Zompist Bboard Again

How do I originate a vowel system for my conlang?

Some design principles:

1. Phonemic distinction between long and short vowels are mandatory. There is a human reason for this.

2. {a, i, u, aː, iː, uː} is too simple of a vowel system.

3. The common vowel system lengthened {a, i, u, e, o, aː, iː, uː, eː, oː} is also too simple.

By 'originate' do you mean how to derive it diachronically or just how to design it?

Zju wrote: ↑Sun Nov 10, 2024 11:29 am By 'originate' do you mean how to derive it diachronically or just how to design it?

Good question. Thank you. I meant the latter.

As far as I recall, there was this general rule of thumb for a naturalistic vowel system:

1. Pick one of: /a i u~o/, /a e~ə i u/, /a e i o u/, /a ɛ i ɔ u/, /a ɛ e i ɔ o u/
2. Make up to two changes. A change is one of the following: dropping a vowel, adding a vowel, substituting a vowel with another.
3. Make sure that surface realisations / allophones still cover all of [a i u]
4. For length, anywhere between one and all vowels can have long counterparts. Long vowels tend to have more central realisation

Zju wrote: ↑Sun Nov 10, 2024 11:35 am 4. For length, anywhere between one and all vowels can have long counterparts. Long vowels tend to have more central realisation

I presume that was a mistype and you mean short vowels tend to have more central realization.

I meant to say that low long vowels tend to heighten: [ɛː ɔː] → [eː oː]

Zju wrote: ↑Sun Nov 10, 2024 11:41 am I meant to say that low long vowels tend to heighten: [ɛː ɔː] → [eː oː]

I.e. then tend to me tenser, which in many ways is the opposite of being more central. Also, long low vowels tend to be lower than short low vowels.

Zju wrote: ↑Sun Nov 10, 2024 11:35 am As far as I recall, there was this general rule of thumb for a naturalistic vowel system:

1. Pick one of: /a i u~o/, /a e~ə i u/, /a e i o u/, /a ɛ i ɔ u/, /a ɛ e i ɔ o u/
2. Make up to two changes. A change is one of the following: dropping a vowel, adding a vowel, substituting a vowel with another.
3. Make sure that surface realisations / allophones still cover all of [a i u]
4. For length, anywhere between one and all vowels can have long counterparts. Long vowels tend to have more central realisation

For number 2 I'd instead suggest "up to two systematic changes", although for "interesting" results you may need more than two. This could include processes such as umlaut, compensatory lengthening, split, and merger, Have a look at the evolution of the English vowel system from PIE for something particularly "interesting".

Also, it is extremely common to have one or two non-open central vowels (e.g. classical Modern English's schwa and schwi) on top of the basic systems mentioned above, and one should not have to expend one's "two changes" to get these.

Zju wrote: ↑Sun Nov 10, 2024 11:35 am As far as I recall, there was this general rule of thumb for a naturalistic vowel system:

1. Pick one of: /a i u~o/, /a e~ə i u/, /a e i o u/, /a ɛ i ɔ u/, /a ɛ e i ɔ o u/
2. Make up to two changes. A change is one of the following: dropping a vowel, adding a vowel, substituting a vowel with another.
3. Make sure that surface realisations / allophones still cover all of [a i u]
4. For length, anywhere between one and all vowels can have long counterparts. Long vowels tend to have more central realisation

This is an excellent continuation. Thank you for your lispy cleverness.

I can’t believe no-one has yet linked this guide from the old board. It’s a great overview of what kinds of vowel systems are found in natlangs.

bradrn wrote: ↑Sun Nov 10, 2024 8:25 pm I can’t believe no-one has yet linked this guide from the old board. It’s a great overview of what kinds of vowel systems are found in natlangs.

With how much people keep referencing Alice's work here despite it having succumbed to linkrot multiple times I feel like it needs to be better preserved, perhaps on zompist.com, with some commentary and redaction - I distinctly remember some criticism of the practice of giving "the short vowel system of X" as an example of a particular vowel system "shape" because vowels don't exist in a vacuum like that.

And once again my past comes back to haunt me. Did nobody notice that Lass's Phonology got there first?

That section of my copy of Lass is well-thumbed, but I thought the increased detail you provide is more relevant for conlanging.

HolyKnowing, what is holding you back from picking a vowel system? Are there other design criteria that would prevent you from just having a seven vowel tirangle or something?

Ketsuban wrote: ↑Mon Nov 11, 2024 4:56 pm That section of my copy of Lass is well-thumbed, but I thought the increased detail you provide is more relevant for conlanging.

That's nice to know

One thing to keep in mind is one can be creative with vowel system's allophony. For instance, my latest language has a vowel phoneme system of /i æ ɑ u/ and /iː æː ɑː uː/ but a vowel phone system of [i e ɛ æ ɐ ɑ ɔ o u] and [iː eː ɛː æː ɑː ɔː oː uː], as well as the realized diphthongs [æw ɑj ɛːj æːw ɑːj ɔːw]. Also, while vowel phonemes of any quantity can appear in any syllable, only one long vowel can be realized in a word, and then only in the stressed syllable, which is the leftmost heaviest of the last three syllables of a word. There are vowel reduction rules for vowels in unstressed syllables and vowel retraction rules for vowels before uvular consonant phonemes.

An easy and fun way to mess with the typical /a e i o u/ long + short system is to keep the short vowels conservative and let the long vowels go on an adventure.

Imagine /uː/ fronting to /yː/. /oː/ then raises to /uː/ and /aː/ moves back and rounds to /ɒː/. Finally, /eː/ lowers to /æː/, completing the chain shift.

To take it even further, you could have /iː/ lower to /eː/ as /yː/ unrounds, yielding a new /iː/. Or maybe you could let those two vowels merge. Or maybe /iː/ remains the same while /yː/ becomes /ɨː/. Or maybe various dialects do different things and the prestige variety of the language ends up taking from all of them, leaving you with six long vowels.

äreo wrote: ↑Wed Jan 08, 2025 11:59 pm An easy and fun way to mess with the typical /a e i o u/ long + short system is to keep the short vowels conservative and let the long vowels go on an adventure.

Imagine /uː/ fronting to /yː/. /oː/ then raises to /uː/ and /aː/ moves back and rounds to /ɒː/. Finally, /eː/ lowers to /æː/, completing the chain shift.

To take it even further, you could have /iː/ lower to /eː/ as /yː/ unrounds, yielding a new /iː/. Or maybe you could let those two vowels merge. Or maybe /iː/ remains the same while /yː/ becomes /ɨː/. Or maybe various dialects do different things and the prestige variety of the language ends up taking from all of them, leaving you with six long vowels.

These sorts of chainshifts really are very common. For instance, they happened in Oïl, Occitan, Greek, Swedish, and Norwegian to varying degrees.

Another vowel chain you could take inspiration from: Halkomelem and Northern Straits (Salish) have u > a > e, from an earlier *i u ə a system. However, both languages (AFAIK) have /u/ as a marginal phoneme from things like vocalization of /w/ and loanwords.

It makes sense that these kinds of shifts would be common with vowels, since languages seem to like to "fill out" the vowel space as much as possible.