Lithuanian “būti” with the infinitive as a modal expression and its Latvian counterparts

ACTA LINGUISTICA LITHUANICA XLIV (2001), 71–87 Lithuanian būti with the infinitive as a modal expression and its Latvian counterparts AXEL HOLVOET Lietuvių kalbos institutas, Vilnius In Baltic, būti + infinitive is now used as a modal construction expressing necessity. It developed out of an existential construction with būti, expanded with an infinitive of purpose (namas yra statyti ‘there is a house to be built’ → ‘a house can/should be built’). This construction, originally vague between possibility and necessity, subsequently became specialised in the meaning of necessity. The original nominative with būti was reanalysed as an object and replaced with the accusative. Whereas the type expressing necessity is productive in both Lithuanian and Latvian, only Lithuanian has retained the original construction with a nominative as an expression of possibility (man namai matyti ‘I can see the house’). Latvian counterparts to this construction are sometimes cited in the literature, but they represent a different syntactic type which is an innovation of Latvian: it is characterised by a nominative instead of the original dative for the agent/modal subject and a different status of the negation: zirgam vezumu nepavilkt → zirgs vezumu ne pavilkt. 1. THE CONSTRUCTION būti + INFINITIVE IN THE LITERATURE The use of Baltic būti and Slavonic byti with the infinitive as an expression of possibility or necessity is well known. In the grammars as well as in monographs on historical and synchronic syntax it is sometimes referred to as it is done here (i.e., as a construction involving the verb ‘be’), e.g., Miklosich (1926: 859) writes ‘das subjectlose verbum substantivum bezeichnet in verbindung mit einem infinitiv die notwendigkeit oder möglichkeit und negativ die physische oder moralische unmöglichkeit einer handlung’. A similar account can be found in Brugmann (1916: 923–926). In other cases, the lack of an overtly expressed verb in the present tense (estb, yra, ir etc.) leads the authors dealing with the type under discussion to refer to it as an instance of the predicative infinitive (for Latvian, cf. Endzelin 1951: 991; for Lithuanian, cf. Ulvydas, ed., 1976: 323)1. If, alongside a construction 1 In some (mainly Russian) works, the basic structural identity of constructions with and without overt forms of the verb ‘be’ is not disregarded, but both are described as instances of an ‘independent infinitive’ (cf. Borkovskij, ed., 1978: 278–283). Such a treatment is apparently based on the assumption that the past and future tense forms of the verb byti used with the infinitive are not instances of an autonomous verb but mere tense markers. 72 | AXEL HOLVOET tion without an overtly expressed present tense of ‘be’, there is a comparable construction with an overt tense form (say, Lithuanian Man buvo eiti ‘I had to go/was to have gone’ alongside Man važiuoti ‘I must/should go’), then it is clear that the former must contain a zero realisation of the 3rd person form of the present tense of ‘be’ (this zero realisation is a feature well known from both living Baltic languages as well as from several Slavonic languages), and that we are dealing with varieties of one single construction rather than with two different constructions. However, not all tense forms are used with the same regularity (for example, the construction may be virtually restricted to the present tense, i.e., the variety without overt form of ‘be’); furthermore, additional functional differences may appear that may obscure the fundamental identity between those constructions that contain some form of the verb ‘be’ and those that do not2. For Lithuanian, both Palionis (1972) and Ambrazas (1995) conclude that there is no reason to set the two types apart. In other words, man važiuoti should be interpreted structurally as man (yra) važiuoti (for Old Russian this was emphasised already by Potebnja 1958: 394). Another reason why particular varieties of the type discussed here are sometimes set apart as independent constructions is that there occur objects in different case forms. In Lithuanian (vestigially in Latvian) and in some Slavonic dialects we find a nominative object alongside an accusative object with the infinitive: man namai statyti alongside man namus statyti ‘I have to build a house/should build a house’ (the examples used to represent these models are cited after Ambrazas 1995). If the nominative is interpreted as a subject and the accusative as an object (which is, in a way, accurate but means putting phenomena from different stages of language development on one level, as the reanalysis of an original subject as object is involved here, v. infra), then there is no way of accounting for both types in a similar way. This led Fraenkel (1928: 118–119) to say that the infinitive ataiti in kaip bua iam ataiti is the subject of būti whereas in iumus bus rustas sudas pas Diewa kelti the infinitive expresses a predicate, the subject of which is sudas (Fraenkel 1928: 14; both examples from the Wolfenbüttel Postil, without precise location). The interpretation of the infinitive as the subject of būti (suggested by Fraenkel for kaip bua iam ataiti) cannot be an accurate description of the oldest stage of the construction dealt with here because the Baltic and Slavonic infinitive was originally the dative of a verbal noun in -ti- and its original scope was therefore restricted to what is now described as the infinitive of purpose. In reconstructing the oldest stages of the Baltic infinitival constructions one should always be aware of the relatively recent origin of the infinitive as a verbal form and of the range of possible syntactic functions the infinitive could initially perform by virtue of its morphological nature, viz. that of the dative of a verbal noun. Wherever an infinitive functions as a subject in the sense of occupying the place of a nominative subject (as in Turėti daug dėdžių [...] nemažą garbės ir malonumo ‘to have many uncles ... is no small honour and pleasure’ – Vaižgantas, cited in Ulvydas, ed., 1976: 292), this state of affairs is a relatively recent development. Ambrazas (1995) distinguishes, within the syntactic type under consideration here, two subtypes: (man) yra / Ø / buvo / bus namai matyti and (man) yra / Ø / buvo / bus namai 2 As an instance, one could cite the functional differentiation between constructions with a zero form of the verb ‘be’ (present tense forms) and those with the future būs in Latvian (v. infra). LITHUANIAN b
6bti WITH THE INFINITIVE 73 statyti. Bearing in mind the origin of the Baltic infinitive, he analyses the construction (man) yra /  / buvo / bus namai matyti as an expanded variety of a construction with a substantive verb and a nominative subject. The nominative subject subsequently underwent a reanalysis and began to function as the object of the infinitive. This is one of the sources of the nominative object characteristic of infinitival constructions in Baltic. The replacement of the nominative object with an accusative object gave rise to the newer variety (man) yra /  / buvo / bus namus statyti. Alongside the above-mentioned models with transitive verb, Ambrazas discusses a model with intransitive verbs: (man) yra /  / buvo / bus eiti. This model is undoubtedly of more recent origin, and it must have developed on the basis of the transitive model after the reanalysis had taken place. Ambrazas’ account of this process as an instance of reanalysis (basically reflecting that of Brugmann 1916: 925) is quite convincing, though alternative explanations have been offered as well: Timberlake (1974) explains the nominative object as an areal convergence linking Finnic, Baltic and North Russian. As Ambrazas has pointed out, these two explanations are not necessarily contradictory. Ambrazas’ reconstruction is the most plausible explanation for this expression of necessity anyway, as a construction containing the verb ‘be’ combined with an infinitive of purpose or some equivalent construction expressing purpose is a common means of expressing possibility and/or necessity, cf. English what is to be done, German die Frage ist nicht zu lf6sen etc.). We may assume that such constructions originate as existential constructions and that they are subsequently reanalysed as modal, i.e., the original model is yra namai statyti / matyti, which would mean something like ‘there is a house to see/build’. In other words, namai was originally the subject of an existential construction. Timberlake’s hypothesis is therefore required not so much for explaining the origin of the use of the nominative as for explaining the retention of the nominative after the construction had been reanalysed as a modal expression, i.e., its retention at a stage when the nominative subject had become a nominative object. And this is precisely the role ascribed to the Fennic substratum/adstratum by Ambrazas (1997: 99). 2. THE PROBLEM OF b
6bti AS A MODAL VERB If the change in the status of the nominative in namai yra statyti (i.e. its reanalysis as a nominative object and its eventual replacement with an accusative object) is clear, no attention has been paid until now in the literature to the changes in the status of the verb ‘be’. Originally, it was a substantive verb expressing an existential predicate. Before the reanalysis of the nominative subject as an object there must have been a certain period when the original syntactic structure of the clause was retained but a shift had already occurred in semantic structure, the whole construction having acquired the meaning of an expression of possibility or necessity. A similar situation can be observed in the case of other constructions that have developed or are developing a modal meaning. So, for instance, a modal meaning (that of lack of necessity) has developed in Baltic and Slavic constructions where an infinitival relative clause is adjoined to the implicit subject of an existential construction. Lithuanian n
1ra ko skaityti, Latvian nav ko las
2t means ‘there is nothing to read’, but also ‘there is no point in reading (anything, or some particular text)’. 74 | AXEL HOLVOET The latter reading leads to the use of this construction in modal function, in which case the relative pronoun loses its grammatical relation as object of the infinitive skaiytyti, lasīt and the infinitive may take an object of its own: Nēra ko skaiytyti tokių knygų, Nav ko lasīt tādas grāmatas3. The introduction of an object (distinct from the relative pronoun) with the infinitive is an instance of the actualisation following upon the modal reanalysis of the construction. Those instances where no new object is introduced though the infinitive is transitive (partly as a result of ellipsis, e.g., nėra ko skaiytyti ‘there is nothing to read or ‘there is no point in reading [it]’, viz. the book you just mentioned) are now syntactically ambiguous, but there must evidently have been an intermediate stage, prior to syntactic changes, in which only a shift in meaning had occurred. A similar situation may be reconstructed for namai statyti ‘there is a house to build’, ‘a house must be built’. This new modal meaning led, in its turn, to a syntactic reanalysis in such a way that yra started functioning as an expression of modality, i.e., as representing a modal predicate, the argument of which was represented by the embedded structure [namai statyti]; this, in turn, led to the reanalysis of namai as a nominative object and to its eventual replacement with the accusative. The process is exactly parallel to that involving the verb turėti ‘have’. We may assume the construction turiu duonos valgyti to have originally contained a dative of purpose as well: ‘I have some bread to eat’ → ‘I have to eat bread’. In fact, the original meaning (presupposing the segmentation [(turiu duonos) valgyti])) still exists alongside the new one ([turiu [duonos valgyti]]). Nowadays the meaning of a modal verb (‘have to, must, should’) is ascribed to turėti as a separate meaning alongside the basic possessive one, and in view of the basic semantic similarity of ‘habeo’ and ‘mihi est’ (which should be described as syntactic converses, as emphasised by Benveniste 1966) it should be possible to ascribe a similar modal meaning to ‘be’ as well. The interpretation of būti as a modal verb calls, of course, for a further explanation. The verb turėti corresponds to our notions about modal verbs because it displays some features characteristic of what is described as the AUX component of generative grammar, viz. it stands, as it were, apart from the syntactic structure of the sentence and does not influence it. In this sense, it is similar to a tense and mood marker, and indeed tense and mood are often expressed by auxiliaries as well. Thus, we may speak of a modal auxiliary here. Obviously the case of būti is different because it seems to influence syntactic structure. However, the requirements and conventions connected with the generative mode of description are not relevant here. A modal verb typically represents a modal predicate with one propositional argument, the syntactic representation of which is an embedded clause. I will assume here that such a predicate can influence the assignment of case forms. It should be noted that such an assumption is necessary if we are to account for the syntactic properties of the Latvian debitive: this is a form the function of which is purely modal, i.e., it expresses a modal predicate. It changes, however, the assignment of case forms. The system of modal predicates (as distinct from sentence modality) is based on the central meanings of necessity and possibility. Modal verbs proper tend to be specialised in 3 Cf. also Russian нечего читать такие книги, where the negated substantive verb нет has fused with the relative pronoun into what is now considered, in this particular use, to be a predicative expression with the meaning ‘there is no point (in doing something)’. LITHUANIAN būti WITH THE INFINITIVE 75 terms of either necessity or possibility. These two modal meanings are, of course, interrelated in several ways through the action of the negation. In particular, the following equivalences are to be observed: ‘□¬p’ = ‘¬◇p’ (‘it is necessary that not¬p’ = ‘it is impossible that p’) ‘¬□¬p’ = ‘◇p’ (‘it is not necessary that not¬p’ = ‘it is possible that p’) Though languages tend to show various restrictions of a formal nature as to the introduction of a negation in constructions with modal verbs, there is, in principle, a possibility of adding a negation either to the modal verb (EXTERNAL NEGATION) or to the infinitive (INTERNAL NEGATION), as illustrated by the following Lithuanian examples: jis negali ateiti (‘¬◇p’) ‘it is impossible for him to come’ ‘/it is impossible that he will come’ jis gali neateiti (‘◇¬p’) ‘it is possible for him not to come’ ‘/it is possible that he will not come’ jis negali neateiti (‘¬◇¬p’) ‘it is impossible for him not to come’ ‘/it is impossible that he will not come’ With expressions of necessity, however, the negation may often be internalised, i.e., it may formally be attached to the modal verb while being semantically interpreted as belonging to the infinitive, e.g. the structural meaning of jis neturi ateiti is ‘¬□¬p’, i.e., ‘it is not necessary for him to come’ = ‘he need not come’, but it can as well mean ‘□¬p’, i.e., ‘it is necessary for him not to come’, ‘he should not come’. In addition to the modal verbs briefly characterised here we may note a series of alternative means of expressing necessity and possibility. In Latin, for instance, the meaning of necessity is often expressed by means of a construction with the predicative gerundive: Karthago delenda est ‘Carthage should be destroyed’. A structural and functional counterpart of the Latin construction with the predicative gerundive is the Latvian construction with the present passive participle: Siens ir pļaujams ‘the hay must be mown’. The Latvian construction is interesting in that it may serve as an example of an expression that is vague between necessity and possibility. A participle or, to be more precise (as the purely participial use of this form is but rarely observed) verbal adjective in -ms marks a general predisposition of an object to undergo a certain type of action, but the precise interpretation of this predisposition in terms of possibility (‘that can be mown’) or necessity (‘that must be mown’) is context-dependent. This indeterminacy subsists in those cases where the verbal adjective is used in a copular construction, and Siens ir pļaujams may also have both meanings, that of possibility and that of necessity. In the case of constructions with the verb ‘be’ and an infinitive of purpose, a similar modal vagueness is attested from several languages, cf. German Die Frage ist zu lösen ‘The problem can be solved’ or ‘The problem must be solved’. The specialisation of such constructions as expressions of necessity is probably associated with their increasing grammaticalisation. It is probably in this way that one should explain the specialisation of the Latin construction with the predicative gerundive as a means of expressing necessity. The Latvian construction with present 76 | AXEL HOLVOET passive participles seem to be undergoing a similar process. In those cases where the construction with a passive participle is copular, it is also vague between possibility and necessity, but when it ceases to be copular and begins to function as a MODAL CONSTRUCTION proper, only the necessity reading is retained. This is the case when a construction of this type is derived from an intransitive verb, so that there is no nominative subject and the construction cannot therefore be interpreted as copular any more: Ikviens jau zināja, kurā baznīcā viņam ejams (P. Rozītis) ‘But everyone of them know to which church he was to go’. It is possible to find more examples of modal constructions that are originally vague between necessity and possibility and subsequently undergo specialisation4. The constructions under discussion in this article are one of them. 3. TWO STAGES IN THE GRAMMATICALISATION OF būti + INFINITIVE AS A MODAL CONSTRUCTION The Baltic constructions with būti and the infinitive seem to confirm the notion of an originally undifferentiated construction, susceptible of a twofold modal interpretation (possibility or necessity) developing towards a modally unambiguous construction expressing only necessity. Instances of the meaning of possibility should be interpreted as residual and improductive. I will start my exposition with Latvian, where the constructions discussed here are more frequent than in Lithuanian (though they also show traces of some new, specifically Latvian developments, to be discussed below). In Lithuanian, the constructions with overt forms of būti are now exceptional; they are well attested in Bretkūnas’ writings (cf. Palionis 1972: 125–126). In the Latvian Academy Grammar (1959: 622), the ‘related meanings’ of inevitability and impossibility are said to be expressed by the predicative infinitive. The following examples are cited: (1) Ātrāk par nākamo pavasari ... naudu neredzēt. ‘Before next spring there is no hope of seeing any money’. (2) Jums manu ilgu mūžam nesaprast. (Rainis) ‘You will never understand my longing’. It is clear that we are not dealing with two related but distinct constructions here, as the formulation of the Academy Grammar seems to suggest. Two different formulations are, of course, possible, as the impossibility of p may also be formulated as the inevitability of ¬p. Now if we assume that the constructions cited here are instances of būt with the infinitive rather than of independent predicative infinitives, then the occurrence of the negation with the infinitive suggests that we are dealing with an internal negation, and that ‘¬◊p’ (‘it is impossible that p’) is expressed as ‘□¬p’ (‘it is necessary that not¬p’), and the Of course, this does not mean that such modally vague constructions must necessarily undergo such a specialisation, or that they must necessarily develop into expressions of necessity rather than possibility. The German ‘modal infinitive’ (ist zu lösen), for instance, shows no signs of undergoing an evolution in either direction: it appears to be quite stable as an expression of either possibility or necessity (cf. Van der Auwera & Plungian 1998: 100–103). LITHUANIAN būti WITH THE INFINITIVE | 77 meaning of the verb būti in [BŪT] + NEG-INF is thus that of necessity, not possibility. (1) and (2) thus represent the same construction expressing necessity as (3): (3) Ai vecais bāleliņ, Tev manam tēvam būt, Tev manim zirgu pirkt, Tev atvest līgaviņu. ‘Oh, my elder brother, you should be a father to me, you should buy me a horse and bring [me] a bride’ (BW 13796) When the negation is added to the modal verb būt, one may theoretically expect two meanings: ‘¬□p’ or, with lowered or internalised negation, ‘□¬p’. Actually only the latter meaning is attested in those cases where the verb is overtly expressed: (4) Nevienam svēšam nebīj redzēt, ka Slaucēja madāma ēda [...] (R. Blaumanis) ‘No stranger was supposed to see Madame Slaucēja eating [...].’ (5) Tev nebūs raudāt, mīļā dvēsele... (Rainis, cited after Bergmane e.a., 1959: 623) ‘You must/should not weep, my dear soul’. Formal attachment of the negation to the infinitive (*tev būs neraudāt) does not seem to be attested, and it is evidently superfluous because the corresponding meaning is expressed by the construction with negation lowering (tev nebūs raudāt). On the other hand, the latter construction does not have any equivalent in the present tense: there is no construction *Tev nav raudāt ‘you should not weep’. In the meaning we should expect such a construction to have (negation lowering being taken into account), the future of būt is always used. Here, evidently, a specialisation of tense forms has taken place in the negated constructions. The use of the future būs in constructions like Tev nebūs zagt ‘Thou shalt not steal’ is not motivated semantically, because in a generalising statement of deontic necessity this modality is not normally viewed as arising in the future while not encompassing the moment of speaking; the future of the modal verb is normally used only in the case of dynamic modality (cf. You should not steal vs You will have to steal). Evidently the negated future has superseded the negated present tense in deontic meaning. In the present tense, only an internal negation is possible. This distribution of deontic and dynamic meanings has evidently spread from the negated constructions to the affirmative ones, so that Tev būs strādāt ‘You should work’ is now also specialised in deontic meaning5. Lithuanian seems to have the same pattern with negation. The negation can be added to būti as a modal verb, the meaning is deontic and receives the interpretation ‘□¬p’ (with negation lowering). The whole construction is not productive any more in modern Lithuanian. The following example has nebuvo ‘one shouldn’t have’: (6) Kaip nuseidau nuo žirgelio, nebuvo vadinti, kaip įėjau į seklyčią, nebuvo sodinti. (Ambrazas 1995: 94) 5 In older Latvian grammars, būs with the infinitive is sometimes described as a future imperative, cf. Andronov (1999: 161–163). 78 | AXEL HOLVOET ‘When I dismounted from my horse, you shouldn’t have invited me to come in, when I entered your house, you shouldn’t have given me a seat’. When the negation is internal, the meaning is dynamic, and ‘□¬p’ is used to convey impossibility (‘¬◇p’). The negation is then added to the infinitive, and the verb ‘be’ has no surface realisation: there is therefore no possibility of tense variation. The construction is described as one with a predicative infinitive in Lithuanian grammar (cf. Ulvydas, ed., 1976: 323): (7) Seserys atšoko: jo (žalčio) nebįsvaryt! (Salomeja Nėris, cited after Ulvydas, ed., 1976: 323) ‘The sisters sprang back: the grass-snake was not to be driven away any more’. (8) Arkliui nepavykti vežimo. ‘The horse is unable to draw the cart’. The impossibility reading can, of course, apply only if the negation is a VERB PHRASE NEGATION, not if its scope does not encompass the verb phrase at all. In (9), for instance, the negation is internal in the sense of not belonging to the modal verb (the unexpressed verb būti), but it has only the adverbial visuomet within its scope, not the verb: (9) Ne visuomet gi ir mums jungą vilkti. (A. Vienuolis, cited after Ulvydas, ed., 1976: 616) ‘And we will not always have to pull the yoke either’. While the meaning of impossibility can be derived from that of necessity in the way described above, possibility obviously cannot. Therefore the modal verb būti(i) ‘must’ cannot be contained in constructions of the type namai matyti, or else this verb must also be ascribed another meaning, viz. possibility. But there is probably no reason to do so. The type namai matyti has its syntactic as well as its semantic peculiarities: the most important of these is the consistent retention of the nominative (Ambrazas 1995: 90). As the nominative represents an older stage in the development of the constructions we are dealing with here, it seems legitimate to assume that the modal meaning of this construction also represents an older stage in the process of its grammaticalisation as a modal expression. Vagueness between possibility and necessity is not frequent in specialised lexical expressions of modality, i.e., in modal verbs. But it does seem to be characteristic of various constructions in the initial stage of their grammaticalisation in modal function. It was stated above that the type yra namai matyti / statyti must have originated as an existential construction: ‘There is a house for building/seeing’. As long as the construction continued to be existential, it was also vague with regard to modal meaning: yra namai statyti could mean ‘there is a house to be built’, whereas yra namai matyti could at the same time be interpreted as ‘there is a house to be seen’, i.e., ‘a house can be seen’. At a later stage, when the modal meaning prevailed and namai statyti was interpreted as an embedded infinitival clause (with a nominative object), this ambiguity was eliminated in favour of the meaning of necessity. As a parallel, we may cite the Latvian constructions with present passive LITHUANIAN būti WITH THE INFINITIVE 79 participles mentioned above: in their original shape (as copular constructions) they are vague between possibility and necessity, but at a more advanced stage of grammaticalisation they become specialised in the meaning of necessity. The type namai matyti is thus a residual one, and this fact is also reflected in the lexical restrictions to which it is subject: the type now seems to be restricted to a small group of verbs of sensory perception6. For Latvian, Endzelin (1951: 992) cites one example that can be compared to the Lithuanian type namai matyti: (10) pazīt man(i) brāļu māsas: pilni pirksti gredzentiņu BW 6259, 4 ‘I can recognise my kinsmen’s sisters: their fingers are full of rings’. This type, like the corresponding Lithuanian construction, is a relic from the functional point of view: the original meaning of possibility is retained in the case of one (or perhaps a few) verbs expressing sensory perception or cognition. In Latvian, the construction has completely disappeared from the modern language, whereas its Lithuanian counterpart, though now subject to lexical restrictions, is still fully alive. Syntactically, the Latvian construction has been assimilated to the productive type expressing necessity, and the nominative has been replaced with the accusative, which cannot be seen in (10) but is revealed by the other variants of BW 6259 (e.g., pazīt bija kunga riju ‘one could recognise the squire’s barn’; pazīt man brāļa māsu ‘I can recognise my kinsman’s sister’) as well as by other contexts (pazīt bija pora egli melnajām skujinām ‘one could recognise the fir-tree growing on marshy soil by its black needles’). This assimilation has not taken place in Lithuanian, where the nominative is retained in the residual type expressing possibility. 4. PECULIAR DEVELOPMENTS IN LATVIAN Example (10) cited above is an exact counterpart of the residual Lithuanian type expressing possibility. But Mühlenbach (Endzelin & Mühlenbach 1928: 194) and Endzelin (1951: 992) cite a whole series of examples allegedly corresponding to the type namai matyti. Among these, only (10) is affirmative and can be plausibly interpreted as expressing possibility. The remaining examples are all negative, so that there is no reason to set them apart from constructions like (1–2), which actually represent the meaning of necessity. The reason why they are mentioned together with the use of pazīt in (10) is that they also contain verbs of perception. On closer scrutiny, however, the Latvian constructions, 6 In Slavonic, certain traces of the restriction of byti + infinitive as an expression of possibility to verbs of sensory perception can be found as well. For Old Russian, cf. Borkovskij, ed. (1978: 280). In Polish, the original infinitives widać ‘one can see’, słychać ‘one can hear’, czuć ‘one can smell’, znać ‘one can notice’ are still used with być ‘be’ (omitted in the present tense), as in widać było ‘one could see’ etc. They show the replacement of the original nominative with the accusative, e.g., słychać muzykę ‘music is heard, can be heard’. Synchronically, however, they are not treated as infinitives any more, partly because the verbs from which they derive (widać, słychać) have fallen into disuse. In contemporary Polish grammar, widać and słychać are described as non-inflecting verbs using być as an auxiliary of tense and mood. 80 | AXEL HOLVOET or at least part of them, appear to be of a different nature. We observe a superficial similarity to (10) in: (11) ķēniņa dēla ne redzēt, ne dzirdēt ‘The prince is nowhere to be seen or heard’. (12) no tās reizes vilku ne redzēt, ne dzirdēt ‘From that time onward the wolf was not to be seen or heard’. (13) sauli nekur manīt ‘The sun is nowhere to be seen’. (14) uzsviedis tik augsti, ka gandrīz ne saredzēt ‘He threw it up so high that it was hardly visible.’ These constructions are of a very peculiar nature when compared to (1, 2) and (10). What is striking is that in (11–12) and (14) there is no negation prefixed to the verb. In (13) we would expect nemanīt rather than manīt (cf. es viņu nekur nemanīja) and instead of ne saredzēt ([neˈsaredzēt]) in (14) we would expect nesaredzēt ([ˈnesaredzēt]). The negation is a clause negation and as such it should be proclitic⁷ and be prefixed to the verb, as in the case of neredzēt in (1) and nesaprast in (2). This is not the only irregularity. What is most puzzling is the use of case forms. (10) has a dative subject, just as (1) and (2). But there are constructions where a nominative appears which, semantically, does not function as the object of the infinitive (as in the Lithuanian construction jau namai matyti) but as its agent. Again, I cite the relevant examples from Endzelin (1951: 992): (15) zirgs tādu vezumu neparko pavilkt ‘There is no chance of the horse moving such a cart from its place’. (16) ķēniņa dēli ... tuvumā ne rādīties ‘There can be no question of the princes coming anywhere near the place’. If (11–14) are structurally equivalent to the Lithuanian construction jau namai matyti, in which the agent, if overtly expressed, can only be in the dative (man bus namai statyti), then we expect the agent with the infinitive to surface as a dative as well. If this is the case, then we must separate the type (15–16) from the type represented in (11–14). But are we really entitled to separate these two groups of constructions in Latvian? Note that there are cases where no overt nominative subject-agent appears, but the appositions with the ‘zero subject’ are in the nominative all the same: (17) bij tāda pirts, kur vēlā vakarā ne domāt viens pats rādīties ‘There was a bath-house were one couldn’t even think of going on one’s own late in the evening’. 7 Here, as in many other cases, proclitic does not mean unstressed. The Latvian sentential negation forms an accentual unit together with the verb, and as stress is normally initial within the stress unit, it is actually the negation that is stressed. LITHUANIAN būti WITH THE INFINITIVE 81 The nominative viens pats (instead of vienam pašam) clearly indicates that we are dealing with a NOMINATIVE ZERO SUBJECT: if it were overtly expressed, it would appear in the nominative, just as in the Lithuanian construction jau namai matyti the agent, if expressed, would appear in the dative. In both cases, a zero subject evidently appears instead of an overt subject to indicate that the agent is generic, i.e., namai matyti means ‘anyone can see the house’, and Latvian ne domāt means ‘one could not even dream of doing something, one could not even consider doing something’. A specific (non-generic) subject would obviously appear as a nominative, which would yield (18), a structure analogous to (15–16): (18) ķēniņa dēls ... ne domāt viens pats rādīties And a full syntactic representation of (17) should account for the presence of a zero subject marked as a nominative (the subject having no surface realisation but imposing the use of the nominative rather than the dative as a case for agreement of the predicate nominal): ∅nom ne domāt viens pats rādīties whereas for the Lithuanian construction we could propose an abstract model of the following kind: ∅dat namai matyti Now it becomes clear that for (11) we must posit an underlying structure completely different from that of the Lithuanian construction, viz. ∅nom ķēniņa dēla ne redzēt, ne dzirdēt It is true that, among the examples with zero subject cited by Endzelin, only one, viz. (17), contains syntactic evidence for a nominative zero subject. In other examples, such as (13) and (14), we have only oblique evidence in the sense that the place and status of the negation allows us to identify these sentences as belonging to the same type as (17) and therefore also to the same type as (15), (16). Such a conclusion does not seem to be unwarranted, as all the examples suggest a close connection between the place and status of the negation on the one hand and the nominative subject on the other. The nature of this connection will be discussed below. The question to be answered at this point is whether clauses of this type, quite unknown to Lithuanian (as well as to the Slavonic languages), are historically connected with the Lithuanian type (man) namai matyti or whether they represent a completely independent development of Latvian. In view of the overall functional similarity it seems tempting to assume that the Latvian type arose from an older type corresponding to the Lithuanian one, in which an agent with the infinitive could be expressed by a dative. What gave rise to the clause type with a nominative subject was evidently a reanalysis of the variety with a 82 | AXEL HOLVOET generic zero subject: ∅dat was reanalysed as ∅nom, after which overt subjects in the nominative could be introduced. This reanalysis of the zero subject as a nominative (i.e., a zero subject occupying the place of a nominative subject and imposing the nominative as a case for agreement of the predicate nominal) is no doubt connected with the rise of generic singular zero subjects, a feature characteristic of Latvian, without counterpart in Lithuanian. The Latvian zero subjects are dealt with in Holvoet (1995), but the main points should be repeated here in brief. Both Lithuanian and Latvian have indefinite zero subjects, i.e., constructions in which the position of a nominative subject with a personal verb form remains unoccupied, the subject being interpreted as indefinite, but referential, e.g. kur (∅nom) medžius kerta, ten skiedros lekia, Latv. kur (∅nom) malku cērt, skaidas lec. But Latvian distinguishes two kinds of indefinite subjects: one of them is syntactically plural8, and denotes a referential subject (typically a group of persons); the other is syntactically singular, and denotes a non-referential (generic) subject. Such a generic singular zero subject is often used with a modal verb like varēt ‘can, be able’, as illustrated in (19). Note that the zero subject behaves, for purposes of agreement, as a nominative masculine singular subject (cf. the nominative masculine singular traks). (19) Varēja traks palikt no dusmām. (A. Eglītis) ‘One could have got mad with anger’. It is conceivable that singular zero subjects of this kind extended their scope to the construction būt + infinitive, giving rise to a structural model as illustrated in (11–14). One more peculiarity in this process calls for a syntactic explanation: the puzzling functioning of the negation. As we observe in the above examples, all of them are in some way negative, but the type of negation is different from what one would expect on the basis of such examples as (1), (2). The type with dative subjects (the one with structural equivalents in Lithuanian) has a normal proclitic negation attached to the infinitive (neredzēt, nesaprast). Why do we have Zirgs tādu vezumu ne pavilkt rather than *Zirgs tādu vezumu 'nepavilkt with a proclitic negation? This negation is similar to the emphatic negation ne ‘not even, not as much as’ as illustrated in Kārlēnu nevarēja vairs ne pazīt ‘It was impossible as much as to recognise little Kārlis’. But this emphatic negative particle ne is used only if a normal (proclitic) negation is added to the verb (neteica ne vārda ‘didn’t say a word’), or to the main verb if the particle ne is added to an embedded infinitive (nevarēja ne pazīt ‘couldn’t even recognise’). Moreover, no trace of an emphatic negative particle can be detected in (13). It is therefore hard to explain the functioning of the negation in (11–14) unless we assume it to be attached to some abstract verb not appearing in surface structure: zirgs ne-∅ pavilkt vezumu. When an indefinite pronoun or correlative adverb occurs, the negation is attached to it, as in (13) and (15). That the zero subject is syntactically plural can be seen from the form of the participle in compound verb forms, e.g., Mani apzagaši ‘They robbed me’, ‘I was robbed’. In most cases, singular and plural zero subjects are not formally distinguishable because of the homonymy of singular and plural 3rd person forms, but the distinctness of both types of zero subjects manifests itself in instances of nominal agreement. LITHUANIAN būti WITH THE INFINITIVE | 83 A possible explanation could run as follows. It should be noted, first of all, that a structure with the dative still exists in contemporary Latvian. It is illustrated by (2) above. Its underlying structure could be represented as follows9 (note that the main clause contains a modal verb with zero realisation—a modal verb which we identify with the verb būt and to which we ascribe the lexical meaning ‘be necessary): In the variety in which the dative was replaced with a nominative, the verb būt as a modal verb was replaced with another abstract modal verb with zero realisation, differing from būt with regard to syntactic properties: with būt, the modal subject (the person who is subjected to a necessity) is in the dative; the abstract modal verb succeeding to būt combines with a modal subject in the nominative. It behaves, in this respect, like a modal verb such as varēt ‘can, be able’. At this stage, the negation is raised, i.e., it occupies the position reserved for the modal verb and functions as an external negation. This modal verb would have to be glossed ‘can, be able’, because, the negation being external, the meaning of impossibility cannot by derived from that of necessity any more. This raising has to be assumed because a negation ascribed to the embedded clause (an internal negation) should have the form characteristic of a clause negation: i.e., it should be proclitically attached to the infinitive. The rules on the placement of the negation in those cases where an indefinite pronoun or adverb occurs are the same as in constructions with an overt modal verb (Sauli nevar nekur 9 In this tentative representation of underlying syntactic structure, the dative subject is assigned to the matrix clause, and the subject of the infinitive is assumed to be deleted as being coreferential with this main clause dative subject. It would also be possible to assign it to the embedded clause and to assume the modal predicate to have just one propositional argument. Either choice is more or less arbitrary, and the same can be said of other impersonal modal verbs such as vajadzēt: In Tam vajag notikt ‘This must happen’ we could either assume the dative to be governed by vajag (as in Man vajag naudas ‘I need money’) or to represent the subject of the infinitive. 84 | AXEL HOLVOET manīt), but there evidently is a rule of deletion eliminating one of the negations, hence Sauli nekur manīt instead of *Sauli nekur ne manīt. But if a reanalysis of this type really occurred in the sentence type corresponding to Lithuanian arkliui vežimo nepavilkti, then why was the original model (with a dative subject) retained alongside the new type (with a nominative subject)? A possible explanation could be that the new construction was not, in all respects, functionally equivalent to the old one. In fact, Endzelin mentions the construction SUBJ + ne + INF twice, without considering the question whether there could be any link between the two uses10. Apart from the constructions expressing possibility it is also mentioned in connection with the narrative infinitive (Endzelin 1951: 991). The narrative infinitive replaces a finite verb form (a preterite) with a nominative subject, e.g. (20) Šis ... tūlin no dūkstes ārā, zirgam mugurā un skriet nu uz mājām (folk tale) ‘He jumped out of the bog, mounted his horse and galloped home’. But Endzelin adds that ‘there are also cases where the verb хотеть could be added in a Russian translation’ („citūr infinitīvam krievu tulkojumā var priekšā likt verbu хотеть“) (Endzelin 1951: 991), and he gives the following examples: (21) Brencis bija tāds pamulkis, un māte neparko viņu laist (folk tale) ‘Brencis was rather a simple-minded fellow, and his mother wouldn’t hear of letting him go’. (22) izteic visu, kā noticis, ķēniņš gandrīz ne ticēt (folk tale) ‘He tells everything exactly as it happened, and the king won’t believe it’. (23) saukuši, lai nevedot..., bet jaunkungi ne dzirdēt (folk tale) ‘They shouted they didn’t want to be taken along, but the young gentlemen wouldn’t listen’. As can be seen, this type is, with regard to formal features, absolutely identical with the type discussed above, which was stated to express (im)possibility. It is distinct, however, from the narrative infinitive illustrated in (20) by the obligatory presence of a negation. This negation is exactly parallel to that which we observe in the above-mentioned constructions expressing impossibility, viz. it is (at least formally) identical with the emphatic non-proclitic negation ne ‘not even’. The way in which both constructions, the one expressing volition and the one expressing impossibility, are presented suggests that Endzelin had no difficulty in establishing when he was dealing with which type. Is the difference really so neat? Can one establish beyond doubt that ķēniņš ne ticēt in (22) means ‘won’t believe’ rather than ‘can’t believe’? Probably not, because there are no formal differences associated with either of these interpretations. If both interpretations are possible, then the construction is evidently ambiguous. In other cases, only one interpretation is acceptable, 10 Unlike Mühlenbach (Endzelin & Mühlenbach 1928: 195), who cites the volitional type in close connection with the type allegedly expressing possibility, without suggesting a link with the narrative uses of the infinitive. LITHUANIAN būti WITH THE INFINITIVE 85 e.g., in (23) only ‘wouldn’t listen’ makes sense. Thus, there appears to be one syntactic model (SUBJnom + ne + INF) which is, in principle, vague between negative volition and impossibility. Which of the two readings is appropriate can usually be established from the context, but there are probably no formal criteria enabling disambiguation. Of course, the question must be posed whether there is not a genetic link between the volitional type illustrated in (21–23) and the narrative type illustrated in (20), as suggested (though not explicitly stated) by Endzelin. In that case, there would be no reason to assume a reanalysis as suggested above. The type with the nominative subject would have developed from the narrative infinitive, which is well attested in other languages as well. The existence of a link between the narrative infinitive and the volitional type is suggested, in Endzelin’s exposition, by the existence of what appears to be a transitional type: in the following example, Endzelin (1951: 991) argues, the infinitive denotes an intended action, not one actually performed: (24) Skrien divas stirnas; dēli tūlīn šaut, bet stirnas lūdzās. ‘Two roes ran by; the sons were about to shoot, but the roes began to beg mercy’. One could imagine, in principle, a volitional meaning evolving from the narrative infinitive describing an ‘intended action’, but in that case one would have expected an affirmative type to have evolved first, with a negative variety as a possible further development. Actually, the volitional type seems to be exclusively negative. A negative narrative infinitive is hardly imaginable, and therefore a link between the volitional type and the narrative infinitive is at least problematic. Moreover, even if we assume a volitional type *dēli nešaut ‘the sons did not want to shoot’ to have existed, there is still no explanation for the type with nominative zero subjects, for constructions with nominative zero subjects expressing negative volition are not attested: the only attested type is the one expressing impossibility. We are therefore compelled to look for a connection with the other sentence type that corresponds exactly to the model ‘SUBJnom + ne + INF’, viz. the one expressing impossibility. It is difficult to establish whether the constructions with the narrative infinitive could have provided a model for the remake of constructions originally characterised by the use of a dative subject. If there was any influence, it could have consisted in the rise of a sentence type with a syntactic zero replacing a modal verb: zēni šaut ‘And then the boys started shooting’. zēni ∅ šaut ‘And then the boys wanted to start shooting’. But to go further still and to assume the type expressing impossibility to have evolved from the narrative type as well seems unwarranted, as there is, after all, a close resemblance and a functional correspondence between the two types expressing impossibility, viz. (a) zirgam vezumu nepavilkt (with proclitic negation) 86 | AXEL HOLVOET (b) zirgs vezumu ne pavilkt (with non-proclitic negation) ‘The horse cannot draw the cart’. I therefore assume (a) and (b) to be genetically related, i.e., I assume (b) to have arisen from (a), a process to which certain other constructions, not genetically related to (a), could have contributed as far as the formal make-up of the construction and its unexpected volitional reading are concerned. One further point should be made here with respect to relative chronology. Note that in order for the process of substitution of Ønom for Ødat to occur, the syntactic position originally reserved for the nominative expressing the object of the action (as in Lith. namai matyti) had to be freed first. In other words, the replacement of the original nominative residually retained in the Lithuanian type man (yra) namai statyti with an accusative (as in Lith. man (yra) namus statyti) had been completed before a new nominative noun phrase was introduced. This means that, syntactically, no construction of the type namai matyti as a distinct residual type alongside the grammaticalised expressions of necessity could have been retained at the time when the reanalysis Ødat as Ønom occurred. Semantically, some traces of the original vagueness of the type represented by Lith. namai matyti seem to have been retained, if pazīt in (10) expresses possibility. As noted above, the original construction with the dative was retained alongside the new one. What we observe is the process of a syntactic structure splitting up into two separate ones rather than of the development of one structure into another. For the original construction to subsist, the new one had to be sufficiently distinct from it formally and functionally. The differences could be summed up as follows: (1) case marking (nominative as against dative), (2) status of negation (non-proclitic), (3) semantic vagueness (impossibility or negative volition as opposed to impossibility as the only possible reading), (4) perhaps there might be differences, within the general meaning of impossibility, between several types of modality that could perhaps be expressed by the new type but not by the old one or vice-versa; this, however, must be investigated separately. 5. CONCLUSIONS The construction discussed in this article is a common feature of Lithuanian and Latvian. Its oldest shape is retained only in Lithuanian: it is a construction with a nominative subject used with the substantive verb and an infinitive of purpose: namai (yra) statyti ‘the house is to be built’. This construction was modally vague between possibility and necessity, a feature frequently observed in modal constructions based on ‘be’ with the infinitive in several European languages. The use of this construction as an expression of possibility is a syntactic relic, retained mainly in Lithuanian, and subject to lexical restrictions (verbs of perception, as in namai matyti ‘the house can be seen’). In the course of the further development of this construction, būti began to function as a modal verb expressing exclusively necessity, and this is the only productive use in historical times. There is no longer a possibility reading, only an impossibility reading derived from necessity (‘□¬p’ = ‘¬◇p’). A syntactic concomitant of this process was the reanalysis of the nominative as an object of the infinitive and its replacement with the accusative (though Lithuanian has residually retained the nominative object). LITHUANIAN būti WITH THE INFINITIVE 87 Within the domain of negated constructions expressing impossibility, Latvian also shows a peculiar development not attested in Lithuanian, viz. the rise of constructions with a nominative subject instead of a dative subject, e.g., zings vezumu nepavilkt instead of zirgam vezumu nepavilkt. This construction is also characterised by certain peculiarities in the functioning of the negation. The mechanism of its rise is not clear. Most of the constructions cited in the literature as equivalents of the Lithuanian construction expressing possibility (namai matyti) actually belong to this type, and should be set apart (as an innovation) from the Lithuanian constructions from which they differ structurally. It is important to note this in connection with the Lithuanian type namai matyti, because scholars relying for their Latvian data on the publications of Mühlenbach and Endzelin (cf., e.g., Ambrazas 1995: 90) have been misled into believing that there is no fundamental difference between the Lithuanian and Latvian constructions expressing (im)possibility. REFERENCES AMBRAZAS, V. 1995: Lietuvių kalbos bendraties konstrukcijų raida. Lietuvių kalbotyros klausimai 33, 74–109. AMBRAZAS, V. 1997: Lietuvių kalbos papildinio vardininko su bendratimi kilmė. Lietuvių kalbotyros klausimai 38, 93–103. ANDRONOV, A. 1999: „Vajadžibas izteiksme“ latviešu valodas gramatiskajā tradīcijā (jeb The fluctuating fortunes of the Latvian debitive). Baltu filoloģija 8, 1998, 154–177. AUWERA, J. VAN DER, PLUNGIAN, V. A. 1998: Modality’s semantic map. Linguistic Typology 2, 79–124. BENVENISTE, E. 1966: ‘Etre’ et ‘avoir’ dans leurs fonctions linguistiques. Bulletin de la Société de Linguistique de Paris LV, 1960; here cited after: Problèmes de linguistique générale, Paris: Gallimard, 187–207. BORKOVSKIJ, V. I., red. 1978: Историческая грамматика русского языка. Синтаксис. Простое предложение. Москва: Наука. BRUGMANN, K. 1916: Vergleichende Laut-, Stammbildungs- und Flexionslehre [...] der indogermanischen Sprachen. II. Lehre von den Wortformen und ihrem Gebrauch. 3/2, Strassburg: Trübner. ENDZELĪNS, J. 1951: Latviešu valodas gramatika, Rīga: Latvijas valsts izdevniecība. ENDZELĪNS, J., MÜHLENBACHS, K. 1928: Latviešu gramatika. 4. iespiedums, Rīga: Valters un Rapa. FRAENKEL, E. 1928: Syntax der litauischen Kasus, Kaunas: Valstybės spaustuvė. HOLVOET, A. 1995: Indefinite zero subjects in Latvian. Linguistica Baltica 4, 153–161. MIKLOSICH, F. 1926: Vergleichende Grammatik der slavischen Sprachen. IV. Syntax. Manuldruck der Erstausgabe von 1868–1874, Heidelberg: Carl Winter. PALIONIS, J. 1972: Über einen Typ des Infinitivsatzes im Litauischen. Baltistica, I priedas, 125–129. POTEBNJA, A. A. 1958: Из записок по русской грамматике. Том I–II. Общая редакция... В. И. Борковского, Москва: Государственное учебно-педагогическое издательство. TIMBERLAKE, A. 1974: The nominative object in Slavic, Baltic and West Finnic, München: Sagner. (Slavistische Beiträge, 82.) ULVYDAS, K. ed. 1976: Lietuvių kalbos gramatika 3. Sintaksė, Vilnius: Mokslas. Axel Holvoet Lietuvių kalbos institutas Antakalnio g. 6, 2055 Vilnius, Lietuva [email protected] Gauta 2001 05 26