Orthography: Vowels, Glides and Diphthongs

[chris_notts]

One thing I always struggle with is the choice about how to analyse and write vowels, glides and diphthongs, and what the best choice is. In natural languages, vowels and their associated semi-vowels often seem to be in almost but not quite complementary distribution, so that in most but not all cases you could predict the syllabification and whether the (semi-)vowel was the nucleus or not. If the choice is made to write vowels and semi-vowels differently, there's then another choice of whether any diphthongs exist (written as sequences of orthographic vowels) or if all potential complex nuclei are a combination of a vowel and a semi-vowel.

I think there are three basic common orthographic combinations, plus some hybrid patterns and non-phonemic orthographies. Spanish, for example, show both patterns sometimes, with <y> and <i> representing basically the same sound in different contexts, and English uses a famously complex system for orthographically representing vowel qualities.

Romanian: Vowel (i) = Semivowel (i) = Diphthong (Vi)
Quechua: Vowel (i) != (Semivowel (j), Combination (Vj))
Tariana: (Vowel (i), Diphthong (Vi)) != Semivowel (j)

Decisions taken in natural languages seem to be a mix of arbitrary choice and for analytical convenience. Possible criteria:

Vowel vs Semivowel

1. If i/j, u/w ... are in complementary distribution, then treat as a single phoneme. Normally written using the vowel letters, e.g. <i>, <u>, but in some languages with wide-spread vowel epenthesis or syllabic consonants it seems to go the other way. PIE often seems to get described in this way, with the high vowels being not quite fully paid up vowels because they show the same distribution as the syllabic consonants.

2. If syllabification is not predictable in some circumstances, then treat as two phonemes. E.g. if both ju and iw are possible then a sequence <iu> might be ambiguous in some contexts. May be ignored if the unpredictable contexts are very limited, which I get the impression is quite common in the wild.

Diphthongs

3. If there are strong collocational restrictions (e.g. ai but no ei, oi, ..., or au but no eu) then treat as diphthongs not as V + semivowel.

4. If part seems to occupy an onset or coda "slot" to the exclusion of other consonants, then treat as V + semivowel. For example, in a language with no onset clusters, if ia / ja cannot take an (additional) onset then analyse as ja.

5. Conversely, if part of a possible diphthong can satisfy a phonological or morphological requirement for an onset, then treat as V + semivowel. E.g. if an epenthetic consonant is inserted between adjacent vowels created by affixation, but no epenthetic consonant is inserted in the context of a possible diphthong.

I've been struggling with this a bit for Sint. On the one hand, there are restrictions on which combinations occur. There are four vowels, but only 3 "diphthong" combinations not the 4x3=12 you might expect. Furthermore, the "diphthongs" take exactly the same possible onsets and codas as any other syllable nucleus. However, the final element of the diphthong can become an onset in certain morphological contexts, e.g.:

tsajanaa
tsai-a-naa
BACK-P-go
"she came back"

A description of the morphology and spelling of morphologically complex forms would be simplified by a V + semivowel analysis, but this would slightly complicate the description of possible syllable structure(s).

How have others resolved this problem? Has anyone else found the intermediate / ambiguous status of semivowels an orthographic issue?

[Pabappa]

What are the three diphthongs? Given the one example , I'd say that yes, these sound like true diphthongs but I'd like to see more.

A diphthong always has a fixed stress contour. If the /i/ in /ai/ can ever serve as the syllable peak, then it undercuts everything else and you would need to analyse it as a sequence of /a/ and /i/, and just say that only certain vowel sequences are allowed. Also relevant for long vs double vowels, for the few languages that make that distinction.

[chris_notts]

What are the three diphthongs? Given the one example , I'd say that yes, these sound like true diphthongs but I'd like to see more.

Well, ignoring epenthetic vowels, the basic vowel system is /i e a o/. The possible "diphthongs" are ai ao oi. This is a bit complicated by the fact that stressed syllables can be short, long, or checked (glottalised).

What I was trying to demonstrate with that example is that there are environments where the "diphthongs" don't behave like monophthongs. If a vowel i, e, a, or o is followed by another vowel, then merger/deletion or consonant insertion would be used to avoid hiatus. With the diphthong, this isn't required because the high end point is already effectively a semi-vowel /j/ which also happens to be a consonant in Sint. There is very little phonetic difference between aj and ai, and even if there were I don't think this resolves the phonological and orthographic question of which is a better representation.

The main problem with the glide analysis is that then you have an ad-hoc rule that /j/ can only be a coda after /a o/, and /w/ only after /a/.

I think plenty of natural languages have similar issues, and different choices have been made. Quechua, for example, consistently writes the second half of what could be considered falling "diphthongs" with <y> and <w>, I think because these exclude another coda (although I'm not an expert on Quechua so I could be wrong). If there were diphthongs, you would have to posit another rule forbidding codas after diphthongs, whereas if they are monophthongs followed by glides then the coda behaviour is natural.

A diphthong always has a fixed stress contour. If the /i/ in /ai/ can ever serve as the syllable peak, then it undercuts everything else and you would need to analyse it as a sequence of /a/ and /i/, and just say that only certain vowel sequences are allowed. Also relevant for long vs double vowels, for the few languages that make that distinction.

You're right that this could be a criterion for a vowel hiatus vs single syllable decision. But even if you have a single syllable, it still doesn't tell you whether a more sensible way to represent it is /ai/ (complex nucleus) or /aj/ (vowel + coda consonant).

[chris_notts]

This is a good discussion about English and why for English the diphthong analysis is favoured:

http://seas3.elte.hu/phono/notes/glides.html