The inexistence of a semantic structure in music

[Guijarro]

I am having the hell of a time in trying to convince a professional musician that, although music can be said to have a syntactic structure of sorts, the one thing that distinguishes music from language is that it definitely has no semantic structure AT ALL.

I have tried to reason about the distinction between the meaning of a linguistic expression, like, say, "He then put it on top of the other one", "Vot vopros!, niuyeli vui nie znaietie?", "Yo, desde luego, del asunto ése pues nada", or "On ne la voit pas dancer le long des golfes clairs", etc., i.e., SEMANTICS,  and the meaning of the speaker (which often uses linguistic props to form premises from which to infer the messages), i.e., PRAGMATICS.

I have tried to show that almost ANYTHING in the world (from a toothpaste tube, to the Empire State Building, from the War of Crimea to Beethoven's  5th symphony, along with our semantic codes) may serve at one moment to help us convey our messages. Thus, I have argued, to say that music communicates is a platitude which amounts to say nothing at all.

To no avail!

Am I getting nuts, or are you all, linguists, in agreement with me here?

And if not, WHY NOT? 

[Guijarro]

According to me, this is the most common "error" of non linguists:

https://www.youtube.com/watch?v=8IxJbc_aMTg

[Daniel]

Music has no semantic structure because it has no conventional mapping from form to meaning. But it does have pragmatic meaning I think, so that's probably the distinction that isn't clear.
(In fact, I might go as far as saying that it has conventional implicatures to some degree such as genre and common sequences of notes.)

[lx]

What is his argument to the contrary? For those of us who don't have the 2+ hours to watch the... oh, I'll probably watch it anyway. 

[jkpate]

Mark Granroth-Wilding is doing extremely cool work using CCG to parse jazz. The "words" are chords, and the semantics is a path through tonal space. The CCG combinations express the build-up and resolution of harmonic tension.

[Daniel]

But then what does "semantics" mean? Shouldn't that just be syntax?

[jkpate]

The syntax is the combination of chords. Different chord combinations can express the same path through the tonal space. This conference publication has some examples.

[MalFet]

Ask him how to say "There is a red hat on the television" in music.

[jkpate]

Oh, I think Mark would agree that the semantics of music are different from the semantics of natural language. I'm just highlighting a reason to contest the opening claim that music "definitely has no semantic structure AT ALL." If we view the syntactic structure of music as encoding a path in this tonal space, then we can devise a well-defined semantics and obtain pretty accurate automatic parses.

[MalFet]

Oh, I think Mark would agree that the semantics of music are different from the semantics of natural language. I'm just highlighting a reason to contest the opening claim that music "definitely has no semantic structure AT ALL." If we view the syntactic structure of music as encoding a path in this tonal space, then we can devise a well-defined semantics and obtain pretty accurate automatic parses.

But what makes it semantics, exactly? Just the fact that it's structured and meaningful? Our friends in anthropology have been arguing for years that those two conditions describe virtually all human behavior.

I enjoyed the paper tremendously, but I'm not sure I see the utility of calling movements through tonal space "semantic" except analogically.

[Guijarro]

@jkpate:

Here is something the author of your linked conference pretends to be the case:

This may be taken as further evidence suggesting that music and language have a common origin in a uniquely human system of interpersonal communication.

I don't think music and/or language are unique systems of interpersonal communication. They can be used (as almost everything else in the world  --so hardly UNIQUE, if you please!) as props to make our thoughts, impressions, attitudes and feelings manifest to others in ostensive acts of communication. The fact that linguistic material is used massively in our communicative acts does not automatically imply that it's the only way we may communicate felicitously with other people.

Take Malfet's last message.

You may point with your finger to a red hat on the TV set and your communicated message will be the same as when you say Malfet's utterance linguistically. The ability we have to make a premise linguistically matches our pointing to a given fact or event. Suppose now that on top of the bloody piano, there is an urinoir, and you say "there is a red hat on the television" pointing to it. Which message prevails? That there is a red hat or that there is an urinoir + that you are somehow joking about it?

As said (and in accordance with Daniel). Any thing can have a PRAGMATIC sense (I wish I would be more conversant with musical theory to believe that it may have conventional implicatures, but in fact, I am not, so I leave this open) when used as prop.

But there is no way yo can say Malfet's expression musically. You may not build a premise to that effect using music, if you see what I mean.

[jkpate]

I'm not sure I see the utility of calling movements through tonal space "semantic" except analogically.

I mean, I think it's analogical to the semantics of formal languages in the same way that Montague semantics is analogical to the semantics of formal languages. They're both using the machinery of formal semantics to build up some representation that gets closer to the point of the, erm, expresser: "why did you produce that surface form (chord sequence, word sequence) rather than something else?" Talkers aren't interested in building an NP representation in their listeners' heads; they want to put a proposition there. Jazz pianists aren't interested in Tonic-Dominant-Tonic, they're interested in taking you on this journey through tonal space.

We know neither of these can be the whole story (Q: what's the meaning of life? A: life').

@jkpate:

Here is something the author of your linked conference pretends to be the case:

This may be taken as further evidence suggesting that music and language have a common origin in a uniquely human system of interpersonal communication.

I don't think music and/or language are unique systems of interpersonal communication. They can be used (as almost everything else in the world  --so hardly UNIQUE, if you please!) as props to make our thoughts, impressions, attitudes and feelings manifest to others in ostensive acts of communication. The fact that linguistic material is used massively in our communicative acts does not automatically imply that it's the only way we may communicate felicitously with other people.

I think Mark's point there was that language and music are uniquely human (is birdsong harmonic? or just melodic?), and may rely on the same cognitive substrate. The statement does presuppose that language and music are inherently communicative, but I don't think that's central to his point.

Take Malfet's last message.

You may point with your finger to a red hat on the TV set and your communicated message will be the same as when you say Malfet's utterance linguistically. The ability we have to make a premise linguistically matches our pointing to a given fact or event. Suppose now that on top of the bloody piano, there is an urinoir, and you say "there is a red hat on the television" pointing to it. Which message prevails? That there is a red hat or that there is an urinoir + that you are somehow joking about it?

Well, language sure is a lot more efficient. We couldn't get anywhere in a discussion this abstract by just pointing. Or, consider a case where we wanted to say that a red hat was not on the television (perhaps we intended to surprise the listener with a gift, but it disappeared).

As said (and in accordance with Daniel). Any thing can have a PRAGMATIC sense (I wish I would be more conversant with musical theory to believe that it may have conventional implicatures, but in fact, I am not, so I leave this open) when used as prop.

But there is no way yo can say Malfet's expression musically. You may not build a premise to that effect using music, if you see what I mean.

Sure, the semantics of music is different from the semantics of language and represents something different. That does not mean music does not have any semantic structure at all.

[MalFet]

I'm not sure I see the utility of calling movements through tonal space "semantic" except analogically.

I mean, I think it's analogical to the semantics of formal languages in the same way that Montague semantics is analogical to the semantics of formal languages. They're both using the machinery of formal semantics to build up some representation that gets closer to the point of the, erm, expresser: "why did you produce that surface form (chord sequence, word sequence) rather than something else?" Talkers aren't interested in building an NP representation in their listeners' heads; they want to put a proposition there. Jazz pianists aren't interested in Tonic-Dominant-Tonic, they're interested in taking you on this journey through tonal space.

We know neither of these can be the whole story (Q: what's the meaning of life? A: life').

I guess that's the rub for me, though. If semantics is just a five-dollar synonym for "meaning", I'm not sure the word does much for us. It's a pretty steep downhill to assume any kind of isomorphic relationship between formal semantics and The Point™ of an expression, and the whole semantic formalism boom of the 60s and 70s really came about as an attempt to reconcile this problem in natural languages with both advances in CompSci and (equally importantly) post-Frege analytic positivism.

I wish I knew more about Montague's intellectual biography, but his critical innovation in many ways was to understand Frege's Sense and Reference in a way that allowed to him to complete ignore the "sense" part of it. It's all reference all the way down for him, and that's I think an extremely important dimension of modern semantics. If it's not referencing, why is it more than meaning?

There's some really interesting work being done on how metapragmatic sensitivities allow speakers to understand patterns of repetition in reported speech as an extension of referential functions. I could definitely be persuaded that something like this is giving music a referential dimension for those savvy enough to pick up on it, but unfortunately there's probably not much hope for musical dullards like me!

[Mark Granroth-Wilding]

Perhaps I should make an appearance here. I have little to add regarding my own work beyond jkpate's excellent explanations of the thinking behind it.

Indeed, music is pretty poor at expressing referential semantics and, though examples are occasionally cited, they are invariably both straightforward and a relatively uninteresting part of the meaning of the music in question. Various authors have written on the sort of meaning that music does generally convey. A well-known example is Meyer's Emotion and Meaning in Music. A particular favourite of mine, though, which may be of interest to Guijarro, is Raymond Monelle's Linguistics and Semiotics in Music, which covers a large amount of theoretical work. Incidentally, it is precisely because the actual semantics (or meaning system more generally) expressed by music is different to language that it is such a powerful medium for communication. The same can be said of other art forms.

To claim the music has no meaning at all would, of course, be preposterous. The connection to language is interesting (and useful) in as far as the relationship between the meaning and the musical surface bears some resemblance to the corresponding relationship in language -- for example, the extent to which this relationship relies on syntactic structure, or exhibits compositionality or long-distance dependencies. Such connections are interesting because (among other things) they suggest parts of the processing mechanisms for unconscious human interpretation that may be shared, and useful because they may lead to the application of well-developed practical processing techniques from one domain to the other.

The distinction between semantics and pragmatics may be a tricky one to get right here, but in our work we have specifically focused on an issue that we consider to be more closely connected to the former -- the structures underlying harmonic expectation. In Montagovian style, we have looked largely at these structures (and the mechanisms required to derive them from the surface), but little at how they relate to other forms of meaning, such as emotional -- the meanings that people really care about! Until we've got a better hold on the structures, we don't stand a chance at formalizing the latter.

And by the way, Guijarro, I agree with your assessment of Bernstein's attempt at music semantics. It's a mess -- I think few would disagree -- though one has to credit his endearing enthusiasm! (And some of the other lectures are very good.)

[Daniel]

Semantics is meaning, a thought message. Music does not convey such a message. It does convey information, but so does any signal-- for example sound waves convey phonemes. But we don't consider phonemes to be a kind of semantics or meaning. Therefore, music does work well as an example of duality in a signal, but that is as far as I can see the association stretching. Syntax certainly has parallels in music but I don't see it, in any functional sense, for semantics. The fact that there is some layer there does not make it the same layer found in language.

These are all very interesting ideas, but I would much prefer them with less confusing terminology. Then the formal similarities could be used productively without odd and controversial implications.


BTW, jkpate, don't dismiss bird songs so quickly. I don't know all the details but at least: 1) they are recursive, and 2) they show critical period effects. In fact they were influential in Lenneberg's original proposal of the critical period and are often cited as a counter example to the uniqueness of human language with recursion.