Sound Change Quickie Thread

[bradrn]

Well the world's your oyster. Any one of these, or conceivably more than one depending on the vowel sequence, might happen. There are probably other processes that could happen that I'm not thinking of, too.

• Vowel deletion (with or without compensatory lengthening): **gaʔen → *gaen → ga(ː)n, ge(ː)n
• Consonant (e.g., glide) epenthesis: **gaʔen → *gaen → gajen
• Coalescence (±compensatory lengthening): **gaʔen → *gaen → gɛ(ː)n
• Diphthongization/glide formaiton: **gaʔen → *gaen → ga͡ɪn, gajn
• Metathesis: **gaʔen → *gaen → gane (This one might be a little cooky)

The problem I have is - exactly what happens? Take vowel deletion: if I have (say) 5 vowels, that makes 25 vowel pairs; which vowel gets deleted? What sort of rules have languages used to resolve this situation? (e.g. always delete first vowel, always delete second vowel, etc.) The same happens with epenthesis (which semivowel gets epenthesised?), coalescence (what do they coalesce to?), and diphthongization (it's unrealistic to have a diphthong for every vowel pair).

If you look at real-world examples of this, then the answer is that every vowel pair is liable to change individually. Sometimes you get patterns, like e.g. all back vowel + i diphthongs do the same thing, but not necessarily. For instance, Mongolic languages lost /j/ before /i/, and in modern dialects the resulting Vi diphthongs change in several ways: In Khalkha, <ai ei oi öi> coalesce into /{: E: 9: 2:/, but <üi ui> retain their original values of /uj Uj/ (except in the negative <-güi> /gwi, gHi/). In Chakhar, on the other hand, <ai> largely continues unchanged, but the front rounded vowels happen; and in Buryat <ai> coalesces but <oi> becomes /oE/ with a mid onglide (<öi> is lost as part of a general merger of short <ö> into <ü> initially and <e> elsewhere).

Thanks! However, I actually did just manage to resolve my problem with vowel sequences: I discovered that Luganda had a nice morphophonological change where a high vowel > semivowel / _V, with all other V > Ø / _V, and so I decided to steal that. (Although in Luganda, the second vowel lengthens, which I'm not including.)

(BTW, what are /{: E: 9: 2: E/? I don't know which transcription method you're using)

[KathTheDragon]

That's Sampa. A lot of people just don't want to spend any extra effort to type IPA.

[bradrn]

That's Sampa. A lot of people just don't want to spend any extra effort to type IPA.

That makes sense. I checked X-SAMPA, but I didn't think to check the original SAMPA. (The {: confused me because I didn't realise the : was a diacritic, so I was looking for {: as a single vowel.) So translating: /{: E: 9: 2: E/ corresponds to /æː ɛː œː øː ɛ/.

[Whimemsz]

That's Sampa. A lot of people just don't want to spend any extra effort to type IPA.

That makes sense. I checked X-SAMPA, but I didn't think to check the original SAMPA. (The {: confused me because I didn't realise the : was a diacritic, so I was looking for {: as a single vowel.) So translating: /{: E: 9: 2: E/ corresponds to /æː ɛː œː øː ɛ/.

X-SAMPA also uses : for length, so that transcription could be either, and was presumably intended to be X-SAMPA (but since they're the same in this instance it's irrelevant). But anyway yes, your readings are correct.

[StrangerCoug]

I'm working on a conlang with the phonology /a e i o u p ɓ t t’ ɗ k k’ q q’ ʔ m n ŋ t͡s t͡s’ t͡ɬ t͡ɬ’ t͡ʃ t͡ʃ’ s ɬ ʃ h l r j w/. The syllable structure is a strict (C)V(C)—vowels can be long, but that means you can't have a coda consonant, and either way, a syllable can't (morphophonemically—see below) end with an ejective or implosive. The only other phonotactic constraint than that is that /ij uw/ don't occur syllable-finally—you have /iː uː/ instead.

The tenuis stops /p t k q/ are defined as being non-glottalic and there is no contrastive voice yet, so [b d] would be interpreted as /p t/ instead of /ɓ ɗ/ by this conlang's speakers. They are supposed to voice between vowels, though. While I'd also like them to voice adjacent to certain consonants, I feel like the exact mechanics need a second pair of eyes: I have them as voicing before a sonorant, but I plan a general metathesis from stop-nasal to nasal-stop later, and I'd eventually like contrastive /mp mb nt nd ŋk ŋɡ/ without getting rid of my ejectives (getting rid of my implosives, however, is fair game, though note that I don't have /ɠ/). Nowhere other than where I said in this post do I explicitly have a voicing environment defined, but geminate glottalic and/or affricate consonants are marked in spelling with a preceding tenuis stop.

What tweaks can I do?

[Pabappa]

But wouldn't you have all those contrasts anyway just from the implosives turning into regular voiced stops, while the voiceless stops stay voiceless after nasals? All six of /mp mb nt nd ŋk ŋɡ/ would contrast so long as original /mp,nt,nk/ stay that way.

Sorry cantbtype IPA on phone

[Nortaneous]

You could just drop some unstressed vowels to create clusters. Maybe only high vowels, since they tend to be shorter than low vowels.

Or do something like Japanese: u > ɿ > ɯ, but mɿ > n (and tɿ dɿ > tsɯ (d)zɯ, all the other stuff you'd expect from a superhigh vowel)

[StrangerCoug]

But wouldn't you have all those contrasts anyway just from the implosives turning into regular voiced stops, while the voiceless stops stay voiceless after nasals? All six of /mp mb nt nd ŋk ŋɡ/ would contrast so long as original /mp,nt,nk/ stay that way.

Sorry cantbtype IPA on phone

Can't get /ŋɡ/ that way since there's no /ɠ/. I plan conditional denasalization of /ŋ/, but I don't find /ŋŋ/ -> /ŋɡ/ plausible without that kind of fortition happening to the other nasals (it is admittedly one way to answer my question, though).

You could just drop some unstressed vowels to create clusters. Maybe only high vowels, since they tend to be shorter than low vowels.

Or do something like Japanese: u > ɿ > ɯ, but mɿ > n (and tɿ dɿ > tsɯ (d)zɯ, all the other stuff you'd expect from a superhigh vowel)

Dropping unstressed vowels to create clusters was something I thought about.

Edit: Remind me to actually drag myself out of bed and boot up my desktop when I want to post on this site...

[StrangerCoug]

Also, really for my information more than anything else: Since there are no voicing contrasts until the implosives become regular voiced stops and merge with the allophonically voiced tennis stops, is it plausible with my phonotactics to have allophonically devoiced implosives? If so, where?

[Pabappa]

I meant that you'd have /Ng/ from primordial /kN/ and /Nk/ from primordial /Nk/. If I'm reading it right. N=eng, of course

[StrangerCoug]

I see now.

[dɮ the phoneme]

How reasonable is a change that deletes unstressed vowels / C1_C2, where C1 is more sonorant than C2?

A more general question: does anyone know of a good crosslinguistic survey of different types of vowel deletion? Either synchronic or diachronic.

[WeepingElf]

How reasonable is a change that deletes unstressed vowels / C1_C2, where C1 is more sonorant than C2?

I see no problems with that. ['kanad] > ['kand], why not?

[dɮ the phoneme]

How reasonable is a change that deletes unstressed vowels / C1_C2, where C1 is more sonorant than C2?

I see no problems with that. ['kanad] > ['kand], why not?

It will also give things like [ra'ka] > [rka] without [ka'ra] > [kra], the latter of which seems like perhaps the more natural developement.

[bradrn]

A more general question: does anyone know of a good crosslinguistic survey of different types of vowel deletion? Either synchronic or diachronic.

Oddly enough, a while ago someone recommended one to me in this same thread:

Here's a whole dissertation on hiatus resolution: Casali (1996) [pdf]. That might give you some concrete ideas.

It's very comprehensive, with lots of information on which vowel is deleted, how often it happens, vowel coalescence, etc.

[mae]

-

[dɮ the phoneme]

Here's a whole dissertation on hiatus resolution: Casali (1996) [pdf]. That might give you some concrete ideas.

It's very comprehensive, with lots of information on which vowel is deleted, how often it happens, vowel coalescence, etc.

Ah, thanks, that looks quite interesting! I should have specified though; in this context I'm looking more for interconsonantal vowel deletion, as I'm mainly trying to create more consonant clusters.

[bradrn]

Here's a whole dissertation on hiatus resolution: Casali (1996) [pdf]. That might give you some concrete ideas.

It's very comprehensive, with lots of information on which vowel is deleted, how often it happens, vowel coalescence, etc.

Ah, thanks, that looks quite interesting! I should have specified though; in this context I'm looking more for interconsonantal vowel deletion, as I'm mainly trying to create more consonant clusters.

Sorry - I saw 'vowel deletion' and immediately thought of that article...

[náʼoolkiłí]

Here's a whole dissertation on hiatus resolution: Casali (1996) [pdf]. That might give you some concrete ideas.

It's very comprehensive, with lots of information on which vowel is deleted, how often it happens, vowel coalescence, etc.

Ah, thanks, that looks quite interesting! I should have specified though; in this context I'm looking more for interconsonantal vowel deletion, as I'm mainly trying to create more consonant clusters.

Here's another dissertation: Gouskova (2003) [pdf] on syncope.

[cedh]

How reasonable is a change that deletes unstressed vowels / C1_C2, where C1 is more sonorant than C2?

I see no problems with that. ['kanad] > ['kand], why not?

It will also give things like [ra'ka] > [rka] without [ka'ra] > [kra], the latter of which seems like perhaps the more natural developement.

You could specify V > Ø / VC1_C2, i.e. the change would only happen if C1 is preceded by another vowel. Which looks like a completely natural (and even likely) condition for this type of change.

If you actually want [ka'ra] > [kra] too, but not [ra'ka] > [rka], the best way to get there IMO would be to have the following two changes one after the other:
V > Ø / VC1_C2
V > Ø / {V,#}C2_C1V
(Note that the second change happens only when C2 is not preceded by another consonant, and C1 is followed by a vowel. This prevents e.g. a word like ['kanakara] from becoming ['kankra] with a CCC cluster.)