The Ridiculous Effectiveness of the Comparative Method
A Case from Daatsʼíin
I started poking around with Daatsʼíin upon finding yesterday a 100-word Swadesh list comparing it with two dialects of Gumuz, and wanting to see how much of a Proto-Gumuz could be wrung out of the 89 matches. A decent bit turns out to be indeed doable.
With data running this low though, it will be inevitable that there will be sound correspondences that are plausible but cannot be proven regular. E.g. Daatsʼíin /ɗa-/ ‘to go’ corresponds to Southern Gumuz /ɗá-/, Northern Gumuz /tsá-/; and D /áɗa/ ‘I’ corresponds to SG /ára/, NG /áɗa/. In both cases we have Daatsʼíin /ɗ/ corresponding to a /ɗ/ also somewhere in Gumuz proper, and thus probably these sound correspondences both go back to Proto-Gumuz(ic) *ɗ. But without further data I cannot know if initial fortition to /ts/ or medial lenition to /r/ should be considered regular or irregular. (In fact I don’t even know if limitation to initial / medial position is an actual rule of conditioning, though from general typology of sound change I can assume that *ɗ > r should be more probable medially, *ɗ > ts more probable initially.)
The data still suffices to demonstrate also some non-trivial regular sound correspondences. One clear case is that velar consonant + w + a in Gumuz proper corresponds to plain velar + rounded vowel in Daatsʼíin, with at least five good examples and one more dubious:
D /bekʼo/ ~ G /béékʼwa/ ‘hair’
D /ɗakʼu/ ~ NG /dúkʼwa/ ‘smoke’
D /kʼófakʼu/ ~ SG /kʼófagwa/ ~ NG /kʼwáʃákwa/ ‘navel’
D /kʼókéé/ ~ SG /kʼoca/ ~ NG /kʼwaca/ ‘eye’
D /mágúŋkú/ ~ SG /magókwa/ ~ NG /magáákwa/ ‘night’
? D /voko/ ~ G /ʒákwá/ ‘bone’ (no parallels for /vo/ ~ /ʒa/)
Which side is more original however? No way to tell off the cuff: both Proto-Gumuz *Kwa > Daatsʼíin /Ko ~ Ku/ (seemingly /o/ in 1st syllables, /u/ in 3rd syllables; both options attested in the 2nd syllable) and PG *Ko > Gumuz proper /Kwa/ would be perfectly reasonable sound changes. One option would be to try to dig out data from the more distantly related Koman languages to see what they might point to. Another would be to note that also Southern Gumuz has /kʼo-/ in ‘navel’ and ‘eye’, which could suggest that /Ko/ is an archaism and /Kwa/ is an innovation. However, another argument still points at the opposite: Gumuz proper also allows /Ko/ sequences. Only two of them seem to find Swadesh list cognates in Daatsʼíin, both telling in their own ways however:
D /kʼaw/ ~ G /kʼóá/ ‘dog’
D /kʼôs/ ~ G /kʼósa/ ‘tooth’
In the first we have a word-final labiovelar /w/ in D. If the PG form was something like *kʼawá, the word-medial *w could probably have conditioned preceding *a > o in G, giving some degree of independent confirmation for spreading of rounding from consonants to vowels. The second then should have become G ˣ/kʼwása/ if the /Ko/ ~ /Kwa/ correspondence came from original *Ko. Again, due to lack of data we cannot show either of these correspondences to be regular. But at least they already suffice to show that the hypothesis PG *Kʼwa > D /Ko/ is better than the opposite.
---
Anyway, today I follow up by reading the actual grammar of Daatsʼíin, as linked in my previous post. Guess what jumps at me around noun inflection? Several nouns are pluralized with a prefix of the shape /Cáá-/, with the initial consonant reduplicated from the noun. There are some complications however…
So why is ‘guests’ /kwáákodar/ rather than /káákodar/? Clearly because, at one time, the word-initial consonant to be reduplicated was not /k/ but *kw! That is, /kodar/ : /kwáá-kodar/ comes from earlier *kwadar : *kwáá-kwadar, and *kwa then develops into /ko/, exactly as I have already predicted on the basis of just a handful of Daatsʼíin–Gumuz comparative data.
This is what I would call “ridiculously effective”: any sound law supported by just half a dozen good etymologies and not contradicted by other data is already going have pretty good odds of being correct. Quantity matters some, but in practice, going from 3 examples to 6 examples is already a much bigger improvement than going from 6 to 60 (or 60 to 600).
(As for why the plural prefix remains /kwáá-/ instead of also turning into ˣ/kóó-/ or ˣ/kóá-/ — per Ahland’s closer description in the grammar, what she transcribes as long /aa/ and short /a/ are phonetically realized as [a(ː)] and [ə] respectively. So “*Kwa > *Ko” is phonetically really rather *Kw[ə] > *Ko, and actual open [a] would seem to be not affected.)
