The first time I encountered the Tocharian word for ‘father,’ pācer, I thought I had found a rather straightforward example of Indo-European linguistic conservatism. It looks remarkably like Latin pater, and, given that Tocharian was discovered thousands of miles east of the Mediterranean, I found the resemblance rather exciting. Then came mācer, ‘mother,’ and procer, ‘brother,’ and the whole thing began to look less like an isolated coincidence and more like a family reunion. There is something rather remarkable about finding words so familiar in a language whose surviving manuscripts were recovered from the oases of the Tarim Basin, in what is now northwestern China, an area whose immediate linguistic surroundings are predominantly Chinese, Turkic, and Iranian. One might reasonably expect its vocabulary to look rather more like that of its neighbors. Instead, a student of Latin can open a Tocharian dictionary and recognize quite a surprising amount of it.
Of course, the resemblance is not accidental. Tocharian is an independent branch of the Indo-European language family, and its relationship to Latin, Greek, Sanskrit, and the other languages is the result of descent from a common ancestor, Proto-Indo-European. What makes it particularly interesting is that it occupies an unusual geographical and linguistic position. It is the easternmost attested branch of the family, separated from the other ancient Indo-European languages by enormous distances, and yet its vocabulary preserves correspondences that are sometimes remarkably transparent. The word for ‘horse,’ Tocharian B yakwe, corresponds to Latin equus, while okso, ‘ox,’ is practically identical to English ox. Tocharian ñem, ‘name,’ belongs with Latin nōmen, Greek ónoma, and English name. Even ṣpane, ‘sleep,’ can be connected with Latin somnus and the reconstructed Proto-Indo-European form *swepnos. These are not isolated resemblances selected from a dictionary of several thousand words; they represent regular historical relationships that linguists can trace through the sound changes separating the languages.
The vocabulary becomes particularly interesting when one moves beyond the most familiar family words. Tocharian B pwar, ‘fire,’ belongs to the same Indo-European root as Greek pyr and English fire, while kem, ‘earth,’ has an older form tkem that can be compared with Greek khthōn and Hittite tekan. The latter is especially interesting because it preserves a consonant cluster that has been simplified in the later Tocharian forms. There is also kokale, meaning ‘chariot’ or ‘wagon,’ which is connected with Greek kyklos, ‘wheel,’ and ultimately with English cycle. The semantic development is not particularly difficult to follow: a word referring to something circular can become associated with the vehicle that moves on wheels. Tocharian keme, ‘tooth,’ is related to Greek gomphos, ‘peg’ or ‘bolt,’ and Sanskrit jambha, another word associated with teeth. The comparison is slightly less immediately obvious, but the underlying connection becomes easier to appreciate when one considers the way teeth are described as fixed projections in the jaw. Finally, there is the verb pik-, ‘to write’ or ‘paint,’ which corresponds to Latin pingere, ‘to paint.’ Given that so much of our knowledge of Tocharian comes from manuscripts, it is an especially appropriate correspondence, although the verb itself presumably had a much wider range of uses than the rather specialized circumstances in which we encounter it.
There is even a word for alcohol. Tocharian B mot, according to Michael Weiss, is probably related to Greek methy, ‘strong drink,’ which also belongs to the historical family of English mead. The exact reconstruction is less straightforward than the comparison between yakwe and equus, but it is another indication of the breadth of the inherited vocabulary. The interesting thing about these words is that they do not necessarily describe the world of the Tocharian speakers in the period represented by the surviving manuscripts. Some are inherited terms for objects and concepts that had existed thousands of years earlier, while others had undergone semantic changes as the language developed. The presence of an inherited word for a chariot, for example, tells us something about the vocabulary of the ancestral language, not necessarily about the transportation available in eighth-century Kucha. Historical linguists are often able to recover vocabulary from a period for which there are no written records at all, simply by comparing the words preserved in their descendants.
The numbers are perhaps even more revealing. Tocharian B counts trai, later trey, stwer, piś, ṣkas, ṣukt, okt, ñu, and śak for three through ten. At first glance, these look like a rather peculiar collection of syllables, but placing them beside their Latin counterparts makes the relationship much easier to see: tres, quattuor, quinque, sex, septem, octo, novem, decem. Three is trai, four is stwer, and five is piś; six and seven are ṣkas and ṣukt, while eight is okt. The resemblance between okt and octo is particularly obvious, as is the relationship between ñu and Latin novem. The numbers also demonstrate something important about the reconstruction of Proto-Indo-European: individual words can change quite substantially while retaining enough of their original structure to be recognized through systematic comparison.
Ten, however, is less cooperative. Latin has decem, while Tocharian B has śak, which initially looks unrelated. The explanation involves a series of regular sound changes. The original Indo-European dental stop d became t in Tocharian; this t subsequently underwent palatalization, developing through an affricate before eventually becoming ś. The final form is therefore the result of several successive developments rather than an unrelated replacement of the original word. The process is particularly interesting because the same sound changes can be observed elsewhere in the language, allowing linguists to establish that the relationship is systematic rather than simply proposing an etymology that happens to produce a convenient resemblance. One can reconstruct the earlier form, follow its development through the stages of the language, and arrive at the attested word. This is the sort of comparison that makes historical linguistics rather more convincing than simply collecting words that sound vaguely similar.
Then there is kante, ‘hundred,’ which is considerably more important than its appearance in a list of numbers might suggest. One of the traditional ways of dividing the Indo-European languages is the centum-satem distinction, named after the words for ‘hundred’ in Latin and Avestan. In the centum languages, the original Proto-Indo-European palatovelar consonants developed into ordinary velars, represented approximately by k and g. In the satem languages, they developed into sibilants, producing sounds like s and ś. Latin centum belongs to the centum group, as do Greek hekaton and Tocharian kante, while Sanskrit śatam, Avestan satəm, and Lithuanian šimtas belong to the satem group. The geographical distribution initially appears rather neat: centum languages are generally found farther west, while satem languages are concentrated farther east. Tocharian, located deep within Asia, is therefore rather inconvenient for anyone hoping that geography alone will explain the classification. Its kante preserves the velar consonant, placing it firmly among the centum languages despite its eastern location.
This is particularly significant because it demonstrates that the centum-satem division is not a straightforward division between western and eastern Indo-European. It is a set of shared sound changes, and its geographical distribution reflects the historical development and movement of different language communities rather than some fundamental division of the family into two geographical halves. Tocharian's position has consequently been important in discussions of Indo-European subgrouping. Its centum characteristics do not establish a special relationship with Latin or Greek, any more than the shared satem developments of Sanskrit and Lithuanian necessarily mean that those languages are especially close relatives. Tocharian is an independent branch, and its similarities with other languages have to be evaluated feature by feature. The fact that it preserves a centum consonant while occupying the easternmost known position of the family is precisely what makes it so useful to historical linguists.
The nouns, however, are considerably less familiar than the vocabulary might lead one to expect. Lane describes the nominal system as behaving in a rather unusual, even ‘highly non-Indo-European,’ fashion. Proto-Indo-European nouns were inflected for several cases, with endings indicating relationships such as subject, object, possession, and location. Latin preserves a relatively recognizable version of this system, with nominative, accusative, genitive, dative, ablative, and other forms, although these categories themselves developed over time. Tocharian has a more reduced set of primary cases, particularly the nominative, oblique, and genitive, with additional grammatical relationships expressed through elements attached to the oblique stem. Among these are -sa, meaning ‘with’ or ‘by means of,’ -mpa, meaning ‘together with,’ -ne, meaning ‘in,’ and -meṃ, meaning ‘from.’ These elements resemble postpositions that have gradually become attached to nouns, rather than the inherited case endings familiar from Latin and Greek.
The distinction matters because it tells us something about the way the language has developed its grammar. In Latin, the ablative ending in gladiō, for example, can express several relationships depending on the context, including means and separation. Tocharian instead makes more extensive use of elements that are attached to the oblique stem to express particular relationships. Weiss points out that these formations are not identical in Tocharian A and B, suggesting that the process of grammaticalization continued after the two varieties had begun to develop separately. This is particularly interesting because it demonstrates that the languages were not simply preserving an unchanged Indo-European grammatical system. They had inherited a recognizable set of structures from their common ancestor, but had subsequently reorganized some of them in rather different ways.
There are also features of the nominal system that suggest considerable contact with neighboring languages. The use of attached postpositional elements has often been compared with the agglutinative structures found in Turkic languages, where grammatical information is expressed through sequences of suffixes added to a stem. It would be tempting to attribute the entire development to Turkic influence, especially given the geographical surroundings of the Tocharian-speaking communities, but the situation is more complicated. Similar grammatical developments can occur independently, and establishing contact influence requires more than simply identifying two languages that happen to use suffixes. Nevertheless, the comparison is useful, particularly because Tocharian existed in a multilingual environment in which Iranian, Turkic, Chinese, and Indo-Aryan languages all contributed to the region's linguistic history.
The verbs, meanwhile, are a rather different matter. Unlike the nouns, which have undergone substantial restructuring, the Tocharian verbal system retains many features that are recognizably Indo-European. There are active and middle voices, several moods, and a distinction between present and past formations, alongside a variety of stem-building patterns. The middle voice is particularly interesting because it is expressed through endings containing -r. In Tocharian B, third-person singular and plural middle endings include -tär and -ntär. Anyone familiar with Latin deponent verbs will recognize the general appearance of these forms: Latin sequitur, ‘he follows,’ and sequuntur, ‘they follow,’ have the same characteristic -r in their endings. The comparison is not accidental. Similar middle endings occur in Hittite and Old Irish, among other Indo-European languages, and are generally understood as preserving an old feature of the ancestral verbal system.
The middle voice itself is a grammatical category that can be rather difficult to explain through a direct English translation. It often describes actions in which the subject is particularly involved in, or affected by, the action, although its precise uses vary considerably between languages. Greek provides a familiar example, distinguishing active forms from middle forms that can express actions performed for one's own benefit or actions in which the subject participates more directly. Tocharian preserves this distinction in its verbal morphology, providing evidence for an older grammatical system that has largely disappeared from many modern Indo-European languages. Its presence is especially useful because it allows linguists to compare Tocharian with languages such as Greek and Hittite without relying entirely on vocabulary.
The number of possible verbal formations is also rather impressive. Weiss identifies eleven classes of present stem, nine classes of subjunctive stem, seven preterite classes, and six imperative classes. These categories reflect different ways of constructing verbal stems, often involving particular suffixes, vowel changes, or modifications to the root. They do not simply represent thirty-three different tenses; rather, they describe the morphological patterns through which different grammatical forms are produced. Still, it is a substantial system, particularly for a language whose surviving documentation is so fragmentary. The forms are not all equally productive, and many verbs are irregular, but the variety demonstrates how much of the inherited Indo-European verbal complexity Tocharian managed to retain while developing its own distinctive structures.
The sounds of the language are another source of interest, particularly because the script in which Tocharian was written had originally been developed to represent Sanskrit. The Tocharian scribes adapted the Brāhmī-derived writing system to accommodate sounds that did not exist in Sanskrit, creating additional signs where necessary. Among these was a vowel that Emil Sieg and Wilhelm Siegling initially called the Fremdvokal, literally the ‘foreign vowel.’ The name is rather descriptive, although not especially imaginative. It is now conventionally represented as ä and is generally understood as a reduced central vowel, broadly comparable to the schwa in English about. Its behavior is particularly interesting because it can disappear in unstressed syllables and reappear when the structure of a word changes.
For example, Tocharian B yakwe, ‘horse,’ has the related form yäkweñe, ‘equine.’ The vowel in the root behaves differently depending on the formation, illustrating the way stress and syllable structure affected the development of the language. A similar process can be seen in cämel, ‘birth,’ whose plural form cmela involves the loss of the vowel in the first syllable. This is not simply a matter of scribes deciding that they could not be bothered to write every vowel. The alternations reflect an actual phonological process, and they can be observed in a variety of inflectional and derivational forms. The existence of a vowel that Sanskrit lacked also demonstrates how extensively the original writing system had to be adapted for the language it was being used to record.
Consonants undergo their own changes, particularly when they occur before front vowels. Tocharian has a process of palatalization in which certain consonants develop different pronunciations depending on the vowel that follows them. The result can be seen in the relationship between lyak, ‘thief,’ and lyśi, ‘thieves.’ Rather than simply changing the ending to indicate plurality, the consonant itself changes as a consequence of the following vowel. There are parallels in many other languages, including English, although the particular historical processes are different. English foot and feet, for instance, preserve the effects of an older vowel alternation, while Tocharian exhibits consonantal alternations that similarly reflect earlier phonological environments. These changes are particularly useful to linguists because they preserve traces of earlier forms that would otherwise be difficult to reconstruct.
Some of the grammatical distinctions are also remarkably restricted in their use. Tocharian preserves a dual number, used to refer specifically to two objects or people, alongside the singular and plural. The dual is familiar from Ancient Greek and Sanskrit, where it was once a productive grammatical category, but its use in Tocharian is considerably limited. Lane notes that it is primarily attested with natural pairs, particularly body parts such as eyes, ears, feet, and knees. This is a rather interesting survival of an older grammatical category: the language retains a distinction that other Indo-European languages have largely lost, but no longer uses it freely across the entire nominal system. The dual is therefore not simply a general-purpose way of referring to two things. Its distribution suggests that it had become associated with particular semantic categories, perhaps through the gradual restriction of an originally more productive form.
There is also the distinction between Tocharian A and B, which introduces another layer of complexity. Although the two varieties are closely related, their differences are substantial enough that they cannot simply be treated as regional pronunciations of a single language. Weiss identifies approximately thirty-five regular correspondences between them, many of which reflect systematic sound changes that occurred after the two varieties had separated. One particularly noticeable difference is that Tocharian A tends to lose final vowels that survive in Tocharian B. Thus B yakwe, ‘horse,’ corresponds to A yuk, B walo, ‘king,’ to A wal, and B luwo, ‘animal,’ to A lu. These are not isolated differences but part of a much broader pattern of phonological development.
The treatment of diphthongs provides another example. Tocharian B preserves ai and au in places where Tocharian A has monophthongized them to e and o. Consequently, B trai, ‘three,’ corresponds to A tre. The same tendency toward the reduction of vowels can be observed in several other forms, making Tocharian A appear, in many respects, more phonologically reduced than Tocharian B. This does not necessarily mean that B is an entirely conservative version of the ancestral language, or that A is simply a later stage of the same development. Both have undergone their own changes, and the surviving evidence is insufficient to reconstruct every detail of their separation. Nevertheless, the regularity of the correspondences allows scholars to identify shared innovations and distinguish them from features inherited from their common ancestor.
One of the more unusual differences concerns grammatical gender. Tocharian A distinguishes masculine and feminine forms of the first-person singular pronoun, with ñäś used by men and ñuk used by women. This is particularly striking because the first-person pronoun is generally not a grammatical category in which Indo-European languages distinguish the gender of the speaker. Latin ego, Greek egō, and English I do not change depending on whether the speaker is male or female. Tocharian A's distinction is therefore unusual, and it is difficult to find a direct parallel elsewhere in the family. It also demonstrates that the two Tocharian languages had developed differences that extended well beyond the simplification of vowels and consonants. Whatever the original source of this distinction, it had become part of the grammatical system of Tocharian A.
Taken together, these features make Tocharian particularly valuable for understanding the development of Indo-European. Its vocabulary preserves recognizable cognates with languages scattered across Europe and Asia, while its phonology and morphology demonstrate the extent to which an independent branch can develop after separating from the ancestral language. It retains certain archaic features, such as the middle voice and the centum consonants, while also displaying substantial innovations in its nominal system, vowel structure, and verbal morphology. It is neither an unusually conservative relic of Proto-Indo-European nor simply an eastern version of Latin or Greek. It has its own history, and that history is precisely what makes it so interesting.
What I find most remarkable is how much of this can be established from a corpus consisting largely of damaged manuscripts, many of which preserve only a few lines of text. The languages themselves were spoken in communities that participated in an extensive network of trade and religious exchange, and their surviving literature reflects a world in which Indian Buddhist traditions, Iranian cultural influences, and local Central Asian practices existed alongside one another. We can identify the Indo-European ancestry of a word for ‘horse,’ reconstruct the historical development of a consonant in the word for ‘ten,’ and compare the middle endings of a verb with forms preserved in Latin and Hittite. We can even distinguish two closely related languages through the systematic differences in their vowels and pronouns. None of this requires Tocharian to have been particularly similar to any one of its better-known relatives; in fact, its importance comes largely from the ways in which it differs from them.