The points of interest in my reconstruction are the reflex of
*ñh /ŋ̊/, the evolution of the
*ṭr /t̪ɾ/ consonant cluster and the developments underwent by the
*-àgi- /ɤgi/ sequence.
The main isogloss is the reflex of
*ñh /ŋ̊/ which divides the language family into two branches: the
ŋ-branch that includes Language 2 and Language 3, and the
x-branch which is larger and comprises the rest of the family.
The reflex of /x/ is divergent in the x-branch: Language 6 is the only language that preserves the original reflex, whereas original /x/ has weakened to /h/ in Language 7 and Language 8. Other languages have done away with /x/ entirely. Among those languages, Languages 1, 5 and 10 have lost the word-initial minor syllable, but it is most likely that the three languages don't constitute a separate phylum; while Language 1 is closely related to Language 5, Language 10 is considerably divergent.
There are two separate processes that took part in the evolution of the
*ṭr /t̪ɾ/ consonant cluster.
One process in question is fortition, which is attested in Language 7 and Language 8 where
*ṭr is attested as /ttr/. Overall, these two languages are the most conservative ones in the entire language family.
In other languages, the cluster underwent lenition to /θɾ/, which in turn has a stunning variety of reflexes across the language family.
Loss of /ɾ/ is a very common occurrence as it is attested in a majority of this language family, six languages in total. It is attested both in the ŋ-branch and in the x-branch.
In the ŋ-branch, the cluster simplified to /ð/. This is preserved in Language 3, whereas the original /ð/ changed to /ɣ/ and then /x/ in Language 2.
The x-branch can be divided along the lines of whether it was fortition or lenition what took place.
The branch where fortition took place (the
F-branch) is a closely-knit one, consisting of Language 7 and Language 8. Besides fortition, similarities between these two languages are preservation of the word-initial minor syllable and collapse of the remaining two syllables into one. The reconstructed ancestor of the F-branch might have been something like
*hatrɤn.
This branch is firmly within the x-branch and its distinctive feature is the change of Proto-X
*x to /h/.
The branch where lenition took place (the
L-branch) is a much wider one, with many smaller sub-groupings inside the branch.
Languages 4 and 6 display affinities to the fortition branch as they both preserve the minor syllable, /θ/ and one of /ɾ n/ while losing
-g- in some way. These two languages aren't likely to be mutually related though.
In Language 4, word-final /n/ lengthened the vowel in front of it; Language 6 displays loss of /ɾ/ (possibly via /w/, explaining the rounding on the second vowel) and a probable stress shift towards the first syllable.
Languages 1 and 5 are closely related and constitute a subgrouping (the
Fr-branch) within the L-branch. Similarities between those two languages include loss of the minor syllable, change of /θ/ to /f/, and preservation of /ɾ/ and /n/. In Language 1, intervocalic /g/ lenited to /j/ or an equivalent as the reflex of the
*-àgi- /ɤgi/ sequence is /oi/. In language 5,
*à /ɤ/ underwent raising to /ɨ/ and later to /i/ while /i/ was reduced to /ʌ/ in the final syllable.
Considering the changes affecting the
*-àgi- /ɤgi/ sequence, stress is presumed to have been penultimate in the ancestor of this branch. I've hence reconstructed the Proto-Fr word as
*fɾɤgin, with stress located in the penultimate syllable.
In Language 10 both the minor syllable and /r/ were lost, whereas the reflex of Proto-L */θ/ is /ð/. It is unclear whether this is a result of voicing assimilation with /ɾ/ or of intervocalic voicing of /θ/ after loss of /ɾ/. It is likely that
*r /ɾ/ changed to /w/ at some point before it was lost and that it helped
*à /ɤ/ become a rounded vowel.
The evolution of the
*-àgi- /ɤgi/ sequence in Language 10 might've been as follows: /ɤgi/ → /ɤɰi/ → /ɤːj/ → /oː/, with /ɤːjn/ palatalizing to /ɤːɲ/.
Languages 9 and 11 (the
Z-branch) are the other subgrouping within the L-branch. They are the most divergent languages in the L-branch and in the entire family whatsoever. Similarities between those languages are losses of minor syllable and /ɾ/, loss of /θ/, fronting of /ɤ/ and palatalization of original
*g /g/ before /i/ and emergence of nasal vowels.
Losses of /ɾ/ and /θ/ amount to a total disappearance of the
*ṭr /t̪ɾ/ cluster. This cluster was lost after the loss of the minor syllable, and the first part of the loss was probably elision of /θ/. The rhotic /ɾ/ probably changed to /w/ before being lost altogether according to evidence from Language 9 where the reflex of
*rà /ɾɤ/ is /y/ (via **/we/)
Regarding Language 11, the presence of word-initial /ɛ/ might be explained in two ways. The first explanation is that after /ɤ/ was fronted, it fused with /w/ into something like /ø ~ œ/ which later unrounded. The other explanation is that /ɾ/ or /w/ was lost without a trace. A possible proto-form for the Z-branch is one of
*øʑĩ,
*weʑĩ or
*ɾeʑĩ depending on the explanation chosen.
All this data can be reconciled into a single proto-form for the entire L-branch, which I reconstruct as
*xaθɾɤgin.
All L-branch languages other than the Fr-subgrouping and Language 4 display loss of /ɾ/ with /w/ as a possible intermediate form and a possible relationship cannot be discounted.
When reconstructing Proto-X it is best to consider /x/ and retention of /g/ as a conservative feature, which leads us to the conclusion that the ancestor of Proto-X is something like
*xatɾɤgin.
The ŋ-branch is yet another closely-knit subgrouping. Similarities include loss of /ɾ/, prevalence of back vowels, and loss of the final syllable.
Before the final syllable was lost, intervocalic /g/ has most probably already lenited to /ɣ/. In Language 1 it underwent palatalization in front of /i/ and it is attested as /ʒ/, whereas it became a guttural /ʁ/ in Language 3.
This branch also displays lenition of
*ṭr /t̪ɾ/ to /θr/. This cluster probably changed to /ð/ via voicing assimilation to /ɾ/ as there is no evidence of /ɾ/-loss by means of /w/ (no vowel rounding). In Language 2, /ð/ has backed to /ɣ/ presumably after /ɣ/ was palatalized. It lost voicing which results in the Language 2's /x/.
The vowel in the minor syllable is the most difficult to explain. A likely development is raising of original
*a /a/ to a mid to high vowel which assimilated in backness in Language 2 and spontaneously rounded in Language 3.
A probable ancestor of the Language 2 and Language 3 words is
*ŋəðɤɣɪ, with stress on the penultimate syllable.
Considering the developments in the x- and ŋ-branches, I have reconstructed the proto-word as
*ñhaṭràgin /ŋ̊at̪ɾɤgin/.
A tree of this language family according to my view is displayed below.