Stability of vowel systems

[Abun]

During a discussion on a different thread I once more realized that my knowledge about phonetics is still quite limited, especially with respect to vowels. So I thought I'd start a new thread to ask, what constitutes a more or less stable vowel system.

I know that phonetic systems tend to symmetry and "want to" fill phonetic gaps, but what exactly constitutes a gap? For example, it seems to me that many languages seem to prefer back and front vowels to come in pairs, but many seem happy with a front low vowel /a/ and don't develop the urge to get /ɑ/ as well. Also, how big does a gap have to be for becoming unstable? For example, /a i u/ is apparently stable; I guess this is because it's "in balance" (there's a rather symmetric triangle around schwa)? And why is /a ɛ ɔ i u/ with a gap at close-mid level apparently unstable but /a e o i u/ with a gap at the open-mid level isn't?

EDIT: Just noticed that the /a e o i u (y ø)/ languages I know all seem to have [ɛ ɔ (œ)] as allophones of /e o (ø)/, so is that the reason they're stable? And /a ɛ ɔ i u y œ/ are unheard of because those would have allophonic [e o (ø)] and would therefore be essetially the same thing as /a e o i u/ with allophonic [ɛ ɔ (œ)]?

And what about rounding? Back vowels seem to be more likely to be rounded and front vowels (especially lower ones) more likely to be unrounded. But suppose my /a e o i u/ language develops /y/ (as a merger of a dipthong /ui/ maybe), does this mean the phonetic system becomes unstable and is likely to develop /ø/? And how come languages seem to round front vowels starting high to low (/y/ before /ø/ before /œ/)? It doesn't seem the same with unrounded back vowels (Mandarin for example has [ɤ] as its own unrounded back vowel, although it's only an allophone of schwa)...

And then there's nasalization. Do languages tend to develop nasal variants of every oral vowel? French seems to have started that way but today only has nasal /ɑ̃ ɔ̃ œ̃ ɛ̃/ (e.g. "sans, son, brun, brin") but oral /a ɑ e i ɛ œ ɔ ə e ø o i y u/. The Chinese Minnan/Hokkien language has oral /a e i o ɔ u/ (there are dialects with /ɯ ə ɛ/) but only nasal /ã ẽ ĩ ɔ̃/ (some dialects even reduce this to /ã ĩ ɔ̃/). Again, it seems to me that the assymmetry is due to merges, not because some vowels resisted nasalization to begin with. So does this mean nasalization is likely to affect the whole vowel system at once but nasal vowels are somewhat more susceptible to merging (and that the resulting asymmetry doesn't cause the system to become unstable)? If so, are there vowels which are more likely to prevail in the merging? (The only common one between French and Minnan is /ɔ̃/ but I hesitate to draw basic conclusions from a comparison of just two cases^^')

[Ryan of Tinellb]

The bottom of the mouth doesn't have as much room as the top, so there is much more space between /i/ and /u/ then between /a/ and /ɑ/. Rounding of vowels has similar effects on the formants (~frequencies) as backing, so they tend to go together. Nasalisation tends to open vowels, and seems to make it more difficult to distinguish them.
As for open-mid vs. close-mid, I'm not sure. Again, there's more space around the close vowels. On the other hand, different dialects have different phonetic vowels without the speakers necessarily realising it. A BrEng [e] is the same as as AmEng [e] (mostly), even if one is closer to /e/ and the other to /ɛ/. So I'm wondering if close-mid is more stable than open-mid, or if it's just that <e> tends to be on more keyboards than <ɛ>.

Just an interesting side-note: I'm a native speaker of Cultivated Australian English. I've found that Wikipedia's article on AusEng disagrees with all my dictionaries as to where exactly my vowels are. My LOT vowel, for instance. In dictionaries, it's /ɒ/, and according to Wikipedia, it's /ɔ/. So I end up with no idea how to pronounce my own conlangs. Once I get a good enough microphone, I'll be checking my formant chart.

[Nortaneous]

Languages never have more nasal than oral vowels, and tend to have fewer.

An /a e o i u/ language that develops /y/... probably wouldn't happen in the first place, unless surrounding languages have /y/. (Souletin Basque has /a e o i y u/ but it's spoken in France.) It'd be more likely to lose /y/ than to develop /2/, I think.

There's no real difference between /a E O i u/ and /a e o i u/. Slavic languages generally have /a E O i u/, plus /1/ in a lot of them.

/y/ before /2/ before /9/ probably isn't true.

[vokzhen]

As Ryan suspects, I'm fairly certain it's not that /a e o i u/ is more stable than /a ɛ ɔ i u/, it's that the former is easier to type. The IPA includes rules for using former to describe a set of [a ɛ ɔ i u] due to ease of typing, and simply clarifying in text that the actual values are [ɛ ɔ]. Similar to how a whole bunch of different sounds are subsumed by the cover symbol /r/, and how almost everyone uses /a/ rather than the more accurate /ä/ to represent a central vowel.

Another thing to keep in mind that the less filled the space is, the more room for allophony. For example /a i u/ vowel sets with /q/ among the consonants generally the vowels the allophones [ɑ e o] or [ɑ ɛ ɔ] (Arabic, Quechua, Inuit).

Also, out of curiosity, does anyone actually have a link to something showing evidence of nasalization leading to open vowels? I've heard it before, but none of the evidence I've seen actually supports it. It just supports some kind of vowel changes, but that can be all kinds of different redistributions - opening, closing, or other changes. French and Polish are two big, European languages that have only open nasals, and I have a feeling that has given undue weight.

An /a e o i u/ language that develops /y/... probably wouldn't happen in the first place, unless surrounding languages have /y/.

I dunno about that. Afaik that's mostly what's posited for happening in various Sino-Tibetan languages (Lhasa Tibetan, Qiangic, rGyalrongic, Standard Mandarin), often from coronal-triggered fronting, and one or two Inuit dialects have [y].

[sangi39]

Also, out of curiosity, does anyone actually have a link to something showing evidence of nasalization leading to open vowels? I've heard it before, but none of the evidence I've seen actually supports it. It just supports some kind of vowel changes, but that can be all kinds of different redistributions - opening, closing, or other changes. French and Polish are two big, European languages that have only open nasals, and I have a feeling that has given undue weight.

WALS has something on this topic. From what they've briefly said, nasal vowels, as Nort has said, are typically fewer in number than oral vowels within a given language (they provide one counter-example, Koyra Chiini, as the result of borrowing from French, but that seems like it might need some more research by those interested on what's going on there) and they go on to say:

There is evidence to suggest that a reduction in the number of nasal vowels has a phonetic basis: the acoustic effect of nasalization is known to affect the perceptual distinctiveness of vowels (e.g. Beddor 1993). However, it is difficult to predict the shape of reduced vowel inventories. Lakhota (1) has only maximally peripheral nasal vowels. In other languages, such as Yoruba (Niger-Congo; Nigeria); the distinction between close and open mid vowels is reduced in favour of nasal open mid vowels; but in Seneca (Iroquoian; New York State) the nasal vowel system is limited to mid vowels /ɛ̃ ɔ̃/ only

The basic tendency seems to be that phonemic nasal vowels undergo a series of mergers, but the direction in which those mergers occur (either towards high or low vowels) seems to be somewhat unpredictable.

[Abun]

Thanks for your answers, you made things a lot clearer to me

An /a e o i u/ language that develops /y/... probably wouldn't happen in the first place, unless surrounding languages have /y/. (Souletin Basque has /a e o i y u/ but it's spoken in France.) It'd be more likely to lose /y/ than to develop /2/, I think.

Really? Then how did /y/ develop in Europe in the first place? Latin didn't have it and neither did proto-Germanic as far as I can see, yet French and quite a few Germanic languages developed it. Or is this Finno-Ugric influence which has spread via Germanic languages to France and part of the Basque Region? As another case study, neither Middle Chinese nor Classical Tibetan had rounded front vowels, yet both Lhasa Tibetan and Mandarin (as far as I hear Wu as well but I don't know enough about Wu languages) developed them, and from quite different sources. Afaik Mandarin monophthongized /iu~iw/, whereas in Tibetan certain final consonants (namely /n d s l/, the middle two of which have since been reduced to /ʔ/) caused /a o u/ to umlaut to /ɛ ø y/. Is this to be attributed to contacts with Mongolian then (not sure when Mongolian lost its rounded front vowels)? I'm not sure how likely it would be for one to cause the rounding on the other seeing as I don't know whether this feature of Lhasa Tibetan extends to the North-Eastern Tibetan languages which would have been in regular contact with early Mandarin.

Also I just thought that nasal colouring is not the only colouring possibility. Are there example for, say lateral colouring (vowels pronounced while the tongue was in [l] position)?

And then there's diphthongs... From what I encountered so far, /ai/ and /au/ are quite common, /ɛi~ei/, /ou/, /ɔi~oi/, /iu/ and /ui/ maybe a bit less, some languages also have /ae/ (not sure if Latin <ae> was originally a diphthong but I guess so since the Romans didn't invent a new letter for <ɛ>). Strikingly, all of these are going in a direction from low to high, except for the high-high combinations /iu/ and /ui/. Is this a common principle then? Do high->low vowel combinations tend to have the first vowel become semivowels instead then (not that the second vowel in the low->high combinations never become semi-vowels)?

[Zaarin]

Germanic languages got front rounded vowels from ablaut.

[Abun]

Germanic languages got front rounded vowels from ablaut.

Really? At least in Contemporary German I can't think of a single ablaut form which features front rounded vowels. Most of those are part of umlaut paradigma (say, nom.sgl. "Vogel" --> nom.pl. "Vögel") which as far as I know is historically an assimilation to an /i/ in the suffix: /u/ and /o/ get fronted, /a/ raises to /ɛ/. There are a few words which never alternate with non-umlauted vowels, "Tür" for example. But I guess it's not at all impossible that this was once something like [tu:ri]-->[ty:ri]-->[ty:r].

But be the case as it may be, as far as I understood Norteanous, he was suggesting that languages don't develop /y/ (or front rounded vowels at all?) except if they are in contact with languages which already possess it (right?). So I wondered where he was suggesting German and French got the idea from, and my only idea would be a spread from Proto-Finnish-->Old Norse-->Middle High German-->French.

[vokzhen]

An /a e o i u/ language that develops /y/... probably wouldn't happen in the first place, unless surrounding languages have /y/. (Souletin Basque has /a e o i y u/ but it's spoken in France.) It'd be more likely to lose /y/ than to develop /2/, I think.

Really? Then how did /y/ develop in Europe in the first place?

European languages a) had more complex vowel systems than /a e i o u/ and b) often didn't just gain /y/ but got a whole set of front-rounded vowels. Greek and French as fronting of /u/ (later filled in with /o/, in turn filled in by /ɔ/), French further through diphthong coalescence (eu > ø), Germanic through i-mutation (uCi > yCi, etc) and to a lesser extent u-mutation or labial spreading (iCu > yCu, bi > by). Finnish has /y ø/ as far back as we can trace. However, I agree with you that as far as I know the front-rounded vowels in Sino-Tibetan languages can't at least all be attributed to contact, unlike French which afaik is generally attributed to Germanic influence (though, interestingly, the front-rounded vowels didn't appear until several centuries after Frankish died out in favor of French), and the genesis of front-rounded vowels is afaik posited from a simpler system, closer to /a e i o u/ (unlike e.g. Proto-Germanic's /a e i o u a: ɛː e: i: ɔː o: u: ɛ:: ɔ:: au ai eu iu ɔːu ɔːi/ plus some nasals, or pre-Classic Greek's system something like /a e i o u a: ɛː e: i: ɔ: o: u: ai au eu oi aːi ɛːi ɔ:i/).

As for diphthongs, pretty sure /eu oi/ are of similar commonness as /ei ou/. Diphthongs of intermediate height /ae ao ea oa/ are pretty rare, I think, and unstable as well. /iə uə/ or /ia ua/, where the first element is stressed (roughly British tear, tour), aren't common in Europe but are in the Himalaya - East Asia - Southeast Asia region (along with /ɯə ɯa/, since non-front unrounded vowels are pretty common there). A number of languages have /ie uo (yø)/, with the first element being stressed - Romance, Finnish, a few Turkic languages iirc, Yukagir - though I believe they tend to be fairly unstable (Spanish and Italian shifting to stress on the second element, Turkic and Yukagir are just allophonic).

Germanic languages got front rounded vowels from ablaut.

Really? At least in Contemporary German I can't think of a single ablaut form which features front rounded vowels.

Ablaut has multiple meanings, one of them being regressive vowel harmony, i.e. i-umlaut.

[Abun]

Ablaut has multiple meanings, one of them being regressive vowel harmony, i.e. i-umlaut.

Oh ok, my bad then. I had only heard the term as referring to the "strong" verbal paradigms like "sink, sank, sunk" before

[... T]he genesis of front-rounded vowels [in French] is afaik posited from a simpler system, closer to /a e i o u/ (unlike e.g. Proto-Germanic's /a e i o u a: ɛː e: i: ɔː o: u: ɛ:: ɔ:: au ai eu iu ɔːu ɔːi/ plus some nasals, or pre-Classic Greek's system something like /a e i o u a: ɛː e: i: ɔ: o: u: ai au eu oi aːi ɛːi ɔ:i/).

Ah. So languages with a rather full vowel charts are likely to develop rounding contrast in order to get another dimension for vowels to develop into?

/iə uə/ or /ia ua/, where the first element is stressed (roughly British tear, tour), aren't common in Europe but are in the Himalaya - East Asia - Southeast Asia region (along with /ɯə ɯa/, since non-front unrounded vowels are pretty common there).
With my limited experience I would at first have expected those to be allophones of [ɯ], but maybe that's because they remind me of the diphthongization of [ɤ] to [ɤə] which is common in northern dialects of Mandarin

[vokzhen]

[... T]he genesis of front-rounded vowels [in French] is afaik posited from a simpler system, closer to /a e i o u/ (unlike e.g. Proto-Germanic's /a e i o u a: ɛː e: i: ɔː o: u: ɛ:: ɔ:: au ai eu iu ɔːu ɔːi/ plus some nasals, or pre-Classic Greek's system something like /a e i o u a: ɛː e: i: ɔ: o: u: ai au eu oi aːi ɛːi ɔ:i/).

Ah. So languages with a rather full vowel charts are likely to develop rounding contrast in order to get another dimension for vowels to develop into?

Not necessarily, it's just that that's what happened in Europe. (Also, real quick check names again so you don't get things confused. Nort and I were both talking about this). Nort's claim, at least how I took it, was specifically that a language was unlikely to develop /y/, without /ø/, from a simple vowel inventory of /a e i o u/, except via outside influence, using Basque of an example of that. In your followup to that claim, you seem to have mistaken that for a claim that /y/ is unlikely to arise except via outside influence, missing the other parts. Europe is a good basis for making Nort's claim, because they all either had a (much) more complex vowel system, or developed /y ø/ both (and off the top of my head all of them had the former). French and Greek both developed /y/ (and only /y/ for Greek) by changing a system of /u o ɔ/ to /y u ɔ/, necessitating a larger vowel inventory than /a e i o u/. The others ended up with /ø/ as well, not just /y/. The problem is I'm not sure Sino-Tibetan backs this up as a general tendency, as they developed front-rounded vowels with at least what's often reconstructed as a simpler vowel system, and at least in some cases with only /y/. And, as I said, some Inuit dialects from what I read have [y] as an allophone of /i/ after labials, or /u/ before coronals (I can't seem to find a source on the latter where I thought it was, though), despite a simple vowel system of /a i u/+length.

[gach]

Finnish has /y ø/ as far back as we can trace.

Since you were involved in gathering vowel systems and histories, let's correct this one. Finnish /y/ goes indeed back to Proto Uralic */y/ but /ø/ is a latecomer that has appeared independently in various Uralic branches and in Finnic only at the Proto Finnic stage. Even in Proto Finnic it most probably didn't appear past the first syllable. Finnish /øy/ and /yø/ (< Proto Finnic */ø:/) have ancient etymologies like

Fi löyly < PFU *lewli
Fi syödä < PFU *sewi-
Fi vyö < PFU *üwä
Fi yö < PFU *üji

but cases of short /ø/ are limited to words with limited distribution and can often be related to affective vocabulary (coined words with emotionally marked connotations) or nativising loans. All cases of long /ø:/ in Finnish have to be new anyway since any older ones would have turned into /yø/.

You might say that /ø/ appeared in Finnic because there already was /y/ to support it. The vowel system reconstructible to the Proto Finno-Saamic stage is /a o u æ e i y/ which can't be even as common as the /a e i o u y/ system. Because of this and the unknown origins of the Uralic /y/ you probably won't find this development path particularly enlightening for your discussion. I agree, though, that if you have a process leading to the creation /y/, there are good chances that a nearly symmetric process will also lead to the creation of /ø/.

[Haplogy]

Finnish has /y ø/ as far back as we can trace. However, I agree with you that as far as I know the front-rounded vowels in Sino-Tibetan languages can't at least all be attributed to contact, unlike French which afaik is generally attributed to Germanic influence (though, interestingly, the front-rounded vowels didn't appear until several centuries after Frankish died out in favor of French), and the genesis of front-rounded vowels is afaik posited from a simpler system, closer to /a e i o u/ (unlike e.g. Proto-Germanic's /a e i o u a: ɛː e: i: ɔː o: u: ɛ:: ɔ:: au ai eu iu ɔːu ɔːi/ plus some nasals, or pre-Classic Greek's system something like /a e i o u a: ɛː e: i: ɔ: o: u: ai au eu oi aːi ɛːi ɔ:i/).

French gained /y/ due to Gaulish influence, I'm pretty sure.

[Zaarin]

Finnish has /y ø/ as far back as we can trace. However, I agree with you that as far as I know the front-rounded vowels in Sino-Tibetan languages can't at least all be attributed to contact, unlike French which afaik is generally attributed to Germanic influence (though, interestingly, the front-rounded vowels didn't appear until several centuries after Frankish died out in favor of French), and the genesis of front-rounded vowels is afaik posited from a simpler system, closer to /a e i o u/ (unlike e.g. Proto-Germanic's /a e i o u a: ɛː e: i: ɔː o: u: ɛ:: ɔ:: au ai eu iu ɔːu ɔːi/ plus some nasals, or pre-Classic Greek's system something like /a e i o u a: ɛː e: i: ɔ: o: u: ai au eu oi aːi ɛːi ɔ:i/).

French gained /y/ due to Gaulish influence, I'm pretty sure.

What makes you think that? I've never seen anything to suggest that Gaulish (or any Celtic language) had front rounded vowels.

[Haplogy]

Finnish has /y ø/ as far back as we can trace. However, I agree with you that as far as I know the front-rounded vowels in Sino-Tibetan languages can't at least all be attributed to contact, unlike French which afaik is generally attributed to Germanic influence (though, interestingly, the front-rounded vowels didn't appear until several centuries after Frankish died out in favor of French), and the genesis of front-rounded vowels is afaik posited from a simpler system, closer to /a e i o u/ (unlike e.g. Proto-Germanic's /a e i o u a: ɛː e: i: ɔː o: u: ɛ:: ɔ:: au ai eu iu ɔːu ɔːi/ plus some nasals, or pre-Classic Greek's system something like /a e i o u a: ɛː e: i: ɔ: o: u: ai au eu oi aːi ɛːi ɔ:i/).

French gained /y/ due to Gaulish influence, I'm pretty sure.

What makes you think that? I've never seen anything to suggest that Gaulish (or any Celtic language) had front rounded vowels.

My IE teacher, for one, and from a quick Google search Wikipedia at least seems to support it.

[WeepingElf]

Ablaut has multiple meanings, one of them being regressive vowel harmony, i.e. i-umlaut.

Oh ok, my bad then. I had only heard the term as referring to the "strong" verbal paradigms like "sink, sank, sunk" before

In IE studies, the term ablaut is commonly used only for a specific vowel alternation in PIE and its reflexes in the daughter languages, of which English sink, sank, sunk is an example. This ablaut doesn't create front rounded vowels in PIE (it had none), nor in any daughter languages. The front rounded vowels in German are the products of i-umlaut, which is a Germanic process that has nothing to do with PIE ablaut.

[Zaarin]

Finnish has /y ø/ as far back as we can trace. However, I agree with you that as far as I know the front-rounded vowels in Sino-Tibetan languages can't at least all be attributed to contact, unlike French which afaik is generally attributed to Germanic influence (though, interestingly, the front-rounded vowels didn't appear until several centuries after Frankish died out in favor of French), and the genesis of front-rounded vowels is afaik posited from a simpler system, closer to /a e i o u/ (unlike e.g. Proto-Germanic's /a e i o u a: ɛː e: i: ɔː o: u: ɛ:: ɔ:: au ai eu iu ɔːu ɔːi/ plus some nasals, or pre-Classic Greek's system something like /a e i o u a: ɛː e: i: ɔ: o: u: ai au eu oi aːi ɛːi ɔ:i/).

French gained /y/ due to Gaulish influence, I'm pretty sure.

What makes you think that? I've never seen anything to suggest that Gaulish (or any Celtic language) had front rounded vowels.

My IE teacher, for one, and from a quick Google search Wikipedia at least seems to support it.

How does that make sense? The attestation for Gaulish is extremely sketchy past the sixth century; its likely influence on French/Late Romance must be small and mostly lexical. Not to mention that the fronting of /u/ /y/ seems to have occurred quite late, long after the last remnants of Gaulish would have been certainly gone.

[Dewrad]

Eeeh, it's possible, but not incontrovertible. The evidence that Gaulish /u/ was either realised [y] (or tended toward a more front realisation than Latin's /u/) is in the main derived from:

- the fact that in the Brythonic languages we see universal fronting of the corresponding phonemes. If you accept that Gaulish and Proto-Brythonic were essentially mutually intelligible varieties (which is what all the evidence points towards), then it's not implausible the same realisation in Gaulish.
- fronting of Latin /u/ only occurs in regions with a Gaulish substratum. This argument has the deficit of being circular, and frankly can be accounted for in other ways.
- Greek transcription of Galatian /u/ with υ (e.g. Δρυνεμετον for /drunemeton/). This falls down given that the Greek transcription of /u/ in southern Gaul was ου.

So it's possible, even plausible, but far from proven.

[vokzhen]

According to Wikipedia, fronting of /u/ in French didn't happen until the 12th century, more than half a millennia after Gaulish attestation.

[Dewrad]

It intrigues me how they arrived at that date: it seems implausibly late, just on the evidence of early post-conquest English scribal habits.

[Zaarin]

Eeeh, it's possible, but not incontrovertible. The evidence that Gaulish /u/ was either realised [y] (or tended toward a more front realisation than Latin's /u/) is in the main derived from:

- the fact that in the Brythonic languages we see universal fronting of the corresponding phonemes. If you accept that Gaulish and Proto-Brythonic were essentially mutually intelligible varieties (which is what all the evidence points towards), then it's not implausible the same realisation in Gaulish.
- fronting of Latin /u/ only occurs in regions with a Gaulish substratum. This argument has the deficit of being circular, and frankly can be accounted for in other ways.
- Greek transcription of Galatian /u/ with υ (e.g. Δρυνεμετον for /drunemeton/). This falls down given that the Greek transcription of /u/ in southern Gaul was ου.

So it's possible, even plausible, but far from proven.

Interesting information; thank you.

[gestaltist]

I think we should rename the topic to „origins of «y» in French“ at this point.

My question: why did Germanic languages develop such a complicated vowel system in the first place?

[gach]

why did Germanic languages develop such a complicated vowel system in the first place?

Reinterpretation of vowel length?

[gestaltist]

why did Germanic languages develop such a complicated vowel system in the first place?

Reinterpretation of vowel length?

I am more interested in the economy of that change. People are lazy creatures and they don’t tend to complicate their lives if they don’t have to. Given the prevalence of simple vowel systems like /a e i o u/ what was the advantage of complicating the system? PIE had a much smaller vowel inventory, right?

[jal]

I know that phonetic systems tend to symmetry and "want to" fill phonetic gaps, but what exactly constitutes a gap?

There's two processes going on:
1) Pushing vowels from a crowded area into a less crowded area
2) Lots of allophony into less crowded areas

The first happens with "unstable" vowel inventories: vowel inventories where the POA of two or more vowels, and hence the accoustic differences, are close. In order to maximize the differences between those vowels, speakers will (completely subconciously, mind you) change the POA away from that of the other vowel(s), preferably into an area that's less crowded. If there is no such area, a chain shift can happen, that may lead to the creation of diphthongues (think English i [ai] once [i:]), when there's no place to flee to anymore, and/or mergers of vowels. Note that I'm not sure whether the direction of movement, in case there's more than one place to go, is predictable or not.

The second follows from the fact that, as long as there's still enough distinction (i.e. accoustic difference), it is less important to always use the same POA, as long as there's no significant overlap (though overlap may occur, as with the English tendancy to schwa-ize all unstressed short vowels, when there's no interpretational problem). In those cases, speakers are happy to accomodate the pronounciation of the vowels to that of the surrounding sounds, so that, say, /ki/ gets to be pronounced [kɨ], which is fine if there's nothing that sounds like [ɨ] otherwise. The allophones in turn can give way to new phonemes due to other sound changes (say you have [ti] and [kɨ] from /ti and /ki/, then /k/ changes to [t] before front vowels, you might well end up with [ti] and [tɨ]).

For language change, this means that 1 can lead to interesting chains shifts and vowel mergers, and 2 can lead to more complex vowel inventories.


JAL