Looking for sound changes

[Chengjiang]

Lately I've noticed that I know of quite a few sound changes that replace a particular sound but far fewer that create it. I want to keep the sound changes behind my languages thoroughly naturalistic and based on sound changes that are attested in reality. Therefore, I'm creating this thread so I (and others in my position) can post a type of sound change they're looking for and others can share sound changes (from natlangs) that fit that description, as a resource.

To start, I'm looking for changes that generate velar stops. I know of tons of different kinds of change that destroy them (e.g. whole spectra of palatalization and lenition alone) but not a lot of changes that create new velar stops. Anyone got some historical changes that derive [k], [g], or varieties thereof from something that isn't a velar stop? (Preferably not from uvular stops either.)

[Nortaneous]

http://kneequickie.com/archive/Most_wan ... nd_changes

r > g
l > g has probably happened somewhere
p_j > k_j > k
ŋ > x > k
is ŋ > g attested?
S > k is claimed there as a possibly-unconditional change
some Athabaskan languages apparently had k q > tS k

[----]

One thing that's happened frequently is that /k/ is lost or shifted to glottal or something, and then another sound replaces it. Arapaho has p > {k, tS}. and a whole ton of languages have t > k. Armenian also had w > k~g in some situations, apparently (cf. *dwoh > erku)

[Nortaneous]

I think I came up with a set of sound changes once that shifted all the stops forward by one... or was it back?... in certain environments. t > T > f > p, p k > k_w k_+, k_+ > tS > t pushed by k_w > k maybe.

Eskimo-Aleut had some fortition of voiced fricatives to voiceless stops, so any way to get G R [assuming the series work out right for this] can get you k q.

also could use epenthesis maybe. say velar nasal + fricative sequences turn to velar nasal + velar stop with the voicing of the following fricative + fricative, then widespread cluster deletion processes have the nasal drop, nasalize the vowels, nasal vowels get fucked with and then merge into oral vowels (how stable are nasal vowels?), and also the fricative gets deleted, so Ns Nz > k g or something

as for getting rid of k, there's some PNG lang that shifted both k and g to glottal stops, but maintained the fortis/lenis contrast

h > N_0 > k / a_a maybe? rhinoglottophilia could be useful

or you could just not worry about it. some khoisan languages don't allow velars anywhere but word-initially, though they usually have a very reduced inventory of word-medial consonants in general (n||ng /m n B r l/)

[Tropylium]

is ŋ > g attested?

Yes, in Hungarian (via *ŋk, as attested in Ob-Ugric).

S > k is claimed there as a possibly-unconditional change

*ʃ > *x is a frequent unconditional change, and you could probably feed that further into *x > k, so I'll buy that idea.

I think I came up with a set of sound changes once that shifted all the stops forward by one... or was it back?... in certain environments.

OT: I recall inventing one too. Mine operated on voiced stops and went backwards: b > β > w > gʷ > g; d > ð > v > b; dz > (d)z̻ > (d)ð > d; dʒ > dz; g > ɣ > j > ɟ > dʒ.

In my impression, glide fortition is probably the most common sound change that creates velars (uvular fronting might be more probable per se, but uvulars are themselves usually rare, and tend to originate from previous velars anyway). As an extreme example, an attested change from Selkup is *j, *w > *kʲ, *kʷ (in all prevocalic positions), possibly followed by later shuffling into /tɕ, tɕʷ, k, q, qʷ/ in different dialects depending on the following vowel… there is even a handful of cases where *j >> /qʷ/.

But most of the time dorsal stops don't "regenerate". If a lang has gotten low on them due to, say, a couple different waves of palatalization, a common way of recuperating is not sound changes that create some. It's sound changes that eliminate other consonants (t > d > ð > ∅, tʃ > ʃ > h > ∅, l > ɫ > w > ∅, etc.), and the resulting increasingly eroded word roots being gradually extended by compounding or derivation (also: is it just me or is -ka or something similar a common denominal suffix in an incredibly large number of languages?) — so that what velars remain, end up relatively more common again.

[Chengjiang]

But most of the time dorsal stops don't "regenerate". If a lang has gotten low on them due to, say, a couple different waves of palatalization, a common way of recuperating is not sound changes that create some. It's sound changes that eliminate other consonants (t > d > ð > ∅, tʃ > ʃ > h > ∅, l > ɫ > w > ∅, etc.), and the resulting increasingly eroded word roots being gradually extended by compounding or derivation (also: is it just me or is -ka or something similar a common denominal suffix in an incredibly large number of languages?) — so that what velars remain, end up relatively more common again.

That is true, and presumably is true for many segments that are made scarcer, rather than simply eliminated, by sound changes. I admit that this particular question I kicked off the thread with is largely about reduction of frequency of velar stops rather than their complete disappearance, so yes, there isn't as much "need" for sound changes to replace them as there could be. I'll also note that velar stops tend to be frequent in onomatopoeia and sound symbolism, two constant sources of fully new morphemes.

That said, I have a new request: Sound changes that increase the number of VOT/phonation distinctions made by a language, especially those that increase the number of such distinctions made in plosives beyond two. (E.g. a sound change that converts a /t d/ system to a /t d ɗ/ one.)

[CatDoom]

Stops (and, much more rarely, other types of sound) are believed to have fused with /h/ and /ʔ/ to produce aspirated and ejective consonants in some languages. Meanwhile, proto-Bantu is reconstructed with voiced and voiceless stops, but it's daughters developed other stop series from clusters of a nasal with a stop. Nasals followed by voiceless stops became aspirated stops in some languages, and I think a similar change produced ejectives in at least one. I think that implosives may have come from voiced prenasalized consonants in at least some Bantu languages.

[Terra]

is ŋ > g attested?

Proto-Japanese to Japanese had the following changes: (Note that PJ had no voiced stops or fricatives. They are innovations of Japanese.)
mp -> b
nt -> d
nk -> g
ns -> z
I recall reading somewhere that some dialects still have 'mb' for 'b', 'nd' for 'd', 'ng' for 'g', and 'nz' for 'z'. I've only ever rarely heard 'ng' and 'N' for 'g' though. I don't know if this really answers your question, because I don't know if N was part of this progression. Maybe it was 'nk -> ng -> g', or maybe it was 'nk -> ng -> N -> g'.

Note: Just pretend all the single quotes are square brackets. I'm too lazy to edit them now.

[Pole, the]

is ŋ > g attested?

Proto-Japanese to Japanese had the following changes: (Note that PJ had no voiced stops or fricatives. They are innovations of Japanese.)
mp -> b
nt -> d
nk -> g
ns -> z

Hungarian had almost the same, except: /ns/ /ɟ/.

[Nortaneous]

N for g is p common in Japanese.

Korean developed tense ( ~= unaspirated) stops and /s/ from cluster simplification.

I posted a list of papers on aspirated fricative origin a while back. presumably those changes could be used for stops. though I think most of them were shuffling already-existing distinctions

look up Tsakonian -- it and Spanish both developed aspirated plosives from sP clusters (so the opposite of what happened in Korean)

[CatDoom]

I posted a list of papers on aspirated fricative origin a while back. presumably those changes could be used for stops. though I think most of them were shuffling already-existing distinctions

I think I know the paper you're talking about. It explains several different origins, including PS clusters > S and plain S > Sʰ, aspirated stops and affricates that lenite to fricatives but keep their aspiration, a chain shift that goes something like TS > S > Sʰ, fusion with /h/ and/or /ʔ/, fricativization of aspirated voiceless sonorants, and geminate simplification (eg. SS > Sʰ, attested in Chumash as a result of consonant affixation).

[TaylorS]

Is [ɻ] > [l] plausible?

What about [ɰ] > [x]?

[Zju]

And how about changes in stop systems? Is any of these plausible:?

T Tʰ D > Tʰ Tʰ T
T D > T T'
T T' D > T Tʰ T'
T T' D > T Tʰ D

Are ejectives and aspirates able to change into one abothers? And ejectives and implosives?

Are there any hypotheses about how various stop systems in Armenian dialects came to be?

And could T D system somehow change into a three way contrast without merging with h or ʔ (or any other sound for that matter)?

[KathTheDragon]

I have a sneaking suspicion that that's about PIE.

[Tropylium]

Is [ɻ] > [l] plausible?

What about [ɰ] > [x]?

This is not the Sound Change Quickie Thread, but yes, sounds plausible.

Are ejectives and aspirates able to change into one abothers?

Probably not.

Are there any hypotheses about how various stop systems in Armenian dialects came to be?

I read a paper once that listed most of these and charted their historical development. The aspirated stops, interestingly, do not change in any of the Armenian dialects. The other two series have the quirk that there is a dialect group ("Middle Armenian", according to a handy table they've recently added at Wikipedia) that appears to reverse Classical Armenian /T/ and /D/ series, which makes a headache for reconstruction. They cannot be shuffled past one other just using steps in a T ↔ D ↔ Dʰ ↔ Tʰ ↔ T cycle because the common aspirates get in the way. The paper I had was from Kortlandt who suggested that a glottalized *D' should be reconstructed for Proto-Armenian; which becomes T in most dialects including Old Armenian, and could voice to D later, but went directly to D in these.

[Zju]

Then how do you get ejectʼives without using gloʔal stops and without imɓlosives?

[Zaarin]

Then how do you get ejectʼives without using gloʔal stops and without imɓlosives?

I'm not 100% certain on this, but I'm pretty certain that changes like

tt > tʼ

are attested. Also, since ejectives apparently became pharyngealized consonants in certain Semitic languages such as Arabic, it might be worth investigating if the inverse is possible.

[Zhen Lin]

is ŋ > g attested?

This is just denasalisation and should be about as common. For example, Minnan, and as already mentioned, Japanese (but notably in kan-on).

t k > 0 / V_V (Marathi) probably with voiced stops as intermediates, since they also became silent

Latin intervocalic /t/ (and /d/) chain-shifts to nothing in French: mutare > muer, natus > né, putere > puer, *tutare > tuer, vita > vie... but there are probably as many non-examples, due to vowels disappearing in the interim.

p > w / V_V (Marathi)

Also famously in Japanese, via an unvoiced (!) fricative stage.

[Zju]

I have a sneaking suspicion that that's about PIE.

Kind of. I'm fairly certain *t d dh were [t ɗ d]. Implosives then changed to ejectives in some branches and finally ɗ/tʼ changed to ʔd/ʔt in most branches.
Implosives also easily change to sonorants and that's why [ɓ] was so rare - it changed to [m] and [w]. That's also hinted by syllable structure:

See how m and w seem to pattern with stops.

[R.Rusanov]

http://kneequickie.com/archive/Most_wan ... nd_changes

It's incredibly funny how some people still believe y > u is impossible

[Chengjiang]

Awesome!

ejectives and implosives?

I don't know much about Afro-Asiatic outside of Semitic, so it's quite possible I'm wrong about this, but I thought there were some correspondences between implosives in some branches of Afro-Asiatic and ejectives in others. I thought Chadic implosives corresponded at least in part with ejective consonants in some other branches of the family, for instance. I'm having trouble finding a paper that examines this subject directly, though. Tell me, guys: Am I dimly remembering something genuine or is it a load of bollocks?

OK, new type of sound change: Any change that creates a lateral vs. rhotic distinction where no such distinction existed previously. Off the top of my head I can't think of much beyond [ɲ] > [ɻ] in Mandarin. And yes, I do realize that "rhotic" isn't a terribly well-defined category, but oh well.

[KathTheDragon]

Well, Latin had some d > l cf. *dn̩gʱu- > lingua

[Cedh]

If you use a narrow definition of "rhotic" as "a coronal tap, flap, trill, or approximant" (excluding at least "bunched R" and maybe also retroflex [ɻ]), then the American English shift {t d} → ɾ / V_V might be an example. In any case, it represents a sound change that creates a new "rhotic" consonant, and if it operates in a language that only has lateral liquids, it will result in the type of distinction you're looking for. There are probably several natlangs which did this (even if AmE doesn't count).

[Tropylium]

OK, new type of sound change: Any change that creates a lateral vs. rhotic distinction where no such distinction existed previously. Off the top of my head I can't think of much beyond [ɲ] > [ɻ] in Mandarin. And yes, I do realize that "rhotic" isn't a terribly well-defined category, but oh well.

IIRC there are Bantu languages where first *l > r / _{i u}; after this loanwords happened and the contrast became phonemic.

Many Indo-Aryan varieties had *ɖ, *ɖʱ > /ɭ/, /ɭʱ/ between vowels and I'm sure that's been in at least some place a similar motivation to start using laterals in loanwords.

Several branches of Iranian had *d > *ð > /l/ (sometimes also from *t). In Middle Persian and several other southwestern languages this only occurred in the Old Persian cluster /rð/. In some "East Iranian" languages (Pashto, Munji, Yidgha & Bactrian) the change was unconditional, leaving the languages with no /d/. (*ð > *l is also attested from Proto-Bantu, but it had no previous rhotic either; and from Tavastian Finnish, which already had both /l/ and /r/.)

Something like *s > *ɬ > /l/ ought to be possible as well…

[Pole, the]

Some Germanic had conditional /s/ → /z/ → /r/, this is, however, already having a rhotic.

Japanese /r/ is [ɺ] before back vowels and [ɾ] before front vowels, so it could effect in a split in some indefinite future.

Also, I think, assimilation and dissimilation can make wonders, e.g. l → ɭ → r / _[+retroflex] or l → r / _Vl.