HENRY the Human Evolution News Relay

Punctuated Equilibrium and the evolution of languages.

Over the past few weeks the science-blog cadre has been squabbling about punctuated equilibrium (PE). Today, in a rather brilliant example of good timing - if I say so myself - some colleagues (Quentin Atkinson, Mark Pagel, Andrew Meade, Chris Venditti) and I have a paper showing that languages evolve in punctuational bursts.

Atkinson, Q.D., Meade, A., Venditti, C., Greenhill, S.J., & Pagel, M. (2008) Languages evolve in punctuational bursts. Science, 319, 588. (doi:10.1126/science.1149683)

.

More on this in a minute.

So, what is punctuated equilibrium?

Most orthodox views of evolution argue that new species arise out of a gradual accumulation of genetic changes. Eldredge and Gould (1972) disagreed, and argued for a model of punctuated equilibrium where evolution shows a pattern of "bursts" of change and periods of little overt change ("stasis").

This did not go down particularly well. Richard Dawkins and Daniel Dennett, for example, tried to explain it away. Dawkins, in The Blind Watchmaker complained that this was what everything thought all along, and no-one was really a gradualist in the way Eldredge and Gould had proposed. He argued that their conception of gradualism was akin to believing that in the biblical story of the children of Israel, the kids literally moved one yard an hour over the 40 years it took them to cross the Sinai desert. Dennett made a similar argument in Darwin's Dangerous Idea, which tried to explain away punctualism as an incorrectly scaled graph (i.e. increase the time axis and everything looks gradual!).

Others focused instead on the types of mutations that could cause punctuational evolution. Are there certain mutations which have a large effect on the phenotype? This argument has popped up repeatedly under various names like 'saltationism', 'hopeful monsters', and 'macromutations', etc. Most recently, last week, Olivia Judson revived it again, and was promptly stomped by Jerry Coyne.

However, punctualism does not require macromutations, and Gould largely saw these as separate debates. As T. Ryan Gregory points out, punctualism is more about the rates of change. Are there periods in evolutionary lineages where rates of evolution are fast, and are there periods where they're slow? If there are differences in rates of evolution over time, then we can detect these using phylogenetic methods.

Punctuation and Language Evolution

The punctualism debate has also entered linguistics - Dixon wrote a recent and extremely controversial book, The Rise and Fall of Languages, where he argues that language evolution is, in many cases, punctuational. Dixon's not alone - other linguists have also noted the potential usefulness of this idea (Claire Bowern has an excellent paper reviewing this debate: Punctuated equilibrium and language change (PDF)).

So, how do languages evolve? The languages we see in the world today are the outcome of an evolutionary process of descent with modification. Over time, languages accumulate changes in their grammar, phonology (sound system) and vocabulary. Compare, for example the differences between modern English, Shakespeare or Chaucer. Or, to take a simple example, the words for "back" (the body part) in the Tahitian language, recorded at different points in time:

• 1773 - etoo
• 18th Century - tue
• Modern - tua

Occasionally, new languages are born, perhaps as a result of some members of the population of speakers migrating elsewhere causing the language lineage to split in two. The two descendant lineages then begin to accumulate changes independently, and over time become mutually unintelligible. Eventually, these become languages in their own right. For example, when the Saxons crossed the channel from Europe, their language separated from the ancestral Germanic language and eventually became English.

Using phylogenetic methods, we can reconstruct the evolutionary history of these languages. In this study, we used word "cognates" to build family trees of the languages from three large language families. These cognates are words of similar form and meaning that have descended from a common ancestor. For example, the German "vasser" and Hittite "watar" are cognate with the English "water", but not with the French "eau". For an even bigger example, you can see some of the data used in this study below (and it's all on my research website). This shows some of the words meaning "hand" in a number of languages. Note the first cognate set of related words like "lima" found in languages like Bali and Sasak (all the entries coded as "1" in the right-most column). There's a second and third cognate set for the forms related to "tangan" (18) and "lenge" (19):

...and just to re-use one of my favourite images, here's a map of where these words are spoken. This time, it's words meaning "hand" in the Austronesian languages of Indonesia and Papua New Guinea. The cognate sets of related words are color-coded. More on this here: Rates of word evolution.

Our model of language evolution aims to find the (set of) family trees that best explains the distribution of these cognate related words. So, if two languages - say, English and German - share a lot of cognates, then they are closely related. If they share few cognates, then they're further apart on the tree and separated by longer branches.

In this paper, we analysed the cognate sets found in 200 items of basic vocabulary from hundreds of languages in three of the largest language families in the world: Austronesian, Bantu, and Indo-European. You can see a good chunk of the data at my research website here.

Here's the fun part - once we've built these trees, the length of the branches is proportional to the amount of change between languages: the longer the branch the more change has occurred. This gives us a "path length" from the root of the tree to each language, which represents the total amount of change that each language has undergone.

What we show (after a few statistical corrections to avoid node density artifacts), is that as the length of these paths increase, then there are more languages born along that lineage. This is indicative of punctuated equilibrium: splitting events seem to be promoting language change. In contrast, if lexical evolution was gradual, then the path lengths would be independent to the number of divergence events along that path. In fact, across the three families, we show that 31% of Bantu, 21% of Indo-European, and 9.5% of Austronesian vocabulary arose at or around the time that the languages diverged.

Ok, so what could be behind this? We argue that there are two possible reasons for this punctuational effect in language evolution. First - as languages diverge, they tend to split into smaller populations. Like small biological populations, small speech communities tend to be at the risk of things like founder events and drift, which increase the rates of change. To check this, we looked at the Polynesian languages, which underwent a long chain of founder events as they settled the Polynesian triangle. These languages showed a huge increase in this punctuational effect (33%) when compared to their parent family, Austronesian (9.5%)

The second possible cause of punctuational language evolution is that people tend to use their languages as a way of providing a social "emblematic" function. That is, people use their languages as a tool for promoting social cohesion and group identity (we speak like this, they speak like that). To leave the last word to Noah Webster, the lexicographer behind the American English Dictionary: "as an independent nation, our honor requires us to have a system of our own, in language as well as government".

More Links:

Nature News coverage, the Telegraph