Linguistic paleontology and catastrophism

In the previous post (31st Aug 09) I reviewed the first part of an article by Andrew Garrett: «Convergence in the Formation of Indo-European subgroups: Phylogeny and chronology», 2006), in which he reaches a series of interesting conclusions about the philogeny of Indo-European (IE) by analysing some evidence from ancient Greek dialects. Today, I'll review the rest of the article.

In the second part, Garrett focuses on the origins and dispersal of IE. Using linguistic paleontology as the main argument, he concludes that the date of IE dispersal cannot be earlier than 4,000 BC., aligning therefore with the traditional, also called Kurganic theory. According to him, the existence of common Proto-Indo-European (PIE) words for plough, wheel, wool, yoke and other technological innovations invalidates Renfrew's theory of a Neolithic, and therefore earlier, dispersal of IE. At first sight, the argument of linguistic paleontology seems quite strong. It is one of the pillars of traditional PIE methodology, and has often been used as a way of reconstructing PIE society, economy, religion, etc. A remarkable example is Émile Benveniste’s (1969) Le Vocabulaire des institutions indo-européennes (in this post you can find some comments on this book). Linguistic Paleontology aims to reconstruct the vocabulary of a proto-language by analysing and comparing the linguistic material of the descendant languages. It relies on the reconstructed linguistic forms and the inherent assumptions that lead to those reconstructions. In this blog I have shown various examples of how misleading this type of traditional reconstruction, based on branches, subgroups, laws and other theoretical entelechies can be. Linguistic paleontology can easily be proved wrong as a way of dating or reconstructing the homelands, the dispersals or the societies of the speakers of a given proto-language. For example, how can we know exactly what a given reconstructed word actually meant 6,000 years ago? This is only one of many questions that could be asked to the advocates of this method. In Chapter 4 of his book The Puzzle of the Indo-Europeans (1987), Colin Renfrew offers some curious examples of how the use of linguistic paleontology can lead to seemingly ridiculous results. In a previous post I mentioned the example of the word television. If we analysed the various words for 'television' in Romance languages from the perspective of language palaeontolgy, we could reach the conclusion that ancient Romans actually had TVs in their villas! For a complete evaluation (and I would say complete demolition) of linguistic paleontology I suggest reading an article by the British linguist Paul Heggarty (2006): Interdisciplinary indiscipline?: Can phylogenetic methods meaningfully be applied to language data - and to dating language? - Funnily enough, both articles, Garrett's and Heggarty's, are published in the same book: Forster, P., and C. Renfrew, eds. (2006) Phylogenetic methods and the prehistory of languages, Cambridge: McDonald Institute for Archaeological Research (further details here; a list of Heggarty's publications, here).

In the third part of his article, Garrett tackles the difficult question of IE dispersal, one of the leitmotivs of my blog. As we saw earlier, he champions the traditional chronology of IE, which means he is obliged to produce a convincing explanation for the dispersal of IE after 4,000 BC. Obviously, as we are in the 21st century and the traditional tales of horse-riding invaders from the steppes are no longer in fashion, he proposes a different type of explanation for the phenomenon of IE dispersal. In fact, he uses three different patterns: (p. 146): "One is steppe spread that led to the dispersal of Tocharian and Indo-Iranian. A second pattern is characteristic of the IE spread into Europe (...): dispersal was associated with systems collapse (...) and the social reorganizations of the secondary products complex (...) The third pattern is not widely noted but seems quite robust: a north-south spread into the interactional spheres of the urbanized zone that runs from the Aegean through Anatolia and the Near East to Bactria-Margiana". What we find here is a remnant of the traditional steppe migration plus a couple of relatively new ideas. The second pattern is peculiar: the idea of systems collapse reminds me of other catastrophic explanations for IE: there must be something catastrophic in order to explain the intercontinental expansion of this language group in just a couple of millennia (otherwise, how can you explain it?). 19th century scholars imagined a world of invasions and massive migrations. New developments, like Garrett's, put forward a more realistic scenario, but the problem still remains: in the systems collapse theory, there is a group of people, the speakers of PIE, who seem to have the secret of success. They wait in silence for centuries until the opportunity arises, and then, they are so irresistible that they impose their language causing the disappearance of any other previous language, an event which is repeated everywhere they go: Italy, Greece, Central Europe... IE opportunists I would call them. Maybe some elements in Garrett's proposals are acceptable and reasonable; very possibly they could explain some aspects of language spread or hybridization in the context of IE-speaking areas, but they can hardly be acceptable as a general explanation of the spread of PIE. As I have variously suggested in this blog, a much earlier date for IE dispersal is required.

Bibliography:
- GARRETT, Andrew (2006). «Convergence in the Formation of Indo-European subgroups: Phylogeny and chronology», in Forster and Renfrew, eds. Phylogenetic Methods and the Prehistory of Languages. Cambridge: McDonald Institute for Archaeological Research, 139-151.
- HEGGARTY, Paul (2006). «Interdisciplinary indiscipline?: Can phylogenetic methods meaningfully be applied to language data - and to dating language?», in P. Forster, and C. Renfrew, eds. Phylogenetic methods and the prehistory of languages, Cambridge: McDonald Institute for Archaeological Research, 183-194.
- RENFREW, Colin (1987). Archaeology and Language. The Puzzle of Indo-European Origins. London: Pimlico.

Language family trees: what are they good for?

Genealogical trees have been used extensively in historical linguistics. They are a visual representation of the relationships between an original language, normally extinct, and its descendants, through a series of divisions and subdivisions that have taken place during long periods of time. A familiar example would be the genealogical tree of Romance languages, stemming from Latin. Another example is the IE familiy tree, which has been portrayed in a variety of forms. In this post you can find a couple of pictures depicting the IE 'family' (the first one is really beautiful, with its leaves and branches; the second one is a more prosaic diagram).

The idea behind family language trees seems quite reasonable. It is based on analogous types of representation found in other sciences, especially zoology and botany, where the evolution of species is shown as a succession of mutations and adaptations whereby new species are born out of the older ones. It seems that the use of genealogical trees in biology is justified, or at least acceptable. Now, can languages be compared to living organisms? Do they undergo mutations? Can they be classified into family trees? In this blog, I have already stressed the fact that there are no inherent components in languages that make them change. It is speech communities that change, causing hybridization and other phenomena that affect language. It is hard to imagine how a given speech community would just split into two without the actual intervention of external factors. Following this line of thought, it is clear that language family trees are just an inaccurate account of how languages evolve. However, an advocate of these trees may argue that the new approaches to language change, like the ones proposed by Mario Alinei, can be incorporated to the old theory, so that the idea of family relationships, with the whole array of divisions and branches, can still be maintained, at least to a certain extent. This way, for example, the division of language A into A1 and A2 is not just the consequence of a simple ‘mutation’ but the result of a combination of social and historical factors. Family trees are saved: there’s nothing specially wrong about them, and they’re visually attractive and very useful for pedagogical reasons. No problem then... Really? I’m not so sure.

First, a question: what can we see when we look at an IE genealogical tree? Does it depict the history of IE languages or, rather, the history of the written languages belonging to the IE group? The main concern of traditional historical linguistics has been the study of the written texts that have reached us from the past. Most of the principles and rules that have been proposed to describe the genealogical relationships between the various IE languages are based on this type of analysis. Is this an acceptable way to understand the ‘evolution’ of languages? I think it isn’t.

Let’s imagine a given geographic area, where there is written evidence of an old, now extinct language (Language A): by carefully studying the texts and all the available information, we can produce a description of the grammar, vocabulary and phonology of this language. We also find a second 'language' in the area (we will call it Language A2), whose written documents belong to a different period, about 500 years later. This language has a lot in common with its predecessor (Language A) but shows some clear signs of ‘evolution’. In Language A, for example, we find the words “pata” and “bolún”. Five centuries later, in Language A2, we have "peita" and "bulún". Apparently, the facts are clear: Language A has evolved into Language A2, and we can even devise a series of 'laws' that govern these changes. No problem. End of story. Or maybe not... As I said earlier, traditional IE theories and chronology are mainly based on the 'evolution' of written standards. However, if we apply a sociolinguistic approach to diachronic studies we get a different picture. Written languages originate in the dialect spoken by the ruling elite. Their social prestige and their association with power turn them into a very influential factor that affects the language of the whole society, but we can imagine that in a stratified society this written standard is not ‘the language’ (take a look at this post for a similar analysis). Let's turn now to Language A2 in our example. What is it exactly? Is it the result of an evolution process from Language A? Or is it a standard language connected with a different ruling elite that emerged later in the historical record? How do we know that the features of Language A are 'older' than the features of Language A2? The only thing we know is that they’re different standardized languages used at different times. Mario Alinei, in the framework of the Continuity Theory, is one of the first linguists who have analysed historical linguistics from a sociolinguistic perspective. He takes into account the written evidence and the traditional analysis but he puts them in a completely new context. His research on Italian dialects, for example, shows that some traditional assumptions about the emergence of Romance dialects in Italy cannot be accepted any more, as we have already seen in this post. A similar type of analysis could be applied to other languages.

The speed of change

In 2003, Morten Christiansen and Simon Kirby edited a volume called Language Evolution (Oxford University Press), which is an excellent, and quite comprehensive account of the main lines of research in the field of the origin of language and language evolution. (You can find more information about this book and an excerpt here).

The authors who contributed articles to this volume are among the leading researchers in the area, e.g. Steven Pinker, Derek Bickerton, Michael Tomasello, Philip Lieberman or James Hurford. They come from a variety of scientific backgrounds (linguistics, archaeology, psychology, anthropology, etc.) and they’re mainly concerned with the evolution of the language human capacity, rather than the evolution of particular languages or language groups. However, from time to time they use data from historical linguistics (mainly from Indo-European studies, which is the best-known, most studied language group) in order to develop their own theories. They’re no experts in historical linguistics, so they naturally turn to the main theories and most recent developments in this area. What they find at the beginning of the 21st century is not much different from what they would have found fifty years earlier, especially in the field of Indo-European studies: traditional chronology and traditional assumptions about language change. I have variously shown in this blog that these ideas are quite dated and basically wrong. But they are still the generally accepted framework, which means that they lead other authors, who are not specialists in historical linguistics, to assume wrong things about the history of languages. In Christiansen and Kirby’s volume I have found some examples of these induced errors. Let’s take a look at them:

- Frederick Newmeyer, What can the Field of Linguistics tell us about the Origins of Language? (op. cit., pp. 58-76).
(p. 62): “The problem is that languages change so fast that (by most estimates) the method allows a time depth for reconstruction of about 6,000-8,000 years (...). So if Proto-World coincided with the appearance of Homo Sapiens, let us say arbitrarily 200,000 years ago, then the intervening 192,000-194,000 years would have been more than enough time to obliterate any signs of what it might have looked like”.

- Michael Tomasello, On the Different origins of Symbols and Grammar, (ibid, pp. 94-110).
(p. 103): “... grammaticalization and syntacticization is somewhat speculative - (...) for example, the main diversification of the Romance languages took place during some hundreds of years - and thus there is no reason why they could not also work to make a simpler language more complex syntactically in some thousands of years”.

- Michael Arbib, The Evolving Mirror System: A Neural Basis for Language Readiness, (ibid, pp. 182-200).
(p. 182): “The immense diversity of the Indo-European languages (...) took about 6,000 years. How can we imagine what has changed in ‘deep time’ since the emergence of Homo sapiens some 200,000 years ago, or in 5,000,000 years of prior hominid evolution?”.

- Robin Dunbar, The origin and Subsequent Evolution of Language (ibid, pp. 219-234).
(p. 230): “Ancestral Indo-European, the language spoken by a small group of (probably Anatolian) agro-pastoralists around 6,000 BC, has given rise to around 150 descendant languages (...) over an 8,000-year period. Interpolating these values into the standard Gaussian logistic growth equation for biological population growth suggests that the Indo-European language family has evolved at a rate equivalent to the budding off of a new language from each existing language, on average, about once every 1,600 years”.

It is clear from these examples that the traditional chronology for Indo-European (IE) creates the wrong impression that IE languages changed at a very high speed. According to the Paleolithic Continuity Theory, which is a new paradigm in historical linguistics, these dates are wrong. IE languages took many more millennia to develop and change than commonly accepted. Language change is not an inherent or measurable feature in languages, it depends on historical, social factors, which means that there are certain circumstances which accelerate the process of change. In the absence or of these factors, human languages tend to remain unchanged. This conservatism of languages is beneficial in terms of communication effectiveness and learnability and it can easily be observed in all human communities. As I said, there are many factors which trigger language change. It is obvious, for example, that the Neolithic revolution is one of them. Another one is the migration of human groups or the invasion and colonization of territories. But these are relatively recent phenomena in the history of humanity. Let’s think about the Paleolithic period, which lasted more than a hundred thousand years and where the socioeconomic and demographic characteristics of human populations were quite different from the ones we know today. The main factor in language change is hybridization, that is, when there is contact between speakers of different languages. It is obvious that in a Paleolithic environment there were fewer opportunities for mobility and for exchange between human communities than in the Neolithic or Metal Age periods, and therefore fewer opportunities for language contact. On the other hand, Paleolithic societies were not highly stratified, which means that pressure from elite groups was not a relevant factor in language.

As we have seen, the assumption that language change happens at a given rate, or that it can be understood regardless of external factors, is a serious mistake that has pervaded historical linguistics for centuries. Evidence for this comes from a variety of sources, e.g. the fact that (relatively) isolated populations, like the ones we find in islands, speak languages with markedly archaic traits if compared to other related languages. This can be seen in Icelandic, for example. There is no internal clock in languages signalling the moment for divergence. In fact, the idea of divergence from the common tree is also very poor when applied to historical linguistics. Languages are not like animal or plant species that can be distributed along a genealogical tree in the form of branches or sub-branches that are the consequence of mutations and adaptation. Languages are something quite different.

Further reading:
- ALINEI, Mario. (1996). Origini delle lingue d'Europa. Vol. I: la Teoria della Continuità. Bologna: Il Mulino.
- ALINEI, Mario (2000). Origini delle lingue d'Europa. Vol. II: Continuità dal Mesolitico all'età del Ferro nelle principali aree etnolinguistiche. Bologna: Il Mulino.
- DIXON, R. M. W. (1997). The Rise and Fall of Languages. Cambridge: CUP.

Word diffusion

In the framework of traditional historical linguistics, also called Comparative Grammar, the reconstruction of Proto-Indo-European (PIE) words has been the basis for the reconstruction of PIE society and culture. An analysis of the economy, social structure and technological level of this human group is a relevant tool in order to determine the chronology of IE languages and peoples. For example, by determining the presence of common PIE vocabulary for agriculture or stock breeding, some authors have concluded that the speakers of PIE were the members of a post-Neolithic society. A classic in this type of research is Émile Benveniste’s Le Vocabulaire des institutions indo-européennes (1969), in two volumes. Even though it is written in the framework of the traditional IE paradigm, I must say this book is worth reading, especially because of the author’s impressive erudition and depth of analysis. More recently, J.P. Mallory has also summarized the main points in the reconstruction of PIE society (see J.P. Mallory's In Search of the Indo-Europeans, 1989, pp. 110-127). He gives examples of reconstructed PIE words which can be considered ‘general’, that is, present in a relevant number of IE languages. These words belong to the fields of environment (for example the names of trees), the economy (words for “sheep”, “pig”, “horse”; only a few words connected with agriculture, e.g. *agros), the concept of ‘settlement’, technology (pottery, metallurgy, sewing, words for “wool”, “wheel”, etc.) and social organization (kinship terms, words for “chief of the clan”, etc.). After reviewing these examples, Mallory reaches the following conclusions:

(p. 126): “... we can see that the presence of words for pottery, domestic animals and agriculture in the Proto-Indo-European lexicon argues that the community was at least Neolithic and that it would be senseless to assign the Proto-Indo-European to the earlier hunter-gatherer societies of the Mesolithic”.

(p. 127): “..., we might then assign a notional date of about 4500 BC as the earliest probable time for the culture reconstructed from the inherited vocabulary of the Indo-European languages”.

(p. 127): “In the broadest terms then, the Proto-Indo-Europeans were a Late Neolithic or Eneolithic society which began to diverge about 4500 to 2500 BC”.

These conclusions seem to be based on solid ground. Or maybe not...

First, a brief excursus. We all know that television is a 20th century invention. Like many other technological innovations, it spread quickly from the place were it was invented (Great Britain) to the rest of the world. And the name of the invention travelled as well, adapting to the various languages. This phenomenon is called diffusion in linguistics. There are many examples of the diffusion of the word “television” in many languages: Spanish televisión, Tagalog telebysion, Mongolian телевизор (“televisor”), Swahili televisheni, Bahasa Indonesian televisi, Turkish televizyon, etc. Something similar happens with other recent technological developments (computer, video, DVD, etc.), which have quickly spread around the world. It is obvious that the similarity between English television and Tagalog telebysion is no proof of the common origin of both languages. Let’s try to imagine a linguist defending that proposal on the basis of this vocabulary! But diffusion is not only a recent or contemporary phenomenon. There are many examples of terminology diffusion in the 19th century, or the 18th century, or in the Middle Ages, or in the time of the Roman Empire, and in prehistory. We can deduce that the diffusion of technological innovations such as agriculture or the wheel, which had a major impact in pre-historical societies, was accompanied by the corresponding terminological diffusion. Could it have happened otherwise? By applying this concept to the study of PIE, it is becoming clear that some of the supposed common vocabulary in pottery, livestock, sewing, etc, can be explained quite satisfactorily in terms of diffusion. In other cases, it is not even necessary to use this method. Authors such as Mario Alinei, Xaverio Ballester, Francesco Benozzo and Gabriele Costa have given many examples of how the list of ‘common PIE society/culture terms’ can be reinterpreted in terms of diffusion or other criteria. In fact, we can say that all the classical interpretations, one by one, can be refuted. And not only that. This new paradigm (the Paleolithic Continuity Theory) also seems to provide answers for cases which were not explainable following traditional criteria.

The main result of this re-interpretation of PIE society/culture terms is obvious: they offer no confirmation of traditional IE chronology, for which they were the main basis. In previous posts I have mentioned evidence from archaeology, population genetics and anthropology which also seems to point in the direction of a Paleolithic origin of PIE. Which means, for example, that the first humans who populated Europe from about 40000 BC were already speakers of Indo-European languages (Celtic, Italic, Germanic, etc.), or speakers of languages belonging to the other families (Uralic, Basque, etc.) which are nowadays present in Europe.

Mario Alinei

Born in 1926, Mario Alinei is an Italian dialectologist who has made impressive contributions in the field of historical linguistics (you'll find his CV here, and some of his articles here). I have had the opportunity of reading some of his writings, including his opus magnum: Origini delle lingue d'Europa, published in two volumes (1996 and 2000).

On the whole, Alinei's ideas are really interesting, and revolutionary. Maybe the fact that he is not actually an expert in historical linguistics has given him the right perspective to evaluate things properly. The core of his approach is to apply sociolinguistic methods to the study of language evolution. He focuses on dialects, both modern and old. He is not worried about establishing rules or classifying languages in genealogical trees, and his aim is not to reconstruct protolanguages. It would be difficult to summarize the main points of his theory (for a good introduction, see here) but I'll try anyway:


- The traditional chronology for Indo-European languages is wrong. Proto-Indo-European didn't start to expand at around 4000 BC, as traditionally accepted, but much earlier, in the Paleolithic period.

- The dialects and populations of Europe are characterized by their continuity throughout the ages, at least from the Upper Paleolithic. This is confirmed by comparing linguistic, anthropological, archaeological and genetic data. There is no evidence for massive invasions or migrations in recent prehistory (Neolithic or post-Neolithic periods) .

- Languages don't change at a given speed or because of intrinsic mechanisms. Language change depends on external, historical factors. The most relevant one is hybridization, that is, when speakers of different dialects mix.

- The lexicon is the most reliable type of data in order to study language change. In the traditional paradigm, linguists were primarily concerned about grammatical and phonological rules.

In future posts I will discuss some of Alinei's writings and ideas. I'm planning to write a review of his book Origini delle lingue d'Europa.

Last edit: 12 July, 2008