"Sound–meaning association biases"

Discussion of natural languages, or language in general.
User avatar
Tropylium
Avisaru
Avisaru
Posts: 512
Joined: Sun Aug 07, 2005 1:13 pm
Location: Halfway to Hyperborea

Re: "Sound–meaning association biases"

Post by Tropylium »

As far as pure statistics go, this is a pretty good study.

On the other hand, if we really wanted to show that there's some kind of an onomatopoetic motivation for given words, we really should be looking at historical innovation events, not just at whatever currently exists. When we're analyzing basic vocabulary of this kind? The default explanation is that the words have whatever type of phonetic shape they have because its source language (usually a parent, sometimes a loangiver) had a word of shape *X, followed by sound changes Y. Nothing about this stops working when we're dealing with timescales too deep to reconstruct full family-size relationships — it's entirely possible that some correlations that come up simply indicate deep etymological relationships in basic vocabulary. We already know that a few cases are just this; e.g. Uralic *śarwə 'horn' (Finnish sarvi, Hungarian szarv, etc.) is an Indo-European loanword, and therefore not a fully independent piece of evidence from horn etc. Presumably some other, thus far unidentified cases are lurking in the data as well. I'm not convinced they're ruled out entirely this as a source of their results.

(Even though we could maybe take the hard-line position that all language families not shown to be related are indeed entirely unrelated and go back to individual language evolution events, it will be still unreasonable to suggest that their entire lexicons would have been closed books for the dozens of millennia that language has been around. If Ket can have still-identifiable cognates with Navajo, or Malagasy with Cham, then over 50k years and allowing also loaning — sky's the limit.)

But we can also dig up known events of word replacement. Consider e.g. how English replaces hound by dog. The existence of /h/ in the former is due to Grimm's Law, and /u/ due to Germanic zero-grade epenthesis, and they are therefore clearly not due to onomatopoetic motivation; however, we do gain here some (miniscule)amount of evidence that languages might prefer their word for "dog" to have /g/ and /o/ rather than /h/, /u/ and /n/ in it.

I'd also perhaps want to start this kind of an analysis from some segment of vocabulary where we know that innovations happen a bit more often. That would allow getting some basic grasp for how often "onomatopoetic drift" happens. There is still a lot of vocabulary that is "basic" in the sense of having stable semantics, even though it might be far from Swadesh-level stable etymologically. (At the extreme, there are Wanderwörter like tea and potato, but these of course would run into the problem of too few independent innovation events.)
[ˌʔaɪsəˈpʰɻ̊ʷoʊpɪɫ ˈʔæɫkəɦɔɫ]

Post Reply