by Paul Heggarty, Max Planck Institute for Evolutionary Anthropology, Leipzig.
1. Towards the End-Game at Last?
An ‘ancient DNA revolution’ is now sweeping through genetics. Suddenly, ancient population migrations can be recovered far more clearly than before. For linguists, this holds out the prospect of ‘closure’, at last, on the Indo-European question. And that is quite some prospect, for agreement on the origins of Indo-European has eluded us ever since linguistic science began, when Sir William Jones first posed this very question in 1786.
Today’s issue of Nature (11th June 2015) publishes two major papers based on Bronze Age ancient DNA from the Eurasian Steppe — one of the two leading candidates for the original homeland of the Indo-European family.
- Haak, W. et al. [David Reich’s group, Harvard] 2015. (online since 2015-03-01)
Massive migration from the steppe was a source for Indo-European languages in Europe.
Nature 522 (7555): p.207–211. http://dx.doi.org/10.1038/nature14317
- Allentoft, M.E. et al. [Eske Willerslev’s group, Copenhagen] 2015.
Population genomics of Bronze Age Eurasia.
Nature 522 (7555): p.167–172. http://dx.doi.org/10.1038/nature14507
Both papers interpret their results as leaning towards the Steppe hypothesis, albeit rather tentatively and superficially in places. On closer inspection, indeed, all is by no means so clear cut. The new data actually turn out to be equally compatible, if not more so, with the Steppe as the immediate origin of just a few branches of Indo-European (notably Balto-Slavic and perhaps Tocharian). These Bronze Age movements would thus be only secondary to an original Neolithic expansion of the Indo-European family as a whole, with farming, out of the northern arc of the Fertile Crescent (i.e. the ‘Anatolian’ hypothesis). The ancient DNA data also reconfirm the spread of farming as the dominant shaper of the genetic make-up of Indo-European-speaking southern, Mediterranean Europe, with relatively little Steppe impact.
Now there is of course one big caveat to all of this: languages do not always ‘go with genes’ in any case. So at the end of this blog, Addendum 1 takes up just that issue. Still, the claim in both Haak et al. (2015) and Allentoft et al. (2015) to support the Steppe hypothesis is indeed founded specifically on their new genetic data, so it is on that basis that this blog will assess them.
2. Asking the Wrong ‘Indo-European Question’
The mistaken impression that the results support the Steppe hypothesis comes from what seems to be a failure to grasp the scope and scale of the Indo-European question. And from the an apparent assumption that if one can detect some population movements originating in the steppe, then that means they must support the steppe hypothesis, no?
Yes, Haak et al. (2015) report a strong case for a population movement in the Early Bronze Age (c. 4500 BP) by which people of the Yamnaya culture on the Steppe provided a major genetic input to the people of the Corded Ware culture of north-eastern Europe.
But the Indo-European question is not “Was there or was there not a Yamnaya > Corded Ware population movement?” Rather, the question is: what explains the entire Indo-European family? That means (see map below) Indo-European languages not just in northern and eastern Europe, but also right across southern and western Europe, ancient Anatolia, Armenia, much of the Middle East (Kurdish-speaking areas, Iran, Afghanistan), and most of Pakistan, India and Bangladesh. That far greater scale is ‘the Indo-European question’. And that is what the two main rival hypotheses — pastoralism out of the Steppe, or farming out of Anatolia — claim to answer: where all of Indo-European comes from.
3. History of an Argument: Pastoralism or Farming?
One should not be under any illusions, then, about the scale of what needs to be explained. Indo-European is the linguistic lineage of far more of the world’s population and land area than any other (and has been since long before European expansionism). For a map of present-day distributions in Eurasia, one example is Steve Huffman’s here: the full resolution version is at: www.worldgeodatasets.com/index.php/download_file/view/1553
For all three contexts of the where, when and why of first Indo-European expansion, the two main hypotheses propose radically different answers.
- They place the homeland either on the Eurasian Steppe just north of the Black and Caspian Seas — or in the northern arc of the Fertile Crescent, through central-eastern Anatolia.
- They set the time-depth when it first began to disperse and diverge at around either six millennia ago — or nine.
- And their answer to the why ‘causation‘ question is either highly mobile horse-based pastoralism, or the ‘Neolithic Revolution’ and spread of farming.
For doubters of the Steppe hypothesis, the whole problem has always been that it only ever seemed plausible to explain the languages of north-eastern Europe, and perhaps Tocharian. For the great majority of the family’s branches and distribution, however, Steppe pastoralism never convinced: across the early farming heartlands of the Mediterranean (i.e. Greek, Latin and its Italic sisters, Albanian, other extinct branches of Indo-European in the Balkans and western and central Anatolia); and south of the Caspian and Himalayas, in the Iranic and Indic branches.
Yes, the new genetic data do point to a Steppe role in north-eastern Europe, and perhaps Tocharian. But that is not the real point, then. Rather, to the extent that these papers discuss the other, main Indo-European-speaking regions at all, the data here seem only to reconfirm the impression that Steppe pastoralists had no great impact there.
4. All Indo-European, or Just Some? Primary Migrations, or Secondary ?
The new papers rather misleadingly talk of ‘the’ hypothesis of steppe origins, and ‘the’ Indo-European languages, as if referring by default to the entire family, and to what explains that family as a whole. But they go on to refer mostly to regions that host but a small proportion of the sub-branches of Indo-European, which is all that their new data really bear on. This is also why their results actually fit also with a secondary sub-expansion (out of the Steppe) of just those few parts of Indo-European. This Bronze Age movement would have been a second stage, long after the primary stage in the Neolithic which had already seen the much wider expansion of most of the family, with farming, out of an original homeland in the northern arc of the Fertile Crescent (i.e. central-eastern Anatolia, hence the ‘Anatolian’ or ‘farming’ hypothesis).
The title of Haak et al. (2015) writes of the steppe as “a source for Indo-European languages in Europe” [emphasis added], and the abstract claims “These results provide support for the theory of a steppe origin of at least some of the Indo‐European languages of Europe.” But this is not the established Steppe theory for Indo-European, which is instead that the steppe is the origin of all Indo-European languages, and not just of Europe, but everywhere else too: Iran, India, (formerly) Anatolia, and so on. So if the steppe is the source of only some of the Indo-European languages of Europe, then that contradicts the steppe hypothesis. And it is compatible with the theory that Indo-European languages in the rest of Europe arrived there earlier, as in the farming hypothesis (version A2, see below).
Allentoft et al. (2015), likewise, make statements such as “Our genomic evidence for the spread of Yamnaya people from the Pontic-Caspian steppe to both northern Europe and Central Asia (Fig. 1) corresponds well with the hypothesized expansion of the Indo-European languages”, when it does not. Northern Europe and Central Asia fit only with a subset of those languages, which is not the Steppe hypothesis. Their title is “Population genomics of Bronze Age Eurasia”, but it too only really addresses northern Eurasia.
Mediterranean Europe, and Asia south of the Black Sea, Caspian and Himalayas, are core areas of Indo-European speech, the large majority of the family, and crucial to answering the Indo-European question. Neither paper does justice to these areas, and neither reports any significant new genetic data for them for the appropriate time-periods. Their ancient DNA data from northern Eurasia are already revolutionary achievements; nobody imagines getting these results was easy. But until we have fuller geographical and chronological spans of genetic data, those from the north alone will not be able to answer the question of Indo-European origins, and it is unwise to imagine that they will.
5. The Alternative: Farming Hypothesis A2
These papers also do not look to the deeper origins of the pastoralism that eventually developed on the Steppe.
A Yamnaya > Corded Ware population movement is in fact entirely compatible with the farming hypothesis, as a later, secondary movement within the overall spread and intensification of food-production. The main first phase saw farming (both crops and livestock) spreading originally out of the northern arc of the Fertile Crescent: westwards into Europe; and eastwards through the Middle East, to the Indus and eventually the Ganges. Colin Renfrew’s (1987) original ‘hypothesis A’ for Indo-European had both of these movements spreading the Indo-European family too, both eastwards and westwards.
But farming also spread out of Anatolia ‘northwards’ onto … the Steppe: see the map below from Balaresque et al (2010), their Figure 1A on the chronology of the spread of agriculture through Europe.
Or for another source, Bellwood’s (2005) Figure 4-1 on the spread of farming through Europe:
Whether this spread to the Steppe was predominantly through the Balkans and/or Caucasus remains debated. It may hint at a Caucasus route that the Balto-Slavic branch of Indo-European in northern and eastern Europe, overlapping heavily with the Corded Ware territories, shares certain linguistic characteristics with the ‘eastern’ Indo-European languages (Indo-Iranic), and not with the other branches of Indo-European in Europe.
Once on the steppe, farming eventually specialised there into pastoralism, in this highly suitable grasslands environment (although some crop-farming continued in more western regions). Several millennia after the first spread of farming, this new and predominantly pastoralist ‘package’ on the Steppe made for an intensification of food production, most viable and ‘competitive’ in environments where farming was not yet especially productive, including temperate north-eastern Europe. It seems no surprise that pastoralists spread successfully there in the early Bronze Age, then.
But this Steppe pastoralism was in any case ultimately a development out of farming from Anatolia. That is, this Yamnaya > Corded Ware movement was indeed secondary within the overall spread of food production. And there is very little evidence in the new papers that this movement brought into Europe any branches of Indo-European other than Baltic, Slavic and perhaps Germanic, the only braches of Indo-European with high Steppe input. (The case of Germanic is more complex anyway: the former Corded Ware regions whose populations today speak German were actually mostly Slavic and Baltic-speaking until Mediaeval times.)
6. Conveniently Ignored: What about Uralic?
Haak et al. (2015) make much of the high Yamnaya ancestry scores for (only some!) Indo-European languages. What they do not mention is that those same results also include speakers of other languages among those with the highest of all scores for Yamnaya ancestry. Only these are languages of the Uralic family, not Indo-European at all; and their Yamnaya-ancestry signals are far higher than in many branches of Indo-European in (southern) Europe. Estonian ranks very high, while speakers of the very closely related Finnish are curiously not shown, and nor are the Saami. Hungarian is relevant less directly since this language arrived only c. 900 AD, but also high.
These data imply that Uralic-speakers too would have been part of the Yamnaya > Corded Ware movement, which was thus not exclusively Indo-European in any case. And as well as the genetics, the geography, chronology and language contact evidence also all fit with a Yamnaya > Corded Ware movement including Uralic as well as Balto-Slavic.
Both papers fail to address properly the question of the Uralic languages. And this despite — or because? — the only Uralic speakers they report rank so high among modern populations with Yamnaya ancestry. Their linguistic ancestors also have a good claim to have been involved in the Corded Ware and Yamnaya cultures, and of course the other members of the Uralic family are scattered across European Russia up to the Urals.
7. Yamnaya > Corded Ware: How Big an Impact?
Haak et al. (2015) write of the Yamnaya > Corded Ware movement in terms of a “massive” population migration and near-total replacement. Allentoft et al. (2015) row back considerably, however, on the scale and impact of this movement. They write instead explicitly of admixture with existing farming populations:
Populations in northern and central Europe were composed of a mixture of the earlier hunter-gatherer and Neolithic farmer groups, but received ‘Caucasian’ genetic input at the onset of the Bronze Age (Fig. 2). This coincides with the archaeologically well-defined expansion of the Yamnaya culture from the Pontic-Caspian steppe into Europe (Figs 1 and 2). This admixture event resulted in the formation of peoples of the Corded Ware …
… the resulting Corded Ware culture in Europe was the result of admixture with the local Neolithic people.
Allentoft et al. (2015) are also more insistent on the limited geographical scope of this impact, and that the Corded Ware samples do not stand in for the wider population of northern and eastern Europe.
… European Late Neolithic and Bronze Age cultures such as Corded Ware, Bell Beakers, Unetice, and the Scandinavian cultures … display a cline of genetic affinity with Yamnaya, with highest levels in Corded Ware, lowest in Hungary, and central European Bell Beakers being intermediate …
Indeed, although Haak et al.’s (2015) Figure 3 can seem to present strikingly high rates of Yamnaya ancestry in some parts of Europe, things look much less clear-cut in the alternative presentation in Allentoft et al.’s (2015), Figure 2: “Genetic structure of ancient Europe and the Pontic-Caspian steppe.”
Note also the relative scale of the main Neolithic (A) and minor Bronze Age (C) impacts recorded in Brandt et al.’s (2013) Figure 3: “Development of mtDNA components from the Late Mesolithic to present day”.
8. Europe: North vs. South
Above all, the Yamnaya > Corded Ware impact is much less widespread in Europe than Indo-European languages are. Much of southern Europe has spoken Indo-European languages from our earliest records (Latin and its ‘Italic’ relatives, Greek, Albanian and various other Indo-European languages of the Balkans, now extinct).
Some (low) proportions of apparent ‘Yamnaya’, ‘Corded Ware’ and north European ancestry do appear in present-day populations of southern Europe (Haak et al. 2015 Figure 3b). But such north to south population admixture is in any case expected from the historical period. The collapse of the Roman Empire and the migrations of the early mediaeval period were defined by major invasions and settlements of Slavic and Germanic-speaking populations into southern Europe. To clarify whether the ‘northern’ component already appears in southern European populations along with the Yamnaya > Corded Ware impact in the north, we need not modern samples but much fuller ancient DNA from Mediterranean Europe. Especially illuminating would be samples from the period when this region (and Anatolia) are already documented as speaking Indo-European languages: Mycenaean Greek, Italic, and Anatolian Indo-European.
For now, these papers do not yet provide new ancient dna from the southern regions at the appropriate periods. Allentoft et al. (2015) do at least report these new data from northern Italy:
the Copper Age Remedello culture in Italy does not have the ‘Caucasian’ genetic component and is still clustering genetically with Neolithic farmers
Chronologically, Copper Age Remedello is borderline too early to be contemporary with the Yamnaya > Corded Ware movement, so it is not necessarily conclusive. Nonetheless, it does reinforce the chronological concerns that attend the Steppe hypothesis on this score…
9. Chronology of Indo-European Divergence in Europe
Traditionally, the Steppe hypothesis has proposed a time-frame for Indo-European expansion of c. six millennia. For the Yamnaya > Corded Ware movement, however, Brandt et al. (2013) and Haak et al. (2015) give a date of just 4500 BP. But is a time-span of just 4500 years or so sufficient to allow for all the divergence between all European branches of Indo-European: Slavic vs. Baltic vs. Germanic vs. Celtic vs. Italic vs. Greek vs. Albanian, and so on? That is, we are left with little more than two millennia to take us, for example, from the early Latin of the last few centuries bc back to Proto-Italic, then further back to Proto-Italo-Celtic (if one accepts that clade), and then back again to allow for a sufficiently deep split from other branches such as Greek. Early Latin and Greek texts document what were, already by 2500 years ago, sub-lineages far diverged from each other, into fully-fledged, mutually unintelligible languages. And even amongst Greek dialects, their own divergence within their single clade already takes us back to at least 3000 bp, on standard thinking. That leaves just 1500 years for divergence vis‑à‑vis all other branches of Indo-European in Europe. As an informal but informative yardstick, consider how similar modern Italian and Spanish remain, some 2200 years after Roman expansion to Iberia. From that perspective, an expansion at barely double that time-depth, 4500 bp, looks suspiciously shallow for the entire, far greater diversity of Indo-European within Europe.
10. Asia: North vs. South, Tocharian and Indo-Iranic
Looking to Asia, meanwhile, Allentoft et al. (2015) point to the extinct Tocharian branch of Indo-European as the linguistic legacy of Afanisievo, as if that supports the Steppe hypothesis:
the existence of the Afanasievo culture near Altai around 3000 BC could also provide an explanation for the mysterious presence of one of the oldest Indo-European languages, Tocharian in the Tarim basin in China.
(Technically, it betrays some misconceptions to talk here of “one of the oldest Indo-European languages”. Languages have no age or date of birth: they are lineages constantly changing through time. Nor is attestation of Tocharian particularly old. To be correct, what this should say is that Tocharian’s sub-lineage is among the first to branch off from the rest of the family — at least on some analyses.)
In fact, in the Steppe hypothesis Afanasievo and other eastern Steppe cultures are supposed to be the linguistic ancestors of far more than just Tocharian: principally, of the main Iranic and Indic branches of Indo-European. If Allentoft et al. (2015) suggestion of Tocharian is all that the eastern Steppe cultures left us, then it is bad news indeed for the Steppe hypothesis. (One of its main weaknesses has always been the lack of a convincing archaeological scenario by which Steppe speakers crossed southwards into Iran and India.)
It also does not work well to suggest that Tocharian and Indo-Iranic could both be descendants of the Steppe cultures, since Tocharian is a ‘loner’ branch with nothing particular in common with Indo-Iranic (not even the traditional kentum/satem criterion). This, and Tocharian’s isolated, far-flung location in Xinjiang, has always implied an explanation in terms of a one-off, long-range, migration. If Tocharian is to be identified with Afanasievo, then on that count too, Afanasievo would seem an unlikely source of Indo-Iranic as well.
Moreover, far from leaving a great legacy that might include the huge Indo-Iranic branch of Indo-European, Allentoft et al. (2015) explicitly report that the genetic legacy of these Steppe cultures tended more to fade out (mirroring how Tocharian speech also died out?):
Andronovo was replaced by the Karasuk, Mezhovskaya, and Iron Age cultures which appear multiethnic and show gradual admixture with East Asians
11. Summary: Which Hypothesis Do the Genetic Data Really Fit Best?
Naturally, Haak et al. (2015) and Allentoft et al. (2015) have an interest in pushing the potential relevance of their results to the Indo-European question — and rightly so, since they bring revolutionary new data on some past population movements in Eurasia. Understandably, it can also seem attractive that their steppe DNA data may superficially appear to give a simple and powerful ‘result’. Nonetheless, both papers misconstrue the true nature and scale of the Indo-European question.
The latest ancient DNA data are no real support for the Steppe hypothesis, in fact. The ancient population movements they detect fall far short of identifying the Steppe as the origin point of the main, primary expansion of all branches of the entire Indo-European family, which is what the Steppe hypothesis claims. The ancient DNA record, hugely impressive as it already is, for now remains too limited, in time and space, vis‑à‑vis the much broader scope of Indo-European as a whole.
On closer inspection, these papers’ findings are just as compatible, if not more so, with the steppe movements they uncover being part of a secondary phase of expansion, which may well have further spread a few branches of Indo-European (plus some Uralic) into north-eastern Europe, and perhaps Tocharian eastwards — but only after most of the rest of the family had spread earlier with farming. The new data are highly coherent with the wider farming hypothesis, particularly its version A2.
12. What Next to Answer the Question? Ancient DNA from the South?
The Indo-European question remains very much open, then — but perhaps not for many years more. Spectacular advances in ancient DNA data, as published in papers such as these, are enriching and revolutionising our understanding of prehistory at a bewildering pace. They are advancing us in great strides towards an eventual resolution of the Indo-European question.
But these latest papers have focused thus far on the Steppe itself, and northern parts of Eurasia. To resolve the origins of humanity’s single most widely spoken language lineage, that is not yet enough. To that end, we still need more ancient DNA data, from a wider chronological span, and from much more of the Indo-European-speaking world, above all the ancient heartlands south of the Alps, Black Sea, Caspian and Himalayas. It was “the Sanskrit, the Greek and the Latin” that first opened Sir William Jones’s eyes to the Indo-European question; 229 years on, linguists should await those new genetic data with bated breath.
But hold on: can’t languages also spread ‘culturally’, through prestige, without migrations of people? In the modern world, a majority of human languages are already heading for extinction as their speakers shift en masse to other languages. But this is an abrupt reversal: a collapse and disappearance of the language diversity arisen over the preceding millennia. It is driven by the revolutionary changes of the Modern Era: the nation state, language standardisation, printing, mass literacy and transport, telecommunications and now globalisation. Those are no valid model for deep Indo-European prehistory, when human populations were just tiny fractions of today’s, and societies radically different in scale.
Even in fairly recent historical times, the ‘elite dominance’ model turns out to be a very poor predictor of language shift. Examples are nowhere more obvious than in historical steppe incursions upon farming Europe, India and China. Incoming, conquering steppe elites — e.g. the Bulgars, Mughals and Yuan Dynasty — far more often abandoned their own elite language, and switched to that of the existing local majority instead.
In publications as recent as Anthony & Ringe (2015: 210), however, prominent supporters explicitly reiterate that the Steppe hypothesis sees Indo-European spreading from the Steppes to Europe by …
“language shift … [that] must have depended little if at all on demographic advantages, as no strong demographic advantage can be assigned to any particular region in Copper and Bronze Age Europe … most steppe-origin scenarios for the IE expansion begin with the shift of this large Old European population to IE languages. But steppe herders certainly held no demographic advantage over the Old European population”.
There is some irony, then, to see genetics papers now ostensibly supporting the Steppe hypothesis, but only out of a conviction that there was in fact a Yamnaya > Corded Ware demographic replacement.
Certainly, the generalised appeal to ancient, mass language shift, without demographic advantage, to explain the most dominant of all the world’s language families has always struck sceptics of the Steppe hypothesis as extremely implausible. The doubts will hardly be dispelled if its advocates now change their mind to claim that demographic advantage does after all explain Indo-European — or at least, it does in those regions that do show powerful genetic impact from the steppe, but everywhere else, any contrary genetic signal can be disregarded, and elite dominance conveniently still invoked instead?
The scenarios in which major language shifts may be more or less plausible, and how those may have changed through prehistory, is a big topic in itself, but covered here (pp. 599 & 614-623). In this blog, the focus is just on assessing Haak et al. (2015) and Allentoft et al. (2015). Their claim to support the Steppe hypothesis on the basis of genetic data is of course effectively premised on the basic assumption that so vast a linguistic impact as Indo-European, at so early a stage in prehistory, would indeed have involved major population expansions. So it is on that basis that this blog has proceeded.
Addendum 2: What About the ‘Wheel’ Words?
While not to do with genetics, it is worth clarifying one of the main arguments made in favour of the Steppe hypothesis. Its supporters often point to a set of superficially beguiling claims about what linguistics can supposedly do not just to reconstruct linguistic characteristics of ancient languages, but even to engage in “cultural reconstruction” (also known as “linguistic palaeontology”).
Among the words that linguists can ‘reconstruct’ back to the Proto-Indo‑European language, for example, are a very few that in various branches of Indo-European languages came to refer to various parts of wheeled vehicles, although the specific meanings are actually often inconsistent: wheel, chariot, (life-)cycle, and so on. Enthusiasts for the Steppe hypothesis have taken this to prove that the speakers the Proto-Indo‑European must have already had wheeled vehicles. And since those did not develop until after the first spread of farming, this would mean that Indo-European could not have spread so early as with farming.
Unfortunately, the logic does not work. Linguists could also ‘reconstruct’ the word mouse back to Proto-Germanic on the basis of its computing sense alone; not that Proto-Germanic speakers had computers. Nor did their houses and ships look much like ours. Nor did the Romans have battery chargers, although those words reconstruct back to the Proto-Romance language.
“Cultural reconstruction” is no hard science. It is open to a great deal of subjective interpretation and interminable arguments, and not probative at all. For while linguists can reconstruct sounds reliably, by near exceptionless ‘sound laws’, we have no equivalent ‘meaning laws’ to be able to reconstruct *exact* meanings, especially not where referents themselves are necessarily changing, in processes of domestication or technological development. The supposed ‘wheel’ words actually go back to more general words for turn, rotate and walk, formed into words that look literally like ‘turn-turn (thing)’, for example.
Here are some other linguists’ views on “cultural reconstruction” / “linguistic palaeontology”:
It looks as if ‘wheel’ was not in the proto-lexicon and the various words for it were created independently after the dispersal. (Coleman 1988: 450)
If Renfrew were able to convince … that the first farmers were the only possible bearers of Proto-Indo-European, then philologists could probably explain away all the shared vocabulary that has seemed to imply later phases of civilization. (Sims-Williams 1998: 510)
The whole doctrine of making cultural inferences from linguistic evidence, known as linguistic palaeontology, has rarely enjoyed particularly high repute. (Anttila 1989: 373)
In practice, however, there are major pitfalls … and the reliability of the approach is questioned by many linguists. (Trask 2000: 198)
“naiveté [that] seems to enjoy the status of high acumen” (Pulgram 1958: 145-7)
For more on the pitfalls of linguistic palaeontology, see pages 607-610 of this paper.
Allentoft, M.E., Sikora, M., Sjögren, K.-G., Rasmussen, S., Rasmussen, M., Stenderup, J., Damgaard, P.B., Schroeder, H., Ahlström, T., Vinner, L., Malaspinas, A.-S., Margaryan, A., Higham, T., Chivall, D., Lynnerup, N., et al. 2015. Population genomics of Bronze Age Eurasia. Nature 522(7555): p.167–172. http://dx.doi.org/10.1038/nature14507
Anthony, D.W., & Ringe, D. 2015. The Indo-European homeland from linguistic and archaeological perspectives. Annual Review of Linguistics 1(1): p.199–219. http://dx.doi.org/10.1146/annurev-linguist-030514-124812
Anttila, R. 1989. Historical and Comparative Linguistics. Amsterdam: John Benjamins.
Brandt, G., Haak, W., Adler, C.J., Roth, C., Szécsényi-Nagy, A., Karimnia, S., Möller-Rieker, S., Meller, H., Ganslmeier, R., Friederich, S., Dresely, V., Nicklisch, N., Pickrell, J.K., Sirocko, F., Reich, D., et al. 2013. Ancient DNA reveals key stages in the formation of Central European mitochondrial genetic diversity. Science 342 (6155): p.257–261. http://dx.doi.org/10.1126/science.1241844
Coleman, R. 1988. Review of Renfrew (1987): Archaeology and Language. Current Anthropology 29(3): p.449–453. www.jstor.org/stable/2743460
Haak, W., Lazaridis, I., Patterson, N., Rohland, N., Mallick, S., Llamas, B., Brandt, G., Nordenfelt, S., Harney, E., Stewardson, K., Fu, Q., Mittnik, A., Bánffy, E., Economou, C., Francken, M., et al. 2015. Massive migration from the steppe was a source for Indo-European languages in Europe. Nature 522 (7555): p.207–211. http://dx.doi.org/10.1038/nature14317
Pulgram, E. 1958. The Tongues of Italy. Cambridge, MA: Harvard University Press.
Renfrew, C. 1987. Archaeology and Language: The Puzzle of Indo-European Origins. London: Jonathan Cape.
Sims-Williams, P. 1998. Genetics, linguistics, and prehistory: thinking big and thinking straight. Antiquity 72 (277): p.505–527. http://antiquity.ac.uk/ant/072/Ant0720505.htm
Trask, R.L. 2000. Dictionary of Historical and Comparative Linguistics. London: Routledge.