By comparison to the Modern English system, the verbal system of Proto-Indo-European is quite ornate. In Modern English, the verb give merely displays 5 forms: give, giving, gives, gave, given. In Proto-Indo-European, by contrast, any particular verb can disply dozens of distinct forms. We turn now to a discussion of the categories which describe the formal verbal system.
In many modern, western European languages, we customarily distinguish a number of verb tenses. Tense denotes the placement of an action in time, relative to the moment of utterance. That is, when I say in English "he gave," we say gave is in the past tense, and that this signifies that the action of giving occurred prior to the moment the sentence was uttered. "He gives," by contrast, employs the present tense, and this is supposedly concurrent to the moment of utterance. But here we run into a peculiarity of English: the simple present form typically denotes a habitual action, among other types of action, so that "he gives" typically means "he gives regularly" or "he is always giving," i.e. "he is a generous person." What it does not mean is "he is giving right now." Historically, in earlier stages of English, this situation did not obtain, and "he gives" was truly an action concurrent with the utterance; in Modern English we must say "he is giving" to achieve the desired effect of a true present tense. This brings out a subtle point: not everything that is called a "tense" actually is a tense in the strict sense; at times, however, we will bow to tradition and still employ the terminology.
Verbal aspect contrasts with tense. Aspect denotes the quality or shape of an action in time. That is, aspect does not say, even relatively, when an action takes place, but rather describes something about how it takes place, or rather how the speaker views it taking place. For example, contrast Modern English "he gave" vs. "he was giving." Both verb forms mark the past tense, i.e. they both mark the action as occurring prior to the moment of utterance. They therefore have the same tense. But they do not show the same aspect: in particular, "he gave" represents the action as a complete whole, while "he was giving" does not. To make the distinction more clear, realize that with "he gave," we know that the act of giving began, perhaps progressed, but definitely terminated before the moment of utterance; with "he was giving," we similarly know that the action began and progressed, but we cannot be sure that the action finished before the moment of utterance. We might distinguish these two aspects by saying that "he gave" is completive -- or punctual or perfective -- whereas "he was giving" is progressive -- or non-punctual or imperfective. The perfective aspect views the action as a point in time, whereas the imperfective aspect views the action as an interval of time (with indeterminate endpoint).
Clearly the concept of verbal aspect is more nebulous than its cousin tense, and there are many finer distinctions to be made, not all of which enjoy universal agreement within the linguistic community. As an example of another type of verbal aspect, we might consider the verbal unit "be green," as in "he is green," or even more simply "be" or "exist." These actions -- well, verbs at any rate -- do not sit conveniently within the above framework of completive (denoting a point in time) or incompletive (denoting an interval of time). Linguists assign such verbs to a class all their own called stative verbs, aptly enough comprising verbs which denote a state of being, i.e. something almost out of time altogether.
Many languages, such as German, have morphology which predominantly marks tense: er geht, formally "he goes," but really equivalent to "he is going", vs. er ging "he went." Some languages, such as Spanish, morphologically mark tense distinctions, e.g. dice "he says" (really, "is saying") vs. dijo "he said" (cf. Latin dīcit vs. dīxit); and even mark some aspectual distinctions morphologically, e.g. dijo "he said" vs. decía "he was saying" (cf. Lat. dīxit vs. dīcēbat). Still other languages predominantly mark aspectual distinctions, such as Yucatec Maya: k-inw-il-ik "I am seeing (something)" vs. tz'o'ok inw-il-ik "I was seeing (something up till just now)" vs. tz'o'ok inw-il-aj "I saw (something)." In this last example, the root il- means "see (something);" the prefixes k- and tz'o'ok denote, respectively, aspectual distinctions of ongoing and ongoing-to-the-present-moment, while the suffixes -ik and -aj denote, respectively, incompletive and completive aspect.
Communis opinio, but by no means the only opinio, currently holds that the early stages of Proto-Indo-European demonstrate an aspectual, rather than tense, system. This system comprises three basic aspects: durative (imperfective), punctual (perfective), and stative. Any one of these aspects could be inherent to the particular verbal root's (more on which below) meaning; but the Proto-Indo-European speaker could also shift categories by applying morphological affixes with aspectual content. This structure eventually gives way, at least in large part, to a tense system before the appearance of the documented languages. But the variety of tense structures we find in the daughter languages arises from the intricacies of the original aspectual system of PIE. For example, the fact that linguists may reconstruct (Sihler, p. 446ff)
|*dʰegʷʰ-ti 'is burning'||(!)gʷem-ti|
|*e-dʰegʷʰ-t 'was burning'||*e-gʷem-t 'arrived'|
suggests that what distinguishes the imperfect tense (ongoing, past-time action) from the aorist (punctual, past-time action) was not necessarily the morphology, but often the verbal aspect inherent in the root. The PIE root *dʰegʷʰ- 'be burning', 'be on fire' has an inherently non-punctual meaning, so that when a speaker applies hic-et-nunc (here-and-now) morphology in the guise of *-ti, the natural interpretation is as a continuous, present-time action, i.e. as a present tense. When the speaker applies, let's say, the there-and-then morphology *e-...-t, the resulting interpretation is a continuous, past-time action, i.e. an imperfect tense. But with PIE *gʷem-, the outcomes are different: applying *e-...-t produces a form which the speaker interprets as punctual and there-and-then, i.e. a simple past, or aorist, tense; the difference in interpretation can not derive from the morphology, and so must lie in the root itself. The root *gʷem- must inherently have a punctual (non-durative) meaning, "arrive" or "reach this point" (either you've reached it or you haven't; either you've done the verb or you haven't). In this sense, the absence of here-and-now morphology for the root *gʷem- derives from a natural incongruity between the sense of the hic-et-nunc morphology and the inherent aspect structure of the root.
But PIE speakers could also employ a number of suffixes which modify the aspectual structure. For example, we do find evidence to reconstruct
and even the aorist *e-dʰēgʷʰ-s-t with additional punctual marker *-s-.
Of course the data supporting the reconstruction of the PIE aspectual system is never as clear-cut as one would hope, and exceptions to the above tendencies abound. However the tripartite aspectual distinction durative-punctual-stative goes a long way to making sense of the idiosyncrasies which permeate the tense systems of the daughter languages. To provide a common footing, nevertheless, we may say that Proto-Indo-European itself showed, in its latest stages, the emerging dominance of a tense-based verbal morphology. Generally, scholars identify the following tenses.
|Present||durative action, concurrent with utterance||he is eating|
|Imperfect||durative action, prior to utterance||he was eating|
|Aorist||punctual action, prior to utterance||he ate|
|Perfect||action completed as of utterance||he has eaten|
There also arise further tenses whose inclusion in PIE is more problematic.
|Pluperfect||action completed prior to utterance||he had eaten|
|Future||action subsequent to utterance||he shall eat|
The perfect, pluperfect, and future enjoy a special status. The perfect, on the one hand, most archaically derives from the stative, signifying the state resulting from prior action: "I have done" is equivalent to "I am in a state of having completed." In this and other senses, the notion of tense only loosely applies to this category. On the other hand, in some daughter languages, what is termed the perfect is actually a conflation of tenses. In particular, the Latin verb dīcere "to say" forms the perfect dīxit. This form may have two interpretations in Latin, either "he said" or "he has said." Only the latter properly expresses the perfect tense; the former denotes a simple past. In the case of this verb, the sense of the simple past results from the historical morphology: dīx-it < PIE *deiḱ-s-, where the *-s- provides the overt marker of the aorist. In Latin, the twofold interpretation for perfect tense forms derives from the coalescence of both perfect and aorist markers into the same tense category.
The notion that the pluperfect, or past of the perfect, was a truly PIE category meets with some skepticism in the scholarly community. Similarly scholars debate whether the future itself should be counted among the PIE inventory. In particular, as a tense it lacks anything near uniform attestation in the daughter dialects, both in terms of existence as a distinct morphological category and, if it does happen to exist in a given language, in terms of its morphological realization. It is likely that the subjunctive mood filled the role of a future in PIE, and that the future as such was not a morphological tense in PIE itself.
Thus saying that PIE had such a replete system either throughout its history, or throughout all regions ultimately leading to the earliest attested languages, is a gross oversimplification. Hittite for example has a greatly reduced verbal morphology, in many ways more in line with the aspectual system discussed above. Since Hittite is to date the earliest attested Indo-European language, this furnishes the necessity of distinguishing stages on an aspect-tense continuum throughout the Proto-Indo-European period.
The classical moods ascribed to Proto-Indo-European are the indicative, imperative, subjunctive, and optative. A verb takes the indicative mood when the speaker wishes to relate the action as a fact. The imperative mood denotes a direct command. Both the subjunctive and optative combined constitute the so-called irreales moods, that is, the speaker employs a verb in the subjunctive or optative when the depicted action does not (yet) hold a factual status from the speaker's perspective. Among these two, subjunctive and optative, the optative generally has the further connotation of a wish or desire. As formal categories, these continue to distinguish themselves in English. Consider the following examples.
|Indicative||I was looking...|
|Subjunctive||Were I looking...|
|Optative||Would I were looking...|
Their morphological distinction, however, is another matter. In general, PIE does in fact distinguish all of these categories based on morphological features. In particular, the imperative distinguishes itself by employing the basic verbal stem and a special set of endings. The subjunctive exhibits a lengthened vowel grade between verbal root and ending, while the optative employs the special suffix *-ih₁. In fact, the morphological distinction between the subjunctive and optative is often clearer than the semantic distinction: for example, the Germanic languages as a whole generally lost the subjunctive and replaced it with the optative. The subjunctive seems also to have served as the future tense. Homeric Greek and Vedic, moreover, provide evidence for a further irrealis category: the injunctive or conjunctive. These tend to show past tense morphology, but often have an irrealis force. Scholars still debate the status of this category, but likely it represents an under-marked form of the verb that takes its interpretation from the preceding verb.
Proto-Indo-European verbal morphology generally proceeds by adding affixes to a verbal root. For the most part, these added elements are suffixal, coming at the end of the word. Scholars find it convenient to further analyze these terminal elements into two parts, called the suffix and the ending. Thus a PIE verb form as encountered in an utterance breaks down into the following constituents:
This is often simply denoted R+S+E. For example, a Sanskrit form such as bʰárati 'carries' would break down as follows:
For a given form or a given verb, however, one or both of the suffix or ending may be Ø, as in asti 'is':
The root conveys the inherent meaning of the verb. The root and suffix taken together form what linguists term the verbal stem:
Any one of the elements in a verb, be it the root, the suffix, or the ending, can undergo alterations of various types to further modify the meaning of the form as a whole.
The thematic vowel is a suffix of particular interest. This suffix takes the form of either *-e- or *-o-, depending on the particular form within a verbal paradigm. For example, consider the PIE roots *bʰer- and *h₁es-, antecedent to Skt. bʰar- and as- mentioned above. The former verb employs the thematic vowel in the present indicative, while the latter does not, as the following forms help illustrate.
The presence or absence of the thematic vowel plays an important role in PIE historical morphology. Accordingly linguists divide all verbs into two main classes, thematic and athematic, depending on whether a given verb employs or does not employ the thematic vowel in its conjugation.
Of course PIE languages also display a wide variety of prefixal elements, or preverbs. Sometimes these elements only find application in verb forms, such as the past tense augment *(h₁)e-. But more frequently these preverbs take the form of pre- or postpositions that gravitate strongly to the verbal root and more or less substantially color its meaning. In this latter case, these generally undergo no further modification and become part of the lexical entry in a dictionary. Thus, in terms of the R+S+E analysis of verbs, it is customary to include these with the verbal root under the heading of R.
Ablaut refers to the use of vowel gradation to change meaning within a root. Examples abound in English: sing, sang, sung, song. With each change of vowel, the root s-ng referring to some type of melodic speech act changes grammatical function: sing, present indicative; sang, past indicative; sung, past participle; song, noun. Such vocalic shifts form the bane of any non-native speaker's attempt to learn the language, since the particular type of vowel gradation often depends strongly on the verb: fling, present indicative; flung, past indicative; flung, past participle -- this employs some of the same vowels as sing, but at times in different roles. The assumption of parallelism, can run one into trouble: flang won't beget anything but a chuckle. All hope is not lost, however, since the possible ablaut patterns realized by English verbs actually fall into only a handful of categories: for example hang, hung, hung follows the same pattern as fling; so the problem reduces to one of identifying the few possible ablaut patterns and then remembering which verbs belong to which class, something which proves difficult even for native speakers! (Do I say you sank my battleship or you sunk my battleship?)
Speakers of Germanic languages, however, may at least find it interesting that this system of vocalic alternation, ablaut, is as old as the hills from a linguist's perspective. That is, verbal ablaut is an original feature of Proto-Indo-European itself. Certainly the Germanic ablaut categories do not exactly match those of PIE -- at times they don't even match between Germanic languages -- but the essential system remains in many respects intact. Linguists customarily illustrate the system with an example from Greek: the Greek form leípō means 'I leave,' where -ō is the present indicative ending for the first person singular. Consider now the grade of the root vowel in the following forms.
We see the appearance of three distinct grades of the root vowel e: full grade e, zero grade Ø, and o-grade o. The above correlation between grade of the root vowel and grammatical function happens to be emblematic of the situation in PIE as a whole. In fact, the above ablaut pattern correlates nicely with the threefold aspectual distinction mentioned earlier: durative correlates with full grade, punctual with zero grade, and stative with *o-grade.
Of course nothing in Indo-European linguistics is ever that easy, and the above threefold system does not depict the whole story. For example, the verb leípō also forms the aorist éleipsa, with the full grade of the root. But we see that this has in addition the punctual marker *-s-, and so this too concords with our notion that the durative stem leip- can be made punctual by other morphological features, as here with the suffix *-s-. Neither are the above the only vowel grades one encounters. There is strong evidence for a lengthened-grade stem in the past tense formations: Latin present indicative leg-ō 'I read' < *leǵ-, but perfect (i.e. preterite) lēg-ī, with lengthened grade of the root vowel.
We also see in the form léloipa the feature of verbal reduplication, that is, the appending of the initial consonant (cluster) and vowel of the root to the front of the root itself. Thus the initial *le- of the root *leip- are prefixed to the verbal root, and then ablaut is applied to the root syllable itself. Such verbal reduplication is a characteristic, but not defining, feature of the PIE stative or perfect. But it is not wholly confined to the perfect, appearing also in the present of some verbs: Gk. dí-dō-mi, from PIE *deh₃. Present reduplication is not nearly as systematic as its perfect counterpart: neither is it as pervasive as in the perfect, being confined to a small number of roots; nor is the formation quite the same, since Gk. dí-dō-mi and Skt. dá-dā-mi show the reduplicated vowel does not follow any clear pattern.
The actual semantics of reduplication still remain obscure. One in fact finds another type of reduplication in which the root as a whole (including root-final consonants) is prefixed to the root itself. This reduplication falls under the heading of intensive, but it remains unclear as to how this relates to present and perfect reduplication.
Linguists have been able to reconstruct three basic sets of verb endings for PIE. The first set, called primary, consists of those endings which generally pertain to non-past tense conjugations of verbs in the daughter languages. The second set, called secondary, consists on the other hand of those endings which generally pertain to past conjugations. The third set, the perfect endings, naturally comprises those endings which pertain to the PIE stative or perfect. Each of these sets further divides into active and middle endings, pertaining to the respective voices of the PIE verb. The following charts illustrate the endings.
|1 Sg.||*-mi, H₂||*-m||*-H₂e|
|3||*-or, tor||*-o, to|
|1 Pl.||*-medʰH₂ (?)||*-medʰH₂ (?)|
|2||*-dʰ(u)we- (?)||*-dʰ(u)we- (?)|
|3||*-ro(r), ntor||*-ro, nto|
The paradigms clearly show that we have more secure knowledge about the singular forms than about those for any other number. Clearly the active primary and secondary endings bear a closer resemblance to each other than either does to the perfect system. In part, the active primary endings recapitulate their secondary brethren, but with the addition of the hic et nunc (Latin "here and now") particle *-i. Interestingly, however, the perfect active endings and the middle endings in general bear an unexpected similarity. The exact reasons for this similarity remain unclear, and the relation between the systems still forms an important area of research.
Participles are verbal adjectives, that is, adjectives built from a verbal root. PIE possesses a large variety of participial formations. The *nt-participle shows reflexes in a number of daughter languages. This participle was formed by adding *-(e)nt- to the verbal root. This combined in athematic verbs with the weak form of the present stem: *h₁s-ent- > Lat. ab-sent- 'being away'. In thematic verbs the zero grade *-nt- combined with the *o-grade of the thematic vowel: *bʰer-o-nt- > Gk. pʰéront- 'carrying.' As seen in the examples, this formation generally takes on the role of a present active participle. In the Anatolian branch, however, the same formation conspicuously carries the meaning of a past participle: Hitt. iyant- 'having gone'.
Scholars also reconstruct a mediopassive participle in PIE, either *-m(e)no- or *-mh₁no-. This survives in such forms as Gk. pʰeró-menos 'carrying (oneself), being carried'.
PIE likewise contains a perfect participle in *-wos-/*-us-. This was added to the zero grade of the perfect. For example: Gk. (w)eid-ṓs (m.) and (w)iduĩa < *widusih₂ (f.) 'knowing'.
One of the more pervasive formations was the *tó-participle. This was added to the zero-grade of the root and provided the semantic equivalent of the modern past participle in English. In particular, when suffixed to a transitive root, the *tó-participle took on a passive sense, while when suffixed to an intransitive root, the sense remained active: *gʷʰen- 'kill' gives *gʷʰṇ-tó- > Skt. hatá- and Gk. -pʰatós '(having been) killed'; but *gʷem- 'come' yields *gʷṃ-tó- > Skt. gatá- and Gk. -batós '(having) come'. Filling an analogous role is the *nó-participle, which carries a similar connotation: PIE *bʰid-onó- > Eng. bitt-en.
The following Tocharian A text is A255, now THT 888 in the new Berlin numbering system. This was the first text published by Sieg and Siegling, appearing in 1908.
The text contains a list of the deeds of past Buddhas, a list also found in the Turkic text Maitrisimit nom bitig. Though there is as yet no reason to suppose that the present Tocharian text forms the basis of the Turkic version, they evidently treat the same material. The general form of the text as a whole is as follows.
Chapters 1--4 relate the story of how Maitreya meets the historical Buddha Śakyamuni, who tells Maitreya that he is the future Buddha. Chapters 5--9 tell of Maitreya's preparation in the city of Katumati. In chapter 10 Maitreya descends from Heaven to Earth. There follows a biography of his youth and the typical three encounters of a Buddha. In chapter 12 Maitreya leaves the city and begins his life as an ascetic. In chapters 16--19 Maitreya proceeds to convert various people he encounters. In the "chapters on health", chapters 20--25, he counsels people concerning the perils of Hell. Maitreya converts King Singha in chapter 26, who subsequently becomes a monk. In the last chapter Maitreya converts his mother, and he himself finally enters Nirvāṇa.
26 - okät-tmāṃ puklā wrasaśśi śolaṃ Vipaśyi ñomā ptāñkät ṣeṣ.
27 - säm käṣṣi āṣānik tmāṃ-ñu-wälts-puklyi puttiśparäṃ kälpāt.
śtwar-tmāṃ päñ-wälts puklā puttiśparṣinās wlesant wleṣāt.
tmāṃ ṣäk-wälts puklā śol lyalyipuräṣ ksaluneyaṃ kälk.
28 - ṣpät-tmāṃ puklā wrasaśśi śolaṃ Śikʰi ñomā ptāñkät ṣeṣ.
säm penu kāruṇik tmāṃ-ṣäk-wälts-puklyi puttiśparäṃ kälpāt.
śtwar-tmāṃ puklā puttiśparṣināṃ wles wleṣāt.
tmāṃ śtwar-wälts puklā śol śkā lyalyipuräṣ ksaluneyaṃ kälk.
26 okät-tmāṃ puklā wrasaśśi śolaṃ Vipaśyi ñomā ptāñkät ṣeṣ. 27 säm käṣṣi āṣānik tmāṃ-ñu-wälts-puklyi puttiśparäṃ kälpāt. śtwar-tmāṃ päñ-wälts puklā puttiśparṣinās wlesant wleṣāt. tmāṃ ṣäk-wälts puklā śol lyalyipuräṣ ksaluneyaṃ kälk. 28 ṣpät-tmāṃ puklā wrasaśśi śolaṃ Śikʰi ñomā ptāñkät ṣeṣ. säm penu kāruṇik tmāṃ-ṣäk-wälts-puklyi puttiśparäṃ kälpāt. śtwar-tmāṃ puklā puttiśparṣināṃ wles wleṣāt. tmāṃ śtwar-wälts puklā śol śkā lyalyipuräṣ ksaluneyaṃ kälk.
26 For eighty thousand years in the life of beings there was a Buddhalord Vipasyin by name. 27 This praiseworthy teacher, possessing nineteen thousand years, attained Buddhahood. For forty-five thousand years he performed the deeds of Buddhahood. After sixteen thousand years, having left life behind, he attained Nirvana. 28 For seventy thousand years in the life of beings there was a Buddhalord Sikhin by name. This compassionate (man) as well, possessing sixteen thousand years, attained Buddhahood. For forty thousand years he performed the service of Buddhahood. After fourteen thousand years, having in addition given his life, he attained Nirvana.
The foundation of all modern historical linguistics lies in the study of phonological development. Two given languages cannot be shown to be either genetically related or genetically unrelated if not by phonological rules: to show linguistic relationship one must establish, either by documentary evidence or by hypothesis, the phonological system of a parent language, and then show how this phonological system develops by distinct rules, with little to no exceptions, into each of the phonological systems of the languages under consideration. The introductions to preceding lessons have discussed the phonological system of Proto-Indo-European in some detail.
The task remains to establish how the PIE phonological system thus develops into the phonological systems of the two Tocharian languages. Due to the relatively recent discovery of Tocharian relative to the other archaic Indo-European languages, the process of pinning down how the PIE system developed into that of Tocharian is very much alive and an area of active debate. These lessons will try to walk a middle path, for the most part quoting standard phonological developments where there is general consensus, but also providing some less than standard conjectures when deemed useful for better understanding, and even on occasion leaving out the phonological history altogether if there is no real consensus or if the rules are technical and of dubious pedagogical value for a first pass through the language.
For clarity, it is easiest to break the changes undergone by vowels into stages, first demonstrating how the sounds of Proto-Indo-European (PIE) resulted in the phonetic inventory of Proto-Tocharian (PToch or PT; also Common Tocharian, CToch), and then showing how these sounds in turn developed into those of the documented languages. This procedure of breaking the historical evolution into stages actually helps shed light on the possible historical relationships Tocharian A and Tocharian B can have with one another. Specifically, one can say that B descends from A only if the sounds of B can be regularly derived from those of A. To put it another way, in deriving Tocharian B from Proto-Tocharian, we would have to be able to consistently insert Tocharian A as an intermediate stage. In current research on the Tocharian languages, this does not seem to be feasible; there are situations in which the Tocharian B reflex must develop directly from Proto-Tocharian, and Tocharian A cannot be interpolated as an intermediate stage. This supports the conclusion that Tocharian A is not somehow an 'archaic' form of Tocharian, and Tocharian B its 'vulgar' or 'popular' or 'modern' form. The two languages most likely developed separately from Proto-Tocharian.
The starting point, therefore, is the Proto-Indo-European vocalic system, as depicted in the following chart.
|High||*i, *ī||*u, ū|
|Mid||*e, ē||*o, ō|
In Proto-Indo-European, vowel quantity (length) was an important distinction, each of the vowels having a lengthened counterpart, here denoted by a macron. In some sense, these lengthened counterparts denote a secondary development, whereby a sequence of short vowel plus laryngeal simplified into a single long vowel. For example, PIE *eh₂ > *ā. Thus the above depicted long vowels may be understood as a simplified notation for certain sequences of the form VH, where V is some vowel, and H some laryngeal.
Scholarly opinion still fluctuates as to the exact phonological system which must be hypothesized for Proto-Tocharian. For much of what these lessons treat, the following system will suffice.
It bears repeating that, where Tocharian is concerned, ā does not denote quantity (length), but rather quality. The sound represented by ā is short, or to be more precise, indifferent to length. One can see then in the above chart that Proto-Tocharian did not distinguish vowel quantity; length was not a phonemic characteristic in the Tocharian vocalic system.
We also see in the above chart the emergence of two vowels denoted *æ and *å. The position of these vowels within the above chart remains, to date, an educated guess -- there is little approaching scholarly consensus on the actual phonetic qualities of these vowels. In particular, though *æ uses the archaic English orthography for the letter ash, it does not necessarily represent the sound denoted by a in Modern English hat. Why then use this symbol? Simply put, it provides a convenient mnemonic: the Proto-Tocharian vowel represented by *æ results in Tocharian A a and Tocharian B e. The symbol merely combines its outcomes in the respective daughter languages. Similarly for *å: the actual phonetics of the vowel are not altogether clear; but nevertheless PToch *å results in Tocharian A a and Tocharian B o. Their participation in Tocharian umlaut, discussed below, does however lend credence to the idea that their relative phonetic values may be similar to those depicted in the chart.
It remains to describe how the PIE vocalic system descends into that depicted above for Proto-Tocharian. We outline this below, but first we briefly discuss how the Proto-Tocharian system above develops into the systems of the respective daughter languages. This will facilitate understanding of the examples quoted for the development from PIE to PToch.
To guide one through the maze of historical phonology, there are some rules that, though not exceptionless, may be broadly stated and on which we may hang our hats. These are the following:
These rules play an important role in the evolution of the majority of Tocharian words.
For the most part, the remaining PToch vowels develop into what one would expect in Tocharian A and B. That is, as mentioned above, PToch *æ and *å have reflexes which differ among the documented languages; but with the cleverly chosen mnemonic orthography, the outcomes are transparent. As for the other PToch vowels, they essentially pass unchanged into the daughter languages. One exception to this tendency, however, is the effect the accent has on the series of central a-type vowels. This forms the subject of the following section.
The reader may have noticed the parentheses surrounding PToch *a in the above chart depicting the PToch vocalic system. Certainly this vowel appears in the documents of both Tocharian A and B; recall the three-way distinction displayed in the chart below.
In that sense AB a may be supposed to form part of the PToch phonetic inventory. But as we will see in the examples to follow, the presence in the documented Tocharian languages of the vowel AB a rarely results directly as a development from PIE to PToch. Many etymologies allow the appearance of PToch *a to derive from synchronic accentual processes within PToch itself, rather than from historical inheritance from PIE.
Specifically, the vowels PToch *ä and *ā show an important interplay with the PToch accent. In fact in many examples they provide our only clue regarding where the PToch accent resided. Note the following alternations.
|*ä vs. *a||*ä||*á|
|*a vs. *ā||*a||*ā́|
Thus we may say that PToch *ä, when accented, yields *a; and PToch *ā, when unaccented, yields *a. In this way, PToch *a appears as the result of secondary processes, and not directly as the descendent of a PIE vowel.
Palatalization plays an important role in Tocharian phonology, and it is useful to distinguish palatal vowels, which share certain changes in historical phonology, from the non-palatal vowels which do not. We treat first the non-palatal vowels.
11.4.1 PIE *a
The Proto-Indo-European short *a, whether the (rare) original *a or the later result of laryngeal coloring PIE *h₂e > *a, yields Proto-Tocharian *ā. Consider the following chart.
|*a||*ā||*h₂elyo-||*āl'æ + kä||A ālak||Lat. alius|
|*h₂ént-o-||*āntæ||A ānt||Hitt. hant-|
|'in front'||B ānte||Gk. ánta-|
|*sak-ro-||*sākræ||A sākär||Lat. sacer|
We note in the first example the role played by accent in deriving the Tocharian B form: *h₂elyó- > *ā̀l'æ + kä > B alyek, showing the emergence of B a as a result of the PToch unaccented *ā̀. Lest this seem a purely academic matter of historical interest and of lesser import from a synchronic perspective, consider the following example.
In the above example, perhaps inherited directly from PIE *bʰag-o- or perhaps a Middle Iranian loanword borrowed during the Proto-Tocharian period, we see that the Proto-Tocharian accent (often extending back to the PIE accent) has direct consequences in the vocalic alternation exhibited within nominal paradigms. Here we see it is not enough to view the plural form B pakenta as merely the synchronic singular with the addition of a plural suffix; rather we find a change in the root vowel which results from a shift of accent in PToch itself. In the singular, the accent fell on PToch *ā, while in the plural it fell on PToch *æ.
Some scholars suggest that PIE *a < *h₂e may possibly yield PToch *ä in absolute final position, for example in the present tense, second person singular ending, coopted from the PIE perfect ending: PIE *-t h₂e > PToch *-tä > A -t B -t(o) (with 'moveable-o').
11.4.2 PIE *ā
Proto-Indo-European long *ā (including PIE *eh₂ and in general *aH) results in PToch *å. Consider the following examples.
|*ā||*å||*bʰrā́tēr||*pråcær||A pracar||Eng. brother|
|'brother'||B procer||Lat. frater|
|*wā́stu||*wåstä||A waṣt||Ved. vā́stu-|
|'dwelling'||B ost||Gk. (w)ástu|
Note in the second example above the typical Tocharian A shift -st- > -ṣt-.
There is evidence that PIE *ā may also result in PToch *ā, for example: PIE *peh₂-sḱe / o- > PToch *pāsk- > A pās- B pāsk- 'protect'. This however may be alternately explained as deriving from an original zero-grade, PIE *pʰ₂-sḱe / o-, and the general rule that a vocalized laryngeal yields PToch *ā. As another example, PIE *wleh₂nt-ih₂ > PToch *wlāntsā > B lāntsa 'queen' may be alternately explained as PIE *wlh₂nt-ih₂ > PToch *wlāntsā > B lāntsa, given that the sequence PIE *-h₂nC- vocalizes the nasal and yields PToch *-ānC-.
11.4.3 PIE *o
Proto-Indo-European short *o, including *h₃e > *o, results in PToch *æ. Consider the examples in the following chart.
|*o||*æ||*h₃ékʷ-s||*æk||A ak||OE ēage|
|'eye'||B ek||Gk. ṓps|
|*ḱṇtóm||*käntæ||A känt||Lat. centum|
|'hundred'||B *kánte||Gk. hekatón|
|*ǵómbʰo-||*kæmβæ||A kam||Skt. jambʰa-|
|'tooth'||B keme||Gk. gómpʰos|
We also find an example in the pronoun: PIE *so > PToch *sæ > A sa- B se. The addition of suffixes prevents the a- of A sa- from standing in final position, thereby preventing its loss. The following example again shows the importance of the accent in Proto-Tocharian.
|'a rushing, moving forward'||'river'||B cáke||Av. táčati|
|Pl.||> B *c(ä)ké-ntā̀||B ckenta|
In Tocharian B, PIE *mózgo- 'knot' becomes PToch *mæskæ > B meske 'joint'; but the development in Tocharian A shows an example of the insertion of -ä- to break up a consonant cluster: PToch *mæskæ > A *mask(a) > A masäk.
In some instances PIE *o remains PToch *o, but this generally results from assimilation of the normal result PToch *æ to a following PToch *u or *o < PIE *ō or *ā (umlaut). For example, PIE *(d)oru > PToch *æru > PToch *or(u) > AB or 'wood'.
11.4.4 PIE *ō
Proto-Indo-European long *ō, including *eh₃ and *oH both giving *ō, results primarily in PToch *ā. Consider the following examples.
|*ō||*ā||*ṇ-ǵneh₃-tih₂||*ā̀knā́tsā̀||A āknāts||Lat. ignōtus|
|'not knowing, foolish'||B aknā́tsa|
A more detailed explanation of the first example proceeds as follows: PIE *ṇ-ǵneh₃-tih₂ > PToch *æn-knā-tsā by normal treatment of the vocalized nasal and the semivowel PIE *-i- followed by the laryngeal *h₂. The initial *æ subsequently assimilates to the following *ā, and the cluster *-nkn- simplifies by dissimilation to *-kn-. Thus the entire sequence becomes PIE *ṇ-ǵneh₃-tih₂ > PToch *æn-knā-tsā > *ān-knā-tsā > *ā̀knā́tsā̀ > A *āknāts B aknā́tsa.
PIE *ō however has a varied treatment in Tocharian. In absolute final position, PIE *-ō# becomes PToch *-u#. This PToch final *-u# is subsequently lost in most situations. A similar situation obtains when PIE *ō precedes a final *-s, *-y, or *-(n). Again we find the development PIE *-ōs#, -ōy#, -ō(n)# > PToch *-u# (and subsequent loss). Occasionally PIE *-on# gives a nasalized vowel PToch *-õ, often causing umlaut of the preceding vowel. PIE *ō before a final *-r#, by contrast, results in the normal PToch *ā, and the final *-r remains. The following chart provides examples of these developments.
|'I drink'||> *yoku||B yoku||Lat. ēbrius|
|'having drunk'||> *yoku||B yoku|
|*ōn#||*-õ#||*ḱléu-mōn||*klyāumõ||A klyom||Skt. śrutá|
|'possessing fame'||> *klumo||B klyomo||Gk. klutós|
|*ōr#||*-ār#||*h₁ésh₂-ōr||*yäsār||A ysar||Hitt. ēšhar|
Note in the first two examples that the form B yoku is ambiguous, taken by some scholars as the first person singular of the subjunctive, but taken by others to be the nominative masculine singular of the preterite participle. The fact that both PIE *-ō# and *-ōs# yield PToch *-u# allows the form to descend from either a first person singular or from the PIE perfect participle. Note also that the final *-u causes umlaut of the preceding vowel: PToch *æ > *o.
11.4.5 PIE *u
Proto-Indo-European *u has two outcomes in Proto-Tocharian, *u and *ä, in some sense both equally likely. Consider the following chart.
|*u||*u||*yudʰ-sḱe / o-||*yut-sk-||A yutk-||Ved. yudʰ-|
|*ä||*h₁rudʰ-ró-||*rätræ||A rtär||Lat. ruber|
|'red'||> A *rätär||B|
|*dʰug h₂tēr||*täkācær||A ckācar||Ved. duhitár-|
|'daughter'||B tkā́cer||Gk. tʰugátēr|
|*h₁ludʰ-e-t||*läcä||A läc||OIr. luid|
|'(s)he went out'||B lác||Gk. ḗlutʰon|
In the derivation of Toch A rtär we again find the appearance of -ä- employed to break up a consonant cluster: PIE *h₁rudʰ-ró- > PToch *rätræ > A *rätr > A *rätär > A rtär. With PIE *dʰug h₂tēr, we find in Tocharian A assimilation of the initial PToch *t- to the following *-c-: PToch *täkācær > A *tkācar > A ckācar. The forms A läc B lac show how the preterite forms of the root AB lät- 'go out, emerge' still preserve features of the corresponding PIE form also found in Greek and Old Irish.
The development of A por and B puwar poses problems. According to the above rules, we might surmise B puwar < PToch *púwār < PIE *puwōr. However the connection between this PIE form and PIE *peh₂wṛ > Hitt. pahhur 'fire' remains difficult. The form A por may come from an altogether different form in Proto-Tocharian itself.
In initial position, PIE *#u- may develop a labial on-glide (similar to the situation we find for palatal vowels). The seeming necessity of such a rule becomes most evident in the development of A wär B war 'water'. In addition to the labial on-glide, the development seems to employ laryngeal hiatus: PIE *uh₁-ṛ > PToch *wä(H)är > *wäär > *wär > A wär B war.
11.4.6 PIE *ū
Most cross-linguistic evidence suggests that PIE long-*ū only results from the simple semivowel *u followed by a laryngeal: *uH. Some scholars suggest that Tocharian may treat this combination differently depending on which laryngeal followed the semivowel. It appears that PIE *uh₁ > PToch *wä, while both PIE *uh₂ and *uh₃ result in PToch *wā. Compare the following example.
|*uh₂||*wā||*wāstu- + -h₂||*wåst(u)wā||A waṣtu (pl.)|
|(u-stem) + (nt. nom/acc. pl.)||B ostuwa (pl.)|
We see in this example the origin of a new Tocharian plural suffix A -u B -uwa, which is then extended beyond historical *u-stem nouns. (Compare the singular form of the same noun, listed in the examples for PIE *ā.)
It should be clearly understood however that, at present, this non-uniform treatment of *uH is an inferral based on
Thus, to date, Tocharian has only furnished clear examples of a development PIE *uH > PToch *wā. For the development of PIE *iH, see below.
The palatal vowels provide the catalyst for the palatalization of consonants which plays such an important role in both the diachronic and synchronic aspects of the Tocharian languages. We must therefore have a clear understanding of the historical development of the original PIE palatal vowels. We turn to their treatment in this section.
11.5.1 PIE *i
Proto-Indo-European *i results in Proto-Tocharian *ä. Palatalization of the preceding consonant accompanies this change, regularly when the consonant is a dental stop or *n, but less so when the consonant is a velar. Consider the following examples exhibiting palatalization.
|'fist'||B maśc||Skt. muṣṭi-|
|*kuH-ti-||*kwācä||A kāc||Lat. cutis|
|'show'||> *tsäk-||B tsk-aññ-||Ved. diś-|
Here C' denotes a palatalized consonant corresponding to the consonant C. Note in the example of *mus-ti-m, the *i causes palatalization of the immediately preceding *t > *c, and the resulting *c causes palatalization of the immediately preceding *s > *ś. Note also in the example of *kuH-ti- the general rule *kw- > A k- in Tocharian A. Both examples show that the previously assumed development PIE *ti > PToch *si > *ṣi does not generally hold true.
In certain environments, the consonant preceding the *i does not palatalize. This generally holds for velars, as in the following examples.
|'shadow'||B skiyo||Gk. skiá|
|*kʷi-so||*kʷäsæ||A kus||Lat. quis|
|'who'||B kuse||Hitt. kuiš|
|*wis-o-||*wäsæ||A wäs||Skt. visá-|
|'poison'||B wase||Lat. vīrus|
|*-mesi||*-m'äsä||A -mäs||Ved. -masi|
|1st pers. pl.||B|
One might surmise from the example of *kʷi-so that the labiovelar *kʷ simply does not exhibit palatalization, but this conclusion would be erroneous. We do in fact find palatalization when the labiovelar is followed by *y: PIE *kʷy > ś. Moreover, we see in the same example that PToch kʷwäsæ must have had the accent on the the second syllable; if not, the development into Tocharian B would have been *kʷäsæ > B (!)kwáse, which we do not find.
11.5.2 PIE *ī
As mentioned in the discussion of the development of *uH, *iH has a twofold development within Tocharian. The evidence suggests that PIE *ih₁ > PToch *yä, while both PIE *ih₂ and *ih₃ result in PToch *yā. Compare the following examples.
|*ih₁||*yä||*kolh₁ + -ih₁-mi||*kælā-yä-m||A||Skt. cárati|
|'move' (sbjnct.) + (opt. marker)||B kaloym||Lat. colere|
|'queen'||> *wlāntsā||B lāntsa||OIr. flaitʰ|
We see in the example of *kolh₁ that the Proto-Tocharian subjunctive marker combined with the reflex of the PIE optative marker *-ā-ih₁- > A -e- B -oy- to provide the origin of the optative and imperfect markers in Tocharian.
11.5.3 PIE *e
Proto-Indo-European *e has two different reflexes, depending on whether it is word-initial (perhaps following an initial laryngeal *h₁, which does not color the *e) or directly follows a (non-*h₁) consonant. Specifically, we have the following.
Here C stands for any arbitrary consonant, and *C' stands for its palatalized variant. We see that, understanding *C' as the result of *Cy, these reflexes essentially result from the same process, whereby the original *e develops a palatal onglide. Compare the situation in English with u: uvula [yu-vyu-lā́] and nature [ne-cṛ]. The following examples serve to illustrate the historical development.
|*#(h₁)e||*yä||*h₁eḱw-o-||*yäkwæ||A *yäkʷ > yuk||Skt. áśva-|
|'horse'||B yákwe||Lat. equus|
|'in the eyes'||> *yänæś||B yneś||Eng. eye|
|*Ce||*C'ä||*ḱlew-o-s||*kl'äwæ||A klyu||Gk. klé(w)os|
Note in the example of *h₁eḱwo- that, in Tocharian A, the labiovelars trigger rounding of a neighboring vowel: yäkwæ > A *yäkw > yuk.
Though there are similarities between the developments of *i and *e, the two do in fact generally distinguish themselves. Compare the development of *si with *se in the root *loḱs- 'salmon', with zero-grade *lḱs-:
In general, all consonants are subject to palatalization preceding *e, whereas only certain types of consonants are subject to palatalization before *i.
11.5.4 PIE *ē
Proto-Indo-European *ē, or *eH, also has two different reflexes, depending on whether it, or *h₁e, is word-initial or whether it directly follows a consonant. Specifically, we have the following.
Thus *ē follows its shorter cousin in developing an onglide which results in initial *#y- or palatalization of the preceding consonant; the difference lies in the vocalism of the underlying vowel, which in the case of *ē gives PToch *æ. This, of course, is also the Proto-Tocharian reflex of PIE *o > PToch *æ; thus PIE *ē and *o only distinguish themselves in Proto-Tocharian by the presence or absence, respectively, of preceding palatalization. The following examples will serve to illustrate the situation.
|*#(h₁)ē||*yæ||*h₁ēgʷʰ-||*yækʷ-||A yok-||Hitt. eku-zi|
|'drink' (long grade)||B yok-||Lat. ēbrius|
|*Cē||*C'æ||*meh₁-nēs||*m'æ-ñæ||A mañ||Gk. mēn|
|'measure'||> *mæ-ñæ||B meñe||Lat. mēnsis|
|*h₂wéh₁-nt-o-||*w'æntæ||A want||Ved. vāta-|
|'wind'||B yente||Lat. ventus|
Note in the example of the long grade *h₁ēgʷʰ- of the root *h₁egʷʰ-, the Proto-Tocharian form *yækʷ- apparently undergoes a process of dissimilation, resulting in the common -o- vocalism of the root in both Tocharian A and B. In the example of *meh₁-nēs, note that both *m and *n palatalize due to the following *ē, but PToch *m' regularly depalatalizes to AB m. Observe in the example of *h₂weh₁nto- that the reflex of palatal PToch *w' in Tocharian B is y, whereas the same phoneme depalatalizes in Tocharian A to give w.
Diphthongs exist in both Tocharian dialects; only in Tocharian B, however, do the original PIE diphthongs remain diphthongs into the daughter language. In Tocharian A, all such original diphthongs undergo monophthongization. The diphthongs that do arise in Tocharian A, therefore, always constitute innovations from a diachronic perspective. The evolution of PIE diphthongs is outlined in the following chart.
The following table provides some examples illustrating a number of the above developments.
|'town'||B *yike > ike||Lat. vīcus|
|*ay||*ai||*h₂ei-sḱ-e / o-||*ais(k)-||A esam||Hitt. pāi-|
|'give' (pres.)||B aiskau||Gk. aínūmai|
|*oy||*æi||*pe-póiḱ-||*p'ä-pæik-||A papeku||Skt. piṃśati|
|'paint' (perf.)||B pāpaikau||Lat. pingō|
|'light'||B lyuke||Ved. rócas|
|*ow||*æu||*pout-sḱo-||*pæutk-||A potkam||Lat. putāre|
|'prune' (subjnct.)||B pautkau|
One major result of the evolution of PIE diphthongs in Tocharian is the blurring of the original PIE ablaut patterns. If we consider *e-, *o- and zero-grades of a root such as *peiḱ- 'paint,' we find the following reflexes for the root in Tocharian.
The zero grade eventually loses out, and one finds a two-way, rather than three-way, ablaut pattern: A pik-/pek- B pik-/paik-. More specifically, the present B piṅkäṃ shows a nasal infix. In terms of PIE, however, this should be infixed to the zero grade: *pi-n-ḱ- > (!)p'äṅk-, which we do not find. Instead, Tocharian has interpreted AB pik- as the new zero-grade, obscuring the original ablaut pattern.
Moreover we see f rom the above chart that the monophthong A e and the diphthong B oy found in the optative and imperfect do not descend from PIE *oy, but rather from the coalescence of the subjunctive marker *ā with the PIE optative marker *ih₁. See the discussion concerning the development of *iH.
Umlaut describes the process whereby the particular articulation of one vowel affects the articulation of another vowel, usually in the immediately preceding syllable. In these terms, this is a general process found in the act of speech throughout languages the world over. However, in some languages, the assimilation is strong enough and pervasive enough to cause a phoneme shift in certain environments. Such processes have been at work throughout the Germanic languages for thousands of years, assimilating the *o of Proto-Germanic *konungaz to the following *u to produce Old English kyning, and further shaping the y to finally produce Modern English king.
Similar processes run through the Tocharian languages. In particular, Proto-Tocharian *u, *ā, and *o each can trigger the umlaut of vowels in the immediately preceding syllable. The following chart illustrates the process.
|'wood' (sg.)||B or|
|'wood' (pl.)||B ārwa|
In the last example, we see that the *o-vocalism of *å can trigger *o-umlaut. As a further example, consider PIE *kléu-mo(n) 'to be heard' > PToch *klyäu-mo > *klyumo > A klyom B klyomo 'noble', cf. Skt. śrutá-, Gk. klutós. Thus the *o-vocalism that produces *o-umlaut can either come directly from PIE, or it can arise within the Proto-Tocharian period itself.
Where Tocharian failed to retain the original Indo-European cases, it seems to have overcompensated for the loss by generating new cases. These secondary cases are generally the result of the merger of the oblique (old accusative) case with postpositions that followed. A few of these postpositions survive as free-standing particles in the documented Tocharian languages, though these particles often do not have the same meaning as the related secondary case endings. The endings themselves are the same in both singular and plural, the difference in forms often coming from the use of the oblique singular or plural, as context demands, and accompanying sound changes at morpheme boundaries.
The development of the secondary cases seems to have been fairly recent in the history of Tocharian. This is suggested by the fact that only a few of the cases employ the same ending or postposition in the two languages, the majority in Tocharian A differing from those in Tocharian B. The new cases are far more regular in form than the primary cases, and have led to Tocharian acquiring in part an agglutinative language structure in nominal morphology.
Both Tocharian languages show reformed ablative and locative cases. These express the same notions of source and location, respectively, as their PIE predecessors, but were early lost and subsequently reformed in the Tocharian languages. The two languages also possess the following new types of secondary cases: perlative, denoting 'by' or 'through' someone's agency; comitative, expressing accompaniment; allative, expressing the 'place towards which'. Tocharian A alone has a reformed instrumental case, expressing means or instrument; on the other hand Tocharian B alone has a causal case, though the function is similar to the instrumental of Tocharian A.
The following table gives the secondary case endings in the two Tocharian languages, as well as the Proto-Tocharian antecedents where they are discernible. The last two columns show particles etymologically related to the case endings.
|Secondary Endings||Toch. A||Toch. B||PToch.||Pcle. A||Pcle. B|
|Ablative||-äṣ, -aṣ, -āṣ||-meṃ||ṣu||mante|
There are several classes of nouns in the Tocharian languages. They are distinguished based on the forms of the nominative and oblique plurals. The basic system of declension is nevertheless fairly uniform and simply illustrated. Each noun has synthetic forms for the primary cases, where a given case will generally have a different form for singular and plural. Though there is some uniformity in the endings of the genitive, the nominative may have a wide variety of forms. Some nouns have oblique singular identical to the nominative singular, while others do not. Human nouns have oblique singular ending -ṃ.
The complexity of the primary cases is counterbalanced by the simplicity of the secondary cases. These are agglutinative, so that each case is represented by one form and one form only -- the same suffix is applied to both singular and plural. The secondary case endings are suffixed to the appropriate form of the oblique: the singular of the secondary cases is formed by adding case suffixes to the oblique singular; the plural of the secondary cases is formed by adding case suffixes to the oblique plural.
The basic paradigms are illustrated by the nouns A yuk B yakwe 'horse' and A oṅk B eṅkwe 'human'. These are shown in the following tables. Note that the instrumental case, present only in Tocharian A, is typically not used with human nouns. The causative case is only found in Tocharian B.
|Non-Human||Toch. A||Toch. B|
Tocharian suffers from what one might call the poverty of an inflectional impulse. That is, Tocharian does not inflect nouns more than it has to. When a nominal phrase is in a secondary case, generally all the constituent words appear in the oblique case, except for the last word, which carries the proper ending of the secondary case. This of course makes sense from the point of view of seconday case development: the secondary cases were in origin usually the oblique case of the noun followed by a postposition. The words then are all in agreement in the underlying oblique case, and the postposition (i.e. secondary case ending), postposed to the final element, governs the phrase as a whole. This situation is termed Gruppenflexion, group inflection. Consider the following example from Tocharian A:
|'with the powerful mighty artificial beings'.|
And from Tocharian B,
|'with body, word, (and) thought'.|
For further discussion, see the section on Coordination.
Adjective Class I comprises originally thematic adjectives. That is, the morphology of adjectives in this class ultimately derives from PIE adjectives with thematic *-o- in the masculine and neuter. This particularly shows itself in the plural endings.
|Obl.||*-o-ns||*-æns||*-a(i)ns > -es||-eṃ|
The singular shows greater variability. In general, the adjective endings show a conflation between simple endings of the PIE thematic declensions and "extended" endings ultimately derived from PIE *n-stems. The latter lacks a completely regular distribution, at times providing the unique ending for a particular gender, case, and number; at times providing an alternate long ending. The masculine oblique provides the most salient example: because of the general loss of final consonants in Tocharian, the masculine oblique singular ending -äṃ cannot derive from the ubiquitous PIE masculine accusative singular ending *-m, since this was word-final and therefore lost. Ultimately the Tocharian ending must derive from the nasal suffix itself of *n-stem adjectives.
The feminine ending *-eH₂ or *-iH₂ also appears among the thematic adjectives. This leads to the vowel *-ā- characterizing the feminine of Tocharian adjectives. The PIE *-i- remains as *-y-, which often results in the (secondary) palatalization of the preceding consonant.
As generally with Tocharian nominal endings, these forms derive from various sources. The oblique singular shows the intrusion of the *n-stems; the plural, in addition to showing the athematic ending *-es, also shows possible traces of the thematic ending *-ns. Both of these endings exhibit augmentation via *+nā, ultimately deriving from the neuter plural ending. Moreover the feminine plural in *-H₂ > *-ā demonstrates a direct holdover from the neuter plural.
These endings divide into three major groups. These distinguish themselves according to the presence or absence of -y- (which palatalizes a preceding consonant, according to secondary palatalization), and according to the ending of the plural.
|Feminine||Group i||Group ii||Group iii|
|N Pl.||-ana||-yana, -ona||-a|
The general thematic adjective case endings appear as follows.
|Them. Adj.||A Masculine||A Neuter||A Feminine||B Masculine||B Neuter||B Feminine|
|N Pl.||-e||-aṃ||-aṃ||-i||-ānā, -onā||-ānā, -onā|
|Obl.||-es||-aṃ||-aṃ||-eṃ||-ānā, -onā||-ānā, -onā|
In the above the term neuter represents the so-called alternating gender in Tocharian, that is, the forms associated with nouns taking masculine concord in the singular and feminine concord in the plural. Tocharian A āṣtär B astare 'pure' illustrates the declension of primary adjectives.
|Class 1 Primary||A Masculine||A Feminine||B Masculine||B Feminine|
|N Sg.||āṣtär||āṣtri||astare (āstre)||astarya|
|N Pl.||āṣtre||āṣtraṃ||astari (āstri)||astarona (āstrona)|
|Obl.||āṣtres||āṣtraṃ||astareṃ (āstreṃ)||astarona (āstrona)|
Tocharian greatly extends its adjectival inventory by means of a number of suffixes which it productively applies to other nominal elements to produce new adjectival forms. These largely follow the thematic declension described above, but often show greater influence from *n-stem formations.
The suffix *-i(y)o- (*-iH₂-o-) which survives in Latin adjectives in *-ius finds a reflex in Tocharian A -i B -(i)ye. Consider the following examples.
In the last example, the adjectival formation to a Proto-Tocharian nominal *ñäkät is straightforward, though the problem of relating this root to reflexes outside of the Tocharian family remains problematic. The extant forms are listed in the following chart.
|*ñäkäciyæ-||A Masculine||A Feminie||B Masculine||B Feminine|
In general the declension of such adjectives follows the pattern provided for adjectives built from the suffix A -ṣi B -ṣṣe, which we now describe.
Tocharian employs three suffixes with particular regularity in the formation of derived adjectives. First among these is the suffix A -ṣi B -ṣṣe < PIE *-syo-. An adjective X-ṣi or X-ṣṣe has the connotation "consisting of X." The forms for the various genders and cases are as follows.
|A Masculine||A Feminine||B Masculine||B Feminine|
|Obl.||-ṣi(ṃ), -ṣinäṃ||-ṣi(ṃ), -ṣināṃ, -ṣyāṃ, -ṣṣāṃ||-ṣṣe||-ṣṣai|
|N Pl.||-ṣiñi||-ṣināñ, -ṣṣāñ||-ṣṣi||-ṣṣana|
These forms are always suffixed to the oblique form of the appropriate noun, generally the singular oblique, but occasionally to the paral or plural:
|Noun||Sing. Obl.||Paral Obl.||Plur. Obl.||Adjective||Meaning|
|A||*kanwe 'knee'||kanweṃ||kanweṃṣi||pertaining to both knees|
|B||keni 'knee'||kenine||kenineṣṣe||pertaining to both knees|
Such adjectival formations generally replace in Tocharian what would be the first member of a compound noun in Sanskrit.
The next major type of derivational suffix is A -ñi B -ññe < PIE *-nyo-. Adjectives with the form X-ñi or X-ññe connote "relating to X" or "pertaining to X." For example, B perne poyśiññe 'the glory pertaining to The All-Knowing', where poyśi < po 'all' and aik- 'know' is a term commonly applied to the Buddha. Consider also the following examples.
|B ost||household||ostaññe||member of the household|
The endings follow the above paradigm, with A -ṣ- B -ṣṣ- replaced by A -ñ- B -ññ-, respectively. Interestingly, Tocharian frequently employs this adjective formation instead of an expected genetive. This parallels usage found in other IE languages, e.g. Homeric Odusḗion es dómon 'to the house of Odysseus.' For example, Dʰarmasomäññe Udānālaṅkār-ne 'in the Udanalankara (title of a treatise) of Dharmasoma,' where here the adjective (rather than the expected noun) Dʰarmasomäññe remains in the oblique case according to group inflection.
The last major type of derivational suffix is A -ts B -tstse < PIE *-tyo-. Adjectives with the form X-ts or X-tstse have the connotation 'related to X.' These adjectives often denote possession, as in B kokaletstse 'having a chariot,' from B kokale 'chariot.' The adjective A tsopats 'big' and its synonym B orotstse serve to illustrate the declension.
|A Masculine||A Feminine||B Masculine||B Feminine|
|N Pl.||(wākmtse)||tspoktsāñ (wākmtsaṃ)||orocci||orotstsana|
Tocharian A provides few examples of the feminine forms, and of plural forms in general. The above chart illustrates some of these forms with the word A wākmats 'outstanding.'
Note also the palatalization in the Tocharian B paradigm in the masculine, in particular in the oblique singular and the nominative and oblique plural. Interestingly, this palatalization does not arise from historical phonetic considerations; rather the palatalization seen in the above paradigm arises from similar palatalization in other paradigms (due to true historical morphology). In this sense, one can see here the beginnings of a morphological palatalization, a palatalization due to grammatical function rather than phonetic environment.
The past system in Tocharian essentially comprises the preterite. That is, the preterite is the only past tense formation which generally shows its own unique stem. The imperfect and optative both fall under the heading of 'past', in fact employing the same i-marker; however this marker generally applies to the present, respectively subjunctive, stem to derive the imperfect, respectively optative. Moreover, while the optative employs special endings in the active singular, the remaining forms employ the non-past endings. The same holds for the imperfect, except in Tocharian A, which employs the preterite (i.e. past) endings.
In a word, the preterite marker is PToch *-ā-, except when it's not. Less coyly, Class I provides the dominant preterite pattern, and the verbs comprising this class largely come from Indo-European verbs with root-final laryngeal. According to regular phonological changes, this PIE *H becomes PToch *ā. Note however that this vowel is lost in Tocharian A when word-final, as with all word-final vowels in Tocharian A. For example we find the mediopassive PIE *krih₂-medʰh₂ 'we buy/are bought' > B käryāmte, cf. Gk. priato.
Class II is heterogeneous across the two dialects. Tocharian A preserves the system of reduplication found for example in Gk. épepʰnon 'I slew'. In similar fashion, A käl- 'endure' shows preterite kakäl. Tocharian B on the other hand likely derives CLASS II preterites from an original *ē-grade perfect similar to that found in Latin lēgī: PIE *kērs- > B śārsa 'made known'.
The sigmatic aorist leaves remnants in the Class III preterites. As common in Greek, Sanskrit, and even Latin, these preterite forms add an *-s- to the verbal root: A arsāt B ersate 'was evoked' from A ar- B er-, cf. Gk. õrsa, aorist of órnūmi.
Classes IV and V appear to be new formations within Tocharian. Class IV preterites show a suffix AB -ṣṣā- appended to the root, while the more sparsely attested CLASS V exhibits a suffix AB -ñ(ñ)-.
Class VI only contains two verbs, AB lä-n-t- 'leave' and B käm-. They contain traces of an original thematic past tense formation. The following chart summarizes the Tocharian preterite classes.
|Preterite||Type||Stem Shape||Toch. A||Toch. B||Comparanda|
|I||athematic||PIE *-H-||katar||śtare||Gk. skídnēmi|
|II||PIE *CV-CVC, CēC||cacäl||cāla||Lat. tollō, Goth. þulan|
|III||PIE *-s||arsāt||ersate||Gk. õrsa, Skt. ṛṇóti|
|IV||PToch *-ṣṣā-||kākätkṣuräṣ||kakātkäṣṣu||Gk. gētʰéō|
|V||PToch *-ñ(ñ)-||weñār||weñāre||Gk. eĩpon, Lat. vōx|
|VI||thematic||PIE *-e / o||läc||lac||Gk. ēlutʰon|
Clearly some of the above formations, such as the stem-suffix AB -ā-, in fact overlap with other tense/mood formations: e.g. AB -ā- also characterizes subjunctive formations, and AB -s- variously marks both present and preterite classes. Scholars have not as yet outlined such thoroughgoing rules as, say, in Sanskrit, which would elucidate why a given verb with one present class would have a preterite of another given class. To date the most satisfactory explanations rest on historical grounds for those verbs which we can trace back to Indo-European antecedents, and synchronic extension to other verbs with similar phonology and/or semantics. In this regard, each verb must be considered in the light of its entire conjugational system from PIE down to Tocharian in order to reveal patterns of association between various tense/mood formations. That is, one seeks to ascertain if the PIE antecedent contains a root-final laryngeal, and if that root formed a sigmatic or lengthened-grade aorist, etc.; thereby one hopefully uncovers naturally the association between its present, subjunctive, and preterite classes.
The past endings pertain to the preterite verb forms of all classes, save Class VI, which shows some variation. Tocharian A also employs the past endings in the imperfect, with the exception of the imperfects A yem and ṣem. Tocharian A optative endings differ in the active singular. Tocharian B shows slightly different endings in the active singular of both optative and imperfect; the remaining forms employ the past endings. The following table shows the Tocharian past endings.
|Past||Active A||Active B||Mediopassive A||Mediopassive B|
For reference, we record here the optative endings, which depart from the past endings only in the active singular.
|Optative||Active A||Active B|
The Tocharian A endings therefore mimic the non-past endings in the active singular, whereas the corresponding Tocharian B forms are almost altogether different. In the remaining active forms, and in all mediopassive forms, they are identical with the past endings.
The table below presents the preterite forms of the verb AB läk- 'see'. As in the present system, A pälk- completes the suppletive paradigm in Tocharian A.
|Past||Active A||Active B||Mediopassive A||Mediopassive B|
Unfortunately the active forms of A pälk- are unattested, unlike the mediopassive. Tocharian B on the other hand shows a rather full paradigm.
A noun and accompanying adjective must agree in grammatical gender. Most Tocharian adjective endings distinguish clearly between masculine and feminine endings; few Tocharian nouns, by contrast, have terminations that signal grammatical gender with any regularity. In practice, therefore, one determines the grammatical gender of a noun by the grammatical gender of the associated adjective. Some nouns, however, take masculine concord in the singular and feminine in the plural. Such nouns therefore constitute a third grammatical gender, generally termed alternating in Tocharian handbooks. For example:
|A||tsopats (m.) wäl||śāwe (m.) lāṃś|
|B||orotstse (m.) walo||orocci (m.) lāñc|
|A||sās (f.) ytār||toṣ (f.) ytāräṃ|
|B||sā (f.) ytārye||toy (f.) ytariñ|
|A||säs (m.) oko||toṣ (f.) okontu|
|B||se (m.) oko||toy (f.) okonta|
Tocharian B, however, occasionally breaks this rule when convenient. Specifically, Tocharian B will at times employ a masculine adjectival form when it should properly use a feminine to ensure that the words fit the poetic meter.
A noun and accompanying adjective must agree in case. That is, if a noun in, say, the genitive case is accompanied by an adjective, that adjective too must be in the genitive case. Such concord pervades the ancient Indo-European languages. Tocharian, however, has decided that such a rule is more of a guideline, and therefore remains a bit lax. Specifically, recall that the secondary nominal cases derive from postpositions appended to the oblique case form of the underlying noun. This underlying oblique form leads to so-called group inflection, or Gruppenflexion: an adjective modifying a noun in a secondary case will generally be in the oblique. Consider the following examples.
|A||āṣāniāṃ||Metraknaśśäl||together with the worthy Maitreya|
|B||cämpamñecceṃ||orocceṃ||wnolmeṃmpa||together with the great (and) able beings|
As mentioned in the previous lesson, Tocharian has interestingly extended this even to the genitive case, even though this case does not derive from the union of the oblique case with a postposition. Consider the following examples.
|A||ṣomāp||lānt||of one king|
|B||cwi||yāmorntse||of this deed|
|A||śāwes||käṣṣiśśi||of the good teachers|
|B||añmalāṣkeṃ||käṣṣintse||of the compassionate teacher|
Tocharian similarly employs the oblique in a parallel string of nouns. That is, given a sequence of nouns -- be they parallel items in a list or equated to one another through apposition -- in a secondary case, typically only the last item carries the secondary case ending. Consider the following.
|A||kuklas||yukas||oṅkälmāsyo||with wagons, horses, and elephants|
|B||kokleṃ||oṅkolmaṃnpā||with wagons and elephants|
We see from the examples above that an adjective must agree in number with the noun it modifies. Similarly, a verb and its subject must agree in number.