bigideas

As prep for last night’s event I wrote up some notes based on some rather ancient research work, largely to limber up and remind myself of some things. I’ve posted them below — be warned, this is a bit long and pretty sketchy, but at least those who were there might find it useful.

It’s Henry Longfellow who’s usually quoted with having said that music is a universal human language. He didn’t know much about music, though, and anyway he doesn’t mean much by it. It comes from a remark in his prose travelogue Outre-Mers, in the context of a chapter about Spanish folk-songs which is more interested in the words than the music, and which is relentlessly, drearily, dishearteningly gushing.

Still, plenty others have echoed the same sentiment. Many of them have been musicians or music critics rather than philosophers of music, so perhaps we shouldn’t take the statement too seriously, but it’s worth using it as a jumping-off point.

In starting to look at the statement “Music is a universal language”, we obviously need to think about what we mean by “music” and “language” and probably also “universal”.

Music

We can get into a debate about what music is later. All I want to say here is that we need to be careful with ethnocentrism. Most philosophy of music, for example, takes all of its examples from the classical music of the eighteenth and nineteenth centuries. As a result, it’s easy to reach conclusions that only make sense in that context. Any account of music as a universal language has to work for West African drumming, football terrace songs, North Indian classical music, the Chinese Noh theatre, samba, Inuit throat-singing and Kylie Minogue.

It’s hard to come up with a definition of music that covers all this, although it’s certainly fun to try. Let’s borrow Supreme Court Justice Potter Stewart’s pragmatic definition from another context: we know it when we see it.

Hockett’s Design Features

Music clearly isn’t a “language” in the sense that Russian is a langauge. It’s not what linguists call a “natural language”, which is a term of art for any general-purpose communication system that can be a person’s first language. Yet there are many things that are language-like without being natural languages: computer programming languages, “body language”, sign systems like road markings and some forms of animal communication might all fit the bill. The question is, how much like these sorts of things is music? Is it language-like enough to warrant its description as “a universal language”?

One way to think about these sorts of things is Charles Hockett’s “design features”. His view is that natural languages evolved gradually from simpler forms, and that there’s a continuum between them and animal communication. So he devised a list of the features of human language usage. They seem to characterise it uniquely – that is, there are no animal communication systems that have all of these features. Yet most of the features can be found in at least one non-human form of communication. This is designed, in part, to remind us that human beings are unique in their use of language, but also to make us feel less isolated from other species. We’re different, but not completely different.

Music, at least as we mean it for now is unique to humans, but we can repurpose Hockett’s approach. If music shares many of the design features of natural languages, we might be impressed enough to call it a language. If it doesn’t, we may feel that even if it communicates it does so in a non-linguistic way. I’ve split up the design features into categories because some are more relevant for us than others.

This list and the descriptions were adapted from the ones given here.

Design Features that Describe what Speaking is Like

Interchangeability — The speaker can receive and broadcast the same signal. This is distinctive from some animal communications in which males and females produce different signals, and only respond to one another’s signals. It’s a feature of music, since a musician is normally also a member of the audience.

Total feedback — The speaker can monitor their language performance as they go. This differs from some other simple communication systems, such as traffic signals (a red light can’t tell when the bulb is burned out). As with interchangeability, music does have this. A competent violinist who’s out of tune can adjust their playing to fix the problem.

Productivity — Human languages allow speakers to create new, never-before-heard utterances that others can understand without being given some kind of definition. This is another place where road signs fall down. If music communicates at all then it must surely exhibit this feature.

Traditional Transmission — Human languages aren’t innate; they’re acquired by socialisation. Many, though by no means all, non-human communication systems fail to exhibit this feature, but music surely does.

Design Features About Meaning

Semanticity — Specific signals can be matched with specific meanings: we can attach the word “ball” to a certain kind of spherical object, “jumping” to a certain vigorous vertical activity, “tired” to a certain physical state and so on. All competent speakers of a language attach more or less the same meanings to the same signals. The process by which this happens even in natural languages is philosophically fraught, though.

Displacement — We can talk about places other than here, people and things that aren’t present, counterfactuals (things that didn’t happen but might have) and purely fictional things. Natural languages give us a way of saying something like “If I’d caught the 187 yesterday I wouldn’t have been late for Dave’s party”.

More Technical Design Features

Arbitrariness — “Miniscule” is quite a big word, “big” isn’t. The connection between signifier and what it signifies is conventional, not based on similarities. You can’t guess the meanings of words from their sounds or their look on a page.

Discreteness — Natural languages are made of a finite number of “atoms”, such as phonemes or words. If I want to say that someone was somewhere between a boy and a man, I can’t say a word that’s, say, two-thirds of the way between the two. It’s not a continuous scale, in other words.

Duality of Patterning – We have a small repertoire of meaningless, discrete atoms – eg letters or phonemes, but we can combine them in a huge variety of ways to make meaningful words and phrases.

Design Features Relevant to Human Speech Only

For the sake of completeness, Hockett also gives the following features that are not relevant to music and that really just characterise human speech. Written English, for example, fails to exhibit any of them:

Vocal-auditory channel
Broadcast transmission and directional reception
Rapid fading (transitoriness)
Specialization

Arbitrariness

It’s well-known that Ferdinand de Saussure argued for the “arbitrariness of the sign” – that is, there’s nothing doggy about “dog” or happy about “happy”. The words of a natural language are arbitrary, conventional and therefore learned.

If a sign-system “refers” or has Semanticity in any very strong sense, we expect it to have arbitrariness. I’m going to switch semioticians for a minute and talk about Peirce, who did more work than anyone to try to categorise the difference ways something can mean something else. He came up with three main categories.

First is an icon. X is an icon of Y if X is just like Y in some respect. The obvious example is a photograph of me, which is an icon of me because it looks like me. Actually, though, there are not many “pure icons”. A photograph of me doesn’t really look much like me at all; you’d be unlikely to mistake one for the other. In fact we know how to “read” a photograph as a representation of a person. In this respect something more conventionalised – something more arbitrary – is going on.

Second is an index, which refers to something simply by pointing to it. If you ask me which person is John and I point with my finger, my finger is acting as an index of John. If I see smoke coming from downstairs, I might take it to be an index of a fire. An index doesn’t have to bear any relation to the thing it’s indicating except one of proximity, or some other kind of pointiness.

Third, and most interesting, is a symbol. A symbol represents its object by convention only, by what Peirce calls “habit”. A word is a symbol; so is a red traffic light. When I wear a suit for a job interview, the suit functions as a symbol of my willingness and ability to confirm to certain norms. When a film cuts from a boy playing with a ball in the street to a man hiding in some bushes, the cut is a symbol that tells us that the man is probably watching the boy. These are arbitrary signs.

Only symbols are truly arbitrary and, we might be tempted to say, properly semantic. After all, anything at all can be an icon of anything else just by happening to share some property with it and by my noticing that fact. And this is very important: for Peirce the interpretation makes the meaning, at least in the first two cases.

So, if a green traffic light reminds me that my lawn needs mowing the light really has acted as an icon for my lawn. Likewise, if a car going past my window draws my gaze, and my gaze then falls on someone walking the other way, the car acted as an index of the person. This isn’t the kind of thing we mean by “musical meaning”. We mean something socially or culturally shared, conventionalised, properly semantic. For that, we need arbitrariness.

The Revenge of Discreteness

Let’s take two vowel-sounds, |i:| and |u:|, as in “feed” and “food” respectively. Try making the first sound and then slowly changing it into the second; you’ll get something like the diphthong sound in “feud”, especially if you say it like Brian Sewell would.

The continuous gradation between |i:| and |u:|, however, contrasts with the fact that here we have two distinct vowel sounds, and words that use them have different meanings (“feed” and “food” are different words). You can’t just pick any sound somewhere between the two.

Similarly, I can say “I like bats” and “I like cats”, but not something with a sound in between |b| and |k| that indicates that I like both bats and cats, or that I like the rather Lovecraftian creatures known as bat-cats.

Discreteness is the feature of language that enables us to use a continuous medium, such as sound, to form the discrete “atoms” of a language. We can tell the difference between “bat” and “cat” because |b| and |k| are discrete phonemes, and swapping one for the other changes meaning, or sometimes obliterates it (“canana” isn’t a meaningful English word).

This feature is so important because we need to have a finite number of atoms in our language if we have any hope of learning it. We have to be able to learn them if their reference is arbitrary, meaning we can’t rely on intuition to tell us what some new sound means without having heard it before. All the atoms must be learned before we can develop any real competence in a language.

It seems to me that music doesn’t have Duality of Patterning in general even if it sometimes seems to exhibit it in specific cases. That’s something we can talk about in the session. If I’m right then music can’t have an arbitrary semantics – that is, can’t be a symbol in Perice’s sense – and the whole idea that music refers to things in any strong way goes out of the window.

Communication

Hockett’s design features tell us how language-like a communicative act is. Before we can make much use of them, though, we need to characterise music as a communicative act.

By “communication” we might mean something like “the passing on of information”. We don’t need any intention to pass it on – for example, my body language might be communicating something to you without my knowing. On this view there’s a sense of a little conceptual package of some kind being made by me and passed on to you through the communicative “medium”.

This is a very classical view of communication based on our intuitions about natural language. We may have the feeling that when I speak to you I first formulate a thought, then package it up in words, put it into a little vehicle made of sound waves and send it to you; your brain unwraps the package and “gets the message” – that is, if the communication was successful, you have the same idea as I did to start with.

This is a pretty dubious idea, though, both philosophically and psychologically, and it’s not clear that it accounts for all communication even if it can account for some. We might take a more radical view and think of communication in general as a shared process of making meaning.

My body language is a good example. If neither you nor I are aware of it, it makes no sense to say that communication has taken place. If you become aware of it and I’m not, then in a sense you’re engaged in an act of interpretation – that is, the manufacture of meaning. If I become aware of my body language and start trying to use it to influence your interpretations then you and I are colluding in an act of meaning-making. This is what it means, on this account, when we say “we’re communicating”.

It’s useful to make a distinction between an idiosyncratic interpretation and a collaborative process like the one just described. If the way I’m sitting reminds you of your mother, and makes you happy, that doesn’t quite mean that I’ve communicated happiness to you. Communication implies co-operation – your interpretation has to be, in some sense, “what I meant”. In the context of music this kind of thing can get very murky very quickly.

Emile Benveniste, in his famous study of pronouns, suggests that the idea of individual subjectivity comes about as a result of using language; that, in a reformulation of Descartes’ famous phrase, “ego is he who says ‘ego’”. I suspect there’s something in this if the collaborative model of communication is right, because our sense of ourselves is also something we develop through our interactions with others, and those interactions are overwhelmingly communicative.

That might give us something to think about when we think about musical communication. What is it that we’re constructing when we construct musical meanings together? Music is a social and cultural phenomenon, and we’re certainly constructing social and cultural meanings. Those meanings are in turn social and cultural facts, and they help determine other social and cultural facts. So it’s important what kinds of meanings we produce.

Communication of Emotions

We might well ask what kinds of things music is supposed to communicate. Probably not propositions like “John went to the shops yesterday” or “The price of coffee has gone up again”. The most common thing music is thought to communicate is emotions. The model is that the musician is thinking of something like “sadness” when creating the music, and it’s supposed to make us think of sadness when we hear it.

Note that the musician doesn’t have to be sad, and the music isn’t supposed to make us feel sadness. Rather it’s supposed to evoke the feeling of sadness – make us feel it at one remove, as it were. The universalist claim is that this kind of communication, because it happens at a non-verbal level, is effective across cultural, social and linguistic barriers; that we can communicate with one another musically even where such barriers are insurmountable by ordinary language.

Studies have been done (mentioned in, eg, http://tinyurl.com/mgxwpd) in which people with little experience of Western classical music were played examples that are widely agreed by Western audiences to communicate emotions, with the result that the communication was unsuccessful – that is, music evoking (for us) sadness, anger or love wasn’t recognised as evoking those things by people who hadn’t had much exposure to that kind of music.

We can try this kind of thing on ourselves, for instance, by listening to performances of North Indian classical music in which the ragas are associated with different times of day. It’s unlikely that anyone not familiar with that musical tradition would reliably “receive the message”.

There are a lot of difficulties with coming up with an account of how music encodes and passes on this supposed emotional content. Yet this is very clearly something it can do, even if it needn’t always do it.
Metamusic

One popular claim is that music does have semantic meanings, but the things it refers to are musical things. Hence, perhaps, a variation on a theme refers to the original, or a snare drum pattern in a Haydn symphony refers to the tradition of military drumming.

This is Hockett’s Reflexivity design feature, and it’s an important and powerful feature of natural languages that they can talk about themselves, as Benveniste did when he wrote about pronouns. But it’s not clear to me how we can say that a system that can only refer to itself is in fact doing that rather than just exemplifying itself. Is anything semantic really happening here?

We can formalise this observation into a rule that we may or may not find convincing: a sign-system only exhibits Reflexivity if it also exhibits Semanticity in which the referents aren’t elements of the sign-system. In English: a sign-system can only talk about itself if it can also talk about other things. If it can only talk about itself then at the very least we need an explanation of how we know it’s doing that rather than simply existing.

Music as Expression

Musical meaning is, of course, very often said to be about emotions, and is often stated in a phrase like “music expresses emotions”. There are lots of theories about how this might work. One way to chunk them up is as follows:

First: what’s being referred to? A theory is physiological if it talks about on the effects of an emotion on your body. Otherwise it must be associative – that is, it must be about what you think about the emotion, not what you physically feel.

Second: how does the reference take place? A theory is an arousal theory if it says that music changes your emotional state. If not, a theory must be a connotation theory – that is, it causes you to think of the emotion rather than feel it directly.

These four terms give us a four-way classification of theories of musical expression of emotions. I think every theory fits somewhere in here, although it’s not the only way to carve things up and some theories may sit on the fence.

Physiological Arousal – Hearing the Moonlight Sonata directly changes my physical state into one where I feel sadness
Associative Arousal – Hearing the Moonlight Sonata reminds me of being sad, and that in turn causes me to feel sad.
Physiological Connotation – The Moonlight Sonata has certain physical properties that are also present when sadness is experienced, and so sadness is brought to mind.
Associative Connotation – The Moonlight Sonata has certain properties that are associated more abstractly with sadness, and so again sadness is brought to mind.

You probably find one of these more appealing than the others. Nobody denies that music can elicit emotional responses in listeners: the question is, are those responses meanings? Only, I think, if these “expressions” are conventions that competent listeners agree on. If the Dance of the Sugar Plum Fairy makes me angry, that’s my problem. It’s only an example of communication if it reliably makes listeners angry (arousal), or at least makes them think of anger (connotation), which in this case I don’t suppose it does.