My G.A. Phonemic Transcription

12.01.2011 § 2 Comments

For purposes of both Phonetic Inversion and Sound Poetry I need a standard for transcribing General American English phonemes.

The Ñ Standard

Phonemic Transcription

It is important to stress that this is a phonemic, not a phonetic, transcription. That is, it does not distinguish between allophones within phonemes. In other words, I rely on the reader to be familiar with the behavior of these phonemes in context and thereby employ the appropriate allophone. I believe that even when used outside of familiar phonetic contexts, as often occurs in both my phonetic inversion and my sound poetry, the results are intuitive enough that to encode them would impede rather than expede understanding.


One of the main things I was striving for in my phonemic system was to make it context-free. That is, I wanted each letter to stand for one sound and one sound only, and not rely on its neighboring letters for pronunciation. I am extremely frustrated that there are 22 consonant sounds (ignoring affricates) in G. A. English, yet only 21 consonants to represent them. I finally resorted to incorporating that extra letter in the Spanish alphabet with which many Americans are familiar: ñ.

The matches I made between consonants and their sounds is not always perfectly intuitive. The ñ, for a good example, doesn’t make the “ny” sound as it does in Spanish, but the “ng” sound, /ŋ/. There is no reason for ‘q’ and ‘x’ to make the soft and hard “th” sounds other than that the sounds these two letters most commonly make are extraneous combinations of other sounds — “kw” and “ks”, respectively — and, being rare enough, were less likely to interfere with readers’ attempts to pronounce my work by asserting their normal pronunciations.

As for ‘j’ and ‘c’, they were assigned as a result of my approach to the two affricates: not treating them as distinct enough from a clustering of their constituent sounds that they would need separate definition. Now, I would not take this approach with diphthong vowels, since treating their constituent sounds as separate would imply separate syllables. But in a sound poetry world with frequent clusters of four, five and more consonants, splitting affricates wasn’t going to be any trouble.

What I mean is, since the “ch” sound is created by “t” followed by “sh,” and I didn’t have a letter assigned yet to “sh,” and ‘c’ was free — its two most common pronunciations “k” and “s” already being covered by ‘k’ and ‘s,’ respectively — it was nice to let it participate in the pronunciation of the sound that “ch” makes (and it’s not unheard of for ‘c’ to make the “sh” sound, as in “ocean” or “facial”). And since the “j” sound is created by “d” followed by “zh,” and I didn’t have a letter assigned yet to “zh,” and ‘j’ was free, it was nice to have it stand for the part of its traditional sound which was not already covered (and it’s not unheard of for ‘j’ to make “zh,” as in French loanwords such as dijon or déjà vu).

Futility of Context-Free Vowel Transcription

It should be no surprise that there are far more vowel sounds in English than there are vowel characters. I was thereby forced to allow vowels to combine up to two at a time, which meant that I had to correct for ambiguity in some other way.

I found that one simple rule solves for almost all ambiguity in the resultant strings of vowel characters: assume the minimum number of vowel sounds. That is, in a sequence of four vowel characters, the least possible number of vowel sounds would be two, each sound created by a pair of vowel characters. If one of these two pairs of vowel characters did not code for a vowel sound in the transcription system, then one would need to move on and see if the sequence of four vowel characters could be interpreted as a sequence of three vowel sounds: two created by single vowel characters, sandwiching one created by a pair of vowel characters in the middle. If even this failed, then the four vowel sequence would be read as a sequence of four vowel sounds, each produced by a single vowel character.

For example, consider ‘aoeu.’

All three of ‘ao,’ ‘oe,’ and ‘eu’ are found in the phonemic transcription system. However, we must assume that this is sequence is to be read as ‘ao’+’eu,’ since the interpretation ‘a’+’oe’+’u’ is three sounds, while the former is only two.

For another example, consider its retrograde, ‘ueoa.’

There are no pairs of vowels in this sequence that are members of the phonemic transcription system. Therefore, it can only be read as ‘u’+’e’+’o’+’a.’

It is still possible for ambiguous vowel sequences to exist. There are ten different three vowel sequences which can be interpreted two different ways, either one pair followed by one single, or one single followed by one pair (e.g. ‘aei’ => {‘a’+’ei’, ‘ae’+’i’}). There is even one five vowel sequence, ‘aoeui,’ which can be interpreted three different ways: ‘ao’+’eu’+’i’; ‘ao’+’e’+’ui’; ‘a’+’oe’+’ui’. What must be done in these cases is divide them up with apostrophes. Most of these ambiguous sequences should be exceedingly rare, but here they are nevertheless:

a’ei / ae’i
a’eu / ae’u
a’oe / ao’e
a’oi / ao’i
a’ou / ao’u
a’ui / au’i
e’ui / eu’i
o’ei / oe’i
o’eu / oe’u
o’ui / ou’i

Choice of Vowel Pairs

Since I know you’re dying to know why I picked each and every one of my pairs, as much as my mother was dying to hear why I put each and every Lego piece on my constructions when I was a kid, I will tell you why, and why every other option was less ideal. 🙂

It was a lot like one of those logic puzzles you did as a kid, you know, process of elimination on steroids.

I wasn’t sure if I was down with including double vowels. First of all, they look awkward and ugly. Second of all, they breed ambiguity. Third of all, only three of the five of them are useful, and since I’m going for regularity here as well, it would be subpar to include some but not all double vowels. So the double vowels were out.

Then there were a bunch of vowel character combinations that normally lead to two vowel sounds, such as ‘ia’, ‘ua’, ‘io’, ‘iu’, and ‘uo’, so I took them out of the running.

The next thing I did is just go through the vowel sounds and decide for each one, ignoring all the others, what would be the best vowel character or characters to represent it in the most contexts, just to get a rough first pass.

Six of the vowel characters and pairs were no sound’s bests: ‘ae’, ‘ie’, ‘eo’, ‘oe’, ‘ui’, and ‘ou’, which was not alarming, since I had five extra characterizations (2o of them to 15 sounds), so only one of my sounds necessarily was going to get paired with a less than ideal characterization.

Another result of this was that all of the single vowel characters were multiple sounds’ bests, so I was going to have to do some deciding between them. Those single characters, being the most attractive, were most important to get assigned to sounds that were common and really well represented by them.

I started with Schwa. Schwa, /ə/, is mission critical. As the sound of unstressed vowels, it is the most common sound in English, so a weird schwa would doom the transcription. I had to give schwa its only ideal, ‘e,’ even though that meant that /ɛ/ would have to take a less than ideal characterization.

Looking at ‘u’, it could be given to /ʊ/, /ɐ/, or /u(ː)/, but /ʊ/ and /u(ː)/ have other ideals, so I decided I should give it to /ɐ/ for now, and ‘eu’ to /ʊ/ and ‘ue’ to /u(ː)/.

Similarly, looking at ‘i’, it could be given to /i(ː)/, /ɪ/, or /aɪ/, but /i(ː)/ and /aɪ/ have other ideals, so I decided I should give it to /ɪ/ for now, and ‘ea’ to /i(ː)/ and either ‘ai’ or ‘ei’ to /aɪ/. Interestingly, ‘Ai’ and ‘ei’ were equally good for either /eɪ/ or /aɪ/; using ‘ai’ for /aɪ/ and ‘ei’ for /eɪ/ resembled the IPA transcription better and was as I was used to in the strict Japanese phonology, so I was leaning toward that, but if other more pressing needs of other matches conflicted, I was prepared to accept the other way.

‘O’ was weird because I hardly needed it. ‘Au’ for /ɒː/ and ‘oa’ for /o(ʊ)/ were each less ambiguous characterizations. I was just going to have to come back to assigning it.

‘A’ was sad since neither of the other ways to represent /ɑ(ː)/ or /æ/ were ideal. One of these two sounds was going to get the characterization shaft.

As for diphthongs, ‘ao’ for /aʊ/ and  ‘oi’ for /ɔɪ/ seemed to be pretty natural pairings.

So this is where I was at now:

The biggest problems were /ɛ/, /ɑ(ː)/, and /æ/. And boy were they problems.

It was becoming clear that there was not going to be any way to assign characterizations such that all of them would be intuitive in almost any context. So I began to change my values. I was going to start valuing regularity in my assignment to rival individual characterization appropriateness. In other words, it occurred to me that if I had 15 vowel sounds, they could be covered by each single vowel character plus precisely every combination of vowels, that is, every two vowel subset of the five vowel set. To shake things up, I tried combining the columns that were permutations of the same combination.

At first it would seem that I’d given myself more problems. Now ‘ao’/’oa’ is the best combination for two sounds, as is ‘eu’/’ue’, and this hasn’t seemed to have done anything to help with /ɛ/, /ɑ(ː)/, or /æ/. Well the first issue is easily solved: ‘ao’ knocks /o(ʊ)/ back to ‘o’. And the third issue is actually helped a bit: only ‘eo’/’oe’ is left as a possible characterization of /ɛ/.

Now with ‘o’ and ‘ae’/’ea’ taken, both /ɑ(ː)/ and /æ/ needed ‘a’, but they weren’t both going to get it. One of ‘o’ and ‘ae’/’ea’ were going to have to give their sound up. At first it seemed like I should give ‘o’ /ɑ(ː)/, since /o(ʊ)/ would be fine as ‘ou’, and if I gave ‘ae’ /æ/, /i(ː)/ would have nowhere to go. With ‘ou’ taken, that would leave /u(ː)/ to leave ‘eu’/’ue’ for ‘ui’.

So now here’s where I was at:

It would seem that I was done: I have a one-to-one match between each vowel sound and characterization. But I wasn’t satisfied. One of my problems was that while ‘ui’ does make /u(ː)/ in a handful of words like ‘fruit’, ‘juice’, ‘cruise’, ‘recruit’, ‘suit,’ etc., it’s just not an intuitive enough characterization, though I had to admit that the only other sound it could make, /ɪ/, was just as awkward and rare, in ‘build’, ‘guild’, ‘guilt’, ‘biscuit’, ‘circuit’; I did really like this maximal regularity thing, but it was inevitably going to lead to somebody getting this wacky ‘ui’.

That wasn’t the main problem though. The main problem was that ‘o’ may make /ɑ(ː)/ regularly, but word final, really wants to make /o(ʊ)/. I really wanted to give /o(ʊ)/ back to ‘o’, so that /ɑ(ː)/ could take ‘a’, since I was okay giving /æ/ to ‘ae’ (the characterization matches the IPA symbol, plus the name of that digraph in English is “ash”, pronounced the same as the tree, i.e. with the /æ/ sound) even though it’d knock /i(ː)/ off of ‘ae’/’ea’ and it’d have nowhere left to go other than knocking /ɪ/ off of ‘i’, which would be okay since I could give /ɪ/ to ‘ui’, knocking /u(ː)/ off ‘ui’, but I could put it on ‘ou’ where /o(ʊ)/ had come from. Phew.

The last thing of note was that either ‘eo’ or ‘oe’ could be used to make /ɛ/: ‘eo’ as in “leopard” or “jeopardy,” or ‘oe’ as in “oedipal,” “roentgen, “or a bunch of words for which the ‘œ’ digraph was reduced to an ‘e’ such as “ecology,” “ecumenical” and “esophagus.” I eventually decided to go with ‘oe’, for two reasons: primarily, that ‘eo’ outside of very specific contexts is usually pronounced as two vowel sounds, /iɒ/, as in “neoteny,” “geometry,” or “ideology;” and secondarily, that the sound ‘œ’ makes as an IPA symbol, /œ/, is the rounded version of /ɛ/.

And that’s where I ended up.

The H Standard

I developed a second standard which I believe most English speakers would be able to pronounce correctly with much less coaching than the first standard. The H standard is based on the heavy use of the letter ‘h’ as a modifier on the letter before it (and giving the /h/ sound to x, as in Spanish). Most of the h-modified consonants should need no explanation; only ‘dh’ as /ð/ might not be obvious, but is arrived at by analogy < t : d :: th : dh >.

Neither the consonants nor vowels are context-free in this system. I made an attempt to treat some vowels as modifiers in the same way as ‘h’, which, had it been successful, would have made the H Standard more beautiful than the Ñ, insofar as the consonants would be context-dependent in the same manner as the vowels were. Unfortunately, it was not possible, since the maximum number of sounds one can code for by converting vowel characters into modifiers is only 12: convert one vowel into a modifier, and you can use each remaining vowel character once plus once per modifier, of which there are two, h and that former vowel, for a total of 12. Convert two vowels into modifiers and the remaining three can be alone or modified by three different modifiers, again for 12 sounds. Go any further and you’re down to 10 or even 6 sounds.

The availability of ‘h’ to modify vowels did cause some reassignment. ‘Ah’, ‘eh’, ‘ih’, ‘oh’, and ‘uh’ take /ɑ(ː)/, /ɛ/, /ɪ/, /o(ʊ)/, and /ɐ/ respectively with very little hesitation. That freed ‘u’ to take /u(ː)/ and ‘a’ to take /æ/, most immediately. For some measure of regularity, I made it so that even though my converted vowel modifiers were not truly modifiers since they themselves could be modified and stand alone, the system otherwise behaved like they were modifiers, that is, in character combinations not involving ‘h’, only ‘i’ and ‘u’ occur at the end. This changed the characterization of /aʊ/ to ‘ou’, which is just as acceptable as the ‘ao’ in Ñ (and ‘ou’ is as good with either one for that matter, e.g. route vs. route). Also, plain ‘o’ was free, so I gave it /ɔ/, a sound which seems to be more important in Received Pronunciation, but I still seem to use it myself whenever /o(ʊ)/ is kind of short.

Other reasons I find the H Standard ugly is that ‘q’ is not used at all, ‘c’ is only used in combination with ‘h,’ and since beauty was out of the picture at a certain point, I decided to use ‘g’ like ‘u’ and ‘i’ as a partial modifier as well.

I will probably eventually write everything I write in one standard in the other as well, but I will default to the Ñ standard.

§ 2 Responses to My G.A. Phonemic Transcription

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

What’s this?

You are currently reading My G.A. Phonemic Transcription at cmloegcmluin.


%d bloggers like this: