The Hills Are Alive With The Semantics of Music
June 22, 2004 12:56 AM   Subscribe

Tunes create context like language : "musical notes are strung together in the same patterns as words in a piece of literature". Full paper. On a related note, hone your musical comprehension by playing with Impromptu. Better yet, co-ordinate it with this MIT OpenCourse - Developing Musical Structures.
posted by Gyan (21 comments total)
Excellent post.

I don't know if I'll ever forget when the prof of the first composition class I took brought out the book "See Spot Run," and used it as a prime example of certain kinds of context/structure in composition, albiet a very simple one. I think he'd enjoy this. I will.
posted by weston at 1:09 AM on June 22, 2004

I'm not sure I agree that single notes in music have the same context within a piece as words have in a book, but halfway through the paper above, the author highlights these issues, agrees this is a very controversial thing to do, and then doesn't convincingly argue away his misgivings. After reading the paper, I'm not sure the author has faith in their own method.
posted by BobsterLobster at 3:03 AM on June 22, 2004

BobsterLobster: The author prefers individual notes because it is a computationally convenient choice and somewhat resonating as an analogy:

pauses are typically between words < -> pauses between notes;

word stems can be combined as prefixes, suffixes or just adjuncts < -> notes can be combined to form a continuous played note;

words are composed of letters whose independent existence is mostly acknowledged only during error creeps and other attentive/distracting events < -> notes are superpositions of simpler waves, which aren't normally acknowledged.

Besides, using notes, the law was remarkably upheld.
posted by Gyan at 4:01 AM on June 22, 2004

A word has a meaning on it's own < -> A note needs other notes in certain rhythms to achieve a similar context to a word.
posted by BobsterLobster at 6:31 AM on June 22, 2004

BobsterLobster, at this stage it is just about identifying the probability distribution. The mechanism for power-law it is yet completely understood. Here is another example that leads to the same result.
posted by MzB at 6:37 AM on June 22, 2004

Very nice, thanks Gyan! I'm not too happy though with the drawing of analogies between 'language' and 'music' that seem to privilege the former (is Simon's model a model of language, or of information?). Language and music do both have the same Zipfian characteristics, but that could be because they are both examples of information spaces, which in general have Zipfian characteristics, and which in general show emergent patterns of context. But very nice, anyway; I like Arxiv finds!
posted by carter at 7:24 AM on June 22, 2004

Voss and Clarke did something very similar back in 1978. There has also been a long-running debate as to whether finding this kind of statistical structure indicates anything other than randomness. For example, Zipfian structure has been found in so-called junk DNA but there is considerable debate over whether that means anything - see this SciAm article for details. See also Citations and the Zipf-Mandelbrot's law (pdf) which gives a nice overview of the occurrence of this kind of structure in a range of natural phenomena.
posted by jamespake at 8:05 AM on June 22, 2004

I'm always wary of papers that say "this is why 'tonal' music is better to listen to" (more comprehensible, etc. whatever the framework of the day is) because there are different kinds of 'tonality,' including the difference between Bach and Debussy. And having read through the paper (breezily), I'm not sure the author is actually saying that to the degree the Nature article implies.

One thing missing from both the paper and the article is the recognition that Shoenberg's piano piece was written to showcase atonality, i.e. to throw off the shackles of conventional tonal patterns, so it's not surprising that it wouldn't mimic the patterns of conventional tonal music. I would have appreciated some consideration of Berg's later works, or even the early serialists, where they were grappling with the question of how to create comprehensibility, or this expectation that you can grasp what's about to happen, while still avoiding tonal relationships. And how would this apply to something like In C or Music for 18 Musicians, two of the most 'tonal' works ever?

It would also have been useful to acknowledge different kinds of language patterns throughout the world and see if the different 'tonalities' of those parts' respective musics had any correlation. The author concludes the paper suggesting "It would be interesting to consider alternative extensions, at the level of melodic phrases, harmonic sequences, or rhythmic patterns, and thus explore the concept of musical context at different scales." I'd second that, especially rhythmic patterns. I think it goes without saying that music with repeated, recognizable rhythmic patterns is more "comprehensible" to the average listener than the opposite. But do those patterns also group themselves analogously to language? Given that rhythm has such a more prominent place in most of the world's (non-Western) music, I'd think this would be something worth looking into rather soon.
posted by soyjoy at 8:07 AM on June 22, 2004

Milan Kundera had an interesting take on this near the end of the Book of Laughter and Forgetting.
posted by rushmc at 8:47 AM on June 22, 2004

First of all, let's agree that when they say "tonal" they mean "centric." The difference has to do with both tuning systems and use of harmony. Centric is a much looser term, and can be applied to anything with a clear home "key" whether the music is tonal, modal, minimalist, built on an alternative tuning system, or otherwise post-tonal.

Second, when the byline says "it has to do with math" I say "DUH! That's what Pythagoras said thousands of years ago!" Simple ratios (tweaked for modern tuning systems that came into use in the late Baroque Period) are the entire basis of simple scales and triads. I furthermore suggest that it isn't that centric music is "just better," but rather that most people don't know the language of atonal music. It's like watching a foriegn film without subtitles.

The "Nature" article is further filled with such "DUH!" statements as "For example, it is more likely that the rest of this article will contain the word 'music' than the word sausage,'" and "The key in which a piece of music is written is one factor that influences which notes are more or less likely to come next." The author goes on to talk about how melodies can be organized into "phrases" just like language!

Then they use math to show that centric music repeats more notes than serialism. Serialism, by definition, must use all 12 tones of the equally tempered scale before repeating any of them. Frankly, they could have saved a lot of number crunching by reading a few decent music theory books.

The MIT course described sounds suspiciously like a beefy "musical form and structure" course much like that required for music majors everywhere. Having passed such a course, I will venture that of course music can be organized much like literature. However, there is a wide range of style and language use in each genre. Poetry -- much like short songs -- can range from Shakespearean sonnets to Haiku to e. e. cummings. Novels -- much like symphonies -- can range from Tales of Genji to the works of Charles Dickens to the latest science fiction and romance titles.
posted by ilsa at 9:48 AM on June 22, 2004

BobsterLobster: A word has a meaning on it's own.

So does a note. Like the author says, music has no functional semantics, but an individual note does encapsulate information like absolute pitch, duration, timbre, component waves...etc On a more abstract level, we *could*, in theory, build up a nuanced memory of each unique permutation as an unique "word". We don't, because 1)there are theoretically infinite such permutations and not all of us have perfect pitch to appreciate Zipfian distribution in action. 2)Consequently or otherwise, we don't assign consistent extra symbolic meaning to music, if at all.

ilsa: Then they use math to show that centric music repeats more notes than serialism.

The pattern of repetition, not just the fact they repeat more. Duh.
posted by Gyan at 10:06 AM on June 22, 2004

Gyan: It's not just the repetition, it's that serialism's mandate is to avoid the repetition patterns of tonality and forge entirely new patterns for people to supposedly grow to expect. It's like saying that the word "voici" occurs a lot in French texts, but almost never in English ones. As the saying goes, "Duh."

ilsa - good points, but please remember, as the page you linked shows, it's "E.E. Cummings."
posted by soyjoy at 10:50 AM on June 22, 2004

Gyan: a word has meaning more than it's constituent pitches, timbres, durations, etc. A word opens up an abstract world of 'meaning'. A note on it's own doesn't mean anything, it needs to be part of a musical phrase to have an equivalent state of 'meaning'. We also could in theory build up a nuanced memory of each unique permutation of letters as an unique "word". Why would we? Not many of these would be words that we would all recognise, or look up in a Dictionary. I think the analogy holds.
posted by BobsterLobster at 10:59 AM on June 22, 2004

BobsterLobster: The essence is the same. Words also exist only within contexts. You were taught (or heard), say, a particular word at a certain age. At that time, you associated the graphical representation of the calligraphy with its onomatopoiea(the phoneme combination), you associated the word itself with its extra-symbolic counterpart (emotion, perception..whatever). Then, with subsequent interactions and experiences, you built a rich topological network around the word. This topological network is ingrained for words of a language and virtually not maintained for individual musical notes (mostly since most of us aren't taught as kids to discriminate and categorize each note in the same manner). But the nature of words and notes within their respective domains is analogous. The meaning of individual notes just isn't that rich or appreciated by many.

soyjoy: I don't see what changes. Repetition patterns in Serialism are unconventional. Agreed. Deliberately. Agreed. The paper's not just saying that music creates some context. That's 'Duh'. The paper is showing that tonal music of the variety examined, in some way, mimics context created by language. From a broader perspective, this is also a 'Duh'. Since music and language are both information carriers created and comprehended by the same organ, it's not really surprising that language and music are instances of some deeper neural grammar and hence analogous in some primal sense. What this paper shows is a nice concrete attempt at a mapping between them.
posted by Gyan at 12:19 PM on June 22, 2004

The paper is showing that tonal music of the variety examined, in some way, mimics context created by language.

And what I'm saying is that the examination process had certain (Western) tonality-based assumptions built into it so that the conclusion is practically a restatement of the premise - but that if they took other features of language and found different pattern-based analogies in music they might also find the same to be true of serialism. I don't know that we have any point of disagreement here, except, perhaps, on exaclty how often to use the word "Duh."
posted by soyjoy at 12:28 PM on June 22, 2004

Mind you, I can't stand serial music and think the lackluster efforts of composers to find patterns that would resonate with listeners the way tonal patterns do was both misguided and ill-managed, and accounts for much of why the music never really caught on. But that doesn't mean it couldn't happen in theory.
posted by soyjoy at 12:35 PM on June 22, 2004

soyjoy: And what I'm saying is that the examination process had certain (Western) tonality-based assumptions built into it so that the conclusion is practically a restatement of the premise.

I did read the full paper, but could you elaborate?
posted by Gyan at 12:44 PM on June 22, 2004

Gyan - I don't think I need to go into too much detail, but just for example the focus on keyboard works means that the set of notes will be limited to an artificial construct of Western tonality. And the limitation of consideration to pitch and duration shows a built-in bias for Western tonality. The stated reasoning behind having individual notes be the "units of context" states this plainly:
    "The contribution of notes to the creation of musical context, determining tonality and the basis for rhythm, is particularly transparent. In addition, the choice of single notes has several operational advantages. In the first place, the collection of notes available to all musical compositions – or, at least, to all those compositions that can be written on a staff using the standard note types – is the same." (my emphasis)
So I'm saying that if they had looked at some other patterns - the repetition of notes of a certain volume or rhythmic placement within a meter, for instance - then they might have found a stronger "language" correspondence for Balinese Gamelan than for Debussy. Or if they had looked at the total "vocabulary" of the piece in terms of the array of pitches, dynamics, and note durations, and examined the large-scale syntactic patterns therein, serialism probably would have looked more "linguistic" than Mozart.

Nothing wrong with the way they decided to do it, of course, but as I say, it's a little tautological. Someone approaching this question using Mandarin, with Chinese music as their starting point, might find more of a "language" correspondence there than with Bach.
posted by soyjoy at 2:18 PM on June 22, 2004

soyjoy: And the limitation of consideration to pitch and duration shows a built-in bias for Western tonality.

I don't see what's wrong with this. He's trying to find a consistent pattern. He found one. He didn't know beforehand that notes would follow the Zipf Law. It's not a complete isomorphism, but its' a good start.
posted by Gyan at 3:08 PM on June 22, 2004

I'm not saying there's anything wrong with it. The paper itself is certainly of interest, but as I tried to make clear in my first comment, I think it's being mischaracterized somewhat in the Nature article. Something that shows one kind of correspondence with one particular established pattern winds up confirming the inherent value and accessibility of "traditional" Western music. Obviously the researcher couldn't do all the stuff I've proposed in one pass. But it would take doing that, and more, to support the sweeping claims the Nature writer makes about why people do or don't grasp so-called "atonal" music - which label would fit a great deal of music played, and enjoyed, by ordinary people worldwide.
posted by soyjoy at 9:28 PM on June 22, 2004

soyjoy and gyan: I'm really enjoying your discussion.

I didn't read the paper, and I know little of music theory, but the OpenCourseWare syllabus was intriguing.

I'm really interested in what this says about our experience of sound in general. Think birdcalls, war drums, Teletubbies, radio shows.

Or: When we feel stirred by an impassioned speech, how much of our appreciation is for the text of the speech and how much is for the speaker's body language & vocal patterns, along with the sequence of emotional responses they trigger?
posted by zerolucid at 9:33 PM on June 22, 2004

« Older Voices from Naropa   |   Iran blocks Movable Type Newer »

This thread has been archived and is closed to new comments