This content is archived, it could be outdated.
Some language aspects change slower than others
On September 1, 2010, MPI researcher Dan Dediu's paper, 'A Bayesian phylogenetic approach to estimating the stability of linguistic features and the genetic biasing of tone', was published online in the Proceedings of the Royal Society B. This research furthers work Dediu has done with D. Robert Ladd of the University of Edinburgh. Three years ago, they claimed that the geographical distribution of tone languages might be linked to two genes involved in brain growth and development. This 'genetic biasing hypothesis' predicts, among other points, that tone tends to be more stable than other language properties. Recent studies, including Dediu's new paper, have supported this prediction.
September 6, 2010
We all experience language change, making the speech of younger generations sound 'funny' (or 'sloppy', depending on your take) and producing disagreements about the meaning of words such as 'peruse'. Given enough time, these small changes will add up to produce new languages, as any English speaker trying to understand Beowulf, or Italian speaker reading Virgil will be able to confirm. Languages not only differ in the words or sounds they use, but also in their grammar – their structural properties.
It has been long known that some words or sounds can change in a matter of tens or hundreds of years, and while it was suspected that others would be more conservative, it was not clear how slowly they actually would change. Recently, there has been a small revolution affecting the way we study language evolution and change, fueled by the adoption of methods developed in evolutionary biology and genetics to deal with the way complex biological systems (such as species or genomes) change through time. When applied to language, these methods helped to identify words and language structures that seem to change as slowly as genes, having weathered, in some cases, tens of thousands of years.
More language families
The only potential drawback to these previous studies is that, due to the intrinsic difficulties of collecting good quality data for many languages, they have focused on a few intensively studied language families, like Indo-European, Austronesian and Bantu. However, there is no guarantee that patterns detected in these cases can be generalised: Dediu's paper tries to avoid this bias by using data from many more language families. Moreover, given that specific methods used to infer such rates of change might have their own biases when applied to language data, the present study uses several software packages and ways of coding the data and linguistic classifications. These all tended to produce similar results, giving a bit more robustness to Dediu's conclusions. It was found that there do seem to be certain aspects of language which tend to resist change.
Genetic biasing hypothesis
Three years ago, Dediu and D. Robert Ladd (University of Edinburgh) claimed that the geographical distribution of tone languages – those, such as Yoruba or Chinese, which use the voice pitch to convey linguistic information – might be linked to two genes involved in brain growth and development, ASPM and Microcephalin (see link). This genetic biasing hypothesis makes the prediction that tone should tend to be more stable than other language properties, being 'anchored', as it were, in the slower-changing genetic landscape below the fast and furious cultural sea above. And, indeed, tone was found to be one of the most stable properties studied, supporting the prediction.
This raises an interesting question: what about the other stable features? And what makes them stable? 'It is, indeed, possible that tone is not the only aspect of language in which genetics has a word to say and the method introduced in my paper could help to identify them', says Dediu. 'It is also possible that purely cultural and historical processes play a role in their stability. Nevertheless, this is a question which can be meaningfully asked and the answers will help us better understand language.'
'This is only the beginning', Dediu continues, 'and these results must be seen, as for many products of scientific inquiry, as suggestive at best.' Dediu explains: 'We need to gather more high-quality linguistic data and make them available in public databases. We also have to develop new methods to extract the information from such data, tailored to the specifics of language. In this spirit of openness and freedom, the source code of the software developed for this study is made available under the GNU General Public License (GPL). I hope it will serve as a framework for testing new ideas.'
See also: Dan Dediu