kitchen table math, the sequel: IPA English vowel chart, oriented towards learners (not documentation)

Monday, January 11, 2010

IPA English vowel chart, oriented towards learners (not documentation)

So this is basically the IPA chart I used with Ali, just that it looks cleaner and there's less scribbling. (Click for larger version).

In my opinion (as in not backed by double blind clinical trials, rigourous empirical evidence or anything but simply my own intuition and experience with others), IPA charts are useful. They provide a "skeleton" in which to arrange the seemingly vast array of English vowel sounds.

I have not really been happy with the way the English vowel system is generally presented to learners, both native children and ESL learners. For one, the vowels are often presented in random lists, and learners aren't told how the vowels are related to each other. An IPA-based vowel chart is organised by two physical traits often used to classify vowels (among others): vowel height, and vowel backness.

Here, height and backness are relative. If you look up the IPA vowel chart on Wikipedia, you find distinctions like "mid-open", "mid-front", etc. but those are w/respect to "absolute" cardinal vowels defined by phoneticians -- and there is no single language that uses all of them. Most languages distinguish 2 or 3 heights, and distinguish 2 or 3 levels of backness. There are always exceptions -- but languages that distinguish 5 heights for example, may only have a handful of consonants, etc. It's a trend you find among the world's languages -- complexities in one area tend to be compensated by simplifications elsewhere.

The more "open" a vowel is, the higher its F1 formant frequency is (more close ==> lower F1 formant frequency). F2 formant frequencies are primarily influenced by vowel backness (more front ==> higher F2 formant frequency). Formant analysis in phonetics goes all the way to formants like F9 ... but their effects are more subtle. And of course formant frequencies have interfering effects on each other, but there are compensating functions to do that -- and I won't discuss any more phonetics because I really want to discuss literacy.

A brief vowel chart orientation (an initiation to IPA). In IPA, /j/ is used preferentially for the "yod" (as in yogurt), since /y/ is used for a vowel sound (that English doesn't use). English has (more or less) eight pure vowels:

æ -- "bad"
ɛ -- "bed"
ɪ -- "bid"
ʊ -- "good"
ɔ -- "law"; for people who make the cot-caught distinction. This pure vowel does not exist by itself in my dialect (New England rhotic)
ɑ -- "father"
ə -- "kernel"
ʌ -- "bud"

Now long vowels. Long vowels aren't actually pure vowels, though they used to be. They are actually diphthongs. A diphthong is a combination of vowel + semivowel (aka vowel-like consonant). Not all combinations of two vowel-like sounds are diphthongs. For example, if you pronounce "towel" with two syllables, there are three-vowel like sounds in there, but there is only one diphthong and there is no "triphthong". My rule of thumb:

1. A diphthong has to occupy the time length of one syllable. The semivowel component is generally shorter and more "clipped" than the vowel, thus the VV sequence is really behaving like VC. The preferred syllable structure in English (if not universally) is the structure CVC, which lasts for the length of one standard syllable. Deviations from this structure generally result in compensation.

For example: the word it. Kinda simple, but you never knew you implemented so many phonological corrections when you used this word! If you use it by itself (just say it to yourself), notice how, the /ɪ/ phoneme is stressed, and is slightly longer than say, the word in "sit". But put a consonant before it -- like in a sentence, and "it" will actually pull a consonant off your preceding syllable, and /ɪ/ will revert back to being an unstressed vowel of normal length. Thus, "make it so" can be analysed as: /mɛj.kɪt.sɔ w/; all three syllables have the structure CVC. This is part of what we usually regard as "fluency" -- our mind is so automatic and flexible, we unconsciously rearrange phonemes around based on sound laws we don't even think consciously about. These sound rules will also make consonants appear out of thin air:

by themselves: "you" "can" "see" "it" -- /jʊw/ | /kɛ~n/ | /sɪj/ | /ɪt/ [note stresses in bold]
in a sentence: "You can see it." /jʊw.kɛn.sɪj.jɪt/ (extra consonant suddenly duplicated from previous syllable: length compensation!) [stresses in bold, extra consonant italicised]

You can observe this as a rule in children. Young children and toddlers have incomplete length compensation -- the result is a sort of sing-song sentence structure we associate with two-year-olds. But analyse the sound structure of a six-year-old or an eight-year-old, and many elaborate sound rules suddenly appear.

Okay, that was a rather long aside. Second rule of thumb:

Diphthongs generally "obscure" their component sounds. Only by conscious analysis do you realise that the sound in "ay" (like stay) is made up of ɛ (like in bed) and the yod (the consonant of yogurt). Same goes for "how" (æ+w). I include rhotics in here too, because they have a tendency to change the vowel, but you can also make an argument for excluding them.

So here are the diphthongs of English:

æw -- "how"
æj -- "my"
ɛj -- "bake"
ɪj -- "bee"
ʊw -- "moose"
ɔw -- "bow"
ɔj -- "toy"
ɑw -- "sock"
ɑr -- "far"
ɑwr -- "war", "core" **
ər -- "wicker"
ʌr -- "kernel", "rehearse" **

** There's further discussion about these diphthongs but it would make this post too long.
There's also nasal vowels (like the vowel in sand) but I've decided to skip that for now.

IMO, this is better than the traditional short-long vowel system often presented, because that system really only allows for ten vowel phonemes and the rest of the vowels are randomly placed. Was always confused about the vowel system from day 1 when it was presented in first grade, until grade 10, which is when I learnt phonetics.

A pure vowel / diphthong contrast can be used to explain English spelling, whereas it's not apparent with a short-long system. So you have to explain why silent-e often generates diphthongs ("long vowels"). OK. Short answer: It's related to length compensation, like the syllable exercises we've been doing above. But why does double-o ("long o") not actually sound like a long "o" (like in stone)? Can you actually make sense of English spelling? Ah, that is for an upcoming post on the Great Vowel Shift.


Sara R said...

Hooray for the Great Vowel Shift!

I learned about this in my college language classes. I always wondered what the actual evidence for the Great Vowel Shift is. How do we know how middle English was pronounced, for example? There are no surviving sound recordings. Of course moderns can make guesses based on how poets of that age rhymed words, and so on, but is that sure evidence?

le radical galoisien said...

It's pretty easy with English; up until the invention of the printing press, people wrote phonemically (more or less). When the printing press was invented, the unification in print (or lack thereof) of all of England (which spoke like 100 variations of dialects) it got really weird (to make the story short), but it's still analysable.

Poetry and rhyme guides contribute evidence. Rhymes that made sense in the 16th century don't make sense now. If you read Chaucer, you see him rhyming words that no longer rhyme anymore -- liquor and flour; beak with break... etc.

There is a great deal of puzzle-solving and logical (if not mathematical) analysis in historical linguistics --- which is why I got pulled into it.

It's much harder for the Chinese languages, since Chinese writing isn't phonemic (though phonetic in basis). The pronunciation of the characters can diverge in a multitude of different dialects without the character changing.

For example, the same character that read "swra" (such a great sound) is now "hua" in Mandarin. It makes historical analysis much harder, but it can be done (of great help again are poetry and rhyme books that people published).

The same tools also come in handy when you're doing genetics -- both in terms of protein sequencing a contemporary sequence and the evolution of a genetic sequence. Funny how historical linguistics training is so useful in biochemistry LOL.

le radical galoisien said...

The method use to determine what people spoke in the past is called the "comparative method". Basically, a language tends to diverge into a multitude of dialects, and you can use this contemporary information to work backwards, because some dialects conserve feature A and B but lose feature C, some dialects conserve B and C but not A, etc. and you toss this in with analysis of historical texts and historical rhyme goods you can get a pretty robust method -- up to a point.

We can only reconstruct for the past 5000-7000 years usually, because past that you get a lot of "multiple solutions" (to put it in math terms) and there are way too many competing possibilities and the p-value for your hypothesis becomes really high (to put it in statistical terms). With Old Chinese the reconstructions are only reliably certain for Middle Chinese -- Shang Dynasty Old Chinese reconstructions are quite tentative.

It's a giant logic problem. For example, if you know that language B and C descended from some predecessor X not shared with language A, then you can rule out any hypotheses posing genetic transmission from A to B but not A to C.

It's a great mix of literature-corpus-reading and cold, hard logic. In the past (before the advent of modern linguistics) this kind of thing was called philology, but has since been absorbed into historical linguistics. One of the great "Eureka" moments of historical linguistics was when a British Governor of India noticed various structural and lexical similarities between Greek and Latin and the Indic classical language Sanskrit (and its vernacular descendants like Punjabi and Hindi), but noted that these similarities did NOT exist for the Southern Dravidian languages like Tamil.

le radical galoisien said...

okay not so good with my biographies -- William Jones, the person who posited a hypothesis that Latin, Greek, Gothic, Celtic, Sanskrit and Persian Farsi were all descendants of PIE -- was a British judge in India, not a governor of the colony. But still.