#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).

How Do We Know How Languages Are Related?

Quantitative social scientists have something of a catechism when it comes to providing evidence for causal assertions. If we want to assert that A is a contributing cause to B (for example, living in a neighborhood with many sub-standard housing units is a cause of higher rates of delinquency), we need to conduct a study involving a reasonably large number of cases and then assess whether cases…

Ian Morris has written a pair of books that are intended to contribute to a particularly important set of disagreements in comparative economic history: what accounts for the advantage in economic development that seems to be enjoyed by Western Europe at various points in history? The key arguments are presented in  Why the West Rules–for Now: The Patterns of History, and What They Reveal About…

The open texture of the social world

What is involved in arriving at scientific knowledge about the social world? The position I have consistently taken emphasizes contingency and heterogeneity of the social: the social world is a mixture of diverse processes and structures; it is constituted by socially constituted and socially situated actors, leading to ineliminable features of contingency and heterogeneity; and that there are no…

First on the comparative method, external reconstruction and I’ll add internal reconstruction as well: in the Slovene tradition (at least; it should be similar to the German tradition, I guess), I have often heard it being phrased as “the comparative method with the method of reconstruction” – implying reconstruction itself is separate from the comparison alone. In that view, then, external reconstruction is not the same as the comparative method. However, as I’ve seen it used in English (and also elsewhere in Slovene), the term comparative method comprises both the process of comparison and the process of reconstruction, so when used that way it can be equated with external reconstruction. I’ve further also seen it used (even) more generally as comprising all of external reconstruction, internal reconstruction, analysis of borrowings (using substitutional phonology on different diachonic stages), typology ... that is, everything that comparative linguistics is, basically.

I pointed out internal reconstruction in the beginning because I would like to note here that I’ve also seen it used in the sense of a more limited application of external reconstruction. Internal reconstruction is commonly understood as an analysis of a single system (language/dialect) based on the variation the system exhibits within itself (e.g. the alternation in otrok/otroci/otroki can tell you something about the origin of c and also that the two word-final i’s are likely not the same in origin (of course then that has to be based on a wider observance of similar patterns within the system, not just one single example)), but in the case I wanted to point out it was used to mean external reconstruction within the scope of Slavic languages (that is only an example of course, it could be any other (more narrow) language family) as opposed to external reconstruction as in using comparative data from outside Slavic, that is, Indo-European (again, it could be any other (wider) language family). This is not the usual use of the term, though, but it seems prudent to mention it here if you ever encounter it anywhere else.