Mallory's Revised Views On Tocharian Origins

Months ago, I mentioned that leading Indo-Europeanist Mallory had a meaty new paper on Tocharian origins. But, I didn't dig into his conclusions at the time.

In essence, he has lost the courage of his convictions in his earlier solution, in part because he has simply become more skeptical, and in part, because he is troubled by the source of cereal growing in the Tocharian culture when there is no archaeological evidence that the predecessor culture he has assigned to the Tocharians midway into their route, engaged in this kind of cereal growing. He expounds on this particular problem in a recent short paper:

The critical issue for these models is that while any and all of them could explain the distribution of domestic animal names, there are serious problems involved with the spread of arable agriculture. As Anthony remarks in this symposium, there is really no serious evidence for arable agriculture (domestic cereals) east of the Dnieper until after c 2000 BCE (see also Ryabogina & Ivanov 2011; Mallory, in press:a). This means that there is also no evidence for domestic cereals in the Asiatic steppe until the Late Bronze Age (Andronovo etc). From the perspective of the Pontic-Caspian model, the ancestors of the Indo-Iranians and Tokharians should not cross the Ural before c 2000 BCE at the very earliest. Hypotheses linking the Tokharians to earlier eastward steppe expansions associated with the Afanasievo or Okunevo cultures of the Yenisei or Altai (Mallory and Mair 2000) become very difficult if not impossible to sustain (as long as there is no evidence of arable agriculture in these cultures) as Tokharian retains elements of the Indo-European agricultural vocabulary. Of course, it should be emphasized that sites of the Afanasievo and Okunevo cultures are overwhelmingly burials that hardly provide the context in which one expects to recover the remains of domestic cereals; moreover, there is no evidence that any of these sites have been excavated in such a way that the recovery of seeds is likely. On the other hand, domestic cereals have been recovered from the site of Begash in the Jungghar mountains at dates of c 2300 BCE (Frachetti 2012) although this site is not connected (so far as we know) with the steppe trajectory of sites (Afanasievo, Okunevo). 

If this were not bad enough, it is also difficult to map the agricultural vocabulary across a Pontic-Caspian homeland within Europe itself. Main elements of the scheme suggested by Nikolai Merpert in 1977 still appear to be valid in current models of the evolution of steppe cultures involving an east (Volga-Ural) to west (Dnieper) cultural trajectory but if there was little or no agriculture east of the Dnieper, then how can we describe the eastern archaeological cultures of the Don (Repin), Volga (Khvalynsk) or the entire Don-Ural region (Yamnaya) as Indo-European if they lacked arable agriculture? That the steppe populations exploited wild plants such as Chenopodium and Amaranthus is well known and while this might explain the ambivalence of some of the cereal names to reflect a specific cereal type (rather than just ‘grain’) we would still need to explain why the semantic variance among cognate words is largely confined to ‘wheat’, ‘barley’ and ‘millet’ as if at least one of these was the original referent (and not some wild grain). All of the above problems would also be inherent in Renfrew’s revised version of the Anatolian homeland model that requires the eastern IndoEuropeans (Indo-Iranians, Tokharians) to pass through the Pontic-Caspian steppe. 

Conversely, the Near Eastern model, that requires the ancestors of the ‘ancient European’ languages to wander through Central Asia, cannot place the ‘Europeans’ north of south Central Asia before c 2000 BCE at the earliest. This is going to render the Indo-Europeanization of Corded Ware horizon that in almost every way imaginable would appear to be archaeologically, spatially and culturally a part of the Indo-European world. More importantly, it creates a ‘bottle-neck’ for the Northwest (?) Indo-European languages dated to about 1500 BCE where they all should have passed from east to west across the Pontic-Caspian and on into Europe. To propose a common secondary home and time depth for Balto-Slavic, Germanic, Celtic and Italic so late leaves hardly any time at all to explain the phylogeny of the European languages and how they arrived in their historical seats. If supporters of this model sought an escape route from the situation they seem to have created for themselves, one might possibly propose the route north through the Caucasus to explain not only Iranians (at Sintashta in Grigoriev’s account) but the rest of the Europeans. However, this is hardly without problems as well as one must also explain how the ancestors of most of the European languages managed to pass through the Caucasus without leaving a trail of European languages. 

If there are any lessons to be learned, it is that every model of Indo-European origins can be found to reveal serious deficiencies as we increase our scrutiny. One is reminded of Daniel Kahneman’s observation: 

“It is the consistency of the information that matters for a good story, not its completeness. Indeed, you will often find that knowing little makes it easier to fit everything you know into a coherent pattern” (Kahneman 2011, 87). 

The problem here, of course, is that over time we have come to know more and more and that our earlier, simpler and more alluring narratives of Indo-European origins and dispersals are all falling victim to our increasing knowledge. We have obviously moved on from the time when Nikolai Merpert first published his analyses of the role of the steppelands within the context of the Indo-European homeland but it is evident that we still have a very long way to go.

J. P. Mallory, "Twenty-first century clouds over Indo-European homelands" (Conference Presentation in Moscow, September 12, 2012).

Another major issue which this conference talk discusses is the relationship of the Anatolian languages to the other Indo-European languages.

Pontic-Caspian. In this model the linguistic ancestors of Anatolian are seen to depart earliest from a homeland north of the Black Sea where they pass through the Balkans (Mallory 1989: 241; Anthony 2007: 259) and, by the beginning of the Bronze Age (depending on which archaeological scenario one wishes to invoke) they enter Anatolia to settle and eventually dominate local non-Indo-European populations such as the Hatti. Later, within the PonticCaspian homeland, Brugmanian or mature Proto-Indo-European develops. Subsequent migrations carry ancestors of most of the European languages into central and northern Europe while ultimately the linguistic ancestors of the Greeks and Indo-Iranians disperse both west and east during the Bronze Age. These later migrations would also include the ancestors of the Phrygians and Armenians, two other language groups that occupied Anatolia but cannot be regarded as ‘Anatolian’ in the linguistic sense. Whatever the archaeological merits of this argument, this homeland does account for the division between Anatolian and the other IndoEuropean languages.  

Near Eastern. Although the supporters of this theory may differ in detail, they are at pains to provide a model that allows Anatolian to develop independently of the rest of the IndoEuropean languages who can evolve together. For example, in Grigoriev’s model, the ancestors of the Anatolians move from Anatolia into the Balkans while there are subsequent linguistic developments in eastern Anatolia that can account for the shared development of the other IE languages (Grigoriev 2002: 354–357, 412–415). Later, Anatolian relocates back to Anatolia during the Bronze Age while the ancestors of the Greeks (at least some of them) may have made their way through the Caucasus and into the Balkans. Thus this model also meets the minimal requirement of explaining the first element of Indo-European phylogeny, the separation of Anatolian from the rest of Indo-European although the subsequent movements of the other IE languages appear far more complicated than those proposed in the Pontic-Caspian model.  

Anatolian Neolithic. In Renfrew’s (1999) revised model (Plan B), Anatolian remains within the homeland while the rest of the Indo-European languages disperse into Europe which would again permit Proto-Indo-European to evolve separately from Anatolian and Phrygian and Armenian could later ‘return’ to Anatolia. As for the Asian languages, this model is not significantly different from the Pontic-Caspian model. This model then also provides a possible spatial solution to the initial break-up of Indo-European.  

On the other hand, the recent hypothesis of Bouckaert et al (2012) deals with the split between Anatolian and the other Indo-European languages in a very different way. It appears to situate the homeland (and Proto-Anatolian) in Anatolia. With Anatolian emerging in the centre, the European Indo-Europeans (ancestors of Greek, Latin, etc) disperse westwards into Europe through the Aegean and Balkans and the Asiatic Indo-Europeans (Indo-Iranians) move eastwards towards the Indus, i.e. there is a symmetrical ‘big bang’ from a homeland identical to the later historical seats of the Anatolian languages. It seems to me that there is no attempt whatsoever to deal with the division between Anatolian and the other Indo-European languages that, according to the authors’ own chronology, arise millennia later. Particularly noticeable is that this model appears to situate the ancestors of Greek in Greece to the west of Proto-Anatolian and the ancestors of Indo-Iranian far to the east of Anatolian thus preventing both branches from sharing the 2500 years of common development that is required by this model’s own chronology and phylogeny. How is one to explain parallel linguistic innovations both to the east and west of the region assigned to proto-Anatolian? The statisticians who devised this model seem to require some form of mutual contact at a distance, one of the stranger aspects of quantum theory that Einstein once dismissed as Spukhaftige Fernwirkung. It is difficult to see how one can resolve this problem without either revising the model so that all the rest of the Indo-European languages are ejected in a single direction from Anatolia or creating a complexity of movements within Anatolia. The first solution is indeed the one that Renfrew adopted in his revision that appears to be contradicted by the Bouckaert model; the second solution is more difficult to imagine as the time depth involved would appear to anchor the model with the spread of agriculture to Greece in the 7th millennium BCE (and hence force the ancestors of the Indo-Iranians into the same process if they must evolve along with the Proto-Greeks?) or disassociate the Greek movements from the Neolithic to a later period (the Bronze Age?) which will still require one to somehow connect their putative development with the Proto-Indo-Iranians. One might try to employ Robert Drews’ (1988) ingenious chariot model with the spread of Greeks and Indo-Iranians set to c 1600 BCE and linked to chariot warfare but this would bring Bouckaert’s entire chronology of the Indo-European languages into total disarray. In short, this is the one model that does not seem to address the only feature of Indo-European phylogeny that has near universal support. 

There is naturally an alternative view of Anatolian that does not support its relatively great antiquity but rather explains its absence of features found in other IE languages as ‘losses’. As these are generally explained by losses occasioned by the impact of a non-IE substrate on Anatolian within Anatolia itself, this alternative model can hardly support the notion that the Indo-European homeland was within Anatolia.

I personally feel that the archaeological evidence strongly undermines an origin for the Anatolian languages that is more than a century or two older than Greek or Indo-Iranian, and that a very different non-Indo-European substrate for the Anatolian languages than the substrate of the other Indo-European languages (or almost no substrate at all in the case of Tocharian) is the best explanation. The empirical evidence also supports rather strongly the hypothesis that language drift alone is very modest in the absence of contact with other languages.

Likewise, I personally think that he is not justified in abandoning the conclusion that the Tarim mummies are early Tocharians, and doubting his original hypothesis that Indo-Iranian influences are late contacts with the Saka. This still requires us to find a source for their cereal technology sometime between the arrival of the Tarim mummies in the Tarim basin, ca. 2000 BCE, and their departure from the proto-Indo-Europeans, realistically sometime between 4000 BCE and 2100 BCE, which a linguistic bias towards an older date, and a cereal technology bias towards a later date. But, this is a far more manageable riddle than the epic questioning of all of his assumptions that he raises in this 2015 paper.

A long, rapid cross county march of thousands of miles at a more or less constant latitude and climate, across the steppe, by people who can ride horses and have wagons has multiple ample precedents. The Turks and the Mongols did so in the other direction. The Russians reversed the tide. Pioneers in the 19th century in the United States did the same thing. A smaller temporal distance between the Tarim Basin mummies viewed as early Tocharians and the predecessor culture shared by the Celtic and Italic Indo-Europeans who surged into Western Europe and Italy respectively, also makes cultural similarities between the two like "witch hats" and primitive tartan style weaving patterns, much easier to understand.

Renfrew’s Neolithic Anatolian hypothesis, in contrast, is confounded by the existence of common vocabulary for horses and wagons that weren't invented at the time, by the glacial pace of linguistic change it implies contrary to all empirical data on the subject, by the historically attested predominance of non-Indo-European languages in Anatolia prior to 2000 BCE, and by ancient DNA evidence which strongly supports a steppe origin and demic diffusion model of Indo-European expansion.

This paper has not only failed to provide a solution to the problem of Tocharian origins—it has even helped undermine the author’s earlier solution (Mallory and Mair 2000). Many of the inadequate solutions to the problem of Tocharian origins probably stem from a tendency to take unacceptable shortcuts in developing arguments (e.g., Tocharians are “Westerners,” the Tarim mummies are “Westerners,” therefore, the Tarim mummies must be Tocharians). The tendency has been, at least to some extent, driven by the sheer lack of archaeological evidence, but this is now being dramatically redressed by archaeologists working in Xinjiang. In tackling the issue anew I have tried to approach the entire problem more systematically by listing the criteria that I believe are required of any archaeological solution to the problem. I briefly revisit these below. To reach a solution, it would be necessary:  

1. To establish the physical and cultural remains of known historical Tocharian-speaking peoples. It can be seen that at the present we lack the type of Tocharian archaeology that would permit us to discern the historical Tocharians through the veneer of urbanism or Buddhism and the international styles in which these were expressed. This might be resolved, at least to some extent, if we had better knowledge of the formation of the northern oasis towns comparable to what we now have for Yumulak Kum.  

2. To trace the physical and cultural remains of historical Tocharians retrospectively into the prehistoric period. The key obstruction here is that there does not seem to be a valid way by which we might distinguish between a prehistoric speaker of Iranian and one of Proto-Tocharian. The material culture of the Iron Age cemeteries and the few Bronze Age settlements of the region can all be attributed to the Iranians. This, it should be emphasized, is not an issue of “fact” but rather the paradigm within which we seem to be trapped, i.e., any cultural connection with the north or west appears to be with a region where we expect to find Iranian speakers. The reconstructed lexicon also does not serve us very well in determining any critical differences between the reconstructed culture of the Iranians and Tocharians. To be sure, we could adduce finer levels of cultural comparison. For example, we reconstruct for the Indo-Iranians a sacred drink (soma/haoma) which, if its botanical identification among the Iranians be a guide, was ephedra, which is known from special vessels in the BMAC of Turkmenistan, as well as its ubiquity in the Xiaohe horizon (Parpola 2012a, 250–251). If we applied the use of ephedra as a cultural indicator of the Iranians, we would then have to place them in the Tarim Basin by 2000 BCE, which would further reduce any windows for identifying Proto-Tocharians.  

3. To trace the ancestors of the “prehistoric Tocharians” to a location outside of Xinjiang. The obvious issue here is that if we cannot identify prehistoric Tocharians within Xinjiang, then it will be impossible to tie them to an external origin. This paper has suggested that we seem to be dealing with one of the following phenomena: 

a. The prehistoric Tocharians are already well known to us from Iron Age cemeteries such as Chawuhugoukou, and their external origins can be traced, perhaps, to the Eurasian steppe cultures of the Iron Age, where we find similar burial practices. In this way the Tocharians are simply a linguistic group who occupied the Eurasian steppe and maintained their language although they apparently absorbed much of their material culture and behavior from their Iranian neighbors. Although Tremblay’s conclusion that the earliest evidence for Tocharian-Iranian contacts coincides with his “Old Sakan,” this does not necessarily require that these contacts took place exclusively within the Tarim Basin, as these contacts could have occurred further north in the steppelands during the period that Parpola (2012a, 223) assigns to Late Proto-Iranian (1500–1000 BCE), which should coincide with the date of early Eastern Iranian. 

The problem with this model is that it implies that the ancestors of the Tocharians were situated somewhere where they avoided contacts with earlier stages of Indo-Iranian and only began their association with this branch later, when the Eastern Iranians had emerged. In short, it does not indicate where the Tocharians might have been before they came into contact with the Saka, so it does not really get us much further toward Tocharian origins. 

b. The prehistoric Tocharians are already well known to us from Iron Age cemeteries such as Chawuhugoukou. The Tocharians had appeared south of the Tianshan much earlier than the Iron Age but absorbed material culture and behavior from Iranians over the course of the Early Iron Age within Xinjiang, so that by the Iron Age they were indistinguishable from them archaeologically. Any search for Tocharian origins should be rooted in earlier cultures, e.g., Xiaohe, and not those of the Iron Age. In short, just as the historical Tocharians are viewed through an urban Buddhist filter, the Iron Age Tocharians are obscured by an Iranian filter. 

One of the obvious problems with this model is that it requires us to establish the origin of the Xiaohe horizon, and, while there are a few comparisons that can be made with cultures from the Altai and Minusinsk Basin, the evidence is not really convincing. Moreover, it is also clear that Xiaohe is entangled with the spread of domestic cereals from west to east or, at least, a recipient of such exchanges, and we are still far from establishing the precise route of this dispersal. 

c. The prehistoric Tocharians are basically unknown to us. They moved into the oases along the Tarim river and are buried under the foundations of the now increasingly modern towns of Xinjiang. In short, we have been looking in the wrong place. 

The problem with this approach is that we do not expect the urban centers of the northern Tarim to date much earlier than the first centuries BCE, and it seems unlikely that Proto-Tocharian entered the Tarim Basin so late. It seems far more likely that they entered earlier than this and are to be accounted for in hypothesis (a) or (b) above.  

4. To trace the cultural path of the prehistoric Tocharians back to a geographical source congruent with their position (temporal and spatial) within the Indo-European language family. Without a firm anchor immediately outside the Tarim Basin one can hardly evaluate Tocharian dispersals within the general framework of Indo-European expansions. From an archaeological perspective, there appear to be (at least) three competing models.  

a. The Eurasian steppe model (Early Bronze Age) that sets the Indo-European homeland in the Pontic-Caspian region and identifies the ancestors of the Tocharians as members of the earliest eastward expansion of steppe pastoralists from the Urals eastwards to the Altai and Yenisei, i.e., the Afanasievo culture (Mallory and Mair 2000; Anthony 2007, 307–311). This model satisfies those who regard Tocharian as a very early departed language, geographically peripheral to the other Indo-European branches, and eliminates the problem of dating contacts between Tocharians and Indo-Iranians to any period earlier than the entry of the Saka into the Tarim Basin. Among its major problems are: 1) it lacks any evidence of the suite of domestic cereals which the ancestors of the Tocharians should have known; 2) while there may be some Afanasievo artifacts associated with the Qiemu’erqieke culture in the Junghhar basin, these are really totally different cultures, so there is no evidence for an Afanasievo migration south through the Junghhar Basin towards the land of the historical Tocharians; 3) the archaeological case for contacts between the Afanasievo and later Okunevo cultures with the Early Bronze Age culture of the Tarim Basin (Xiaohe) is, other than burial posture, generally weak and circumstantial.  

b. The Eurasian steppe model (Middle/Later Bronze Age) that sets the Indo-European homeland in the Pontic-Caspian region but identifies the ancestors of the Tocharians as a later Bronze Age phenomenon that followed after the Afanasievo culture, e.g., an element of the Andronovo culture or some other later culture (Kleyn’s Fatyanovo> Karasuk; also Kristinsson 2012). This type of solution might satisfy those who prefer to see the ancestors of the Tocharians more closely related to the European languages, and if the Tocharians had adopted the material culture of steppe Iranians, that makes it easier to argue for an immigration of Eurasian steppe populations into the Tarim Basin, since there is abundant evidence for Andronovo and Karasuk material culture in Xinjiang. By the later Bronze Age period one might also believe that the issue of cereal agriculture could be more easily addressed, as there is evidence for cereal agriculture among the Andronovo tribes. Among its major problems are: 1) it sets Tocharian origins in Europe in geographical areas (northern Europe, central Europe, the Balkans, the forest area of Russia) from which we would far more easily derive the ancestors of the various European branches; 2) other than Kleyn’s attempt to tie the material culture of Fatyanovo with Karasuk, there is no archaeological evidence adduced to support such late migrations of Tocharians across the eastern steppe; 3) it generally results in the Tocharians occupying an archaeological staging area (Andronovo, Karasuk) that we would otherwise naturally assign to the Iranians or some other group. It is true that Andronovo is a higher taxonomic label than “culture” and could well embrace a variety of languages or language groups over its vast area, but, if the Proto-Tocharians were an element of the Andronovo cultural historical region (Kristinsson 2012), we will need to explain why they borrowed vocabulary only from a presumably later sub-branch of Eastern Iranian rather than Indo-Iranian itself. Moreover, we would also need to describe how they came to be absorbed into the Andronovo world and where their place of origin was before this happened.  

c. The Central Asian model sets the Indo-European homeland anywhere from eastern Anatolia (Gamkrelidze and Ivanov 1995) to Central Asia (Sverchkov 2012 / Сверчков 2012). The potential advantages of this model are that it locates the ancestors of the Tocharians closer to the Tarim Basin and so reduces the length of any migration; it can accommodate cladistics that either place the ancestors of the Tocharians on the periphery (but only if one presumes that the Afanasievo culture is more closely tied with Central Asia than the European steppe), or it can position the Tocharians geographically adjacent to Europeans, who are then presumed to enter Europe from southeast of the Urals. As the earliest wheat in China is identified as bread wheat, the same type as grown in south Central Asia, it provides a more convincing link between China and the West in terms of cereals than does the Eurasian steppe. Among its problems are: 1) it fails to provide convincing evidence that the European steppe cultures (which are integral to this model, which also must explain the spread of the Indo-European languages all over Europe) are derived from east of the Caspian Sea— there are, for example, far more proximate and believable sources for the earliest domestic animals and cereals of the steppe region in the areas adjacent to the European steppe than can be supplied by the Central Asian model (Mallory 2014); 2) the earliest horizon of Central Asian expansions across Central Asia and south Siberia are associated with the Kelteminar culture, which suffers from the same absence of domestic cereals as does the Early Bronze Age Eurasian steppe cultures.  

d. The combined Steppe and Central Asian model that sets the Indo-European homeland in the Pontic-Caspian but argues that steppe populations intruding into the indigenous agricultural societies of Central Asia adopted many elements of material culture without undergoing language shift. This model is employed, for example, to explain the Indo-Aryans or segments of the Iranians as the hybrids of the Andronovo and the BMAC of Turkmenistan (e.g., Mallory 1998). The advantage of such a model would be that it would allow Eurasian pastoralists to maintain contacts with settled farmers and, presumably, assist them in the retention of the inherited Indo-European vocabulary concerning agriculture and domestic plants. Moreover, it might also provide a “steppe” connection to any model such as Sverchkov’s that derives the Tocharians from Central Asia. This would require the entry of steppe people prior to the Andronovo culture.  

Evidence for earlier contacts between Eurasian steppe cultures and the settled farmers of south Central Asia is sparse, but it does exist. Afanasievo sherds, for example, have been reported from Gonur Depe in Turkmenistan (Avanesova and Dzhurakulova 2008, 28 / Аванессова и Джуракова 2008, 28) and, more importantly, there are the finds of both Afanasievo (of Yenisei provenance) and Yamnaya remains in a ritual complex at Zhukov, 16 km from Samarkand (Avanesova and Dzhurakulova 2008 / Аванессова и Джуракова 2008), coupled with material compatible with Sarazm II (c 3200–2900 BCE). The complex can be compared with similar Afanasievo cult sites in both the Yenisei and Altai (described in Parzinger 2006, 191, 197). The excavators also see such evidence as adumbrating the Afanasievo-compatible burials of Sarazm IV (c 2300–2000 BCE). No one seems certain precisely how one might link the European steppe, the Zervashan Valley of Tajikistan and the Minusinsk Basin together (mobile traders from the European steppe, a single interaction sphere of exchange relationships, Frachetti’s “Intermountain Corridor”?), but there is clearly evidence in both the Afanasievo and subsequent Okunevo periods for some form of mutual contact. As I indicated above, the reason for suggesting this model is that it places steppe populations in an area where cereal agriculture was well established, so it reduces both the spatial and temporal lacuna between their homes in the Pontic-Caspian region and their possible approach to the Tarim Basin. Unfortunately, the spatial and temporal lacuna with respect to domestic plants now appears not merely between the Urals and the Altai but even farther, between the Dnieper and the Altai (Mallory 2014). I do not know how we are going to be able to resolve these issues, but if we really want to trace the Tocharians to their origins we might paraphrase the immortal lines of ‘Deep Throat’ and “follow the cereals.” 

J. P. Mallory, “The Problem of Tocharian Origins” Sino-Platonic Papers, 259 (November 2015) at 46-52.