Well, I'm bored. So let's give it a shot too. First just some thoughts I get from staring at the map:
First of all, I am not convinced the final is a cluster. It would certainly be convenient, readily explaining the alteration between the scattered coronals in the west and the more persistent velars in the east. Treating /tS/ as a single consonant, there is not a single form showing any hints of a coda cluster. I'd reckon based on the existence of both coronal and dorsal outcomes that the original coda was palatal, perhaps *c or *kʲ
Almost exceptionless, the form of the word is CVCVC. In some cases it's CVC:VC, never more, and any deviations can be quite easily shown to be derived from one of these forms.
Something curious is going on with rounding: There are three distinct instances of rounding found on the map:
Red: Final VC rounded
Green: Initial V rounded (remains front)
Blue: Initial V rounded and backened
Blue is in an area in which the initial V tends to be /a/, in which case a change of the sorts of a→ɑ→ɒ(→ɔ) is not unlikely. The other two demand an explanation though. The simplest explanation would be that the medial C was actually rounded - a possibility, but there's no further data to back it up. Alternatively, there could've been some other rounded segment in the language. The rounding also appears to corellate with the existence of /S, tS/ rather than expected /s, ts/ in that in the eastern two rounded branches, those rounded vowels are always adjacent to an esh. Another possibility is that the roundedness of one of the vowels was there to begin with.
It seems quite obvious that the second V was nasalized.
This is a map of all vowel isoglosses:
However, this doesn't seem very useful. Instead, here are two more maps:
In the west, this is very straightforward. In the east there's a bit of a mess (summarized as "not a"), but generally speaking the eastern vowels are [+front][+high] compared to the western ones.
Much more straightforward. All vowels appear to be either /ɛ̃/ or simple derivations of it (lowered in blue, raised in red, rounded in purple)
Next, I decided to see if there are any subgroups with trivial reconstructions and found that I could find lots of pairs. I assume this is due to how the map was constructed (presumable along the lines of a binary branching tree, perhaps with some overlaid isoglosses):
Some of the reconstructions aren't exactly amazing (and I've put a question mark before the least certain ones). In particular, the far east one is odd as it has a dorsal-coronal alteration in an otherwise very clearly related form. That protoform may also have had a development along the lines of *k→kʲ→c→cç→tʃ, initially motivated by the preceding front vowel. In fact, after thinking about this while typing it out, I now prefer that solution. The other questionmark is about the reconstructed *tɬ yielding t and ɬ. This form probably derives from an older *tʃ, which would be reasonable considering the surrounding forms.
I'll now see if I can get anywhere by working with the new "reduced" dataset. First, the forms, from west to east. I'll omit the nasalization marker as combining diacritics and monospace fonts don't get along and I'm absolutely certain this is an original feature of the proto-form. With the exception of (3), where the final vowel has devoiced, all final vowels are nasalized in the reconstructed forms. (8) is not a reconstruction.
Code: Select all
Forms 1-5 seem very closely related. All share the onset tʰ; 2-5 have an initial vowel that can be traced back to /a/; all reflect a medial consonant /s/. 1-2 have a final vowel /a/, the rest /ɛ/. Additionally, among 1-2, three synchronic forms show a debuccalized coda, the last has a voiceless n, which I first assumed to be the older form, but now think might also be derived from h, "absorbing" the nasality of the preceding vowel.
It is quite clear that 3-5 form a coherent group. The placement of 1 and 2 is more problematic - they may form their own subgroup, with the initial vowel in 1 being an innovation, or 2 might be closer to the other group, with 1 being the outlier. I assume the former is the case, leading me to the following reconstruction (now assuming (1) to end in /h/ instead, but ultimately it wouldn't matter anyway for the final step of the reconstruction):
Then taking data from outside this group into account, the V is likely to be the /ɛ̃/ of the protoform. The S is some voiceless fricative - it may be anything from ʃ~ɕ~ç, but I doubt that it's retroflex due to the lack of retroflexes anywhere else.
(6) can fit into this reconstruction without requiring any adjustment to the proto-form (the final consonant, whatever it was, drops, leaving creaky voice; the aspiration turns into affrication), but it could just as well be connected to another neighboring form instead (which at least superficially it looks like it does - I will form a conclusion later).
7-8 also definitely form a coherent group. I reconstructed (7) as tʃɶsɛk, (8) is a synchronic word ʃasɛk. *tʃasɛ̃k strikes me as the most likely reconstruction here, with the rounding being an innovation of 7. I don't like just passing off a feature as an "innovation", but I don't see it as a likely feature of the protolang. (6) again neatly fits into this reconstruction, requiring merely tʃ→ts and Vk→V[+creaky]. Basically, I'm just going to ignore (6). It isn't relevant to the reconstruction either way.
The two forms I've reconstructed thus far (*tʰasɛ̃S, *tʃasɛ̃k) may well be a coherent group. I don't know what the initial consonant of the proto-form would be, but the coda is probably *k, which may have palatalized in 1-5 (k→c→ç) under influence of the front vowel. This would give a Proto-(1-8) of *Tasɛ̃k. Whether 1-8 are actually a complete subbranch remains to be seen though.
(9) is a bit weird. It completely dropped its initial vowel and has a higher second vowel than most other forms, a feature shared only with neighboring (10). However, unlike (10), the coda is reflected as t. There are two possible explanations here: One or both forms of (9) is very divergent (loglorn's hypothesis) or the -t is actually a reflex of earlier *k. This may seem a bit unlikely at first, but consider the following: Coronals are acoustically more neutral; a change velar->alveolar is weird from an articulatory standpoint, but not unreasonable acoustically, especially next to a front vowel - and on top of that word-finally, where plosives are often harder to distinguish anyway. It may also be motivated by the coronal in the onset. I will for now assume that k→t happened. Under this assumption, 9-10 readily form a group. The lateral in 10 poses a bit of a problem - I assume it comes from an earlier tʰ, which is a bit odd but I can accept it (the alternative is assuming it's part of the proto-form and perhaps derives from an even earlier tʃ. I will keep this possibility in mind but not use it for now). To recall, the forms we're working with are tsːẽt and tɬIsːẽk. These then readily provide a proto-form in *tʰIsːẽk.
This leaves us with two forms, (11) təʃːɔkʷ and (12) təSːɛk. Again, the rounding wants an explanation, and here I'm much more inclined to say that it might've been original in some way, as it comes with backing and all. Of course, the development could be similar to (5) instead, but it just seems a bit out of place, and in both words in (11) the rounding seems to be the primary feature, spreading over the whole VC segment, while back in (5) it looked much more like some rounding caused by backing. I assume therefore that the final consonant carried rounding, which was lost in 12 but spread to the vowel in 11. This leads us with *təʃːɛ̃kʷ for Proto-(11-12)
Those are then the three major groups: 1-8 with *Tasɛ̃k; 9-10 with *tʰIsːẽk and 11-12 with *təʃːɛ̃kʷ.
Bringing it together:
The first consonant clearly had some form of aspiration or affication. Seing as we find *tʰ in two branches (turns to tʃ in one major subbranch) I posit that as the proto-form
The first vowel could be many things *a(→ə(→ɪ)) is a possibility, but I don't like it. *e with changes in all subbranches is a possibility too. My favourite is *ə though, which lowered (perhaps due to a stress change onto the first syllable) in 1-8 and got retained in the others, with 9 then dropping it and 10 raising it.
The middle consonant is definitely a geminate. I'm going to assume it was sː, with a palatalization in 11-12.
The final vowel I've long established as "very likely to be ɛ̃" and my proto-forms agree with that sentiment.
The coda looks from this data set to be kʷ, but this is all dependent on a correct grouping and reconstruction of 11-12, otherwise it's probably k.
Thus my guess: *tʰəsːɛ̃kʷ
Edit: While I wrote this, three new comments appeared. Looking at it, I appear to have had mostly the same conclusions as ixals, albeit a slightly different grouping in the eastern parts.
Edit: Also damn, this got a bit rambl-y