2157

LESSWRONG
LW

2156
Language & LinguisticsMusicPhilosophyScholarship & LearningRationality
Frontpage

18

On music and language

by Joey Marcellino
7th Jul 2025
10 min read
6

18

18

On music and language
2cousin_it
1Joey Marcellino
3cousin_it
1Trevor Hill-Hand
1Joey Marcellino
1Trevor Hill-Hand
New Comment
6 comments, sorted by
top scoring
Click to highlight new comments since: Today at 4:30 AM
[-]cousin_it3mo*20

Amazing post. But I want to maybe push it a bit further.

From the perspective of creativity, any given domain (like music, writing, drawing, mathematics and so on) can be seen in two ways:

  1. A target of translation. You come up with things that are "good" in other domains, then translate to this one.

  2. A creative medium of its own. You first learn it by imitation, then try to synthesize little bits, and gradually learn what's "good" or not.

You approach jazz saxophone as (2), and say poetry is more (1). But from what I understand about poetry, and writing in general, it's also much more (2) than (1). Good writers use language the way you use the saxophone. Annie Dillard mentions a young writer who is asked "do you like sentences?" and becomes confused by the question - but liking sentences is precisely the right way to good writing! It's not so much about having cool thoughts and translating them into sentences, but more about directly creating cool sentences, and even cool individual words. The poet Mayakovsky said a rhyme is a barrel of dynamite, and the line leading up to the rhyme is the fuse.

So the question "is music a language?" is a bit of trick question. When treated as a target of translation, music is poorer than language: things like mathematics can be somewhat translated into language, but not into music. But as a creative medium, language feels similar to music and other creative media.

Reply
[-]Joey Marcellino3mo10

Thank you!

That's an interesting perspective. I wouldn't have described poetry writing as being a clear case of 1), since on my model ordinary thoughts already spawn in language, and so wouldn't require "translation" rather than just massaging or reshaping into cool sentences. The model you suggest, where for experienced writers their poetic thoughts spawn ex nihilo rather than being the result of this sort of massaging, seems plausible as well.

Reply
[-]cousin_it3mo31

One example I like is Eminem's line "I make elevating music, you make elevator music". The meaning behind the line is unremarkable: "I'm better at music than you". But it works so well on the level of language, it's clear that it was born in the form of language straight away. I think all good writing (rap, poetry, prose) is full of this kind of thing.

Reply
[-]Trevor Hill-Hand3mo10

I'm with you on "music feels language-like", I think even just looking at spectrograms of music and speech, and comparing those to the spectrograms of random soundscapes, makes it visible that music at least plays with the same types of rhythmic, pitch, and formant patterns in our "sound view", but they have a difference similar to the difference between how a textbook "plays with the patterns of letters and punctuation" to convey an idea, and a Celtic knot "plays with patterns" to create a pattern that is sort of just... intrinsically nice to have around? I 've started thinking of melodies as sort of "null sentences", sentences with the semiotics intentionally stripped out, leaving behind just the feeling. You can of course sort of "stretch" or "push" most systems of patterns and make it Turing complete (do some crazy stuff with Conway's Game of Life or something), but with music that's usually not the point, music feels like... the "RNA" to language's "DNA", in a way. Or rather, music is an artform that plays with the RNA.

However, I have a counter-anecdote to how you feel like your thoughts "spawn" in either language or not-language: I speak only English, but began speaking very late. When my thoughts spawn, especially when I have in high intensity mode and thinking clearly, they spawn as... I've best been able to describe them as "clouds"? A "concept-graph" is also close. Not just strings of words, but also complete spatial understandings in my head (i.e. similar to the fidelity and depth of memories I have of real places), sounds, tastes, smells, emotions, connections and causal graphs, I'm able to rotate this new place around in my head and see how it operates and watch rules connecting together and causing little behaviors... not at all once though, at first it's very vague impressions, and as it gets more 'filled in', none of it involves me thinking any words like 'spatial understanding' or 'systemic behaviors' though. It's more like gradually more and more understanding "appears from the mists" as I think. I don't even really know what my brain is doing step to step other than I sort of can "feel" the effort.

But then, as a secondary process, I try to translate those thoughts into English. When I am alone, and don't need to do this, I think orders of magnitude faster and more clearly. When I am with others, I cannot stop myself from always preparing and translating my thoughts into English "just in case". I do this habitually most of my waking time anyway. However, while I do feel that part of me which does the translating is resource-hungry (a little emotionally draining even, on bad days), it is also very well-developed and practiced at this point and can translate or even predict what I would say as fast as I can think most of the time, but for anything I haven't seriously thought through before, I full on have to simply stop and shut down my language processing for brief moments to let my thinking get ahead a bit more so I can understand what I'm trying to say and figure out how to say it. Sometimes there is no way to say it, but a diagram helps. But for the most complex thoughts, often direct demonstration is the only feasible way to explain it. Things like modular synthesis really opened up music for me because I have so many different ways to go from the musical ideas I have into some sort of real world outcome of sound, behavior, pattern, sequence, shape, etc. which can all be organized in flexible ways.

EDIT: A more straightforward metaphor: I feel like I am a Diffusion model, not a token predictor model. Being able to edit my writing is crucial to me being able to function, because I'm just sort of incrementally getting the final product out of my head piece by piece as it is ready. Non-linear editors in general are also helpful, as is asynchronous communication.

Reply
[-]Joey Marcellino3mo10

Thanks for reading! Could you realize the same thought-cloud twice using, for example, language and music? And if so, do you think the end results would count as "translations" of each other in some sense? If the answer is yes I'd be very curious to see/hear an example.

Reply
[-]Trevor Hill-Hand3mo10

The clouds are sort of permanent- as they are filled in, whatever degree of detail it's at is where it stays in my head, wherever they're "stored", and it just sort of sticks in there. It feels a lot like just putting energy into some sort of "progress bar", my best thinking I really do in just a meditation sort of behavior- closed eyes, relaxed pose, slow breathing [I'll shift back and forth into this as I'm reading challenging books for instance, or when working on a project there's a lot of "lean back and sit quietly for a moment, staring into the middle distance], but there's diminishing returns (you can't make a bad idea better than it is by thinking it harder, and very simple ideas even just... "finish"). Then it's just sort of "in there", available to be translated out into the word in various physical forms, be it a sentence, an Excel sheet, a process design, a sculpture, a music video, etc. It feels like the underlying "data" is multimodal or maybe even typeless, and it's the expressions of it I create are all imperfect models of it, but can all be different and make different features visible or not, or obvious or not. But the line gets blurred, because those physical manifestations themselves, my memories of speaking them, etc. do end up getting fed back in and become part of the cloud.

And really, even the above is simplifying, because it doesn't feel like I have separate clouds for separate thoughts, I just have one gigantic internal cloud of everything I've ever thought, and I'm just adding to that with what 'comes out of the mists'. And I use that metaphor 'comes out of the mists' only to describe the feeling- the thoughts don't seem like they are literally coming from "nowhere", they feel like they come from me, they just come in very diffusely in across a lot of different conceptual dimensions besides just words, all at once and gradually and in a very "diffuse but coming into focus" way.

Reply
Moderation Log
More from Joey Marcellino
View more
Curated and popular this week
6Comments
Language & LinguisticsMusicPhilosophyScholarship & LearningRationality
Frontpage

My (unsuccessful) submission to the 2025 ACX Anything-But-A-Book Review Contest, preserved here for posterity. Originally framed as a review of the aphorism "music is a language."


Relevant context: I play jazz saxophone at a high level.

 

Here are two situations that I’ve been in a million times and that feel strikingly similar:

  1. Conversing in Portuguese, which I speak mediocrely, and running into something I don’t know how to say.
  2. Improvising on the guitar, which I play less-than-mediocrely, and attempting to play something requiring a note outside of the rudimentary scales I’ve practiced.

In both cases, the feeling is that there’s a nice, well-formed idea in my head that I could express perfectly well some other way, but I’m blocked by my lack of ability. 

I want to contrast these with a third situation. I’ve been taking salsa classes for about half a year, with no prior experience with dance beyond doing the White Boy Shuffle™ at college parties. At my classes, the teacher puts on music and calls out a series of moves that we perform in real time, like the world’s least intimidating drill platoon. I’m great at this and nail it almost every time. Then I go to the club, where I find myself:

     3. Trying to dance a whole five-minute salsa tune with no one telling me what to do,              and completely drawing a blank.

This is a much different feeling. The problem is no longer that there’s something in my head that I can’t get out; it’s that there’s nothing in my head to begin with, or at least no dance moves (my inner monologue, on the other hand, keeps a running commentary along the lines of “what was that spin we learned last week?” and “oh my god that’s the fourth time we’ve done that move in the last 30 seconds, surely she’s getting bored by now”).

I want to use the example of these three scenarios to argue against one interpretation of the statement that “music is a language.” Clearly it’s a pretty decent analogy at a high level; both involve transmitting some information over auditory channels, are constrained by culture-specific and essentially arbitrary conventions, and serve as media for creative expression. Language can be musical, music can feel like a conversation, and none of what follows should be taken as denying that. My aim is to set an upper bound on how far we ought to push the metaphor, rather than to outright reject it. 

All that said, the way of thinking that I claim is wrong goes like this:

“There are thoughts in my head. They spawn there in English, but since I also speak Dothraki and Kazakh I can translate them, or if I’m in Essos or Almaty sometimes they just spawn in those respective languages. Since music is a language, if I learn how to play the guitar I’ll be able to translate my thoughts into music, the way learning a bunch of words in a new language lets me translate my thoughts into that language.”

Instead, I claim:

  • The kind of thoughts you express when you improvise music (or dance) are completely distinct from the kind of thoughts you express with language (call them “music-thoughts” vs “dance-thoughts” vs “language-thoughts”)
  • There is no way to directly translate between these different kinds of thoughts
  • Non-prodigies without experience improvising will generally not have music or dance-thoughts, and there is no easy, fast, or obvious way to start having them (or to have better ones)

These strike me as the kind of claims that seem either obviously true or obviously false depending on your intuition, so let me first argue that they’re true and then that they’re not obvious. For the moment, I’m distinguishing between thoughts and feelings/emotions, and just talking about thoughts; I’ll come back to feelings at the end. 

In his essay “Musical thinking,” philosopher of music Jerrold Levinson distinguishes between two conceptions of music as thought…..just kidding, I’m arguing mostly from personal anecdote[1]. You’ll have to take my word for it, but I promise that, in general, nothing remotely resembling natural language thought is going through my head when I’m improvising on the saxophone. This actually wasn’t clear to me before I started preparing this review; I’m usually just not paying any attention to what my brain is doing while I’m playing. As an experiment, I sat down, played for a bit, and tuned in to my thought process mindfulness-style as best I could. 

Like I said, I didn’t find an inner monologue per se, but it turns out that if I’m playing well, I can kind of “hear” what I’m about to play in my head a split second before it comes out of the horn. Those little phrases are what I want to call music-thoughts; they spawn without any particular effort or concentration (if the tune isn’t too complicated), and change in response to both the underlying harmony and whatever I played last. If I do have a language-thought (about the music, not just my mind wandering), it’s usually something along the lines of “I should incorporate that new thing (pattern, scale, etc) I’ve been practicing.” This can either trigger a corresponding music-thought[2] or not; if not, I can shoehorn in the new thing anyway but whatever I play will invariably sound forced and mechanical[3], like a 45-year-old trying to sound hip with some bit of Gen Z slang. 

My misadventures dancing further convince me that English-to-X translation is not what’s going on when I improvise. I have plenty of language-thoughts in my head while I dance, including thoughts about seemingly relevant things like the music, and I know a solid handful of moves, but despite my best efforts I can’t link the two up in any coherent way. 

Given that I’m pretty new to dancing, you could dismiss my problem as a lack of “vocabulary” (maybe I just don’t know the dance translation for “I hope the sound system explodes in the middle of this song before I run out of ideas”), but I think there’s at least one clear counterexample: classical musicians, who despite knowing their instruments inside and out and attaining levels of mastery I can only marvel at are generally unable to improvise[4]. This contradicts the model where you start with general, undifferentiated Thoughts and Ideas, then get them into the world via your outlet of choice; were that the case, we’d expect superb technical skill on an instrument to be more than sufficient for realizing thoughts via music. If a few months and a couple hundred words on Duolingo are enough to start improvising sentences about el gato en la biblioteca, but twenty years of practicing violin until your fingers bleed aren’t enough to make it through 32 bars of “Fly Me to the Moon,” clearly the analogy is flawed. 

At this point, you might have noticed that, to the extent that domain-specific, untranslatable thoughts of uncertain genesis are a feature of the world, they’re surely not confined to language, music, and dance. Anyone who’s been a novice at an activity requiring creative decision-making, especially in real-time, is familiar with the feeling of paralysis in the face of a vast, apparently featureless possibility space, and anyone who has progressed past this level knows that ideas tend to arrive seemingly ex nihilo, rather than after a process of rational deliberation (and sometimes directly contradict what rational deliberation tells you!). Even in a basically objective, calculation-intensive activity like chess, there’s a difference in kind between an amateur who decides their next move after five minutes of mental simulation and a pro who sees a position and instinctively knows the best line. The latter doesn’t reduce to the former. 

In accordance with the theorem that nothing written by a rando on the internet can be simultaneously true, general, and novel, you can pretty fairly accuse me here of just reinventing something like System 1 vs System 2 thinking, and indeed we can tidily classify cases where you directly have an X-thought, like a musical idea, as utilizing System 1, and cases where you try to engineer an X-thought through deliberation as utilizing System 2. Still, I think there are some subtleties when it comes to music and language. First, I usually thought about System 1 and 2 in the context of practical decision making, as the difference between a gut reaction and a considered choice. Upon reflection, improvising is just making a continuous stream of micro-decisions and so should be subject to the same analysis, but on the surface creative activities seem like a different sort of thing than e.g. picking a washing machine based on vibes.

Second, the relationship of language to both systems is complicated, in that we often use language for System 2 thinking, but the process of forming thoughts into language itself almost always uses System 1 (unless you’re struggling to find the precise words for a thought, or translating into a language you don’t know well); it’s natural, spontaneous, and effortless. Thus the rough heuristic of “if I’m inner-monologuing about it, I’m using System 2” fails when what we’re considering is exactly the activity of inner-monologuing, or the music/dance/whatever equivalent. 

Technical points aside, my main defense against the charge of stating the obvious again comes from my personal experience. All of the beginner improvisers I’ve given lessons to ask some variant of the question “how do I decide what notes to play?” I usually say something about listening to the harmony, forming a coherent melodic idea, etc, but the real answer is of course that you don’t “decide,” at least at any conscious level; it’s like asking “how do I decide what words to say?” Some part of this question is probably wishful thinking – the kind of knowledge that you can learn from a textbook is always easier to acquire than the kind requiring raw experience – but I think at least some part reflects genuine confusion about what improvising actually entails. Moreover, despite my decade-and-a-half of experience improvising on the saxophone, I still fell into exactly the same trap when starting to dance! I spent months trying to figure out how I’m supposed to decide what moves to do before making the connection and realizing that the question is ill-posed[5]. 

Thinking about music and dance as involving their own distinct kinds of thought also clarified the concept of “expressiveness” for me. To give a mildly self-incriminating example, I play fighting games competitively, especially the Super Smash Bros. series. There have been various editions of Smash throughout the years for different consoles, and all have their devotees, but the version for Nintendo GameCube (“Melee”) is widely regarded as having the highest skill ceiling and allowing for the most precise control of your character. Melee diehards love to talk about how “expressive” the game is, and for a while I rolled my eyes whenever I heard this[6]. Like, what could you possibly be expressing with a combo in Super Smash Bros. (other than maybe anger)? 

I now think that this skepticism stemmed from the view of language-thoughts as privileged or fundamental, and of expression in other domains as requiring translation from language to the medium of choice. On the view where Smash-thoughts (or maybe fighting-game-thoughts) are their own kind of thing, it’s obvious how to interpret the claim that Melee is expressive; it’s simply an overall better vehicle for realizing these sorts of thoughts, the same way that a saxophone is a better vehicle for realizing music-thoughts than a kazoo[7]. A more everyday domain replete with non-linguistic expression is fashion; self-expression in the context of an outfit doesn’t mean somehow transmuting your opinions about pizza toppings into a choice of shirt and pants. You just have an outfit-thought, and realize it to the best of your ability and wardrobe.

The other phenomenon that I feel like I understand a little better in light of the view I’ve articulated is talent. Lurking in the background this whole time has been the question “how do you start having X-thoughts/have more interesting X-thoughts?” It’s seemingly paradoxical; you want to train your unconscious mind/System 1/Primordial-Soup-Whence-Your-Thoughts-Emerge via directed effort, but directed effort is pretty much by definition a conscious activity. It turns out that the best you can do is some combination of ingesting/analyzing good examples and just trying to do the thing yourself, with a mix of pre-planned ideas and randomness, and eventually your brain does magic and starts generating ideas of your own. It’s intuitive that something like this must be true when it comes to improving ideas — there’s no deterministic recipe for having better language-thoughts, after all — but since most of us don’t remember learning our first language, we don’t have a good reference for the process of going from having no ideas to some ideas (“gaining consciousness,” if you will). Having a step clearly labeled “the brain does magic” provides a convenient slot for an otherwise-nebulous concept like talent; it consists at least in part in how efficiently your brain does that magic. 

So where does all this leave us with respect to “music is a language”? Clearly in some ways music is like language, but those are mostly ways that music and language are also like everything else. I think the main link propelling the aphorism is the similar-seeming relationships of music and language to emotions, which I’ve been ignoring until now. It’s true that songwriters can write sad songs to reflect or give voice to their sadness in much the same way that poets can write poetry, and listeners/readers might find that the two media resonate with their own emotional states in similar ways[8]. That said, I think what’s going on when someone writes e.g. sad poetry is closer to description of or inspiration from an inner state, rather than the brain-to-reality transfer that happens when I play a musical idea. It’s hard to see how someone could express their sadness with the same sort of directness; emotions just don’t seem to have a corresponding medium in the exterior world the way that music-, dance-, or language-ideas do. Still, if you’re unconvinced, you can just take the preceding arguments as applying to the dare-I-say majority of ideas across types that aren’t intended to reflect an emotion. 

I’ll conclude with a caveat. I’m a pretty good musician and improviser, but the talent distribution has a long tail, and it’s definitely possible that my claims here no longer apply past a certain level. Maybe Chris Potter[9] really can play a Shakespearean sonnet on his horn. I also don’t have synesthesia and have never tried psychedelics, both of which could conceivably scramble sensory modalities enough to make translation between thought-types possible (or at least seem reasonable from the inside). For those groups, music might genuinely be a language. For the rest of us, it picks up right where language leaves off. 

  1. ^

    His essay is excellent and touches on many of the ideas to follow, if you’re into that sort of thing. https://www.musicandmeaning.net/issues/showArticle.php?artID=1.2

  2. ^

    The corresponding music-thought is still not a musical translation of the English-thought, the same way that the Portuguese word cão is not a translation of “I should write the Portuguese word for a furry four-legged friend.”

  3. ^

    How can I play something without having a music-thought about it? In the end, notes on instruments are produced by series of physical actions. If I know what buttons I need to press to play a certain scale, I can play it even if I don’t instinctively know what it will sound like before it comes out.

  4. ^

    Am I claiming classical musicians who can’t improvise don’t have music-thoughts? Kind of; clearly they have beautiful and nuanced thoughts vis-a-vis interpretive elements of music like dynamics, tone, tempo, vibrato, etc. Still, I think it’s fair to say that the content of e.g. a language-thought is (mostly, usually) distinct from its verbal delivery, and so I’m comfortable restricting the content of music-thoughts by definition to more constitutive elements like melody, harmony, and rhythm.

  5. ^

    This is the main reason I think I’m not purely fighting a strawman here. Even if few people would defend the view that music is literally a language the way that Portuguese is, I think many people implicitly expect learning music to be like learning language in a way that presupposes similarities that don’t exist.

  6. ^

    Some putative examples of self-expression in Melee: https://www.youtube.com/watch?v=ui0ExdoEBNA

  7. ^

    This doesn’t mean that anyone who plays the other versions of the game is somehow wrong; I accept that the piano is on balance a better instrument for expressing music-thoughts than the saxophone (for one, you can play multiple notes at the same time), but the saxophone has a unique character that I like, and excels at certain aspects of expression, such that it’s not strictly dominated by other instruments. The same goes for Smash. 

  8. ^

    The standard idiomatic English here would be that artists are “expressing” their emotions, but I’ve already introduced the idea of expression to mean the concrete realization of an idea. 

  9. ^

    As good as it gets. https://www.youtube.com/watch?v=UeRQWQnCuuY&t=174s