Some Theoretical Stuff

by Margaret Magnus
copyright 1997, 1998 by Margaret Magnus
all rights reserved

Margo's Magical Letter Page

See also the introductory chapters of my dissertation:

Abstract and Introduction
Theoretical Preliminaries

It is commonly believed that the letters that make up a word are unrelated to its meaning, that the relationship between the form and the content of a word is arbitrary. People think that what we call 'weird' could just as well be called 'grand', and what we call 'grand' could just as well be called 'weird'. But this commonly held belief is... well... simply wrong. This can be demonstrated by legitimate scientific means, i.e. by repeatable experiments. But more on that anon. I begin by simply stating the essence of my findings with no attempt to substantiate them.

So why does nearly every student of language think that a person of my conviction (a so-called sound symbolist) must be off their rocker? Well, they have in fact very good reasons. Consider, for example, that here in New Hampshire and in Tennesee and in Colorado, we all speak the same 'language' (American English), and we all have a word 'creek' (or pick any other word you like), but we all pronounce it differently. If sound affects meaning, this would be impossible. But this is small potatoes. In English, we call it 'creek', and in Norway, it's pronounced 'bekk', and in Russia it's pronounced 'rechka', not to mention how they say it in Yoruba and Chinese. It would be quite a stretch to say that the sounds of these words have anything predictable in common. And even this is just part of the story. For within one and the same dialect of our American English you can just pop a suffix on a word and change the color of the vowel in something like 'Newton' and 'Newtonian'. And we can have two different words for one and the same species like 'pigeon' and 'dove'. In fact, languages can sometimes evolve so fast that people can't even understand what their ancestors 150 years before were writing. In some tribes in Australia, languages can evolve so fast that if a member leaves and comes back 50 years later, she no longer readily understands the language her family is speaking. How can I be claiming that the sounds in a word have anything to do with its meaning in the face of such overwhelming and obvious counter-evidence?

Well, I begin with the claim that 'creek' does not mean the same as 'bekk' and neither mean the same as 'rechka'. They do overlap on one very obvious point... they all refer to a rather small natural stream of water. But there are subtle differences between them. I claim that typically when we use the expression 'the meaning of a word', we have very little idea of what we're really talking about. To begin with, there are at least two aspects of a word's meaning, which you've probably heard about. One aspect is what the word denotes, and a word's pronunciation has relatively little to do with this. But the other aspect of its meaning is something like what it connotes, and sound plays a much bigger role in this.

Sound affects the meaning of every word, but it is given much less power in words with very, very specific referents that you can point to and define clearly, like 'enchilada' and 'Cheshire' and 'dalmatian' and 'pi'. Sound really shows its stuff in verbs that describe a peculiar motion (twist) or adjectives that describe a state of mind (sad) or prepositions... the kinds of word that you have a hard time explaining to a foreigner because they're so touchy/feely.

Connotation vs. Denotation

Consider what connotation and denotation are in essence. Denotation I understand as reference. You take a sequence of speech sounds, and you map it to an object or quality or action in the world. On the other hand, I have found connotation to be tightly tied in with something Ullman calls 'inherent meaning' and Peirce calls 'iconic meaning'. Inherent meaning is a myterious thing. It is the meaning that is implicit in the form of a thing.

I think there are several reasons that we tend to emphasize denotation over connotation. One reason is that we assign denotation to words consciously, so we obviously know consciously what the denotation of a word is. But connotation happens subconsciously and is assigned by natural and universal psychological processes rather than by a conscious human decision. It also lies deeper within the world than denotation, so that a very well-defined denotation tends to mask it. And there is another reason for this. Denotation is what makes a word a word. When is it that we are convinced that a child is no longer merely babbling, but has said its first word? That moment happens when we are convinced that the child is no longer merely uttering the word for aesthetic effect, but is using it to refer to or denote something.

We humans have many forms of communication -- dance, music, art. These can express a vast variety of things with a tremendous degree of precision. But none of them can voice an articulated thought like, "Please open the windows in the kitchen," or "what is the speed of light?" The reason for this is that they are confined to the purely connotative dimension of expression. To express an articulated thought, you need reference or denotation. And denotation is in some essential sense arbitrary. You don't know it automatically when you are born. You have to be taught it. Only once a group of people agrees that the word 'light' is going to refer to light and 'speed' is going to refer to speed, and 'what' is going to be a question word can you ask the question, "What is the speed of light?" You could also ask that question using dance or music if you agreed that a certain sequence of steps or notes referred to 'light', 'speed' and 'what'. But barring this arbitrary assignment of referent, you can express a great deal, but you cannot say anything. This is obvious in a way, but I've never seen it mentioned in the linguistics literature.

Three Types of Sound Meaning

In the middle of the 18th Century, Von Humboldt distinguished 3 types of sound meaning. I will call them onomatopoeia, clustering (or phonosemantics association) and true iconism. It's most unfortunate that everybody ignored him.

1. Onomatopoeia is the least pervasive of these forms of sound-meaning, and I have not particularly concerned myself with it at this Web site. Onomatopoeia is the imitation of a sound using the consonants and vowels in a word. It therefore only applies to words which either refer to a sound, or which refer to something that makes a sound in some salient way.

2. Clustering is an effect I also call Semantic Association. If a speech sound is used disproportionately often in a certain context, then we tend to use that speech sound productively in that context. That is Semantic Association. That's how a child learns a word. The child hears the word 'table' frequently in the context of tables, and then she starts using 'table' in the context of tables herself. As soon as she does, we say she as 'learned' the word. We say that the word has meaning. It is quite generally accepted that Semantic Association applies on the word and morpheme level. In other words, it is accepted that words, prefixes, suffixes and roots have meaning. But we deny that vowels and consonants alone do. But there are a few ways to verify that vowels and consonants do have meanings in this way. Consider this. The other day, while playing a flicking game with my daughter, I was wadding up bits of paper. She said to me at one point, "Smish it down real tight." Now what sort of a word is 'smish'? She invented it on the fly. It was not derived from some Indo-Germanic root by some regular vowel alteration the way all good words are supposed to come down to us. No. She created by analogy with words like 'squish' and 'smash'. She wanted the smallness implicit in the /i/ sound without the wetness implicit in the /w/ of 'squish'. So she invented 'smish'. And in the process, she is both doing deference to the meaning of individual consonants and vowels (/i/ and /w/ at the very least), and she is also reinforcing these meanings by using them in a new context. She is using Semantic Association not on a word or a prefix, but on consonants and vowel. She is imbuing consonants and vowel with meaning.

3. The third type of sound-meaning is the direct, unmediated, visceral effect of the sound on us. I call this True Iconism, in the sense that Charles Sanders Peirce (who coined this use of the term 'iconism') used it. Observe that any form whatever has some sort of meaning. If you see a car, you have a sense for what sort of person owns it. The shape of a piece of furniture or the body movements of a cat or the pattern on the wall paper all convey some experience that on some level informs us. We by definition can't articulate what the effect is, because articulation always requires denotations, and this is happening on the level of connotations. The same is true of speech sounds. If I say a nonsense word, 'sping' or 'toip' or 'glorp', these sounds are all different and convey different things to our feeling. Once the word is 'assigned' to a referent and gets put into service, like 'spring' and 'toy' and 'gorp', that effect is masked by the referent, but it does not cease to act altogether. That effect is still influencing the word. There are several ways to test for this too. One way is to look at a group of words that all refer to the same thing, and that differ only in their sound, like 'stamp', stomp', 'tamp', 'tromp', 'tramp', 'step'. Or 'gleam', 'glimmer', 'glitter', 'glisten', 'shine', 'shimmer', 'flash', 'flame'. You find in these cases that the differences between the meanings of the word is connotative rather than denotative, and that it correlates nicely with the form of the word. 'Stamping' comes down more forcefully if there is an /m/ before the /p/ (slam, bam, wham, smack, mash,...). The /r/ and the beginning of the word allows the motion to go forward, especially after a /t/ (rush, run, travel, train, trip,...) The /p/ in all those words is what emphasizes the individual steps. An /m/ in connection with light makes the reflecting surface smooth and suggests the beginning of something (make, mother, mold, form,...). /gl/ light is indirect. /fl/ light is direct (front, forward). All the sounds pronounced at the back of the mouth have something hidden about them (/g/, /k/, /h/, /ng/).

Clustering is language-dependent. If the basic word for 'house' in a given language starts with a /h/, then by clustering, you expect disproportionately many words containing /h/ to concern housing: hut, home, hacienda, hovel,... But whether 'house' begins with /h/ is a matter of reference/ denotation, and it is therefore arbitrary. So in a language where the word for 'house' begins with /d/ instead of /h/, you'll probably get clustering of house words to /d/ instead of /h/. The clustering meaning of /d/ will therefore differ from language to language, and will be more similar in closely related languages. But the tendency to cluster is universal. All languages have a clustering dynamic. But true iconism is universal. The phonaesthetic effect of a /d/ is the same whether you speak Swahili, French or Tagalog. So whether a lot of light words will begin with /fl/ will vary from language to language. But if the light word begins with /fl/, it will be direct light. Consonants pronounced at the lips like /f/ make things direct. Thus whether /f/ brings to mind 'light' will vary from language to language. But how it affects the light will not vary from language to language.

Inherent Meaning, True Iconism

Notice that although our bodies are different from those of dogs, we can read a dog's emotions. When it growls, we know it is angry. When it wags its tail, we know it is happy. When it puts its tail between its legs, we know it is frightened or ashamed. We know this despite the fact that we ourselves have no tail, and hence never use this mode of expression. We know it, because there is something implicit in the form of the wagging that evokes happiness. The happiness is inherent in the form. It is not really reference. Wagging does not refer to an emotion 'happy' in dog language. A dog doesn't learn to wag. A dog just pops out of the womb barking, growling and wagging its tail.

So as it turns out, English (and every other language) also uses just this sort of sensitivity to form to a very high degree and in very subtle ways. Each sound ­ each vowel and consonant ­ has an inherent feel to it, and that feeling informs the meaning of every word that contains it. The sequential flow of the vowels and consonants in a word also affect the meaning. 'fl' at the beginning in 'fly' and 'flee' and 'flow' is different from 'f' followed by a vowel followed by 'l', as in 'fall' and 'fail'. The inherent meaning of the 'f' and the 'l' remain constant, but the role they play in the word differs significantly depending on whether they precede or follow the vowel. Before the vowel, the form the backdrop, the stage on which the event in the word occurs. After the vowel, they express the punch line, the conclusion.

Because this 'inherent' meaning is implicit in the form, it is universal. The feeling of a 'b' will be the same in every language that uses the sound 'b', and that feeling will directly reflect how 'b' is pronounced. It involves a barrier. It is loud, sudden, explosive. It's high pressure within, low pressure without. It involves a breach, a violation of the integrity of the barrier. Languages vary on how they pronounce 'b', and consequently the feeling of 'b' in various languages also varies. Slavic languages have a palatalized (soft) and a non-palatalized (hard) 'b', and these two consequently have different feelings associated to them and occur in different kinds of words.

Also, because the inherent meaning is implicit in the form, you cannot express directly what it is. I cannot write down what the meaning of 'p' is, because whatever I would write down would be the meaning of the words I wrote down, and not the meaning of 'p'. Whereas referential meaning is indirect, inherent meaning is direct, unmediated. You can only really get to it by communion, by feeling into the words deeply.

Clustering and Archetypes

As we've seen, within a given language, certain basic concepts tend much more strongly toward one letter than another. For example, words concerning 'babies' and 'birth' are much more likely to begin with a 'b' in English than with any other sound. Words concerning 'gardens', 'greenery' and 'growth' are very likely to contain 'g' followed by 'r' in English. Words concerning 'travel' fall in /tr/. Clustering, unlike the basic feeling-tone of a sound, is language-specific. Birthing goes with 'b' in English, but not particularly in Russian. Still, some clusterings are surprisingly universal. For example, in languages as diverse as Japanese, Polynesian, Russian and English, 'p' is associated with 'pricking' (a long pointed object piercing a surface), whereas 'b' is associated with 'beating' (violent contact which does not necessarily involve an instrument and does not necessarily break the surface). (This suggests to me that these languages ultimately all have a common origin someplace.)

As it turns out, if you look at all the words that contain a given sound, you find that they cohere into a semantic space that has a sort of archetypal quality to it. Bateson says that we think in terms of stories. Each consonant has implicit in it a story, which is different from the story of other consonants. The English /s/ centers around what you might call the 'serpent'. Not only does it have tons of 'snake' words out of all proportion to the actual number of snake words in English, the remainder of basic vocabulary which surrounds /s/ can easily be thought of in 'Garden of Eden' terms, as centering around knowledge, betrayal, slyness, sex and expulsion.

Linguistic Ecology

When we speak, the Word we use is affected by the Sounds who form them. But the Sounds are affected by how we use the words in which they live. We saw that the clustering meanings of /i/ and /w/ concerning 'smallness' and 'water' were reinforced by inventing 'smish'. And when you look a word up in a dictionary, you find listed a number of senses. Consider what each sense is. It is a description of one of the word's primary functions. And what is this function in essence? It can be thought of as a set of contexts in which that word commonly appears. Every time you use a word, you place it in a context among other words. The sense listed in the dictionary is a summary of some of these contexts.

So each time you use a word, if you 'say something new', you are putting it in a new context relative to other words. You are giving it a slightly new job description. And if this little experiment in juxtaposition is accepted by others, we say your sentence was 'meaningful'. And so once you have done that, you have changed the meaning of the word ever so slightly, and consequently, you have influenced each of the sounds that make up that word just a teeny little bit. As a result, every word that contains those sounds is ever so slightly affected by your new sentence.

So one important effect of speaking English is to cause English herself to evolve. The more prone you are to place your words in the already existing patterns, in 'government jobs', the less likely is your speech to have an effect on the language itself. But if, like Shakespeare, you rarely use a word in the same way twice, but feel into its peculiar genius and just let it fly, you will have a much more powerful effect on the language. Your words will dance, and we fellow speakers of English will dance with them.

Now, some words like 'modem' have almost exclusively worked 'government jobs', whereas others like 'funky' have held very nearly none. You can't point so clearly to what these itinerant words do with their lives. You have to feel them to use them. This is true also of all the really basic words of English, like the prepositions and conjunctions, verbs like 'take' and 'get' and 'be'. It's much less true of concrete nouns.

But both the hobos and the government employees in a language deserve our respect. English would be unusable if either class ceased to exist. The concrete nouns consent to remain firmly anchored to specific things in the world, because otherwise we couldn't say anything. And furthermore, all the rest of English is anchored to the world through these. 'House' is anchored firmly and unambiguously to a very specific set of objects in the world. That is, there is very little ambiguity among English speakers concerning what things in the world are 'houses'. But 'mansion', 'hut', 'cabin', 'cottage', 'hacienda' are anchored to the word 'house'; they only qualify it a little. A 'mansion' is a house... a big house. On the other hand, if all the words were as stuck as concrete nouns and scientific terms, we couldn't talk either. The one word 'put' in 'put up with', 'put away', 'put down', 'put through', 'put past',... can be fully understood really only by knowing the clustering meanings of English /p/, /u/ and /t/. These words are almost impossible to explain to a foreigner, for example. You just hang out with English for awhile and start to learn how /p/ and /t/ cluster, and when you know what they mean, you can use 'put' fluently. 'Put' begins with 'p' (a specific point), and ends with 't' (directedness toward a goal).

Of course, the effect of the sounds on the meanings of the words you use is very powerful and obvious by comparison with the effect of your individual sentences on the English language. Your novel sentence will carry relatively little weight compared to the vast semantic ballast of the English language. So a new context for a word will affect the English language just a little, whereas the English language affects the usage of the word a lot. Nonetheless, the way we talk and think as a nation matters a great deal, for it is literally incorporated into the language which our progeny inherit. There's a linguistic ecology just as surely as there's a planetary ecology.

Concrete Nouns

The most common way to go about trying to identify the clustering meanings of vowels and consonants is to form phonesthemes. That is, you identify classes of words, like 'gleam, glimmer, glisten, glow, glare,...' or 'glue, glom, glop,...' in which a particular sound or sound sequence appears more often than you'd expect were sound-meaning correspondences purely random. These sound/meaning classes are called phonesthemes. Now a certain type of words have a harder time fitting into a sound-based classification than other words. These uncooperative words are those like 'house' which have relatively unambiguous referents. They fit in a surprisingly small set of classes of a different kind. They are words referring to people, body parts, clothing, games, animals, plants, plant parts, food, tools, furniture, musical instruments, weapons, buildings, rooms, vehicles, containers, types of cloth, minerals, colors, symbols, titles and units of measurement. Most of these words are what linguists call 'concrete nouns'. And whether or not they are concrete nouns, what characterizes these words is that they have a very specific referent which is very clearly defined. That is, we may easily differ over what things in the world are 'interesting' or 'masterful'. We may not be clear on whether I smashed, smushed, squished, smished, squashed or mashed that piece of paper. But we aren't confused about which things are giraffes, and we don't get confused about whether to classify something as a giraffe, an antelope or a moose. That is, we agree nearly unanimously on what set of objects or people fit in the classes of concrete nouns, that is, on what set of objects in the world constitute 'hammers' or 'pilots' or 'maples' or 'sushi'. As Plato puts it:

Socrates: When someone says the words 'iron' or 'silver', we all have the same objects before our minds, haven't we?
Phaedrus: Certainly.
Socrates: But what about the words, 'just' and 'good'? Don't we diverge and dispute not only with one another, but with our own selves?

However, these concrete words are 'detector' words. They know exacly what they refer to. They are anchored fast in the world.

Levels and Context

Language, like nearly every other complex system, is composed of a series of levels. Elements at each level are placed in a context among other elements on its level. The whole is more than the sum of the parts. The whole then constitutes an element on a new and higher level. The phonemes (speech sounds/letters) are placed in context to form morphemes (prefixes, suffixes and roots). Morphemes are juxtaposed to create words. Words are juxtaposed to form phrases, and phrases are put together to form sentences. Whenever a word is placed in a sentence, only part of it is used. It is serving a particular function in that sentence which is only a subset of all the roles that it can play. Nonetheless, the entirety of the sound of the word is present every time it is used. This is the case on every level of language.

When you place a consonant in a word, only part of it is being used. In the words 'squish', 'squeeze', 'smash', the constricting and powerful aspects of 's' are emphasized most. But in 'slip' and 'slide' and 'slick', its smoothness is more pronounced. Its a curious phenomenon that within any given consonant, it is very common to find a thing and its exact opposite strongly represented. For example, in 't', you find many words of 'travelling' (trip, train, trail, track) but also many words 'tying' (tack, tether, trap), words of 'tenderness', but also 'terrifying' words. In 'd', you find day/darkness, do (drive, drill)/dam up (dampen, die, douse). In 'b', black/blank, blame/bless, build/break. There are thousands of examples of this nature. As it turns out, opposites are very close in meaning. The opposite of 'dark' is not 'cheesecake'. The opposite of 'hard' is not 'turquoise'. The opposite of 'man' is not 'carpet'. Indeed a word and its opposite are in general the same in every way but one. In fact, the words which are closest semantically are antonyms.

So these various levels of language, like the letter, the morpheme (prefix, suffix or root), the word, the phrase, the sentence are like prisms that split the light. What was one thing on the level of the phoneme is apparently split into many things by its placement in context, and it appears on the surface as a thing and its opposite. What was merely 'extent' at the level of the sound 'l', becomes 'long' in one set of contexts and 'little' in another. To find 'l', one must look at what 'light' and 'land' and 'liquid', what 'blowing' and 'flowing' and 'clustering' and 'glomming' have in common.

Testing the Hypothesis Formally

So we want to see whether the relationship between the sound and the meaning of a word is arbitrary. What form would such a test take? Well... for clustering data, you'd have to do it one language at a time. To test it for English, you'd have to look at every word that contains a 'b' in English and see if you could find an element of meaning that is absent in every word that does not contain 'b'. You'd have to do this for every consonant and vowel. You'd have to pay attention only to how the word is pronounced, not how it's spelled.

How would one do this? Let's start this way... I'll simplify the data by taking into account only words which contain no prefixes or suffixes. Then I'll show that all those English words which contain the letter 'b' fit into a tight classification scheme­ a classification scheme that reflects the pronunciation of the letter 'b' itself, a classification scheme that reflects the meanings I have given for each of the consonants. This classification, if it's to pass as a serious test, must have the following characteristics.

1. You must agree with me that the words in the classes (called phonesthemes) have something in common that other words don't have. For example, you must agree with me that 'blimp' and 'ball' and 'balloon' and 'bulge' and 'bubble' and 'bloat' and quite a few others have an element of meaning in common that 'sardine' and 'rug' and 'wise' do not. This is the foundation of all the clustering experiments. You do not have to agree on what bulge and bubble and bloat and bubble have in common that distinguishes them from 'ship' and 'happy'. You only have to agree that they belong in a class together which 'ship' and 'happy' don't belong in. On that basis, I am off and running.
2. You must agree with me that these words all contain a 'b' sound and an 'l' sound.
3. The classes must be fairly tightly defined. That is, a classification scheme of 'big things', 'colored things', and 'verbs of sound' doesn't show much at all. A much better scheme would be 'things which are bloated' (relatively high pressure within a completely enclosed membrane), 'words for shades of brown', and 'verbs for exploding noises'. Those classes do in fact exist in 'b', but not 's' or 'g'.
4. Words which contain no 'b' sounds should not fit in the classification scheme, or put another way, words without 'b' that do seem to fit in the classes should fit differently. For example, 'puff' is sort of like 'bulge', but it's often fluffy. It doesn't have to involve a membrane. Besides, 'p' is pronounced very much like 'b'.

I need to produce such a classification for every consonant in English. Look through the consonants to review my Classification Scheme. My classification is of course not the only right and true one. The classes are only a side effect of a deeper process, and so there can exist no one right set. But the fact that it is possible to devise at least one such scheme seems to me very strong evidence that this underlying principle is operative in English.

And here's your chance to check out a part of my Sound Dictionary. Pick a non-compound word beginning with B and containing no prefixes or suffixes. Then look it up in the index, and find there the label or index of the classes that word fits into. The average word fits in 3.7 classes. I have broken the dictionary into two major types of classes, the concrete classes I mentioned above and classes containing words which are not necessarily concrete. This latter set of classes appears first. All the B-words in my vocabulary have been put into some class.

Part of the test for the effect of sound on meaning must take the form of such a classification scheme. But another part of the proof consists in showing that other words don't fit in that scheme. The most straightforward way to carry out the second part of the proof is to choose a large number of fairly narrow semantic classes, each of which contains words with a wide spectrum of pronunciations. Then you compare all the words that fit in these classes and show that their meaning is affected by their pronunciation in a regular way. That is, all the 'good' people beginning with 'b' are good in one way, and those beginning with 'p' are good in another, and so forth.

Further Examples of Tests

* Phonosemantic Dictionary
* Bias in the Labials
* /str/ in Several Languages
* Place words in /b/, /g/, /n/
* Comparison of /b/ and /d/ Classes
* Syllables
* Inversions
* Basic Words
* Some Semantic Classes

* /r/ in second position
* /r/, /l/ in All Positions
* People, People, People!
* Fire
* Animals


Interested in More?

See also the introductory chapters of my dissertation:

Abstract and Introduction
Theoretical Preliminaries


Try My Dictionary

Or My Book


Excepts from my book The Gods of the Word
from Truman State University Press


Oh, music!
If you abandon us
What will become of us?

Nicolai Vasilevich Gogol


Why Language Can't Function without Inherent Meaning

After quite a bit of communing with words in this way, it becomes pretty obvious that, for example, a tendency toward a goal really is introduced by /t/. And furthermore, that it very fundamentally affects how you use the words that contain /t/. It's not just a peripheral phenomenon. Anyone who has tried to learn a foreign language knows how difficult it is to understand exactly how to use the basic words like 'get' and 'take' and 'move' and 'find'. You have to hang out with the language for quite some time before you really start getting a sense for how it works. And it's really hard to give any good explanation for how it works. If you want to learn German, and someone tells you 'geben' means 'give', you quickly discover that it sort of does, but you certainly can't use 'geben' exactly like you use 'give'. Inherent semantics holds the key to understanding an important part of the usage of these words...

"Whenever the essential nature of things is analysed by the intellect, it must seem absurd and paradoxical. This has always been recognized by the mystics, but it has become a problem in science only very recently."

Fritjof Capra, The Tao of Physics


'Phoneme Physics'

I'd like to present a crazy idea first articulated by Paget(1930) which may allow us to make more sense of the differences we find between the two closely related phonemes, /b/ and /p/, and their clustering dynamics. This idea could perhaps be refined to such a degree that one could use it to quantify phoneme interactions and make predictions, but for now, I just think of it as a conceptual tool that helps me think about how phonemes interact.

If we are correct about the usefulness of drawing the distinction between inherent and referential semantics, then the mouth can be viewed as a system whose physics in part determines the meanings of words. Imagine that each phoneme has inherent physical properties. Speaking roughly, the stopped consonants /b/, /d/, /g/, /p/, /t/, /k/, form walls at different points in the mouth. /r/ like 'fire' has force pointing out of the mouth. The muscles in the mouth push outward when you say /r/. /l/ is like liquid or air. The mouth is very relaxed when /l/ is pronounced in isolation, but /l/ tends to move a little toward the point of articulation of its neighboring consonants. It can so to speak be induced to gain momentum when a force acts upon it, but its energy in the absence of other phonemes is 'potentional' rather than 'kinetic' like /r/.

Let us apply this very rough characterization to what we know about the semantics of /b/, /d/, /g/, /l/ and /r/. Assume for the moment that the magnitude of /r/ is greater than that of the voiced stops. That is, the muscles are more tightly constricted when pronouncing /r/ than /b/. Then, when /r/ hits /b/, it 'breaks' it into 'branches'. When it hits /d/, it 'drains' through into 'drops'. And when it starts from /g/, it is 'ground' into 'grains'. Imagine that /l/, on the other hand, cannot pierce /b/. That is, the magnitude and direction of force used to pronounce /l/ is less than that of /b/, so /b/, so to speak 'wins out'. So when /l/ swashes up against /b/, it 'bloats' it into a 'blister', but /b/ still manages to 'block' it. /l/ also does not have an inherent direction, unlike /r/. Therefore, /g/ can pull /l/ back into the mouth, so that whereas /gr/ grows out of the mouth, /gl/ stays stuck back in the throat like 'glue'. Also, /l/ seems to be able to 'leak' through /p/, and so we get a 'plane' rather than a 'bulge'.

In other words, such a 'physical' description of these sounds would be consistent with the fact that it is /gr/ and not /gl/ which grows, that it is /bl/ and not /br/ which is bloated, that it is /gl/ and not /bl/ which is gluey.This description does seem to make some intuitive sense of the phonosemantics of a large percentage of English words. I deduce the very general physical properties of the phonemes by how they interact in words, and this may or may not prove to correlate with muscle tension, air pressure in the mouth, etc. during articulation. Unfortunately there is probably no very reliable way to measure this muscle tension or air pressure. If there were a real physiological basis to this 'wild idea', one could imagine using word semantics to make educated guesses about muscle tension. This in turn could perhaps be used to make predictions about syllable structure and other phonological processes, as well as word semantics.

Let me suggest another way in which 'phoneme physics' might be used to give a unifying picture of the situation. Recall that we observed that although /b/ and /p/ have corresponding semantic classes in many domains, these classes cluster among themselves differently within each of the two phonemes. Most of the 'balls' of /b/ contain an /l/ and are bloated. Most of the corresponding 'peas', 'points' and 'pits' of /p/ contain no /l/ and are not bloated. When /l/ encounters /b/, it is 'blocked' and it 'bloats' up. However, when /l/ encounters /p/, it forms a 'plane' (plate, platter, plateau, etc.) Thus, the little round things in /p/ can't contain an /l/, because /l/ makes a plane of /p/. So although the corresponding classes of 'balls' and 'pebbles' exist in /b/ and /p/, the physical dynamic which forms them is on the whole very different.

Perhaps the most convincing evidence that evolution is a mental process is in its slowness, its fits and starts, its errors and stupidity. In a word, its conservatism. In a universe conceived by physics, there could be no stupidity, no conservatism, no tragedy and no humor.

Gregory Bateson
A Sacred Unity


A Biological Analogy

There was a view proposed primarily by Schleicher and also Bopp in the middle of the 19th century that a language is analogous to a living being. Schleicher was strongly influenced by Darwin. His view was overrun by the enthusiasm of the Junggrammatiker in the latter half of that century, who held that Schleicher's view, which allowed for an individual history for each word, was inconsistent with precise and general predictions about how languages evolve. Mendel had done his research, but was not recognized until the 20th century, and by then the Junggrammatiker approach had taken such firm hold in linguistics that Schleicher was laughed out of existence. I too hold with the Junggrammatiker and all of 20th century linguistics that all natural laws ­ linguistic and otherwise ­ must be stated in general and universal terms. God thinks abstractly. But I disagree that this is inconsistent with Schleicher's view. Perhaps Schleicher and the Junggrammatiker were both right.

Once again in the interest of thinking globally and unprofessionally, let me draw some obvious superficial comparisons. A language shares many significant traits with biological systems. In Campbell's (1990) introductory biology text, life is characterized as having the following fundamental properties:

1. It has a hierarchy of organization (atoms, molecules, organelles, cells, tissues, organs,...). Language has a similar hierarchy of organization into phonemes, morphemes, words and sentences.

2. Each level involves emergent properties or properties which cannot be determined by the sum of the parts. In linguistics, we have also observed emergent properties. These are largely due to idiomatization/reference/interpretation. That is, because of reference and its clustering dynamic, a monomorphemic word is more than just a string of phonemes.

3. The basic unit of life is the cell. Language also has a basic unit, namely the word. The minimal organism is a cell. The minimal utterance is a word. This is the level of 'identification'.

4. The correlation of structure and function is very important in biology. Phonosematic data shows that there is a similar correlation of structure and function in linguistics. The mapping between sound and meaning is not arbitrary. Inherent semantics limits referential semantics.

5. Organisms interact with their environment. Language does also and each linguistic unit interacts with its context. It is this which gives rise to its emergent properties and this is the driving force in its evolution.

6. Organisms inherit genetic material. It is clear that languages and words do inherit something from their ancestors. We have observed that aspects of a phoneme's meaning are highlighted or suppressed in varying environments. This reminds one of the fact that every cell carries all the genetic material for the entire organism. But factors in the cell's environment serve to repress or promote the expression of certain genes. In biology, the factors which determine which genes are expressed revolve around the function of the cell in its environment. For example, a developing cell may be determined by its environment to be a muscle cell or a nerve cell. Shorter term environmental factors also affect gene expression, such as cellular response to a viral attack. So also in language, the factors that highlight or repress aspects of a phoneme meaning revolve around the function of the word within which it is contained, i.e. its referents, which can be thought of as its long-range contexts.

7. Unity in diversity ­ there is a tremendous variety of life, but biological systems on all levels fall into fairly clear classes. In like manner, linguistic units are diverse, but also fall into a fairly clear taxonomy.

8. Organisms evolve. Languages obviously also evolve as a result of their interaction with the environment.

In correlating biological systems with linguistic systems, we correlate what is essentially matter (organisms) with what is essential thought (language), and this has historically been a sticking point in drawing such an analogy. But the correlations are so strong, that it seems short-sighted not to try to learn what we can from the biologists and to use it to best advantage.

Margo's Magical Letter Page