Periphrastic perfect / resultative in Tocharian Текст научной статьи по специальности «Языкознание и литературоведение»

Ilja A. Serzant

University of Leipzig

PERIPHRASTIC PERFECT / RESULTATIVE IN TOCHARIAN

1. Introduction

Tocharian is known as the most eastern Indo-European language originally spoken in the northern part of Tarim Basin (province Xinjiang, China), for which a coherent grammatical description was only provided at the beginning of the twentieth century by two German indologists Emil Sieg and Wilhelm Siegling. Tocharian is divided into two major dialects or languages: Tocharian A (henceforth: TA) and Tocharian B (henceforth: TB) with the latter exhibiting further variational subdivision. These languages were spoken at least during the last centuries of the first millennium AD and have, since then, been extinct.

In this paper, I analyze the Tocharian perfect/resultative that is periphrastically formed by means of the auxiliary 'to be' (omitted under certain conditions) and the past participle (PP), henceforth the PP construction. As will be argued in the paper, this Tocharian PP construction encodes a grammatical category that can be referred to as perfect with both resultative and more general perfect meaning. Both resultatives and perfects are characterized by a complex temporal structure invoking two temporal layers: past (the event time) and present (the reference time).

Thomas (1957: 245) was the first to conclude that the PP construction primarily has the perfect meaning in Tocharian. The aim of this paper is to elaborate on his findings. Regarding the data underlying this study, I draw on Thomas' (1957: 244-306) work with a large body of examples and their interpretation. In addition to this collection, I have confined myself to Maitreyasamiti-Nataka fragments (TA) which retained larger text pieces (as edited in Ji et al. (1998)), in order to have the context available when determining the temporal-aspectual function of the PP. Additionally, I have used the extensive digital text collection CEToM. For statistical purposes, I have selected 97 in-

stances of the PP in total from Maitreyasamiti-Nataka excluding the following: repetition, lack of context and, hence, no possible interpretation, the passage regarding the 32 marks of the Great Man (namely, Buddha) as it is heavily influenced by the respective Buddhist Sanskrit or Middle Indic notions. In what follows, I refer to these 97 examples as the sample.

The paper is structured as follows. First, I provide a brief description of the morphological (section 2) and some syntactic properties (section 3) of the PP and the construction based thereon. Then, I discuss the meaning (section 4): purely resultative uses of the PP in Tocharian including lexicalized resultatives are treated in section 4.1; section 4.2, in turn, is devoted to the more general and more frequent meaning of the PP, the perfect. Section 5 examines the meaning of the PP headed by the imperfect auxiliary (5.1) and by the preterite auxiliary (5.2). Finally, section 6 summarizes the main results.

2. Morphological properties of the PP construction

Tocharian uses the complex resultative form strategy (according to the classification in Nedjalkov & Jaxontov (1988: 19)) to encode not only the resultative, but also, as will be demonstrated below, the perfect: the tense and mood are marked on the auxiliary to be, while resultativity is expressed by the PP of the lexical verb. The auxiliary is not obligatory and typically dropped in the present tense.

The past participle (traditionally participium preteriti) is formed from the verbal root by means of the weak grade of the root (if the root is capable of having vowel gradation / ablaut), the suffix (TA) -u / -o and (TB) -u / -au in the nominative case and, in most instances, by the reduplication of the initial consonant (with special rules for initial consonant clusters).

Out of a total of 71 occurrences of the PP in Maitreyasamiti-Nataka, as the head of the clause (i.e. excluding 24 instances of the NP-internal use and 2 ambiguous instances), PPs were not accompanied by the auxiliary 56 times. In 15 instances, there was an auxiliary. The following table provides the overview of the tense-aspect-mood distribution of the auxiliary (given the low total number, the figures must be considered preliminary):

Table 1: The tense-aspect-mood distribution of the auxiliary in the sample

Auxiliary Present Imperfect Preterit Future/ Subj unctive Optative

Overtly expressed 8 (9%) 3 (3%) 0 4 (4%) 0

Absent 56 (62%)

The Tocharian PP construction is, thus, not particularly special in the cross-linguistic perspective with regard to its coding strategy. It adheres to the cross-linguistically common pattern in which the auxiliary provides the meaning of the (present) state while the past participle (PP) refers to a dynamic situation in the past (cf. Bybee et al. 1994: 67-68).

The following table provides a brief overview of the morphological forms of the auxiliary found in texts, based on Krause & Thomas (1960: 196), Thomas (1957: 251f):

Table 2: The third singular form of the auxiliary based on the verb TA nas- / TB nes- 'to be' and TB mask- 'to be found/placed'

Present Imperfect Preterite Future/ Subjunctive Optative

TA nas, mäskatär ses täk tas täkis

TB (nesäm) ste/star-/stäre', mäsketär sat täka takam täkoy

In Tocharian B, different verbs denoting 'to be' are used interchangeably, and the third singular form nesam does not seem to occur in the PP construction at all (Itkin, p.c.).

3. Syntactic properties of the PP construction

Differently to many languages (e.g., most of the European languages), resultative / perfect PPs are not part of the passive voice system in Tocharian. Although PPs may have P-orientation in Tocharian, surfacing thereby syntactically quite similar to passives, they are rarely used as purely passive forms to some active ones in terms of a syntactic and information-structure device not altering temporal-as-

1 Cf. MSL 19, 160 (Thomas 1957: 251).

pectual semantics. Instead, various tense, aspect and mood forms of the Tocharian passive are formed inflectionally by means of a dedicated middle/passive inflection (the r-endings set). The PPs primarily have functions from the domain of aspectuality, which implicationally determines their syntactic orientation (see section 3.1).

At the same time, Tocharian PPs are not restricted regarding the syntactic positions they may occupy: they may be used attributively to an NP, be the head of an NP itself (and then inflected for case), can head independent clauses, and can be used in various kinds of subordinate clauses (complement clauses, adverbial clauses and, perhaps, relative clauses), cf. section 3.2.

3.1. Orientation

Syntactic orientation is the effect of the semantics or, more precisely, of the event structure of the PPs whose original function was resultative. Resultatives describe a state — a situation that generally involves only one core participant. They highlight the state of affairs after the action denoted by the lexical verb has already taken place, and focus on the participant affected by this preceding event. Even if the preceding event involved two participants, the presence of the participant other than the affected one is not necessary at this poststage. This is why resultatives typically have only one core participant in their event structure. This makes the recovery of the agent participant quite vague if not impossible. Thus, the resultant state denoted, e.g., by the PP (TB) neneku 'X is destroyed' is potentially compatible with the following three types of preceding events:

(a) X destroyed itself (on purpose) [intransitive agentive, corresponding to the middle-voice inflection of the finite verb]

(b) X vanished (involuntarily, due to X-external factors) [intransitive non-agentive, corresponding to the middle-voice inflection of the verb]

(c) Y destroyed X [transitive agentive, corresponding to the active-voice inflection of the verb]

In (a), X is both the agent and the affected participant, while the agent is Y in (c) and there is no agent implied in (b) at all. The event structure denoted by resultatives does not contain the information about the event structure of the preceding event. To give an example,

consider the verb tsalp- 'to liberate (from smth.)' (cf. Serzant 2014: 77):

(1) winasam-ci kasu-tsalpont praise.PRS. 1sg-2sg.obl well-liberate.PP.OBL.SG risakam2

wiseman.OBL.SG

'I praise you as a well-redeemed wise man.' [TA 230 a2]3

(2) brahmayu pacar sasalpu ni Brahmayu.NOM father.NOM.SG liberate.CAUS.PP.NOM.SG samsar-as

1SG.OBL samsara.SG-ABL

'[My] father Brahmayu has been liberated by me from Samsara.' [Thomas 1957: 263; TA 258 b5]

Now, both forms tsalpo and sasalpu do not differ from each other in the participial morphology: the only difference is that the first PP is formed out of the inagentive verb base (meaning 'to become free') while the second PP is formed from the respective causative verb base (involving palatalization of the initial ts- into s-) with the meaning 'to make someone become free'. In effect, the former PP presupposes a less controlled (at least not externally controlled), intransitive preceding event of 'becoming free' while the latter implies a preceding transitive event of 'making free'. Crucially, the very morphological form of the PP does not favor the first or the second meaning, only the underlying word stems which themselves contain diathesis-related morphology.

Many languages have either ergative-like alignment in the re-sultative domain, grouping P- and S-oriented resultatives together, or the accusative-like alignment with a distinction between A/S-oriented resultatives, on the one hand, and P-oriented resultatives, on the other

2 Note that, for the sake of clarity, I skip all the auxiliary philological signs like brackets for amendments, square brackets for an unclear reading, etc. in all examples.

3 Here and elsewhere, I indicate the source of the example by providing the fragment number in a simplified manner (without the provenience sigla for TB) preceded by the abbreviation for the dialect: TA (for Tocharian A) or TB (for Tocharian B) in square brackets.

(I adopt the macroroles A, S, P as in Comrie 1989; Lazard 2002). To give an example, in Baltic, e.g. in Latvian, the resultative/perfect participles align accusatively: iedzer-i-s (drink-A/S.RESULT-NOM.SG.M) 'drunken (e.g. a person)' and iemidz-i-s (fall.asleep-A/S.RESULT-nom.sg.m) 'fallen asleep' with the suffix -i- (and its allomorphs) vs. izdzer-t-s (drink-P.RESULT-NOM.SG.M) 'drunk up (e.g. a beer)' with the suffix -t-. In contrast, Tocharian exhibits the neutral alignment in this domain with no formal distinction between A-, S- or P-oriented uses of the PP. The Tocharian PPs are contextually oriented participles in terms of Haspelmath (1994: 154), which means that if the underlying lexical verb admits more than one argument, its PP can be dependent on and select any of these arguments for agreement. The only verbal information that is encoded by the PP's morphology is their aspectual function.

The head noun of the PP may be missing and the PP itself be nominalized and inflected for case but, at the same time, exhibit a verby, accusative government:

(3) klesas wawikuntap el

klesa.OBL.PL disappear.CAUS.PP.GEN.SG gift.OBL.SG

wawura

give.PST.CONV

'Having made a gift to someone who has driven away the Kle-sas4' [Ji et al. 1998: 185; TA YQ 1.20 1/1 b8]

There is an asymmetry in how orientation appears in different participles of Tocharian: while present participles typically show active (A/S) orientation (this is valid for all active present participles), the PP does not show any significantly frequent orientation. Consider the following table on the basis of the sample:

Table 3: Orientation distribution of PPs in the sample

hits in percentages

S 53 55 %

A 20 23 %

P 24 25 %

4 Klesas are parts of human psychology that always cause sufferings to human beings.

Note that the unproportionally high number for the S orientation does not prove anything on itself as it is just the effect of the fact that there were approximately as many intransitive (53) as transitive (20+24=44) verbs in the text sample. Thus, Tocharian adheres only partly to the cross-linguistically frequent asymmetries whereby present participles tend to have active orientation while past participles exhibit passive orientation, cf. Comrie (1981), Haspelmath (1994; 2008: 200). The PPs do not show any significant inclination towards P or A orientation.

3.2. Adverbial and subordinate clauses headed by bare PPs In addition to the NP-internal, attributive use (i.e. the corefer-ential NP and the PP form one constituent), the Tocharian PPs are also employed in various subordinate clauses, mostly without conjunctions.

Table 4: Syntactic distribution of the PP in the sample (total — 97 hits, 2 hits are unclear)

Main clause Subordinate clause NP-internal, attributive

22 (23%) 50 (53%) 23 (24%)

The presence of the copular auxiliary in subordinate clauses is highly infrequent (Thomas 1957: 273). Thus, among the 50 uses in subordinate clauses, there was not a single hit with an auxiliary expressed except for the following relative clause introduced by the relative pronoun kusne:

(4) nmuk sak pt wakn-a kusne

ninty six PRT way.OBL.SG-PERL 3SG.REL.NOM wast-as lantus nenc

house.SG-ABL leave.PP.NOM.PL be.PRS.3PL 'Those who left home in ninety six ways'

[Ji et al. 1998: 23; TA YQ 1.30 1/1 b3]

Often it is quite difficult to tell the following two syntactic constellations apart:

(i) same constituent: the PP is used attributively or the nominalized PP forms a constituent on its own, or, alternatively,

(ii) different constituents: the PP is the predicate of a subordinate clause; the agreeing NP and the PP are parts of different constituents (cf. Thomas 1957: 244)

Given that the language is dead, constituency tests cannot easily be carried out here. The surface structure in (i) and (ii) is quite similar — in both cases there is typically an overt or pro-dropped NP somewhere in the sentence that agrees with the PP in case, number and gender.

There are just few criteria that may be helpful to distinguish between these two: (a) word order and (b) case inflection. Regarding (a), Tocharian word order is not rigidly constrained and is sensitive to information-structure related considerations. However, on average, the default word order within the adjectival phrase is clearly adjective / participle — noun (cf. examples in Krause & Thomas 1964: 91-92) and not the other way around. Reversely, any deviations from this word order may be plausibly interpreted as an indication for the PP heading a subordinate clause (ii).

Moreover, as regards (b), the position of the inflectional affixes of the so-called secondary cases, that is, cases that are attached agglu-tinatively to the singular or plural form of the noun / adjective / participle / determiner in the oblique case, may be decisive. Crucially, in Tocharian, the affixes need not be attached to all members of the constituent but may be realized only once ("Gruppenflexion") and, in this case, on the constituent final noun (Krause & Thomas 1964: 91). Here, a PP that semantically belongs to a noun in a secondary case and agrees with it by its oblique case (which yields the morphological base for the secondary cases affixes) has to be analyzed as (i).

I will not concentrate on the same-constituent uses of the PP here, cf. (6) below. What follows aims to provide evidence for the use of the PP as the predicate heading a subordinate clause. In the following example from TA, the PP namtsus 'having become' together with weyem 'astonished' may potentially be either interpreted as attributive to manarkan 'the brahmin youth', namely, as 'the astonished brahmin youth', or as an adverbial subordinate clause encoding an anterior process lit. 'having become astonished again, they say':

(5) kaklyusuras manarkan wtakot weyem

hear.PST.C0NV youth.NOM.PL again astonished.INDECL namtsus trankinc

be.PP.NOM.PL say.PRS.3PL

'Having heard (that), the brahmin youths, astonished again, say...' [Ji et al. 1998: 121; TA YQ 1.11 1/2 a2]

The analysis of the PP as an attributive adjective is unlikely in view of the word order found here: the PP occupies the position next to the main verb and not next to the coreferential NP, which is the regular word order in Tocharian for adjectival use of the PP as illustrated by the following example:

(6) pissankis el wawu community. GEN. SG gift.NOM.SG give.PP.NOM. SG pissank-am kalko el community. SG-LOC go.PP.NOM.SG gift.NOM.SG tranktra

say.PRS.3SG.PASS

'A gift that is given to the Community is called "gift gone to the Community" (i.e. skt. samghalambana or Pali samghagata-).' (cf. Ji et al. 2001: 177, 180 fn. 10-12; YQ 1.41 1/1 b3)

Semantic considerations might also speak against the attributive interpretation of the PP in (5). While in (6) the restrictive meaning of the PP is found, the restrictive meaning *'the again astonished youth' is not felicitous in (5). It is, of course, theoretically possible that the PP was used in (5) non-restrictively. However, given that attributive PPs are mostly used restrictively and the PP's position next to the main verb, the subordinate interpretation is much more likely.

In a similar way to the adverbial subordination, it is difficult to entirely set apart control constructions such as accusativus-cum-participio from PPs used attributively. Consider the following example with a verb of perception:

(7) ptañkatt ats wawrunt Ikam Buddha-god.OBL.SG prt awake.PP.OBL.SG see.PRS.1SG 'I see the Buddha-god awakened indeed.'

[Ji et al. 1998: 57; TA YQ 1.3 1/1 b3]

Verbs of perception are typical candidates for taking subordinate complement clauses cross-linguistically. The meaning found here is not the one of a non-restrictive relative clause *'I see the Buddha-god who has awakened', but rather about the fact that the very state of being awakened is being observed: 'I see the Buddha-god and I see that he has awakened'. This interpretation is suggested by the context of (7) which is, in short, about the Brahmin Badhari who has just heard about Buddha's fate and awakening. The following example from a medical text in TB is comparable:

(8) po yentem kektsen-ne all.INDECL wind.OBL.PL body. OBL. SG-LOC stmauwa lk-ass-am

stand. pp . obl . PL see-PRS-3SG 'He sees all winds standing in the body.'

[Carling 2000: 345; TB 41 b6]

The resultative stmauwa 'standing' (as a result of having been vertically placed/put) is also part of the observation referred to by the matrix verb 'to see', i.e. lit. 'He sees the winds and he sees that they have stood up and are standing now'.

The following example contains a conjunction that provides unequivocal evidence for the subordinate-clause analysis of the PP papassos (here, the head of a conditional clause):

(9) anaisi kwri papassos walke klyentar carefully if care.PP.NOM.PL long stand.PRS.3PL.MID kokalyi

cart.NOM.PL

'If carefully preserved, carts hold long.'

[Thomas 1957: 271; TB 5 b2]

In other cases, PPs are used in the way similar to (restrictive) relative clauses as regards the word order and semantics (cf. [Cristofaro 2003: 195]). Thus, in the following example, the object NP of the main verb 'head' is provided additional information by the PPs karsont 'known' and aksimnnunt 'taught' with an overtly expressed agent (by the genitive case on the noun 'Buddha-god'), thereby se-mantically yielding a complete proposition:

(10) mantne yas manarkan mrac

as you.NOM.PL youth.NOM.PL head.OBL.SG

p-karsas ptannakte

IMPV-know.2SG Buddha-god.GEN.SG karsont aksimnnunt

know.PP.OBL.SG teach.PP.OBL.SG

'Therefore, you, oh brahmin youths, recognize that as the "top" which is known and taught by the Buddha-god.'

[Ji et al. 1998: 121; TA YQ 1.11 1/1 b2]

Even though no relative pronoun is used here, from the functional point of view, both PPs seem to have the restrictive meaning here 'that top that was known and taught by Buddha'.

When it comes to the aspectual function of the PP in subordination, it is notoriously difficult to tell whether PPs just denote anterior events that are completed by the time before the main event begins, or, alternatively, whether they denote both the completion and the resul-tative state lasting over the time period of the matrix event. Thus, there is no way to differentiate between the resultative meaning 'to stand' for kakatkus (TA) derived from katk- 'arise' and the anteriority meaning 'having arisen' with no post-state of standing:

(11) asan-as kakatkus lamtsam kenenc

throne.SG-ABL rise.PP.NOM.PL queen.OBL.SG call.PRS.3PL

'Having risen from the throne they call out to the queen.'

[Ji et al. 1998: 165; TA YQ 1.24 1/1 b8].

Having said this, the anteriority meaning is unlikely in certain cases, e.g. in (9), for pragmatic reasons: *'after having been carefully preserved, the carts hold long'.

3.2. Part of Speech of the PP

Haspelmath (1994: 152) defines participles as "... words that behave like adjectives with respect to morphology and external syntax, but are regularly derived from verbs." He furthermore lists additional features typical of participles such as verbal valence and being part of the verbal paradigm. Tocharian PPs fit this description in many respects: they represent an inflectional category morphologically derived from verbs only and have actional meaning; they behave as adjectives syntactically if used attributively and typically retain the verbal semantic valence (not necessarily in terms of argument realization). Furthermore, the PPs in Tocharian are inflected for case and can be nominalized. At the same time, a number of properties that clearly distinguish PPs from adjectives are found. First, the argument realization pattern need not be retained: it might be inverted into an ergative one, e.g. genitive-marking for the A argument and nominative-marking for the P-argument. Secondly, they are at least partly different from adjectives with respect to the external syntax since they can

themselves head clauses and, thus, pattern with finite verbs in heading clauses as well as with converbs in modifying the main predication.5

3.3. The use of the PP with the present indicative auxiliary While the presence of the auxiliary is necessary to signal non-indicative mood as well as the non-present tenses, the present indicative is the default interpretation and, therefore, the auxiliary can and is often dropped. In certain contexts, however, the auxiliary nevertheless also appears in the present indicative where it does not seem to contribute any additional semantic component to the meaning. One potential reason for having the present indicative auxiliary might be related to reference tracking on the inter-clausal level. The subject NP is typically dropped in Tocharian whereas the verb desinences are used as weak referential devices (in terms of Kibrik 2011) to provide the reference for discourse-activated referents and speech act participants. Therefore, in clauses with subject shift, there must be at least one referential device indicating the shifted subject referent: the subject pronoun or noun phrase, or, alternatively, a finite verb carrying the referential desinences. To give an example, consider the following sentence:

(12) ma kaswone kaklyusu na-st

neg virtue.OBL.SG hear.PP.NOM. SG be-PRS.2SG ma tunk nas=si metrakn-am

NEG love.NOM.SG be.PRS.3SG=2SG.0BL Metrak.OBL-LOC 'You have not heard of [his] virtue. You have no love for Metrak' [Ji et al. 1998: 40; TA YQ 1.17 1/1 b1]

The sentence consists of two coordinated clauses or two simple sentences with non-coreferential subjects: the subject of the first clause 'you' is not identical with the subject of the second clause '(your) love'. The omission of the present indicative auxiliary nast 'you are (sg)' might create ambiguity in interpreting the referent of the first clause, because the referent of the PP will automatically be construed then as co-referential with the referent of the second clause's subject, namely, tunk 'love'. Moreover, the present auxiliary

5 Tocharian has dedicated converbs - these are non-agreeing and indeclinable deverbal nouns derived from PPs in the ablative (rarely in the perlative) case. However, as has been shown, PPs themselves equally can head subordinate clauses.

indicates the syntactically independent status of the PP clause which otherwise might also have been ambiguous between the main and subordinate clause interpretation. Thus, in terms of truth conditions, it indicates an independent assertion.

4. Aspectual semantics of the PP construction

The PP construction of Tocharian — although referring to events in the past — is different from other past tenses such as the imperfect and preterite. There are two main objective arguments to claim this: (i) its combinability with time adverbials and (ii) its cooccurrence with other tenses.

Regarding (i), as far as I can tell, the PP construction is found with the time adverbials referring to the present time relative to the speech time such as TA tapark 'now', while I have not come across any example where the PP construction would have been used with such time adverbials as TA tmas 'then' (except for its second meaning 'when'). The PP construction is not used in a narrative discourse for enumeration of events that happened after one another. In these cases, the imperfect or, more frequently, the preterite is used. Moreover, concerning (ii), I have examined all examples of the PP construction headed by auxiliary in the present tense, mentioned in Thomas (1957: 246-255) with both the "active and passive use" of the PP construction regarding which tense follows the PP construction. Unfortunately, 16 examples thereof were not valid for two reasons: fragment lacunas and being at the end of a discourse chunk:

Table 4: Co-occurrence with other tenses

Non-past reference Past reference

Present Subjunctive/Future Optative Preterite

10 2 1 2

Total 13 2

Although the number of examples is not significant, one may observe that the PP construction clearly favors contexts with present tense reference. One of the two examples showing co-occurrence with the preterite (i.e. a seeming exception) does, in fact, also have present tense reference:

(13) waltsant-yo pukl-a katkont thousand.PL-INS year-PL.NOM pass.PP.NOM.PL na=m tapark swatsiyis AUX.PRS.3SG=2PL.0BL now.ADV eat.INF.GEN.SG nom was makk ats klyosamas: swatsi name.OBL.SG 1pl neg prt hear.PRET.1PL eat.INF palko ma opyac na=m see.PP.NOM.SG NEG memory.SG.ALL be.PRS.3SG=2PL0BL 'In [their] thousands the years have now passed on us, the word "food" we have not heard at all. We do not remember anyone who has seen food.' [TA 340 a6]6

The time adverbial tapark 'now' clearly indicates the present tense reference (relative to the speech time). Additionally, the third clause contains a full verb in the present tense (the same one is used as the auxiliary in the PP construction). Regarding the other example with PP co-occurring with a preterite [TB 17 a2], it seems that the presential meaning is also found there, but I do not have any non-in-terpretational evidence to corroborate this. In sum, there is evidence suggesting a present-tense reference of the PP construction.

I turn to a detailed discussion of the exact nature of this category. In what follows, I consider two typologically established categories that both refer to some past event concurrently exhibiting a presential meaning: the resultative (4.1) and the perfect (4.2). I argue that both meanings can be found with the PP construction but the perfect meaning clearly prevails, as has already been suggested in Thomas [1957: 245].

4.1. Resultative

Nedjalkov & Jaxontov (1988: 6) define resultatives as follows:

(14) "The term resultative is applied to those verb forms that express a state implying a previous event."

Thus, predicates such as to hang and to be hung denote the same state, but only the latter additionally provides information about how this state came about (Nedjalkov, Jaxontov 1988: 6), which is why to be hung is only felicitous in the contexts that are compatible with the preceding event they entail:

6 Adapted from Gerd Carling, CEToM, retrieved on 14.01.2015; (Thomas 1957: 246).

(15a) There hangs a picture on the wall

(15b) A picture is hung on the wall

(16a) There are apples hanging on branches

(16b) *Apples are hung on branches

In order for a verb to have a potential to form resultative, its lexical semantics must entail some inherent (cf. Depraetere 1995) or intrinsic (cf. Sasse 2002) endpoint constituting a change-of-state and yielding a new, resultant state at one of the event participants. From this, it follows that one of the verb's arguments must be "affectable" by the action, because if an argument is affected by the action, a new, resultant state with this argument can emerge (Haspelmath 1994: 160— 161). As Haspelmath [ibid.] notes, the affectable participant is typically the patient argument (e.g. the destroyed house), but not always, and such verbs as to wear or to learn also entail affectedness on the part of the agent. The resultant state of the agent triggers the "active" orientation of the resultative participle.

While the definition in (14) neatly differentiates resultatives from simple states, it is only a necessary but not a sufficient condition. Contrasting with perfects, the resultant state must be integrated into the lexical semantics of the verb denoting the previous/preceding event:

(17a) John has broken a stick

(17b) The stick is broken

Thus, if (17a) is true, (17b) must also hold true regardless of the context because the resultant state in (17b) is integrated lexically into the verb to break. The following definition of resultatives incorporates this insight and will be used in this paper:

(18) Resultatives denote a state that entails a preceding event whereby this state is the lexically integrated and context-independent consequence of the preceding event

This definition helps to delineate resultatives from perfects — the latter denote a situation whose results are context and situation dependent (see below, section 4.2). Moreover, it also provides the basis for distinguishing lexicalized resultatives. Lexicalized resultatives equally entail a resultant state and a preceding event. The relation between the

resultant state and the preceding event, however, is not semantically straightforward and requires additional semantic operations such as semantic extension or semantic narrowing. Crucially, even with lexi-calized resultatives, the resultant state is context-independent — which can be observed in the fact that they always have one and the same resultant state in different contexts. Thus, although the resultant state and the preceding event are not straightforwardly related with lexicalized resultatives logically and semantically, there is, nevertheless, one-to-one correlation integrated into the lexical semantics of the verb.

While resultatives may generally impose restrictions regarding tense (typically present tense), person (e.g. in Selkup (Samoyedic / Uralic, Siberia), third person only is possible), and polarity (e.g. Nivkh (isolate, Outer Manchuria & Sakhalin) has affirmative resulta-tives only) (Nedjalkov & Jaxontov 1988: 36-37), to the best of my knowledge, there are no such restrictions in Tocharian.

4.1.1. Regular resultative meaning. As has been noted above, the resultative meaning is more frequently found when it is used as part of an NP constituent, as an NP modifier. Clear instances of the resultative meaning in the predicative position are less frequent. The resultative meaning of the PP is found in the following examples:

(19) cincrone puk kalymentw-am charm.NOM.SG all direction.PL-LOC satko tni expand.PP.NOM.SG 2sg.poss

'Your [scil. Bhadra's] charm is expanded into all directions.'

[Thomas 1957: 257; TA 66 a2] 'expanded to all regions' > 'is in all regions'

(20) kuse no ost-mem ltu yassu who.NOM but house.SG-ABL leave.PP.M.NOM alms sawanca kwri su eat.PARTC.PRS.ACT if this.NOM.SG.M

krauptar waipeccenta

collect.PRS/suBJ.3sG property.PL.OBL

'The one who is a monk (lit. left from home) and eats alms

bread (lit. the eating-alms one), if he collects properties, (then

he) [Thomas 1957: 269; TB 33 a6-7])

'having become a monk' > 'is a monk'

The verb la-n-t- 'to leave' if combined with ostmem 'from house' has the meaning 'to become a monk', i.e. lit. 'to leave the house'. In turn, the resultative (TB) ostmem ltu (lit. house.SG-ABL leave.PP) just means 'being a monk'. Note the coordination with a present participle sawanca 'eating' that is also indicative of the present-time reference of the PP ltu.

It has been noted in the literature that resultatives are often restricted with regard to tense, primarily occurring in the present tense only. Thus, Nedjalkov & Jaxontov (1988: 36) provide the following hierarchy entailing the probability of a particular tense form of the resultatives: (21) present tense > past tense > future tense

Tocharian PPs, in turn, are flexible regarding tense. They are compatible with the future time reference (coded by the auxiliary in the subjunctive). Furthermore, past tense auxiliary in the imperfect tense (cf. subsection 5.1 below) and, somewhat infrequently, the preterite are also found (cf. subsection 5.2). The conditioning factor here seems to be that the stative meaning of the PP should match with aspectual meaning of the auxiliary. While there is not enough knowledge about aspectuality in Tocharian on the functional differences between the preterite and the imperfect7, the correlation of the imperfect with the imperfective viewpoint-aspect meaning and, respectively, of the preterite with the perfective interpretation is, grosso modo, not unjustified.8 Having said this, it is not unexpected that states are rather compatible with the imperfective viewpoint.

7 Exceptions are Thomas (1957), Itkin (2014).

8 I refer to the bi-dimensional approach to aspectuality (inter alia, (Bertinetto 1997, Smith 1997, Sasse 2002). There are two dimensions that are orthogonal to each other: the viewpoint aspect or aspect1 in (Sasse 2002) and actionality (also termed, e.g., as Aktionsart or aspect2 in (Sasse 2002). While both dimensions operate within the boundaries of an event, they differ from each other as to whether the boundaries are (a) inherent (cf. Depraete-re 1995 or intrinsic (cf. Sasse 2002) in case of actionality or aspect2 or (b) just temporally established in case of viewpoint aspect or aspect1 (Sasse 2002: 205-206). Actionality refers to the inherent organization of an event such as, e.g., referred to by Vendler's classes (activity, achievement, accomplishment or state (Vendler 1967) or by such compositional properties as telicity (telic vs. atelic) or dynamicity. In turn, the viewpoint aspect is more subject to the speaker's construal and that pertains to such domains as discourse organization or pragmatics and not to the very semantics of the event.

4.1.2. Development of the simple stative meaning. While both resultatives and simple statives such as to sit or to stand are states in terms of Vendler's classes (Vendler 1967), the main distinction is whether or not the state is conceived of as a result of some preceding event: simple statives do not entail a preceding event while resultatives do. Yet, in the course of semantic development, resultatives may lose this preceding-action-entailment and become in no way different from simple states except for the morphological form which, in turn, generally tends to be more conservative than semantics. This development is especially likely if the resultant state expresses conventional states like to lie, to stand, etc. and not states like to be destroyed. Moreover, the conventional states tend to expand on the respective inchoatives. Recall examples (15) and (16) with the English verb to hang and its resultative to be hung above. Both these predicates denote exactly the same state. The difference between these two is that the former additionally has the completive dynamic reading 'to place smth. into a hanging position' from which the resultative meaning to be hung is derived, while the latter entails the preceding action described by this reading which its stative counterpart to hang does not. In turn, the verbs like to destroy are less likely to develop the simple stative meaning, because destroyed does not encode a conventional state.

There are a number of instances in Tocharian attesting this development. Thus, the PP stmau (TB) 'standing' is derived from the verb kaly- / stam- 'to stand oneself, to place oneself and should be semantically a regular resultative with the meaning 'to be / stand somewhere after having been placed / having placed oneself there'. However, examples of TB stamau are found where the context rather excludes the second part of this meaning, namely, the preceding-ac-tion-entailment:

(22) wreme emalyasse kektsem-ne

thing.NOM.SG hot.OBL.SG body.OBL.SG-LOC stmaus avasth yainmu

stand.PP.OBL.SG state.OBL.SG achieve.PP.NOM.SG 'A thing that has achieved the hotness state [the temperature] that is found in the body.'

[Thomas, Krause 1964: 70 fn. 1; TB 197 b4]

The temperature found in the body is nothing that could have placed itself or that could have been placed there by someone else at some earlier point but rather, so to say, is always there. Thus, there is no implication of some preceding event of placing the temperature into the body. The next example is equally incompatible with some preceding event: a meaning 'they had placed themselves with tearful eyes and now they are standing with tearful eyes' is semantically unlikely:

(23) stmos akärnunt asä(nyo) stand.PP.NOM.PL.M tearful.OBL.PL eye.PL.lNS9

'They are/were standing with tearful eyes' [TA 44 b2]10

Compare the PP (TB) wewenu 'to be called' (cf. German heißen) in (24) and the finite middle form in (25), which both have identical meaning:

(24) krenc tne wewenos kuse ... good.NOM.PL here call.PP.NOM.PL who.NOM ' [Those] are called the good ones here who ...'

[Thomas 1957: 269; TB 15a6=17a7]

(25) somo-ainyai somo ytarye alone-passable.NOM.SG.F alone.NOM.SG.F way.NOM.SG.F ka we-s-tär

why call-PRS-3SG.MID

'The only traversable [way], why it is called the only way?'

[Adams 2013: 658; TB 29 b1]

Both are derived from the verb (TB) we- 'to speak, say, state, tell' (cf. Adams 2013: 658). The verb to say, tell is normally an activity verb and does not entail any inherent result (cf. 'to speak') unless it is used with an object, the latter use being quite infrequent. Equally, the meaning 'to be called' found in the finite middle-voice forms of the same verb (as in (25)) is rather a state and not a verb entailing some inherent result. One can, of course, argue that the PP wewenu is

9 According to Burlak, Itkin (in print), there is no paralis / dual in Tocharian A and the form must be regarded as plural.

10 Gerd Carling, Fanny Meunier and Michaël Peyrot, in: CEToM, retrieved Nov. 11, 2014.

semantically a resultative and is opposed to the finite verb westar 'is called (prs.3sg.mid)' in the same way as English to be hung is se-mantically opposed to the finite verb to hang. However, there is a difference: while the English predicates are indeed different with respect to the preceding-action entailment, logically there can be no such difference with the PP wewenu 'called' and the finite form westar 'is called'.

In the next example, the PP yutkos does not entail a preceding

event:

(26) tamyo yutkos lmos

therefore be.anxious.PP.NOM.SG.F sit.PP.NOM.SG.F

ses

be.lMPF.3SG

'Therefore, she was sitting there anxious.'

[Thomas 1957: 302; TA 111 a1]

The PP yutkos 'anxious' from the experiencer verb yutk- 'to be anxious' cannot be analyzed as semantically resultative because it does not imply any new situation that would have emerged as a result from some previous situation of 'being anxious' encoded by the finite verb. Differently to to sit down, which is an achievement verb, the latter verb is a stative one and, hence, should not be compatible with a resultative at all:

(27) natak tam surmas tu mar

lord this because-of 2sg.nom neg.prohib

yutk-a-tar kuyalte

worry-PRS-3SG.MlD because

'Oh lord, do not worry because of that, for ...' [TA 66 a6]n

There are also other experiencer verbs that do not distinguish between the meaning of the finite verb forms and the PP, both denoting the experiential state with no preceding-action-entailment. Consider the following examples with the verb spant- 'to trust, to have confidence in something':

11 Gerd Carling, Georges-Jean Pinault, Melanie Malzahn, in: CEToM, retrieved Nov. 20, 2014.

(28) larem patar-s somske ra dear.OBL.SG father.OBL.SG-ALL son.NOM.SG also sartai po su spant-e-tar encouragment.OBL.SG all this.SG.M trust-PRS-3SG.MlD 'Like the son for his dear father, [so] everyone has confidence in [your] encouragement.' [TB PK AS 17C b4]12

(29) sak wat tu ste wakitse luck.NOM.SG or here be.3SG supreme.NOM.SG larem lak-a-m kakaccos dear.OBL.SG see-suBJ-3SG rejoice.PP.OBL.SG.M spantos skwasont

trust. PP.OBL.SG.M happy.OBL.SG.M 'Or, this [is] supreme happiness, if one sees the dear one rejoicing, trusting, and happy.' [TB PK AS 17C a5-6]12

Note, furthermore, that the verb katk- 'to rejoice' is in no way semantically different from the PP kakaccos 'rejoicing' in the above example.

The meaning of the PP (TA) namtsu is 'having become' (cf. example (33) below). There are, however, instances in which the PP namtsu 'having become' has lost its entailment of some preceding change of state, although such contexts excluding a resultative interpretation are quite rare (for example, all attestations of namtsu in Maitreyasamiti-Nataka are compatible with the resultative meaning):

(30) cami alasuneyis nu tsrassune this.GEN.SG inertia.GEN.SG PRT energy.NOM.SG pratipaks namtsu

hindrance.NOM.SG be.PP.NOM.SG 'For the inertia, energy is a hindrance'

(cf. Sieg 1944: 5; TA 2 a6)

The stative non-resultative translation 'is', i.e. German 'ist', is given in Sieg (1944: 5), and it is motivated by the context which does not presuppose that things have ever been different before: the state-

12 Georges-Jean Pinault (in collaboration with Melanie Malzahn), in: CEToM, retrieved: Nov. 28, 2014

ment in (30) is a rather generic one and is not to be interpreted as the result of some preceding event. The following example is fully aligned:

(31) wast-lma-lune-yis nakcy arkisosi-s house-sit-NMLZ-GEN.SG divine.OBL.SG world-GEN.SG skam tsmar namtsu amok

and root.NOM.SG be.PP.NOM.SG virtuosity.NOM.SG 'For the household and for the divine world, the virtuosity is the root.' (cf. Sieg 1944: 5; TA 3 a1)

Moreover, the PP namtsu 'having become' from (TA) nas- 'to be, to exist' is used as the resultative/stative of the light verb with a number of indeclinable adjectives including the indeclinable present middle participle (TA) katkmam 'to be happy' from katk- 'to rejoice':

(32) trankas klyom metrak say.PRS.3SG noble.NOM.SG Metrak.N0M katkmam nasam

rej oice. partc. prs . mid. indecl be.PRS.1sG '(Badhari) says: Noble Metrak, I am fully of joy.'

[Ji et al. 1998: 23; TA YQ 1.30 1/1 b1]

The following indeclinable predicatives13 are found in the sample: weyem 'surprising, surprised', joyful', lyutar 'more, over', pakar 'evident', in combination with the PP namtsus with the respective meaning 'having become surprised', 'having become joyful', 'having outweighed', 'having become evident'. In many instances, the meaning is indeed resultative and the implication of some preceding change of state is likely as in the following example:

13 I refer to these words as predicatives and not conventionally as adjectives because they seem to mostly occur in predicative positions and not as noun attributives as far as I can tell. Moreover, they often head clauses themselves in the present indicative without a finite form of the copula. To encode the respective change of state they take the PP namtsu 'having become' of the verb nas- 'to be, to exist' for a change of state in the past and, interestingly, not the respective present form nas 'be.3sG.ACT' (which only has the copular meaning here), but rather the present form of the synonymous verb mask- (TA) maskatar 'be.3sG.MiD', otherwise also meaning 'to be, to exist' (cf. also Batke 1999: 42ff). Occasionally wsok 'happy' may appear as an indeclinable predicative, but it may also be inflected (Ilya Itkin, p.c.).

(33) sisk-i ypesumnts-amn kats-a klawra

lion-NOM.PL tigress-NOM.PL belly.SG-PERL fall.IMPF.3PL14

wsok namtsunt

happy.ADJ.INDECL be.PP.NOM.PL

'Lions and tigresses fell down on their bellies, full of joy.'

lit. 'having become full of joy'

[Ji et al. 1998: 102; TA YQ 1.5 1/2 a6],

Yet, the meaning of the respective finite forms of the verb nas-is only 'to be, to exist' and not 'to become' (cf. Thomas & Krause 1964: 109), for which the otherwise synonymous verb mask- 'to be, to become' can be used. The semantic relationship between the finite forms of the verb nas- 'to be' and its PP namtsu is, therefore, no longer transparent. Most probably, the original meaning of nas- was not only 'to be', but also 'to become' — a development very frequently found across Indo-European languages, cf. Sanskrit bhu- 'to be' or Baltic and Slavic *bu- 'to be' which originally stem from Proto-Indo-European verb *bkueh2- 'to grow, to become', cf. Ancient Greek phy-o-mai 'grow.PRS.1SG.MlD'. The meaning 'to become' was lost with TA nas- (and its TB correlate nes-) but was partly retained in the resultative meaning of its PP. The latter, however, as will be shown in the subsection immediately below, also gradually loses the resultative meaning in favor of the simple stative meaning 'to be', cf. (31).

Nedjalkov & Jaxontov (1988: 14) call this type of expressions quasi-resultatives. However, it seems to be misleading to refer to these verb forms as resultatives in terms of a semantically well-defined category since they violate the definition given in (18) above by not entailing any preceding event. Apart from the morphological form, they do not distinguish themselves from simple statives.

4.1.3. Lexicalized stative meaning

There are some PPs that have a stative meaning that is not directly derivable from the lexical meaning of the underlying verbs. There are two types of semantic shifts here: (i) lexicalization of the resultative meaning and (ii) loss of the preceding-event-entailment (as in 4.1.2 above). Consider the PP kaknu (TA) 'to be endowed with, to

14 Identified as imperfect in Malzahn (2010: 264) and Itkin (2014: 34) with more details.

have' formed from the verb kan- 'to arise' and governing the instrumental case:

(34) warssaltsune-yo kaknu

energy.SG-INS arise.PP.NOM.SG '[The being / person] endowed with energy'

[Thomas 1983: 19; TA 214 b1]

The Tocharian B counterpart kekenu 'endowed with (perlative)' has the same meaning, not presupposing a preceding action of 'providing with' (cf., e.g., the utterance in PK AS B5 a4).

The alleged meaning of the PP should have been and assumedly was 'arisen' which is suggested by the meaning of the verb. However, in some instances, as in the above, it does not entail any preceding event whatsoever. Moreover, the meaning 'to be endowed with' does not directly follow from the lexical meaning of a verb to arise. I assume the following semantic development:

Table 5: Development of PP TA kaknu / TB kekenu 'endowed with'

Resultative Lexicalized Resultative Lexicalized State

arisen > arisen with X > having acquired X > being endowed with X

In turn, PP rittau (TB) / rito (TA) — originally a resultative with the meaning 'connected with' (from the verb TA ritw-/TB ritt-'to connect') — patterns as a preposition, governing the comitative case in TB with the meaning 'related to'. It equally does not entail any preceding event, nor is it semantically, straightforwardly related to the event of connecting. Parallel to the PP kaknu above, there is an extension or bleaching of the meaning from a very specific form of being related, namely being connected, into any kind of being related. In the same manner as with kaknu, there is a subsequent loss of the preced-ing-event-entailment. Thus, both the PP rittau (TB) / rito (TA) and PP kekenu (TB) / kaknu (TA) involve both changes mentioned above: (i) lexicalization of the resultative meaning and (ii) loss of the preceding-action entailment.

4.2. Perfect

There are a few remnants of the old morphological perfects in Tocharian A inherited from Proto-Indo-European. These perfect forms were originally formed morphologically by means of the reduplica-

tion, the *o vs. *0 (zero grade) ablaut in the verbal root and dedicated perfect endings. The old aorist (perfective past) and the old perfect merged into the preterite in Tocharian having mostly aorist functions (very much in the same way as it happened in Latin). Nevertheless, there are some "synonymous" preterite forms for one and the same verb which I interpret as remnants of the old aorist and the old perfect, respectively (Serzant 2014: 93-97). Thus, the Tocharian A sa-sruka-t (REDUPL-kill-3SG.MID) '[he] has killed himself and sruk-sa-t (kill-pret(=aor)-3sg.mid) '[he] killed himself from the root sruk- 'to die/to kill'15 differ in that the former seems to still preserve the perfect meaning of current relevance while the latter, historically an aorist, still functions as an aorist, cf. the following example with the present time adverbial (unlikely to be used with an aorist):

(35) sam tapark kip surmasi sni

3sg.nom.m now shame.OBL.SG because rfl.poss

ancam sasrukat

self.OBL.SG kill.PERF.3SG.MID

'He now has committed suicide because of shame.'

(adapted from Sieg 1952: 12; TA 9a5)

There are very few verbs that unequivocally attest old perfects in Tocharian A (Serzant 2014: 93-97). Having said this, I now turn to the PP construction that replaces the old perfects functionally as has already been suggested in Thomas (1957: 245).

Resultatives frequently undergo the expansion of their meaning, developing thereby into perfects (inter alia, Kurylowicz 1962: 141-155; Serebrennikov 1974: 234-236; Nedjalkov & Jaxontov 1988: 41). The development of perfects from constructions containing a copula with a past participle of the lexical verb is widely attested cross-linguistically

15 It is quite difficult to determine the basic meaning of this verb: in the dedicated inagentive, middle-like paradigm (present III, subjunctive V, preterit I) it has the meaning 'to die' while in the paradigm of preterit III it has the meaning 'to kill'. I have argued in detail that the paradigm of present VIII, subjunctive I/II and preterit III constitute the basic paradigm in the Tocharian verbal system (Serzant 2014: 1-16). Hence, the meaning 'to kill' is likely to be basic here, while the meaning 'to die' is a derived meaning due to the inagentivizing morphology of the paradigm of present III, subjunctive V, preterit I.

(Bybee, Dahl 1989: 678). This process has been described by Lind-stedt (2000: 368) as "the generalization of meaning from 'current result' to 'current relevance'". Lindstedt (2000: 366) gives the following two definitional criteria of perfects:

(36a) "the relevance of a past situation from the present point of view

and

(36b) detachment from other past facts, i.e., non-narrativity."

Dahl & Hedin (2000: 392) elaborate on the notion of current relevance (first introduced in McCoard 1978) in (36), stating that — in contrast to resultatives (resultative perfects in their terms) — current relevance perfects imply repercussions that "are not directly derivable from the meaning of the verb" but rather from "the specific knowledge about the situation or about some conventions." To illustrate this point, they adduce the following example:

(37) The gong has sounded.

Dependent on the context, this utterance may have current relevance by indicating the beginning of the dinner time or by signaling that a round in a boxing match is over (Dahl & Hedin 2000: 392). That is to say, while resultatives denote a specific result that is rooted in the lexical semantics of the verb, current-relevance perfects, in contrast, encode a result that is derivable from the situation and general conventions associated with the event denoted by the verb. One of the most obvious consequences of this is that verbs that do not entail an inherent endpoint (atelic verbs) become compatible with the category (cf. Bybee et al. 1994: 68-69; Dahl & Hedin 2001).

Indeed, this state of affairs is found in Tocharian. As far as I can tell, Tocharian does not seem to impose any selectional restrictions onto which verbs may form the PP. Prototypically atelic verbs such as to go are frequently attested here in the PP form. The following table provides an overview over the proportional distribution of the Vendler's classes in the PP construction in Maitreyasamiti-Nataka. Note that the assignment of a particular verb to one or another class is preliminary, based only on the meaning of the verb and not on any aspectual tests:

Table 6: The distribution of the PPs across the actional classes in the sample

accomplishment gradual accomplishment achievement activity state

15 (15%) 19 (19%) 47 (49%) 10 (10%) 6 (6%)

Accomplishments: ritw- (caus.) 'to connect, to compose'; yam-'to make' (6)16; läm- 'to sit down' (2); wät- 'to put'; twank- 'to pack, to wrap'; nask- 'to spin'; wik- (caus.) 'to expel'; wäs- 'to wear, to put on' (2).

Gradual accomplishments: kur- 'get old' (3); krop- 'to gather, to collect' (5); näm- 'to bend' (3); yät- (caus.) 'to adorn' (3); katk- 'to rise'17 (4); päk- 'to ripen'.

Achievements: e- 'to give' (2); lä-n-t- 'leave' (2); war- 'to wake up'; ents- 'to grasp' (3); tärk- 'to release'; wäl- 'to die'; täm- 'to be born, to arise' (4); ar- 'to evoke, to cause, to produce'; käm- 'to come' (13); *weyem nas- 'to become amazed'18 (2); katkmam nas- 'to become joyful' (5); wsok nas- 'to become joyful'; pakär nas- 'to become evident'; nas- *'to become'; kän- 'to become' (4); kälp- 'to attain, to obtain' (4); tsälp- 'to liberate'.

Activity verbs: päl- 'to praise'; wnisk- 'to oppress, to torment'; nu- 'to roar'; aks- 'to declare, to teach' (2); art- 'to praise'; yärka yam- 'to honor'; we- 'to say'; kärs- (caus.) 'to instruct, tell';prutk- 'to crowd'.

States: kärs- 'to know'; klyos- 'to hear' (2); spänt- 'to trust'; lyutar nas- 'to outweigh, outdo'; kätk- 'to exceed'.19

I am thus inclined to categorize the PP construction of Tocharian primarily as that of a perfect with properties of a resultative to use

16 Number of occurrences if more than one is indicated in the brackets.

17 This verb has two meanings: gradual accomplishment 'to rise' and achievement 'to arise, to appear'.

18 The verb nas- means 'to be, to exist'. However, the PP namtsu presupposes the older meaning 'to become', see section 4.1.2 above.

19 More frequently, this verb has the meaning 'to cross something' (e.g. the ocean) and is rather an accomplishment. However, in the context it is used in here, it has the stative meaning: "a piece of clothing having exceeded human dresses in refinements" [Ji et al. 1998: 169; TA YQ 1.25 1/2 a3].

Nedjalkov & Jaxontov's terminology (1988: 43). On the whole, they list the following six operational criteria (C) that can be used to distinguish between a resultative and a perfect:

C1: "The after-effects of the action expressed by the perfect are non-specific, and they are not attributed to any particular participant of the situation" (Nedjalkov & Jaxontov 1988: 15);

C2: The perfect can be derived from any verb, unlike the resultative (only telic verbs), because the current-relevance meaning of the perfect does not require any inherent endpoint in the semantics of the verb, cf. John has sung (Nedjalkov & Jaxontov 1988: 15);

C3: The perfect of transitive verbs is usually transitive, while the resultative thereof is usually intransitive, because "the state resulting from a previous event is attributed only to one participant (either the underlying subject or object of this event)" (Nedjalkov & Jaxontov 1988: 15);

C4: Time duration adverbials such as for two hours, since morning, all day long measure the preceding event with perfects but rather the resultant state with resultatives. Moreover, only resultatives combine with adverbs of unlimited duration, such as still or as before, cf. *She has still gone vs. She is still gone (Nedjalkov, Jaxontov 1988: 15; Lindstedt 2001: 367).

C5: The same is true of momentary time adverbials such as at 7 o'clock in the morning: "[w]ith resultative, such an adverbial can only indicate a moment at which the state is in existence", whereas with perfects they indicate "the moment at which the action took place" (Nedjalkov & Jaxontov 1988: 16).

C6: In the same vein, resultatives of verbs of motion rather combine with essive locations (at which place?) whereas the respective perfects are used with lative expressions (towards which place?) (Nedjalkov & Jaxontov 1988: 16-17).

In what follows I examine the criteria C1-C6 against the Tocharian data to show that the PP (construction) has acquired the meaning of the perfect here.

I do not distinguish between the finite and non-finite uses of the PP with regard to aspectual properties because — as far as I can tell

— there are no differences relating to the perfect vs. resultative meaning between these two. Note that the function of anteriority is expressed by the means of a dedicated anteriority converb — that is related to perfect/resultative only derivationally: the anteriority converb is formed from the deverbal noun in the ablative (or rarely perlative) case, morphologically derived from the PP by means of the suffix -r.

4.2.1. C1: Current relevance. As has been mentioned above, the meaning of current result is extended to a more general concept of current relevance that also subsumes the former (Comrie 1976: 52; Lindstedt 2000: 368; Dahl & Hedin 2000). The consequence is that there is no requirement for the verb to have a lexically inherent boundary after which the same action cannot continue as such and verbs that are atelic become temporally delimited such as in the sentence Someone has been here (Lindstedt 2000: 368).

Thomas (1957: 245) was first to argue for the current relevance function of the PP construction in Tocharian. He speaks about aftereffects ("Nachwirkungen") subsequent to the action having been completed. Consider the following example in which the PP tatmu-s 'born' (TA) signals the current relevance and not just the resultative state of being born:

(38) Some wild animals speak to Metrak and ask him for a permission to follow him in order to understand the Samsara and to be freed from all woes. Metrak approves this and tells them to avoid evil deeds and explains:

omaskenam lyalypur-a yas cam

evil.OBL.SG deed.obl.sg-perl 2pl.nom.pl this.OBL.SG son-am tatmu-s

bad.form.of.existence.OBL.SG-LOC give.birth.PP-NOM.PL na-s

be.PRS-2PL

'Because of your bad actions you have been born in this reincarnation class.' [Ji et al. 1998: 103; TA YQ 1.5 1/1 b2]

This sentence is not just about animals being in a particular reincarnation class (resultative meaning), but rather about the lasting after-effects of this fact, namely, the woes that animals have and want be released from at the reference time (current relevance meaning). Note that these repercussions are not part of the lexical meaning of the

verb tam- 'to arise, to be born' and can only be retrieved from the context. The meaning of tatmu- is thus the one of the current-relevance perfect, despite the fact that the verb tam- 'to arise, to be born' is telic and is, in principle, capable of yielding a resultative. The following example is similar:

(39) badhari trankas perak, ni se

Badhari.NOM.SG say.PRS.3SG credible 1SG.POSS son k^yalte kaklyusu ni nesinas

because hear.PP.NOM.SG 1SG.POSS early.OBL.PL knanmanancas kassis-as

know. prs . PARTC. obl .PL teacher. OBL. PL-ABL maskatar be.PRS.3SG.MID

'Badhari says: My son, it is to be believed, because I [have]

heard it from the clever wise men of the past.'

adapted from [Ji et al. 1998: 69; TA YQ 1.2 1/2 b2-3 + A 214 a7]

The meaning of kaklyusu is not resultative *'it is heard by me (to me)' or *'it is known (to me)'. There is no particular focus on the result at the P argument (the thing heard, namely, the fact that in a short lifetime a Buddha will appear in the world, cf. Ji et al. 1998: 69) here, but rather on the whole event including the source of the information, namely, the wise men of the past. This is all relevant because it provides the motivation for why 'it should be believed'. The reference time coincides with the speech time but not with the event time that precedes both. Moreover, note that the clause containing the PP construction has the subjunction k^yalte 'because'20 and hence provides an explanation for the content of the preceding clause. It has been shown that, for example, the English perfect is often used in clauses which provide "a causal explanation of state-of-affairs referred to in another clause" (Dahl & Hedin 2000: 393). Dahl & Hedin (ibid.)

20 This interpretation has been suggested independently by Itkin (p. c.), while Ji et al. [1998: 69] interpret kuyalte as a sentence containing just this word and translate it as 'Why?'. However, even with their interpretation, the clause with the PP must be interpreted as the answer to this question and would still conform to the typical use of perfect as the "causal explanation" suggested in Inoue (1979) and Dahl & Hedin (2000).

rely on Inoue (1979) who even claims "that a sentence in the Present Perfect conveys an explanatory sense."

As has been mentioned above, time adverbials that are preferably used here are TA tapark 'now' in contrast to, e.g., tmas 'then' (this.ABL.SG) that is employed to link subsequent events. Thus, among the 97 examples in the sample, I found 2 examples with the adverbial tapark 'now' and no examples with any other time adverbial. Consider the following example: The teacher Badhari, too old to go and see Buddha, sends his 16 disciples to Buddha and tells them they should become Buddha's disciples but is so desperate about the necessity to be separated from his disciples and particularly from Metrak that he figuratively says:

(40) nu tapark wtakota walu nasam prt now again die.PP.NOM.SG be.PRS.1SG 'But now for a second time I have died.'

[Ji et al. 1998: 83; TA YQ 1.8 1/2 a7]

The two adverbs 'now' and 'for the second time' scope over this exact change-of-state and not the result. Thus, the time adverbial 'now' refers to the change of state and not to a time at which the resultant state holds true (C5). The situation referred to is not about Badhari's being dead, but about the fact that it happened right now (due to the necessity to be separated). It is about the emotional consequences thereof. I take this as evidence for the perfect interpretation of this PP and not a resultative one * 'now I am dead for the second time'.

Another example comes from the verb pal- 'to praise', which is an activity verb and does not presuppose any natural inherent endpoint. Hence, no resultative reading is possible with this verb for semantic reasons:

(41) sne emtsal(u)ne pe sokyo paplu

without clinging also extremely praise.PP.NOM.SG

sastrantw-am

sastra.OBL.PL-LOC

'[He says :] ... Not clinging to property is also highly praised in the Sastras (i.e. the tenets, instructions).'

(cf. Ji et al. 1998: 45; TA YQ 1.16 1/2 a2)

Consider now the finite (present-tense) use of the same verb in the following example:

(42) we-s-enc-ai-mpa seme wantre

say-PRS-PARTC-OBL.SG-COM one.OBL.SG thing.OBL.SG mask-e-mar ma tu-sa katkau ma

be-PRS-1SG.MID NEG this-INS rejoice.PP.NOM.SG NEG tu pal=la-mar this praise=PRS-1SG.MlD

'I am together with someone who tells a thing [and] I am neither happy about it [scil. the thing] nor am I praising it.'

(cf. Thomas 1954: 728; TB 596 a3-4)

Except for the P-oriented diathesis of the PP paplu in (41) as opposed to the active use of (formally middle form) of the verb pal- in (42), there seems to be no difference. I assume that the difference here is exactly the one made between a stative current-relevance meaning of a perfect in (41) and the dynamic activity meaning in (42). The current relevance is seen in the fact that the sentence is used to corroborate the speaker's argument, very much in the sense of Dahl & Hedin's (2000: 393) "causal explanation", while the present tense in (42) indicates the very process of praising. Note that this verb is a deponent one which means that the passive meaning cannot be formally distinguished from the active one. Possibly, the additional reason for the use of the PP paplu in (41) instead of the synonymous finite form is the wish to demote the A argument, which is also natural given the uncertainty about the authorship of different Sastras.

4.2.2. C2: Atelic verbs. This property is notoriously difficult to assess precisely. The reason is that telicity is a language specific property and may vary across languages even with verbs that are otherwise synonymous. In order to have certainty, one has to carry out aspectual tests. Thus, the verb i- / kalk- 'to go' may be considered as a typically atelic verb, for example, in English. Nevertheless, it can or could form the resultative in English, cf. He is gone. For the resultative, a slightly different, metonymically altered meaning of the verb must be assumed, namely, to go in the sense of 'to leave' which, in turn, is telic. The same holds true for Tocharian as well. The PP kalko (TA) and yku (TB) derived from the verb 'to go' have rather a

telic meaning of 'go from/to', cf. examples (52), (53) and (54) as well as Itkin (2014: 41-44). Nevertheless, there are unequivocal atelic verbs attested in the PP construction, e.g., the atelic verb pal- 'to praise' discussed in details in (41) and (42) above. Moreover, there are no input restrictions on the formation of the PP in Tocharian, whatsoever. This suggests that atelic verbs were possible in the PP construction as well.

In what follows, I provide evidence separately for PPs used as predicates of main clauses (4.2.2.1) and subordinate clauses (4.2.2.2). I leave out evidence where the PP is used as an NP-internal modifier.

4.2.2.1. Subordinate clause. In the following example, there is no inherent result that would be entailed by the very semantics of the verb (TA)pas- 'to take care of, retain, guard'. Instead, a repetition of the delimited actions of preservation and taking-care-of has the aftereffect in that the carts are in the appropriate condition. This aftereffect is not part of the lexical meaning of the underlying verb thus violating the definition in (18):

(43) anaisi kwri papassos walke carefully if preserve.PP.NOM.PL long klyentar kokalyi stand.PRS.3PL.MiD cart.NOM.PL

'If carefully preserved, carts hold long.'

[Thomas 1957: 271; TB 5 b2]

The following example has already been discussed above as (26) and is repeated here for convenience:

(44) tamyo yutkos lmos therefore be.anxious.PP.NOM.SG.F sit.PP.NOM.SG.F

ses

be.lMPF.3SG

'Therefore, she was sitting there anxious.'

[Thomas 1957: 302; TA 111 a1]

The PP yutkos (TA) is derived from the verb yutk- 'to be anxious' which is an atelic verb.

Example (10) repeated here for convenience contains two PPs, both of which are derived from atelic verbs, namely, the stative verb kars- 'to know' and the activity verb aks- 'to teach':

(45) mantne yas manarkan mrac

as you.NOM.PL brahmin.youth.NOM.PL head.OBL.SG

p-karsas ptannakte karsont

IMPV-know.2SG Buddha-god.GEN.SG know.PP.OBL.SG aksimnnunt teach.PP.OBL.SG

'Therefore, you, oh brahmin youths, recognize that as the "top" which is known and taught by the Buddha-god.

[Ji et al. 1998: 121; TA YQ 1.11 1/1 b2]

Note that the PP aksimnnunt is intended here as an action performed for a certain period of time and then ceased, i.e. as a delimita-tive 'he taught for a period of time / he has been teaching'. An arbitrary temporal delimitation of the preceding event is only compatible with perfects and not with resultatives.

4.2.2.2. Main clause. In the next example, the verb kars- 'to know' is used in the causative form sasars- 'to inform' (lit. 'to let someone know smth.') which is rather an activity verb with no inherent endpoint:

(46) Haimavati trankas tapark sakkats klyom Haimavati.NOM say.PRS.3SG now surely noble.NOM.SG metrak suddhava=si-nas naktas-a Metrak.N0M Suddhavasa=ADJ-0BL.PL god. obl. pl-ins sasarsu

know.CAUS.PP.NOM.SG

'Haimavati says: Now, surely the noble Metrak was instructed by the Suddhavasa gods.'

[Ji et al. 1998: 23; TA YQ 1.30 1/2 a]

The translation provided by Ji et al. (1998: 23) above cannot be correct as regards the rendering of the PP, it also clashes with the time adverbial 'now'. There are two translations possible depending on the time of the instruction: (i) the adverbial 'now' has the preceding event in scope, or (ii) the adverbial 'now' has the resultant situation in scope and the instruction took place sometime in the past. The problem here is that the context preceding this sentence is missing. If one accepts the claim that the PP has perfect meaning, the second interpretation is then more likely: 'Now the situation is such that the noble Metrak is instructed by the S. gods (and knows all the things he has to know)'.

The context of the following example is more unequivocal in suggesting the interpretation of a perfect:

(47) ma kaswone kaklyusu nast ma

NEG virtue.OBL.SG hear.PP.NOM.SG be.PRS.2SG NEG tunk nas=si metrakn-am

love.NOM.SG be.PRS.3sG=2sG.OBL Metrak.SG-LOC 'You have [not] heard of his virtue. You have no love for Metrak.' [Ji et al. 1998: 40; TA YQ 1.17 1/1 b1]

The PP construction is used on the experiential or non-referential reading of perfect here. In its narrower definition, an experiential perfect presupposes an animate agent since it expresses the fact that "certain qualities or knowledge are attributable to the agent due to past experiences" (Bybee et al. 1994: 62). In a broader definition, it only means that "a given situation has held at least once during some time in the past leading up to the present" (Comrie 1976: 58). Indeed, the PP construction in (47) does fit the narrow definition, expressing the property of the A argument of not having heard about the virtues of Metrak until the moment of speech. Moreover, there are other indications typical of perfects and not of resultatives in (47), e.g., there is no particular result that can be attributed to one of the arguments of the verb klyos- (TA) 'to hear' (C1). The reference point is, in turn, the moment of speech which is confirmed by the coordination with the following present-tensed possessive clause.

4.2.3. C3: Transitive PPs. With regard to the orientation of the Tocharian PPs, it also provides evidence for the development from a pure resultative category into the perfect. Thus, there are a number of occurrences with A-oriented PPs with verbs that typically do not allow A-resultatives because they do not entail any affectedness on the part of the A. One of the PPs that is most frequently used in the A-orientation is the TA emtsu 'taken' with a canonical object. Note, however, that the A-orientation of the resultative of this particular verb is not unexpected on semantic and typological grounds: resultatives select the argument of the verb which is affected by the result of the action (Nedjalkov & Jaxontov 1988: 23). The subject of the transfer verbs such as to take or to receive can indeed be construed as affected because it acquires another participant into its possession which might be interpreted as affectedness. Therefore, the NOM-ACC (i.e. NOM-OBL) alignment of the PPs from

transfer verbs does not necessarily represent a violation of the definition of the resultatives (e.g. such as (18) above). Transitive resultatives formed out of these verbs, called possessive resultatives, frequently occur in grams used to encode resultatives only with no semantic affinity to perfects as, for example, in Nivkh (isolate, Outer Manchuria & Sakhalin) (Nedjalkov & Jaxontov 1988: 23). In turn, the existence of A-oriented resultatives out of causatives in Tocharian is indeed a violation of C3 and a strong indication for the development into a perfect:

(48) from Buddhastotra, a praise of Buddha:

puk klesa-si-nam wraske

all Klesa-ADJ-OBL.SG sickness.OBL.SG

wawiku

disappear. caus.pp.nom. sg

'[You] have removed the sickness of all Klesas.'

(cf. Thomas 1957: 280; TA 246 b2)

The A participant is hardly affected by the preceding event of removing, while it is the P participant (the sickness) that is affected and, hence, should have been the agreeing NP of the PP on the resul-tative reading of the latter. Notably, the transitive use of the PP makes it semantically and syntactically indistinguishable from the respective finite forms of this verb.

4.2.4. C4: I have no evidence corroborating or speaking against this property.

4.2.5. C6: Combination with essive vs. lative locations. Certain criteria are less applicable to the Tocharian data. Thus, C6 is not a strong predictor due to the more general essive-lative homonymy in Tocharian. In what follows, I examine the verb (TA) i- / kalk- 'to go'. The finite forms of this verb are, as expected, always used with the allative case with animate NPs, and mostly with inanimate NPs:

(49) orkmac kalkac darkness.SG.ALL go.suBJ.2SG

'You will go to the dark.' [Ji et al. 1998: 103; TA YQ 1.5 1/1 b3] Occasionally, the locative case with inanimate NPs is also found:

(50) wart-am y-mam lwa forest.SG-LOC go-PARTC.PRS.MID animal.OBL.PL tunkinna

love.PRET/iMPF.3SG

'When going to the forest, he was kind to the animals...'

[Ji et al. 1998: 103; TA YQ 1.5 1/1 b4]

Theoretically, the sentence may also mean 'while walking in(side) the forest' (essive), but cf. the following example where this interpretation is not a possible option:

(51) k^yal ma nas sol

why NEG 1SG.NOM life.OBL.SG raryuras ksaluney-am kalk-i-m

give.up.PST.CONV extinction.SG-LOC go-OPT-1sG 'Why shouldn't I, having given up life, go into nirvana?'

[Peyrot 2013: 242; TA YQ 1.36 1/2 b1]

Fully parallel to this, the respective PP kalko 'gone' (TA) is found with both the locative and the allative case. Consider the following example with the NP-internal use of the PP kalko in which the PP, despite its being a verb of movement, takes a locative case-marked location phrase:

(52) pissank-am kalko el community. SG-LOC go.PP.NOM.SG gift.NOM.SG 'gift gone into the Community'21

[Ji et al. 1998: 180.9; TA YQ 1.41 1/1 b1]

I have checked all the attestations in CEToM (around 20 in total). In most of the cases, indeed, the PP kalko 'gone' (TA) and PP yku 'gone' (TB) are used with the locative case (the essive strategy), which is in contrast to the use of the respective finite forms, because the latter have a clear preference for the allative case (the lative strategy), cf.:

(53) om no ceu kalal-ne ykuwes there then 3SG.OBL womb.SG-LOC go.PP.OBL.SG kau-tsi-sco speltkeyam-as-am

kill-INF-ALL effort.OBL.SG do-PRS-3SG

'There it [scil. the lie] makes efforts to kill him who entered the

womb (lit. 'gone in the womb').'

(cf. Thomas 1954: 755; TB 333 a4)

21 An idiomatic expression rendering Pali samghagata 'supporting the Community' (Ji et al. 1998: 180 fn. 9).

There was, however, one exception to this. Here, the PP yku 'gone' (TB) is combined with a location adverbial marked by the bare oblique case. The oblique-case-marked adverbials exclusively denote directions to a landmark and never a position within a landmark and represent just a more conservative instance of the lative strategy than the allative case-marking (Thomas 1983):

(54) tswaine ka yku past kremnt

just prt go.PP.NOM.SG away good.OBL.SG samanne-mem san oskai

monk.OBL.SG-ABL RFL.POSS house.OBL.SG 'Just after having gone from the good monkhood into his house.' [TB 44 b6]22

To summarize, there is indeed some preference for the essive strategy with PPs as opposed to the preference for the lative strategy with the respective finite verbs of movement. Even though this distinction is not a clear-cut rule, as a significant tendency it is another piece of evidence for my claim that PPs have stative aktionsart as opposed to the actional / dynamic aktionsart of the respective finite forms of the verbs of movement.

5. Pluperfect

In this section, the PP construction consisting of the PP itself and a past-tense auxiliary is examined: section 5.1 is devoted to the PP construction with an imperfective past auxiliary (traditionally and in the glossings: imperfect), while section 5.2 is about the perfective past auxiliary (traditionally and in the glossings: preterite). This alleged aspectual relationship between the preterite as the past perfective and imperfect as the past imperfective in Tocharian is a description advocated for in (Thomas 1957) in detail, and also accepted in (Batke 1999: 55, Malzahn 2010). It indeed accounts for most of the cases; see, however, (Itkin 2014) on the imperfect of the verb i- 'to go' which may be used as a perfective, that is, as a preterite and some other perfective-like uses of the imperfect.

Following Krause (1957: 296ff) I refer to this category as the pluperfect: it codes an event that took place before the reference time

22 Ed. by Hannes A. Fellner, CEToM, retrieved 15.01.2015; cf. also Sieg & Siegling (1949: 64-65).

which, in turn, lies in the past relative to the speech time. I have not found any secure attestations of an anteriority use of the pluperfect nor a remote past use. In many instances, it is just a perfect or resultative with the time reference in the past (relative to the speech time) and does not represent a grammaticalized independent category on its own.

5.1. Imperfective pluperfect Consider the following example:

(55) Upanande-m-sc ajivike sem

Upananda-OBL-ALL Ajivika.NOM come.PRT.3SG ratrem kampas ausu sai

red.OBL.SG coat.OBL.SG dress.PP.NOM.SG be.iMPF.3sG 'Ajivika came to Upananda. He had worn a red coat.'

[Thomas 1957: 297; TB 337 a4]

Pragmatically, the pure anteriority interpretation is less felicitous: (?) 'Having worn a red coat, he (then) came to Ajivika'. This sentence must be interpreted differently. The resultant state of 'having a red coat on' is a background state, and its reference time includes the reference time of the perfective event of 'coming' (coded by the preterite). In effect, this, of course, does imply that the very event time, i.e. the change-of-state of wearing, is anterior to the event of coming, but I claim that this is not the primary function of the imperfective pluperfect here. The primary function is to communicate that the resultant state of 'being dressed in a red coat' provides a background situation to the main action (coming), i.e. I interpret (55) as 'A. came to U. and he had a red coat on'. The function of the PP with the imperfect auxiliary is, thus, one of past resultative here. Given that this is a stative predicate, it is natural to not be temporally delimited and, therefore, extend across the whole time frame of the respective discourse chunk.

The following example is analogical. It is about the Sakya women being originally excluded from the preaching of the Buddha-god which, later in the narration, results in the inclusion of the women. The main events are described by perfective past events iteratively coded by the same verb in preterite kalpant 'they acquired for themselves'. In turn, the last clause with the PP and the imperfect auxiliary provides the background situation that holds at every event of preaching and acquiring:

(56) okat tmam (...) param kalpant

eight ten_thousands glory.OBL.SG acquire.PRET.3PL.MID

trit aksinna stwar tmam sakkin

third preach.PRET.3SG four ten.thousands Sakya.N0M.PL

param kalpant (. ) okak

glory.OBL.SG acquire.PRET.3PL.MID up_to

sakki-si kuli param

Sakya-ADJ.NOM.SG.F woman.NOM.SG.F glory.OBL.SG

ma kalpos ses

NEG acquire.PP.N0M.SG.F be.IMPF.3SG

'Eighty thousand [scil. of Sakya men] ... acquired the glory. He

preached for the third time, [and] forty thousand Sakya [men]

attained the glory. not a [single] Sakya woman [lit. 'not up to

one woman'] had attained the glory.'

[Ji et al. 1998: 151; TA YQ 1.22 1/2 a4-5]

The story continues by stating that the women complained about this and demanded to be taught the Law, as well. The PP construction has the following meaning: '[while] not a [single] woman had the glory to listen to Buddha's preaches [as the result of having acquired the glory]'. Note that the context does not allow the interpretation of the PP construction in terms of an event in a sequence of events of acquiring, like *'80000 (S. men) acquired ..., then 40000 (S. men) acquired ... then not a single Sakya woman acquired'. Rather, the meaning of the sentence is that for each instance of acquiring the glory achieved by the Sakya men there was no respective acquisition by the Sakya women.

The following example, this time from TB, illustrates the same effect of the PP with the auxiliary in the imperfect form. The imperfective pluperfect eros seyem 'they were in the state after having evoked the wrong idea', that is, 'they had the bad idea' also introduces the background scene in the past against which events like naksante 'they destroyed' (perfective past) and tesar 'they put' (perfective past) took place:

(57) wnolmi tallanco naksante being.NOM.PL miserable.NOM.PL destroy.PRET.3PL.MlD san anm eros pilko RFL.POSS self.OBL evoke.PP.SG.PL view.SG.OBL ankaim seyem tesar s false.lNDECL be.lMPF.3PL put.PRET.3PL and

naki krentamt-sa

blame.SG.OBL good.0BL.PL-iNS

'The suffering beings destroyed themselves — they had evoked the false idea — and put blame on the good ones.'

(cf. Thomas 1957: 297; TB 17 a6)

Notably, in the three examples (55)-(57), the clause with the PP follows the clause with the perfective past (preterite), even though the action described by the preterite (such as 'they destroyed' in (57)) is logically posterior to the event referred to by the PP clause. Since there is no conjunction (such as 'before'), this order would have been misleading for the interpretation. However, exactly because the PP clause does not denote the preceding event but rather does denote the resultant state which lasts during the whole discourse chunk, there is no mismatch between the sequence of the events and the sequence of their clauses in the discourse.

The next example is similar. Here, the resultative situation after 'having descended at my house', i.e. 'staying at my house' holds true in the past and is not bounded in any way — it stretches across the whole period of the narration. Note that the painter feigned committing suicide while staying at mechanic's house:

(58) The host [the mechanic] went to the king to inform him about the fact that the painter who stayed at the painter's place committed suicide:

natak alu ype-si pekant

Lord other.GEN.SG land-ADJ.NOM.SG painter.NOM.SG

yes ni wast-a kakarpu

go.lMPF.3SG 1SG.POSS house.SG-PERL descend.PP.NOM.SG

ses sam tapark kip

be.lMPF.3SG 3SG.NOM.M now shame.OBL.SG

surmasi sni ancam

because rfl.poss self.OBL.SG

sasrukat

kill.PERF.3SG.MID

'Oh Lord, a painter of another country came, he was staying (after having descended) in my house. He now has committed suicide because of shame.' (cf. Sieg 1952: 12; TA 9a4-5)

Example (12) repeated here for convenience is somewhat different:

(59) ma kaswone kaklyusu

neg virtue.OBL.SG hear.PP.NOM.SG

nast

be.PRS.2SG NEG

metrakn-am

Metrak.OBL-LOC

ma

tunk

nas=si

love.NOM.SG be.PRS.3SG=2SG.OBL tamyo tsam ma kakmu

set

therefore here neg come.PP.NOM.SG be.lMPF.2SG 'You have not heard of his virtue. You have no love for Metrak. Therefore, you did not come here.'

The first two clauses have the present time reference: the experiential perfect 'you have not heard', that is why 'you don't know', and the present state 'you don't have love' because both situations are still true at the moment of speech. The resultative-stative meaning 'you were not in the situation of ever having come here' holding true in the past provides a scene or background information for the present events here: 'you had not come here', i.e. 'you were not here'. The only difference to the preceding examples is that the reference time of the preceding clauses is not prior to the moment of speech (past tense) but contemporary to it: 'You don't know [after not having heard] the virtues, you don't love Metrak, therefore, you [never] were here.' As in this English rendering, the fact of 'not having been here' might also be true for the speech time (present). However, this is not due to the temporal-aspectual semantics of the imperfect PP construction. Rather, this is just an effect of a conventional implicature. Note that the meaning of the PP construction headed by the imperfect auxiliary is not committal to any temporal boundaries.

To conclude, the PP accompanied with the auxiliary in the imperfect tense typically denotes a resultative state or after-effects with the time reference in the past relative to the speech time. This resultative state or the after-effects are construed as not delimited temporally in the past, and, therefore, they typically embrace the reference time of the other actions in the discourse. The event leading to the result/aftereffects takes place prior to the time reference, hence the relation to the category of pluperfect. Due to the effect that the result/after-effects tend to encompass the whole time frame of the respective discourse chunk, the imperfect PP has a flavor of providing background information. Indeed, this is not unexpected, since the imperfect in general is often found to provide circumstantial information, cf. Pinault (2008: 569).

[Ji et al. 1998: 40, TA YQ 1.17 1/1 b1]

5.2. Perfective pluperfect

In this section I discuss the PP construction formed by means of the preterite forms of the same auxiliary. Generally, the Tocharian preterite encodes perfective past or perfects (Thomas 1957). Occurrences of the PP with the preterite auxiliary are, thus, somewhat unexpected given the clash between the durational and stative (and hence imperfective-like) actionality of the PP and the perfective, punctual aspect of the Tocharian preterite. Indeed, they are quite rare (cf. Thomas 1957: 287).23

While the construction with the PP with imperfect auxiliary typically provides information about a background state that holds true prior to the moment of speech and is not temporally delimited, the construction containing the preterite (i.e. past perfective) auxiliary, not unexpectedly, refers to temporally delimited results/after-effects from the preceding event. It provides foreground information.

Consider the following fragment which is about a painter visiting the mechanic at his house. There he sees a beautiful girl. The whole discourse chunk can be divided into three parts: first, (a) there is narration of events (all in past perfective, i.e. preterite), then, (b) there is a historical present to highlight the culmination (though the time reference is still in the past), finally, (c) the perfect with current relevance:

(60a) tmas pekant sla tunk yantarsinam somine tsara emtsuras tmak sam yamtar wekat pkant pkant kratswan sorkmi sutkmi klar-am somim ma skam tak tam palkoras pekant 'When the painter then, full of love [lit. with love], reached (past converb) for the hand of the artificial girl, she immediately broke apart (pret.), and her rags, ropes [and] pins fell apart (pret.), and there was (pret.) no longer a girl. When the painter saw (past converb) this,'

(60b) wiyo=ki lakey-as kakatku

frighten.PP.N0M.SG=as_if bed.SG-ABL rise.PP.NOM.SG anenci palkoras trankas: precisely see.coNV say.PRS.3SG

'as if distraught (PP), risen (PP) from his bed and after regarding (past converb) it well he says (pres.):'24

23 Thus, no examples of this have been found in the sample.

24 The English translations by Gerd Carling (CEToM, retrieved 01.12.2014) were adjusted to the analysis suggested here.

(60c) hai sokyo nu kakatwu taka

oh very PRT fool.PP.NOM.SG be.PRET.1SG

yamtracarem kassin-a

mechanic.OBL. SG teacher.OBL.SG-PERL

'Oh, I have been highly fooled by the foreman-mechanic!'25

[Thomas 1957: 288; TA 7 b1-2]

This is followed by some sentences in the present tense with generic reference. The sentence in (b) is a historical present used to highlight the culmination of the story. Hence, both PPs in (b) are formally present resultatives but with past time reference, exactly like the third singular 'says' in the same sentence. Both resultative PPs kakatku 'risen' and wiyo 'distraught' encompass all situations in (b) and (c) and are not different from the imperfective pluperfects in 5.1 above with regard to their tense and aspect properties. As for (c), the whole sentence is dynamic with the focus on the preceding event, meaning: 'an event of fooling had taken place'. Judging from the context, it was not intended to communicate the bare fact that the painter was in the state of a deluded person but rather to focus on the very event that led to this state. Moreover, this resultative state 'I was a fooled one' is delimited temporally, because the context refers to the particular happenings with the artificial girl and the after-effects thereof: 'I was in situation for a while in which I was fooled and suffered from the consequences.'

Just a few lines later in the same story, there is another PP with the auxiliary in the preterite and its interpretation is aligned: highlighting a particular event that took place prior to the speech time:

(61) sas wram nu yamtracarem

this.NOM.SG thing.NOM.SG PRT mechanic.OBL.SG kassin-a sni amok-am opassune

teacher.OBL.SG-PERL RFL.POSS art-LOC.SG skills.NOM.SG ns-ac lalaksu tak kyal ma

1SG-ALL see.CAUS.PP.NOM.SG be.PRET.3SG why NEG nas penu cami sni amok

1SG.NOM also 3SG.GEN RFL.POSS art.OBL.SG lkatsi ayim see.lNF give.OPT.1SG

25 English translations are by Gerd Carling, CEToM, retrieved 01.12.2014.

'This thing [scil. the mechanical girl] has been shown to me by the mechanic-teacher as virtuosity in his [scil. the mechanic's] art. Why shouldn't I also show him my own art?'

(cf. Thomas 1957: 288; TA 8 a1-2)

A resultative meaning of the PP lalaksu 'shown' can be securely excluded. Although the PP is P-oriented, the A argument is equally salient which is evidenced by the causative morphology and the overt agent phrase. Recall that resultatives introduce a state with just one participant (except for the possessive resultatives not found here). In turn, the interpretation as a perfect is suggested by the context. There is current relevance of the preceding event of showing with the reference time in the past, meaning something like 'I was exposed to this thing for some period of time in the past'. Similarly, as in the example above, the resultative state of 'being exposed to the mechanic's art' is terminated because the mechanical girl fell asunder immediately after this moment. The after-effects, namely the need for a response on the part of the painter, are still valid at the reference time.

The following example is also likely to be interpreted as a perfect with a current-relevance meaning as the reference time is in the past but simultaneous and not prior to the reference time of the preceding discourse:

(62) onkalmassi natak sokyo nu maski

elephant.GEN.PL lord.NOM.SG very prt hard yamlam wram yamast sni

make.GER thing.OBL.SG make.PRET.2SG RFL.POSS kaknu tak te cam tni

become.PP.NOM.SG be.PRET.3SG QST this.OBL.SG 2SG.P0SS saras puskas swal ankaras

vein.OBL.PL sinew.OBL.PL flesh.OBL.SG tusk.OBL.PL (raswaluney-am klop-yo) sarkinco rip_out.NMLZ. SG-LOC pain.SG-INS finally sinlune

satisfaction.NOM.SG

'O lord of elephants, indeed a thing extremely hard to do you did. Was there, by any chance, satisfaction for you (in) this (pain of ripping out) your veins, sinews, [your] flesh [and your] tusks?' [TA 67 a3-4]26

26 Amended by Sieg (1952: 14-15); the translation is adapted from Gerd Carling, CEToM, retrieved 01.12.2014.

Again, the PP construction headed by the preterite auxiliary has to be interpreted as a perfect, namely, the experiential perfect. The question of whether there has been satisfaction or not does not refer to some particular event in the past but rather generically to any kind of event that fits the description. Analogically to the previous examples, the reference time here is the same as of the discourse introduced by the preceding clause 'you did'. G. Carling coherently translates this example with the English present perfect (as quoted above). This has, though, one undesirable implication that the after-effects are still felt at the present. It seems that it is more likely that the question is about the after-effects of the satisfaction felt at the time introduced by the preceding clause '.you did'. A true pluperfect (e.g. in English) would imply that the satisfaction would be something resulting from an event before the action of 'you did' clause which is obviously also not intended here. One should, therefore, translate this example with 'Was there ... satisfaction for you .'. The past tense reference of the after-effects is visible in the answer in which the auxiliary is repeated in the preterite, that is, indicating past reference:

(63) trankas ma ontam taka-ni

say.PRS.3SG NEG at_all be.PRET.3SG-1SG.OBL 'He says: by no means was there [satisfaction] for me.'

[TA 67 a4]27

To summarize, the crucial difference from PPs accompanied by the auxiliary in the present, imperfect tense or just bare PPs heading a clause is that the preterite auxiliary additionally indicates that the after-effects or results from the preceding event hold only for a certain period of time at the reference time.

In the same way as for the imperfect PP construction, the preterite PP construction does not always fit the notion pluperfect. The PP with the imperfect auxiliary may be compatible with present-time reference (or contemporaneity) as well (cf. ex. (59) above) due to not committing to temporal boundaries. This is, however, excluded with "real" pluperfects, e.g., in English. In turn, the PP with the preterite auxiliary may denote situations which are the result of events that are

27 Translation is adapted from Gerd Carling, CEToM, retrieved 01.12.2014

contemporary and not prior to other past events in the discourse chunk (cf. ex. (62) above). This is again a deviation from pluperfects like English.

6. Conclusions

I have argued that the only meaning the Tocharian PPs have is their aspectuality-related function sensu lato. They are neither directly related to the grammatical category of passives, nor are they in any way marked for their orientation, exhibiting thereby neutral alignment. Moreover, there are no restrictions for the PPs whatsoever regarding which syntactic position they occupy: (i) the attributive slot within an NP, (ii) head of the NP, (iii) main verb (with or without auxiliary), (iv) adverbial clause heads (in terms of an agreeing converb), (v) relative clause head (with or without the auxiliary, and with or without the relative pronoun). In this respect, it is quite difficult to determine the part-of-speech association of the PPs with regard to their syntactic properties. They can pattern as nouns, adjectives, converbs and finite verbs. The three types (i)+(ii), (iii) and (iv) do not show any significant frequency differences.

There is evidence for the fact that PPs could head various types of subordinate clauses such as adverbial, complement and possibly relative clauses; subordination conjunctions are relatively rare here. In the subordinate function, PPs can have resultative, perfect and the taxis-related, anteriority meaning, while in the attributive position to an NP they tend to have the resultative meaning.

Regarding the aspectuality function of the PP, I have argued that present-tense PPs (i.e. with a present-tense auxiliary or with no auxiliary at all) combine with time adverbials referring to the moment of speech such as TA tapark 'now' and typically occur in the context where the neighboring clauses are headed by non-past-tense predicates. The preference in selecting the essive strategy with the PP but the lative strategy with respective finite forms of the verb 'to go' in Tocharian indicates that there is a change in actionality between these two: the former are rather states and hence the essive strategy, while the latter are dynamic actions naturally selecting for the lative strategy.

Furthermore, although PPs may have a resultative meaning, e.g. in the attributive position to an NP, on the whole, the category denoted by the PP has to rather be defined as perfect with properties of a resultative according to Nedjalkov & Jaxontov's typology (1988: 43), cf. also Thomas (1957: 245). Indicative for this are the following criteria: (i) there are no selectional input restrictions concerning which verbs may or may not form the PP construction: even though telic verbs are much more frequent, atelic verbs are also found here; (ii) in many instances, the respective contexts suggest the current-relevance interpretation of the PP construction; (iii) the PP construction has uses typical of perfects in other languages: as a discourse device it provides explanations to the next clauses (cf. Inoue 1979), it attests experiential or generic meaning; finally, other criteria are found, such as the occasional transitive alignment (rather atypical for resultatives outside the possessive resultatives).

The so-called "pluperfect" — if formed with the auxiliary in the imperfect tense — typically provides background information that holds true for the whole time frame of the narrowest discourse chunk. In turn, in the rare case where the auxiliary is in the preterite (i.e. perfective past), the overall meaning belongs to the foreground information crucial for the given discourse chunk. It denotes a resultative state or just any kind of repercussions holding for just a delimited period of time at the reference time which is prior to the speech time. This aligns very much with the findings about the difference between the imperfect and preterite use in Tocharian of Pinault (2008: 569) who claims that the imperfect itself codes typically circumstantial meanings.

The imperfect and preterite auxiliaries invoke considerably different temporal constellations: while the imperfect auxiliary indicates no temporal boundaries of the resultative state or the after-effects in the past, the PPs headed by the preterite auxiliary typically denote after-effects or results that are temporally delimited. The PP construction headed by an auxiliary is, thus, fully compositional in meaning and not grammaticalized into a grammatical category in its own right. Except for the inherent past time reference, the PP construction with an imperfect or preterite auxiliary deviates from typical pluperfects: it does not have any taxis-related conditioning (anteriority vs. posteriority); such meanings as remote past or the counterfactual meaning have not been found in Tocharian as yet.

Acknowledgments

I cordially thank the reviewers and the editors of the present volume for all their comments, help and patience: first of all Timur Maisak and Ilya Itkin without whom this paper would not have been possible. I also thank Vladimir Plungian for inviting me to this project. All shortcomings are mine.

Abbreviations

abl — ablative, act — active, adj — adjective, adv — adverb, all — allative, aux — auxiliary, caus — causative, com — comitative, conv — converb, f — feminine, fut — future, gen — genitive, impf — imperfect, impv — imperative, indecl — indeclinable, inf — infinitive, ins — instrumental, loc — locative, M — masculine, mid — middle, neg — negation, nmlz — nominalization, nom — nominative, NP — Nominal phrase, obl — oblique, opt — optative, partc — participle, pass — passive, perf — perfect, perl — perlative, PL — plural, poss — possessive, pret — preterite, prt — particle, pp — past participle, prs — present, pst — past, qst — question, rel — relative pronoun, result — resultative, rfl — reflexive, SG — singular, subj — subjunctive, TA — Tocharian A, TB — Tocharian B.

References

Adams 2013 — D. Q. Adams. More thoughts on Tocharian B prosody. //

Tocharian and Indo-European Studies 14, 2013. P. 3-30. Batke 1999 — C. Batke. Das Präsens und Imperfekt der Verben für "Sein" im Tocharischen, Tocharian and Indo-European Studies 8, 1999. P. 1-74.

Bertinetto 1997 — P. M. Bertinetto. Il domino tempo-aspecttuale: Demar-

cazioni, intersezioni, constrasti. Torino: Rosenberg and Sellier, 1997. Breu 1987 — W. Breu. Resultativität, Perfekt und die Gliederung der Aspektdimension. Slavistische Linguistik 1987, P. 42-74. Burlak, Itkin, in print — S. A. Burlak, I. B. Itkin. Formal'naja grammatika toxarskogo A jazyka: Fonologija. Morfonologija. Morfologija. Moskva. [A Formal Grammar of Tocharian A: Phonology. Morphonology. Morphology.], (in print). Bybee, Dahl 1989 — J. Bybee, Ö. Dahl. The creation of tense and aspect systems in the languages of the world, Studies in Language 13 (1), 1989. P. 51-103.

Bybee et al. 1994 —J. Bybee, R. Perkins, W. Pagliuca. The Evolution of Grammar: Tense, aspect and modality in the languages of the world. Chicago: University of Chicago Press, 1994.

Carling 2000 — G. Carling. Die Funktion der lokalen Kasus im Tochari-schen, Berlin / New York: de Gruyter, 2000.

CEToM — A Comprehensive Edition of Tocharian Manuscripts, URL: http://www.univie.ac.at (retrieved: Nov. 12, 2014)

Comrie 1967 —B. Comrie. Aspect. An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press, 1967.

Comrie 1981 — B. Comrie. Aspect and voice: some reflections on perfect and passive. // P. J. Tedeschi, A. Zaenen (eds.). Tense and Aspect. (Syntax & Semantics, 14) New York etc.: Academic Press, 1981. P. 65-78.

Comrie 1989 — B. Comrie. Language universals and linguistic typology: Syntax and morphology. 2nd ed. Oxford: Blackwell, 1989.

Cristofaro 2003 — S. Cristofaro. Subordination. Oxford: Oxford University Press, 2003.

Dahl & Hedin 2000 — Ö. Dahl, E. Hedin. Current relevance and event reference. // Ö. Dahl (ed.). Tense and aspect in the languages of Europe, Berlin: Mouton de Gruyter, 2000. P. 385-401.

Depraetere 1995 — I. Depraetere. On the necessity of distinguishing between (un)boundedness and (a)telicity, Linguistics and Philosophy 18(1), 1995. P. 1-19.

Haspelmath 1994 — M. Haspelmath. Passive participles across languages. // B. Fox, P. J. Hopper (eds.). Voice: Form and Function. Amsterdam: Benjamins, 1994. P. 151-177

Haspelmath 2008 — M. Haspelmath. Creating economical morphosyntactic patterns in language change. // J. Good (ed.). Language universals and language change. Oxford: Oxford University Press, 2008. P. 185-214.

Inoue 1979 — K. Inoue. An analysis of the English present perfect, Linguistics 17, 1979. P. 561-590.

Itkin 2014 — I. Itkin. Nekotorye nabljudenija nad formami imperfekta v Toxarskom A jazyke. // Voprosy Jazykoznanija 2014 (2). P. 27-45. [Some observations on Tocharian A forms of the imperfect]

Ji et al. 1998 — Xian-li Ji, W. Winter, G.-J. Pinault. Fragments of the Tocharian A Maitreyasamiti-Nataka of the Xinjiang Museum, China. Transliterated, translated and annotated by Ji Xianlin in collaboration with Werner Winter, Georges-Jean Pinault. Trends in Linguistics. Studies and Monographs 113. Berlin/New York: De Gruyter, 1998.

Kibrik 2011 — A. A. Kibrik. Reference in Discourse, [Oxford Studies in Typology and Linguistic Theory]. Oxford: OUP, 2011.

Krause & Thomas 1960 — W. Krause, W. Thomas. Tocharisches Elementarbuch, Band I. Grammatik. Heidelberg: Carl Winter Verlag, 1960.

Kurylowicz 1964 — J. Kurylowicz. The inflectional categories of Indo-European. Heidelberg: Carl Winter, 1964.

Lazard 2002 — G. Lazard. Transitivity revisited as an example of a more strict approach in typological research, Folia Linguistica 36(3-4), 2002. P. 141-190.

Lindstedt 2000 — J. Lindstedt. The perfect — aspectual, temporal and evidential. // Ö. Dahl (ed.). Tense and Aspect in the Languages of Europe. Berlin, New York: Mouton de Gruyter, 2000. P. 365-383.

Malzahn 2010 — M. Malzahn. The Tocharian Verbal System. [Brill's Studies in Indo-European Languages & Linguistics, 3]. Leiden / Boston: Brill, 2010.

McCoard 1978 — R. W. McCoard. The English perfect: Tense choice and pragmatic inferences. Amsterdam: North Holland, 1978.

Nedjalkov & Jaxontov 1988 — V. P. Nedjalkov, S. Je. Jaxontov. The Typology of Resultative Constructions. // V. P. Nedjalkov (ed.). Typology of Resultative Constructions. Amsterdam / Philadelphia: John Benjamins, 1988. P. 3-63.

Peyrot 2013 — M. Peyrot. The Tocharian subjunctive, A study in syntax and verbal stem formation. (Brill's Studies in Indo-European Languages & Linguistics 8.) Leiden / Boston: Brill, 2013.

Pinault 2008 — G.-J. Pinault. Chrestomathie tokharienne. Textes et grammaire. Leuven-Paris: Peeters, 2008.

Sasse 2002 — H.-J. Sasse. Recent activity in the theory of aspect: Accomplishments, achievements, or just non-progressive state? Linguistic Typology 6, 2002. P. 199-271.

Serebrennikov 1974 —B. A. Serebrennikov. Verojatnostnye obosnovanija v komparativistike. Moscow: Nauka, 1974.

Serzant 2014 — I A. Serzant. Das Kausativ im Tocharischen. [LINCOM Studies in Indo-European Linguistics 44] München: Lincom, 2014.

Sieg 1944 — E. Sieg. Übersetzungen aus dem Tocharischen I, in: Abhandlungen der preußischen Akademie der Wiss., Philosophisch-historische Klasse, Jg. 1943, Nr. 16.

Sieg 1952 — E. Sieg. Übersetzungen aus dem Tocharischen II. Aus dem Nachlass herausgegeben von Werner Thomas, // Abhandlungen der deutschen Akademie der Wiss. zu Berlin, Klasse für Sprachen, Literatur und Kunst, Jg. 1951, Nr. 1.

Sieg, Siegling 1949 — E. Sieg, W. Siegling. Tocharische Sprachreste. Sprache B. Heft 1. Die Udanalankara-Fragmente. Text, Übersetzung und Glossar, Göttingen: Vandenhoeck & Ruprecht, 1949.

Smith 1997 — C. S. Smith. The Parameter of Aspect. 2nd ed. Dordrecht: Kluwer, 1997.

Thomas 1954 — W. Thomas. Die Infinitive im Tocharischen. Asiatica. Festschrift Friedrich Weller. Zum 65. Geburtstag, gewidmet von seinen Freunden, Kollegen und Schülern, Leipzig: Harrassowitz, 1954. P. 701-764.

Thomas 1957 — W. Thomas. Der Gebrauch der Vergangenheitstempora im Tocharischen. Wiesbaden: Harrassowitz, 1957.

Thomas 1983 — W. Thomas. Der tocharische Obliquus im Sinne eines Akkusativs der Richtung. Akademie der Wissenschaften und der Literatur. Abhandlungen der Geistes- und Sozialwissenschaftlichen Klasse. Jahrgang 1983, Nr. 6. Mainz / Wiesbaden: Verlag der Akad. der Wiss., 1983.

Thomas, Krause 1964 — W. Thomas, W. Krause. Tocharisches Elementarbuch. Band II. Texte und Glossar, von Werner Thomas unter Mitwirkung von Wolfgang Krause. Heidelberg: Carl Winter, 1964.

Vendler 1967 — Z. Vendler. Linguistics in Philosophy. Ithaca, N.Y.: Cornell University Press, 1967.