UPDATE: Technical aspects of the two papers are discussed at length at Language Log (here for the Dunn et al. paper, and here for the Atkinson paper). My post below takes the results at face value and discusses implications, in particular how the results of the two papers seem to interact. I am not competent to discuss the solidity of the results themselves, so what I'm doing here is what I think philosophers do best: unpacking consequences of claims and theses.
________________________________________________________________________________
This has been a lively week in the blogosphere, with at least three major ‘hot topics’: a politician making statements which are not intended to be factual statements (and a comedian having a ball making statements not intended to be factual statements); an ad for a clothing brand featuring a boy having his toenails painted pink by his mother; and the publication of two computational linguistics papers with remarkable, potentially groundbreaking results.
This blog has already covered the politician and the comedian, and my plan is to cover the other two topics in two different posts (there must be some interesting connections between boys with pink toenails and anti-Chomskian claims against the universality of language, but I won’t go as far as that…). So let me start with the two computational linguistics articles, which both used techniques developed for computational modeling in evolutionary biology to investigate the evolution of human languages. The Economist has a good summary of the two articles here. One (published in Nature) is co-authored by four researchers: Michael Dunn, Simon J. Greenhill, Stephen C. Levinson & Russell D. Gray (working in Nijmegen and Auckland); it presents a novel approach to the good old Chomskian idea of there being universal features common to all human languages. The team studied the characteristics of word order in four language families: Indo-European, Uto-Aztec, Bantu and Austronesian. As reported by the Economist,
The results were unexpected. Not one correlation persisted across all language families, and only two were found in more than one family. It looks, then, as if the correlations between grammatical features noticed by previous researchers are actually fossilised coincidences passed down the generations as part of linguistic culture.
The second paper, which emblematically appeared in Science, Nature’s arch-enemy, is by Quentin Atkinson, also of the University of Auckland, (good days for Auckland!). It is a study of the distributions of phonemes in different languages, and what the variety of phonemes in a wide range of languages and their geographical distribution could tell us about the geographical origin(s) of human languages. Again as reported by the Economist,
It has been known for a while that the less widely spoken a language is, the fewer the phonemes it has. So, as groups of people ventured ever farther from their African homeland, their phonemic repertoires should have dwindled, just as their genetic ones did.
So the basic idea is that the range of phonemes present in the different languages and their geographical locations might indicate epicenters, i.e. regions concentrating languages with more phonemes, and a gradual decrease in the number of phonemes as one moves further away from the epicenters. On the basis of computational modeling, Atkinson considered many possible epicenters, 2.500 to be precise, and the conclusion is surprising: the model suggests that there is only one such epicenter, in central or southern Africa (certainly not a coincidence!). This implies that all modern human languages do indeed have a common root, a unique ur-proto-language.
How do the results of these two papers relate to one another? At first sight, it might seem that they are somewhat in conflict. Atkinson’s results suggest that the emergence of human spoken language was a unique, isolated event, which in turn would support the idea of it being caused by a specific genetic mutation (possibly involving many genes), which then increased the fitness of the mutated individuals. In other words, that all languages ultimately stem from the same language may be viewed as corroborating the thesis that human language is a straightforward adaptation with a strong genetic component, and thus essentially ‘innate’ (an instinct, as Pinker famously claimed). The results by Dunn et al., by contrast, would strongly suggest that “cultural evolution is the primary factor that determines linguistic structure”, as the authors themselves claim in their abstract – nurture, not nature. These results contest the Chomskian idea of the presence of universal features in all human spoken languages, which would presumably reflect the innate, internal structure of our ‘language module’.
But of course, things are not nearly as black-or-white as the nature-nurture dichotomy would have us believe (as also articulated by M. Ridley in his Nature via Nurture). Not even the staunchest supporter of non-universalism about language would deny that there is a biological basis for our ability to speak, and this biological basis may very well have emerged from a single, isolated evolutionary path. In fact, trivially there must be a biological substrate for everything that humans do as a species, including the potential for cultural evolution. What the non-universalist claims (at least some of them) is that, given this starting point, there is still a wide range of possible paths that could have been taken, leading to the significant variations observed in human languages. To this day, we have never discovered human populations who do not engage in practices that we have no issue describing as speaking and talking, no matter how different they are from ours (e.g., languages lacking words for numbers). What the biological basis determines is that we, as a species, have a predisposition to engage in this particular form of communication, speaking, but it does not have to determine specific characteristics and in particular invariant patterns across human languages.
If indeed all human languages originate from the same proto-language, as claimed by Atkinson, then even if there were language invariants, they could have emerged as the contingent outcome of having a common starting point, rather than being an indication of strong biological constraints on the human ability to speak. But if Dunn et al. are correct, human languages have all evolved in different directions in spite of having a common starting point, which suggests that variation and change (essentially cultural processes) really are at the very core of our linguistic abilities (“it’s variation all the way down”, as well put by Alva Noë).
In other words, these two articles taken together actually become even more interesting than they would be if taken individually. If all human languages originate from the same rather than from multiple focal points, one might have expected that there would be universal (but possibly contingent) features common to all languages. Thus, the claim that there is no universal pattern across the four main groups of languages currently recognized becomes even more interesting and non-trivial if there is indeed this unique focal point from which all languages evolved. Conversely, given the variation observed, it would have been natural to expect multiple distinct starting points for the development of human languages, which makes Atkinson’s results all the more surprising and interesting. In sum, taken together, these results strongly suggest that there is at least one universal feature of human languages: variation.
Recent Comments