no difference in pleasure for tone combinations that probably you and I would experience as being dissonant and unpleasurable, so that's a really concrete example, and we've got other examples of listeners from Pakistan, for example, or Papua New Guinea, where this is also the case, so they showed different patterns of pleasure, and it looks like those patterns of pleasure are related to their experience of the music of their culture, so this is this is sort of the the entry point of the book. If we think about music evolving over time within cultures, and then different cultures kind of branching out to different endpoints, the listeners within those cultures have to be able to learn the structure of the culture, and also then do this in a kind of dynamic way, each generation, because the music is changing over time, they have to be able to, you know, accept those changes, learn them, and then, and then find them pleasurable, or learn to find them pleasurable, and you know, if the
Load full transcript
music changes too much, then the audiences will reject it. I mean, we've got good examples of this in Western classical music, for example, where you know composers kind of take a step too far, you know, Stravinsky being a good example, and you know, and then audience don't, don't accept those changes, and yet over time those pieces become standards of the repertoire, so listeners take some time to kind of accommodate changes in musical structure over time, so really, and that's what I want to understand in the book, is that process of learning that leads to perception of musical structure, and then, and then I can explain it in a bit how that links into pleasure.
PJ 7:28
Part of the reason I was interested in your book is I was in college, and I had a friend who'd grown up in Cambodia, and so I mean, this is, you said that this research just happened in the last 10 years. This is a little bit further than 10 years ago, and we were talking about music and beauty and music, and she grew up in a western household that moved over to Cambodia, but she talked about the experience, she most of her childhood was there, and the difference of that, the different, like tonal resonance, dissonance, harmony, and the funeral music. I went back and listened to it after she said the funeral music, they play it, and the people just weep. It's so clearly connecting to them. And, of course, to me, it sounds like I sounds like my six year old hitting a steel drum, right? I can't, I can't parse it at all. So, of course, it takes - it's a lot easier to have an anecdote then than to do all this research. So, I want.. that's why I'm into.. part of the reason I'm interested in.. because it's like, okay, what is the scientific backing behind this? It's really fascinating to me.
Speaker 1 8:38
Yeah, so I think this is a really good case in point, and so my approach to this has been to to build a computational model using AI methods that that learns the structure, the syntax, if you like, of a musical style through exposure, and and then over time it comes to simulate the perception of a kind of mature listener within that culture, and the two kind of hypotheses, I suppose, the psychological hypotheses underlying the model are firstly statistical learning, which is this idea that we have this, there's a kind of obligatory process that's implicit, and you know we can't control it. Our brains just absorb the regularities in the sounds that we hear, and they find structure in them, and they build internal models of that structure, as they do with language, for example, but also with music. And then the second psychological principle is probabilistic prediction. So, see, the idea here is that these models are used to generate expectations for what's going to happen next in the music. Well, while we're listening, and again, this is a kind of obligatory, implicit process, and the reason that that we generate expectations well is useful from a survival perspective. You have to be able to predict what's going to happen in the future, and this is particularly true, I think, in the auditory domain, where sounds are changing rapidly in time, so that there's a kind of link to general evolutionary psychology there, and so part of the first part of the book really is looking at the modeling process, so building the model that learns this structure, internalizes the regularities, and then is able to generate predictions for what's going to happen next, and then really the rest of the book is then evaluating that model, so we can, you know, we can give the model pieces of music and see how it behaves, and then we can see whether that behavior corresponds to listeners when they present them with pieces of music,
PJ 10:48
and now when you talk about modeling this with AI, is that is the model of AI you're using, not the model I need, the is it, is it an AI name that we would recognize, or is it specific to what you're doing, like your scientific field?
Speaker 1 11:06
So, it's AI, I suppose. It's AI in the, in the general sense of the term, as in it's part of the field of AI that's been in existence since, since the 50s, I guess. So, it's not a neural network model, that's it's a, it's a probabilistic model.
PJ 11:24
Okay,
Speaker 1 11:24
and the reason it's a probabilistic model is, is really for reasons of interpretability. So, it's a neural network models, are you know, are more powerful, but they also require more data, and then they're harder to understand exactly how they're doing what they're doing, whereas a probability is
PJ 11:42
that what they call like the black box effect with some of the okay, sorry, tracker with you. Okay, please continue.
Speaker 1 11:50
Yeah, exactly. So it's it's difficult to interpret exactly the kinds of representations that neural network is learning, but whereas with a probabilistic model, it's all from a scientific perspective, kind of, we can analyze it and look into what's going on in a much more, much easier way, which is helpful scientifically, but the kinds of predictions that it's generating are really, I mean, kind of analogous to the kinds of predictions that a transformer model would be, like, you know, the large language models that we were all using these days, it's a very similar process, it's, you know, it's, it's going through a process of statistical learning, learning the structure of the data that it's exposed to, and then generating predictions while processing new sequences. So it's conceptually very similar
PJ 12:33
when we talk about survival, and I don't know what to call.. I think immediately we have this auditory example of its survival. We think about the person in the woods who's like, is that an interesting twig snap, or is that a, you know, or a boring twig snap? Do you also think, and I wonder, when we talk about rhythm, we talk about harmony, and what music does, do you think that this idea of learning to predict what other
people will do is part of social survival from an evolutionary psychology standpoint, this idea of learning to predict how other humans will behave, because that's one of our big evolutionary advantages.
Speaker 1 13:12
Yeah, I'm pretty sure it is. It's in the book. I don't really go into social listening to music, so it's not something that I've kind of covered in detail, but yeah, I think it absolutely is. Yeah, I mean, I think prediction is just an extremely important tool, really. You know, if you're able to kind of learn from your environment and then predict what's going to happen next, now you're from a from an adaptive perspective, an organism is at a great advantage, and that, that applies to the social realm, just as it does to the environmental realm.
PJ 13:46
That would obviously be probably another bigger step than what, like, that's that's a whole nother work. And you're providing kind of the next step before that, the foundations for before, like, you know, understanding individual psychology is important as well, right? But there's also that social side, which seems to be from what you're, when you talk about enculturation, is a big part of what music does. Where am I? Am I okay? Okay.
Speaker 1 14:10
No, I think it is important. I think it is, and I think it's very important, but it's, I guess you're right. So, what the book does is like the foundation for maybe, maybe that'll be the next book.
PJ 14:22
Yeah. Well, I hope you come back for that one too. No. And now underneath computational model, and again, and I appreciate your patience, because I am far, you know, I - my background is in philosophical hermeneutics. I am far afield here. So, you've been very gracious and kind. What is the Markov assumption, and why do we need to get beyond it? Right,
Speaker 1 14:47
so the model is trying to predict what's what's going to happen next. It's listening to a piece of music, and it's got a context of the existing music that has heard so far, and it's trying to predict what the next event is going. To be know what the next note or the next chord or whatever, and in order to do that, it has a has a kind of model of which events have followed similar contexts in the past, and and it has a has a way of kind of putting all that learning that it's that it's that it's done over sort of similar contexts into trying to estimate what the, what the probability is of different next possible events. The Markov assumption is that the prediction only depends on that immediately preceding portion of the musical sequence, so it doesn't, so what it can't do is say, oh, I'm going to disregard what I've, what I've seen here, and actually, you know, five minutes ago in this piece that there was something happened that actually, and that's what determines what's going to happen here, so you have what they're called non-local dependencies, and that's the kind of thing that happens in it, happens in language, it also happens in music, and there's a question in music of how to what extent listeners are representing those non-local dependencies. I think they are to some extent, but perhaps not to the same extent as in language. It's an open question, and this is something that transformer models are quite good at picking up on those with given an enormous amount of training data has to be said but that they can pick up on those kind of non local dependencies whereas whereas idiom this the probabilistic model can't do that explicitly at least the surprising thing is, so the, as I said, the model's interpretable, and it's relatively simple in some ways, and what's what's interesting about it is that actually, as the results in the book show, you can get a long way in explaining perception of music with actually a relatively simple model, so then probably we do need some extra power in terms of capturing these non-local dependencies, but perhaps it's not as important as some people might say think it is.
PJ 17:12
So I'm trying to follow you there. When you're talking about non-local, are we talking about non-local, are we talking about the kind of frame, like cultural framework of music, or are we talking about within the piece, further back in the piece?
Speaker 1 17:26
No, just within the piece.
PJ 17:27
This is, oh, just within the piece. Oh, okay. Okay, because obviously there's some training, and that's going to do a lot of work. Okay, so we, I just took my kids to see Beethoven's ninth symphony, and so obviously he's going to start themes way back here, and then he's going to bring them, and so that would be when it's like 1015 minutes earlier, that's that's a non-local, that's a
Speaker 1 17:50
non-local, yeah, or for example, if a piece starts in one key and then modulates into another key, then you know it has to kind of return to the same key at the end, for example, that would be an example of a non-local dependency. It's interesting you bring up thematic structure, though, because the one of the.. I think one of the interesting things about the approach is that it accounts for not just kind of low-level aspects of perception. So, if we give people melodies to listen to, and we say, you know, how predictable do you think this melody is, or how complex, how complex is it? Then the model simulates those judgments really well, so that that will be low-level perception, and the same for harmony, it simulates perception of predictability in harmony very, very well, but it also captures kind of higher level, what we think of as cognitive processing of music and thematic perception of thematic structure. I think is relevant there. So we've done experiments where, where we get people, for example, we get two themes from from a piece of music, so a theme and a kind of restatement of it that's imprecise, it's not an exact repetition, and we say to what extent do you think these two come from the same piece of music, and some of them, you know, some of them are
quite similar, and some of them are not very similar. The model accounts for those those judgments pretty well, actually. The other thing is, if we look at a longer time scale, so you have pieces of music that are sort of 234 minutes long, and we take a passage in the second half. We say, do you think this passage is heard before? Is it a restatement of a theme that you'd hope you've heard before? Actually, the predictability of the passage for the model that's been kind of been trained incrementally on the piece, it predicts the listeners' judgments. It also predicts their sense of the overall unity of the piece. So, I think, I think this suggests that to model perception, actually, maybe we don't need these, these sort of very kind of structured grammar-like models of. Of music perception that they can maybe be modeled in a, in a sort of simpler way, using this, this kind of dynamic model of learning, that there's this just, it's just kind of learning regularities in an ongoing way, dynamically, whatever it's exposed to,
PJ 20:16
so the Markov assumption is that it only is local, or is it that it needs Markov, is only local, and then when we talk about beyond, we're talking about that non-local, yes,
Speaker 1 20:27
yeah, and this is, there's a kind of debate about this in the field, but whether, whether, whether listeners do have these sort of kind of hierarchical models of non-local dependencies, and I think that the approach I've taken in the book is just kind of start simple, start with a simple model, and see how far you get with it, and then if you need to make it more complex, then you can, so you can build in these non-local dependencies if you need to, but I think the results we've got so far suggest, you know, maybe it's not as important as we might have thought,
PJ 20:57
would, and I, this is me coming from my background, so this might be just a completely ignorant question. Is it possible that it depends on the listener whether the you have to go past the Markov assumption? Because when I look at the conductor doing Beethoven's ninth symphony without music, I assume that he might need, would that be? Is that I mean, again, this is outside necessarily what your experiment was, but that seems like there might be an interesting direction. Am I right in thinking that?
Speaker 1 21:25
Yeah, yeah, absolutely right. So, there are different ways of experiencing music, and if we, if we sit down and have studied a piece of music, committed it to memory, perhaps you know, we know how to perform the piece or parts of it, then I think that's a different, a different kind of experience of music to the experience that you might have, you know, you sit down and you're just listening to it, to a piece of music that you haven't heard before, or even, you know, even, even if it's, if it's familiar, that's kind of different thing, but I think if you're, you know, musically trained and you're analyzing the music explicitly and consciously, that's, that's quite a different thing, I think, from the kind of musical experience that most people have most of the time, and I think really what the book is about, in the most part, is about that the musical experiences that people have, most people have most of the time, which is, which is sort of non-musicians just enjoying listening to music,
PJ 22:20
yes, yes, and that, that makes total sense. And I could see immediately how you have, like, there's listening to music, there's playing music, and then I'm sure even composing music is like, I don't know what the mathematical model would be for someone like Beethoven's brain, but I assume that would be, maybe, maybe it would just be probabilistic, but it seems like that would be a little bit more, more complex, I think it would be
Speaker 1 22:42
more complex. Yeah, and I come to this. Well, actually, I come to the.. there's a chapter on musical training, actually. And perhaps we should talk about pleasure first, and then we can complete question musical training, because there's an interesting connection between the two. Okay,
PJ 22:57
yeah. So
Speaker 1 22:57
the connection with pleasure is that we have, we enjoy listening to music, and different people enjoy listening to different, different kinds of music, and we want to try to understand what are the musical properties that produce pleasure. We think that there are lots of things that go into a pleasurable experience of music, and some of them have not that much to do with the music, they're more to do with, you know, one's experience of music. So we all have pieces of music that, you know, remind us of particular times in our life, and so that, you know, by evoking those kinds of memories, then you know we find them pleasurable. And or sometimes we just, we've heard a piece of music in a particular context, and it, you know, it kind of acquires the emotional character of that context, so if it's a positive context, then you know that makes us gives us a positive feeling, for example. So those kind of characteristics of musical pleasure, I think, are important, but they're somehow kind of idiosyncratic to each individual, and what I was more interested in with the book is what might be common across people, and a key kind of concept here is that there's a relationship between the pleasure we experience when listening to music and predictability, so this process of prediction that we've talked about means that some pieces of music are predictable and some are unpredictable. In fact, I mean, some passages within particular pieces of music are predictable and some are unpredictable, and composers sort of play with this process by generating unpredictable events and passages, and then in order to create a sense of kind of unease or tension, and then you know bring back sort of familiar or expected elements that reintroduce a sense of resolution, so you know the music takes you on a journey, and this is one of the
techniques that composers use in order to. Produce that journey, so the so what we, what we hypothesized, and this is a kind of long-standing hypothesis that actually goes back all the way to the beginnings of experimental psychology in the 19th century. Wilhelm Wundt was a German psychologist, proposed that the kind of intermediate degrees of arousal would be pleasurable, so intermediate degrees of complexity would be experienced as pleasurable, and this has come back into various forms over the over the years since, including most famously by Daniel Berline, and it's, it's proposed as a kind of inverted U-shaped relationship between complexity, so if you think about complexity on the x axis and pleasure on the y axis, and think, then think about an inverted u, so that what that means is that an intermediate degree of complexity is experienced as being most pleasurable, and we've run quite a few experiments now, and found actually that is the case. So, if we characterize using the model that I explained, pieces of music in terms of their predictability, then we get people to listen to them and say, How pleasurable do you find listening to these pieces of music? Then we do find this inverted U-shaped relationship, so people like things that are, you know, a bit predictable, a bit unpredictable. Sorry, but you know, not too unpredictable, because I guess too predictable is somehow boring, and too unpredictable is too complex. So this is this is interesting, because it starts then to relate perception of music to the pleasure, which is kind of where we started. The question really is, why? Why do we have that inverted U-shaped relationship? And there are several different theories in the literature. The one that I'm most partial to is called learning progress, and the idea is that whenever we're listening to a piece of music, and let's say it's an unfamiliar piece of music, for the sake of argument, then our brains are trying to figure out how the music works. What are the kind of, what are the patterns in it? What are the regularities? Can we kind of relate it to existing things that we've already heard in our existing model that we talked about of musical structure, and in that process there are two things going on.
Speaker 1 27:32
The first thing is that prediction, or predictability, is rewarding, so we like things that confirm our predictions, and the reason we like them is because they confirm our model, so a model that generates accurate predictions should be rewarded, so there's a process of reward that that is positive for events that are predictable, however, there's there's also a sense in which learning is pleasurable, so and the reason is I think you can think about this in different ways. One way I think about it is from a sort of foraging perspective, in terms of exploitation and exploration, so we want an agent in an environment that, where things are changing over time, you want to reward an organism that finds resources and then exploits that niche, but the resources might be finite, or the context might change, and so you also want to reward exploration, learning of exploration outside that niche, so that then that builds in resilience, I guess, so that then if the resources that no food or what have you do disappear, then it has a plan B, and I think when listening to music, these two things are going, and I think, as I said before, this is a kind of obligatory process, not something we can switch off, it's kind of very deeply hardwired, ingrained in us that learning is rewarding and predictability is rewarding, and so if you think about a piece of music that has moderate unpredictability that kind of balances the reward from learning with the law of the reward from prediction confirmation, so if it's if the music is too complex, then there's no prediction confirmation, but also there's not really any learning, because it's hard to form a model if the music is too simple. Then there's reward from prediction confirmation, but there's actually no learning at all. And that's, you know, we, we all have pieces of music that initially you think, oh, that's quite catchy, but then it kind of gets stuck in your head and becomes an earworm, and then it's,
PJ 29:58
yeah, it's
Speaker 1 29:58
no longer pleasure. We will have that experience
PJ 30:02
with,
Speaker 1 30:03
yeah, go ahead.
PJ 30:04
I was gonna ask, so is this similar to the idea of the zone of proximal development?
Speaker 1 30:11
I think it is, by the sounds of it. Could you, could you explain exactly what you mean? So,
PJ 30:15
this comes out of child psychology and pedagogy, the name Vygotsky can't remember his first name right now, but this idea of there's there's a threshold, and this would actually go back to what we were talking about earlier with a composer and the players, and then even children, right, like if we sat here and like twinkle, twinkle, little star, our audience is already the listeners are like, please don't do that, that's, you know, that's that's so, but for a child, that's pleasurable, because the child's ability to learn the music is, is less right, there's there are the skills we've mastered, there's the skills that we have no hope of mastering, and then there's the skills that we can master with help, and that in that, in that sounds that zone of where, oh, that's something I can learn, is just beyond me. Yes, it sounds a little bit like what you're talking about, and it gives us, you know, that's why children love things like Baby Shark, and adults don't, right? Like, I mean, this is kind of, and this is why, you know, Stravinsky took off amongst composers and musicians, and the general audience was like, why are you doing this to us, and some of that is where their level of enculturation, maybe, and I might be jumping too high, there is, but is that, am I tracking with you there?
Speaker 1 31:33
Exactly, exactly right, and it also relates to the
concept of flow, which you may have come across in, which is applied in lots of different areas, but in sports psychology and gaming, and so generally I suppose the psychology of optimal experience, the idea that you want to be in a, in a space where you're, where you're challenged sufficiently to kind of, that you're experiencing progress, but not so much that you just can't do it, and not too little that you're bored. It applies exactly developmentally, it applies in lots of areas, so it's exactly the same concept. Yeah, and it's interesting, actually, that you bring it back to that to cultural evolution, because I think there are there are variations in this sort of inverted U-shaped function relating complexity and pleasure between individuals, and that's something we can talk about, which will bring us back to the question of musical training, but there are also changes over time, and that's I think what is happening, exactly what's happening when pieces of music, you know, composers are continually striving for for novelty, and because they want to produce pieces of music that are interesting and novel in striking ways, and sometimes, sometimes they kind of go beyond what their audience can cope with and yet over time through this process of learning that successive generations of listeners go through what was once kind of beyond the optimal peak of complexity becomes actually the norm and so I think I think this this process of learning and finding intermediate degrees of predictability pleasurable can account for a process of change in cultural evolution of musical styles.
PJ 33:34
If you don't mind, I'd love to ask about a slightly different example, and this is again anecdotal, but it's from people who are trained in the industry. One of the things in film that's become apparent over the years is that the audience is becoming more visually astute, more visually skilled. So, jump cuts are faster, there's less - you had to provide more time for audiences to realize what was happening in the past, and as you watch the development of film, and this has come from people who do it right, not necessarily scientists, I'm sure someone's done something, or you know they're about to, but it seems to be the same thing, where you know people seeing Looney Tunes in the first time for the first time in in theaters in the 40s versus kids today, and the way that you know they are used to jump cuts, they've always had jump cuts, you know, from one thing to the other, and they can have a successive flash of images, and they can interpret it. Would that be analogous to what we're talking about?
Speaker 1 34:32
Yeah, I think so. Exactly the same thing, it takes time for audiences to kind of catch up and kind of learned the structure of their perception environment, and that applies to the cultural domain, just as it does to the environmental domain.
PJ 34:49
And let me ask you this, and I, because I didn't want to miss this, this is from earlier, and it's been kind of sitting on the paper here, and I really, but if I understand you correctly, when you're talking about making a simpler model that. Is interpretable versus using one of these big, you know, kind of black box cognitive networks. What's that neural neural networks?
Speaker 1 35:10
Neural networks
PJ 35:12
that, because the this statistical model is works that allows us to have, and I, we, you didn't use this exact term, but the multiple viewpoint representation it allows us to go cross culturally, that leaves it more interpretable and simpler, and it still works, and so it allows us to move cross culturally, and so that's a big part of your work here, that's like a great value that that actually actually does work, right? Am I in right? And understand it like I think that was kind of like hinted at, but I wanted to make that explicit. Am I understanding it correctly?
Speaker 1 35:45
Yeah, that's absolutely right. I mean, just to be clear, we are also doing, doing this similar work with neural network models, and then trying to figure out, trying to figure out how to interpret them, so that's it's not that that's not a valid approach, but we know we have existing models that are interpretable and seem to actually capture a lot of perception pretty well, so in a way we're using those models at the moment, or I guess you know over the past sort of several years when we've been doing this work, we've been able to go further experimentally than we would have done with a neural network model, so I think both things are interesting and both have kind of complementary pros and cons, and one of the things we can do, actually, is use, you know, use the probabilistic models to try to help interpret what neural networks are doing.
PJ 36:28
Got it? Okay. Thank you. Yeah, I was.. I didn't want to miss that. That seemed like just a really important link in the chain of what we were talking about there. Yeah, and I wanted to just to clarify, for like a, as a concrete example, and I want to make sure I'm tracking with you when you talk about the idiosyncratic pleasure, because you're specifically talking about complexity, you know, too much complexity, too little complexity, and then you know the Goldilocks zone of complexity, but that kind of that idiosyncratic pleasure could also be communal, so for instance, however good, like the song may or may not be, if it's for instance an anthem, like a national anthem, people will have a pleasure attached that has nothing to do with the complexity, and same thing with the earlier example of the funeral music, like people are going to start crying because it has no matter what the complexity is, that so that there are these broad trends that we attach to it, but that's that would be the, you know, the possibly the next book, right? That's more like the social attachments we make, I think. So that's what that means. Okay,
Speaker 1 37:40
yeah, and I don't want to sort of
downplay the importance of those processes, but it's yes, this book is not about those things, but I think they're also important.
PJ 37:52
Yeah, so I wasn't thinking you were saying that, I just wanted to make sure I understood what you meant by idiosyncratic pleasures. Okay, thank you, that's really helpful.
Speaker 1 37:59
I mean, I can explain a little bit more about idiosyncratic pleasure, because then it brings us also back to the question of musical training that was raised earlier. So,
PJ 38:07
absolutely,
Speaker 1 38:08
so we have this, this inverted U-shaped function that that we think is kind of related to these reward-related processes for for learning and prediction, and what we would, we, so we know that different people differ in their perception of complexity. So, what we might predict, what we have predicted is that musicians, for example, would have a peak at a higher level of unpredictability, so they would prefer the more unpredictable music than non-musicians say, and we have some evidence that that's true. However, there are also just quite a lot of individual differences in that individual that peak where it lies, and so one of the things we've been trying to do is, and this is this is just within a culture. We've been trying to figure out what determines that those those peaks. Why is it that some people have preferences for greater complexity than other people? And in that work we've done, we've looked at personality factors like openness to experience, we've looked at age, we've looked at creativity, we've looked at cognitive factors like preference for complexity in general, we've looked at music reward, and what we found is really that none of these factors kind of reliably predict this peak, that was not what I was
PJ 39:47
expecting you to say, right? Sorry, go ahead.
Speaker 1 39:51
This is not what we've expected to, and I mean it's possible that we, you know, we need larger sample sizes, or the effects are just quite. Small, but, but we've struggled to find anything consistent that really, you know, we've looked at hundreds of participants, so to find anything really that predicts these, these differences. What we do know is that they, they seem to co-vary with stylistic preferences to some extent, and which sort of makes sense, I guess. There's someone who, whose favorite musical genre is classical music or jazz music, you know, seems to have a kind of higher peak unpredictability than someone who prefers listening to pop music, something like that. But that doesn't really explain why, because it could be it could work both ways. It could be that you know by listening to that kind of music, they, their pleasure, you know, they develop pleasure for more unproductive music, or it could be that there's some kind of pre-existing preference for complexity, and we don't really know which of those is the case, or indeed both of them.
PJ 40:57
Correspondence is not causation. Let's see, right. It's right, so, and I want to be respectful of your time, but I do want to, is that, that, that was the musical training, like that we'd kind of gotten to there, correct?
Speaker 1 41:16
Yeah, I think so.
PJ 41:17
Yeah, yeah, so I mean, it's not like we're going to be able to dive deeply into your book, people. If you want to know more, read Dr. Pierce's excellent book. But the.. when you talk about the emotion of it, how do you.. because that's kind of the final kind of culmination of the book, if I can put it that way. What does this tell us about emotion, and how it's connected with music.
Speaker 1 41:44
So, emotion is one of the reasons why many of us listen to music in many cases, in order to either regulate our mood or empathize, perhaps. So, it turns out it's complicated. There are lots of different mechanisms for the way in which music can elicit an emotion, and expectation that we've been talking about is one of those mechanisms, but it's not the only one, and I kind of alluded to this earlier when talking about nostalgia, or sometimes music just kind of evokes memories for events in the past, and there are other, other mechanisms involved, like, yeah, the social mechanisms that we've talked about. So, there are, I think, lots of things going on. One of the interesting things about, about listening to a piece of music is that, you know, that there are kind of maybe seven or eight thing mechanisms involved, all of them could be at play, you know, all at the same time. If you're listening to a familiar piece of music, which, which can produce a kind of really kind of complex, nuanced emotional experience, and I think that's perhaps, you know, one of the reasons also that the music is so powerful with expectations, specifically, you know, we think that this, these, these processes of prediction learning are associated with, with reward, so they're associated with, with pleasure, and that's because of their relationship with survival. They're also related to arousal, so more complex things will tend to put us, or unpredictable things will tend to put us into a state of higher arousal, and we generally think of emotion as varying along two dimensions, perhaps more, but at least two dimensions: arousal and valence. So when, if we're listening to a piece of music that's varying in terms of its predictability, then that the music is going to be moving us through this valence arousal space into into different states of positive and negative valence and high and low physiological arousal. What we've, you know, we found that that unpredictable music does increase a does increase arousal, and that's perhaps also related to that, to this inverted U-shaped relationship. In that, generally speaking, we know we want to, and this is context dependent, depends on what we're doing while we're listening to Museum, but generally speaking, we want to try to maintain a, like, a moderate level of arousals, so that's kind of tailored to the
task, so if we're too aroused, then we're going to be agitated and not effective at whatever it is we're trying to do, and if we're too under aroused, then equally that's the case, so so I think that there's an emotional context also to this process,
PJ 44:38
and if you, and as we see, if you're just way too, if it's just way too stimulating, you can even go catatonic, right? Like that's where I don't think that's necessarily something that happens with music, but that's that, that's kind of how that plays out. I forgive me, I don't know what the word valence means in this context. When you say valence, what are we talking about,
Speaker 1 44:59
Seville? Just means positive or negative.
PJ 45:02
Oh, and
Speaker 1 45:04
it's, it's a technical term for something that's actually very simple.
PJ 45:09
Okay, that's awesome. Yeah, it's easier than saying positive or negative, so I get it. Yeah,
Speaker 1 45:16
it's interesting in a, in a cultural context, because what I think in general the real world situations psychologists often talk about valence being synonymous with pleasure, and I think that makes sense in real world situations. However, in cultural contexts, in the arts, you know, we often will listen to a piece of music that is expressing sadness and actually making us feel sad, and yet we still find it pleasurable, and same thing with films, right? You know, we go to films that know that they express sadness, and this is well, this is kind of a paradox, if you like, there's been discussed from and investigated from various different angles, think one way of approaching it has been to think about empathy, so you know if we're in a sad state, we might choose to listen to a sad piece of music to empathize with, and the other way of thinking about it, I think, which, which I think is interesting, is if the arts is a sort of like a playground for trying out different kind of emotional states, you know what would happen if you know you were in this situation, and you then you have a kind of vicarious experience of the fear or the sadness that you might experience in those situations, and yet you know in a safe context, so it's like a training ground for those kinds of emotional experiences, and I think the same thing goes for these processes of prediction. Music is kind of like a training ground for honing these processes of prediction, and there is an argument that that's in development, that's one of the reasons why, for an adaptive function of music, that it provides a kind of safe training ground for all these kinds of psychological processes, prediction, learning, memory, sort of cutting things up into phrases, scene analysis, also interactional development, shared attention with the caregiver emotion, all these psychological processes can be rehearsed through music.
PJ 47:32
Phenomenal answers. You've been so gracious with someone who is very new to this. I want to be respectful of your time, but I could ask you one final question. For someone who's listened to us for the last hour, besides buying and reading your excellent book, Learning to Listen, Listening to Learn, what would you recommend someone do or meditate on in response over the next week?
Speaker 1 48:00
I think find some find music from from another culture that initially feels challenging and listen to it and see how your experience of it changes with as you become more familiar and start to figure out how the music is working.
PJ 48:16
A beautiful and practical answer, Dr. Pierce. Thank you so much. It's been a joy talking to you.
Speaker 1 48:23
Thank you so much.
Unknown Speaker 48:24
Thank
Transcribed by https://otter.ai