Evolution, family trees, and language

language.cs.auckland.ac.nz/media-material/

This article appeared in the Spring 2013 issue of Current Exchange Magazine.

The steppes of Southern Russia and Ukraine are dotted with lone hills, each one sticking out like a sore thumb against the flat, grassy landscape. Their presence is not some quirk of local geology: these hills conceal giant burial mounds, built by an ancient horse-riding tribe for its fallen chieftains.

Little else remains of this people, but when archaeologists began looking for the origins of the Indo-European culture – whose languages are spoken by some three billion around the globe today – their suspicions quickly fell on these mound builders. Five thousand years ago, researchers surmised, the horsemen left their ancestral home, carried by a powerful invention: the wheel.  As they swept across all of Europe and much of what is now Iran and northern India, the horse tribes spread their own culture and language throughout these lands, where they persist today.

Perhaps the most amazing thing about this popular hypothesis is how drastically different the alternative is. Many linguists and archeologists have argued that the birthplace of the Indo-European mother tongue was not southern Russia, but Anatolia, in what is now Turkey; that its first speakers were farmers rather than herders; and that they began their spread through the world some nine millennia ago, propelled not by the chariot, but by the plough. It may seem uncanny that we know so little about the cultural ancestors of much of Western civilization, Hinduism, and the ancient Persian empires. But the one factor uniting all the descendants of these mysterious ancient people – language – is not preserved in archaeological artifacts. And that makes the study of Indo-European origins incredibly tricky.

Last year, however, a team of researchers, headed by Dr. Quentin D. Atkinson from the University of Auckland, began to wonder whether the powerful tools developed by biologists to study evolution could help solve the mystery of the origin of Indo-European languages. As the ancient Indo-Europeans – whoever they were – settled into their new lands, their language began to split apart and change, much like living things evolve over time. Just like in biological evolution, such changes in language take time to accumulate. British and American English, which split apart a couple of centuries ago, are almost identical; German and English, separate for ~1500 years, are incredibly different. We know German and English are descended from one ancestral language because of many cognates, or words of common origin, that they share (for example, in German phrases like “das Bier ist gut” and “das Wasser ist warm”). So if it seems surprising that Russian, Hindi, and English are all descended from the same Proto-Indo-European language, that’s because each of them has had a very long time to evolve; nevertheless, their cognates give away their ancient common roots.

For biologists, the problem of studying the long-gone ancestors of today’s living things is akin to trying to figure out your family tree at a reunion. You might start by grouping relatives that have certain unique characteristics in common: for example, the only two redheads are likely to be siblings. From there, you can begin to reconstruct ancestors: one of the parents of those redheads had red hair too, you may guess. Sophisticated versions of this analysis allow biologists to determine the relationships between living things. In recent decades, the wealth of DNA sequence data from a huge variety of organisms has provided millions of characteristics that can be used to build family trees for everything from primates to viruses.

Dr. Atkinson and his team decided to apply the same methods to construct a family tree for 103 Indo-European languages, using cognates as the shared characteristics. For example, English shared twice as many words of Indo-European origin with German as it did with Russian; hence, English is more closely related to German. The resulting tree of languages itself held few surprises – linguists had determined these relationships using different methods a long time ago. But for Atkinson’s team, the tree was just the first step.

Biologists often use fossils – which come with a chemical “timestamp” – to determine how long ago an ancient creature lived, and apply this information to calibrate the time scale on an evolutionary tree. (Imagine you had to guess the ages of a friend’s grandparents; you’d do a much better job if you knew how old his parents were.)  In place of fossils, Dr. Atkinson’s team decided to use ancient languages. They knew, for example, that Old English was spoken in Britain around 1000 years ago. Placing Old English on the tree – along with Sanskrit, Latin, and many other ancient Indo-European languages – allowed Atkinson’s team to approximate when many groups of Indo-European languages first began splitting up. Extrapolating to the rest of the tree resulted in a predicted date for the first split in Indo-European languages – the time when Proto-Indo-European’s ancient speakers would’ve parted ways – at around 8500 years ago. This is also the time when the farmers of Anatolia are thought to have begun to spread out of their homeland.

This date already argued against the hypothesis that the first Indo-European speakers were the mound-builders of Russia, who lived only 5000 years ago. However, the researchers also wanted to directly test the competing claims about where Indo-European languages come from. Scientists studying the evolution of viruses have recently been able to use DNA-based evolutionary trees of virus samples, together with information about the locations at which these samples were collected, to predict where each ancestral strain of virus originated. This provided precious information about where outbreaks began and how infections spread. Atkinson’s team again decided to borrow from the biologists, using the locations where each language on their tree is spoken today to pinpoint ancestral languages on a map. This approach has a number of caveats, and to test it, the group decided to look at where their analysis would place Latin, the ancestor of Romance languages such as Spanish, French, and Romanian. As expected, the statistical analysis placed their common ancestor smack on top of Rome, where Latin originated. Satisfied, the researchers let the algorithm run some more steps back in time, to the origin of all Indo-European languages. This language was placed in southern Turkey. The limitations of Atkinson’s analysis notwithstanding, it appears that the languages of Europe, Iran, and northern India don’t come from a tribe of chariot-riding raiders, but from a group of ancient farmers.

While this study weighs in on an important controversy, perhaps even more exciting than the answers it provides are the ways in which it arrived at them. Over the course of 150 years of studying evolution, biologists have developed a set of incredibly complex family tree-based statistical tools to investigate the long-extinct ancestors of today’s organisms. They have managed to learn where and when these ancient creatures lived, and even to reconstruct pieces of their DNA. By tapping into this amazing tool set, Dr. Atkinson’s research helps usher in a new age for linguistics. So far, their analysis and family tree were based on relatively simple characteristics – a set of shared words. More complex future studies may include elements of grammar and pronunciation. Along with more tools borrowed from biology, these analyses can be used on language families from all over the world. Recently, they have even allowed linguists to more accurately reconstruct ancient languages – much like biologists have been doing with ancient genes – lending a voice to the hereto silent pre-historic remains of our ancestors.