Danny Yee >> People >> Mike Salovesh
Date:         Thu, 13 Apr 2000 04:56:24 -0500
From:         Mike Salovesh <t20mxs1@CORN.CSO.NIU.EDU>
Subject:      Emic and etic?

Since Ron Kephart hasn't taken up the real challenge of this discussion, I guess it falls on me. God knows I've spent long enough talking with both Marvin Harris and Kenneth Pike about what they meant by "emic" and "etic". My conclusion is that there was no way those two were ever going to agree: Marvin Harris simply did not, and does not, and probably will never, understand what linguists like Kenneth Pike do. I therefore doubt that he will ever come close to understanding what Pike meant by "emics" and "etics".

As I recall, the terms "emic" and "etic" first appear in an article Kenneth Pike wrote for something called a "unified encyclopedia of the sciences". I remember the publication date as somewhere around 1938 to 1940; the publisher was the U of Chicago Press. The planned encyclopedia was abandoned, but a single volume was published containing all the articles that were on hand when the project died.

I got my copy of that volume as a gift from George L. Trager, and I used to keep it in a place of honor in my office. I can't cite it properly and exactly because it's still in one of the crates I had to pack in a hurry when I was evicted from my office. The NIU anthropology department decided that retired professors don't need offices. Or connections to the fiber-optic computer network, for that matter. All of a sudden, four of us who still live here and are very active in the profession found ourselves out in the cold. Moral: screw them.

The article was Pike's attempt to tell non-linguists about the procedures that underlie the practice of structural linguistics. His purpose was to suggest that other social scientists would do well to follow the same conceptual model.

Pike pointed out that linguistic procedure begins with fine-grained description. To linguists raised in the traditions of structural linguistics (or what some call Bloomfieldian linguistics), that had a very direct meaning. What the linguist was supposed to do was listen to human speech with an ear specially trained to note fine distinctions between one speech act and another. The training involved a great deal of practice in making marks on paper according to a system that provides a way of noting any sound that humans make with the speech apparatus. Part of the reason for all that practice -- always under the direction of an experienced practitioner -- was to calibrate one linguist's perceptions to standards shared with all other linguists.

The specific notation system that most phoneticians use started with the definitions attached to symbols in the International Phonetic Alphabet, or IPA. Recognizing that IPA was originally developed by people who spoke European languages for the purpose of recording European languages, descriptive linguists modified IPA to cover the entire set of all sounds that can be made in any human language. (Why a "set", not just "all sounds"? Because linguists agreed that they would only deal with sounds of a particular kind. They'll analyze a burp, a hiccup, or even a sneeze, but not a fart.)

The underlying drive was to ensure that linguistic analysis would begin with a corpus of observations about language that would be the same when made by any two competent investigators. The alphabet itself is immaterial; what is important is that two observers attend to the same behavioral phenomena and agree about what they have observed. The alphabet -- the symbolic notation system -- is simply a means of representing the essential elements of the behavior under observation. It is based on detailed knowledge of what happens to noises you make with your mouth when you vary any of the factors that produce sound differences.

The distinctions to which a phonetician attends include, among other things, noting whether the sounds, the vibrations in a medium, are produced by air being drawn into the lungs or other cavities or by air being expelled from those cavities. They go on to whether the vocal cords are vibrating during the production of the sounds, and where the path from the lungs or other cavity is modified (by closure, by constriction, etc.) by which parts of the vocal apparatus. (Side note: In essence, whatever the symbols used, their purpose is to record a set of formulas that can be played back to reproduce the sounds heard by the linguist. "Played back", here, means read aloud by someone else trained in phonetics and judged as meaningfjul by the original speaker.)

Faced with an actual speech event, two phoneticians would agree that a particular sound is produced by the two lips coming together to shut off the stream of sound, and that the vocal chords are vibrating when this happens. When they talk about the resultant sound, they'll call it a voiced bilabial stop. One might write the Roman letter B while the other writes a Greek beta, but what the symbols in their notebooks mean in terms of how the sound was produced would be in one-to-one agreement. Both would try to note the finest details of any little variation they could perceive. Every detail is a relevant phonetic fact.

Phonetics, in short, uses a universal system for organizing observation that is independent of the observer. Two observers should produce records that back-translate to exactly the same sounds. The system produces a record, on paper, that is sufficient to represent the sounds any linguist should hear in a single stream of speech. The fineness of detail in that record is limited only by the training of the observer in applying the universal scheme.

Hence "etic", for Pike, means the kind of detail that any well-trained observer would record for any language event -- regardless of the language in which it occurs or the language the observer speaks.

Let me repeat Pike's main point, so far. "Etics", at least in theory, is universal in two senses: it can be applied systematically and fruitfully to any body of language behavior in the real world, and any two properly trained observers will come up with essentally the same record of that behavior.

Linguistic analysis doesn't end with a phonetic notation of how a vocal apparatus made a bunch of noises. That's just the initial observational data.

Speaking now of the physics of the vocal apparatus, nobody ever pronounces a word the same way twice. Not if you slice the components finely enough. The position of the tongue in producing a vowel-like sound, for example, isn't determined by locking a gear into a tightly defined detent. In trying to say the same word twice, the tongue raises or lowers to slightly different points -- if you measure the difference in millimeters or even finer measurements. The sounds produced depend on the exact shape and volume of the space defined by the position of the tongue, and an extremely small difference in those elements produces a different sound. A skilled phonetician can notice lots more differences of that sort than the rest of us: that's what phonetic training is about.

Clearly, if most of us don't even know how to notice all the differences a phonetician can hear, then all differences aren't equal. Some differences make a difference and some don't.

I'm a guy who speaks two languages comfortably, and I can make myself understood one way or another in perhaps half a dozen more. Each of those languages requires that I pay attention to a different set of differences between sounds.

Let me take a shortcut to what that means. When I speak English, the differences in sounds between the words we spell "beat" and "bit" are quite important. They unequivocally show that those are two different words. When I speak Spanish, however, I can use either one of those sounds to replace the other without being misunderstood. The differences between the sounds never produce a difference in meaning. If I had spent most of my life speaking Spanish as my only language, I would have a terrible problem if I then tried to learn to speak English. Just think of the trouble I could get into by substituting the vowel of "bit", instead of the vowel in "beat", when talking about "the sheet on my bed".

After a phonetician has recorded an extensive body of data about how a speaker produces the sounds of a language, the next job is to find out which of the differences heard by the phonetician makes a difference in the specific language of that speaker. That's the job of phonemic analysis -- or phonemics, in short.

Condensing a lot of linguistic theory and assumptions, I'll assert that the application of phonemic analysis begins with the assumption that any specific language makes use of only a limited set of the sound differences the mouths of its speakers could produce. Most sounds that they could make are simply not used for language purposes. (How do we know they could make those sounds if they don't use them? Because other people, whose vocal apparatus is essentially the same, do make use of those sounds.) Within the range of sounds that are used in a single language, some differences always make a difference.

To find out what differences make a difference *within a single language*, it's necessary to reduce the almost infinite number of distinctions that a trained phonetician might attend to down to a much smaller number. That means lumping bunches of sounds into classes of sounds. A meaningful linguistic analysis aims to lump all the sounds of a particular language into sets whose members, though different, are equivalent to each other. The equivalence is defined by saying that a native speaker of the language will accept two words as "the same" when their only significant difference is that one sound has been substituted for the other.

A set, or class, of sounds that don't make a difference when substituted for each other in a single language are members of a single phoneme in that language.

That's only the beginnning of an operational definition of "phoneme", however.

Sometimes two sounds that are quite similar can't be substituted for each other, but still never produce a difference in meaning. One clasic example includes the sounds standard English spelling represents with the letter "p" in the words "pot" and "spot". Hold your hand in front of your mouth and say those two words, and you'll feel a puff of breath against your hand after the p of pot that you won't feel after the p of spot. (I used to show this in a classroom by holding a lighted match in front of my mouth and saying spot, which had little effect, then saying pot, which often blew out the match.) The two sounds clearly are different. As it happens, they never occur in the same place.

That is, I can write a rule that says "if the sound environment is at the beginning of a word and the following sound is a vowel, the unvoiced bilabial stop is aspirated. If the sound occurs between word-initial /s/ and a following vowel, the unvoiced bilabial stop is not aspirated." In other words, /p/ at the start of a word is followed by a puff of air if the following sound is a vowel; if the word begins with /s/, followed by /p/ and a vowel, there is no puff of air. (Note about standard linguist's shorthand: /p/ indicates "the phoneme this system indicates with the symbol 'p'.")

In doing phonemics, one rule is that two similar sounds which never produce a contrast in meaning can be included in a single phoneme. I won't bother specifying the rest of the rules and procedures followed in reducing a body of phonetic data to a phonemic analysis of the data: this is enough to give the general idea. The whole idea of such an analysis is to reduce a detailed and complicated set of observations expressed in a universally applicable framework to a much smaller set of classes, each of which includes a group of similar sounds whose differences don't make a difference IN ONE SPECIFIC LANGUAGE.

Side note: Please take what Ron said seriously. Linguists are trained to take their notes in ink, and never erase what they write. Mistakes are part of the data, and they often turn out to reveal more than all the things that are recorded "correctly". Similarly, phonetic data are never discarded in the process of phonemic analysis. All the different sounds written in the phonetic transcription of a body of linguistic data are retained. Phonemic analysis is a way of classifying those phonetic data into larger units on a functional basis.

"Phonemics" -- and what Pike meant by "emics" -- is just the classification of data that anyone, outsider or insider, can be trained to perceive into sets of data that are treated as significant units within a single language. The sets, the classes, of data that are significant units in one language are unique to that language.

Marvin Harris muddied the waters considerably when he tried to make "emics" and "etics" fit into the framework of the views of "insiders" and "outsiders". That is just not what Pike's kind of linguist -- or anybody trained in classical, Bloomfieldian linguistic analysis -- would understand by the words "emics" and "etics" that Pike created.

Wow. I've stayed up nearly all night trying to reduce most of a course in phonetics and phonemics into a single message. Obviously, I didn't succeed in condensing it very much. I hope what I've said here helps at least some Anthro-L folks get more of a handle on why Harris's words "emic" and "etic" just doesn't come out of the same universe of discourse as the same words as used by Kenneth Pike. (And that's why at least some of us avoid Harris's words entirely.)

-- mike salovesh <salovesh@niu.edu> PEACE !!!