There is something kind of amusing about the fact that the language we call Tocharian is, by most accounts, not actually the language of the Tocharians. It is one of those historical accidents that, having survived long enough, has acquired the authority of an established fact: a name proposed by scholars in the early twentieth century, based on an identification that subsequent scholarship has largely rejected, which nevertheless remains in use because, by now, changing it would probably cause more trouble than continuing to tolerate it. I find this particularly entertaining because Tocharian is already a rather extraordinary language to study. It is an extinct branch of the Indo-European family, known almost entirely through fragments of manuscripts discovered along the northern edge of the Tarim Basin in what is now northwestern China, thousands of miles from the regions with which most people would associate the ancient Indo-European languages. And yet, despite the remarkable amount of work that has gone into reconstructing its grammar, vocabulary, and literary history, we have managed to give it a name belonging, at least in all probability, to somebody else.
The story begins, as so many stories about the history of Central Asia seem to, with people travelling across the desert in search of things that other people had left behind. During the late nineteenth and early twentieth centuries, the oases of the Tarim Basin attracted a succession of European and Japanese expeditions, whose members were interested in the remains of the Buddhist monasteries, ruined settlements, and ancient libraries scattered throughout the region. These were places that had once occupied an important position along the Silk Road, where merchants, pilgrims, monks, and diplomats had moved between China, India, Persia, and the Mediterranean world. The languages and religious traditions that had passed through them were correspondingly diverse, and the manuscripts preserved in their ruins reflected a rather remarkable mixture of cultural influences. Among the discoveries were texts written in Sanskrit, Sogdian, Old Uyghur, Chinese, and several languages that European scholars had not previously encountered.
The expeditions themselves deserve a little attention, partly because the history of the languages we study is often inseparable from the rather less academic circumstances in which their documents were collected. The Germans organized four expeditions to the region between 1902 and 1914, led by Albert Grünwedel and Albert von Le Coq, while Aurel Stein undertook several journeys on behalf of British institutions. The French archaeologist Paul Pelliot, the Russian scholar Sergei Oldenburg, the Japanese Count Ōtani, and the Swedish explorer Sven Hedin also participated in the wider exploration of Central Asian antiquities. Their discoveries eventually found their way into collections in Berlin, London, Paris, St. Petersburg, and Japan, among other places. The Berlin collection alone contains materials recovered from sites across Turfan, Kucha, Hami, and Qarašahr, written in approximately fifteen languages and thirty different scripts. Michael Weiss, in his introduction to Tocharian, describes these expeditions with the compound expression “looting/research,” which is perhaps a more candid description than the rather grand language of archaeological discovery usually employed in accounts of the period. The scholarly achievements that followed were considerable, but so was the movement of cultural material out of the places where it had originally been produced. The Staatsbibliothek zu Berlin's account of the Turfan Collection provides a particularly interesting overview of the German expeditions and the materials they collected.
The first fragments of the language that would eventually be called Tocharian began attracting scholarly attention in the 1890s, although at the time nobody had a particularly good idea what they were looking at. This is not altogether surprising. The manuscripts were often damaged, sometimes consisting of little more than a few surviving lines, and were written in a script that had to be deciphered before the language itself could be studied. The writing system was derived from the Indian Brāhmī script, which had travelled northward along with the spread of Buddhism and had been adapted to represent the sounds of the local languages. It is rather interesting that a language spoken in the oases of northwestern China should have been written in a script of Indian origin, although this is hardly unusual in the history of the Silk Road, where religious and commercial networks carried writing systems across enormous distances. The same manuscripts also demonstrate how thoroughly Buddhist intellectual culture had travelled into Central Asia: many contain translations of Indian religious works, commentaries, monastic regulations, and accounts of the Buddha's teachings, while others preserve original compositions produced by local authors.
The decisive linguistic breakthrough came in the first decade of the twentieth century. Emil Sieg, working with his former student Wilhelm Siegling, began studying the Berlin fragments and recognized that the unfamiliar language belonged to the Indo-European family. This was a rather extraordinary discovery. Indo-European languages were already known from much of Europe and parts of western and southern Asia, but here was another branch preserved in documents from the easternmost reaches of the Tarim Basin. Its relationship to the other languages of the family was not immediately obvious, and its geographical position made it especially interesting to historical linguists attempting to reconstruct the prehistoric movements of Indo-European-speaking populations. In 1908, Sieg and Siegling published their work under the title Tocharisch, die Sprache der Indoskythen, or “Tocharian, the Language of the Indo-Scythians.” They also established that the material represented not one language but two closely related varieties, which they conveniently designated Tocharian A and Tocharian B. These labels were initially intended as practical scholarly distinctions, but, like many temporary academic arrangements, they became rather more permanent than anyone seems to have anticipated.
The two languages were not distributed evenly across the region. Tocharian B, which is also sometimes called Kuchean, is associated particularly with Kucha, although manuscripts have been found farther east as well. Tocharian A, sometimes called Agnean, is associated with Qarašahr and Turfan. The surviving texts suggest that Tocharian B was already in use by approximately the fifth century CE, perhaps earlier, while the attested Tocharian A material is somewhat later, dating predominantly from the seventh to tenth centuries. Both were written using varieties of the Brāhmī-derived script, and both preserve a substantial body of Buddhist literature. The distinction between them is not merely geographical, either. They are recognizably related languages, sharing a common ancestor and many grammatical features, but they have developed in different directions, with differences in vocabulary, morphology, and sound systems that make it impossible to treat one as simply a dialect of the other. The Encyclopaedia Iranica article by Michaël Peyrot gives a particularly useful account of their distribution and historical development.
There is also a rather lovely story about Sieg, which Michael Weiss relates through a chain of scholarly recollections originating with Don Ringe and Warren Cowgill, although, as Weiss himself acknowledges, it is almost certainly apocryphal. Sieg had apparently encountered a Buddhist text whose contents he could identify from the names it contained, but he could not determine which particular Sanskrit work it represented. One night, he dreamed that he got out of bed, walked to his library, retrieved a particular book, and opened it to a particular page, where he found the Sanskrit original of the text he had been studying. Upon waking, he supposedly went to the library and repeated precisely what he had done in the dream, discovering the passage exactly where he had expected it to be. I am not especially inclined to believe that this happened, not least because the story becomes more impressive with each person who retells it, but I rather wish it were true. There is something appealing about the idea that, after spending years trying to reconstruct a language from scraps of damaged manuscripts, a scholar might be rewarded with the answer arriving in a dream, complete with a page reference.
What makes the story particularly improbable, however, is the condition of the material with which Sieg and his colleagues were actually working. The surviving Tocharian corpus consists of approximately twelve thousand fragments, with the precise totals depending on how individual pieces and newly identified materials are counted. Most are exceedingly small. Some preserve only a few words; others contain portions of sentences whose beginnings and endings have disappeared entirely. The original manuscripts must often have been substantial works, written on long sheets of paper or other materials, but the libraries in which they were preserved have rarely survived in anything approaching their original condition. We almost never possess a complete book, and even complete pages are uncommon. Scholars are therefore frequently forced to reconstruct a passage by comparing several fragments, identifying repeated formulas, and consulting parallel versions in other languages. The difficulty is compounded by the fact that much of the surviving literature consists of translations from Sanskrit, which can provide useful clues but also introduce questions about how closely the translator followed the original. Melanie Malzahn's work on the Tocharian manuscript corpus discusses just how much of this material remains unpublished or only partially understood, and the University of Vienna's Comprehensive Edition of Tocharian Manuscripts project, CEToM, makes many of the surviving texts accessible for further study.
And then there is the language's actual name, which is where matters become considerably more complicated. The identification that gave us the word Tocharian originated with Friedrich W. K. Müller, who, in 1907, noticed something interesting in an Old Uyghur Buddhist manuscript known as the Maitrisimit nom bitig. The work is a translation of the Sanskrit Maitreyasamiti-Nāṭaka, a Buddhist drama concerning the future Buddha Maitreya, who is expected to appear in the world and bring about a new era of Buddhist teaching. The Old Uyghur version contains a colophon, essentially a note explaining the history of the text and its translation, which states that the work had been translated from an Indian language into a language identified by a term conventionally transcribed as twγry. Müller connected this word with the ancient Tocharians, known to Greek writers as the Tókharoi and to Indian sources through forms such as Sanskrit Tukhāra. These people were associated with Bactria, a region encompassing parts of modern Afghanistan and neighboring Central Asia, rather far to the west of the Tarim Basin. Müller nevertheless proposed that the newly discovered language was theirs, and Sieg and Siegling adopted the identification in their terminology.
It was a reasonable enough hypothesis given the information available at the time, particularly since the history of Central Asia contains numerous migrations, political transformations, and changes of ethnonyms that make geographical identifications rather difficult. The ancient Tocharians were associated with the Yuezhi, a people known from Chinese historical accounts who moved westward and eventually became connected with the establishment of the Kushan Empire. The Kushans themselves presided over an enormous territory stretching across parts of Central and South Asia, and their rule was particularly important in the development and transmission of Buddhism. It would not have been unreasonable to imagine that a population associated with them had moved eastward, eventually settling in the Tarim Basin. Unfortunately, the linguistic evidence does not cooperate quite so neatly with this interpretation.
The problem is that the historical Tocharians of Bactria appear to have spoken Bactrian, an Eastern Iranian language, rather than either of the languages preserved in the Tarim manuscripts. Bactrian is itself an Indo-European language, but it belongs to the Iranian branch, making it quite distinct from Tocharian. There are certainly connections between the various peoples and languages of Central Asia, and the region was never a collection of isolated communities, but the simple identification of the manuscript language with the language of the Bactrian Tocharians has not survived subsequent research. As the University of Texas's Introduction to Tocharian explains, the original identification is now generally regarded as untenable, even though the name derived from it remains standard.
There is a further complication, which is that even the word twγry is not quite as straightforward as Müller seems to have assumed. The form occurs in Old Uyghur, but its interpretation depends partly on how the consonants are read and how the word is related to similar forms in other Central Asian languages. Walter Bruno Henning, writing in 1938, examined related forms in Uyghur and Sogdian sources and argued that the term referred to a population or geographical region along the northern Silk Road, rather than necessarily to the inhabitants of ancient Bactria. His work distinguished between the people associated with the name in Bactria and another group whose name appeared in the records of Central Asia. Whether these groups were historically related, and how their names developed, remained uncertain. This is precisely the sort of problem that makes historical linguistics rather enjoyable, although I imagine it can become considerably less enjoyable when one has spent several years trying to establish whether a particular consonant was originally voiced. The name was not simply borrowed from a perfectly identifiable ancient people and attached to an unrelated language; it emerged from a series of uncertain identifications involving Greek, Sanskrit, Iranian, Sogdian, and Turkic forms, each with its own history of transmission and alteration.
What makes all of this especially unfortunate is that the speakers of the two languages seem to have had rather more sensible names for themselves. For Tocharian B, there is evidence of the adjective kuśiññe, meaning something like “Kuchean,” derived from kuśi, the name of Kucha. The city was known under several different names in the languages of the region, appearing as Kucha or Kuča in modern transliteration and Qiuci in Chinese historical sources. It was an important oasis settlement, and its position along the northern Silk Road made it a significant center of trade and Buddhist learning. The surviving texts suggest that its inhabitants were quite conscious of their own linguistic and cultural identity. One particularly interesting expression contrasts kuśi-pele, meaning “in the manner of Kucha,” with yentukäññe-pele, meaning “in the Indian manner,” apparently referring to Sanskrit. It is a small distinction, but I find it rather revealing. Whoever wrote the passage could assume that the reader understood what was meant by the local manner of doing things, and could contrast it with the Indian tradition without needing to explain either one. The distinction suggests a community participating in a wider Buddhist intellectual world while retaining its own linguistic identity.
There is something rather satisfying about the fact that the name of the language might simply have been derived from the city where its speakers lived. We have a tendency, when discussing ancient peoples, to imagine that they must have possessed elaborate ethnonyms and clearly defined national identities, but the surviving evidence often suggests something much more local and practical. A person might identify with a city, a monastery, a religious community, or a particular region, and the language spoken there might be described accordingly. The adjective kuśiññe is therefore not necessarily a grand declaration of collective identity, but it does give us something more directly connected to the people who actually used the language than the name imposed by twentieth-century European scholars.
Tocharian A offers a slightly different possibility. Its speakers used the expression ārśi-käntu, generally interpreted as “the language of Arśi,” with ārśi apparently referring to the region around Qarašahr, known in Sanskrit as Agni and in Chinese as Yanqi. The precise interpretation of the word is not entirely settled, and the relationship between ārśi, the geographical name Agni, and other forms recorded in neighboring languages has generated its own considerable discussion. Nevertheless, it provides a plausible indigenous designation for the language. The distinction is particularly interesting because Tocharian A and B, despite their close linguistic relationship, seem to have been associated with different geographical communities. In other words, the two languages that European scholars conveniently grouped together as A and B may themselves have had rather different names, reflecting the places with which their speakers identified.
There is also an important distinction between what the speakers called their languages and what their neighbors called them. The Old Uyghur term twγry appears to have been an external designation, while kuśiññe and ārśi-käntu provide evidence of local terminology. This is hardly unusual. We routinely encounter historical peoples whose names in foreign records differ from the names they used themselves, sometimes because neighboring populations adopted an older geographical term, sometimes because a political name became attached to an entire linguistic community, and sometimes because the original designation has simply disappeared from the surviving record. The difficulty with Tocharian is that the external name became the standard scholarly name before anyone had properly established whom it referred to. Once the terminology had entered textbooks, dictionaries, and academic publications, it became increasingly difficult to replace.
Indeed, the situation is made more confusing by the fact that the two languages do not appear to have been distributed exactly where their conventional names might lead one to expect. Tocharian B, despite sometimes being called West Tocharian, is attested across a wider area, including sites farther east, while Tocharian A is concentrated in Qarašahr and Turfan. The historical development of the languages suggests that Tocharian B originated around Kucha and subsequently spread eastward, while Tocharian A was associated with Qarašahr and also became established in Turfan. This means that the geographical labels are useful approximations rather than perfectly comprehensive descriptions of where every surviving text was written. It is also worth remembering that a manuscript's place of discovery does not necessarily establish the precise origin of the person who wrote it, particularly in a region where books, monks, and religious traditions travelled extensively. The movement of texts along the Silk Road makes it difficult to draw overly rigid boundaries around the languages and communities involved.
And, while we are already dealing with the problem of names, I think it is worth considering just how unusual the discovery of Tocharian was from the perspective of Indo-European linguistics. Most people encountering the Indo-European family for the first time are introduced to a relatively familiar geographical arrangement: Latin and Greek in the west, the Germanic and Celtic languages farther north, and the Indo-Iranian languages extending across parts of the east and south. Tocharian complicates this picture rather considerably. It is the easternmost ancient branch of the family for which we have substantial textual evidence, and its discovery forced scholars to reconsider what they knew about the geographical distribution and historical development of Indo-European languages.
Its linguistic characteristics are also fascinating. Tocharian has undergone several substantial changes that distinguish it from the other branches, including mergers in its consonant system and considerable developments in its vowels and grammatical morphology. Some of its vocabulary is recognizable in relation to reconstructed Indo-European roots, while other forms have developed in ways that make their historical relationships less immediately apparent. It is particularly interesting because it does not simply preserve an archaic version of Proto-Indo-European, as one might initially expect from a language found so far from the traditional centers of Indo-European scholarship. Instead, it has undergone its own extensive changes, making it an independent and highly informative branch of the family. Its position also gives historical linguists an additional source of evidence when comparing the development of the Indo-European languages, particularly because some of its innovations are shared between Tocharian A and B but not with other branches.
The manuscripts themselves are not exclusively religious, either, although Buddhism accounts for the overwhelming majority of the surviving literary material. Alongside translations of Sanskrit Buddhist texts, there are monastic regulations, commentaries, stories, hymns, and dramatic works, as well as administrative records, legal documents, commercial correspondence, and other more mundane pieces of writing. Some texts contain vocabulary lists and grammatical explanations, while others preserve fragments of technical literature. There are even personal and literary compositions that provide glimpses of life beyond the monasteries. This is important because it prevents us from imagining Tocharian as a language that existed exclusively to translate Buddhist scripture. It was used by people conducting business, managing institutions, writing letters, and participating in the ordinary affairs of their communities. The surviving corpus is overwhelmingly shaped by the circumstances of its preservation, and the large number of religious manuscripts reflects the importance of Buddhist monasteries as centers of literacy and manuscript production. It does not necessarily tell us everything about the language's everyday use.
One of the more charming discoveries in the corpus is a short love poem, which is rather easy to overlook when reading accounts that focus almost entirely on the historical importance of the Buddhist texts. There is something appealing about encountering such a personal piece of writing among documents that are otherwise so closely associated with religious scholarship and the reconstruction of ancient grammar. We can spend enormous amounts of time discussing the relationship between Tocharian and Proto-Indo-European, the historical development of its case endings, or the precise geographical distribution of its manuscripts, and then suddenly encounter evidence that somebody, many centuries ago, was writing about something considerably more familiar. Unfortunately, the fragmentary nature of the material makes it difficult to reconstruct the full circumstances of its composition, but even its existence is a useful reminder of the range of activities for which the language was employed.
So what should we call these languages instead? There have been several proposals, none of which has managed to displace the established terminology. The most straightforward option is probably to use the geographical names: Kuchean for Tocharian B and Agnean for Tocharian A. These have the advantage of being grounded in the historical geography of the languages and, at least approximately, in the terminology used by their speakers. Another possibility is to call them East and West Tocharian, although this retains the very name that caused the problem in the first place. The linguist Bernard Sergent has proposed the combined designation Arśi-Kuči for the language family, while other scholars have suggested variations such as Agni-Kuči. The problem is that none of these alternatives is entirely satisfactory. The geographical names are not necessarily representative of every community in which the languages were spoken, and the combined names are rather cumbersome, especially when one has to use them repeatedly in a textbook. Meanwhile, Tocharian A and Tocharian B are already so firmly established in academic literature that replacing them would require an enormous amount of terminological adjustment.
I suppose this is why the conventional name has survived. There is no particular scholarly consensus about what should replace it, and the alternatives introduce their own difficulties. It is also worth noting that the word Tocharian is not completely devoid of historical significance: it tells us something about how the language was identified by the scholars who first studied it, and about the connections they attempted to establish between the different peoples of Central Asia. The mistake itself is part of the history of the field. Removing the name would not necessarily remove the confusion that produced it, and retaining it does not mean that modern scholars accept the original identification. It simply means that a terminology established more than a century ago has proved rather difficult to dislodge.
I find the whole situation particularly amusing because Tocharian is precisely the sort of language that ought to inspire a certain amount of scholarly humility. We are dealing with an extinct Indo-European branch whose existence was unknown to most European linguists until relatively recently, whose surviving literature consists largely of fragments, whose speakers' historical identity remains incompletely understood, and whose geographical distribution has to be reconstructed from manuscripts scattered across several countries. We have managed to decipher enough of its grammar to compare it with languages as distant as Latin, Greek, Sanskrit, and Hittite, and to establish that it represents an independent branch of the Indo-European family. Yet the name by which we identify it rests on an association that was already being questioned within a few decades of its introduction.
Perhaps the most interesting thing about the entire affair is that the original manuscripts contain evidence of the names their speakers actually used, even if that evidence is itself incomplete and sometimes difficult to interpret. We have kuśiññe, we have ārśi-käntu, and we have the external designation twγry, all of which tell us something about the ways in which different communities understood and described one another. These names are not interchangeable, and they do not necessarily refer to exactly the same geographical or linguistic categories. But they offer a considerably more complicated picture than the simple label that has survived in modern textbooks.
For now, I suppose we are stuck with Tocharian. I cannot say that I particularly mind; after all, changing the name would probably make an already rather specialized field even more difficult to navigate, and there is something useful about having a consistent terminology when comparing the two languages. Still, I do think that anyone encountering the word for the first time ought to be told where it came from. There is a considerable difference between learning that Tocharian is an extinct Indo-European language spoken in the Tarim Basin and learning that the name itself was assigned on the basis of a historical identification that scholars subsequently rejected. The latter makes the whole subject rather more interesting, particularly when one begins to follow the various Greek, Sanskrit, Iranian, Sogdian, Chinese, and Turkic names through which the confusion developed. And, having spent so much time reconstructing the words of people who have been gone for a thousand years, it seems only fair that we should at least try to call their languages something they might have recognized.