Primordial Scratchpad (NP: Dwarvish Consonant Gradation)

[Vardelm]

Happy New Year ZBB!

On the old board, I had my Primordial Scratchpad thread to post about the 4 (maybe 5) proto-languages I'm working on. The past few weeks I've been able to get back to it a bit, so I'm starting a new thread here.

As before, I'll color the titles to help distinguish posts on each language.

Devani
Dwarvish
Jin
Elvish / Yokai

[Vardelm]

Dwarvish Phonology

I've revised my Dwarvish phonology a bit by making it slightly more Gaelic &/or Nordic than it was. The use of /q ʀ/ is the main divergence from that theme which hopefully makes it feel more like its own thing.

Consonants: IPA

	Labial	Dental	Alvelolar	Palatal	Velar	Guttural
Stops	p b pʲ bʲ	t ̪ d ̪	tʲ dʲ		k g kʲ gʲ	q qʲ
Fricatives	f v fʲ vʲ	θ θʲ	s	ʃ	x xʲ
Nasals	m mʲ	n nʲ			ŋ ŋʲ
Continuants			l lʲ	j		ʀ h

Consonants: Orthography

	Labial	Dental	Alvelolar	Palatal	Velar	Guttural
Stops	p b pj bj	t d	tj dj		k g kj gj	q qj
Fricatives	f v fj vj	ð ðj	s	sh	ch chj
Nasals	m mj	n nj			ñ ñj
Continuants			l lj	j		r h

/θ θʲ s ʃ x xʲ/ are probably voiced intervocalically.


Vowels: IPA

	Front	Mid	Back
High	i y	ʉ	u
Mid	e ø		o
Low	æ œ		ɑ

Vowels: Orthography

	Front	Mid	Back
High	i y	ü	u
Mid	e ö		o
Low	æ å		a

All vowels will probably have long variants as well, but I haven't worked out how that will fit with the rest of the phonology yet.

OK, maybe I'm also just having some fun w/ Zompist's new phonology builder.

EDIT: Changed "Scandinavian" to "Nordic" since I include Icelandic & Finnish as influences (and by extension other Uralic languages, although they aren't "Nordic").

[Vardelm]

Dwarvish Consonant Gradation

Since I want my Dwarvish proto-lang to use phonological elements from Gaelic, Nordic, & Uralic languages, the plan is to include a system of consonant mutation & gradation. Over the last week+ I have been trying to understand consonant gradation better than I used to since I'd like to have a more naturalistic system than what I have done in my earlier conlanging activities. Currently my thinking is to have a "base" of consonant gradation like the Uralic languages, and then layer initial consonant mutations at certain word boundaries per Gaelic on top of that.

My design plan right now is:

• 3 grades of consonants, similar to Sami & Estonian
• Most or all consonants are part of the grading, not just stops or stops & fricatives.
• There should be a number of series of grades that overlap each other so it's not obvious what grade series is being used

Based on what I have read at Wikipedia, Language Gulper, and other sources, I determined what consonant grades (strong, median, & weak) should be used for an onset based on whether the syllable is stressed vs. unstressed and open vs. closed.

Consonant Grades per Syllable Type

	Stressed	Unstressed
Open	strong	median
Closed	median	weak

AIUI, the exact source of consonant grades in Uralic langauges is murky. I don't know if my consonant grades would have developed from a pure fortition process of long & overlong gemination (similar to that found in Sami) or from a combination of fortition via gemination for strong & lenition (mostly to voicing?) for weak. I'm sort of proceeding based on the long & overlong gemination origin for now. It could even be some combination, perhaps.

The idea is that the original system of long & overlong gemination degrades (pun intended! ) into grades of long, median, & lenited (voiced?). There may have only been unvoiced phonemes (w/ voiced allophones in some way???) and the short consonants starting to become voiced in more situations. The long & overlong consonants remained fully unvoiced. The long shortened to regular unvoiced consonants since the short consonants became mostly voiced, and the overlong shortened to just long.

That said, I had been thinking there might have been a voiced series of consonants (just stops???) with short, long, & overlong. The long & overlong would devoice & merge with the original voiceless series. I had also thought the short voiced would lenit to fricatives (or other articulation), while the long voiced would devoice & shorten to merge with the short unvoiced, and the overlong voice devoice & shorten to merge with long unvoiced. Something like this:

Development of Graded Stops V1

Grade	Original	Current
Unvoiced Strong	t::	t:
Unvoiced Median	t:	t
Unvoiced Weak	t	d
Voiced Strong	d::	t:
Voiced Median	d:	t
Voiced Weak	d	d

Development of Graded Stops V2

Grade	Original	Current
Unvoiced Strong	t::	t:
Unvoiced Median	t:	t
Unvoiced Weak	t	d
Voiced Strong	d::	t
Voiced Median	d:	d
Voiced Weak	d	n

Version 2 gets things a little closer to what I want with "overlapping" series of grades, where between the two it ends up with a string of t: - t - d - n. I'm not sure it's as likely as version 1, unless the shift of d > n happens 1st, allowing d:: & d: to shorten and then allow the resulting d: > t.

---

I could use input on this, although I don't know that there's enough of the system detailed to comment much yet.

The development is one thing (and not SUPER important since this is just a proto-lang), but the main issue will be what all of the actual grading series are. My questions will mostly center on how voiced stops should be lenited, given that much of my fricative series are unvoiced. I think I'll work on another post at some point to try & hash out what possibilities I see for that & maybe ask specific questions.

[gach]

It's always nice to see different people's takes on consonant gradation systems. I can think of a couple of suggestions you could think abut.

The lenition in the gradation systems is in principle no different from lenition in other contexts. That means that while leniting voiced stops into nasals might make sense in a purely paradigmatic way, you'd still better think about how to justify it historically. More straight forward outcomes for leniting voiced stops would be voiced fricatives or glides.

If you want to keep the ban of most voiced fricatives strict, an original lenition pattern or D > [Ð] (that is, voiced stops into allophonically voiced fricatives) opens up one interesting development path. Once the speakers start to consider the lenition outcomes of the voiced stops as being phonemically fricatives, they may decide that these have to behave exactly the same as the original fricatives and thus loose their voicing. Now you'd have a system where voiceless stops lenite into voiced stops but voiced stops lenite into voiceless fricatives, which would give an interesting complication to the gradation system. You could even decide that in certain cases the voiced fricatives could have been dealt in some other ways possibly leading to some irregular word forms in the proto language.

Do you allow any consonant clusters? If so, have you thought about how gradation affects them? That's something I've been reconsidering on and off for years in the gradation system of Kišta.

You give stress as one of the deciding factors for the gradation. Does this only concern the primary stress or are there rhythmical considerations at play as well?

Lastly, your geminates presumably span the syllable boundary and so make the preceding syllable closed. How does this affect the determination of grades in the surrounding syllables? If gradation has effects like this that spread from one syllable to another, in what way and to which direction do they spread?

[Vardelm]

It's always nice to see different people's takes on consonant gradation systems. I can think of a couple of suggestions you could think abut.

Great! As one of the resident gradation gurus, I was hoping you might take a look at this thread at some point, especially once there is something of substance to describe.

The lenition in the gradation systems is in principle no different from lenition in other contexts. That means that while leniting voiced stops into nasals might make sense in a purely paradigmatic way, you'd still better think about how to justify it historically. More straight forward outcomes for leniting voiced stops would be voiced fricatives or glides.

Yep, I agree. I have an article that extensively covers mutation in the Atlantic African languages (Fula, etc.) and also discusses the Gaelic & Uralic systems. That showed that voiced stops > voiced nasals was at least possible, so I figured I would list that as a starting point. I had concerns over whether voiced stops could realistically lenit to unvoiced fricatives, so I figured it was at least a decent illustration.

If you want to keep the ban of most voiced fricatives strict, an original lenition pattern or D > [Ð] (that is, voiced stops into allophonically voiced fricatives) opens up one interesting development path. Once the speakers start to consider the lenition outcomes of the voiced stops as being phonemically fricatives, they may decide that these have to behave exactly the same as the original fricatives and thus loose their voicing. Now you'd have a system where voiceless stops lenite into voiced stops but voiced stops lenite into voiceless fricatives, which would give an interesting complication to the gradation system. You could even decide that in certain cases the voiced fricatives could have been dealt in some other ways possibly leading to some irregular word forms in the proto language.

Considering the concerns I had over voiced stops > unvoiced fricatives, this is really interesting! I'm glad that this is at least a possible path to consider. I'm not 100% sure on having only unvoiced fricatives. In any event, descendant langs will eventually have voiced/unvoiced fricatives. As I start to develop this more, this will be something to consider.

Do you allow any consonant clusters? If so, have you thought about how gradation affects them? That's something I've been reconsidering on and off for years in the gradation system of Kišta.

I do. It's CCCVCCC, with consonants closer to the vowel needing to be more sonorous. I think the final consonant will only be there if there is gemination. Beyond that, I haven't thought much about clusters as I have just been trying to understand Uralic gradation with just single consonants better. Clusters are something I know I'll need to think more about, though.

You give stress as one of the deciding factors for the gradation. Does this only concern the primary stress or are there rhythmical considerations at play as well?

I'm planning rhythmical stress, pretty much exactly like in Uralic. I just read Helimski 1995, and have a little better understanding of how that affects gradation. Before, I thought open syllables w/ secondary stress had strong grade consonants, but apparently they actually get weak. However, that misunderstanding lead to a potentially interesting twist on what my syllables look like, and so I may stick with the table I have above (Consonant Grades per Syllable Type).

Lastly, your geminates presumably span the syllable boundary and so make the preceding syllable closed. How does this affect the determination of grades in the surrounding syllables? If gradation has effects like this that spread from one syllable to another, in what way and to which direction do they spread?

Yes, correct, they will span boundaries. I think, like Uralic, a syllable becoming closed will lenit the onset. This will probably be regressive.

One thing I have been planning is that prefixes may end in a particular consonant that assimilates to the 1st consonant of the root it attaches to. That would create a geminate, and in descendant langs the geminates will devoice and eventually become aspirates. This will create an additional fortition path that is progressive.

In the descendants, what I would eventually like to have (if possible) is only the initial and final consonant mutating to indicate some inflections/derivations, and then perhaps the next consonant in mutating at least some of the time. I think this is doable, but we'll see.

[Pabappa]

Pabappa forms diminutives by shifting voiced stops to nasals. This arose from generalization of a contrast that was originally of the form /nd ~ d/, shifting to /nn ~ d/, where the morpheme that caused the diminutive ended in /n/. Thus, when this morpheme was removed, the contrast remained.

The African ones may have originated as a contrast between nasal and oral vowels, with the effects precipitating out onto consonants as they became grammaticalized. This could lead to internal gradations even if the morpheme causing them was originally confined to the edge of the word.

The reverse of my Pabappa shift might also work .... you could have pp~pm --- > pp~mm and then delete the morpheme beginning with /m/.

Lastly, mp~mb could shift to mb~mm, although I don't know of a natlang that's done it.

[gach]

I'm planning rhythmical stress, pretty much exactly like in Uralic. I just read Helimski 1995, and have a little better understanding of how that affects gradation. Before, I thought open syllables w/ secondary stress had strong grade consonants, but apparently they actually get weak. However, that misunderstanding lead to a potentially interesting twist on what my syllables look like, and so I may stick with the table I have above (Consonant Grades per Syllable Type).

Yep, at least with the Uralic fortition style gradation the motivation seems to be along the lines of giving more weight to the stressed syllables by strengthening the following consonants. On the other hand, I don't see why there couldn't be a fair amount of language to language dependence of how and where the strengthening of a syllable actually surfaces.

[cedh]

The lenition in the gradation systems is in principle no different from lenition in other contexts. That means that while leniting voiced stops into nasals might make sense in a purely paradigmatic way, you'd still better think about how to justify it historically. More straight forward outcomes for leniting voiced stops would be voiced fricatives or glides.

Yep, I agree. I have an article that extensively covers mutation in the Atlantic African languages (Fula, etc.) and also discusses the Gaelic & Uralic systems. That showed that voiced stops > voiced nasals was at least possible, so I figured I would list that as a starting point. I had concerns over whether voiced stops could realistically lenit to unvoiced fricatives, so I figured it was at least a decent illustration.

If you want an alternation between voiced stops and voiced nasals, you could also try finding a historical explanation that works in reverse compared to the origin of the lenition pattern between voiceless stops and voiced stops. For instance, the original sounds in this series might have been nasals, not plosives, which remained unchanged in those positions where voiceless stops would become voiced, but were strengthened in at least some of those position where voiceless stops would remain unchanged, for instance by becoming geminated nasals, glottalized nasals, prestopped nasals, or prenasalized stops. This "fortis nasal" series could then later merge into the existing voiced stops through an independent sound change, and voilà, there you have a gradation pattern /d/ :: /n/ in the same positions where /t/ :: /d/. If the occurrences of /d/ :: /n/ brought about by sound change are frequent enough (for instance because they appear in a couple of very common words), they can plausibly be extended to most or even all instances of /n/ by analogy.

Also, while I haven't really made a conlang so far that has a fully systematic consonant gradation system, I've worked with lenition-style sound changes a lot, and one specific change I've used several times myself may also be a good option for you: a shift of voiced plosives to approximants. The details may vary, of course. In the context of the phonology you sketched out above, /b bʲ d ̪dʲ ɡ ɡʲ/ > /v vʲ l lʲ ʀ j/ would make a lot of sense.

[Vardelm]

The paper I was referring to with info about African consonant mutation is The Historical Origin of Consonant Mutation in the Atlantic Languages by John Merrill. I've only been able to skim it, but there's lots of good info in there.

[Vardelm]

If you want an alternation between voiced stops and voiced nasals, you could also try finding a historical explanation that works in reverse compared to the origin of the lenition pattern between voiceless stops and voiced stops.

Hi cedh!

If you can recall way back in the 2008-2010 time frame, you provided quite a bit of advice on my 1st conlanging effort, Tibetan Dwarvish. What you're seeing here is what will be the proto-language (or proto-proto-proto-...-language) for a much revised version of that. One of the other "primordial languages" in this project, Devani, will also be a major contributor as well. I'm working on a heavy revamp of that language compared to what's on the old board.

That idea on working in reverse is a good idea, if for no other reason than to not get stuck in 1 mode of thought.

If you want an alternation between voiced stops and voiced nasals, you could also try finding a historical explanation that works in reverse compared to the origin of the lenition pattern between voiceless stops and voiced stops.

Not sure on that yet. I don't necessarily want voiced stops <> nasals, but it's an option. I do like the suggested origin you give, and I could see something like that being used in the more Gaelic features. I sort of have in mind to have 2-4 different types of mutation going on there, so I could see this as being one of them.

...for instance by becoming geminated nasals, glottalized nasals, prestopped nasals, or prenasalized stops.

Introducing glottalization into geminated nasals, approximants, & fricatives was one idea I had for a path to fortition. Prestopped is an interesting suggestion, too.

Also, while I haven't really made a conlang so far that has a fully systematic consonant gradation system, I've worked with lenition-style sound changes a lot, and one specific change I've used several times myself may also be a good option for you: a shift of voiced plosives to approximants. The details may vary, of course. In the context of the phonology you sketched out above, /b bʲ d ̪dʲ ɡ ɡʲ/ > /v vʲ l lʲ ʀ j/ would make a lot of sense.

That's good! I had thought a little about a shift to approximants too instead of fricatives, and that's about what I was thinking. /d ̪dʲ/ > /l lʲ/ were the shifts I was most hesitant about.

The only thing about this is that I would like to have grades for the nasals, fricatives, & approximants as well as the stops, so leniting fricatives to approximants seems most likely, although I could maybe see nasals. Manx, Cornish, & Breton all have /m/ > /v/ or /f/. Granted, that's in the reverse, but might be usable.

[Vardelm]

I did a little work on the strong grade (out of strong, median, & weak). Current idea is that the strong grade is a result of gemination, and the weak grade is lenition. As such, there will only be long consonants, but not overlong. I could, for this proto-lang, just keep the gemination & call it end of story for the strong grade, and leave the sound changes to descendant langs. However, there are a few sounds that I feel would be rather quick to change from geminates to something else, so I'm listing them as changed here. Maybe I'm wrong about that. Here's the plan for sound changes vs no change for geminated consonants for the strong grade:

unvoiced stops = no change
<pp ppj tt ttj kk kkj qq qqj> ⇒ <pp ppj tt ttj kk kkj qq qqj>

voiced stops > unvoiced stops
<bb bbj dd ddj gg ggj> ⇒ <p pj t tj k kj>

bilabial fricatives > bilabial stops
<ff ffj vv vvj> ⇒ <p pj b bj>

glottal fricative > uvular stop
<hh> ⇒ <q>

other fricatives = no change
<ðð ððj ss ssj shsh shshj chch chchj> ⇒ <ðð ððj ss ssj shsh shshj chch chchj>

nasals = no change
<mm mmj nn nnj ññ ññj> ⇒ <mm mmj nn nnj ññ ññj>

palatal approximant > palatalized voiceless velar fricative
<yy> ⇒ <chj>

other approximants = no change
<ll llj rr> ⇒ <ll llj rr>

[Vardelm]

In the context of the phonology you sketched out above, /b bʲ d ̪dʲ ɡ ɡʲ/ > /v vʲ l lʲ ʀ j/ would make a lot of sense.

I looked at my lenition today yesterday, and thinking more about this got some creativity going. What if /v/ was originally /w/ or /β/? If that changed to /v/, that would explain why the bilabials have the voice distinction, but not the rest of the fricatives. It also means that all of the voiced stops originally lenited to continuants, which is nice & organized.

Also, I believe nasals could lenit to approximants. If so, this also adds a nifty bit of Gaelic mutation: /m mʲ/ > /v vʲ/. I'm not if this works, but I would maybe add /n nʲ/ /l lʲ/.

Reading about lenition on Wikipedia also suggested the "opening" type of lenition rather than sonorization. That suggests a different path than unvoiced stops > voiced, which is /p pʲ t tʲ k kʲ q qʲ/ > /f fʲ ð ðʲj ch chʲ h j/. Having /qʲ/ > /j/ in there doesn't follow the rest of the pattern, but I think many of the palatalized consonants will lenit to /j/. The nice part of this is that it avoids the whole path of unvoiced stop > voiced stop > unvoiced fricative.

For the fricatives, I'm thinking /f fʲ s ʃ x xʲ/ > /v vʲ h j h j/. I saw /s/ > /h/ in Scottish Gaelic (IIRC), so these seem realistic.

I see /ʀ/ > /h/ and /h/ just eliding altogether.

That leaves /v vʲ θ θʲ ŋ ŋʲ l lʲ/. I'm not quite sure how to handle lenition of those consonants. I could see /v vʲ l lʲ/ eliding like /h/. The other 2 I'd like to find a different answer for if I can.

Lenition Summary:
/p pʲ t tʲ k kʲ q qʲ/ > /f fʲ θ θʲ x xʲ h j/
/b bʲ d dʲ ɡ ɡʲ/ > /v vʲ l lʲ ʀ j/
/f fʲ s ʃ x xʲ/ > /v vʲ h j h j/
/m mʲ n nʲ/ > /v vʲ l lʲ/
/ʀ/ > /h/
/h/ > ∅

/v vʲ θ θʲ ŋ ŋʲ l lʲ/ > ???

[mèþru]

Why /t tʲ/ -> /ð ðʲ/ instead of /θ θʲ/?

[Vardelm]

Why /t tʲ/ -> /ð ðʲ/ instead of /θ θʲ/?

Whoops, good catch. That was a mix up between IPA & my orthography. Fixed it. I realize using <ð> for /θ/ might not be the best choice, but I much prefer the look of it, and /ð θ/ are allophones, so I'm sticking with it.

[Vardelm]

Small update:

/θ θʲ/ > /l lʲ/ (lines up with the other fricatives)

/v vʲ l lʲ ŋ ŋʲ/ = no change for lenition

Lenition Summary:
/p pʲ t tʲ k kʲ q qʲ/ > /f fʲ θ θʲ x xʲ h j/
/b bʲ d dʲ ɡ ɡʲ/ > /v vʲ l lʲ ʀ j/
/f fʲ θ θʲ s ʃ x xʲ/ > /v vʲ l lʲ h j h j/
/m mʲ n nʲ/ > /v vʲ l lʲ/
/ʀ/ > /h/
/j h/ > ∅
/v vʲ l lʲ ŋ ŋʲ/ = no change


I'm also leaning towards fortition being simple, straigtforward gemination of all consonants: no sound changes. I'll leave those for descendant langs.

I need to review exactly what environments trigger the grades, at least for single consonants, and then think about clusters.