Syntax most broadly treats the system by which the speaker constructs an utterance. Under the rubric of utterance we would include the sentence as a linguistic unit of central importance; but syntax also hopes to explain the principles underlying the use of other elements such as interjections or exclamations as well. Syntax therefore casts a wide net, and many subcategories of the study of language fall within its purview. For example, the study of morphology touches upon syntax, since many inflected languages employ affixes that encode syntactic information: e.g. the case system of early Indo-European languages. Thus, whereas say phonology can be studied more or less independently of other facets of the system of a given language, syntax depends in a crucial way on several other features of that system. Because of this complexity, what linguists mean by the term 'syntax' has often varied over the history of linguistic inquiry, and in particular over the history of Indo-European studies.
Diachronic syntax denotes the study of changes in syntactic principles in a language or language family through time. Diachronic syntax seeks to shed light on the syntax of a language at a given point in time by studying the syntax of the language at an earlier point of time and outlining certain broad principles of change that not only explain the syntactic evolution in the language under consideration, but possibly also in other languages. This forms a primary pillar of IE studies, though it has received far less attention than the other pillars of historical phonology and morphology. In some sense this comes by necessity; for instance we could not hope to say anything cogent about the development of IE syntax without first having some definitive notions of the development of IE morphology, since the former depends quite intricately on the latter. The earliest studies of IE syntax from the 19th century therefore concerned themselves primarily with a description of various morphological forms and their manners of occurrence within utterances.
Because of the complexity of syntactic investigation, as well as shifting notions of just what that study comprises, the study of PIE historical syntax has enjoyed heated debate among scholars and continues to be a rich area of inquiry.
Word order in a larger or more restricted sense has long been a focus for PIE syntactic studies. At times this results from the assertion that it affords a unifying organizational principle for many other syntactic structures; at other times from the assertion that word order has been afforded too much prominence. The forcefulness of the debate generally ensures that focus remains on word order.
In the larger sense word order refers to the order of the principal elements Subject, Object, and Verb in a (usually simple, declarative) sentence: SVO, SOV, VSO, etc. Scholars often restrict attention to the relative ordering of object and verb: VO or OV. Thus certain lines of syntactic investigation seek to understand whether typical word order in PIE is VO or OV, i.e. whether the verb typically precedes or follows its object.
We must ask: how would one go about determining this ordering for PIE? Certainly we must compare sentences across IE languages, focusing on the relative ordering of these elements. Moreover, even a cursory look at the history of English shows that such patterns can change within a language (Modern English generally exhibits SVO order, while Old English generally SOV). We must therefore take care to compare only the most archaic syntactic patterns among the IE languages. We must therefore give precedence to languages such as Hittite and Vedic Sanskrit because of their overall antiquity; and e.g. to archaic Latin rather than Classical Latin because of its greater proximity to the time period which we wish to study. Based on examples such as the following (Fortson 2004)
|Hittite||nu=za MUSilluyankas DIM-an tarahta|
|And the serpent overcame the stormgod.|
|Vedic||maruto ha enam na ajahuḥ|
|Indeed the Maruts did not abandon him.|
|Latin||Eumolpus tamquam litterārum stūdiōsus utīque ātrāmentum habet|
|Eumolpus, so interested in learning, surely has (some) ink.|
|Runic||ek hlewagastiz holtijaz horna tawido|
|I, Hlewagastiz of Holt, made (this) horn.|
|Tocharian A||kāsu ñom-klyu tsraṣiśśi śäk kälymentwaṃ säkatär|
|Good fame of the strong spreads out in ten directions.|
scholars generally agree that PIE itself likely exhibited verb-final word order.
Now one point of contention is what exactly it means to say that PIE exhibits verb-final, or OV, word order. Certainly it should not mean that it only places the verb at the end of a simple sentence: even Hittite, perhaps the most rigidly verb-final among the early IE languages, occasionally puts the verb at the beginning of an utterance. Typically this highlights the verb relative to the other elements of the sentence. If Hittite can show such variation, and if the other early IE languages are even more flexible, then certainly PIE should not be taken to be rigidly OV.
We should mention one important issue. Linguists would say in this context that OV is the unmarked word order, and any departure from that would be marked. By 'marked' we mean that the departure alone from the unmarked order itself carries extra content. We must be careful, as Watkins (1976) points out, not to confuse 'unmarked' with 'normal', or 'marked' with 'abnormal'. Linguists would treat both marked and unmarked usages as normal, each being appropriate for different contexts.
Another subtle question arises: how do we know the above examples don't merely represent several independent developments toward a fairly common word order? That is, how do we know the above examples tell us anything specific about the actual word order in PIE itself? These sentences don't even have the same words! We will return to this issue below.
What are the benefits of assigning any predetermined word order to PIE at all? The word order of Classical Greek and Latin, for example, is relatively free; we might therefore suppose this state of affairs for PIE. As a consequence, this would force us to explain the fairly rigid verb-final syntax of Hittite as an innovation. But free word order would also require us to explain many other features of early IE languages that do seem to show a preferential order. For example, genitive modifiers generally precede the nouns they modify in cognate phrases across IE languages (Clackson 2007):
|*dems potis||dám pati-||déspotis||of-the-house master|
As it turns out, OV languages typically place genitives before the noun they modify. That is, OV structure typically implies Genitive-Head order in noun phrases. We should not interpret this too strictly: the variation exhibited by the daughter languages suggests this was likely not an iron-clad rule in PIE. But what it does provide is a systematic way of evaluating the plausibility of syntactic structures proposed for PIE (Lehmann 1993): so-called typological statements such as 'OV structure implies Genitive-Head order in noun phrases' recapitulate data garnered from modern languages across the globe. Thus the assertion that PIE is verb-final and genitives precede their head nouns would say that, from a present-day perspective, PIE is in these respects a very typical language.
The earliest statements on subordination in PIE asserted that it did not exist as such. That is, PIE supposedly employed a paratactic structure, setting grammatically independent sentences beside one another and essentially employing context to suggest that one clause provided a precondition, or attendant circumstance, for the other. From this evidently developed subordination, or hypotaxis, by interpreting the clause denoting the attendant circumstances as part of the other clause. We may provide a caricature of the process employing Modern English:
|Who loves you? I saw the man.||I saw the man who loves you.|
|You are a darling. I said that.||I said that you are a darling.|
To support the idea scholars generally point to the fact that across IE languages relative and interrogative pronouns are often etymologically related, as well as deictic pronouns and subordinating conjunctions.
Recent research (see the discussion in Clackson 2007, which we follow here) suggests that this idea needs revision. There are two principal reasons:
In the midst of the controversy lie the pronouns *yo- and *kʷo-/kʷi-, which provide two possible reconstructions for the relative pronoun, assuming that this particular type of subordination did in fact exist in the PIE period. In particular, we find relative pronouns derived from *yo- in Greek, Indo-Iranian, Slavic and Celtic; while we find *kʷo-/kʷi- in this role in Anatolian, Latin and Tocharian. Without further information, we have little on which to decide whether both should be reconstructed as relative pronouns in PIE, or whether one should be reconstructed as the relative, while the other would be an innovation.
As it turns out, recent work on types of relative clauses provides a possible path to resolution. We may isolate two different types of relative clause, exhibited by the following English statements:
|Restrictive||The dog (which) I bit growled.|
|Non-Restrictive||The dog, which I later bit, growled.|
The restrictive or defining relative clause serves to delimit its antecedent, providing information necessary for the proper identification of its referent. The non-restrictive, or descriptive or appositional, relative clause, by contrast, adds only non-essential information.
If we now consider Hittite, Vedic Sanskrit, archaic Latin, and Homeric Greek with these relative clause types in mind, we find a striking pattern (Clackson 2007):
|Hittite, Latin||Vedic, Homeric Greek|
|Most common relative clause||restrictive||non-restrictive|
|Most common clause order||relative--main||main--relative|
Thus the syntactic evidence supports not only the possibility of subordination in PIE in the form of relative clauses, but also suggests that the reconstruction of two possible relative markers would fit nicely with two different syntactic substructures.
One of the earliest and most widely accepted laws to be posited for PIE is Wackernagel's Law, named after the scholar who discovered it in the early 19th century. This law asserts that enclitics, which by definition have no accent of their own, are drawn to second position in a sentence, falling behind the first accented word. The Anatolian languages in particular provide numerous examples supporting the law (Clackson 2007):
|Luwian||zam=pa=kuwa DUMU-nin wallindu|
|This child they shall lift.|
The particles appear in boldface, and they follow the demonstrative zam 'this'.
Recent work, however, has forced scholars to rethink some of the details of Wackernagel's Law. Issues arise when elements of the utterance are shifted to the front. This shift to the front, or left given our orthography, essentially occurs in two ways. One way linguists term fronting, or focus, which simply denotes the movement of an important element of the clause to the front to highlight it. In the resulting utterance, the fronted element remains within the clause, with all the attendant morphology. Linguists distinguish this movement from what they call left-detachment, which denotes the addition of material before, i.e. outside of, the clause. Compare the following examples in English (Clackson 2007):
|English||That boy I don't like.||As for that boy, I don't like him.|
Notice how, in the example of left-detachment, the pronoun him recapitulates the noun phrase that boy, which stands outside of the clause.
Wackernagel's Law, simply interpreted, should imply that enclitics follow the fronted element in a clause, regardless of whether this is preceded by a left-detached element (which, by definition, does not form part of the clause). We do in fact find this in Hittite; unfortunately this does not always work, for example, in Greek: enclitics may follow the left-detached element in Greek, preceding the fronted element. As Hale has shown (see the discussion in Clackson 2007), Vedic further complicates the issue, allowing enclitics in both positions:
|Vedic||utá vā yó no marcáyād ánāgasaḥ|
|Or also who would do harm to innocent us...|
Here utá 'also' is left-detached, while yó 'who' is fronted. Hale categorizes enclitics into two groups: those with scope over the whole sentence and connectives follow what he calls the TOPIC (i.e. left-detached) element, while enclitic pronouns follow the fronted element. But unfortunately the data fails to be as clear-cut as one would hope:
|Vedic||sám āmaṃśa sumatíbʰiḥ kó asya|
|Who has attained his good-will?|
Here the enclitic pronoun asya 'his, of him' follows the fronted element, the interrogative pronoun kó 'who'. But if kó is fronted, then what stands before it must be left-detached, and hence must fall outside the clause. However in the above case the clause requires the left-detached element to complete the clause. Thus recent developments have led to a reanalysis of even the firmest of IE syntactic laws.
Let us finally return to the question of just what it means to reconstruct, say, basic word order for PIE. In our initial examples illustrating the verb-final word order we assume for PIE, we noted that the exemplars exhibited parallel structure, but not necessarily parallel semantic elements. Watkins (1976) in a celebrated article pointed the way to a more specific form of syntactic reconstruction, which we outline very briefly here. As point of departure he takes the Old Irish relative forms of the finite verb, special forms that derive from a combination of the PIE verb with the relative marker *yo-: e.g. 3 pl. rel. gontae 'they-who-kill/whom-they-kill' < PIE *gʷenonti + yo-. He focuses on a specific proverb from Old Irish:
The focus on proverbs serves two primary purposes: they preserve archaic syntax, and they preserve ancient thematic content. Focusing for a moment on the content, we see gonas as reflecting a possible PIE
Leaving that aside for the moment, let us turn to a Greek proverb in the Iliad:
|alike to all (is)||the War God,||and||him-who-would-kill||he kills|
Here ktanéonta is a future participle in the accusative, katékta the finite verb form. The future participle lends the connotation of intension or desire, hence the translation 'would kill'. We now note that Greek largely replaces the PIE root *gʷʰen- with the synonymous kten-. Given these two points, and recognizing that participles are formally equivalent to relative clauses, we may see the Greek phrase ktanéonta katékta as the thematic and structural equivalent of an original PIE phrase
In the above the first verb form shows the optional reduplication and the suffix *-H₁se- of the PIE desiderative formation. If we finally take the same phrase, but make the construction passive, we find
This recapitulates exactly not only the morphology but also syntactic structure of the original Old Irish proverb gonas .géntar. That is, we have managed to recover an actual, specific phrase present in PIE itself!
The study of Indo-European diachronic syntax still provides some of the most promising avenues of study of the parent language and culture. Several methods may be brought to bear, and each has something to offer. If we imagine ourselves in the position of a student of some foreign language, such as German, we might wonder how to say 'Someone does something to someone else' in general; statements about typology attempt to reach answers, saying e.g. that the 'does something' part generally comes second. We might also wonder how to say 'It's always toughest at the beginning'; reconstruction of PIE poetics aims towards such answers, saying e.g. that that idea is expressed as Aller Anfang ist schwer. We do a disservice to the discipline it we force one particular method to answer questions of both types; certainly we would not want a description of how to form 'Someone does something to someone' to come in the form: 'If the verb is haben, we say...; if the verb is loben, we say...', etc. Nor would we expect proverbs to employ the same syntactic structures used in the most common discourse, since they by their nature preserve ideas and phrases that are in some sense timeless and immune to change. The study of PIE syntax can only benefit from the skilled application of both methods under appropriate circumstances.
The following selection continues the Tocharian A text A255 (THT 888). The narrative continues with a list of the deeds of past Buddhas, their lifespans, and when they attained Nirvana.
31 - tri-tmāṃ puklā wrasaśśi śolaṃ Kanakamuni ñomā ptāñkät ṣeṣ.
säm penu puk knāṃnmāṃ okät-wälts-puklyi puttiśparäṃ kälpāt.
tmāṃ ṣäk-wälts puklā puttiśparäṃ wleṣāt.
ṣäk-wälts puklā śol lyalyipuräṣ ksaluneyaṃ kälk.
32 - we-tmāṃ puklā wrasaśśi śolaṃ Kāśyap ñomā ptāñkät ṣeṣ.
säm penu āṣānik ṣäk-wälts-puklyi puttiśparäṃ kälpāt.
tmāṃ puklā puttiśparäṃ wleṣāt.
śtwar-wälts puklā śol lyalyipuräṣ ksaluneyaṃ kälk.
31 tri-tmāṃ puklā wrasaśśi śolaṃ Kanakamuni ñomā ptāñkät ṣeṣ. säm penu puk knāṃnmāṃ okät-wälts-puklyi puttiśparäṃ kälpāt. tmāṃ ṣäk-wälts puklā puttiśparäṃ wleṣāt. ṣäk-wälts puklā śol lyalyipuräṣ ksaluneyaṃ kälk. 32 we-tmāṃ puklā wrasaśśi śolaṃ Kāśyap ñomā ptāñkät ṣeṣ. säm penu āṣānik ṣäk-wälts-puklyi puttiśparäṃ kälpāt. tmāṃ puklā puttiśparäṃ wleṣāt. śtwar-wälts puklā śol lyalyipuräṣ ksaluneyaṃ kälk.
31 For thirty thousand years in the life of beings there was a Buddhalord Kanakamuni by name. This all-knowing (man) as well, possessing eight thousand years, attained Buddhahood. For sixteen thousand years he performed Buddhahood. After six thousand years, having given his life, he attained Nirvana. 32 For twenty thousand years in the life of beings there was a Buddhalord Kasyap by name. This praiseworthy (man) as well, possessing six thousand years, attained Buddhahood. For ten thousand years he performed Buddhahood. After four thousand years, having given his life, he attained Nirvana.
The reader may derive some sense of relief from the fact that we have finally arrived at the last major installment of historical phonology: the resonants and laryngeals. They allow a more streamlined presentation than that given for the vowels and the consonants, if only because they are fewer in number.
21.1.1 Medial Resonants
Resonants between consonants change in Tocharian to their consonantal equivalent, preceded by the reduced vowel *ä:
|PIE *CṚC > PToch *CäRC.|
Consider the following examples.
|*CṚC||*CäRC||*bʰṛǵʰ-||*pärk-||A pärk-||Hitt. park-|
|'rise'||B pärk-||Skt. bṛhant-|
|'wolf'||B walkwe||Goth. wulfs|
|*mṇ-sḱe / o-||*mänsk-||A mäsk-||Lat. maneō|
|'remain'||B mäsk-||Gk. ménō|
In the last example, PIE *mṇ-sḱe / o- > AB mäsk-, the consonantal *n is lost regularly before *s.
21.1.2 Final Resonants
With resonants in final position we find the same process as in medial position as described above, except when the resonant is PIE *n. That is, resonants in final position become their consonantal equivalent, preceded by *ä, and thereby remain in final position; *n in final position, however, generally falls away together with the preceding *ä:
Consider the following examples.
|*CṚ#||*CäR#||*uH₁-ṛ||*wä(H)är||A wär||Skt. vāri-|
|'water'||> *wäär||B war||ON vari|
|*Cṇ#||*Ø#||*stH₂-mṇ||*stāmä(n)||A stām||Gk. stēmōn|
|'(something) standing'||B stām||Lat. stāmen|
21.1.3 Initial Resonants
Initial resonants likewise convert into their consonantal twins. But unlike the situation for medial and final resonants, the accompanying vowel is PToch *æ rather than *ä:
|PIE *#ṚC > PToch *æRC.|
This holds for all resonants. Consider the following examples.
|*#ṚC||*æRC||*ṇ-ǵneH₃-tiH₂||*æn-knā-tsā||A āknāts||Lat. ignōtus|
|'unknowing'||> *ān-knā-tsā||B aknā́tsa|
|*H₁ṛgʷ-ont-||*ærkænt-||A arkant||Skt. rájas|
In the first example we witness a change *æ > *ā in Proto-Tocharian due to assimilation to the following *ā.
We have seen already that laryngeals have a dramatic effect on neighboring vowels. In particular, we will not repeat here the development of laryngeals in the positions *VH; the treatment can be found under the appropriate long vowel. Here we treat another laryngeals in a different context, specifically when they carry syllabic content (akin to a resonant). We also discuss the development of laryngeals when neighboring a resonant and when they form a boundary between vowels.
21.2.1 Medial Laryngeals
Laryngeals between consonants yield Proto-Tocharian *ā, and therefore fall together with original PIE short-*a in Proto-Tocharian:
|PIE *CḤC > PToch *CāC.|
Consider the following examples.
|*CḤC||*CāC||*pH₂tḗr||*pācær||A pācar||Skt. pitár-|
|'father'||B pācer||Gk. patēr|
|*(s)tH₂-k-||*tāk-||A tāk-||Hitt. tak-|
|'stand'||B tāk-||Gk. héstēka|
|'protect'||B pāsk-||Skt. pā-|
21.2.2 Final Laryngeals
When in final position, as when between consonants, a laryngeal yields PToch *ā:
|PIE *CH# > PToch *Cā#|
This factors particularly strongly in the reflex of the neuter plural ending for the nominative and oblique.
|'meat'||> *m'äsā||B mīsa||Got. mimz|
|'bone'||> *āstā||B āsta||Gk. ospʰús|
The last example above shows the development of B āsta, plural of the noun B āy 'bone'. We see the action of umlaut in Proto-Tocharian reflected in the change *æstā > *āstā.
21.2.3 Laryngeals with Resonants
Occasionally a laryngeal follows a resonant between consonants. In such positions, the laryngeal is lost without a trace in Proto-Tocharian:
|PIE *CṚHC > PToch *CäRC.|
The resonant undergoes its normal evolution. When the laryngeal stands in initial position before a resonant, however, it might be that the treatment differs depending on whether the resonant is PIE *ṇ or not:
The evidence unfortunately is scant, and scholarly opinion has not yet reached a consensus. Consider the following example.
|*CṚHC||*CäRC||*pṛH₂-wo-||*pärwæ||A pärwat||Skt. pū́rva-|
21.2.4 Laryngeals between Vowels
Generally speaking a laryngeal between vowels falls away with subsequent coalescence of the preceding and following vowels:
|PIE *VHV > *VV > *V̄.|
Such a development gives little cause for surprise, considering the preceding discussion of laryngeals. What is rather remarkable, however, is that some evidence suggests that such laryngeal loss may have occurred close enough to the Proto-Tocharian period for the hiatus between the two vowels to remain and allow further changes. Consider the following example:
|*VHV||*VV > *V̄||*kléu-mo-H₃ō(n)||*klyäumæ(H)u||A klyom||Skt. śrutá-|
|'to be heard'||> *klyäumo-u > *klyumõ||B klyomo||Gk. klutós|
The major point at issue here is that final PIE *-ō generally produces *-u in Tocharian. We do not however find this reflex in the above example. Rather it seems that the former presence of a laryngeal may produce a hiatus that remains long enough for the following vowel, *-u, to influence the preceding vowel, *æ-, according to Tocharian umlaut. Thus *æ > *o, which subsequently coalesces with the following -u to produce a new Tocharian *õ.
In Tocharian previous scholarship typically makes a distinction between the paral and the dual. The paral forms mark pairs of quantities that are in some sense `natural,' that is, of things which naturally come in twos: eyes, ears, hands, etc. The dual forms by contrast mark 'occasional' pairs, things which just happen to be paired at the moment: two boys, two trees, etc. Just what exactly constitutes a 'natural' versus an 'occasional' pair in Tocharian leaves quite a bit of room for debate, however: certainly the paral A aśäṃ B eś(a)ne '(pair of) eyes' is to be expected, but the paral B ñaktene '(a naturally occurring pair of) gods', translated as 'the god and his wife,' might come as a surprise.
The historical perspective further weakens the distinction, since the paral forms generally derive from original PIE duals when such an antecedent can be reconstructed. That is, if we compare the singular and dual for *H₃ekʷ- 'eye', we find the following evolution:
|Dual||*H₃ekʷ-iH₁||*æśä + næ||A aśäṃ|
|'two eyes'||B eśane|
That is, the regular dual in PIE would give PToch *æś(ä), differing from the singular *æk only in the palatalization of the final consonant. This seems to have led to a recharacterization by means of a suffix *+næ (Pinault 2008). Moreover we in fact find the expected dual form B eś in the compound B eś-lmau 'the eyes set,' i.e. 'blinded.'
Because of this somewhat tenuous distinction between the paral and dual, we employ the term dual in these lessons to encompass both formations, unless stated otherwise.
As mentioned above, Tocharian often employs a nasal suffix A -(ä)ṃ B -(a)ne to characterize dual forms. Scholarly opinion is divided as to the origin of this suffix. It may perhaps reflect the PIE *n-stem formation so widespread in Tocharian (Adams 1988). On the other hand, it may reflect a demonstrative adjective in apposition to the substantive and reinterpreted as a suffix (Pinault 2008). Consider the following:
|*no- Demonstrative||*ṇkw-o-H₁e + no-H₁e||*ænkwæyä-næyä||B eṅkwene|
The first example shows the development from a dual ending applied to an *n-stem, while the second shows the development from a dual form built to the normal thematic stem, followed by another demonstrative in the dual. Note that the former argument supposes a change *-ō# > *æ in Proto-Tocharian.
Whatever the origin, the suffix *-næ evidently provides a productive means of deriving a dual form in Tocharian, and it is applied to PToch dual forms which themselves derive from PIE duals. Thus, even if not in origin a deictic adjective, PToch *-næ comes to function in PToch in much the way illustrated in the second example above. Thus we may analyze, for example, B ñaktene '(pair of) gods' as a dual ñakte inherited from PIE, followed by the PToch dual suffix *-næ.
The nominative and oblique dual endings generally follow from their PIE counterparts. The following chart summarizes the developments.
|Nom. anim.||*-o-H₁ > *-ō||*-u > *-ä|
|Voc.||*-o-H₁ > *-ō||*-u > *-ä|
|Acc. anim.||*-o-H₁ > *-ō||*-u > *-ä|
|Nom./Acc. neut.||*-o-iH₁||*-oy > *-äy||A -i B -i|
|Nom./Acc. neut.||*-iH₁||*-(i)yä||AB -yi|
The presence or absence of the thematic vowel will therefore influence the Proto-Tocharian reflex, in addition of course to the division between neuter and non-neuter substantives.
22.2.1 Thematic Nouns
With non-neuter thematic nouns, we generally find the development PIE *-o-H₁ > *-u > *ä according to normal Tocharian phonetic changes. At the same time, however, in the Proto-Tocharian period speakers tended to reform the dual with the suffix *-næ, and so the preceding *ä is either lost or retained depending on accentual patterns:
|Nom./Acc.||*ōmso-H₁||*ānsä + næ||A *āysän > esäṃ||Lat. umerus|
|'two shoulders'||B antsane, āntsne||Got. ams|
In some instances however we find a different outcome, perhaps derived from the PIE vocative and the action of Kuiper's law, whereby a final laryngeal is generally dropped. Consider the following (cf. Pinault 2008):
|Nom./Acc.||*-o-(H₁)||*ñäktæ + næ|
|'two gods'||B ñakte-ne|
We find a similar formation for neuter thematic nouns:
|Nom./Acc.||*H₂(e)nt-bʰo-iH₁||*āntäpäy||A āmpi||Gk. ámpʰō|
|'both'||B antapi, āntpi||Lat. ambō|
22.2.2 Athematic Nouns
The nominative and accusative forms of athematic nouns follow the same basic pattern. The major difference is that the absence of the thematic vowel means that the front vocalism of the vowel in the ending generally leads to palatalization of the preceding consonant. Consider the following examples.
|*-H₁(e)||*-yä||*pod-H₁e||*pæyä + næ||A pe-ṃ||Hitt. pad-|
|'two feet'||B pai-ne||Got. fōtus|
|*-iH₁||*-(i)yä||*H₃ekʷ-iH₁||*æśä + næ||A aśäṃ||Gk. ósse|
|'two eyes'||B eśane||OCS oči|
|*bʰrātr-iH₁||*prātriyä||A pratri||Eng. brother|
Note the effect of palatalization on the *nt-stem. This yields a productive dual suffix in Tocharian B, here illustrated with raso 'span'. In the final example, note that A pratri does not in fact show the expected palatalization.
In the genitive dual we find the endings A -nis B -naisäñ, with palatalization of the preceding consonant: A aśnis B eśnaisäñ 'of two eyes'. The analysis likely follows that given above for the nominative and accusative: the palatalization derives from the original PIE dual ending, and to this we find addition of a demonstrative adjective *no-. The latter shows a form also encountered in the pronouns, perhaps deriving from *-no-i-H₁u-s > *-næyäs > *-nais-.
Adjectives with stems in *nt- form a major category in the Proto-Indo-European inventory. Several of the most common adjectives in Tocharian trace their origins back to such PIE forms. We take as exemplar the adjective A kāsu B kartse 'good'. Below we list the paradigm.
|A Masculine||A Feminine||B Masculine||B Feminine|
The exact PIE root from which this adjective derives remains unclear. In forms other than the masculine nominative singular, we see that the root likely has the basic shape *KṛH-, to which was suffixed *-ont-. Here *K denotes any velar consonant and *H any laryngeal (thereby giving the root a *CVC shape expected in PIE). The Tocharian A masculine nominative singular points to a different root shape, perhaps *ḱHs-, while the equivalent form in Tocharian B actually points to the feminine nominative singular. Below we list some of the forms for which derivations are reasonably clear, given these root shapes. This illustrates the general features of the development of *nt-stem adjectives from Proto-Indo-European into Tocharian.
Note in the masculine genitive singular form B krencepi we have palatalized -nc- following the pattern given by orotstse, G Sg. oroccepi. In general the genitive forms show the inherited base followed by the reformed Tocharian endings. We find the extension of palatalization to the accusative also in the masculine plural form A krañcäs. The feminine singular oblique form in Tocharian A shows the addition of the typical oblique ending A -ṃ to the form of the nominative singular. The nominative singular forms themselves hold special interest. In Tocharian A, the masculine form evidently derives from a different root, while the feminine form is lacking. In Tocharian B, the feminine nominative singular follows regularly, and it appears that this in fact provides the base for a reformed masculine form.
The adjective A puk B po 'each, all, whole' forms another prevalent member of this adjective class. The paradigm follows.
|A Masculine||A Feminine||B Masculine||B Feminine|
Though the above paradigm lists a variety of forms for different cases, numbers, and genders, the Tocharian languages in fact often employ A puk B po as an indeclinable adjective. That is, the forms A puk and B po may modify substantives of any gender, number, or case. In Tocharian B texts alternately employ the declined forms listed in the paradigm above in the plural for both genders; in Tocharian A they alternately employ declined forms in both singular and plural.
When modifying a substantive in the singular, A puk B po can have the sense of 'each, every', as in A puk praṣtaṃ 'every time', or the sense 'all the, the whole', as in A puk kapśañi B po kektseñe 'the whole body'. In the plural the sense is generally 'all (the)': A puk wrasañ B po onolmi 'all (the) living beings'. In prose, this adjective always precedes the substantive which it modifies.
The historical evolution parallels that described for A kāsu B kartse. Take the following for example.
|N Pl.||*péH₂-nt-es||*poñcä||A poñś||Gk. pántes|
|Obl. Sg.||*péH₂-nt-iH₂||*pontsā + n||A pontsāṃ||Gk. pãsa|
|(fem. nom. sg.)||B|
|N/Obl. Pl.||*péH₂-nt-H₂||*pontā||A pont||Gk. pánta|
|(neuter pl.)||B ponta|
In the Proto-Tocharian period, Tocharian speakers borrowed a word *pærnæ from Middle Iranian. The resultant words A paräṃ B perne mean 'glory' in the documented Tocharian languages. At the same time, an adjectival formation arises in Proto-Tocharian, *pærnæ-wænt-, which employs the PIE suffix *-wont- carrying the connotation 'possessing'. The result is the adjective A parno B perneu 'glorious, possessing glory' whose forms are listed below.
|A Masculine||A Feminine||B Masculine||B Feminine|
In Proto-Tocharian contraction evidently gives *pærnæ-wænt- > *pærnænt-, yielding the base for the masculine forms; the feminine forms apparently derive from *pærnæ-wäntsā-, which harkens back to the PIE zero-grade *-wṇt-iH₂. The origin of the masculine nominative singular forms still enjoys some scholarly debate.
We also find a related adjectival formation typified by the declension of A tālo B tallāu 'miserable'. We list the forms below.
|A Masculine||A Feminine||B Masculine||B Feminine|
This formation ultimately derives from adjectives built on roots ending in PToch *ā; in the above instance this derives from original PIE *H: *tḷ-n-H₂- from the root *telH₂-, giving PToch *tälnā- > *tällā-. To this Tocharian adds the same *-wont- suffix, but the root-final *-ā- produces a different vocalism in the resulting forms. Tocharian A further innovates by employing the -o- of the masculine nominative singular throughout the remaining paradigm, thereby removing the distinction between this paradigm and that exhibited by A parno. Tocharian B, however, preserves the original distinction.
Finally we should mention in this context the paradigm typified by AB ymassu 'conscious', evidently an adjective built upon the noun B ime 'conscience' by means of the suffix *-(w)ont-. The following table lists the forms.
|A Masculine||A Feminine||B Masculine||B Feminine|
The forms clearly resemble those of the preceding paradigms, with the notable exception of the masculine nominative singular. Suffice it to say that scholars have yet to come to a consensus on the origins of this form. Whatever its origin, Tocharian A has extended the u-vocalism throughout the remainder of the paradigm.
The lack of a thematic vowel intervening between the root (possibly extended by a suffix) and the personal endings characterizes the athematic presents. Recall that the thematic vowel *-e- develops into PToch *-(y)ä-, causing palatalization of the preceding consonant. Given that the thematic vowel is absent, athematic conjugations demonstrate a lack of palatalization of the stem-final consonant. This provides the single most distinctive feature between the thematic and athematic conjugations.
|3||*-Ø-ṇti||*-änt 'ä||*-äñc > -iñc|
Tocharian B did not remain content with solely athematic forms, and therefore imported some forms from the thematic conjugation. These forms only concern those with original thematic *-o-, specifically the active first and third person plural. Tocharian B also occasionally shows active first person singular ending -au imported from the thematic conjugation.
In both languages anaptyxis often inserts the reduced vowel *-ä- where the consonant of the ending comes into contact with the stem-final consonant. Tocharian B regularly promotes this vowel to -a- when the accent falls on this syllable. Note also in Tocharian A that the reduced vowel in the active third person plural ending is promoted to -i- in the palatal context: *-äñc > -iñc. One should not confuse this *-ä- resulting from anaptyxis with the *-ä- decended from the theme vowel *-e- in the thematic conjugations; where such ambiguity arises, the palatalization or lack thereof of the stem-final consonant should serve to distinguish whether the conjugation is thematic or athematic.
CLASS I comprises root athemtic verbs. The formation parallels CLASS II, which contains root thematic verbs, but omits the thematic vowel. In a sense, the root athematic verbs are the most basic PIE verbal formation. The discussion of the CLASS II presents shows that PIE thematic *-e- leads to palatalization of the root-final consonant. Given the absence of the theme vowel in CLASS I, the root-final consonant suffers no palatalization. Compare the following forms of the CLASS I verb pälk- 'shine' and the CLASS II verb āk- 'lead'.
We do find, however, the insertion of *-ä- to break up consonant clusters. This *-ä- results in B -a- when stressed.
In this root class, one expects to find alternation between full and zero grade of the root vowel. In fact one generally finds zero grade throughout the paradigm, as in PIE *bʰḷg- > *pälk- above (cf. Lat. fulgō < *bʰḷg-, but Gk. pʰlégō < *bʰleg-). Some CLASS I verbs nevertheless show a lengthened *ē-grade: B plyewäṃ < *pl'æw-n < PIE *plēw-Ø+nu (cf. Gk. pléō, Ved. plávate, both from full grade of the root).
The following table depicts the present CLASS I paradigm. The verb AB pälk- 'shine' illustrates the active forms, while A träṅk- 'say' and B kalāk- 'follow' (cf. A kälk-, which forms the non-present tenses of A i- 'go') illustrate the mediopassive forms.
|Pres. Ppl.||träṅkmāṃ||kolokmane (pälkamane)|
Note in particular the interplay of the reduced vowel ä and the accent and/or syllable structure in the above paradigm. In particular, we find A *pälkäṣ-aṃ > pälkṣ-aṃ by regular deletion of ä in an open syllable in Tocharian A. In B palkeṃ we see stress on the initial syllable changing ä to a; however this stress shifts to the second syllable in pälken-ne, and so the vowel ä remains unchanged in this form. Second-syllable stress causes a similar change in the mediopassive participle: PToch *pälk-mānæ > *pälkämānæ > B pälkamane, where stress changes *ä to a, but lack thereof changes *ā to a; by contrast second-syllable stress, or the position in an open syllable, causes deletion of *ä altogether in *kolok(ä)māne > B kolokmane.
For convenience we list here the paradigm of i- 'go', which belongs to CLASS I. Rather than the full grade of the root, e.g. *H₁ei-mi 'I go', as commonly found elsewhere in Indo-European, Tocharian here too employs the zero grade: PIE *H₁i-mi > PToch *yä-m > A yäm B yam.
|AB i- 'go'||A||B||PToch||PIE|
|3||yiñc (yäñc)||*yä-nt 'ä||*H₁i-Ø-nti|
Note that the Tocharian A forms follow regularly according to CLASS I; the Tocharian B paradigm however shows influence from the nasal presents (cf. Old Latin prod-īnunt, Hitt. i-ya-an-na-i). The Tocharian B paradigm nevertheless retains the remnants of the PIE athematic first person singular ending *-mi.
CLASS V comprises those verbs with present suffix -ā-. The origin of this suffix remains uncertain. One likely source of the suffix derives from verbs with root-final laryngeal: PToch *-ā- < PIE *-H-. However few CLASS V verbs of certain etymology exhibit this root structure. Another likely source derives from the suffix PIE *-eH₂-, as found in the Latin first conjugation: cub-ā-re, dom-ā-re, sec-ā-re; or perhaps *-eH₂-ye / o-, with contraction across the *-y-. Verbs with this suffix in PIE generally fall into one of two types: (1) denominative, or factitive; (2) deverbative. Remnants of both types remain in Tocharian.
Denominative verbs may derive from either nouns or adjectives. Several Indo-European languages preserve such constructions, e.g. Lat. novāre 'make new' (cf. Lat. novus 'new') and Gk. neãn 'to replow' (cf. Gk. néos 'new'). The Tocharian texts preserve examples of deverbatives from both nouns and adjectives: B kleṅke 'vehicle' > B klāṅkā- 'to ride, travel'; B swāre 'sweet' > B B!swārā-. The formation klāṅkā- however provides the preterite (Class I), not the present, stem of the verb klānk-; swārā- nevertheless provides the subjunctive (CLASS vi) stem of swār-, which as we will see is generally formally identical to present stem formation in Tocharian. In fact, given this formal identity between present and subjunctive, when an *ā-stem thus formed was coopted as a subjunctive, the present was often reformed according to CLASS IV or CLASS VI: B klautke 'manner, way' > klāutkā- (subjunctive) 'turn, become', and hence klautko- (present CLASS IV); B skeye 'effort' > skāyā- (subjunctive) 'strive', and hence B skāinā- (present CLASS VI).
Deverbative verbs of course are further extensions of roots which are originally verbal themselves. For example, B kwa- 'to call' < PIE *ǵʰu(H)-eH₂-, cf. OCS hŭvati 'to call'.
As the last examples illustrates, Tocharian commonly added the *-ā- suffix to roots with zero grade. Consider also B pälwā- 'bemoan' < PIE *bʰḷw-eH₂-. However Tocharian frequently builds verbs of this class with roots in full grade: B ānā-sk- 'inhale, breathe' < PIE *H₂enH₁-eH₂-(sḱe / o)-
The following table depicts the present CLASS V paradigm. The prime example of CLASS V formation is the verb AB läk- 'see'. In Tocharian A, läk- forms a CLASS V present; in Tocharian B läk- forms a CLASS v subjunctive. The CLASS v subjunctive of B läk- nevertheless follows the same morphological paradigm as any other CLASS V present in Tocharian B; läk- therefore provides an excellent point of formal comparison between the two languages. The following chart also provides the actual Tocharian B present of läk-, which happens to be of CLASS IX.
|Present V||A||B (Subjunctive V)||B (Present IX)|
|Grnd. I||lkāl||lkālle (I/II)||lkaṣṣälle|
Note the shift of accent and concomitant shift in vowel quality in the Tocharian B forms with the addition of the pronominal suffix. The verb B śu- 'eat' provides an example of the Tocharian B present participle for CLASS V, since this form does not appear for B läk- in extant Tocharian texts (naturally so, since B läk- only forms a CLASS V subjunctive, not present).
CLASS VI comprises a number of the so-called nasal presents, a class of fundamental importance in both Indo-European generally and Tocharian specifically. This class, in Indo-European terms, is an infixing class; that is, rather than add a suffix to the end of the root, this class adds an infix within the root itself (i.e. somewhere after the initial consonant, and before the final consonant of the root). The infix in this instance is PIE *-ne-, which as an infix characterizes durative action in PIE; hence its association with the present tense in the daughter languages, and its absence in the past tenses. This still survives in English stand (present, with infix) vs. stood (past, no infix). The PIE nasal infix is subject to vowel gradation, yielding an alternation between full- and zero-grade forms: *-ne- vs. *-n-.
The Tocharian CLASS VI ultimately has its origin in the nasal infixation of PIE roots with final laryngeal, a class found among the seṭ roots of Sanskrit, e.g. pṛṇā́ti vs. pṛṇītá; and elsewhere as in Greek dámnāmi, dámnāto. The Tocharian forms show only the zero grade of the infix, which combined with the root-final laryngeal yields PIE *-n-H- > PToch *-nā- > AB -nā-. Consider the following examples.
|*tḷ-n-H₂-||*tälnā-||tallaṃ||Gk. tela-, talássai, étlā (dor.)|
|tlanatär||Lat. tollō < PIE *tḷ-n-H₂-|
|*(s)kd-n-H₂-||*kätnā-||knāṣ||katnaṃ||Gk. skídnēmi < PIE *skd-ne-H₂-|
|*mus-n-H-||*musnā-||musnātär||musnātar||Skt. muṣṇā́ti < PIE *mus-ne-H-|
In the form A knāṣ we notice simplification of the cluster -tn-: A knā- < *känā- < PToch *kätnā-. Such simplification in common in Tocharian, e.g. *karp-nā- > A kārnaṣ, as is assimilation: *tälnā- > *tällā- > B talla-.
The above Tocharian B form tlanatär also shows another general feature: when PToch inserts -ä- to break up consonant clusters resulting from the nasal, the Tocharian B accent changes this to -a- when it falls on this syllable, and the -ā- following the nasal is reduced to -a-. Hence, Tocharian B shows alternation between -nā- and -ana-.
Perhaps even more interestingly, Tocharian A (only) shows a tendency to truly infix, synchronically, the -n- in roots ending in -tk- and -sk-. This infix always directly precedes the -k-. The -nā- syllable is thus broken up, the -n- preceding the -k of the root, and the -ā- following the -k: A kātk- 'rise' > kāt-än-k-ā-ṣ; A wāsk- 'move' > wās-än-k-ā-tär. We see in these examples that Tocharian A also shows the insertion of -ä- to break up the resulting consonant clusters. And as if this were not interesting enough, the verb AB kätk- 'cross', which obeys this rule in Tocharian A (*kät-än-k-ā- > A ktäṇkāṣ, ktäṇkeñc), shows both possibilities in Tocharian B: the verb forms a regular CLASS VI present with a -nā- suffix, B kätkanaṃ < *kätk-änā-, and a CLASS VII present, B kättaṇkäṃ, which derives from a true nasal infix in the final cluster -tk-.
As mentioned above, the original roots of this class contained a root-final laryngeal. Given that the nasal infix is a PIE durative marker, these roots tended to be originally aoristic or perfective. Hence they tended to form root aorists. In this sense, it comes as little surprise that verbs which form CLASS VI presents often form Class I (root aorists with root-final laryngeal) preterites; similarly, they tend to form CLASS v subjunctives.
Being ultimately roots with final laryngeal (though with a preceding nasal), the conjugation pattern of CLASS VI presents generally follows that of CLASS V. The verb AB kärs- 'know' serves to illustrate the paradigm.
|3||kärsnāṣ||kärsanaṃ (karsnaṃ, kärsnān-ne)|
|1 Pl.||*kärsnāmas||kärsanam (kärsnāmo)|
|3||kärsneñc||kärsanaṃ (karsnaṃ, kärsnān-ne)|
|Grnd. I||kärsnāl||kärsanalle (kärsnālle)|
|Pres. Ppl.||*kärsnāmāṃ||kärsanamane (kärsnāmane)|
This particular verb shows no mediopassive forms in Tocharian A, but such forms do exist in other verbs: A wenaträ < we- 'lead, spring up'. Note also the interplay between vowel quality and accent in the Tocharian B forms.
CLASS VII also comprises nasal presents, but unlike CLASS VI, the verbs of this class do not go back to roots in final laryngeal. This class survives only in Tocharian B. As in CLASS VI, these verbs derive from the zero grade of the nasal infix: PIE *-n- > PT *-n-. Such verbs abound in Indo-European: for example, Lat. pingit 'paints' and Skt. piṃśáti 'carves' correspond to B piṅkäṃ 'writes', all ultimately deriving from the root PIE *peiǵ/ḱ- through nasal infixation. The appearance of -k- rather than -ś- shows that this formation is athematic in Tocharian, a somewhat problematic state of affairs, since forms such as Lat. pingit < *pi-n-g-e-ti show this formation generally follows a thematic conjugation elsewhere in Indo-European.
CLASS VII verbs in final -tk- generally show geminate -tt- accompanying infixation of the nasal. For example, kättaṅkäṃ mentioned above, from kätk- 'cross', and puttaṅkeṃ (3rd pl.), from putk- 'divide' < PIE *put-sḱe / o- (whence, perhaps, *put-eH₂- > Lat. putāre 'prune').
The verb putk- 'divide' serves to illustrate the paradigm of CLASS VII presents.
|1 Sg.||*puttaṅkemär (?)|
Note of course the use the thematic endings in the active 1 Sg., 1 Pl. and 3 Pl. as usually in the athematic conjugations in Tocharian B. However we also find the use of the thematic ending in the first and third person plural mediopassive, which generally does not occur in the other athematic present classes.
The genitive is, most broadly, the case of relationship. Such a specification, naturally, is quite broad; it so happens however that the various uses of the genitive in Tocharian and Indo-European languages in general do span quite a broad range. In that sense, it is rather difficult to provide a detailed list of the many specific uses, nor would it necessarily be particularly fruitful in an introduction to the language. In this section we highlight some of the major uses of the genitive in Tocharian, both in terms of the overlap of usage with that of the genitive in other IE languages, as well as with a view toward idiosyncratic uses within Tocharian that may be less representative of the case's use in other branches of IE.
In the majority of instances the Tocharian languages employ the genitive in functions equivalent to those encountered in other ancient Indo-European languages and their descendants. By and large these parallel the most salient uses of the preposition of in English, some of which were treated in Section 7 of Lesson 2. In particular usages such as Lincoln's assassination and Booth's assassination are classified, respectively, as objective and subjective genitives: Lincoln is the object (patient) of the action represented by the noun assassination, while Booth is the subject (agent) of the same action. Of course grammatically there is no distinction; one must distinguish the two uses based on context or prior knowledge (i.e. one must already know that Booth was the perpetrator, Lincoln the victim, in this particular assassination). Scholars classify usages such as a statue of gold as instances of the genitive of quality, where the dependent genitive of gold describes some quality (here the type of material) of the noun on which it depends. Tocharian commonly employs the genitive in these same functions.
In some uses of the genitive the Tocharian languages either depart from or further restrict uses found in other IE languages. We discuss below some of the more important uses in this regard.
The possessive genitive denotes, straightforwardly enough, possession: A ṣtāmis pältwā B stamantse piltāsa 'the leaves of the tree'; A ptāñkte kapśañi B pudñäktentse kektseñe 'the Buddha's body'. Interestingly, however, Tocharian often opts to employ an adjective where English, say, might prefer a noun in the genitive. Consider the following examples:
Similar constructions occur in other branches of Indo-European, such as Slavic, Italic and Hellenic: e.g. Homeric Odusḗion es dómon 'to the house of Odysseus'; cf. the discussion of derived adjectives in Section 13.2. Additionally, Tocharian often employs the verb 'to be' with a predicate noun in the genitive for periphrastic constructions meaning 'to have':
Below we list other major uses of the genitive in some way peculiar to Tocharian.
One noteworthy function of the Tocharian genitive is its use as a dative case, denoting the indirect object. This usage likely stems from the historical evolution of pronominal case forms, where a probable source for the genitive AB -i lies with the PIE dative case itself (cf. Section 17.4.1). Tocharian generally confines this usage to verbs of giving and speaking. For example:
In the above examples, the italicized phrases are in the genitive in Tocharian, but as the translation demonstrates, these denote the indirect object.
Absolute constructions form an important part of the syntactic constructions in many of the ancient IE languages. The term 'absolute' derives from the notion that such constructions are divorced or removed from the grammatical structure of the remaining constituents of the sentences in which they are found. They continue even in modern English with phrases like 'that said', e.g. 'That said, let us move on to the next order of business.' In such a sentence, the phrase that said is grammatically self-contained and bears no grammatical relation to the rest of the sentence: none of the words 'let us move on to the next order of business' either depends on or modifies in any way either that or said. In this sense we say that that said is absolute.
Such constructions, as in the example of that said, generally constitute a noun or pronoun (here the pronoun that) with an accompanying participle (here the past passive participle said). In inflected languages like Tocharian, as well as Latin, Greek, and others, all substantives and adjectives must be given a case; thus a phrase like that said must occur in some case, be it nominative, accusative, etc. Typically ancient IE languages do not opt for the nominative, since this would necessarily imply a connection to the rest of the sentence; what case is chosen however varies widely across IE. Latin employs the ablative, Greek the genitive or accusative, Sanskrit the genitive or locative, Gothic the dative, and so on. Tocharian for its part employs the genitive: in an absolute construction, the substantive and attendant participle occur in the genitive to isolate themselves from the grammatical constructs of the remainder of the sentence. Of course, one has to isolate such uses from other uses of the genitive such as those described above where the dependent noun just happens to be modified by a participle.
For better or worse, the Tocharian documents illustrate very few genitive absolute constructions. Take for example B mäkte lwasāntso auṣuwaṃts ṣesa lyuketrä yṣīye 'how the night shines (when) the animals gather together', where the absolute construction appears in italics.