consonant mutations

[Vardelm]

I'm working on consonant mutations for my Dwarvish conlang. These will be inspired by both Celtic & Uralic mutations, but I'm starting with Celtic-ish mutations.

The question: Celtic languages have /m/ → /v/, which to me could be considered the "opening" type of mutation. However, I notice there's no examples of a corresponding /n/ → /z/, and I see no example of [n] → [z] in the Index Diachonica. Is having [n] → [z] unreasonable, along with a similar [ŋ] → [ɣ]?

My language also has /s/ with [z] being allophonic, so the lention sound change would actually be /n/ → /s/, or (I think) /n/ → /s~z/ to represent the allophones (not sure on the notation). However, this is a somewhat separate issue from simply having synchronic [n] → [z] and/or [ŋ] → [ɣ] changes.

[Hallow XIII]

Out of those three I would definitely call /n/ -> /z/ the most suspect; the history of m-lenition in Celtic involves an intermediate nasalized labial fricative. That could plausibly also happen with /ŋ/, but with /n/ your problem is the sibilance. That said, there's nothing really stopping you from doing something like [n] -> [ɹ̃] -> [z].

[alice]

I had /n/ mutating to /z/ in one of my conlangs and nobody complained, so if it's good enough for me I don't see why it shouldn't be good enough for you

[Vardelm]

That said, there's nothing really stopping you from doing something like [n] -> [ɹ̃] -> [z].

Could that be a diachronic justification for [n] → [z], where it was originally [n] → [ɹ̃] and then [ɹ̃] universally changed to [z]?

Alternatively, since I have /ʀ/, perhaps I should have /n/ debuccalize to /ʀ/? /ŋ/ could either have the same debuccalization, or it could follow the example of Scottish Gaelic /ɲ/ → /n/ and have /ŋ/ → /n/.

I had /n/ mutating to /z/ in one of my conlangs and nobody complained, so if it's good enough for me I don't see why it shouldn't be good enough for you

I'll see what I can do within the "legal" limits of naturalistic conlanging, but if I hate the result, I'll maybe go this route and just Jedi hand-wave it.

[Nortaneous]

n > l > ɮ > z

[Richard W]

I've long felt that /m/ can be an honorary voiced oral stop. As insular Celtic voiced stops fricativised intervocally, /m/ went along with them.

[Vardelm]

n > l > ɮ > z

Definitely a path. Could that be restricted to places where mutation would be found, as in word-initially and only when following another word that ends in a stop?

I've long felt that /m/ can be an honorary voiced oral stop. As insular Celtic voiced stops fricativised intervocally, /m/ went along with them.

That's how it feels to me after looking at Celtic mutations the past couple days. I was hoping /n/ would be jealous of all the places /m/ got to go and tag along for the ride.

[Richard W]

n > l > ɮ > z

Definitely a path. Could that be restricted to places where mutation would be found, as in word-initially and only when following another word that ends in a stop?

The change n > l can be syllable initial only, thought the examples I can think of are in languages without final liquids.

The behaviour of Korean ㄹ r offers an approximation to the opposite of what you actually asked for. Intervocally, in compound words it's [ɾ] syllable-initially, but otherwise it's generally [n] syllable-initially. Unfortunately for analogy, it appears to derive from a rhotic diachronically.

[Vardelm]

Interestingly, I just found ŋ → x on the Index Diachronica.

https://chridd.nfshost.com/diachronica/search?q=x

#27 Maya to Ixilean, Kaqchikel-Tz’utujil, Core K’iche’, and Mamean. I would suppose that if this can happen without condition, then it could happen intervocalically.

[Estav]

As mentioned, the Celtic /m/ > /v/ mutation seems to be ultimately based on lenition of intervocalic /m/ to something like [β̃~w̃]. It's the same as the mutation of /b/. The way I would expect [n] and [ŋ] to go if similarly lenited would be to [ɾ̃] and [ɰ̃] respectively.

For further developments, you could possibly have [ɾ̃] > [ɾ] > [ɹ] > [z] as mentioned, or [ɾ̃] > [ð]. For [ɰ̃], it could develop to [ɣ] or [ʁ] but I'd expect it to stay voiced intervocalically if /s/ has an intervocalic allophone [z].

Based on the inventory given in your consonants table, given the mutations /m mʲ/ > /v vʲ/, and the stipulation that /n nʲ ŋ ŋʲ/ mutate, I would be least surprised to see /n nʲ ŋ ŋʲ/ > /ð ðʲ x xʲ/ (merging I'd assume with the mutations of /d̪ dʲ g gʲ/) although other outcomes are possible.

[Vardelm]

Based on the inventory given in your consonants table, given the mutations /m mʲ/ > /v vʲ/, and the stipulation that /n nʲ ŋ ŋʲ/ mutate, I would be least surprised to see /n nʲ ŋ ŋʲ/ > /ð ðʲ x xʲ/ (merging I'd assume with the mutations of /d̪ dʲ g gʲ/) although other outcomes are possible.

Ooohhh... I LIKE /n nʲ ŋ ŋʲ/ > /ð ðʲ x xʲ/!!! I do indeed have /d̪ dʲ/ > /ð ðʲ/, so that is nicely parallel to /b bʲ/ > /v vʲ/. I don't know if it's realistic or not (phonology is just not my jam), but it's appealing.

I should mention that /ð/ should be changed to /θ/. My intention was the fricatives from alveolar on back should be voiceless, with optional voiced allophone between vowels. That's an incidental detail, and I don't think affects the suggestions here.

[Znex]

The question: Celtic languages have /m/ → /v/, which to me could be considered the "opening" type of mutation. However, I notice there's no examples of a corresponding /n/ → /z/, and I see no example of [n] → [z] in the Index Diachonica. Is having [n] → [z] unreasonable, along with a similar [ŋ] → [ɣ]?

My language also has /s/ with [z] being allophonic, so the lention sound change would actually be /n/ → /s/, or (I think) /n/ → /s~z/ to represent the allophones (not sure on the notation). However, this is a somewhat separate issue from simply having synchronic [n] → [z] and/or [ŋ] → [ɣ] changes.

There's a few obscure Romance languages that have interesting lenitions of /n/, including Istro-Romanian that has intervocalic *n > r. I dunno about many with n > z, but it doesn't sound implausible.

Generally though, nasals tend to lenite as approximants or similar sonorants. Celtic *m at first shifted to /w᷉/ and then /v/. I haven't seen many examples of lenited ŋ, but I'd imagine something similar.

[Vardelm]

There's a few obscure Romance languages that have interesting lenitions of /n/, including Istro-Romanian that has intervocalic *n > r. I dunno about many with n > z, but it doesn't sound implausible.

Generally though, nasals tend to lenite as approximants or similar sonorants. Celtic *m at first shifted to /w᷉/ and then /v/. I haven't seen many examples of lenited ŋ, but I'd imagine something similar.

Right now, I'm leaning towards /n nʲ/ > /ð ðʲ/ per Estav's suggestion. I think this could originally be something like /n nʲ/ > /r~ɹ~ɾ rʲ~ɾʲ/, then shifting to /ð ðʲ/. In the Index Diachronica, I see Tai-Kadai has {r,r̥} > ð, and Siouan-Iroquoian has r → ð / x_. Additionally, Siouan-Iroquoian has n → ð / _{ã,ẽ,õ}, which seems somewhat similar to the /m/ > /w᷉/ change in Celtic you mention. Proto-Norse to Old Norse has nː → ð / _{r,ʀ}. None of these are exactly what I'm looking for here, but they all are in the ballpark such that /n nʲ/ > /ð ðʲ/ doesn't seem completely unrealistic.

Right now, I also have a lenition of /x/ > /ʀ/. I have it in mind that this was originally /x/ > /ɰ/, with /ɰ/ later changing to /ʀ~ʁ/. If there was a shift of /r~ɹ~ɾ/ > /ð/, it seems logical-ish that /ɰ/ might have become more rhotic.


EDIT:
BTW, I've been looking through The Historical Origin of Consonant Mutation in the Atlantic Languages by Merrill. LOTS of good info on consonant mutation in various languages.

[anteallach]

I've long felt that /m/ can be an honorary voiced oral stop. As insular Celtic voiced stops fricativised intervocally, /m/ went along with them.

What are some examples outside of insular Celtic which can be seen as /m/ patterning with obstruents rather than the other nasals?

[alice]

Perhaps worth mentioning is /n/ > /ð/ before /r/ in Sindarin, as in Caradhras. If it was good enough for JRRT...

[Znex]

I've long felt that /m/ can be an honorary voiced oral stop. As insular Celtic voiced stops fricativised intervocally, /m/ went along with them.

What are some examples outside of insular Celtic which can be seen as /m/ patterning with obstruents rather than the other nasals?

It occurs to me that however likely nasals or other sonorants will pattern with obstruents in lenition or fortition co-occurs quite often with fortis-lenis sets of consonants. Both Insular Celtic and Latin had widespread geminates in their lexicon (curiously, the particular sonorants that later underwent lenition each had their own geminate counterpart).

To elaborate:

More: show

Insular Celtic had
*t k kw
*b d g gw
*tt kk kkw
*m n s r l w j
*mm nn ss rr ll ww

Latin had
/p t k kʷ/
/b d g gʷ/
/pp tt kk kkʷ/
/pj tj kj/
/bj dj gj/
/m n s r l w j/
/nn ss rr ll jj/
/mj nj sj rj lj/

The sound changes we observe:
IC to Old Irish

Code: Select all

initially				medially
kw > k					t k kʷ > θ x x
gw > g					b d g gʷ > β ð ɣ ɣ
					tt kk kkʷ > t k k
					bb dd gg ggʷ > b d g g
m n s r l w j > m N s R L f ∅		m n s r l w j > w̃ n h~∅ r l w~∅ j~∅
					mm nn ss rr ll ww > m N s R L f

Latin to Western Romance (ignoring palatalisation changes)

Code: Select all

initially				medially
					p t k kʷ > b d g g(w)
					b d g gʷ > β ð ɣ (ɣ)w
					pp tt kk kkʷ > p t k k(w)
pj tj kj > pʲ tsʲ tsʲ		pj tj kj ( > ppʲ ttʲ kkʲ) > pʲ tsʲ tsʲ
bj dj gj > pʲ ʝ~dʒʲ ʝ~dʒʲ		bj dj gj ( > bbʲ ddʲ ggʲ) > bʲ ʝʝ~dʒʲ ʝʝ~dʒʲ
n r s l w j > N R S L v ʝ~dʒʲ		m n s r l w j > m n s~z r l w~∅ j~∅
					nn ss rr ll ww jj > N S R L v ʝʝ~dʒʲ
mj nj sj rj lj > mʲ Nʲ~ɲ sʲ~ʃʲ rʲ lʲ~ʎ	mj nj sj rj lj > mʲ Nʲ~ɲ sʲ~ʃʲ rʲ lʲ~ʎ

(later Romance has its own simplication of fortis and lenis sonorants; notably, fortis sonorants are all rendered in Astur-Leonese for instance as palatal)

Middle/New Indo-Aryan languages never really had consonant mutation, but some did have some extensive lenition as well to the point of complete elision, including of /m n/. Notably, this lenition occurs from Middle Indo-Aryan languages that had huge sets of geminate consonants.
eg.
Sans. grāma > Prk. gāma > Apbhr. gā̃və > Hindi gā̃ʋ~gā̃ō {village}
Sans. sthāna > Prk. thāna > Apbhr. thā̃və {station, place}

The broader system:

More: show

Sanskrit had
/p t ʈ tɕ k/
/b d ɖ dʑ g/
/pʰ tʰ ʈʰ tɕʰ kʰ/
/bʱ dʱ ɖʱ (dʑʱ) gʱ/
/m n ɳ (ɲ)/
/s ʂ ɕ/
/ʋ r l j ɦ/

Going into Prakrit or Pali, most to all consonant clusters simplified to singletons and geminates. Also there were a number of consonant mergers.
So Prakrit/Pali had
/p t ʈ tʃ k/
/b d ɖ~ɭ dʒ g/
/pʰ tʰ ʈʰ tʃʰ kʰ/
/bʱ dʱ ɖʱ~ɭʱ dʒʱ gʱ/
/pp tt ʈʈ ttʃ kk/
/bb dd ɖɖ ddʒ gg/
/ppʰ ttʰ ʈʈʰ ttʃʰ kkʰ/
/bbʱ ddʱ ɖɖʱ ddʒʱ ggʱ/
/m n ɳ ɲ/
/s r~l ʋ~j h/
/mm nn ɳɳ ɲɲ ss rr~ll jj/

The sound changes we observe:
Sanskrit to Pali

Code: Select all

initially				medially
					h[T S] > Tː sː
[T N]₁[T N]₂ > T₂ N₂			[T N]₁[T N]₂ > T₂ː N₂ː
dʑɲ > ɲ					dʑɲ > ɲː
					R[T N S v R] > [T N s b l]ː
T₁N₂ > T₁					T₁N₂ > T₁ː
[p t ʈ k]S(N)  > [tʃʰ tʃʰ tʃʰ kʰ]	[p t ʈ k]S(N)  > [tʃʰ tʃʰ tʃʰ kʰ]ː
[TNSv]R > [TNSv]			[TNSv]R > [TNSb]:
[TN]y > T					[TN]y > Tː (but dentals/retroflexes palatalise)
vy > b					vy > bː
S[y v] > s					S[y v] > sː
ST(R) > Tʰ				ST(R) > Tʰː
dv > b					dv > bː
					T+[s y] > sː yː
SN > s(iN)				[S h]N > Nh
						mr > mb
					h[v y] > [v y]h

Prakrit to Apabhramsa

Code: Select all

initially				medially
					p t ʈ tʃ k > β ð ɖ~ɽ ʝ~ʝʝ ɣ > ∅ ɽ dʒ ∅
					b d ɖ dʒ g > β ð ɖ~ɽ ʝ~ʝʝ ɣ > ∅ ɽ dʒ ∅
					pʰ tʰ ʈʰ tʃʰ kʰ > ɸ θ ɖʱ~ɽʱ ç x > ɦ ɽʱ
					bʱ dʱ ɖʱ dʒʱ gʱ > ɸ θ ɖʱ~ɽʱ ç x > ɦ ɽʱ
					pp tt ʈʈ ttʃ kk >> p t ʈ tʃ k
					bb dd ɖɖ ddʒ gg >> b d ɖ dʒ g
					ppʰ ttʰ ʈʈʰ ttʃʰ kkʰ >> pʰ tʰ ʈʰ tʃʰ kʰ
					bbʱ ddʱ ɖɖʱ ddʒʱ ggʱ >> bʱ dʱ ɖʱ dʒʱ gʱ
N s r ʋ j h > N s r ʋ dʒ h		N s r l ʋ j h >> ṽ s~h r~l l ∅
					NN ss rr~ll jj > N s r dʒ

(notice in Middle and New Indo-Aryan, initial consonants behave like geminates)

[bradrn]

I’ve found the following guide extremely useful for designing consonant mutation systems: https://web.archive.org/web/20150503182 ... tation.pdf

[Vardelm]

Perhaps worth mentioning is /n/ > /ð/ before /r/ in Sindarin, as in Caradhras. If it was good enough for JRRT...

EEGADS!!! I can't believe I almost used a mutation from an ELVEN language in my DWARVEN language! The HORROR!!! Changing back to /n/ > /s/ immediately.

(OK, yeah, if the professor can do it....)

I’ve found the following guide extremely useful for designing consonant mutation systems: https://web.archive.org/web/20150503182 ... tation.pdf

Looks like that's a prospectus for the dissertation I linked. Very similar material, but looks like there might be some mutation examples from a few different languages compared to the dissertation, which is always useful.

[bradrn]

I’ve found the following guide extremely useful for designing consonant mutation systems: https://web.archive.org/web/20150503182 ... tation.pdf

Looks like that's a prospectus for the dissertation I linked. Very similar material, but looks like there might be some mutation examples from a few different languages compared to the dissertation, which is always useful.

Huh, so it is. If only I’d known.

[Richard W]

What are some examples outside of insular Celtic which can be seen as /m/ patterning with obstruents rather than the other nasals?

Initial clusters /ml/ and sometimes /mr/, seen in PIE and as /ml/ in written Tai. While other nasal + liquid clusters have been reconstructed for Proto-Tai, they seem much rarer and to have simplified unrecorded, and it's the /ml/ cluster that one sees in written Thai and written Northern Thai and actually hears in Saek, Wu-ming and occasionally Ahom.