|
Post by Admin on May 23, 2018 18:49:07 GMT
David Anthony, Early Indo-European migrations, economies, and phylogenies, presented at the seminar "Tracing the Indo-Europeans: Origin and migration", organized by Roots of Europe - Language, Culture, and Migrations, University of Copenhagen, 12-14 December 2012 THREE REASONS FOR ARCHAEOLOGISTS TO CARE ABOUT THE INDO-EUROPEAN PROBLEM Reason #1: Proto-Indo-European is eloquent and real Language can illuminate the archaeological record in unique ways. Lan-guage is what makes humans human. Vocabularies and grammars contain a rich index of cultural beliefs and practices. Processual archaeology has long been criticized for neglecting the human subject/agent, and historical linguistics could allay some of those criticisms. A reconstructed vocabulary can reconstitute the meanings and cultural models that drove individual and group behavior in the world around the speakers of that vocabulary. Rather than continuing the rather sterile debates about the location of the PIE homeland, we should shift our attention to the cultural models and mean-ings implied by the PIE vocabulary. They are the real prize and the reason for an archaeologist to pursue the question of PIE origins. If we want to understand what the world meant from the subjective viewpoint of a pre-historic PIE-speaking agent, we can acquire a deeper understanding from an examination of his/her vocabulary and grammar (Benveniste 1973; Wat-kins 1995; Mallory and Adams 2006) than from a GIS viewshed analysis— although in an ideal world we would combine both sources of information. This has actually been done by archaeologists in the American Southwest. Here, reconstructed vocabularies and place-names, oral histories referring to prominent mountains, archaeological data including artifact types and architecture, biological skeletal data, and a GIS viewshed analysis of visible horizon features were combined by Bernardini et al. (2013), expanding on earlier work by Ortman (2012), to identify in an entirely new way the origin and route of prehistoric migrations of Tewa-speaking people. Language not only can be used by archaeologists, it is a rich source of information that deepens our understanding of the archaeological record and can be integrated with data gained from other sources such as GIS.But are proto-languages real? Can Proto-Indo-European be mapped and compared with archaeological cultures or other kinds of material-culture distributions? The linguist James Clackson (2007:16, 2012:265) warned against reifying PIE. He suggested that reconstructed PIE is like a constella-tion, composed of elements from different eras located at variable distances, an illusion that doesn’t exist in the real world. This metaphor permits PIE to float across chronological periods, and it reserves the study of PIE to an academically defined domain distant from actual time or geography—and, therefore, also from nationalism. The constellation analogy is reassuring politically, but is also misleading. A constellation has no effect on anything real. Proto-Indo-European is evident entirely through the profound, systematic, and quantifiable effects it had on the world’s languages. An actual language, PIE, must have been ancestral to the daughter IE languages, whose regular derivation from PIE is demonstrable by the comparative method. We can see distinct parts of PIE through different evidentiary lenses—phonology, syntax, morphology, vo-cabulary, poetic conventions, and comparative mythology—and they yield aligned perceptions (Watkins 1995). A constellation would disappear if the observer moved, but PIE gets more interesting when seen through differ-ent windows. Clackson did not say that PIE never existed, but only that it evolved and changed through time and that we cannot separate its earlier from its later materials prior to the appearance of its daughter languages in texts. This text-privileged view of language is not shared by all linguists. The chronological subgrouping problem has been addressed using both traditional (Meid 1975; Lehmann 1989) and new computational methods (Bouckaert et al. 2012; Jäger 2013; Chang et al. 2015). Clackson (2007:17) himself conceded that “it may be possible to assign some absolute dates to items of material culture, such as wheels [emphasis added] or the terminol-ogy for spinning wool.” Can we regard reconstructed PIE as if it were an actual language? First, obviously, the ancient language reflected in PIE is only partially recovered, so PIE is not a whole language; it is like a partial skeleton retaining some desiccated soft tissue. The core skeleton in the case of PIE is composed of reconstructed sounds. The units of sound reconstructed for each word root are accepted as being comparable to the units of sound in the actual ancestral word, although the exact phonetic realization would have varied. The sound system of reconstructed PIE is derived from the accumulated comparisons of thousands of individual cognates and yet presents a rela-tively coherent phonological system. Similar units of sound identify gram-matical categories that are marked in the same way with the same phono-logical endings and conjugations across the PIE grammatical system. PIE grammar and phonology constitute an unremarkable human language that falls within the observed range for known modern languages (Anthony and Ringe 2015). That is a strong indicator that what is reconstructed, while not a whole language, is the skeleton of one. Archaeologists can think of re-constructed PIE as representing the partial remains of a regional linguistic phase, like a regional phase in material culture, that contains traces of both chronological (diachronic) and geographic (dialectical) variation.
|
|
|
Post by Admin on May 24, 2018 18:34:02 GMT
2.1. The geographic distribution of the major Indo-European language branches at about 400 BCE, with the five firm PIE roots referring to wheeled vehicles and their attestation in each daughter branch indicated by symbols. The meaning of a proto-word produces more debate than its sound because meanings don’t change according to regular rules like those that control sound change. The irregularity of meaning-shift events discourages some linguists to the extent that they deny the possibility of ever recovering original proto-meanings beyond rather broad categories such as ‘something that revolves’ (Heggarty 2007:322) where other linguists would reconstruct ‘wheel.’ But most linguists argue that at least some specific proto-meanings can be recovered (Benveniste 1973; Beekes 1995; Gamkrelidze and Ivanov 1995; Fortson 2010; Koivulheto 2001; Watkins 1995; Zimmer 2009). An ex-ample is ‘axle’, which has the same specific meaning in its cognates in San-skrit (ákșas), Latin (axis), Lithuanian (ašìs), Germanic (eax), Celtic (echel), and Greek (áksɔ:n). Comparison of the sound changes in these cognates shows that, in each daughter branch, the sounds of each cognate can be derived in regular, predicted ways from a shared PIE root *h 2eḱ s-, which, therefore, existed in the late PIE parent from which all of these languages evolved (Mallory and Adams 2006:248; Anthony and Ringe 2015). Note that the Anatolian languages did not preserve a word derived from *h 2eḱ s, so Anatolian, the most archaic of the IE daughter branches (Jasanoff 2003), might have split away before axles were invented.The meaning ‘axle’ for the post-Anatolian root *h 2eḱ s- is strengthened by its inclusion in a semantic field, ‘parts of a wheeled vehicle’ (Fig. 2.1), that contains at least four other reconstructed roots with ‘established’ meanings in late PIE: two words for wheel, *Hrotós and *k wék wlos, one noun for har-ness pole, *h 2/3 éyH-os, and a verb meaning to go in a vehicle, *wé ǵ h-e-ti (An-thony and Ringe 2015; Mallory and Adams 2006:247–50; Gamkrelidze and Ivanov 1995:621–27; Garrett 2006:144–45). One of these, *k wék wlos, the PIE root from which English ‘wheel’ evolved, produced ‘wheel’ words across almost all of the IE daughters, but again excluding Anatolian. This root for ‘wheel’ was formed from a PIE verb *k wel-, ‘turn.’ Gray and Atkinson (2005) suggested that it was natural for people to select the verb *k wel-, ‘turn,’ as the root for *k wék wlos, ‘wheel, thing that turns,’ so it could have happened independently many times among long-dispersed IE daughter languages. They implied that the comparative method creates proto-roots and meanings that might not have existed in any real PIE vocabulary and that linguists might be unable to distinguish ‘real’ from apparent ancestral roots, a view aligned with that of Clackson (2007). But reconstructed PIE contains at least four different verb roots in the semantic field ‘turn,’ ‘spin,’ or ‘revolve’ (*k wel-,*h werg-,*wert ,*wel-), so the motivation for a shared ‘wheel’ word based on just one of these, *k wel-, was a specific, shared cultural choice, unlikely to have occurred repeatedly and independently among the dispersed daughters. Moreover, *k wék wlos was formed from *k wel- by reduplicating the initial ‘k’ and altering the sounds around it in an unusual manner, described by Ringe (Anthony and Ringe 2015:205) as: reduplication + zero-grade root + thematic vowel + nominative singular ending. No other noun in the PIE lexicon was made from a verb in that specific way. The duplication of the initial ‘k’ was like making an English noun ‘turn-er’ from the verb ‘turn’ by duplicating the ‘t’ and making ‘turn-ter.’ All of the IE daughters retained this unusual bit of playful reduplication connected with other unusual elements, which together make the formation of *k wék wlos unique within the IE languages. For dozens of widely separated languages to have independently created this unique word in each case when they needed a new word for ‘wheel’ would require a coincidence of staggering improbability (Anthony and Ringe 2015:204–5; Garret 2006:144–45). The alternative explanation, that *k wék wlos was part of the undifferentiated late PIE vocabulary, is the only viable explanation for the shared phonological root, which in each daughter retained specific meanings related to wheels and wheeled vehicles. Over the last 200 years, through debates such as this, linguists have re-fined a body of about 1500 unique roots, not all with established or clear meanings, but all analyzable as parts of the PIE vocabulary. Mallory and Adams (2006) have collected and grouped these very usefully by subject. The PIE vocabulary tells us that its speakers had not just wheeled vehicles, but institutionalized positions of leadership (at least three different roots that later ranged in meaning from ruler/director/king to ruler/master-over and ruler/strong-one) as well as an institution of clientship (literally ‘fol-lowers’). They had a kinship vocabulary most compatible with patrilineal, patrilocal lineages. They recognized the distinction between poverty and wealth and had bards who practiced an institutionalized form of elevated poetry that praised the gifts of immortal gods and mortal chiefs (Wat-kins 1995). They also had a rich vocabulary for strife (Mallory and Adams 2006:277–83), including four kinds of quarrel (accusation, argument, law-suit, and bluster/rage), a root meaning ‘restitution,’ and separate roots meaning harm, strike/wound, stab, fight, destroy, victory, defend, raid/ booty, and army/warrior group (two or three different roots presumably originally referring to different kinds of war-bands). Many of these aspects of the prehistoric past are simply not knowable through the medium of archaeological evidence. Proto-Indo-European is an eloquent source of in-formation about otherwise entirely prehistoric, text-less societies.
|
|
|
Post by Admin on May 25, 2018 18:19:09 GMT
Reason # 2: Proto-Indo-European can be mapped in time and space Of course, in order to associate a proto-language with archaeological evidence, we must be able to confidently locate the proto-language in time and space. The best spatiotemporal evidence comes from two sources. One is located inside the proto-language itself: reconstructed words that refer to faunal, floral, geographic, climatic, economic, and technological (like wheels) aspects of the environment in which the proto-language was spoken. The other is outside the proto-language: borrowed words from or shared inheritance with other proto-languages that, therefore, must have been geographic neighbors. In the case of Proto-Indo-European, words for wool, sheep, cattle, milk and milk products, and plow or ard show that PIE speakers were familiar with an agricultural economy—they were not hunter-gatherers (Mallory and Adams 2006:136–41, 164–69). Even the most archaic phase of PIE, pre-served uniquely in the Anatolian IE languages, was spoken after the inven-tion of agricultural economies. Cognates of PIE *k werus meant ‘cauldron’ in both Celtic and Sanskrit; and with other PIE words for containers this suggests a date after the invention of ceramic cooking vessels. Indo-Euro-pean cognates also referred to institutionalized social inequality as briefly described above (Anthony and Ringe 2015:210–11), and this indicates that PIE probably was spoken not only after the beginnings of agriculture, but also among people who had institutions that recognized and celebrated social inequality. The five late PIE (post-Anatolian) cognates for wheeled vehicles show that the speakers of late PIE knew about wheeled vehicles and used words created within the late PIE language community to name their different parts. Late PIE, therefore, was spoken after wheeled vehicles were invented. The invention of the wheel-and-axle principle is well dated. Wheels did not appear anywhere in the archaeological record or in inscriptions before the fourth millennium BCE, and almost all of the archaeological evidence is dated after 3500 BCE (Bakker et al. 1999; Fansa and Burmeister 2004). Late PIE, containing a not-yet differentiated vocabulary for wheeled vehicles, therefore existed as a single language community after 4000 BCE and prob-ably after 3500 BCE. Turning to geographic neighbors, loans from PIE into Proto-Uralic and structural similarities in pronouns between the Indo-European and Uralic languages show that PIE was spoken near Proto-Uralic (Parpola 2012; Koivulheto 2001; Janhunen 2001, 2000; Kallio 2001; Ringe 1997). Proto-Uralic was a language spoken by foragers, as we can see by the ab-sence of terms for domesticated animals (except dog) or agricultural tools or foods in the reconstructed Proto-Uralic lexicon. The Uralic languages are distributed across the northern forest zone of Eurasia, and proto-Uralic was spoken somewhere near the Urals by forest-zone hunters and fishers (Parpola 2012). In addition, loans into PIE from a pre-Kartvelian language show that PIE was spoken in a region bordering the Caucasus (Gamkrelidze and Ivanov 1995; Nichols 1997). The Pontic-Caspian steppes lie between the Caucasus and the Uralic forest zone, the most plausible location for PIE given these internal constraints from shared loans and/or inheritances with neighbors. The IE languages of Anatolia retained archaic grammar and phonol-ogy and lacked a clear wheel vocabulary. The absence of a wheeled-vehicle vocabulary is consistent with many other signs of archaism in Anatolian. If the language that was ancestral to Anatolian separated before wheels were invented, the archaeological correlate of that separation could be repre-sented by the documented migrations of the Suvorovo/Skelya/Sredni Stog cultures from the Ukrainian steppes into the lower Danube valley and the Balkans about 4400–4200 BCE. These migrations were associated with the abandonment of 500–600 tells and the end of the ‘Old European’ Copper Age tell cultures of the Karanovo VI/Varna/ Gumelnitsa type, after which the steppe-influenced Cernavoda I culture appeared in the lower Danube valley (Ivanova 2007; Bicbaev 2010; Anthony 2007, 2013; Krauss et al. 2016).
|
|
|
Post by Admin on May 26, 2018 18:23:34 GMT
All other IE daughter branches are descended from a language community that remained in the steppes until after wheeled vehicles were adopted, an event that cannot be dated earlier than 4000–3500 BCE. A deeper chronology was suggested by Renfrew (1987; 2002) and Bell-wood (2013) who proposed that the expansion and differentiation of the IE daughter branches began about 6500 BCE with the spread of the Neolithic farming economy from Anatolia to Greece. The Anatolian hypothesis ac-quired strong and independent support from Bouckaert et al. (2012), who refined methods introduced by Gray and Atkinson (2003). Computational phylogenetic linguistics borrows methods from biological cladistics to ana-lyze the evolutionary sequence and relationships of the IE languages as if they were a mutating virus that evolved as it spread through contagion. These methods had successfully identified the origins, relationships, and relative chronologies of mutating viruses and were modified to identify the origins, relationships, and relative chronologies of mutating languages in the IE language family, an interesting but imperfect analogy (Pereltsvaig and Lewis 2015:127ff.). Bouckaert et al. (2012) assigned estimated date ranges to branch separations or splits and concluded that the best date for the initial split between Anatolian and the other daughters was around 6500 BCE, congruent with the archaeological dating of the migration of the first farmers from Anatolia to Greece. But the dates estimated by Bouckaert et al. (2012) for all later branching events in their phylogeny clustered between 4500–3500 BCE, implying an unexplained delay of 2000 years between the first branching event and any later IE branches. These later dates are much too late to represent any of the archaeologically dated migrations of Neolithic populations from Greece to the rest of Europe (beginning 6200 BCE for the Starčevo/Criş migration into southeastern Europe, 5800 BCE for the Cardial-Impressed migration into the western Mediterranean, and 5500 BCE for the Linear Pottery Cul-ture (LBK) spread into northern Europe). Moreover, the dates estimated by Bouckaert et al. for the later branch separations are not congruent with any archaeologically dated expansions outward from Greece—in fact, this is a time when the Greek Late Neolithic population declined precipitously. Overall, the Bouckaert et al. phylogenetic chronology does not correlate well with the archaeological dates for agricultural expansions in Europe. In addition, Neolithic farmers belonging to cultures such as the LBK or Cardial Neolithic certainly lacked wheeled vehicles. Whereas, the speakers of late PIE, after the separation of pre-Anatolian, just as certainly possessed wheeled vehicles. Bouckaert et al. (2012) did not address this seemingly in-superable chronological problem.Other teams of computational linguists (Jäger 2013; Chang et al. 2015) used Bouckaert et al.’s methods with different constraints and smoothing algorithms and found that the chronological aspect of the phylogeny was vulnerable to significant shifts with small changes in the algorithms applied. Chang et al. (2015), using what they saw as more realistic constraints, es-timated a more recent date for the initial differentiation of PIE, after 4500 BCE, in line with the steppe hypothesis and the wheel vocabulary. Compu-tational phylogenetic approaches to language evolution represent a new set of methods that might at some point provide archaeologists with an inde-pendent way to date the nodes and splits in the evolution of the IE daugh-ter branches and could conceivably help to map the geographic spread of languages (if languages can be said to spread by contagion, a problematic assumption), but it is a field that began only in 2002 (Ringe et al. 2002), and Bouckaert et al.’s (2012) refinement of the method contains assumptions that are just now being tested by Chang et al. (2015). Traditional linguistic approaches to identifying language homelands using time-sensitive and place-sensitive vocabulary and inter-language-fam-ily loans and relationships remain central to any solution of the problem of Indo-European origins. They suggest a homeland in the steppes after 4000–3500 BCE for the post-Anatolian phase of PIE. The Yamnaya culture appeared in the steppes about 3300 BCE and introduced a new, mobile form of nomadic pastoralism that depended on a novel combination of wheeled vehicles—which carried food, water, and shelter—and horseback riding— which made large-scale animal herding more efficient (Anthony and Brown 2011). The spread across the steppes of this new economy, with the social and political innovations it required, probably was the vector for the initial spread of the late PIE language within the western steppes. The steppe so-lution has support also from new discoveries in ancient DNA.
|
|
|
Post by Admin on May 27, 2018 18:24:24 GMT
Reason #3: ancient DNA, archaeology, and language In the preliterate past, languages must have spread, at least to some extent, with migrants, and, thus, the study of ancient human DNA (aDNA), which can trace the migration of people, is relevant to evaluating different models of language dispersal by comparing the sequence of language splits to dated population expansions. Because aDNA is derived from graves that are assignable to specific regional-chronological archaeological groups and from human bones dateable by radiocarbon, it can be directly linked to archaeological evidence. It can also be used to distinguish between diffusion and migration, and to specify the degree of admixture at the destination. Archaeologists can identify regional gene pools and movement between them, and the extent of intermarriage or replacement. We no longer have to use artifact styles and types as proxies for people, a correlation we knew was often spurious and un-supported, but instead can consider type/style as an independent set of variables, to be measured against biological identity in specific cases. We have never before had such a precise tool for analyzing prehistoric migrations and kinship. Since about 2005, geneticists have achieved consistently reliable results identifying aDNA from human bones. Many studies between 2005 and 2010 were confined to easier-to-recover mtDNA (which is inherited only in the maternal line, so indicates maternal ancestry), but improved methods rapidly encompassed paternal lineage indicators on the Y chromosome, permitting analysts to compare paternal and maternal lineages in ancient humans (Lacan et al. 2011). The autosomal nuclear genes that contain the code for a whole human genome were more difficult to recover, since the human genome is much bigger than that of our cellular mitochondria, and in ancient samples the human nuclear aDNA is fragmented into tiny microscopic pieces that must be separated from the fragmented aDNA of microbes, which also inhabit the sample (usually composing 99% of the de-tected DNA). Only then can the human aDNA fragments above a minimal size (now about 30 base-pairs) be virtually reassembled into their proper sequences and places on the human genome, a time-consuming procedure with a genome 3.2 billion base-pairs long. A method used in David Reich’s laboratory at the Harvard Medical School and at the Max Planck Institute of Genomics laboratory at Jena, focused on 390,000 (recently up to 1.2 million) single-nucleotide polymorphisms (SNP’s) distributed across the human genome, representing the parts that vary significantly between modern human populations (Haak et al. 2015). This selected sample of genes is obtained and analyzed much more rapidly (a median of 262x less sequencing time) than an actual whole genome of 3.2 billion base-pairs, permitting many ancient individuals to be analyzed at the whole-genome level—230 prehistoric/protohistoric European individuals in the most recent study (Mathiesen et al. 2015). The initial results challenge established ideas about European prehistory, but also open up entirely new areas of interpretation. Preliminary results from what is now the Harvard-Jena group, which shares methods and samples (Haak et al. 2015; Mathieson et al. 2015) and simultaneously from Willerslev’s team in Copenhagen, using different meth-ods and a different sample of prehistoric individuals (Allentoft et al. 2015), agree that European Mesolithic hunter-gatherers can be divided into at least three regional gene pools. It is possible that these represent three regional samples extracted from what was, in reality, more like a cline. The west-ern group (Western Hunter-Gatherers or WHG) occupied most of Europe, from Spain to Luxembourg, Germany, Hungary, and Croatia; while the northern group (Scandinavian Hunter-Gatherers or SHG) lived in Sweden and Norway; and the eastern group (Eastern Hunter Gatherers or EHG) lived in the forest and forest-steppe zone of Russia and extended south into the steppes (Haak et al. 2015) and perhaps into the lower Danube valley. Strikingly similar mtDNA haplogroups (U4, U5) were shared between all three forager populations, but the patrilineal Y-chromosome markers differed regionally, perhaps suggesting that long-distance wife exchanges oc-curred at least occasionally, while men were less mobile. The EHG con-tained a strong ancestral element from an Upper Paleolithic population represented archaeologically by a 22,000-year-old individual from Mal’ta, a site near Lake Baikal (Raghavan et al. 2014), referred to as Ancient North Eurasian (ANE).
|
|