ACTA LINGUISTICA LITHUANICA

LI (2004), 81 – 112

Grammatical Core Areas, Their Reconstruction, and Their Relevance to Historically Attested Stages of Baltic and Slavic

Björn Wiemer

Universität Konstanz

The paper offers a critical account of prominent theoretical approaches to prehistorical developments of basic morphosyntax in Baltic and Slavic: voice, alignment and aspectuality. For comparative reasons the relation between possessive BE- and HAVE-constructions is discussed as well. The methodological practice in reconstructing and explaining the diachronic evolution of ancient Baltic and Slavic is assessed critically, first of all because it has abided by ‘in vitro’-approaches without an account of language contact and typological evidence. It is argued that an adequate theory of grammatical change would further have to try to be more accurate with respect to the relative chronology of ‘layers’ of coding techniques and their possible intersections during long periods of diachronic development. This includes the possibility of the cooccurrence of grammatical coding. As for this issue, speakers of contact languages could trigger the functional repartition of such cooccurrent devices to such a degree that one of these devices would become predominant, fitting their own language’s form:function distribution.

0. INTRODUCTION

In a certain sense, this contribution may be understood as a critical tribute to two scholars who have made outstanding contributions overall to historical-comparative Baltic and Slavic studies: Jurij Sergeevič Stepanov and William R. Schmalstieg. A series of publications by each of them, of more or less influence, will serve here as an occasion primarily to show how much diachronically oriented Baltic studies suffer from the fact that chronological sequences of change in grammatically relevant structures and analogical formations are not examined concretely enough in their mutual entanglement in subsequent periods, and that too little consideration is also given to functional aspects. Added to this are frequent neglect of typologically plausible courses of development and of the language-contact factor, as well as the not infrequent failure to determine the grammatical versus lexical status of morphological distinctions. The areas discussed here are central to linguistic structure and are accordingly important for diachronic morphology and syntax, as well as for morphological and syntactic reconstruction. The topics are diathesis, Aktionsart, and alignment. The discussion of the theses

82 |

BJÖRN WIEMER

and approach of Stepanov (Section 1) and Schmalstieg (Section 2) is followed by reflections on an analogous case from the history of Slavic (Section 3). Taken together, these three broadly outlined “case studies” in historical-comparative argumentation against a typological background are intended to lead to methodological postulates, which are formulated in the conclusion (Section 4).

1. STEPANOV: CONNECTION BETWEEN ACTIONALITY AND DIATHESIS IN STEM FORMATION

Stepanov (1976; 1977; 1978a, b; 1989) undertook an impressive attempt to produce a synthetic outline of developments in the area of Aktionsart and diathesis, drawing on the historically attested facts and reconstructible prehistoric states of Baltic and Slavic in the field of verb morphology. Stepanov’s theses concern pre-stages or early stages of grammaticalization; however, they were formulated according to the “in vitro method” of historical-comparative linguistics, i.e. without reference to contact-linguistic factors. I attempt here a critical summary of his position on this subject. I will not discuss questions that delve further into the details of Indo-European phonology and morphology; nor will I call into question the correctness of Stepanov’s reconstruction in these matters. Rather, my concern is a critical examination of his methodology and of the interpretation of the facts thus presented.

Stepanov starts from facts familiar in Indo-European studies concerning stem changes in the verb systems of the older Indo-European languages, which affected, not least, prehistoric Baltic and Slavic, and from the assumption that early Indo-European syntax had no clear means of marking transitivity (as a sentence-semantic property). Moreover, there was no specialized morphology that would have indicated an expansion or reduction of argument structure on the verb, nor any morphology that would have enabled a more systematic distinction among different actional types (in the sense of aspect-sensitive classes). Rather, actional forms were lexeme-internally determined¹. Starting from these premises, which are also generally accepted in Indo-European studies, Stepanov seeks to uncover “layers” of etymologically shared changes to the verb stem in Slavic and Baltic, which, in the course of the dialectal differentiation of the Indo-European linguistic area, led in prehistoric Baltic and Slavic to more systematic morphological correlations used to mark transitivity, argument reduction or expansion, and actional types. The differentiation of Slavic and Baltic from one another lies in the different ways in which these correlations were functionally elaborated.

According to Stepanov, the affixation processes that were shared by Baltic and Slavic in type and (at least initially) also etymologically (i.e. in terms of form and substance) first formed functional clusters of actional, i.e. aspect-relevant, and diathesis-

₁In the literature, these have sometimes been called “lexical aspect.”

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 83

related features. Stepanov understands ‘diathesis’ as the “relation of the action to its subject” (1978a: 338). This corresponds to the ‘basic voice’ in Klaiman (1991), which also includes the middle voice of historical-comparative linguistics. For Stepanov, by contrast, actional features are verb-inherent features that determine whether or not there is a reference to a boundary (Russian ‘предел’). The issue is thus primarily a distinction between terminative and aterminative verb stems. In Stepanov’s model, stems (and any further affixation they may undergo) can refer to three phases of a situation, each of which focuses on different parts of a complex situation organized around reaching such a boundary. Phase (i) consists of a process that tends toward reaching such a boundary (“процесс, стремящийся к своему пределу”); phase (ii) corresponds to the boundary itself (“наступление предела”); phase (iii), finally, represents a state that necessarily follows from reaching that boundary. This state is itself aterminative, since its end is open (“процесс, ограниченный в начале, исходящий из ранее достигнутого предела и не ограниченный концом”); cf. Stepanov (1977: 145).

It is important to note that, at this point, we are dealing only with a conceptual distinction between actional stages grouped around a change of state. Although Stepanov repeatedly speaks of “aspectual oppositions” or even “aspectual pairs” (“видовые противопоставления / пары”), this initially amounts only to establishing aspect-sensitive classes of verb lexemes, not to a grammatical aspectual opposition (see below). Aspects would be distributed complementarily across these situation phases²—something that cannot be assumed in the language state reconstructed by Stepanov (nor in later periods).

Stem-changing affixes (infixes, suffixes, or reduplication) inherited from the proto-forms of Slavonic and Baltic led in many cases to a shift in stem meaning from situation phase (ii) to situation phase (i) or (iii). Standard examples include reduplication in the formation of the stative perfect and alternations of verbs of position; cf., for example, from the Indo-European root *sed- > OCS sědъl.sg.aor from sěs-ti ‘to sit down’ (= situation phase (ii)) as opposed to sěd-ě-ti ‘to sit’ (= situation phase (iii)); analogously, Lithuanian sėd-u.l.sg.prs from sės-ti ‘to sit down’ as opposed to sėd-ė-ti ‘to sit’ (1989: 221f.). Stepanov now searches within the system of the Indo-European verb for what he calls the “diachronic prototype” of the morphological marking of such a three-phase structure, and considers two “favorites” for this role: namely, (I) the Indo-European distinction between present, aorist, and perfect stems, and (II) the opposition between activa tantum, media tantum, and perfecta tantum (each with its own stem structure); cf. Stepanov (1977: 146ff.; 1978a: 341).

Functional clusters of actional and diathesis-related properties can, among other things, be reconstructed for Proto-Slavic, where ablaut formations were still productive

² Nota bene: Stepanov’s three phases correspond exactly to the three components from which the most complex class is built in an aspect theory that treats aspects as operators on lexeme-inherent actional content (the operands). In Breu’s ILA theory, this is the INCO class (cf. Breu 1998); it includes primarily so-called holistic verbs of the type наполНить / наполнять ‘to fill (up)’, заслонить / заслонять ‘to cover, shield’, спрятаться / прятаться ‘to hide oneself’. On these verbs specifically, see Padučeva / Rozina (1993).

84 |

BJÖRN WIEMER

must have existed, so that, among other things, ablaut formation with stem extension of a root such as *nes- (> Russian nes-ti, Lithuanian neš-ti ‘to carry’) to *nos-i-ti gave rise to both a causative and an iterative formation³. This type of stem or root change later led in Slavonic to the development of iteratives or “indeterminatives” (and thereby later formed one of the formal foundations of the imperfective aspect), while in (East) Baltic it also became a productive type for forming causatives and curatives (cf. Stepanov 1976: 420). Another example: the opposition between short and long root vowels (+ stem extension by means of a suffix) led to the productive contrast between verbs interpreted as punctual or perfective in some other way (short vowel) and verbs denoting a type of situation interpreted as frequentative, iterative, intensive, or progressive in some other way (long vowel). Cf., for example, OCS umrěti ⇔ umirati ‘to die’ (lengthening of the root vowel + stem extension), pogrěbati ⇒ pogreti ‘to bury; to inter’ (ě < *ē because of a < *ā in the stem extension), as well as, in combination with prefixation, vzbъznoti ⇔ bъděti ‘to be awake’ (ě < *ē). Analogous “pairs” are still found in modern Lithuanian; cf., for example, klupti ⇒ klūpoti ‘to stumble’ as opposed to kišti ‘to put (in)’ ⇒ kyšoti ‘to be stuck (somewhere), to remain lodged’ (in the latter case coupled with the [± transitive] opposition).

In contrast to Slavonic, in (East) Baltic this morphological opposition was also increasingly used to distinguish intransitive and transitive stems; cf. Lithuanian kisti ‘to change’ ⇒ keisti ‘to change something’, plisti ‘to spread’ ⇒ plėsti (as well as plėtoti) ‘to spread’ (where ė < *ē), busti ‘to wake up’ vs. ⇒ bausti originally ‘to rouse from sleep’ → ‘to compel to do something’ or ‘to maltreat, beat’, now meaning ‘to punish’ (cf., among others, Stepanov 1978a: 36, 342). Furthermore, in Baltic, nasal infixes and the st-suffix became among the most productive means of forming present stems, primarily of intransitive verbs (Stang 1943: 131ff.), whereas nasal infixes were lost in Slavonic⁴. These intransitives regularly contrast with transitive verbs having a different stem formation and, accordingly, a different tense-stem formation (in part from later “layers,” for example, with the suffix -(d)y- < *ī). For comparison, here is a juxtaposition of typical representatives of this opposition:

(1a) tamp-a.3.PRS vs. tap-o.3.PRT from tap-ti ‘to become’ (originally something like ‘to attach oneself, stick on’).

(1b) tap-o.3.PRS vs. tap-ė.3.PRT from tap-y-ti ‘to paint’.

³ Cf. similarly in Ancient Greek φέρω ‘I carry’ and its iterative derivative φορέω, compared with φοβέω ‘I frighten indm.’, which exhibits the same suffix and the same o-grade ablaut as a causative to the middle φέβομαι ‘I am afraid’ (Stepanov 1978a: 359).

⁴ Cf. meager reflexes in the form of nasal vowels in Polish, or their continuation in East Slavonic, as in Indo-European *sěd- > Proto-Slavic *sě-n-d- > OCS sěd-, or Polish siądzie ‘he/she sits down’, or Russian sjadet ‘same’ (vs. the Polish preterite siadł, Russian šel ‘he sat down’). Unlike nasal infixes used to form present stems, nasal suffixes in the form -n̥q- (< *nā?) played an extraordinarily important role in Slavonic in the emergence of the aspect system and actional functions as stem extensions not restricted to a particular tense (cf. Sadnik 1962).

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 85

(2a) link-st-a.3.PRS vs. link-o.3.PRT

to link-ti ‘incline oneself; show inclination’.

(2b) lenk-ia.3.PRS vs. lenk-ė.3.PRT

to lenk-ti ‘bend, curve’.

(3a) vyk-st-a.3.PRS vs. vyk-o.3.PRT

to vyk-ti ‘happen’.

(3b) vyk-d-o.3.PRS vs. vyk-d-ė.3.PRT

to vyk-dy-ti ‘carry out’.

According to Stepanov, these and a number of other facts allow the following conclusion to be drawn: in Baltic, the diathesis function predominated overall in the initial categorial clusters, whereas in Slavic the actional function gained the upper hand. Stepanov’s theory thus contains, in general terms, the following claims⁵:

(a) The types of affixation that became firmly established in the system of Slavic and Baltic verb morphology (and are now either fossilized there or remain visible to some degree, or even productive) represent an identical morphological technique.

(b) Most of these affixes even share a common etymological origin.

(c) Both these affixes and the types of affixation must date from a period for which intensive contacts between the antecedents of historically attested Slavic and Baltic may be posited (which, in turn, contributed substantially to the hypothesis of a “Baltic-Slavic linguistic community”). Stepanov himself refers here to “Proto-Balto-Slavic.”

(d) In the course of differentiation into Slavic and Baltic, clusters of actional and diathesis-related properties were gradually dissolved; that is, as Stepanov would probably put it, either the actional or the diathesis-related component in them was “suppressed,” while the other came to the fore functionally in the overall system of verb morphology during subsequent development and was continually elaborated or retained for longer. In Baltic, the diathesis-related changes are better preserved (i.e., are more archaic), whereas in Slavic the actional changes are better preserved, and to such an extent that they laid the foundation for the later aspect system.

⁵ Cf. his own words, to which I shall return: “The predominant oppositions in the verb lexicon of modern Lithuanian are paired oppositions between a simple verb of action or state and a causative or factitive verb with the same root. For example: bėgti ‘run’ / bėginti ‘make run (cause milk to flow)’, ėsti ‘eat (of animals)’ / ėdinti ‘feed (animals)’, siūti ‘sew’ / siūdinti ‘have something sewn, faire coudre’, kisti ‘change’ / keisti ‘change something’. Corresponding to this type of opposition is the productive grammatical opposition between ‘soft’ transitive conjugation and ‘hard’ intransitive conjugation; that is, the type with an ia-present and e-preterite versus the type with an a-present and o-preterite. Analogously, the predominant oppositions in modern Russian (cf. also other Slavic languages) are verb pairs of aspect, including pairs of ‘aktionsart’ (modes of action). (...) The basic relations in the category of transitivity/intransitivity and ‘definite’/‘indefinite’ aspect are entirely consistent. The primary opposing pairs represent one and the same action from two different points of view (...). This duality of perspectives on the same action (although the perspectives themselves differ) constitutes the commonality of the Slavic and Baltic systems and the basis for the unity of their prototype ‘aspect–diathesis’. (...) It is tempting to suppose that the Baltic oppositions of diatheses and the Slavic oppositions of aspects arise from one and the same source—from certain constant correlations between aspect and diathesis that existed in Proto-Balto-Slavic and constituted elements of a sufficiently stable ‘aspect–diathesis’ system there.” (Stepanov 1978a: 335–337; emphasis by the author himself, BW).

86 |

BJÖRN WIEMER

As for the formation of clusters of actional and diathesis-related features, Stepanov unfortunately does not specify whether the respective types of stem change simultaneously and closely affect actional and diathetic behavior, or whether, from one verb type to another (or from one lexical group to another), they sometimes produce only an actional modification and at other times only a change in diathetic structure. His sets of examples demonstrate both possibilities. They also partly contradict one another, since it is not always clear what kind of affinities between actional and diathesis-related features lead to cluster formation. Thus, for example, the st-suffix already mentioned above occurs not only in intransitive present stems; Stepanov (1976: 409) also regards it as the bearer of an ingressive function (“начинательность”), which should actually entail a perfectivizing effect. He makes this claim on the basis of modern Lithuanian⁶. However, he does not ask (a) whether an ingressive component might simply be the result of interaction with the lexical actional default (cf., for example, ding-st-a ‘disappears’) and (b) whether the present-day situation can readily be projected onto prehistorical (or at least earlier) Baltic. Another problem arises: how is the alleged perfectivizing effect to be reconciled with Stepanov’s view, expressed elsewhere, that in early phases present stems correlated primarily with progressivity⁷? A further concern is that -st- is thereby placed in opposition to stem extension by *ē (> OCS ē, Lithuanian ė), which very often brought about only an actional modification, namely “progressivization,” while remaining neutral with respect to the stem’s diathetic properties (i.e. it is attested with both transitive and intransitive stems).

Sowohl die *ē-Suffigierung als auch die Suffigierung durch {st} zählt Stang (1943: 118, 132) zu den relativ späten Verfahren. Wenn dem so ist und *ē nur zu einer aktionalen Modifikation des Stamms führte, läge es damit quer zu dem von Stepanov postulierten diachronen Trend; dieser geht ja seiner Ansicht nach im Baltischen zur Seite der Diathese (s.o.). Zu *ē-Erweiterungen treten dann auch intransitive Verben mit nasalem Präsensstamm in aktionale Opposition, indem das Verb mit *ē-Suffix Situationsphase (iii), das Verb mit Nasal Situationsphase (i) oder (ii) aus dem obigen Dreiphasen-Modell bezeichnet, die Diathese aber unangetastet lassen; dies liefe notabene darauf hinaus, daß *ē als Marker der Aterminativität zu einer terminativen Wurzel (Stamm) getreten wäre. Solche von Stepanov postulierten „Paare“ sind gar nicht einmal so selten; vgl. dazu die folgenden repräsentativen Gegenüberstellungen⁸ (mehr dazu in Schmid 1963: 60ff.):

(4a) bu-n-d-a.3.PRS vs. budo.3.PRT

(4b) budi / bud-ē-ja.3.PRS vs. bud-ē-jo.3.PRT

to bus-ti (< *bud-ti) ‘wake up’.

to bud-ē-ti ‘be awake’.

⁶ Sie findet sich auch schon bei Stang (1943: 131).

⁷ Vgl. etwa Stepanov (1989: 180, 186) selbst oder auch Kazlauskas (1968: 289).

⁸ These cases demonstrate not only a difference in phasality but also a lexical distinction with respect to telicity: verbs such as those in the (a)-sentences generally presuppose a telic course (= phase (i)) or an event (= phase (ii)), whereas those in the (b)-sentences do not, since they focus on Stepanov’s phase (iii).

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 87

(5a) švi-n-ta.3.PRS vs. švit-o.3.PRT to švis-ti ‘grow light, become bright/daybreak’.

(6a) kru-n-t-a.3.PRS vs. krut-o.3.PRT to krus-ti ‘begin to move’.

(5b) švit-i / švit-e-ja.3.PRS vs. švit-e-jo.3.PRT to švit-ē-ti ‘shine brightly, shimmer’.

(6b) krut-a.3.PRS vs. krut-e-jo.3.PRT to krut-ē-ti ‘move, shake, fidget’.

The next example is somewhat different in nature: on the one hand, the transitivization of stems in Baltic is linked to the PIE *ē-extension of stems, which is said to have yielded in Lithuanian a past tense with a palatal stem ending (in the 3rd person in -ė, as in pieš-ti ‘draw, paint’: pieš-ė ‘he/she drew, painted’); this contrasts with the past tense without stem palatalization (in the 3rd person in -o), as in bėg-ti, bėg-o ‘run’, ‘he/she ran’.⁹ The same stems are said to have been transitivized in this way in Slavic as well (cf. OCS pas-a-ti ‘write’ ⇒ piš-e-tъ ‘he/she writes’). In both cases, these are said to have been associated with the so-called “indeterminate aspect” (“неопределенный вид”) (Stepanov 1976: 409), and thus to have been potential precursors of imperfective aspect (on this question, see below). Yet elsewhere, transitivization is associated with perfectivization, understood as the imposition of an inherent boundary (i.e. telicization) (cf., among others, Stepanov 1977: 145ff. himself). This is also what one would expect, and it is consistently confirmed typologically (cf., for example, Bybee, Dahl 1989).

Another example: the morphological opposition between i- and a-stems in later Slavic aspect pairs such as Russian оставить / оставлять ‘leave’ is said to go back to an opposition between causative (i-stem) and iterative (or durative, a-stem). Stepanov then generalizes that in “Proto-Balto-Slavic” there was a cluster correlation of ‘intransitive stem + indeterminate aspect + iterative’ vs. ‘transitive stem + determinate aspect + causative’ (Stepanov 1977: 149). In doing so, however, he contradicts himself, as noted above in connection with the extension of PIE athematic stems. For in Baltic (and not only there; cf. Stepanov 1978a: 359f. himself), the i-extension was at the same time the basis for both causative and iterative formations (in the broad sense, i.e. including intensives and frequentatives); cf., for example, Lithuanian ly-dy-ti ‘pour (in a particular way)’ as a causative formation to ly-ti ‘pour, rain’ (see also example 3), as opposed to graib-y-ti ‘(hastily) reach out all around’ as an iterative of grieb-ti ‘seize’. Moreover, i-stem iteratives also exist from intransitive verbs; cf., for example, švisti ‘shine.INTR’ ⇒ švayti ‘fence’ (Old Lithuanian), bristi ‘wade’ ⇒ braidyti ‘1. keep emerging (e.g. the sun from behind clouds), 2. knead by wading’.

Such comparisons give rise to contradictions that stem from the fact that the functional clusters taken by Stepanov as diachronic starting points can resolve in different ways, and the features involved can then recombine differently in different periods. Thus, it can happen that causative and iterative/intensive formed a cluster in an earlier period, to

⁹ Cf. the quotation in footnote 5. This view had, incidentally, already been put forward by Stang (1943: 82f.). Schmidt (1966) provides a clear analysis of the distribution of ė/o in the Lithuanian past tense. Michelini, however, expresses reservations, concluding “that morphological units of different origin have merged in the suffix -ē- of the Baltic past tenses” and “in most cases (…) -ē- goes back to *i̯ā” (Michelini 1990: 845).

88 |

Björn Wiemer

but were separated from one another at a later time and could therefore enter into opposition. The contrast between the above-cited i- and a-stems in Slavic, however, is already a comparatively late development. By contrast, the fact that in Lithuanian i-stems were used for both causatives and iteratives can be interpreted in two ways: either both types of verbs arose at an earlier stage, when these functions were still “paired,” and then represented different “resolutions” of the cluster; or they arose in different periods, when causativity and iterativity no longer formed clusters. Although Stepanov partly places the morphological oppositions he discusses in detail into a rough chronological sequence (see the list in the appendix), the periods during which they were productive apparently overlapped, resulting in an often very opaque interweaving of different strata and the functions associated with them. An open problem, then, is the chronological sequence of the different types of morphological opposition and the duration of their overlap in real time¹⁰.

A related problem, not addressed by Stepanov, concerns analogical leveling, which blurs a clear picture of the consistent sequence of strata in stem changes. To avoid a veritable battle of examples, I will forgo a discussion of examples on this topic here. A careful reading of Stepanov’s works would make it possible to compile corresponding cases, and present-day Lithuanian offers a particularly variegated picture in the case of verbs with nasal infixation in the present tense and of formations derived from the *i-stems (i.e. -(d)in- and -(d)y-), since precisely these affixes entail both actional and diathesis-related modifications.

Some further remarks on Stepanov’s conceptual apparatus and the procedure based on it are in order. It should be emphasized here that Stepanov makes no claims about the degree of grammaticalization of the respective affixation processes (nor was this really his intention). One can, however, infer indirectly that their productivity and the regularity of their formation in the relevant functional domain (diathesis or actionality) could determine how the degree of grammaticalization would be classified, were one to assess it. More importantly, Stepanov groups aspect (вид) and lexical modifications such as Aktionsarten (способы действия), which, among other things, bring about an actional (and hence aspect-relevant) change in meaning, together rather indiscriminately under the global category of actionality (аспектуальность). Accordingly, he understands his three-phase model as a schematic representation of aspect types, although in fact, in itself, it initially represents only a segment of aspect-sensitive vocabulary (see above). Similarly, Stepanov treats the genus verbi and the passive (combined under ‘залог’) alongside changes in lexical diathesis (i.e. argument structure), subsuming them all under the single category of ‘залоговость’. This is legitimate insofar as, as

¹⁰ Particularly in Stepanov (1977), it is unclear how his synchronic classification of transitivity types and the three-phase actional model relate to the sequence of morphological oppositions.

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 89

Stepanov states that his primary concern is demonstrating the systematic conceptual relationships between grammatical categories such as aspect or passive and the functional (conceptual) domains that motivate them. A joint consideration, however, becomes problematic if it blurs the difference in status between the categorial oppositions involved (more grammatical vs. more lexical), or even declares it irrelevant.

Stepanov is basically aware of this circumstance, at least in the domain of diathesis. For one thing, he points out that affixation which, for example, explicitly marks transitive stems (i.e. derives them from intransitive or diathesis-wise labile stems) also results in the derivational and derived stems (verbs) each acquiring different lexical specifications (see, among others, Stepanov 1977: 140). He is likewise aware that one and the same affixation technique may have a purely syntactic function on one occasion (marking the passive) and a lexically derivational function on another (e.g. the formation of decausative verbs); an example is the already familiar Russian postfix {sja} (Stepanov 1978a: 339). Finally, in the same passage he shows that the opposition active vs. passive and an opposition between lexical diathesis distinctions may have quite different, and in some cases even incompatible, consequences for interpreting the state of affairs in question. Thus, the active sentence (7a) can be paired with the denotatively equivalent (i.e. synonymous) passive sentence (7b):

(7a) Лаборант расплавляет вещество

‘The laboratory assistant brings the substance to melting’.

(7b) Вещество расплавляется (лаборантом)

‘The substance is (by the laboratory assistant) brought to melting’.

Neither of the two sentences necessarily entails that the substance being processed actually melts. That, however, is precisely what is denoted (asserted) when the transitive verb расплавить ‘cause to melt, bring to melting’ is contrasted with its intransitive lexical equivalent маять ‘melt.INTR’; cf.:

(8) Вещество тает.

‘The substance melts’,

that is, with respect to the terminative component, (7b) does not necessarily entail (8). Conversely, (8) does not entail (7b) with respect to the agent presupposition: only (7b) presupposes an external causer (it is precisely this component that makes 7b a denotative paraphrase of 7a), whereas (8) excludes one.

The same applies in principle to Lithuanian, where this opposition is still expressed very regularly (though no longer productively) by the relationship between a simple and an extended stem; cf. Krosnis kaista ‘The fireplace gets hot’ ⇒ Šeimininkas kaitina krosnį ‘The owner heats the fireplace up/on’. Paraphrasing Stepanov, one may conclude from this that, in a lexical diathesis opposition of the kind encountered within the so-called ‘inchoative–causative alterna-

90 |

BJÖRN WIEMER

tions’ besprochen wurde (vgl. Haspelmath 1987), häufig¹¹ eine gewisse Verschiebung in der Darstellung der realen Situation stattfindet; beim Passiv ist das, wie gesagt, nicht der Fall.

Lack of rigor in conceptual differentiation is not simply a terminological matter. For in Stepanov it also leads on several occasions to an (presumably unintended) conflation of ‘passive’ and ‘diathesis’ and to misleading, if not outright incorrect, statements¹². This becomes even clearer with regard to his use of ‘aspect’ (вид). He presents Aktionsarten as “constituents fitted into the Slavic aspect system” (“способы действия включены в систему видов как одно из звеньев последней”; Stepanov 1976: 411). If this was intended to correspond to a familiar thesis concerning one of the diachronic sources of the aspect system in present-day Slavic, such a claim could be defended. Stepanov’s further discussion, however, makes clear that he places on the same level, as aspectual oppositions, the present-day correlation pf.: ipf. and the opposition between ‘determinate’ and ‘indeterminate’ stems still discernible in Common Slavic and the earliest historically attested stages of East Slavic (see, among others, Stepanov 1977: 151). As for the latter opposition, it should be clear that it arose more or less directly from the ablaut formation + stem extension of the type *nes- ⇒ *nos-i- ‘carry’ discussed above. Functionally, it was much less abstract than the opposition pf.: ipf. Accordingly, it was also more lexically restricted than the present-day Russian aspectual opposition¹³.

Stepanov also mentions promiscuously in this connection the Proto-Slavic/Common Slavic opposition between aorist and imperfect, the Indo-European distinction between present, aorist, and perfect stems already mentioned above, and certain “actional reflexes” in the Baltic distinction between preterite and present stems. The latter involve (today) tense-dependent stem and root alternations such as those listed in (1-6)¹⁴. Ablaut phenomena belong here as well; compare, for example, perk-a.3.PRS vs. pirk-o.3.PRT for pirk-ti ‘buy’, liek-a.3.PRS vs. lik-o.3.PRT for lik-ti ‘remain’¹⁵.

¹¹ Not always, by any means. For the fact that the implicature of sentence (7b) to sentence (8) is cancellable is due, inter alia, to the referential properties of the highest-ranking argument (here: лаборант): if this were not animate but a referent lower on the animacy scale, the implicature above would be obligatory. This is the case, for example, when the highest-ranking argument refers to a natural force; compare Russian Солнце растапливает снег (corresponding to Lithuanian Saulė tirpdo sniegą) ‘The sun melts the snow’ ⇒ Снег тает / Sniegas tirpsta ‘The snow melts’ (where tirpti.INTR ⇒ tirpdyti.TR).

¹² So etwa in Stepanov (1978a: 338), wo es heißt: „Залог заключается в отношении глагола-сказуемого предложения к подлежащему предложения. Он является, по-видимому, универсальной категорией различных языков [sic!] и в значительной степени связан с поверхностной организацией высказывания.“ Stepanov versteht das Passiv einerseits zwar eindeutig als eine Operation, die nur die morphosyntaktische Kodierung von Argumenten tangiert. Andererseits aber behauptet er hier, daß das Passiv als eine universal in allen Sprachen vorhandene Kategorie anzusehen sei.

¹³ For a comprehensive critique of a direct derivation of the opposition pf.: ipf. from the determinate–indeterminate opposition, see Bermel (1997: 67ff., 84f.).

¹⁴ See further examples such as Lithuanian tek-a.3.PRS vs. tek-ė-jo.3.PRT for tek-ė-ti ‘flow’, mieg-a.3.PRS vs. mieg-o-jo.3.PRT for mieg-o-ti ‘sleep’, tur-i.3.PRS vs. tur-ė-jo.3.PRT for tur-ė-ti ‘have’.

¹⁵ Although these were eliminated earlier in Slavic than in Baltic, there are still a number of verbs whose present and infinitive stems preserve the old distribution according to full grade and

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 91

Comparisons of this kind are nevertheless imperfect, because it makes a great difference whether stem alternations such as ablaut or infixation, viewed synchronically, take place within identical lexemes (i.e., are fully integrated into the paradigm) or lead to a lexical change. If, for example, the two sets of Lithuanian preterite endings (-ė < *ē vs. -o < *ā for the 3rd person) should indeed reflect a distinction between transitive and intransitive verbs inherited from Indo-European (see, however, especially footnote 9), that distinction has now been largely blurred. Today the two sets are morphonologically complementary endings for the same abstract categories of person/number and preterite—nothing more and nothing less. The same applies to the st-suffix and the infixes in the present: these affixes still occur almost exclusively with intransitive stems (and “commute” with the o-preterite in this respect), but otherwise they merely indicate the present. They have neither a specific actional nor a diathetic function.

As for the Slavic imperfect (in -aachъ/-ěachъ), it should be borne in mind that, regardless of whether a stem extension in ē played a role in its formation and, if so, whether this was etymologically identical to the *ē stem extension used to form progressives¹⁶, the grammatical status in these cases is different: the imperfect does not change the structure of the lexeme, whereas the *ē stem extension brings about corresponding modifications—which is why the imperfect is classified as inflection, while the stem extension by *ē is classified as derivation. Conversely, the imperfect suffix is restricted to one tense, whereas the other stem extension by *ē (in the type teka.3.PRS, tekejo.3.PRT tekēti ‘flow’) permits inflection through all tenses.

The prehistory of the verb morphology of older Indo-European languages in particular is full of examples in which the morphotactic composition of word forms remained identical, while scholars posit a shift from derivation to inflection (possibly also in the reverse direction). One example is Barton (1981: 248–250), who (like Stepanov, see above) assumes that Proto-Indo-European had only bare stems (or roots), which, through their lexical actional properties—punctual/terminative versus progressive—each implied the tense in which they alone could be used: punctual/terminative ⇒ past (compare the so-called root aorists) versus progressive present. Later, new stems are said to have arisen through affixation to root aorists; by default, these had an oppositional actional form and could be used in the present; compare *der-t ‘burst, split, tore away, tore off’ ⇒ *dere-ti ‘flays, is tearing’. Barton calls this process

reflect the zero grade (the infinitive behaves here just like the former aorist); compare, for example, OCS berǫ ‘I take’ vs. bъrati ‘to take’, židǫ ‘I wait’ vs. žьdati ‘to wait’. See, among others, Stepanov (1976: 409; 1978a: 343f.) and the quotation in footnote 5.

¹⁶ Gegen eine solche Etymologie äußert sich vor allem Stang (1943: 82–85). Generell gibt es zwei konkurrierende Thesen zur Herkunft dieser ur slavischen Innovation: die eine führt das Imperfekt-Suffix zurück auf die Fusion eines stammerweiternden *ē oder *ā mit einer Vergangenheitsform der Copula *es- (u.a. Stang, Diels), die andere sieht im Imperfekt-Suffix eine Erweiterung des Aoriststamms durch *ē/*ā in iterativer oder progressiver Funktion (vor allem Kurylowicz) (vgl. Birnbaum 1975: 162f.; Stang 1943: 81). Nur die zweite These nimmt etymologische Identität des einen *ē mit dem anderen an. Welchen Standpunkt man aber auch einnehmen mag, sie laufen beide auf eine Stammerweiterung durch ein „iterativ-progressives“ Morphem hinaus.

92

BJÖRN WIEMER

‘Derivation’, since the further course of development in Slavic, among other things, showed that original roots of the type *der- and stems derived from them (*dere- etc.) had been regarded as “lexically distinct.” For both stems would then have produced their own present and/or aorist formations (through ablaut and/or suffixation): root aorist *der-t ⇒ OCS present stem -dbret and *dere-ti ⇒ *d̥rāsom > OCS d̥rachъ. He now calls this process ‘inflection’, because only the temporal reference changes here, and a regular paradigmatic correspondence gradually emerged that left the lexical content of the original stem untouched.

This argumentation is thus based on the relationship of the morphological operations in question to the lexicon. But it also rests on the dissolution of unidirectional functional implications, through which verb roots, owing to their lexical aspectual default, indicate one or another tense. In short, it is founded on the dissolution of lexical-temporal syncretisms, provided that by such syncretisms one understands implicational connections between two (or more) categorical distinctions (prototypes). In the present case, these are grammemes of temporal and diathesis-relevant categories on the verb. Such syncretism presupposes that the respective grammemes already exist as distinct entities; if this is not the case, one should speak instead of ‘(categorical) diffuseness’ (see below). If such a cluster is dissolved so that the grammemes can be marked independently of one another, ‘discrimination’ takes place¹⁷. In the case discussed by Stepanov, discrimination brought about a crosswise combination of lexical meaning and tense. This combination underlies the grammaticalization of tenses (but not of aspect!); see also Lehmann (1990) on this point.

It should be noted, however, that derived stems (“extended roots”) also generally determine the type of inflection (or, to put it neutrally, the type of morphological change at the end of the word form); that is, here too there is an implication of morpheme types following the ablauting roots or suffixal extensions. Thus, nothing essential has changed in morphotactics; what we have before us are examples of stem change, to which various tense, person, and number markers are added¹⁸ and which then combine under certain conditions into paradigms with more or less regular stem alternations. These stem changes are paradigmatic only because they correlate with other categories (here, tense, person, and number) that have become morphologically distinct in some way and thereby warrant the delimitation of grammemes.

If the stem indicates tense, one can again speak of syncretism. The most prominent case is the present stem of perfective verbs in East Slavic and Polish (i.e. in the northeastern part of the Slavic linguistic area), which in-

¹⁷ For these terms and the operations underlying them, see Lehmann (1996; 1999: 229ff.).

¹⁸ These may in turn be fused together syncretically—a phenomenon typical of inflectional morphology.

Rekonstruktion grammatischer Kernbereiche | 93

cidentally also indicates the future and is thus no longer a present¹⁹. This, however, is a very late consequence of the grammaticalization of aspect (which is also not manifested with equal consistency in all Slavic languages); here syncretism occurs at a purely functional level (cf. Kardelis/Wiemer 2002: 53f.; Wiemer 2003: 44–46). It must have been preceded by a phase in which a functional affinity between temporal boundedness and future reference (possibly with modal “shadings”) existed at the lexical level. For as long as a language’s inventory of verb stems is not divided consistently enough into two (or possibly more) classes, each endowed with a clearly delimited set of functions, the correlation with tense (preterite vs. present/future) remains a relatively “loose” matter: the correlation can be annulled “when needed.” In early historically attested stages of Slavic languages and in present-day Lithuanian, precisely the latter is the case; that is, tense, as an already grammatical category, forms (or formed) loose syncretisms with lexically governed aspectual preferences (cf. Bermel 1997: chs. 4–8 passim; Wiemer 2001: 39–44, 47).

But to return to morphotactic processes. Stem changes such as those discussed by Barton are observable (or reconstructible) at every turn in the evolution of early Indo-European languages. And so, on surveying, for example, Stang (1943), the relevant passages in Stang (1966) and Kazlauskas (1968: 287ff.), the works by Stepanov discussed above, and many other authors who have examined in detail the verb morphology of the Slavic and Baltic languages against an Indo-European background, one gets the impression that the entire development of verb morphology in Baltic and Slavic consists almost exclusively of the integration of derivational techniques into paradigmatic relations, including the reinterpretation of aspect-relevant distinctions as temporal ones. The latter, however, are almost never grammaticalized to such an extent that form-function syncretisms are split up (discriminated) and the lexical restrictions arising from aspect-sensitive classes largely disappear.

The upshot of what has been said so far is this: stem alternations oriented toward aktionsart may participate in paradigm formation to a greater or lesser degree. They do so through interaction with other categories developed on the verb, such as tense or those that specify argument structure (diathesis-related categories). Since tense, unlike distinctions pertaining to argument structure, does not bring about lexical changes, close connections with tense distinctions are in principle lexeme-preserving, whereas close correlations with diathesis-related distinctions quickly lead to a change of lexeme. In the extreme case, if the aktionsart function of a morphological marking strategy is strictly bound to a tense (as with the Baltic nasal infixes and {st}, which occur only in the present), this syncretism results in a more stringent paradigmatization of the strategy; that is, it becomes firmly integrated into the paradigm of the category it implicates (here: tense).

¹⁹ Note well: this is not really a matter of derivation, but rather of classification, whereby verbs in these languages are consistently assigned to the class of perfective verbs or to the complementary class of imperfective verbs. Whether a verb is perfective or imperfective does not necessarily follow from its stem structure per se, but from its relationship to the stem of the verb of the opposite aspect that is morphologically closest to it (cf. Wiemer, in press1: section 4).

94 |

BJÖRN WIEMER

I would like to emphasize once again that, although these observations could in principle be identified with a transition from derivation to inflection, nothing changes in morphotactic behavior, for example, in the case of nasal infixes and the st suffix.

In discussing Stepanov’s historical-comparative work on the problem at hand, I have referred to syncretisms between actional and diathetic features. As mentioned, however, syncretisms presuppose categories (or meanings) with distinct oppositions. If Stepanov (along with many others) assumes that, at the beginning of the development of Indo-European syntax and morphology, no transitivity correlation was yet present and neither aspect nor tense distinctions were marked, but only verb roots existed, which must be characterized as activa tantum, perfecta tantum, etc., then these categories existed, at most, in a diffuse form. This means that not only were they not freely combinable (which would have resolved the syncretisms), but there were no categorical oppositions ‘transitive vs. intransitive’, ‘present vs. aorist’, etc. that could have been combined.

Furthermore, Stepanov also subscribes to the thesis that early Indo-European had an active:inactive alignment (see Section 2). If one were to accept this thesis, one would have to assume that the link between argument and predicate was syncretic: both would have to agree with respect to the feature [± animate]²⁰. This syncretism (if it ever existed at all) would, however, be purely semantic in nature and, moreover, left no direct traces in the morphosyntax of the Baltic and Slavic languages. On Stepanov’s premises, the emergence of distinctions between transitive and intransitive stems and of tenses then gave rise to the distinct categories that could be combined with one another and with actional distinctions (which had not yet been grammaticalized), and which seemed to imply one another, or each other, to a considerable extent. Thus, even if one wishes to follow Stepanov’s argument, one must assume that the emergence of such syncretisms would have been preceded by the elimination of the diffuse character of verb roots with respect to the categorical distinctions just mentioned.

Returning now to Stepanov’s treatment of actional distinctions, we find that he confuses grammatical aspect functions with actional defaults and lexical modifications. This is most striking when Stepanov describes secondarily suffixed verbs such as Russian выпивать ‘drink up/empty by drinking’, выделывать ‘produce, manufacture’, разрисовывать ‘paint/draw in detail’, закалывать ‘stab to death’ as “biaspectual” (двувидовые), explaining that these verbs are prefixed and therefore, unlike the unprefixed stems (пить ‘drink’, делать ‘make, do’, рисовать ‘draw’, колоть ‘stab’), are bounded (1977: 151). He does note, however, that the secondarily suffixed and the simple stems also differ from one another lexically and therefore often cannot be used in place of one another. As a further argument for the “biaspectuality” of the aforementioned secondarily prefixed verbs, he contends that they, instead of the correspon-

²⁰ Stepanov (1989: 75ff.) spricht deshalb auch von „semantischer Kongruenz“ („семантическое согласование“).

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 95

perfective verbs ending in … – without the suffix {iva / vva}, that is, вынуть, выделать, разрисовать, заколоть (with the same meanings as the suffixed verbs) – would occur in the narrative present, i.e. quasi in the function of the perfective aspect. It is clear from this that Stepanov conflates lexical and grammatical functions. To be sure, with his observation that the secondarily prefixed verbs and the corresponding non-prefixed verbs do not fully coincide lexically and therefore cannot simply be interchanged in the narrative present, he has in fact touched on a central point of “aspectogenesis,” so to speak its interface with lexical semantics (cf. Wiemer 2001: 48–53).

Overall, it should be added that although Stepanov’s argumentation and theoretical premises may be outdated, his conceptual approach remains relevant. A prerequisite for this is—alongside a better determination of the actual course of the chronological layers and their undoubtedly existing interweaving—the ability to integrate the lines of development he sketched within verb morphology into theories that assume a principled separation of grammatical and lexical categories (oppositions) and are able to operationalize the continuum between them.

2. SCHMALSTIEG: DIE ABFOLGE VON ALIGNMENT-SYSTEMEN UND DIE „ERGATIVE VERGANGENHEIT“ DES BALTISCHEN

Ebenso wie Stepanov (vgl. vor allem Stepanov 1978a: 360–362; 1989: passim) ist auch Schmalstieg ein Repräsentant der These von einem Aktiv:Inaktiv-Alignment in den Vor- oder Frühstufen des Ide. Auf die Diskussion darüber, ob das frühe Ide., genauer: das Ide. nach dem Aufkommen der ersten flektivischen Unterscheidungen, durch ein solches Alignment-System charakterisiert war, braucht hier aber im Grunde nur deshalb eingegangen zu werden, weil sie als Hintergrund für die These vom ursprünglichen Ergativ-Alignment des Vor- oder Frühbaltischen gilt; um diese These geht es im weiteren. Die Hypothese von einem Aktiv:Inaktiv-Alignment des Prä- bzw. Proto-Ide. konnte aufgrund der Unhaltbarkeit ihrer Argumente mehr und mehr in das Reich der Spekulationen verwiesen werden und darf heute als obsolet gelten²¹ (s. Fn. 22). Trotzdem ist es lehrreich, sich etwas näher die Vorgehensweise anzusehen, mit welcher bis vor einiger Zeit noch Entwicklungs-Abfolgen der grammatischen Kernstruktur des Baltischen (und Slavischen) auf indoeuropäischem Hintergrund postuliert wurden. Dies geschah meist ohne Rücksicht auf Sprachkontakt-Faktoren (nicht zuletzt von außerhalb des ide. Dialektkontinuums). Zudem spielt im Fall einer Rekonstruktion des Alignments ein Abgleich der postulierten Entwicklungsrichtung mit typologischen Fakten eine noch größere Rolle als bei der Verbindung zwischen Aktio­nalität und Diathese (s. Abschnitt 1).

²¹ The active:inactive type entered discussion primarily through Klimov’s ‘contensive typology’, although Uhlenbeck and Sapir should be mentioned as important precursors at the beginning of the 20th century. Bauer (2000: 15–22, 32f., 43–92) provides a detailed overview of the history and current state of the discussion concerning the inactive:active and ergative:absolutive structures, as well as whether a particular sequence between the two existed in earliest Indo-European; she herself still argues for a Proto-Indo-European active:inactive system.

96 |

BJÖRN WIEMER

What, then, is the thesis of which Schmalstieg is also an advocate? If an active:inactive system existed, it is supposed to have preceded an ergative system²². Furthermore, the distinction between stems of the active and inactive classes is regarded as the basis both for the emergence of a gender opposition between ‘common gender’ and ‘neuter gender’ and for the dichotomy between aorist/present and perfect stems²³. If the basis for these already extremely old (dating to around 2000 BCE) morphological distinctions is supposed to have been the postulated active:inactive system, this alone shows how far back in time the latter must have existed. Furthermore, the active class is supposed to have formed the basis for the ergative and the latter, in turn, for the nominative; correspondingly, the inactive class formed the basis for the ‘casus indefinitus’ (= absolutive), which in turn formed the basis for the accusative (see below). This hypothesis would thus yield the following sequence of alignment systems: active:inactive > erg:abs (‘indefinitus’) > nom:acc (cf. Schmidt 1980: 103; 1983: 9f.); see, however, Dixon (1994: 185f., fn. 2) for a critique of this sequence.

We shall now focus on the discussion of ergative syntax, or alleged remnants of an ergative system inherited from Proto-Indo-European. I would like to emphasize that ‘ergative form’ or the like cannot simply mean a frequent marking of the agent with an oblique case or a PP²⁴. Such an equation has sometimes led to misunderstandings in Slavic studies as well, when some authors speak of “ergative syntax (construction, etc.).” The confusion arises because it is not taken into account that a morphosyntactically marked expression of the highest-ranking argument does not in itself establish an ergative pattern. On the one hand, this argument must stand in the sentence in opposition to a complementary argument (which would correspond to the absolutive); on the other hand, in contrast to the passive, the oblique marking of the highest-ranking argument must be identifiable as the morphosyntactically unmarked pattern. I will return to this shortly.

For Baltic, and Lithuanian in particular, ergative syntax has been discussed in connection with sentences of the type

(9) Čia žmoni-u gyven-t-a / gyvena-m-a here human.GEN.PL live.PTC.NEUT ‘People have (apparently) lived here / people (apparently) live here’.

(10) Sen-ų mišk-ai mylė-t-a old.GEN.PL forest.NOM.PL love.PTC.NEUT ‘The forests were loved by the old people’

has been discussed. These sentences are notable in that (a) the predicative participle agrees with no

²² Vgl. dazu die sehr kritische Besprechung von Stepanov (1989) durch Krasuchin (1990, insbes. S. 73 und 78) und neuerdings auch Crome (2001: 24–27).

²³ Zu diesen und deren Verkoppelung mit dem Medium vgl. insbesondere Stepanov (1989: 24ff.); s. Abschnitt 1.

²⁴ This is how, for example, Rosén (1979: 165) understands it.

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 97

NP, even though a nominative NP can occur (see example 10), and (b) the highest-ranking (possibly sole) argument of the verb from which the participle is derived is expressed in the genitive. There are at least two hypotheses concerning the origin of this construction, one of which is based on the assumption of a prehistoric ergative syntax. The principal advocate of this hypothesis is Schmalstieg (1982; 1988; 1993; 1994), so I shall briefly call it the ‘ergative’ or ‘Schmalstieg hypothesis’²⁵.

Schmalstieg recognizes the etymological relationship between the *to-suffix of the t-participle and the 3rd person singular of the Proto-Indo-European middle aorist (e.g. in Vedic and Greek)²⁶. In addition to this initial assumption, he proceeds on the basis that early Proto-Indo-European had no agreement between constituents, that no transitivity opposition existed, and that, accordingly, stems were indifferent (or diffuse) with respect to diathesis. Sentences such as (10) should therefore also be viewed by diachronic analogy with other nominal sentences (Schmalstieg 1982: 132f.); compare, for example,

(11)

Mēs-a sveik-a valgy-ti

meat^NOM.SG.F healthy^NEUT eat^INF

‘Eating meat is healthy’.

This would, however, correspond to a neutral alignment, since no clear pattern of case selection or agreement marking can be discerned with respect either to the feature [±active] or to the distinction [±transitive] (cf. Plank 1995: 1185). After all, according to Schmalstieg’s own account, neither existed at the outset. When the transitivity distinction arose, and also manifested itself in the form of case marking, the original ‘indefinite case’ in intransitive sentence patterns was first refunctionalized as the absolutive and then as the nominative, while a special case was used for the highest-ranking argument with transitive predicates. In this way, an ergative pattern initially arose. The former ergative case then developed into the present-day genitive²⁷ (Schmalstieg 1982: 121). It is true that the development of an ergative pattern from a neutral alignment on the basis of transitive sentence types is well attested typologically (Plank 1995: 1194). In the case of the former Proto-Indo-European ergative, however, it seems more plausible that it was reinterpreted as the nominative through extension to arguments of intransitive predicates; these

²⁵ Other proponents of this thesis would be Marvan (1973) and Palmais (1977).

²⁶ Diese Annahme stützt sich auf Hirts Studien zur ide. Syntax; vgl. Schmalstieg (1988: 36f.).

²⁷ Schmalstieg (1982: 121) nimmt hier rekonstruierte Sätze an wie *patér bhr̥-tó vīr-os ≈ ‘Vater getragen des Mannes / durch den Mann’ mit patér als ursprünglichem Indefinitkasus, bhr̥tó als Verbalnomen und vīros als ergativisch zu verstehendem Kasus (später Genitiv). Leider gibt er nicht an, woher diese Rekonstruktion stammt und wie gesichert sie durch historisch-vergleichende Daten ist. Ähnlich wenig gesichert ist seine Behauptung, daß der Genitiv in Sätzen wie lit. ^NOM.SG pribego vandens^.GEN.SG ‘Das Zimmer lief mit Wasser voll’, Žemė^.NOM.SG primirko lietaus^.GEN.SG ‘Die Erde sog sich mit Regen voll’ auf einen Ergativ zurückgehe (Schmalstieg 1984). Heute wird hier der Genitiv zur Angabe einer gewissen Menge einer Substanz bei zweistelligen, aber intransitiven Verben verwendet (s. die gerade zitierten Beispiele). Ein solcher „ergativischer Zusatz“ zu intransitiven Sätzen wäre logischerweise nur für die Zeit denkbar, als es noch keine einigermaßen deutliche Transitivitäts-Opposition gab (s. dazu auch Stepanovs Hypothesen in Abschnitt 1).

98 |

BJÖRN WIEMER

Entwicklung findet u.a. in kaukasischen Sprachen Parallelen (vgl. Schmidt 1980: 103f.). Somit stellt sich die Frage, ob es sich bei Sätzen vom Typ (9–10), welche Ziel der Erklärung Schmalstiegs sind, wirklich um ein ergatives Argument handelt. Dazu sofort mehr.

The Schmalstieg hypothesis, however, presupposes even more. Thus, in the model that must be assumed for the original juxtaposition of sentence constituents, there would have had to be a construction in which the nominative (or, for Schmalstieg, the absolutive) stood alongside a non-agreeing t participle (or verbal noun) of intransitive stems; in Lithuanian this would correspond to a structure such as

(12) *tėv-as gulė-t-a²⁸

father.NOM lie.PTC.NEUT

≈ ‘father lain’ = ‘The father lay’.

Furthermore, because the early Indo-European language is assumed to have lacked a subject–object distinction, this hypothesis also presupposes that forms such as pasnig-t-a²⁹ ‘snowed’, paly-t-a ‘rained’, i.e. forms of avalent verbs (pasnigti ‘to snow’, palyti ‘to rain’), belonged to the original participles (Schmalstieg 1982: 119), rather than having arisen only later through the lexical expansion of a morphological marking pattern. In this connection, the agreementless sentence pattern in (10) would also have to be diachronically prior to an agreeing pattern (corresponding to the passive) such as

(10’) Sen-ų mišk-ai mylė-t-i

old.GEN.PL forest.NOM.PL.M love.PTC.NOM.PL.M

‘The forests were loved by the old’

existiert haben (Schmalstieg 1982: 119f.). Den Versuch einer glaubwürdigen Rekonstruktion, welche diese Annahme stützen würde, unternimmt Schmalstieg allerdings nicht (genauso wenig tun dies im übrigen Matthews 1955 oder Ambrazas, die sich in diesem Punkt mit Schmalstieg decken; zu Ambrazas s.u.). Abgesehen davon wäre es für seine Hypothese eine entscheidende Frage, in welcher (vorhistorischen) chronologischen Relation sich der von ihm postulierte Kasuswandel und die Entstehung des *to-Formans befinden (vgl. Ambrazas 1994: 9). Auch darauf antwortet Schmalstieg nicht.

We can say with some confidence about the *to formant that it acquired a clear diathematic contour relatively late in Baltic (and Slavic), perhaps only with the onset of literacy (in Baltic, therefore, only in the last 300–400 years), i.e. one of the arguments of the derived verb became clearly fixed²⁹. The

²⁸ This illustrates intransitive stems with an inactive argument. But since ergative syntax does not distinguish between active and inactive predicates, sentences such as *Tėvas_.NOM.SG.M eita_.NEUT ≈ ‘father gone’ would hypothetically also have existed (cf. the discussion in Ambrazas 1994: 8).

²⁹ The process of contouring consists in resolving an underdetermined (= diffuse) prototype; this may be represented by a ‘gram’ (in the sense of Bybee, Dahl 1989). Before resolution (contouring), from a logical perspective there is an ‘and/or’ relation between the variants of the prototype, which can be decided (or left open) only under the influence of context. A typical case of a diffuse (i.e. contourable only through context)

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 99

The morphological cases of Indo-European languages, by contrast, are known to be of much older origin. From a typological perspective, one can say the following about the role of morphological cases in a shift from an ergative to an accusative system: such a shift consists in extending the use of the initial ergative case (used only to mark the highest-ranking argument of transitive predicates!) to arguments of intransitive predicates. Through such an expansion, the ergative ceases to be an ergative and becomes what is commonly called the ‘nominative’. Thus, above all, the markedness relations in the alignment are reversed³⁰. And that is precisely what does not happen in the case above (9–10): the genitive remains an oblique case, and the sentence pattern in (9–10) remains “non-basic” against the background of otherwise accusative syntax³¹.

Note also that Schmalstieg’s reconstruction prima facie applies only to constructions without agreement featuring t-participles. If one wishes to explain why m-participles also occur in non-agreeing sentence patterns of type (10–11), one is forced to assume that a far-reaching analogy took place on the model of *to³², and this analogy would have to be related to other relevant processes in the diachronic chronology of alignment systems and associated grammatical phenomena. Likewise, the non-agreeing use of the currently gender-neutral t-participles of intransitive verbs (as in example 9) would have to be explained by analogy. For Schmalstieg’s thesis rests on the claim that an ergative case first arose only in transitive sentence patterns. But again: even if, through analogy, this case had also been used to mark the sole argument of intransitive predicates, such an analogy would have led to a marking pattern that should be classified not as ergative but as accusative alignment³³. Compare (9–10): there the sole argument of the intransitive verb gyventi

The pronoun kas ‘who and/or what’ represents the morpheme in Lithuanian. On the concept of contouring, see Lehmann (1996; 1999: 234), Wiemer (2001: 47ff.). On the diathesis orientation of this and comparable deverbal formatives, see Wiemer (2004: 288–293) with the relevant bibliographical references.

³⁰ Vgl. Plank (1995: 1186f., 1991), der in einem Wechsel des Alignment u.a. „consequences of revaluations of non-basic constructions as basic and vice versa“ sieht, sowie auch Dixon (1994: 182–206, passim).

³¹ Nur als Kuriosum sei noch folgendes vermerkt: Schmalstieg (1982: 122) spricht entgegen seiner eigenen Hypothese – die ja den Ergativ zum Genitiv werden läßt – noch den Standpunkt anderer Verfechter der These von einer ergativischen Vergangenheit des Idg. an, wonach nicht der Genitiv, sondern der Nominativ als Nachfolger des Ergativs anzusehen wäre (dies entspräche der gerade dargelegten, typologisch belegten Entwicklungsfolge). Diese Annahme sieht Schmalstieg als „completely unnecessary“ an (in Schmalstieg 1993: 41 freilich zieht er sie schließlich wieder in Erwägung!). Seine Erklärung endet dann aber in einem Akkusativsystem – ohne die „Zwischenstufe“ des Ergativsystems. Man kann sich bei einer solchen Argumentationsweise nicht des Eindrucks erwehren, daß es sich um inkohärente ad hoc-Versuche handelt.

³² The same would apply to Slavic n-participles, insofar as their non-agreeing sentence patterns can likewise be connected with structures such as those in (10–11). Stepanov (1978b: 348) therefore pointed out that the complementary morphonological distribution of n- and t-suffixes ultimately argues against a hypothesis such as Schmalstieg’s.

³³ This is precisely what distinguishes the Lithuanian case from Indo-Iranian, often cited for comparison (cf. Christen 1998: 58). For the basic marking patterns, see also Comrie (1975: 114f.). Accordingly, the case pattern resulting from Ex. (9–10) would be close to an anti-ergative system, such as is found, for example, in modern Finnish.

100 |

BJÖRN WIEMER

‘live’ (žmoniu) in the same case as the highest-ranking argument of the transitive verb mylėti ‘love’ (senų); syntactically, this corresponds to an accusative pattern, in which, to be sure, the morphological marking relation would be “reversed.” However, neither this nor an Indo-European precursor with a juxtaposition of the type *tēv-as_.NOM.SG.M dirb-t-a _/ gulē-t-a_{.NEUT can be reconstructed plausibly in any way (Ambrazas 1990: 213).}

Finally, the Schmalstieg hypothesis implies that the present-day passive and subject-impersonal in Lithuanian (with the gender-neutral ta/ma-participles) developed from an ergative system. This would, however, be the only case known to typology so far in which a passive developed from an ergative system. Among attested (not reconstructed) cases, only cases are otherwise known in which a passive marker was refunctionalized as an ergative marker (cf. Estival/Myhill 1988 and #1142 from the Konstanz “Universals Archive”).

Stepanov (1978a: 355ff.; 1978b: 348f.) had already taken a position partly opposed to Schmalstieg’s assumptions, discussing the emergence of sentence types (9–10) not on the assumption of ergative case marking, but implicitly by reference to an active:inactive alignment, which, in his view too, was supposedly inherited from early Indo-European by the precursors of Baltic and Slavic (which he calls “Proto-Balto-Slavic”). Accordingly, the oblique marking of the argument in (9) is to be regarded as original, because an animate referent appeared in a role “atypical” for it, namely inactive; correspondingly, it was also marked morphologically and syntactically (by the absence of agreement). On the other hand, Stepanov sees in sentences of type (9) a connection with the Indo-European perfect, which disappeared in Baltic and Slavic. The perfect is generally regarded as a stative category (and, among other things, is therefore associated with the middle voice, or formed a form–function cluster with it). After the perfect disappeared in Baltic and Slavic, a kind of empty slot arose in their systems, which, according to Stepanov, was filled by infinitive forms of the verb and oblique argument marking³⁴. The development of an analytic passive was thus diachronically secondary to the “perfect-like” (more properly: stative) semantics of sentence type (9).

An objection to this thesis is that, like Schmalstieg’s, it raises questions about the verifiability of the chronological sequence: what is the temporal relation among the perfect function, the passive function, and the function in (inferential) evidentiality³⁵? Moreover, Stepanov does not ask whether sentence type (9) really corresponded to a stative or rather to an ‘experiential’ (which the evid-

³⁴ He also makes a broad, direct connection with intransitive sentences with a nominal predicate and oblique argument marking in present-day Russian and Lithuanian; especially with the dative experiencer (cf. Russian Мне холодно, Lithuanian Man šalta ‘I am cold’) or an external dative possessor with the accusative of the lowest-ranking argument, as in Russian (dial.) Мне^.DAT ногу^{.AKK больно} literally ‘My leg hurts’ or Lithuanian Man^{.DAT skauda galva.AKK} literally ‘My head hurts.’ Stepanov overlooks, however, that many of these sentence patterns apparently arose only relatively late, namely in historical times (see Galkina-Fedoruk 1958 generally on this).

³⁵ It is precisely in this function that sentences of type (9–10) are well attested in modern Lithuanian.

Rekonstruktion grammatischer Kernbereiche | 101

would be closer to the evidential function). Furthermore, his assumptions (as well as Schmalstieg’s) are based on hypotheses about a change in alignment, which, in view of the typological counter-evidence, leave him with the burden of proof (see above). Finally, they are based on a purely system-internal reconstruction, which entirely excludes possible influences resulting from language contact.

This leads, especially in Schmalstieg’s account, to wholly unnecessary ad hoc interpretations, which are not supported even by facts that would be relevant to the reconstruction. For example, when he interprets sentences of the type at will

(13) Ant šakel-ių ant žal-ųjų vainik-us kabin-t-a on branch.GEN.PL on green.GEN.PL wreath.ACC.PL.M hang.PTC.NEUT ‘On the branches, on the green ones, wreaths.ACC are hung’

in which the lowest-ranking argument is marked as an accusative object. Schmalstieg (1982: 121f., 129) speculates here that the verb kabinti ‘hang up’ was only later (in the course of the transitivity opposition) clearly reinterpreted as transitive, and that as a result the t-participle it formed, kabinta (today gender-neutral), came to be understood as transitive (in other words, as aligned with the highest-ranking argument). Such an “explanation” ignores the fact that, in the diachronic development of participles with the t-formant, alignment with the lowest-ranking argument prevailed (see Fn. 29). If in individual cases they show alignment with the highest-ranking argument with transitive base verbs, and can otherwise also be formed from intransitive and even avalent verbs (and thus be used “actively”), this could be seen to a greater extent as the result of a far-reaching expansion of this deverbal suffix to stems other than transitive ones, and to a lesser extent as the result of syntactic reanalysis (as in the Polish impersonal with the no/to-participle). Whether Polish influence can be assumed in the case of Lithuanian participles in -ta (as well as those in -ma) from transitive verbs with object government remains an open question³⁶. Apart from this, however, it is much more plausible to assume that the expansion of the deverbal t-formant proceeded not from transitive to intransitive, but rather from intransitive to transitive verb stems—that is, in precisely the opposite direction from that which the Schmalstieg hypothesis would imply.

Die letztgenannte Annahme rechtfertigen ein paar Argumente, welche Ambrazas (1990: 200–214; 1994; 2004) gegen die Schmalstieg-Hypothese vorgebracht hat. Dabei geht auch er von der Annahme aus, daß der frühe ide. Satzbau in erster Linie durch Juxtaposition nominaler Satzglieder mit nur schwach ausgeprägter Wortklassen-Differenzierung geprägt gewesen sei. Der Genitiv in Sätzen wie (9–10) gehe allerdings nicht auf einen Ergativ-Kasus zurück; diesen habe es sehr wahrscheinlich nie gegeben. Vielmehr sei der Genitiv in Sätzen vom Typ (9–10) als Fortsetzung eines sehr alten genetivus subiecti anzusehen, wie er auch in anderen älteren ide. Sprachen durchweg bei Verbalnomina

³⁶ For the diachronic development of the so-called “impersonal” passive in the Baltic-Slavic area, see Wiemer (in press 2).

102 |

BJÖRN WIEMER

(nomina actionis) can be observed³⁷. According to this reconstruction, today’s gender-neutral ta-participles go back to *o-stem verbal nouns, in which arguments first stood in the nominative and did not agree with the participle in -ta (< *to). Such a structure is reflected to this day in sentence patterns of the type

(14) rug-iai sėja-m-a / sė-t-a rye.NOM.PL sow.PTC.NEUT ‘The rye is / has been sown’

reflected. With verbal nouns bearing a t- or m-suffix, the “bearer” of the respective property could be expressed in the genitive (as a genetivus subiecti), both with transitive stems (valstiečio.GEN.SG.M sėjama / seita originally ‘the farmer’s sowing’, i.e. ‘sown by the farmer’) and with intransitive ones (valstiečio.GEN.SG.M gulima / gulėta originally ‘the farmer’s lying / having lain’, i.e. ‘the farmer lies / lay’)³⁸.

This reconstruction has the advantage (a) of being considerably more transparent and (b) above all of making no distinction either between t- and m-participles or between transitive and intransitive stems. Unlike the Schmalstieg hypothesis, this approach therefore does not incur an explanatory burden that would require additional hypothetical assumptions about far-reaching analogical leveling. Its greater transparency is connected with the possibility of assigning the sentence structure in (9–10) an exceptionally long diachronic continuity, in which at most the syntactic status of the genitive NP would have changed (and perhaps only in the last few centuries). Ultimately, Ambrazas’ approach also allows the thesis that the highest-ranking argument marked by the genitive is of possessive origin to be incorporated (see Fn. 37).

Given the highly speculative character of the ergative hypothesis³⁹, Schmalstieg’s decisions, which are generally made ad hoc (see especially Schmalstieg 1988: 30ff.) and are generally not solidly supported by facts, the inconsistencies outlined above, and the fact that Schmalstieg’s approach would require a passive to arise from an ergative system (which would run directly counter to the clear pattern attested in typological cases), Ambrazas’ explanatory approach should clearly be preferred.

³⁷ Dieser Standpunkt ist in so klarer Form erst in Ambrazas (2004) formuliert. Früher schien auch er der These Holvoets nicht abgeneigt gewesen zu sein, dass der Genitiv in diesen Satztypen auf einen possessiven, zuerst nur adnominalen Kasus zurückgehe, welcher später zu einem Adjunkt umgedeutet worden sei. Auch Stepanov nahm an, daß der Genitiv in (9) zuerst possessiv zu interpretieren war. Er sah dies allerdings als eine Weiterentwicklung des früh-baltischen und früh-slavischen „Perfekt-Ersatzes“ an (vgl. 1978b: 347f.).

³⁸ It is noteworthy in this connection that in Lithuanian dialects the genitive with ta- and ma-participles occurs even more frequently with intransitive stems than with such participles of transitive stems (Ambrazas 1994: 9).

³⁹ For other objections to the thesis of an “ergative past” in Indo-European, with further bibliographic references, see Comrie (2001: 24f.) and Matasović (2000: 330f.).

Rekonstruktion grammatischer Kernbereiche | 103

3. AN ANALOGOUS CASE IN SLAVIC: ‘ESSE’ AND ‘HABERE’ IN POSSESSIVE CONSTRUCTIONS

Alongside Schmalstieg, other researchers have until very recently also built their reconstructions on scarcely tenable ground. At times, features of a supposedly archaic morphosyntax, still originating in Proto-Indo-European, are even posited in place of innovations that may have taken hold in individual Indo-European languages only through language contact (including with non-Indo-European languages). A very clear case is, for example, Bauer (2000: 10, 151), who cites the Russian possessive construction of the type u + GEN (у меня книга ‘I have a book’, etc.) as an example of features that were “inherited from the protolanguage.” The question arises whether, in a certain sense, chronology has not been turned upside down here. Since this possessive construction can also be closely connected with the question of alignment (see Section 2), it is worth examining the methodological problem that arises in somewhat greater detail.

Let us therefore consider the thesis that the possessive construction u + GEN (+ ESSE, with the possessee in the nominative) in Russian represents a Finno-Ugric substrate. This thesis was discussed and essentially accepted by Veenker (1967: 117–119), whereas Panzer (1982: 330f.) and Vasilev (1973) have called it into question. Panzer’s skepticism, however, is de facto based solely on the (somewhat dogmatic) assumption that Slavic languages are in fact “BE-languages” and not “HAVE-languages” (following Isačenko 1974), and that it is therefore more likely that Russian merely continues an inherited Proto-Slavic state; if the other Slavic languages use a HABERE construction for possessive expression (OCS iměti and its equivalents), this is said to be attributable to the reinforcing influence of Greek (or, indirectly, of OCS), Latin, or later Western European languages. If Panzer is presumably right about the reinforcing influence, this would, however, presuppose that the HABERE verb was indeed part of the common Slavic inheritance from Proto-Slavic. The evidence in the etymological dictionaries also speaks relatively clearly in favor of this assumption (cf. Vasmer 1953: 479 and Walde 1930: 124). Panzer’s skepticism regarding Finno-Ugric influence is therefore not particularly well founded, and it is neither disproved nor confirmed. Rather, one should consider that both types of possessive construction already existed side by side in prehistoric times, and that the role of a Finno-Ugric substrate may have been to support a redistribution of these competing constructions in favor of the adessive (with ESSE) construction. This is precisely the assumption supported by Safarewiczowa’s study (1964, especially pp. 17, 55f.).

Veenker’s important observation—that the adessive ESSE construction established itself as the predominant possessive construction (with all its functional extensions) vis-à-vis the HABERE construction only in Russian, i.e. on the northeastern periphery of the Slavic-speaking area—must therefore be weighed against Vasilev’s demonstration that the adessive construction is attested already in early documented stages of various Slavic languages, including South Slavic ones, and not only in the narrow sense of possessive function but also in partitive function (with inalienables), the function of ownership, and that of kinship affiliation. Vasilev

104 |

BJÖRN WIEMER

(1973: 363) also rightly points out that, given what is known about the settlement movements of the direct ancestors of the East Slavs, one must assume that direct contact between them and Finno-Ugric tribes can only have begun in the 5th century AD. Since the construction ‘u + GEN + ESSE’ occurs already in the earliest surviving East Slavic texts (cf. Safarewiczowa 1964), its “period of virulence” for becoming established can therefore be estimated at nearly 500 years. One would think that this should suffice for a calque to arise under Finno-Ugric substrate conditions, for which ample prerequisites had existed in Russian since that time (cf. Veenker 1967: 17–21, among others). This would be a further reason to temper Panzer’s and Vasilev’s skepticism. For the onset of the dissolution of Proto-Slavic (ca. 500 AD) more or less coincides with the beginning of the Slavic settlement movement to the northeast.

Observations that ‘u + GEN + ESSE’ occurred practically throughout the entire Slavia in the historically documented period are then of little evidential value. Vasilev himself (in addition to Safarewiczowa 1964 on Common Slavic and Old Russian) points to the fluctuation in the frequency of use of this construction and the HABERE construction. The fact that it remained in the norm over the long term only in Russian is therefore most plausibly attributable to the continued substrate influence of Finno-Ugric speakers. The increased use of this construction in possessive (and semantically related) functions in more recent times is also readily explained by contact with varieties that already used it exclusively or predominantly in this way. This is shown, among other things, by the markedly more frequent occurrence of u + GEN in a possessive function in northern ‘polszczyzna kresowa’ (versus Standard Polish), as well as in colloquial Lithuanian and those Lithuanian dialects that have been in close contact with East Slavic varieties.

In general, these considerations are supported by a comparison with Baltic. Here too, a HABERE verb may be counted among the inherited vocabulary: Lithuanian turėti is to be understood as a pan-Baltic ablaut formation from the root *tuer (cf. its etymological relationship with tverti ‘grasp, seize’; Fraenkel 1965: 1142f.; Pokorny 1959: 1101). As is well known, Latvian no longer has a HABERE verb; the verb turēt, cognate with Lithuanian turėti, has acquired the meaning ‘hold’ and is used primarily with this concrete meaning. We may therefore assume that a HABERE verb also played a part in the formation of basic sentence types in Baltic, and thus determined alignment (as it still does in Lithuanian today). It is also reasonable to assume that Latvian has had substantially more extensive language contacts with Finnic languages and has in many respects been shaped much more strongly by a corresponding substrate than Lithuanian. The displacement of the HABERE verb in Latvian could therefore be viewed as analogous to an identical process in Russian—and this makes more than good sense, both from an areal-linguistic perspective and in light of the known facts of settlement history.

For the time being, all one can say is that this question still needs to be examined closely in its details. The preceding discussion of the ESSE-HABERE problem, as it were “from a bird’s-eye view,” shows only the following: if one wishes to speak of “archaic syntax,” one should first clarify how the relationship of competition between the constructions in question stood in relation to one another—

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 105

which must have been regarded as their morphosyntactic constructions (marking procedures, etc.) and what their chronological relationship to one another was. It could well turn out that a coding pattern that seems archaic from today’s perspective did not exist uninterrupted and unchallenged since the time of a (however posited) proto-language, but could at times have been pushed back, only to be reactivated later as the primary procedure. Language contact (especially substrate influence) would then be the most likely trigger for such a revival, so one should not stop at system-internal attempts at explanation.

4. A FEW CONCLUDING POSTULATES

In the discussion above of case studies involving central morphosyntactic coding procedures and their presumed diachronic changes, ultimately little can be proved; at most, it can be made plausible. The plausibility of explanatory approaches will nevertheless increase to the extent that one can draw on arguments from areas extending beyond system-internal reconstruction. System-internal argumentation may be regarded as a necessary condition for explanation, but it is not a sufficient one—especially when alternative approaches exist. This is all that this article set out to demonstrate. Examining the arguments underlying the theoretical approaches of Stepanov and Schmalstieg was intended, as a representative example of many other cases, to show how essential it is to compare historical-comparative reconstruction with typological facts and findings from contact linguistics, and what may remain of this reconstruction when no effort is made to undertake such a comparison.

Thus, although no authoritative judgments are to be offered here about the Indo-European prehistory of historically documented Baltic (as well as Slavic), a number of conclusions and postulates can nevertheless be formulated concerning it:

1. Speculative assumptions about sweeping changes in alignment are far from convincing; they lead to internal inconsistencies and, for the most part, even contradict typological comparative data. In any case, they leave proponents of the theory of ergative syntax hard-pressed to explain their position.

2. Especially with primarily syntactic problems, taking areal-linguistic factors into account (language contact, substrate formation) advances the explanation, making it substantially more plausible and simpler. It also allows the proposed developments to be assessed under the conditions of real time and actual communication, and on the assumption that variation among functionally equivalent means already existed. Finally, it enables a better assessment of what is archaic and what, by contrast, is innovative. Here, form (etymological “substance”) must be distinguished precisely from functions: it is apparently the latter in particular that can be changed through language contact. In particular, latent functions or those that had previously been prominent can be (re)activated through language contact.

106 |

BJÖRN WIEMER

All this calls for more precise information about the periods under investigation and their chronological sequences.

3. The reconstruction of prehistoric changes in verb roots and stems to explain the emergence of actional and/or diathesis-related distinctions may be essentially correct. However, it is only very loosely connected with the emergence of the modern Slavic aspect system, which arose considerably later. Nor does this reconstruction explain how the present-day aspect system in Russian, Belarusian, or Polish may have supported the system of stem derivation in modern Lithuanian in the formation of aspectual pairs. For these conceivable recent influences can no longer be based on the productive stem-formation processes that are the object of Stepanov’s reconstruction.

4. In assessing the consequences of reconstructed (or, in any case, very remote) processes in stem formation (and morphology in general), it seems essential to allow for the possibility that newer “layers” of productive morphological processes can be superimposed on older ones without the results of the older processes disappearing from the lexicon. This applies par excellence to the opposition between causative and decausative (transitive and intransitive) stems in Lithuanian (see Appendix).

5. The same applies to the occurrence of categorical clusters (actional + diathesis-related features). A further distinction should be made between diffuse clusters with underdetermined categories, which can be given clearer contours, and clusters with already distinct categories, which yield syncretisms that can then be discriminated. The most prominent example from the areas considered here is the association between causativity (diathesis-related) and frequentativity (actional) in the use of the suffix *t.

6. In discussing actional and diathesis-related distinctions in the verb stem, inadequate comparisons between prehistoric processes and developments of more recent times (between which vast stretches of time intervene) can also be avoided by differentiating more precisely between morphological processes along a continuum between grammatical and lexical oppositions.

5. APPENDIX: MORPHOLOGICAL TYPES OF THE TRANSITIVITY CORRELATION IN LITHUANIAN (ACCORDING TO THE CHRONOLOGY OF THEIR EMERGENCE IN BALTIC)

(based on Stepanov 1976: 410; 1989: 170–173)

I. without marking:

Example⁴⁰: degti ⇒ ∅ 1. ‘burn_INTR’, 2. ‘burn up_TR’ dega, degė

⁴⁰Following the infinitive form, the next line gives the 3rd person in the present and preterite, respectively.

Rekonstruktion Grammatischer Kernbereiche | 107

II. suffixal lengthening of a stem with a short vowel:

karti ‘hang.TR’ ⇒ karoti ‘hang.INTR’

karia, korē

CAUSATIVE ⇒ INTRANSITIVE

III. root ablaut + nasal infix and/or sta-suffix vs. change in the inflectional class of the tense: 2 complementary types:

III. a. transitive stems: present in -ia, preterite in -ē

berti ‘pour out’

beria, berē

III. b. intransitive stems: present in -a, preterite in -o

birti ‘be scattered; crumble’

byra (< *bi-n-r-a), biro

DIRECTION OF DERIVATION UNCLEAR

IV. Root ablaut + suffix:

lipti ‘to climb, go’

lipa, lipo

⇒ laipinti ‘to help someone go out/down’

laipina, laipino

INTRANSITIVE ⇒ CAUSATIVE (in the broad sense)

V. unchanged root, but change to the stem at its “edges”: 3 types:

V. a. with RM

ištempi ‘to stretch’ ⇒ išsitempti ‘to stretch oneself’

iš(si)tempia, iš(si)tempē

TRANSITIVE ⇒ INTRANSITIVE

V. b. with suffix -(d)in- (see IV.)

valgyti ‘to eat’ ⇒

valgydinti ‘to feed’

valgo, valgē

valgydina, valgydino

TRANSITIVE ⇒ CAUSATIVE

bėgti ‘to run’ ⇒

bėginti

‘to compel to run, drive on’

bėga, bėgo

bėgina, bėgino

INTRANSITIVE ⇒ CAUSATIVE (in the broad sense)

108 |

BJÖRN WIEMER

V.c. with suffix -dy⁴¹

bed

gulti ‘to lie down’ ⇒ guldÿti ‘to lay someone down, to cause to lie down’

gula, gulē INTRANSITIVE ⇒ CAUSATIVE guldo, guldē

• various combinations are possible!; e.g.

1. mirkī ‘to become damp, wet’ ⇒ mirkyti ‘to moisten’

mirksta, mirko ⇒ mirko, mirkē → Types II + III + V.c. IN BOTH DIRECTIONS

2a. pykti ‘to be angry, to rage’ ⇒ pykinti ‘to anger someone’

pyksta, pyko ⇒ pykina, pykino

2b. pigti ‘to become cheaper’ ⇒ piginti ‘to make cheaper’

pinga, pigo ⇒ pigina, pigino → Types III + V.b. IN BOTH DIRECTIONS

3. lipti ‘to stick (fast).INTR’ ⇒ lipdyti ‘to stick/glue on or fasten.TR’

limpa, lipo ⇒ lipdo, lipdē → Types III + V.c. IN BOTH DIRECTIONS

4. susipažinti⁴² ‘to get to know one another’ ⇒ supažindinti ‘to introduce Y and Z to one another’

susipažįsta, susipažino ⇒ supažindina, supažindino → Types V.a. + V.b. IN BOTH DIRECTIONS

ABBREVIATIONS

ACC – accusative aks. – Old Church Slavonic altlit. – Old Lithuanian AOR – aorist DAT – dative F – feminine GEN – genitive

⁴¹Etymologically practically identical to -in-, since it arose through lengthening of ī after the loss of the nasal. The -d- was originally a hiatus breaker (cf. Stang 1943: 142).

⁴² Strictly speaking, susi- is a single morpheme. This can, however, be disregarded in this context.

REKONSTRUKTION GRAMMATISCHER KERNBEREICHE | 109

IE – Indo-European

INF – infinitive

INTR – intransitive

ipf. – imperfective

lit. – Lithuanian

M – masculine

NOM – nominative

pf. – perfective

PL – plural

poln. – Polish

PRS – present

PRT – preterite

PTC – participle

Russ. – Russian

SG – singular

TR – transitive

Proto-Slavic

1, 3 – 1st, 3rd person

CITED LITERATURE

AMBRAZAS, V. 1990: Sravnitel'nyj sintaksis pričastij baltijskich jazykov, Vilnius: Mokslas.

— 1994: On the interpretation of Lithuanian constructions with neuter passive participles. Linguistica Baltica 3, 7–11.

— 2004: On the Genitive with Neuter Participles and Verbal Nouns in Lithuanian. Baldi, Ph., Dini, P. U., eds., Studies in Baltic and Indo-European Linguistics (in honor of William R. Schmalstieg). Amsterdam, Philadelphia: Benjamins, 1–6.

BARTON, Ch. R. 1981: Notes on the Baltic Preterite. Indogermanische Forschungen 85, 246–278.

BAUER, B. 2000: Archaic Syntax in Indo-European (The Spread of Transitivity in Latin and French), Berlin, New York: Mouton de Gruyter.

BERMEL, N. 1997: Context and the Lexicon in the Development of Russian Aspect, Berkeley-Los Angeles-London: University of California Press.

BIRNBAUM, H. 1975: Common Slavic (Progress and Problems in its Reconstruction), Cambridge, MA: Slavica Publishers, Inc.

BREU, W. 1998: Komplexe aktionale Verbklassen, insbesondere Inchoativa. In: Berger, T., Raecke, J., Hgg., Slavistische Linguistik 1997. München: Sagner, 55–80.

BYBEE, J., DAHL, Ö. 1989: The creation of tense and aspect systems in the languages of the world. Studies in Language 13 (1/2), 51–103.

CHRISTEN, S. 1998: Unpersönliche Konstruktionen und sekundäre Personifizierung. Linguistica Baltica 7, 51–61.

COMRIE, B. 1975: The Antiergative: Finland’s Answer to Basque. In: Grossman, R. E. et al. (eds.): Papers from the eleventh regional meeting, Chicago Linguistic Society 11. Chicago, 112–121.

COMRIE, B. 2001: Typology and the history of language. In: Bisang, W., ed., Aspects of Typology and Universals. Berlin: Akademie, 21–35.

DIXON, R. M. W. 1994: Ergativität, Cambridge: University Press.

110 |

BJÖRN WIEMER

ESTIVAL, D., MYHILL, J. 1988: Formal and functional aspects of the development from passive to ergative systems. In: Shibatani, M., ed., Passive and Voice. Amsterdam, Philadelphia: Benjamins, 441–491.

FRAENKEL, E. 1965: Litauisches etymologisches Wörterbuch, Bd. 2, Heidelberg / Göttingen: Winter / Vandenhoeck & Ruprecht.

GALKINA-FEDORUK, E. M. 1958: Bezličnye predloženi̯ja v sovremennom russkom jazyke, Moskva: Izd-vo MGU.

HASPELMATH, M. 1987: Transitivity alternations of the anticausative type, Köln: Universität zu Köln. (Institut für Sprachwissenschaft. Arbeitspapier Nr. 5, Neue Folge.)

ISAČENKO, A.V. 1974: On ‘Have’ and ‘Be’ Languages (A Typological Sketch). Flier, M. S., ed., Slavic Forum (Essays in linguistics and literature). The Hague etc.: Mouton, 43–77.

KARDELIS, V., WIEMER, B. 2002: Ausbildung von Aspektpaarigkeit in litauischen Grenz- und Inseldialekten (am Beispiel von Sprechverben). Linguistica Baltica 10, 51–80.

KAZLAUSKAS, J. 1968: Lietuvių kalbos istorinė gramatika (kirčiavimas, daiktavardis, veiksmažodis), Vilnius: Mintis.

KLAIMAN, M. H. 1991: Grammatical voice, Cambridge etc.: University Press.

KRASUCHIN, K. G. 1990: Nekotorye problemy rekonstrukcii indoevropejskogo sintaksisa (v svjazi s vychodom knigi Ju. S. Stepanova «Indoevropejskoe predloženie». M., 1989). Voprosy jazykoznanija 6, 71–84.

LEHMANN, V. 1990: Zur ontogenetischen und diachronen Entwicklung grammatischer Kategorien des Polnischen. Breu, W., Hg., Slavistische Linguistik 1989. München: Sagner, 167–188.

LEHMANN, V. 1996: Die Rekonstruktion von Bedeutungsentwicklung und -motiviertheit mit Funktionalen Operationen. Girke, W., Hg., Slavistische Linguistik 1995. München: Sagner, 255–289.

LEHMANN, V. 1999: Sprachliche Entwicklung als Expansion und Reduktion. In: Anstatt, T., Hg., Entwicklungen in slavischen Sprachen. München: Sagner, 169–254.

MARVAN, J. 1973: Baltic and Indo-European Ergative (Based on Professor Pavel Trost’s pioneer work). Lituanus 19, 31–38.

MATASOVIC, R. 2000: Uses and misuses of typology in Indo-European linguistics. Ofitsch, M., Zinko, C., Hgg., 125 Jahre Indogermanistik in Graz. Graz: Leykam, 327–340.

MATTHEWS, W. K. 1955: Lithuanian Constructions with Neuter Passive Participles. Slavonic and East European Review 33, 350–371.

MICHELINI, G. 1990: Die indogermanische Vorlage der -ā- und -ē-Präterita in den baltischen Sprachen. Zeitschrift für Slavistik 35 (6), 841–847.

PADUČEVA, E. V., ROZINA, R. I. 1993: Semantičeskij klass glagolov polnogo ochvata: tolkovanie i leksiko-sintaksičeskie svojstva. Voprosy jazykoznanija 6, 5–16.

PALMAITIS, M. L. 1977: Del baltų kalbų nenominatyvinės praeities. Baltistica (2 priedas), 114–123.

PANZER, B. 1982: Genetische und stratalinguistische Faktoren in der Entstehung der ostslavischen Schriftsprachen. Ureland, P. St., Hg., Die Leistung der Strataforschung und der Kreolistik (Typologische Aspekte der Sprachkontakte: Akten des 5. Symposions über Sprachkontakt in Europa, Mannheim 1982). Tübingen: Niemeyer, 321–337.

PLANK, F. 1995: Ergativity. Jacobs, J. et al., Hgg., Syntax: Ein internationales Handbuch zeitgenössischer Forschung, 2. Halbband. Berlin–New York: De Gruyter, 1184–1199.

POKORNY, J. 1959: Indogermanisches etymologisches Wörterbuch, Bd. 1, Berlin–München: Francke.

ROSÉN, H. B. 1979: Amamini und die indogermanischen Diathesen- und Valenzkategorien. Zeitschrift für vergleichende Sprachforschung 92, 143–178.

SADNIK, L. 1962: Die Nasalpräsentia und das frühurslavische Verbalsystem. Die Sprache 8, 238–249.

SAFAREWICZOWA, H. 1964: Oboczność я имею i у меня есть w języku rosyjskim dziś i dawniej, Wrocław

Rekonstruktion grammatischer Kernbereiche | 111

etc.: Ossolineum.

Schmalstieg, W. R. 1982: The shift of intransitive to transitive passive in the Lithuanian and Indo-European verb. Baltistica 17 (2), 119–134.

Schmalstieg, W. R. 1984: The genitive with verbs denoting ‘to fill’: partitive or ergative? Baltistica 20 (2), 136–140.

Schmalstieg, W. R. 1988: A Lithuanian Historical Syntax, Columbus, Ohio: Slavica Publishers.

Schmalstieg, W. R. 1993: Tevas dirbta?! — A misinterpretation. Linguistica Baltica 2, 39–45.

Schmalstieg, W. R. 1994: Again the Baltic genitive of agent. Linguistica Baltica 3, 13–19.

Schmid, W. P. 1963: Studien zum baltischen und indogermanischen Verbum, Wiesbaden: Harrassowitz.

Schmid, W. P. 1966: Baltische Beiträge IV: Zur Bildung des litauischen Präteritums. Indogermanische Forschungen 71, 286–296.

Schmidt, K. H. 1980: Zur Typologie des Vorindogermanischen. Ramat, P. et al., eds., Linguistic Reconstruction and Indo-European Syntax (Proceedings of the Colloquium of the Indogermanische Gesellschaft’, University of Pavia, 6–7 September 1979). Amsterdam: Benjamins, 91–112.

Schmidt, K. H. 1983: Kaukasische Typologie als Hilfsmittel für die Rekonstruktion des Vorindogermanischen, Innsbruck. (Innsbrucker Beiträge zur Sprachwissenschaft. Vorträge und Kleinere Schriften, 31.)

Stang, Chr. S. 1943: Das slavische und baltische Verbum, Oslo: Jacob Dybwad. (Skrifter utgitt av det Norske Videnskaps-Akademi i Oslo II. Historisk-Filosofisk Klasse, 1.)

Stang, Chr. S. 1966: Vergleichende Grammatik der Baltischen Sprachen, Oslo–Bergen– Tromsö: Universitetsforlaget.

Stepanov, Ju. S. 1976: Vid, zalog, perechodnost’ (Balto-slavjanskaja problema I). Izvestija AN SSSR, serija literatury i jazyka 35 (5), 408–420.

Stepanov, Ju. S. 1977: Vid, zalog, perechodnost’ (Balto-slavjanskaja problema II). Izvestija AN SSSR, serija literatury i jazyka 36 (2), 135–152.

Stepanov, Ju. S. 1978a: Slavjanskij glagol’nyj vid i baltijskaja diateza (problema obščego genezisa i rekonstrukcii). Borkovskij, V. I. et al., red., Slavjanskoe jazykoznanie. VIII meždunarodnyj s”ezd slavistov, Zagreb–Ljubljana, sentjabr’ 1978 g. Doklady sovetskoj delegacii. Moskva: Nauka, 335–363.

Stepanov, Ju. S. 1978b: «Perfektnyj centr» v russkom sintaksise (Vid, zalog, perechodnost’. Balto-slavjanskaja problema IV). Izvestija AN SSSR, serija literatury i jazyka 37 (4), 344–353.

Stepanov, Ju. S. 1989: Indoevropejskoe predloženie, Moskva: Nauka.

The Universals Archive: http://ling.uni-konstanz.de/pages/proj/sprachbau.htm.

Vasilev, Chr. 1973: Ist die Konstruktion ‘u menja est’ russisch oder urslavisch? Die Welt der Slaven 18, 361–367.

Vasmer, M. 1953: Russisches etymologisches Wörterbuch, Bd. 1, Heidelberg: Winter.

Veenker, W. 1967: Die Frage des finnugrischen Substrats in der russischen Sprache, Bloomington: Indiana University.

Walde, A. 1930: Vergleichendes Wörterbuch der indogermanischen Sprachen (herausgegeben und bearbeitet von Julius Pokorny), Bd. 1, Berlin–Leipzig: De Gruyter.

Wiemer, B. 2001: Aspektual’nye paradigmy i leksičeskoe značenie russkich i litovskich glagolov (Opyt sopostavlenija s točki zrenija leksikalizacii i grammatikalizacii). Voprosy jazykoznanija 2, 26–58.

— 2003: Objective repetition types, confused functional assignments, and degenerate inferences in synchrony as a consequence of diachrony. Anstatt, T., Hansen, B., eds., Developments in Slavic Languages 2 (For Volkmar Lehmann on his 60th birthday from his students). Munich: Sagner, 35–62.

— 2004: The evolution of passives as grammatical constructions in Northern Slavic and Baltic languages.

112 |

BJÖRN WIEMER

Bisang, W., Himmelmann, N. P., Wiemer, B., eds., What makes Grammaticalization? A Look from its Fringes and its Components. Berlin–New York: Mouton de Gruyter, 271–331.

— in press 1: On distinguishing grammatical and lexical oppositions in verbal word formation, or: what can aspectologists learn from the example of sja-verbs? To appear in: Lehmann, V., ed., Semantika i struktura slavjanskogo vida. Munich: Sagner.

— in press 2: On the diachronic and areal background of constructions called ‘impersonal’ in Polish, East Slavic and Baltic. To appear in: Manninen, S., Hiiетam, K., Kaiser, E., Vihman, V., eds., Passives and Impersonals in European Languages. Amsterdam, Philadelphia: Benjamins.

Björn Wiemer

Received 2004 11 19

Universität Konstanz

Department of Linguistics

Universitätsstrasse 10, 78457 Konstanz

[email protected]