A little reconstruction game

[bradrn]

Well, some things got in the way on my end, and interest seems to have died down anyway, but if anyone's still engaged with this, I've put up sample sentences for a third language now.

Well, I haven't been too involved in this game (I'm not too good at reconstruction), but at a cursory look it seems like that'll help a lot in clearing up some issues people have been having.

[Moose-tache]

For my part, I'm very keen to sink my teeth into the new material. I think the perceived decline in interest is just because this isn't 4chan and we're not going to shout "bump" at you when you're slow to upload.

[KathTheDragon]

I'm very much still interested, just haven't found the time

[dhok]

Yes, I'm interested. I think a number of us have been waiting for language number three.

[mèþru]

Ejectives?!
The plot thickens

[KathTheDragon]

Found what seems to be a mistake: A has ɘɪ a number of times for what should probably be ɐɪ in 10-12. I'm also wondering if the tɜ in 5A is correct, since no other word has this vowel... Edit: I'm certain that tɜ is an error for tɐ

Edit: 10B has both ɾɛpɑg and ɾɛbɑg, one of which seems to be a mistake.

Edit: 1C has nəse instead of nəːse

Edit: typo in 9C: æid for æɪd

[Salmoneus]

Thank you. And sorry. Those pesky ancient scribes...


[in particular, doing this has really made me hate the fact that IPA has all of ə, ɐ, and ɘ... they make my eyes get crossed just trying to make out which is which...]

[KathTheDragon]

Looks like 12B has dɛʊsɪ instead of nɛʊsɪ.

Edit: I also suspect 9B has næʊ instead of nɛʊ.

Edit: Can we also get confirmation of the correct forms? You seem to be indicating I'm correct with all of those in the first post, but it'd be nice to make it explicit.

[Salmoneus]

Oh, sorry. I had updated the original post, but you're right I should say explicitly here.

Found what seems to be a mistake: A has ɘɪ a number of times for what should probably be ɐɪ in 10-12.

I only found instances in 12, but you're right. Those symbols confuse my brain.

I'm also wondering if the tɜ in 5A is correct, since no other word has this vowel... Edit: I'm certain that tɜ is an error for tɐ

You're right. I hadn't been sure which of the two symbols to use for this, and evidently an uncorrected one slip through...

Edit: 10B has both ɾɛpɑg and ɾɛbɑg, one of which seems to be a mistake.

Yes, they should both have /b/

Edit: 1C has nəse instead of nəːse

Yup.

Edit: typo in 9C: æid for æɪd

Yup.

Looks like 12B has dɛʊsɪ instead of nɛʊsɪ.

No! this one is not an error!

Edit: I also suspect 9B has næʊ instead of nɛʊ.

Halfright. It's actually nɑʊ.

My continuous obsequious apologies for my ineptitude...


In related news: bangs head on desk. So apparently a computer crash eliminated my vowel diachronics diagrams for C (surely I saved that? surely!?), meaning that I get to join in the reconstruction... It isn't a serious problem, since unlike you, as well as the reflexes I've given I also know the phonology of the parent language, and have a general idea of how it develops into C, so I shouldn't have any difficulty redrawing my diagrams. It's just... really irritating.

[dhok]

It is obvious from eyeballing the sentences that B and C form a subgroup opposed to A or, at the very least, that if they do not that A is unusually divergent and that comparisons of B and C are lower-hanging fruit. Let's take a look.

First, initials. These line up well enough that no real table is needed except for differences:

- β~w, e.g. βøɣ~wɛg.

- ʃ~tʃ, e.g. ʃæhɪ:~tʃəʃe: (both of these in 3.) This correspondence also seems to operate finally (mäʃ~mætʃ) and medially (grɔɪʃoz~grə:tʃuz). There is also a pair ʒoɾ~dʒuɾ in 10 which shows B's fricativization was a general change to Proto-B-C *tʃ *dʒ. (However, sentence 3 has mʏh~meʃ. It appears PBC *ʃ gave B /h/ after a vowel...except that there is a cognate set in 3, A~B~C dʒɑhe:~ʃæhɪ:~tʃəʃe:. There are no other examples of C /ʃ/ except for /dɑʃe/ in 10, which has no apparent B cognate but does have an A correspondant /dɐʒɛ/ (not **dɐhe as we might expect). Was at least one word in these two correspondence sets a borrowing?

- However, the pair seg~ʃək in 11 suggests we are dealing with a chain shift *ʃ > s, *tʃ > ʃ in B

- ð~j, e.g. ðuɣ~jog (3)--we can tell these are probably cognate because we have already seen B /ɣ/ corresponding to C /g/ in final position. We also have ðɛh~jɛ (passim), ðuɣɪð~jogd (4). The cognate pair ðɾɪ:gɪ:~dre:k'e: suggests that Proto-B-C *ð gives C /j/ prevocalically and /d/ before a consonant.

- I have hardly been able to keep the lenition processes in my head, but is C alone in never having undergone lenition?

(Perhaps it's really A and B that are the subfamily...)

Other changes in no particular order:

- There appears to have been final nasal loss in B: witness u:~un and fɑ:ʊ~ɸäʊn (5 and 7). And in 10 we have sɔ:ɪjoz~sə:njuz.

- There is some weirdness with /u/ and /o/. We have B /o/ corresponding to A and C /u/ in ɣo~ɣud~ɣot (11), but B /u/ corresponding to C /o/ in dʒɔ:ɾ~ʒor~dʒuɾ (10). The correspondence of the latter is backed up by ʔɔʎɔ~ojo~ulju (5).

Additionally, there seems to have been a process of cliticization in B. We know word-final voiceless obstruents in C correspond to voiced ones in B. See sentence 2:

B: ti βoðʏzo ɔɪdɪː mɔɪd bä wäːz βɑʒɪteɾjɪː
C: ti ɛjɛ wud əːs məːt bæ wæn βäːtsəɾn

Where did C ə:s go in B? It seems likely that an original *wän or *wæn--there is no obvious cognate in A but at least some instances of B /ä/ correspond to A /æ/, so perhaps *wæn was original--underwent final nasal loss and compensatory lengthening in B (c.f. B~C u:~un. There is no other precedent for a final correspondence of z~n, but we expect any cognate of C ə:s to have the form Vz in B. This appears to have happened, if original *wæn Vs gave pre-B wä: Vz, which coalesced to wä:z. The implication is that one or the other of the languages underwent some sort of syntactic change, but it's not yet clear what kind.

[mèþru]

I don't think that there is sufficient analysis to group A,B,C into subfamilies - we need to figure out the order of changes first and which if any could be areal

[KathTheDragon]

Having spent much of the day working on expanding my vowel correspondences to include C and the new sentences, here're my more certain results. I've included a reconstruction where I feel confident of it. (A ~ B ~ C ~ *P-ABC)

• a(ː) ~ ɑ ~ ä
• aː ~ ɑː ~ äː
• aː ~ æɪ ~ æɪ ~ *aɪ (B also has single instances of æːʏ and æjɛ)
• aː ~ ɪ ~ æ
• æ(ː) ~ ä ~ æ
• æ ~ ɪː ~ æ,Ø (both of these are not simple correspondences, more below)
• ɐ ~ ɔ ~ ɑ
• ɐɪ ~ ɔ(ː)ɪ ~ əː ~ *əɪ (length in B is due to loss of coda *n)
• ɐʊ ~ ɛʊ ~ əː ~ *əʊ
• ɑ ~ æ ~ ə
• ɑ ~ ɑ ~ ä
• ɑː ~ ɑ(ː)ʊ ~ äʊ ~ *aʊ (length in B is due to loss of coda *n)
• ɑː ~ ɪ ~ ə
• e ~ ɪ,ʏ ~ e
• eː ~ e(ː) ~ jə(ː) (I'm not yet sure what conditions the lengths here - note that B eː ~ C jəː isn't attested)
• ɛː ~ e ~ (j)ə
• eː ~ ɪː ~ e(ː)
• ɛ ~ ɛ,e ~ ɛ
• ɛː ~ ø ~ ɛ
• ɛ(ː) ~ ɤ,o ~ u
• ɘ ~ ə ~ ʉ
• i ~ i ~ i ~ *i
• o(ː) ~ u ~ o
• ɔ(ː) ~ ɔ,o ~ u
• u ~ ɪ ~ Ø
• uː ~ uː ~ uː
• uː,ʉ ~ ɨ ~ ʉ
• ʉ(ː) ~ ʉ ~ ʉ

The correspondence æ ~ ɪː ~ æ I feel sure must be a conditional development of regular æ ~ ä ~ æ, due to the palatalisation of the following consonant (gj ~ dʒ ~ k’j). Then there's the set of ʔæglɐɪ ~ ɪːglɔːɪ ~ k’nək’ləːn. The A and B forms probably have some kind of anaptyptic vowel, itself causing the vowel corresponding to C ə to syncopate. B then lengthened its ɪ when the *n dropped out. A similar scenario accounts for ʔugaːgʉ ~ ɪlgɑːgɨ ~ k’ləgäːk’ʉ.

[dhok]

I don't think that there is sufficient analysis to group A,B,C into subfamilies - we need to figure out the order of changes first and which if any could be areal

This has not stopped Sino-Tibetanists, Africanists and some Americanists from doing exactly this. It's clear that A is just plain weird--even if it groups more closely with one of B or C than either of them do to each other, it has undergone significant phonological erosion and what appears to be syntactic change. A is simply harder raw material to work with and fit into the larger picture. You can think of my tentative grouping of B and C as a provisional subfamily, a data set carved out of the whole to be easier to work with. It's true that when I wrote "it is obvious" I had just started the post. I would walk that back and say that B and C are easier to work with than A is.

Suppose you got a real-world dataset composed of English, German and Icelandic sentences. Would it be obvious that English and German are the subfamily? I suspect it wouldn't be, and I think most people would want to start with German and Icelandic.

(Or French+Spanish vs. Italian; Bulgarian+Serbo-Croatian vs. Polish; Cantonese+Mandarin vs. Min...there are a lot of real-world triplets that work like this, where family ABC splits diachronically into A+B vs. C but one of A or B is unusually innovative in ways that make it difficult to fit into the picture.)

[KathTheDragon]

Completely irrelevant for reconstruction, but it seems that A vɛːrɑː and vɛːvovræ are the lexical equivalents of BC mäʃ ~ mætʃ and mäʃulɪg ~ mætʃləg respectively.

[KathTheDragon]

- However, the pair seg~ʃək in 11 suggests we are dealing with a chain shift *ʃ > s, *tʃ > ʃ in B

IMO a better explanation of this is that C actually palatalised *s - the vowel correspondences A ɛː ~ B e suggests a pre-C *jə (cf. ɛː ~ eb ~ jəp, ʔɛː ~ eg ~ jək). A similar case of palatalisation are dɛː ~ de ~ dzə (ðɛːɾ ~ ðeɾ ~ jəɾ is ambiguous)

[dhok]

- However, the pair seg~ʃək in 11 suggests we are dealing with a chain shift *ʃ > s, *tʃ > ʃ in B

IMO a better explanation of this is that C actually palatalised *s - the vowel correspondences A ɛː ~ B e suggests a pre-C *jə (cf. ɛː ~ eb ~ jəp, ʔɛː ~ eg ~ jək). A similar case of palatalisation are dɛː ~ de ~ dzə (ðɛːɾ ~ ðeɾ ~ jəɾ is ambiguous)

Aha! I read the B e ~ C jə correspondences more as suggesting original *e unpacking in C, along the lines of Tocharian. ...of course, if that's what happened then you're probably correct: *sek > sjək > ʃək.

[KathTheDragon]

Tocharian is actually a very good analogy, as *e > *ʲə > *ə with palatalisation of the preceding consonant, or *jə initially.

[KathTheDragon]

Where did C ə:s go in B?

Based on the vowel correspondence alone, I wonder if it could actually be the cognate of B ɔɪdɪː. -ɪː seems to be a suffix in B, and based on B ɣɪtɪz ~ C ɣes, I suspect that C underwent a simplification *ts > s at some stage (for B -ɪz ~ C -s after a voiceless stop, see βəgɪz ~ wʉks). So we could have *əɪts for proto-BC.

Edit: here's an interesting parallel: A aː ~ B æɪðɪː ~ C hæɪdz

[KathTheDragon]

@Salmoneus: is bɘ in 11A an error for βɘ?

[KathTheDragon]

Regarding subgrouping, it seems to me that there isn't a simple phylogeny at work here.