Proto-Langs

[Raphael]

If they're protolangs, why are they reconstructed like that? They don't need to have realistic phonologies. PIE doesn't.

Well, perhaps conlangernoob simply wants the protolangs to have realistic phonologies? It might be a simple aesthetic preference.

[conlangernoob]

If they're protolangs, why are they reconstructed like that? They don't need to have realistic phonologies. PIE doesn't.

Well, perhaps conlangernoob simply wants the protolangs to have realistic phonologies? It might be a simple aesthetic preference.

True! My goal for my Proto-Langs is not to be a stepping stone to descendant langs, but to be one puzzle piece is a beautiful system of languages and peoples of my world.

[alice]

Don't forget that the only practical difference between a proto-lang and a normal language is that a proto-lang tends to be older and not always fully reconstructible.

[WeepingElf]

Yep. The way human languages work hasn't changed appreciably during those few millennia historical linguistics can look into at the current state of knowledge, so languages spoken 5,000 or 10,000 years ago are just like modern languages, except, of course, they have no words for things not discovered or invented yet back then (just like attested ancient languages). There is no difference in kind. The notion that such "proto-languages" (I don't like that term - I prefer "common ancestor languages") were more "primitive" belongs to the 19th century, but it occasionally crops up whenever someone misses the fact that historical linguistics and language origins studies are entirely different disciplines and the "protolanguages" in the sense of language origins studies have nothing to do with the "proto-languages" in the sense of historical linguistics.

[Darren]

I think Nort's point was more that reconstructions of proto-langs are often unrealistic or whacky merely by nature of being reconstructed. Proto-Indo European probably wasn't as weird as we reconstruct it, it's just that the comparative method forces us to reconstruct it as such; likewise for such monstrosities as Baxter's reconstruction of Old Chinese phonology with 70 consonants and /qʰʷˤ/. While it's true that ancient languages were not appreciably different from modern languages, it's also true that (reconstructed) protolangs are very different entities and have something of a license to be unrealistic. If a protolangs' descendants consistently reflect four vowels as as /e ə o a/ or clear derivatives thereof, then they can be reconstructed as *e ə o a regardless of whether that was the vowel system at any one stage. However, conlangernoob has pointed out that these aren't reconstructed proto-langs, but rather conlangs that happen to be ancestors of language families, so it's a moot point.

[WeepingElf]

This is indeed a bone of contention among historical linguists. Some consider reconstructed ancestor languages mere formalisms of language relationships, and do not demand typological plausibility; others maintain that a reconstruction represents an actual language, and that an implausible reconstruction is wrong even if it accounts for the known facts perfectly.

[Travis B.]

Just because a proto-lang has been reconstructed in a certain fashion does not mean that it is very plausible as a real language. E.g. with PIE, it probably really had something like [i æ ɒ u] or maybe [i æ a ɒ u] as its set of monophthongs - it is just an artifact of the reconstruction that [i] and [u] have been "explained away" as a syllabic version of a basically non-syllabic glide phoneme, which ignores the fact that this is essentially typologically nonexistent in any real language that we know of. (For the very few languages analyzed with "/e a o/" that we have out of attested real languages, "/e/" turns out to really be /any non-low unrounded front vowel/ and "/o/" turns out to really be /any non-low rounded back vowel/, i.e. it is largely a choice on the analyzer's part that they didn't choose a more conventional "/i a u/".) This is reason I have to be skeptical of the traditional PIE reconstruction, many more traditional Old Chinese reconstructions (seriously, the vast majority of syllables can't include the glide [j], can they?!), or Baxter-Sagart (there has got to be a better way of solving the [j] problem than infesting the consonant system with pharyngealization everywhere?). There is no reason why a reconstructed proto-lang should behave differently from a real, attested one.

[bradrn]

it is just an artifact of the reconstruction that [i] and [u] have been "explained away" as a syllabic version of a basically non-syllabic glide phoneme, which ignores the fact that this is essentially typologically nonexistent in any real language that we know of.

This is attested! I already mentioned Kalam.

[Darren]

There is no reason why a reconstructed proto-lang should behave differently from a real, attested one.

I disagree. In the real world, a protolang is always hampered by the limitations of the comparative method, so it inevitably does behave differently from an attested one. If PIE was attested, we'd know a lot more about ablaut and syntax and verb conjugation and a shitload of other things that we simply don't have enough evidence to work out. Ideally speaking every protolang would be indistinguishable from attested languages, but they always end up being something of an abstraction because when it comes down to it they're basically guesswork. PIE with *e o and *T D Dʰ is still useful as a model even if it never existed.

Getting back on track, having conlangs (proto- or otherwise) with /e ə o a/ for vowels or no voiceless stops is fine, but beware that it's at the outer limits of naturalism and requires some careful treatment if you want to keep them plausible as human languages.

[Travis B.]

it is just an artifact of the reconstruction that [i] and [u] have been "explained away" as a syllabic version of a basically non-syllabic glide phoneme, which ignores the fact that this is essentially typologically nonexistent in any real language that we know of.

This is attested! I already mentioned Kalam.

Oh, I missed that, but even still this is only barely attested, considering the sum of all actually attested (and analyzed) natlangs. When one can point to a all-but-never-attested feature and say "but language X has that!", it might say more about the analysis of language X than anything else.

[Travis B.]

There is no reason why a reconstructed proto-lang should behave differently from a real, attested one.

I disagree. In the real world, a protolang is always hampered by the limitations of the comparative method, so it inevitably does behave differently from an attested one. If PIE was attested, we'd know a lot more about ablaut and syntax and verb conjugation and a shitload of other things that we simply don't have enough evidence to work out. Ideally speaking every protolang would be indistinguishable from attested languages, but they always end up being something of an abstraction because when it comes down to it they're basically guesswork. PIE with *e o and *T D Dʰ is still useful as a model even if it never existed.

That only works if one wants to treat a proto-language as a mere abstraction, a practically fictional language, produced via the comparative method, and not a real language that someone once spoke, where it just happens that our knowledge of it is incomplete due to the inevitable limitations of the comparative method. If we free proto-languages from the bounds of reality, of course we are going to get less-then-sound results when we attempt to reconstruct them. This is why IMO when reconstructing proto-languages we have to keep in mind just how plausible is what we are reconstructing were it a real language spoken by real people today and not a simple abstraction unbound by reality.

[conlangernoob]

Here is a set of sound changes I'm planning from Proto-Lang #1 to one of the descendent languages. All the sound changes are attested but I'm not sure how realistic it is.

ɡ / h / V_V
eu / eo /
s / θ / #_r
i / ə / _#
{f, s, x} / {v, z, ɣ} /
v / θ / _{a,e,i}
ɣ / k / _i
b / w / V_V
ɣ / ∅ / V_u
ɣ / ∅ / u_V
z / j /
i u / e o / _Ca
v / m / medial
{p t k} → {f θ x} /
s / h / #_
m / n / _i
u / w / #_
m / w / #_
∅ / a / w_j
a / ə / _Ca
p / k / _w
f / p / r_
{s, f, θ, x} / {z, v, ð, ɣ} / V_V

Two things:
1) How do I know at what point to call the proto-lang a different language than a descendent lang?
2) How do I know at what point allophonic variation due to sound changes turns allophones into phonemes if my lexicon is too small to produce minimal pairs?

Thanks for all the advice!

conlangernoob

[Darren]

There is no reason why a reconstructed proto-lang should behave differently from a real, attested one.

I disagree. In the real world, a protolang is always hampered by the limitations of the comparative method, so it inevitably does behave differently from an attested one. If PIE was attested, we'd know a lot more about ablaut and syntax and verb conjugation and a shitload of other things that we simply don't have enough evidence to work out. Ideally speaking every protolang would be indistinguishable from attested languages, but they always end up being something of an abstraction because when it comes down to it they're basically guesswork. PIE with *e o and *T D Dʰ is still useful as a model even if it never existed.

That only works if one wants to treat a proto-language as a mere abstraction, a practically fictional language, produced via the comparative method, and not a real language that someone once spoke, where it just happens that our knowledge of it is incomplete due to the inevitable limitations of the comparative method.

Sometimes one does. Both approaches have their merits.

[Travis B.]

Here is a set of sound changes I'm planning from Proto-Lang #1 to one of the descendent languages. All the sound changes are attested but I'm not sure how realistic it is.

ɡ / h / V_V

Unconditional ɡ > ɦ is attested in Czech, Slovak, and Ukrainian (yes, they have breathy voicing here, but that is a minor detail), I should note.

eu / eo /

This reminds me of Old English.

s / θ / #_r

s > θ does happen (e.g. it happened unconditionally in Burmese), but I wonder why /r/ in particular would trigger it (not that there's anything wrong with that.)

i / ə / _#

I might suggest making this dependent on stress, but otherwise fine.

{f, s, x} / {v, z, ɣ} /

I would consider making this conditional, e.g. limiting it to being prevocalic or just intervocalic or when adjacent to a voiced consonant.

v / θ / _{a,e,i}

I would consider [ð] here.

ɣ / k / _i

I would hesitate to do this one, because it is hard to justify why [ɣ] would undergo fortition specifically before [i].

b / w / V_V

A perfectly cromulent sound change.

ɣ / ∅ / V_u
ɣ / ∅ / u_V

This is perfectly fine too.

z / j /

I would expect an intermediate step of something like [ʝ], but all in all this is fine.

i u / e o / _Ca

Simple a-umlaut.

v / m / medial

I would hesitate on this one, because spontaneous nasalization of non-nasal consonants without assimilation to something (e.g. a nasal vowel somewhere else in the word) tends to be limited to rhinoglottophilia, something that seems to be hard to apply to [v].

{p t k} → {f θ x} /

Each of these by itself is perfectly fine, but I would leave a means for voiceless plosives to remain (e.g. preserve them in certain positions, e.g. in certain consonant clusters, or when geminate or like, or don't do all of these at once).

s / h / #_

Also a perfectly cromulent sound change (re: Greek, Iranian).

m / n / _i

I haven't really heard of this one, but it seems to be natural POA assimilation to me, so I am certain it must happen in languages that I just can't recall off the top of my head.

u / w / #_

I would modify this to be when, preconsonantal, to be u > wu instead, to avoid wC clusters, but otherwise it seems fine.

m / w / #_

Reminds me of Irish.

∅ / a / w_j

As you use [ə] elsewhere, for an epenthetic vowel I'd also consider using [ə] here as well.

a / ə / _Ca

I would make this stress-dependent, but otherwise it is fine.

p / k / _w

Seems perfectly fine.

f / p / r_

Normally p > f is more common than f > p, but fortition here seems fine to me.

{s, f, θ, x} / {z, v, ð, ɣ} / V_V

This is so common to not even need any justification at all.

Two things:
1) How do I know at what point to call the proto-lang a different language than a descendent lang?

This is essentially arbitrary, and is up to you really. RL linguists use "proto-" when applied to reconstructed, unattested languages (but sometimes use "proto-" in other cases as well, such as "proto-Romance"), but since this is a conlang you speak of that is a non-factor here.

2) How do I know at what point allophonic variation due to sound changes turns allophones into phonemes if my lexicon is too small to produce minimal pairs?

I would consider allophones as having turned into independent phonemes when their distribution is no longer predictable in a regular fashion.

Travis

[Darren]

Your sound changes on their own are mostly fine, although I'd second all of Travis's suggestions.

v / θ / _{a,e,i}

This is the weirdest one you've got. I can potentially see it working if as Travis said you change it to ð instead. Ideally, you could have something like {m v} / {n ð} /_i, which would make sense as a series (labial > coronal in palatalising conditions is attested multiple times even when other POA's don't change).

Two things:
1) How do I know at what point to call the proto-lang a different language than a descendent lang?

When you feel like it basically. It depends a lot on changes in the lexicon and morphology as well as phonological changes.

2) How do I know at what point allophonic variation due to sound changes turns allophones into phonemes if my lexicon is too small to produce minimal pairs?

When the environment of the allophony becomes obscured by further changes. For example, you might start off with an allophonic rule like this:

u o a / y ø ɛ /_Ci

In which case [y ø ɛ] are allophones of /u o a/. But then if there's another sound change of

{i u} / Ø /VC_#

This will make /y ø ɛ/ phonemic, because you can't always predict where [y ø ɛ] will appear rather than /u o a/ (and this can be evidenced by minimal pairs like *moku moki > /mok møk/).
Alternatively, sounds can be phonemicised by borrowing or by analogy. Say the second sound change I mentioned didn't take place - /y ø ɛ/ could still be made phonemic if the language borrows words like /puri/ or /føna/; or, if -i is a morphological ending it may not trigger umlaut thanks to analogy, e.g. /mok/ + /-i/ → /moki/ rather than */møki/.
It's useful to remember that sound changes tend to encourage other changes that phonemicise them. Going back to the example above, with the first change you get pairs like [møki] : [moku]. This means that speakers can now get away with dropping final high vowels, because there is redundant phonetic information elsewhere in the words, and the final vowel can be dropped without making homophones.

[Rounin Ryuuji]

I agree with most of what's already been said, but to elaborate further —

s / θ / #_r

s > θ does happen (e.g. it happened unconditionally in Burmese), but I wonder why /r/ in particular would trigger it (not that there's anything wrong with that.)

In this case, I would probably expect /r/ to be dental, possibly [ɾ̪].

v / θ / _{a,e,i}

I would consider [ð] here.

This is how I would imagine it happened.

ɣ / k / _i

I would hesitate to do this one, because it is hard to justify why [ɣ] would undergo fortition specifically before [i].

It would possibly need a chain of [ɣ] > [ʝ] > [ç] > [c] > [k], though depalatalisation before [i] would be unusual, and I expect if it were phonemically /k/, it would remain phonetically [c].

{p t k} → {f θ x} /

Each of these by itself is perfectly fine, but I would leave a means for voiceless plosives to remain (e.g. preserve them in certain positions, e.g. in certain consonant clusters, or when geminate or like, or don't do all of these at once).

Possibly doing something like a preceding *s would block the shift (or actually trigger it, with the stop assimilating to the manner of articulation of the preceding fricative, with the fricative then being lost).

m / n / _i

I haven't really heard of this one, but it seems to be natural POA assimilation to me, so I am certain it must happen in languages that I just can't recall off the top of my head.

I can't think of any examples, either, but [m] > [ɲ] before [i] seems plausible enough if Old French can have medial palatal /p b/ become /tʃ dʒ/.

[Darren]

I can't think of any examples, either, but [m] > [ɲ] before [i] seems plausible enough if Old French can have medial palatal /p b/ become /tʃ dʒ/.

That's an example that gets used a lot, but I think the current consensus is that it actually went pjV bjV > ptʃV bdʒV > tʃV dʒV rather than pʲ bʲ > tʃ dʒ; a more fun example is Tolomako, which had p m > t̼ n̼ > t n before front vowels. m v > n ð /_[ie] would be fine, although I'd probably expect it to affect /p/ as well.

[Man in Space]

I can't think of any examples, either, but [m] > [ɲ] before [i] seems plausible enough if Old French can have medial palatal /p b/ become /tʃ dʒ/.

That's an example that gets used a lot, but I think the current consensus is that it actually went pjV bjV > ptʃV bdʒV > tʃV dʒV rather than pʲ bʲ > tʃ dʒ; a more fun example is Tolomako, which had p m > t̼ n̼ > t n before front vowels. m v > n ð /_[ie] would be fine, although I'd probably expect it to affect /p/ as well.

That 1983 paper on Ryukyuan cites some Ryukyuan languages with *p > *ts before at least some front vowels.

[Rounin Ryuuji]

I can't think of any examples, either, but [m] > [ɲ] before [i] seems plausible enough if Old French can have medial palatal /p b/ become /tʃ dʒ/.

That's an example that gets used a lot, but I think the current consensus is that it actually went pjV bjV > ptʃV bdʒV > tʃV dʒV rather than pʲ bʲ > tʃ dʒ

Apparently the forms apcha and abja are attested in Old Occitan, so that theory sounds likely.

...a more fun example is Tolomako, which had p m > t̼ n̼ > t n before front vowels. m v > n ð /_[ie] would be fine, although I'd probably expect it to affect /p/ as well.

What's that little squiggly under the t n? I've never seen it before.

[WeepingElf]

What's that little squiggly under the t n? I've never seen it before.

An IPA diacritic marking linguolabials.