I just noted in the previous post that some internal reconstruction of the structure of Koman roots might be a good idea, e.g. for reducing the large stop inventory of maximally five series /p pʰ pʼ b ɓ/ (broadly retained in Uduk, Dana and Opo; aspirates lost in Koma, both aspirates and implosives lost in Gwama). However, considering for a moment also the Gumuz side, another noticable difference is not even the diversity of stops but their frequency. Proto-Koman has no shortage of stop-final roots, e.g. the smallish set of etymologies found in all five languages includes words like *bɪncʼ ‘fishhook’, *dak- ‘to finish’, *(h)ɔpʼ- ‘to sip’, *kʼʊ́p ‘head’, *kʼu₂t̪ʼ- ‘to cough’, *pʰàd̪- ‘to fly’, *sʼi₂´k ‘rat’, *sɪ`t̪ʼ- ‘to be far’, *ʃukʼ- ‘to wake someone’, *t̪ʰáɓ- ‘to kick’, *t̪ɪ´t̪- ‘to roughen stone’, *úpʰ- ‘to bathe’. This does not look to be the case in Gumuz. While we do not have a proper proto-Gumuz reconstruction to consider in full, Ahland’s grammar presents a list of 52 narrow Gumuz verb stems in appendix C (not a complete list of even the data appearing in the grammar, but an informative initial corpus anyway). Rather few of these end in voiceless stops or affricates: only √(ɓ)átʃ- ‘to hit, kick’ √faat- ‘to fall’, √ook- ‘to heat’. If following my hypothesis that *p > f, there’d be also √ʔa/ef- ‘to wash’. A decent number end in -b- (√ɗáb- ‘to find’, √ɗamb- ‘to try, taste’, √tab- ‘to be thick’, √tib- ‘to kick’) and a few in a voiced velar (√dugw- ‘to run (liquid)’, √fâg- ‘to urinate’), contrasting with a complete absense of -d- though. Among the implosives we have the inverse: no cases of -ɓ-, but three of -ɗ- (√káɗ- ‘to finish, run out’, √koɗ- ‘to skin, strip’, NG √wíɗ- ~ SG √jír-) ‘to see/check’). I suspect that the situation results from a pre-Gumuz loss of many stem-final stops plus new stem-final consonants fossilizing from morphology, yielding an uneven distribution of segments. I don’t currently want to take a stance on how precisely this would have happened though — plausibly this could be cluster simplification *C₁C₂ > *C₂, or medial lenitions, or loss in absolute word-final positions, or even some of each.
For the -b- and -ɗ- groups here I do not have firm proposals (very speculatively, maybe ‘to finish’ and ‘to check’ could contain ɗá- ‘to go’?); but good candidates for this process are provided by the system of incorporated body part terms. They seem to have grammaticalized into a wide range of functions, ranging from formation of locative prepositions / prepositional phrases, to a system of noun classifier agreement on verbs, and the discussion of all of these takes up a large proportion of Ahland’s grammar. The most general-purpose among these would seem to be -kʼwá ‘head > top’; -cá ‘eyes > front’; -sa ‘mouth > opening’; -ʃa ‘hip > base, bottom’; -tsa ‘*body > instrumental’ [1], all with also -aC allomorphs, such that in their grammaticalized functions it is essentially only the consonant (maybe also a floating tone for ‘head’ and ‘eye’) that really carries the meaning. Regular “object-incorporated” verbs maintain these as independent moving parts of the verb construction, such that person marking will follow the stem but precede the “object”, e.g. ʃá- ‘die’ → ʃá-kʼw- ‘kill’ with a 2PL imperative ʃá-cá-kʼw ‘kill them!’. However, also the majority of the basic verb stems in Ahland’s appendix that end in the consonants /kʼ/, /s/ and especially /ʃ/ show semantics compatible with the corresponding incorporated object suffixes:
- korak’- ‘to peel’ (= ‘to peel the top, the surface’), evidently derived from √koɗ- ‘to skin, strip’, given that r is just the medial allophone of /ɗ/.
- (√takʼ- ‘to spit’; I see no grounds to posit division as **ta-kʼ-)
- √gis- ‘to grill’ (= ‘to cook for eating’?)
- √cʼeʃ- ‘to cut’ (= ‘to cut down, thru’?)
- √gaaʃ- (< *garʃ-) ‘to grind’ (= ‘grind down’? as I’ve already hypothesized in the previous post)
- √kʼoʃ- ‘to penetrate’ (= ‘pierce to the bottom, thru’?)
- √ńʃ- (SG also ŋáʃ-) ‘to soak’ (= ‘to become wet to the bottom, thru?’), and perhaps akin to √ŋar- ‘to take, bring’
…so there seems to be good odds that at least some of these derive from shorter verb roots with an incorporated object marker fossilized to the end of the stem. In turn, any *CV roots uncovered this way could prove to correspond to *CVT or similar roots in Koman. So e.g. considering again my previously listed comparison of Koman *kʼi´sʼ- ~ (South) Gumuz cʼeʃ- ‘to cut’, if the latter is rather reduced from pre-Gumuz *kʼʲeC-ʃ-, this would then tell us nothing about what is the default Gumuz equivalent of Koman *sʼ.
If per hypothesis all sorts of pre-Gumuz root shapes like *CVp-, *CVt-, *CVsʼ- could end up reduced to just *CV- (this also would be in principle explorable even without insisting on the relationship with Koman, e.g. thru loanword studies), obviously this also increases the odds of chance correspondences. At most this could be slightly mitigated if loss of stem-final consonants maybe proves to have left some effects on tone; or perhaps yields some instances of long vowels; or has left some morphological doublets around. But it’s also a good reminder that my comparisons remain very tentative and we should keep also waiting for fuller Gumuz data.
[1] Not actually attested as an independent noun, and Ahland’s argument for treating it as specifically a grammaticalized body part rather than in origin a more abstract noun is not very clear to me (is there an assumption here that the other incorporated objects being body parts suggests that this must’ve been too?). I would propose in the first place instead ‘self’, especially in light of examples like ka-tsá-má ‘by him/herself’ (ka- comitative ‘with’, -má 3SG possessive) or fáɗ- ‘to rise’ → fád-(á)ts- ‘to get up’ (‘to rise by oneself’). The freestanding word for ‘body’ is the clearly unrelated ɓaga (also ‘person’).
Freelance reconstruction 
Leave a comment