In the last post I made, I gave some of my thoughts on Quentin Atkinson’s Science paper earlier this year, where he demonstrates that evidence of human migration patterns is still available in the world’s languages if you squint right. His mechanism works on something called the founder effect, which, in terms of evolutionary genetics, provides that new, small offshoot populations will exhibit less genetic diversity than large populations. He assumes the same thing applies to language, and I disagreed. Instead I argued that mechanisms of language contact (which have been observed reducing phoneme counts) would better account for his results, with some possibility that Homo sapiens may have had simplifying linguistic contact with other hominids.
Wellll… it turns out I might have been wrong on both counts. Let’s look at hominids first, since that is easy to tackle. New evidence presented in Science this week (or here for a summary) does indeed confirm that Homo sapiens had repeated contact with other hominids that led to our interbreeding. We are able to surmise this from traces of their genes left in our bone marrow — the evidence providing that we gained defenses against local diseases from the hominids that preceded us in those areas. Although this seems wonderful evidence to back up my wild speculations, a hole in the logic appears — first, these breedings didn’t occur in the Americas, where Atkinson nevertheless finds the continued decline in phoneme inventory size. Secondly, it appears that this research also concludes that the huge amount of genetic diversity found in the African populations that chose to stay in our ancestral birthplace (higher than those that left) is due to continued breeding with other hominid lines that likewise never left the African sub-continent. Thus, the greatest amount of hominid-induced language change should have been in Africa, which is the opposite of what Atkinson found.
The second point on which I was mistaken was in discrediting the founder effect. Though it would seem to work under a different mechanism than the genetic variant, Hay and Bauer (2006) provides a strong correlation between small population sizes and small phoneme inventories. They are only speculative as to the reason for the correlation, but suppose that in smaller populations, two effects are likely: One, higher amounts of variance have been tied with better learning of phoneme distinctions. That is to say, if you hear a lot of people making a distinction between “w-” and “wh-” (witch vs. which, in some dialects), then you will find it easier to make that distinction yourself. The second effect is that in small populations, there is a lot of common background, and it will be less necessary to make as many minute distinctions in order to communicate, leading to loss of some sounds. In any case, the result is that small populations are tied to small inventories, so offshoot populations, which are going to tend towards small, mobile factions, will tend toward simpler sets of speech sounds.
Well, that’s okay. My conclusions, when all’s said and done, were that it seemed more reasonable for speech sounds to become more complex over time as populations age. This, at least, may still be true. McWhorter (1998) points out that very few creole languages make use of lexical tone (i.e. – are not ‘tone languages’ like Cantonese), which is surprising, given that many of the languages that contributed to their creation did make use of lexical tone. If this does hold for historic contacts in general, then we should see evidence of this in population movements, too, where newer languages are more prone to not use lexical tone. (After all, part of Atkinson’s counts were that tonal distinctions counted as separate phonemes. (That is, an “o” with high tone is different from an “o” with low tone.)) This may somewhat be true. Let’s look at the World Atlas of Language Structures chapter on Tone. A peek at the map shows very heavy use of complex tone systems in Africa and Asia, lighter use in North America and even less in South America. Indonesia and Australia are particularly devoid, although there is a fair amount in Papua. Neat. And then there’s Europe. Europe shows almost no use of tone. Why no use in Europe and places like India, which should show some use near that of Africa and Asia? Again, if tone develops as languages age, we should take it as a probabilistic process. Languages may attain tone at a certain level of maturity, but they don’t have to. This may explain why Europe doesn’t see use of lexical tone. In any regard, the overall trend looks about right (without really putting any numbers to it), which would suggest that larger inventories do indeed come with age (as opposed to a large inventory having been established first, and then progressively whittled down).
Cool. So — wrong on two counts, possibly right on one. Not too bad, I guess. Anyway, such is science… such is science.