macedonian mora da

41
is is a contribution from Modes of Modality. Modality, typology, and universal grammar Edited by Elisabeth Leiss and Werner Abraham. © 2014. John Benjamins Publishing Company is electronic file may not be altered in any way. e author(s) of this article is/are permitted to use this PDF file to generate printed copies to be used by way of offprints, for their personal use only. Permission is granted by the publishers to post this file on a closed server which is accessible to members (students and staff) only of the author’s/s’ institute, it is not permitted to post this PDF on the open internet. For any other use of this material prior written permission should be obtained from the publishers or through the Copyright Clearance Center (for USA: www.copyright.com). Please contact [email protected] or consult our website: www.benjamins.com Tables of Contents, abstracts and guidelines are available at www.benjamins.com John Benjamins Publishing Company

Upload: uni-mainz

Post on 06-Feb-2023

1 views

Category:

Documents


0 download

TRANSCRIPT

This is a contribution from Modes of Modality. Modality, typology, and universal grammar Edited by Elisabeth Leiss and Werner Abraham.© 2014. John Benjamins Publishing Company

This electronic file may not be altered in any way.The author(s) of this article is/are permitted to use this PDF file to generate printed copies to be used by way of offprints, for their personal use only.Permission is granted by the publishers to post this file on a closed server which is accessible to members (students and staff) only of the author’s/s’ institute, it is not permitted to post this PDF on the open internet.For any other use of this material prior written permission should be obtained from the publishers or through the Copyright Clearance Center (for USA: www.copyright.com). Please contact [email protected] or consult our website: www.benjamins.com

Tables of Contents, abstracts and guidelines are available at www.benjamins.com

John Benjamins Publishing Company

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

On correlations between categorial restrictions and morphosyntactic behaviour*

Björn WiemerJohannes-Gutenberg Universität Mainz

The article presents a semasiological case study of the Macedonian modal mora, which derives from the common South Slavic modal verb morati ‘must’, and its collocation with the morpheme da, which is the probably most prominent non-factivity marker of Macedonian used in a broad variety of syntactic functions: as a complementiser, as a main clause marker in “insubordination” phenomena (with hortative and optative functions) as well as a morpheme connecting auxiliary and lexical verb in complex predicates. The latter function forms the starting point for the investigation: the author raises the question whether mora and da should not (in many cases) be analysed as one single unit (in functional and lexicographic terms). Mora can be regarded as a full-fledged modal auxiliary, but it tends to lose its inflectional properties (first, person and number, then tense), so that mora becomes petrified as a form isolated from its paradigm and homonymous with the prs.3sg-form of the inflected verb. Furthermore, it needs to be somehow connected to a finite verb and can cooccur only with da. This distribution, in connection with the successive loss of inflectional categories, conditions a transient stage whereby mora+da appear to become a sentence adverb (or particle) with propositional scope (univerbation toward a function word). Together with this, the split between inflected and uninflected mora correlates with a strong tendency to distinguish deontic (with

* The original idea to this article arose from collaboration with Elena Petroska (Skopje). The principal goals of the article have however changed, and in its present shape this paper would not have been possible without the continuous and patient support by Ubavka Gajdova (Skopje). I also want to thank the editors of this volume for their patience and the three reviewers for their critical and helpful reports as well as Zuzanna Topolińska for some discussion. Of course, the usual disclaimers apply, and I am aware that the proposal by which I want to capture the facts does not belong to mainstream treatments among specialists of Macedonian and that it is, in this respect, provocative.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

inflected mora) vs. epistemic (with uninflected mora) readings (dynamic readings have largely been ignored). As concerns this distinction, the feature [± inflected] proves more decisive than the choice of pfv. vs. ipfv. aspect for the lexical verb (being under scope of mora da). Moreover, the only grammatical context for which mora da + Vfin shows a systematic preference for deontic readings, is the pfv. present. All other contexts (and given mora is uninflected) rather favour epistemic readings, and for pfv. present tense (under scope of mora) the deontic reading can be overridden by some other factors. The article divides into two main parts: in the first, the author provides a systematic analysis of mora (da) in virtually all grammatical contexts imaginable for the occurrence of mora (da) in monoclausal structures (Section 2). In the second part (Section 3), the author takes issue with the question whether mora+da should not be treated as one linguistic unit, namely a propositional operator marking epistemic necessity (= ‘strong epistemic support’, in terms of Boye 2012).

.  Introduction and formulation of the task

In Macedonian, mora da serves as a necessity modal with both deontic and epistemic meanings. If used as an epistemic modal it can easily be interpreted as an inferen-tial marker, too, the reason lying in the close conceptual affinity between epistemic necessity and inferential evidentiality, which has been noticed so often in the literature (especially after Van der Auwera & Plungian 1998: 84–86). On first sight, mora da appears to be composed of two elements: the 3sg.prs.ind-form of the verb morati (= former infinitive) ‘must’, and the complementizer da ‘that’. Both have cognates in other South Slavic languages or, as for da, even beyond them.1 However, on closer inspection, when one analyses the functional distribution, syntactic behaviour and morphological restrictions pertaining to both mora itself and the lexical verb forms with which it combines, one gets the impression that mora da, at least in certain cases, has to be treated as an inseparable unit. Moreover, this holistic unit shows functional restrictions to epistemic (and, hence, inferential) usage, which arise from paradigmatic isolation and morphological petrification of the modal. In fact, both phenomena are two sides of the same coin. This implies, of course, that da, in turn, has lost its former status as a complementizer (which it keeps in other contexts).

With this in mind, the present article gives a comprehensive account of the usage types of mora (with and without da), which is meant to acquaint the reader with the relevant facts of Macedonian (Section 2). If not indicated otherwise, the data adduced

.  Among etymologically related units we can mention the Russian “confirmative particle“ da ‘yes’, which is used this way in Macedonian, too.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

have been discussed and more or less systematically been submitted to variation with an informed native speaker (see f. *), but at the present stage no systematic corpus investigation has been conducted. Let me further emphasize that this article takes a primarily semasiological perspective, in contrast to Hansen’s (this volume) typological account of modals in European languages, which by necessity has to look at the data primarily from an onomasiological point of view. In Section 3 I will discuss the lexical vs. grammatical status of different ways of usage of Macedonian (Mac.) mora (da) and show that the behaviour of mora da happens to be ambiguous in syntactic terms. Section 4 presents a summary and some vistas for further research.

.  Mac. mora (da): A survey of its meaning range and syntactic behaviour

The form mora ‘must’ derives from the now extinct infinitive morati and corresponds to its 3sg.prs.ind-form. Since in Macedonian the infinitive has been lost as a form cat-egory of the verb altogether, we will write mora to indicate the lemma. As other modal auxiliaries as well, it belongs itself to the imperfective (ipfv.) aspect and does not have a perfective (pfv.) counterpart (on this aspect distinction see 2.1–2.2).

First and foremost, we have to distinguish between finite, i.e. inflected, forms of mora and its uninflected form. Finite forms belong to a certain type of verbal conju-gation, are organized in a paradigm and show person+number and tense inflections; in this respect we can use the shortcut ‘personal forms’ for them. Note that as ‘finite’ (or: ‘inflected’) I count not only forms that distinguish all three finiteness categories just mentioned, but also forms characterized only by a tense distinction (present vs. imperfect), but not by the agreement categories person and number. In a sense, thus, finiteness can apply fully or only partially (on this issue see in particular Section 3.1). By contrast, the uninflected form looks identical to the 3sg.prs.ind of mora, but it is isolated in paradigmatic terms and has been developing into a morphologically pet-rified auxiliary or even particle; in this respect, we can call it ‘impersonal’. Below we will see that a form shaped mora cannot always undisputably be determined as being either the inflected 3sg.prs.ind of mora or just its uninflected (petrified) form. In other words: On text level its morphosyntactic status can remain ambiguous. For this reason I will gloss this form simply as ‘must’; 1st and 2nd person- and 3pl-forms will be marked as usual fully inflected forms.

The different morphosyntactic status of the two forms of mora becomes obvi-ous if we look at patterns of (non-)agreement. Compare the following examples. In (1–2) either the singular or the plural form of mora in present resp. imperfect tense can occur, however the lexical verb which follows after da is marked with the plural. Although the two forms of mora show clear signs of distributional and functional

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

asymmetry (as will be shown in the central part of this paper), either of them can, at least with certain lexical verbs, contribute to deontic or epistemic meaning:

(1) Mora / mora-at da bid-at / se tamu. must / must.prs.3pl con2 be.pfv.prs.3pl / be.ipfv.prs.3pl there (i) ‘They must/have to be there.’ (ii) ‘They are certainly there.’

(2) Mora-še / mora-a da bid-at / must.impf.3sg / must.impf.3pl con be.pfv.prs.3pl / se tamu. be.ipfv.prs.3pl there (i) ‘They had to be there.’ , (ii) ‘They were certainly there.’

At least theoretically, sentences (1–2) with the 3sg-forms mora and moraše, respec-tively, could be interpreted as biclausal constructions with a modal matrix predicate and its clausal complement headed by an inflected form of a lexical verb and a comple-mentizer da. This would yield a construction with the following “inner form”:

(3) ‘(It) must be/is necessary that P [= they are there]’. ↓ ↓ ↓ mora da [bidat/se tamu] CLAUSE 1 comp CLAUSE 2

However, mora in (1) does not have an independent syntactic status and rather behaves like an auxiliary, if not a particle (see 3.2). This holds even if, within the verbal complex, it does not mark the usual inflectional categories of Macedonian verbs ( person, num-ber, tense, mood). That mora in examples like (1) has to be ascribed uninflected status, becomes obvious if an explicit plural subject appears, as in the following examples (see also (6a, 11–12, 15, 17–21, 26–27, 33)):3

.  con just means ‘connective element’, more precisely: a clitic that links together the aux-iliary and the lexical part of auxiliary complexes. The reason why da is not glossed comp (=  complementizer) will become clear immediately before 2.1, and it will be discussed at length in 3.2. Note that I take complementizers in a narrow sense, namely: as units connecting clausal arguments to complement-taking predicates. They appear thus as explicit markers of junctures between clauses and, thus, participate in a biclausal structure. Contrary to this, mora+da+Vfin constitutes a monoclausal structure with a single argument structure belonging to Vfin and mora as an operator on that structure. See also the “list of abbreviations in glosses” at the end of the text.

.  The same can be seen with 1st and 2nd-person subjects; see examples (22–24, 29).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

(4) Deca-ta mora / mora-at da bid-at / children.def_art must / must.prs.3pl con be.pfv.prs.3pl / se tamu. be.ipfv.prs.3pl there ‘The children must be there.’ → deontic or epistemic (see (1–2))

(5) Deca-ta mora / mora-at da children.def_art must / must.prs.3pl con pristign-at utre. come:pfv.prs.3pl tomorrow ‘The children must arrive tomorrow.’ → deontic

Here mora does not show agreement (nor does it distinguish tense, see below); instead, person+number agreement is indicated by lexical verbs which form a predicate complex together with mora. We may thus treat the combination [mora+da+finite verb] as a complex predicate with mora as a modal auxiliary. On an SAE-background, this “division of labour” within auxiliary complexes looks unusual: contrary to modals, e.g. in German or English, it is the lexical verb (after da) which always indicates the inflectional categories, whereas the modal element can agree with the lexical verb in person and number, but need not.4 In the latter case a petrified default form of the verb (historically deriving from and, as it were, “homonymous” with the 3sg.prs.ind) occurs.

Furthermore, there is a very strong preference for interpreting sentences with mora as deontic if it is inflected (“personal”), while an epistemic reading becomes much more likely if mora is uninflected (“impersonal”). In fact, epistemic readings are possible for the 1st or 2nd person only under specific conditions (see below). This holds for either case, whether mora is inflected (6b) or not (6a). Thus, (6a–b) are interpreted deontically:

(6) a. Mora da dojdam / dojdaš. must con come:pfv.prs.1sg come:pfv.prs.2sg b. Moram da dojdam / Moraš must.prs.1sg con come:pfv.prs.1sg must.prs.2sg da dojdaš. con come:pfv.prs.2sg ‘I / you(.sg) must come.’

.  This pattern is, of course, connected to the loss of the infinitive in the most central “Balkan languages.” On the areal distribution of modal auxiliaries in Europe cf. Hansen and de Haan (2009) and Hansen (this volume).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

For further details on the distribution with respect to ipfv. vs. pfv. verbs see 2.2.2.Moreover, epistemic readings with mora are practically excluded if the proposi-

tion is to hold about future situations, as in (5).5 Thus, if in (4) we would add an adverb indicating that the assertion is valid for the speech interval (e.g. sega ‘now’) this is compatible with epistemic meaning, whereas an adverb like utre ‘tomorrow’ would block this meaning, leaving us with a deontic reading. Note however that it would be inadequate to assume that it is reference to the speech time interval which is crucial for a deontic vs. epistemic preference in utterances with mora (da). Rather, it is relevance for some posterior interval which plays the decisive role in the deontic vs. epistemic interpretation, regardless of whether the reference interval coincides with the moment of speech or not. Therefore, what is at stake is Topic Time in Klein’s (1994) sense: “the time for which the particular utterance makes an assertion” (1994: 37 and passim). This can be seen from examples in which mora da takes scope over a verb marked with the l-suffix. In general, these forms mark anteriority (see 2.1, 2.2.1). With telic verbs they can refer to the resultative state of some future situation, i.e. be used like a futurum exactum; see (7a). Remarkably, these forms are uninterpretable in the scope of inflected mora (7b) for which a strong deontic preference obtains:

(7) a. Deca-ta mora da [pristigna-l-e vo petok]. children.def_art must con come.pfv:l-ptcp.pl on Friday ‘The children must have arrived on/by Friday.’ → epistemic b. *Deca-ta mora-at da [pristigna-l-e children.def_art must.prs.3pl con come:pfv:l-perf.pl vo petok]. on Friday → no interpretation possible

To sum up so far: There are two constructions that differ both in semantic and in mor-phosyntactic terms, and the relation between the semantic and the morphosyntactic dimension is asymmetric.6 Notice furthermore that the two verb forms of the predi-cate complex are “knit together” by the unit da, both with inflected and with petrified mora. If da were to be considered as a complementizer, the sentences it participates in should be analysed as biclausal structures (as paraphrased in (3)). However, such an interpretation has to be rejected.

.  Cf. also Čašule (1989: 108). For the verbs bide (suppletive pfv. present for e ‘s/he is’), ima ‘1. s/he has; 2. there is’ and znae ‘s/he knows’ the situation is different (see bidat ‘they will be’ in (1–2) and (4)). We will not go into these exceptions here.

.  This holds for Macedonian modal verbs in general (cf. Čašule 1989).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

In the following I will be focusing on the semantic differences, more precisely: on deontic vs. epistemic interpretations; other modal subdomains (dynamic etc.) will be neglected. A distinction of semantic functions brings about distributional restrictions with regard to tense-aspect forms of the lexical verb. For a better understanding I first give a comprehensive overview of the Macedonian tense-aspect system together with grammatical marking of evidentiality (2.1), on which background I then analyse the grammatical behaviour of mora (da) (2.2).

.  On the Macedonian system of TAM-forms and grammatical evidentiality

Macedonian demonstrates a rather complex tense-aspect system. It consists of the present, the future (marked with the particle ќe), two aspectually distinguished past tenses (aorist, imperfect), a future in the past (future marker ќe combined with the imperfect) and an old perfect indicated by the l-suffix creating a participle inflected for gender+number and a be-auxiliary (see 8a) as well as a pluperfect consisting of the l-participle and the imperfect of the be-auxiliary. To all this we have to add two newer perfects: one formed by participles with an n/t-suffix (inflected for gender+number) and a be-auxiliary (see 8b), and one expressed by the inflected have-auxiliary plus an uninflected participle (suffixed -n/t- and with -o from the neuter, serving as a default choice in cases of lack of agreement; see 8c):

(8) a. Sum doš-o-l-Ø. / be.prs.1sg come:pfv.l-ptcp:nom.sg.m Sme doš-l-e. be.prs.1pl come:pfv.l-ptcp:pl

b. Sum dojd-e-n-Ø. / be.prs.1sg come:pfv.n/t-ptcp:nom.sg.m Sme dojd-e-n-i. be.prs.1pl come:pfv.n/t-ptcp:pl

c. Ima-m dojd-eno. / have.prs.1sg come:pfv.ptcp_indecl Ima-me dojd-eno. have.prs.1pl come:pfv.n/t_indecl ‘I have come. / We have come.’

The past tenses aorist and imperfect are distinguished as different aspects. A younger aspect system, based on stem derivation, has developed independently from this distinction. It is known as the perfective: imperfective (pfv.: ipfv.) opposition and com-mon to all contemporary Slavic languages (cf. Breu 2000; Lehmann 2009; Wiemer

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

2011: 741f.). Both the distinction of imperfect: aorist and the pfv.: ipfv. opposition represent different types of viewpoint aspects. They have grammatical status, because their functional differences do not consist in (or boil down to) lexically inherent prop-erties (‘lexical aspect’)7 or modifications thereof marked with stem-extending affixes. Since the Slavic pfv.: ipfv. opposition is based on stem derivation, it can be distin-guished in all forms of the verbal paradigm, finite and non-finite ones. But the pfv.: ipfv.- opposition must not be mixed up with lexically inherent aspectual distinctions of verbs, let alone with a distinction between telic and atelic stems, which is another purely lexical property (cf. Breu 2007, among others). Derivational morphology here has developed as a means to “produce“ new stems having (potentially) identical lexical meanings, with which they share into (largely) different sets of grammatical functions (among them event quantification, see below). This is not encountered in Germanic languages, whose verb derivational affixation sometimes superficially remembers what happens in Slavic languages. Therefore the following clarification seems in place.

Perfective-imperfective distinctions in Germanic languages are of a merely lexical nature, since the aspectual difference is practically always accompanied by a change in lexical meaning, mostly carried by prefixes. Affixes mark Aktionsart-distinctions (in the Indoeuropeanists’, not the Vendlerian understanding of this term), but, apart from lexical modification, they do not come in pairs as do, to the contrary, the majority of verb stems in Slavic languages. Thus, for instance, the German simplex (unprefixed) stem teilen ‘divide’ and its derivatives prefixed with zer- or auf- (zer-/auf-teilen ‘divide, split’) may even not much differ in lexical meaning (nor need they in English, e.g. split ⇒ split up, cross ⇒ cross over), but members of the respective simplex and pre-fixed verb pairs are not characterized by a tendency toward complementary distribu-tion over grammatical context types definable by features like [± undivisible event], [± iterative], [± episodic], [± action/event presupposed], but also [± deontic] (e.g. in the scope of modal auxiliaries).8 This is what has happened to Slavic pairs like Russ.

.  In order to avoid a possible misunderstanding (as indicated by one of the reviewers), the following terminological clarification is appropriate. ‘Lexical aspect’ has been distinguished from ‘viewpoint aspect’ by many authors after Smith (21997) and, independently from her, after Johanson (cf. Johanson 2000 and his previous publications). Viewpoint aspect largely corresponds to what I here call ‘grammatical aspect’, while lexical aspect can be captured also as a division of aspectual classes of verb stems as they have been assumed by many typo-logically oriented aspectologists (e.g. Breu 1996, 2000; Lehmann 1999, 2009; Johanson 2000; Tatevosov 2002).

.  This distribution holds to pfv. and ipfv. verbs irrespective of whether they come in pairs or not, because these features apply to either the class of pfv. or the class of ipfv. stems as a whole. As for the contrast with modal auxiliaries, among the standard observations made, first of all, for Russian, one has to be aware that the basic possibility modals moč’ and možno

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

delit’IPFV ⇒ raz-delit‘PFV ‘divide’, rvat‘IPFV ⇒ oto-rvat‘PFV ‘tear (off)’, but also to pairs of simplex and prefixed stem for which the prefix does not alter lexical meaning at all; see, for instance, Russ. pisat‘IPFV ⇒ na-pisat‘PFV ‘write (down)’ (vs. Germ. schreiben ‘write (down) ⇒ auf-schreiben ‘take notes’, which translates into Russ. za-pisat‘PFV). Furthermore, Germanic languages do not exhibit productive aspectual suffixation of verb stems, and where suffixes or umlaut apply to simplicia they modify the lexical meaning (causatives like Germ. trink-en ‘drink’ ⇒ tränk-en ‘water (tr.)’, “diminutives” like Germ. lach-en ‘laugh’ ⇒ läche-l-n ‘smile’), while in Slavic, apart from causatives and “diminutives”, suffixation is productively used to adapt prefixed stems to the inventory of grammatical contexts reserved for ipfv. verbs (e.g. Russ. pi-t‘IPFV ‘drink’ ⇒ po-i-t‘IPFV ‘water (tr.)’ ⇒ [na-poit‘PFV ‘water (tr.)’ ⇒ napa-iva-t‘IPFV ‘water (tr.)’], with […] marking an aspect pair).

Therefore, in Slavic abundant suffixation of prefixed stems is one of the main factors that have conditioned the rise of the grammatical pfv.: ipfv. opposition. Most of the Russian prefixed stems just mentioned can further be suffixed (e.g. razdel-i-t‘PFV ⇒ raz-delj-a-t’IPFV ‘divide, split’, zapis-a-t‘PFV ⇒ za-pis-yva-t’IPFV ‘take notes’) without a change of lexical meaning. As a result, the grammatical pfv.: ipfv. opposition system-atically involves pairs of derivationally related verbal stems which do not differ from each other in their lexical properties (including the feature [± telic]), but only in their grammatical behaviour. This includes a massive number of, on the one hand, “punc-tual” pairs9 (pfv.: ipfv. stems) for which neither stem implies incremental phases or changes of state. Here belong a host of verbs denoting speech acts or epistemic events, but also others, like Russ. spotk-nu-t’sja PFV —spot-y-k-a-t’sjaIPFV ‘stumble’, zamet-i-t‘PFV—zameč-a-t‘IPFV ‘(take) notice (of)’, Pol. dosta-ćPFV—dosta-wa-ćIPFV ‘receive’, dziękowa-ćIPFV—po-dziękowa-ćPFV ‘thank’). On the other hand, productive aspect derivation comprises atelic verbs in which the derived pfv. stems do not imply any resultant state, but only set an external time limit (e.g. Russ. guljat‘IPFV ⇒ po-guljat‘PFV ‘walk’, Pol. leże-ćIPFV—po-leże-ćPFV ‘lie’). Such facts are problematic for approaches to the aspect-modality interface which work with assumptions of incremental processes and resultant states as defining properties of ipfv. and pfv. aspect (cf., for instance,

‘can’ in affirmative sentences combine with ipfv. infinitives for a deontic meaning, while – other things being equal – with pfv. infinitives they associate with a dynamic (circumstan-tial) or ability (dispositional) reading (e.g. Petja možet otkryvat‘IPFV okno ‘Petja is allowed to open the window’ vs. Petja možet otkryt’ oknoPFV ‘Petja is able (or: It is possible/feasible for Petja) to open the window’). This distribution is at variance with standard assumptions about the aspect-modality link, according to which deontic meaning associates with perfec-tivity and epistemic meaning with imperfectivity (cf., e.g. Abraham 2008). For more details cf. Klimonow and Klimonow (2008) and Divjak (2011).

.  Smith’s (21997) semelfactives.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

Abraham 2008).10 All these remarks basically hold for Macedonian, too, with which we now continue.

In Macedonian pfv. vs. ipfv. verbs can be distinguished for the future (marked with the proclitic particle ќe) and the present tense. Actually, pfv. present tense forms practically occur only in subordinate clauses or in the scope of modifiers of factivity, such as da and the future marker ќe (see Section 3.2). The latter one can also be regarded as a conditional marker, with future being only sort of “side effect“ ( Topolinjska 2008a: 53–55 and p.c.). However, in the following I will refer to it as future marker. Regardless of this, pfv. and ipfv. verbs can be distinguished in all types of the perfect (see above) as well as in evidential (= non-confirmative) forms (see 9b, 26). The latter ones are built on the l-participle, and they are basically to be regarded as a func-tional extension of the first (i.e. the oldest) perfect (see (8a) above). Above all there exists an analytical conditional, called potential, distinguished by a combination of the uninflected clitic bi and the l-participle.11 This mood will however be only of minor concern for our further considerations (see the comments in 2.2.4).

In order to get one’s way through this complicated Aspect-Tense-system, its combination with evidentiality marking and the relevance of all this together with the distinction of deontic vs. epistemic mora, one should, first of all, become aware that these paradigms divide into two groups of a pragmatically motivated opposi-tion. Following Friedman’s account, we have to distinguish between confirmative and non- confirmative AT-paradigms (Friedman 2000: 331–336; 2003: 194–196). Non- confirmative forms are functionally marked for their value of indirect evidentiality, i.e. they convey indirect experience of the speaker and, thus, correspond to reportive and inferential evidential meanings as these have been accepted after widely known works

.  For the same reason it is misleading to regard the pfv. aspect as “completed” and the ipfv. as “uncompleted”. For instance, Russ. ipfv. spotykat’sja ‘stumble’, priezžat’ ‘arrive (by some locomotion)’ or brat’ ‘take’ are no less “completed“ than their pfv. counterparts (spotknut’sja, priexat’ and vzjat’, respectively). And, vice versa, Russ. pfv. poguljat’ ‘walk (for a while)’ and pospat’ ‘sleep (for some time)’ are no less “uncompleted“ than their ipfv. equivalents guljat’ and spat’. The latter because the verbs in either aspect are atelic, the former because irrespective of the aspect the verbs denote punctual events (Achievements in Vendler’s terms). After all, “(un)completedness“ is a lexical property, from which grammatical aspect is largely independent.

.  Macedonian grammarians have been calling this the ‘potential mood’ (Mac. ‘potencijal, možniot način’); cf., among others, Koneski (1976: 380, 499–502). To a large extent it functions like a conditional, being exploited mainly in conditional sentences in both protasis and apo-dosis. It predominantly occurs, though, in utterances concerning unrealized events or events judged as doubtful by the speaker (cf. Kramer 1986: 125; Topolinjska 2008a: 51f.; Mišeska Tomić 2012). This construction originated from a Common Slavic optative ( Topolinjska 2008a: 50). Anyway, it is important to distinguish the form bi + l-participle from the historically unrelated, newly developed ‘subjunctive’ marked with da (see 3.2).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

of, e.g. Aikhenvald (2004) or Plungian (2001, 2010).12 In parallel to the (non-)confir-mative opposition Friedman, following here the traditions of Balkan Slavic grammar-ians, distinguishes between the so-called ‘definite’ and ‘indefinite’ tenses. The aorist and imperfect represent definite tenses in the sense that they ascribe definite temporal reference to situations denoted by verbs (Topolińska 2008b [1994]: 178), which is why they are typically used in narration.13 What more, they always convey confirmative meaning (i.e. the speaker takes full responsibility for the asserted propositions), while the perfects divide between confirmative and non-confirmative meaning (cf.  also Friedman 2004: 104–107).

It might seem now that the two distinctions ([± confirmative] and [± definite]) are identical; however they are not, for two reasons. First, the complex form consisting of the l-perfect and the imperfect of the auxiliary (= pluperfect) is not interpreted as non-confirmative, whereas the l-perfect with an auxiliary in the future would create a future reportive meaning – an issue we need not go into here (see however (34)). What is crucial now is that only the present perfect based on the l-participle (as in (8a)) forms the basis of grammatical evidentiality or, in Friedman’s terms, of the non-confirmative paradigms. The entirety of the Macedonian perfect series could equally well be considered as a system of ‘anterior’ grams (in the sense of Thieroff 2000). As such these so-called indefinite past tenses do not carry any sort of non-confirmativity (or indirect evidentiality), but rather denote an unmarked past vis-à-vis the marked confirmative past; they have non-confirmativity only as their chief contextual meaning variant (Friedman 1999). Second, the [± confirmative] and [± definite] distinctions are treated as mutually independent oppositions. This is to be inferred from the fact that the respective forms can (though very rarely) be combined; compare, for instance, the imperfect non-confirmative form in (33b).

What, then, is the essential point behind the non-confirmative interpretation? The crucial thing is that non-confirmative past forms are to be considered as

.  Friedman himself has argued against evidential functions as the proper meanings of the indefinite past (and l-forms in general). Instead, he has developed further Aronson’s (1977) view of these forms as marking ‘status’, a category which he took over from Jakobson (1971 [1967]). For both Aronson and Friedman status is a hybrid category with epistemic, eviden-tial and other associated meaning components (for a critical survey cf. Boye 2012: 40–42). However, for the present concern these distinctions are of marginal significance. Note further that Friedman treats non-confirmative forms as unmarked in morphological terms. However, the non-confirmative l-forms are easily recognizable with stems based on the imperfect. In this case often a vowel change occurs; for instance, prav-e-l.impf from the ipfv. stem prav-i (see also (33b) with pfv. stems).

.  This corresponds to Klein’s (1994) concept of Topic Time referred to in the introduction to this section.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

grammatical paradigms with suspended factivity. This means that the speaker does not take responsibility for the veracity of the message (assertion, statement); cf., for instance, Topolinjska (2001: 41–88). Confirmative tenses mark the utterance as factive, whereas non-confirmative tenses mark the utterance as neither factive nor non-factive, but suspend factivity as being irrelevant. Compare the following Examples: (9a) is uttered in the aorist; the speaker takes complete responsibility for his/her statement. (9b) is an instance of reported speech with a matrix verb in the aorist expressing the reporting speech event (rekoa) and a syntactically subordinate verb in the l-perfect (došol). The latter indicates that the actual speaker does not take the whole responsibility for the contents of the reported statement; s/he assigns its authority to someone else, retells, and thereby refers to indirect evidence. In (9c) the same l-perfect forms the predicate of a non-subordinate clause; again, the speaker does not take responsibility and suspends factivity. By contrast, if instead the n/t-participle (dojden, with an overt auxiliary) is used, the utterance yields factive interpretation (9d):

(9) a. Dojde. come.pfv:aor.3sg ‘S/he arrived.’ b. Mi rekoa deka došol. (reported speech, ± factive) I.dat say.pfv:aor.3pl comp come.pfv.l-ptcp:sg.m ‘They told me that he arrived.’ c. Došol, mu go come.pfv:l-prf:sg.m I.dat he.acc slušam glasot. (± factive) hear.ipfv:prs.1sg voice.def_art.sg.m ‘He came, I can hear his voice.’ d. Dojden e, mu go slušam glasot. (+ factive) come.pfv:n/t-ptcp:sg.m be.prs.3sg ‘He came, I can hear his voice.’

After this overview of the TA-system of Macedonian and the way it interacts with the opposition of (non-)confirmativity, let us turn to the distribution of mora (da) with regard to these forms.

.  mora da and restrictions on AT-forms

The (non-)confirmative distinction is important for explaining restrictions on the use of mora (da). Consider: As an epistemic modifier this modal takes scope over whole propositions, including their temporal and aspectual properties indicated by

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

tense-aspect forms. These forms also bear implications related to factivity and, thus, to the speaker’s responsibility. We may therefore assume that epistemic mora (da) and the values of AT-forms have to be congruent with each other. It is exactly this sort of ‘semantic concord’ which manifests itself in many of the restrictions of mora (da) with AT-forms under its scope.

..  The perfectsMora (da) typically takes scope over non-confirmative forms; the modal and the l-form (without be-auxiliary in the 3rd person) are, as it were, in semantic concord with each other (10a). However, confirmative present perfect forms are not excluded either (10b):

(10) a. Mora da došol, must con come:pfv.l-ptcp:sg.m mu go slušam glasot. I.dat he.acc hear.ipfv.prs.1sg voice.defart.m b. Mora da e dojden, mu go slušam glasot. must con cop.prs.3sg come:pf:n/t-ptcp:sg.m ‘He must have come, I hear his voice.’

In neither case does the speaker take the whole responsibility for the contents of the statement; s/he infers from hearing the voice, thus refers to simultaneous, but indirect evidence for the respective person’s arrival. However, the l-perfect in (10a) more often than not conveys salient evidential meanings, not only inferential, but also reportive ones, which often dominate over the original anterior meaning.

Moreover, mora da can occur with the new ima-perfect; for instance:

(11) Mora da imaat raboteno nešto. must con have:ipfv.prs.3pl work:ipfv.ptcp_indecl something.n ‘They must have worked a little bit.’

(12) Mora da imaat zaraboteno nešto. must con have:ipfv.prs.3pl earn:pfv.ptcp_indecl something.n ‘They must have earned some money.’

(13) Mora da ima zapaleno cigara, must con have:ipfv.prs.3sg lit:pfv.ptcp_indecl cigarette.f mirisa na čad. smell.prs.3sg on smoke.m ‘S/He must have lit a cigarette, it smells with smoke.’

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

See also the following example from an internet source:

(14) Ovoj papagal e sigurno najinteresniot papagal sto ste go videle dosega. Za da go pravi ova mora da go in order it.n.acc do:ipfv.prs.3sg this.n must con it.acc ima videno od nekogo. have.prs.3sg see:(i)pfv:n/t_indecl from somebody.gen ‘This parrot is, surely, the most interesting parrot you have ever seen.

He must have seen it from someone in order to do this.’ (http://www.svastarnik.mk/zabava/klipovi/11172-ova-e-najbezobrazniot-

papagal-shto-ste-go-videle-video)

As Examples (10–14) demonstrate, epistemic (and inferential) meaning of [mora da + lexical verb] can be achieved with all three present perfects, irrespective of pfv. vs. ipfv. aspect of the lexical verb.

..  Aspect (pfv.: Ipfv.) in the present tenseIn general, Macedonian present tense forms of pfv. verbs are restricted to subordinate clauses or structures under the scope of non-factivity operators, such as da (see 3.2).14 This provided, we observe a tendency toward a more clear-cut distribution of present tense forms of pfv. vs. ipfv. aspect; the former ones are associated with deontic, the latter ones with epistemic readings:

(15a) Mora / moraat da dojd-a-t. → deontic must / must.prs.3pl con come:pfv:prs.3pl ‘They must (= have to) come.’ (15b) Mora da doaģaat. must con come:ipfv:prs.3pl Gi sluša-m. → epistemic, inferential they:acc hear:ipfv.prs.1sg ‘They must be coming. I can hear them.’

However, the clear deontic:epistemic bias correlating with pfv.:ipfv. aspect, does not really amount to complementary distribution. First, mora inflected in the present tense with ipfv. lexical verbs often has a deontic meaning, as in

.  This restriction is a general South Slavic feature; cf. Mišeska Tomić (2006), Topolinjska (2008a: 56f.).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

(16) Moraš da ja must.prs.2sg con she.acc pišuvaš domašnata zadača. → deontic write:ipfv.prs.2sg home.defart_f task.f ‘You have to write your homework.’

Second, often epistemic readings need contextual cues able to supply an explicit indi-cation of the basis of inferences, as in (15b). If a hint at the source of information is lacking, mora da + ipfv. verbs in the present tense may be interpreted deontically; cf.:

(15c) Mora / moraat da doaģaat. → epistemic ∨ deontic must / must.prs.3pl con come:ipfv.prs.3pl (i) ‘They must come’ , (ii) ‘They must be coming.’

In the deontic reading the difference between the pfv. (15a) and the ipfv. (15c) verb would be of a merely time-referential nature: mora da dojdatPFV would refer to an obligation for one (imminent) occasion, whereas mora da doaģaatIPFV would mean that the persons in mind are obliged to come regularly to appointments of some sort. The ipfv. verb furthermore opens up a progressive reading (referring to one occasion); see translation (ii) of (15c).

Epistemic readings, in turn, can distinguish semantically embedded aspectual interpretations. Thus, a progressive reading arises with ipfv. present tensed verbs (see  15b), while a resultative state meaning arises with the present perfect (= past indefinite tense), as in (10a–b).

..  Confirmative past tensesImperfect and aorist belong to the confirmative AT-forms. Therefore, as for their usage under the scope of mora da a question might arise: What sense is there to make an inference (and to suspend factivity) and at the same time to assert that the denoted situation obtained (or obtains)? Nonetheless, aorist and imperfect forms do occur and will in both cases contribute to epistemic readings; admittedly, they seem to occur only with uninflected mora.

As for the aorist, apart from its function in narratives it is used to denote resul-tative states (and can in this respect be equivalent to the n/t-perfect; see (8b)). This function is the relevant one under the scope of mora (da); compare:

(17) Mora da dojdoa. → epistemic must con come:pfv:aor.3pl ‘They must have come. ‘

(18) Mora da zarabotija mnogu pari. → epistemic must con earn:pfv:aor.3pl much money ‘They must have earned a lot of money. / They probably have earned…’

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

In turn, the epistemic meaning of [mora da + impf] can at first sight be connected to the basic progressive meaning of the imperfect (which it also has in narratives); compare

(19) Mora da doaģaa. → epistemic must con come:ipfv:impf.3pl ‘It must be that they were coming.’ / ‘They probably were coming.’

This is however not the only possible meaning. Actually, under the scope of mora da the imperfect, as it were, “copies” the functions of ipfv. present tensed verbs by trans-posing them to the past. Compare:

(20) a. ipfv. present Mora da znaat što must con know:ipfv.prs.3pl comp se slučuva togaš. happen:ipfv.aor.3sg at_that_time ‘They must know what happened then.’ → epistemic b. imperfect Mora da znaeja što must con know:ipfv.impf.3pl comp se slučuva togaš. happen:ipfv.aor.3sg at_that_time ‘They must have known what happened then.’ → epistemic

In this case, a present state (20a) is transferred into the past (20b). But the imperfect is also used for iterative contexts; this is a possible reading at least for the next pair of examples:

(21) a. ipfv. Present Mora da odat porano na rabota. must con leave:ipfv:prs.3pl early to work ‘They have to leave early for work.’ → deontic (can be iterative) b. imperfect Mora da odea porano na rabota. must con leave:ipfv:impf.3pl early to work ‘They had to leave early for work.’ → epistemic (can be iterative)

These examples demonstrate two things. First, the epistemic vs. deontic interpreta-tion cannot depend on (grammatical) aspect as such, but appears to be conditioned more firmly by telicity, which is a lexically inherent aspectual feature. Thus, atelic ipfv. znaat ‘they know’ in (20a) can contribute only to an epistemic interpretation, while telic ipfv. odaat ‘they go/leave’ leads to a deontic reading. Furthermore, deontic

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

meaning is strongly associated with reference to some action which is posterior to a reference interval. Second, the iterative meaning of the imperfect can be carried over also under the scope of mora da (see also below).15 And it is iterativity which, in cases of conflict with the telic character of the verbal lexeme, leads to a preference of ipfv. aspect and/or the imperfect irrespective of the deontic:epistemic-distinction. In other words: It is rather that time-referential functions (in particular, event quantification) outrule modal distinctions than vice versa. This observation will be further elaborated on in 2.2.5.

Moreover, the imperfect alone is a very strong factor favouring epistemic mean-ing; it even overrides factors which are otherwise favourable for a deontic reading. Compare 1st person sentences with the lexical verb in the ipfv. present (22a) and in the imperfect (22b):

(22) a. ipfv. present Mora jas da gi vozam decata. must I.nom con they.acc drive:ipfv.prs.1sg children.defart.pl ‘I must drive the children (e.g. to school).’ → deontic (can be iterative) b. imperfect16

Mora jas da gi vozev decata. must I.nom con they.acc drive:ipfv.impf.1sg children.defart.pl ‘I must been driving/have driven the children.’ → epistemic (progressive or iterative interpretation)

This shows that although epistemic meaning tends to be excluded if the modal itself is marked with 1st or 2nd person (mora-m.1sg, mora-š.2sg in the present, mora-v.1sg, mora-še.2sg in the imperfect), epistemic readings seem the only possible ones if past reference is marked with the imperfect on the lexical verb irre-spective of grammatical person. What about the aspect distinction (pfv.: ipfv.); can the associations with deontic vs. epistemic readings be overridden by the imperfect, too? This cannot be tested, because in Macedonian pfv. verbs are not used with the imperfect.

We can however test whether the aspect-modality correlations are overridden with the l-perfect, which combines freely with both aspects (see 2.2.1). Indeed, if the

.  This does not seem to hold for all verbs in the imperfect. The conditions for an iterative function to obtain have to be submitted to further scrutiny.

.  However, in accordance with the compatibility of (in)definite tenses with mora da (see remark at the beginning of 2.2.3), the l-perfect (with be-auxiliary) of the lexical verb would sound more natural: Mora da sum gi voze-l decata.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

lexical verb appears in the l-perfect, the otherwise deontic preference with pfv. verbs and with 1st and 2nd person proves to be invalidated. Compare:

(23) Mora da ste go vide-l-e i porano. must con be(aux).2pl him.acc see:pfv.l-ptcp.pl even earlier ‘You (pl.) must have seen him beforehand.’

(24) Mora da si se naspa-l ubavo. must con be(aux).2sg refl sleep_enough:pfv.l-ptcp.sg.m well ‘You (sg.) must have slept well enough.’

(25) Potoa slušnav drug zvuk, doagaše od šupata, kako glasovi, a sfativ deka nekoj go ostavil radioto vklučeno.

Mora da go namali-l-e glasot must con him.acc.n lessen:pfv.l-ptcp.pl voice.defart_m namesto da go isklučat. instead comp it.acc switch_off:pfv.pts.3pl Tivko se slušaše nekoja stanica. ‘After a while I heard another sound coming from the cellar, like voices, and

I realized that someone left the radio switched on. They must have lowered the sound instead of turning the radio off. We could hear a quiet sound from the radio.’

(http://www.e-books.com.mk/02prose/ernest/tekst.asp?lang=mac&id=8)

Finally, the equivalence between imperfect and ipfv. present mentioned above makes one wonder whether the marking of past can be “switched”, in the sense that the modal would carry past tense, whereas its finite “complement” would be marked with present tense. And indeed, this is possible: (26) can be used as an equivalent of (19) – with the proviso that the epistemic reading would not be the only one imaginable; in fact, the preferred interpretation for (26) would be deontic. This conflicting situation is to be expected, since none of the factors overriding the strong preference for deontic read-ings of inflected mora is present, while the epistemic reading is favoured alone by ipfv. aspect of the lexical verb:

(26) Moraše da doaģaat. must.impf.3sg con come:ipfv:prs.3pl (i) ‘They were obliged to come.’ → deontic (preferred) (ii) ‘They must have been coming.’ → epistemic

Note further that in this construction the two tensed forms (moraše and doaģaat) do not agree in number, and subject agreement (with an implied plural referent) is car-ried only by the lexical verb. We may thus infer that the modal verb, though capable of carrying tense distinctions, tends to be defective in terms of other categories expressed by Macedonian verbs, namely those which are relevant for the syntax. This tendency is

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

realised in particular with respect to epistemic meanings, while, as was already stated above, the pattern in which mora and lexical verb agree in the syntactic categories show a strong association with deontic readings, as in

(27) Moraat da doaģaat. must.prs.3pl con come:ipfv.prs.3pl ‘They must come.’ → deontic

After this discussion of past and present tense forms, combined with mora da, we now come to TAM-forms which are encountered with mora da much less frequently.

..  Future and conditional/potentialAs concerns future tense, marked with the proclitic ќе, and the conditional (potential), marked with the proclitic bi (+ l-participle), (non)confirmativity turns out as irrevel-ant. In principle, since the future per se implies uncertainty, one might assume that it associates with epistemic readings. On the other hand, directive illocutions, desirabil-ity and expectations that arise from social norms (obligations) and, thus, belong to the field of deontic meanings, associate with future situations in an equally natural way. The empirical picture shows that mora under the scope of ќе can yield either interpre-tation, and this interpretation cannot be reliably predicted from the choice of pfv. vs. ipfv. aspect. Сf., for instance, (29) with a pfv. verb, but an epistemic interpretation:17

(28) Κ́e mora da dojde utre. → pfv., deontic

fut must con come:pfv.prs.3sg tomorrow ‘He is obliged to come tomorrow.’

(29) No i toj ќe bide na zabavata. Κ́e mora da

fut must con se sretneme. → pfv., epistemic refl meet:pfv.prs.1pl ‘But he as well will be at the reception. We will (thus) meet (there) for sure.’

(30) Toj e lekar. Κ́e mora da znae. → ipfv., epistemic

fut must con know:ipfv.prs.3sg ‘He is a physician. He probably will know/knows.’

Independently from this, it is impossible to interpret forms of the future or the poten-tial if these, conversely, themselves occur under the scope of mora da:

(31) a. with future: *Mora da ќe kupi / kupuva must con fut buy:pfv:3sg / buy:ipfv:3sg intended reading: ‘It must be the case that s/he will buy/be buying’.

.  Admittedly, this example could get a dynamic reading (‘it is inevitable’), too.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

b. with potential *Mora da bi kupil must con cond buy:pfv:l-ptcp.sg.m / kupuval / buy:ipfv:l-ptcp.sg.m intended reading: ‘It must be the case that he would buy/be buying’.

However, Examples (34–36) below demonstrate that it is nonetheless possible to com-bine the future or potential marker with da, but only provided that da does not occur in immediate contact with one of these markers. In fact, the proper problem does not seem to be mora, but rather the connective da. This can already be inferred from the fact that for može ‘can’, which is the possibility equivalent of mora, no such restrictions hold. One should add that može usually does not take da in order to connect with a finite verb, but it anyway combines freely with future and potential forms:

(32) a. Može znae. / Može doznal. / can know:ipfv.prs.3sg can find_out:pfv.l-ptcp.sg.m Može raboti. can work:ipfv.prs.3sg ‘Maybe he knows/has learnt/is working.’ b. Može ќе kupi. / Može bi kupil. can fut buy:pfv.3sg can subj buy:pfv.l-ptcp.sg.m ‘Maybe he will buy/would have bought (or: would buy).’

Note that Može da kupi (‘He can buy.’) does occur, but in this case da does not occur in line with other so-called particles. Combinations of the kind ?Može da ќе kupi; ?Može da bi kupil (in the meanings given under (32b)) are very unusual.

What is left to be explained are the non-confirmative forms with the l-participle of the perfective stem in the imperfect as in (33b), although this combination is certainly very rare.18 These forms do occur in the scope of mora da and have to be interpreted as reportive accounts of other people’s utterances in the pfv. present as in (33a). This rather complicated (though marked and, thus, rare) combination of func-tions and paradigmatic forms neatly inscribes itself into the rule according to which the imperfect transposes the present to a past scene, while the derivative aspect does not change; the only additional element is the marking of non- confirmativity inferred from the l-form:

.  Not all native speakers accept (33b). Instead of da, they would prefer ќe to combine with the l-participle.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

(33) a. Mora da dojdat. must con come:pfv.prs.3pl ‘They must come.’ (deontic) b. Mora da dojd-e-l-e must con come:pfv:impf:l-ptcp:pl ‘It is said that they must come.’ (deontic under reportive scope)

Moreover, the marking of the reportive function can be exchanged between mora (da) and the lexical verb in the same way as we observed it with the imperfect (see (20a–b)). Thus, (33b) is functionally and semantically equivalent to (33c):

(33) c. Mora-l-e da dojdat. must:l-perf:pl con come:pfv.prs.3pl ‘It is said that they must come.’ (deontic under reportive scope)

This possibility of transferring the marking of non-confirmativity to the modal explains why also future forms can occur under an evidential (reportive) scope:

(34) Toj utre ќe mora-l da odi he.nom tomorrow fut must:l-ptcp.sg.m con leave:ipfv.prs.3sg na rabota. to work ‘He is said to be obliged to go to work tomorrow.’

Note that the finite verb in the “complement” (after da) is imperfective, and yet it gets a deontic meaning. This is in line with the observation made already in 2.2.3 that it is not the distinction of pfv.: ipfv. aspect of the lexical verb which serves as a reliable factor (or indicator) of a deontic (vs. epistemic) reading.

The possibility of transferring the marking of non-confirmativity obtains also for the potential, the only difference in comparison to the future being that the eviden-tial marker (l-participle) does not yield a reportive, but a speaker-oriented inferential reading:

(35) Bi mora-l-a da živeše na ovaa ulica. subj must:l-ptcp.sg.f con live:ipfv.impf.3sg on this.sg.f street.sg.f ‘She probably lived in this street.’

(36) Bi mora-l da go znae toa. subj must:l-ptcp.sg.m con it.acc know:ipfv.prs.3sg this ‘He probably knows this.’

After having given a comprehensive analysis of the behaviour and interpretations of mora da in combination with the TAM-forms of Macedonian, let us summarize the findings assembled so far.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

..  Interim summaryAs an interim summary, we may conclude the following:

1. There are a number of factors which either inhibit or favour the epistemic inter-pretation of the verbal complex [mora da + finite verb]. Inhibitive are: (a) non-3rd persons, (b) reference to posterior situations, (c) inflected, or only partially inflected, mora. Favourable is (d) reference to ongoing actions or (resultative) states.

2. These factors are hierarchized, at least as the imperfect and the perfect are concerned. Both favour epistemic meanings, to an extent that these AT-forms dominate over restrictions of grammatical person and, as for the perfect, deriva-tional aspect (pfv.: ipfv.).

3. Derivational aspect is not per se decisive: Thus we find both epistemic interpreta-tions of pfv. verbs and deontic interpretations of ipfv. verbs under the scope of mora da (see especially Section 2.2.2).

4. By and large, it is not so much that pfv. vs. ipfv. aspect correlate with deontic vs. epistemic meaning, as that the inflected vs. petrified (or only partially inflected) form of mora proves decisive for the modal interpretation.

5. Mora da cannot combine with the future or the potential (bi + l-participle), unless one of the latter markers takes scope over mora da (see Section 2.2.4).

These findings are summarized in Table 1.

Table 1. TAM conditions influencing the interpretation of mora da + Vfin

TAM form of lexical verb Favours meaning Remarks

confirmative

present pfv.ipfv.

deonticepistemic

readily overridden by time-referential factors[± iterative, ± progressive]

aoristimperfect

pfv.ipfv.

epistemicepistemic (exclusively!);see Figure 1 in 3.1.

less common

perfect (n/t-suffix or imam) epistemic irrespective of pfv.:ipfv.non-confirmative take scope over modal

operators (constructions)

l-perfect pfv., ipfv. epistemic[± confirmative] distinction irrelevant

future (ќe) pfv., ipfv. possible only if these distinctions themselves scope over mora da; this provided, interpretation is congruent withpfv.:ipfv. present or perfect (anterior)

potential (bi + l-participle) pfv., ipfv.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

From this summary we see that the only context which per se favours the deon-tic interpretation of a modal complex with mora da obtains when the lexical verb is in the pfv. present. The bias toward this interpretation, as well as the bias toward epistemic readings for mora da taking scope (i) over ipfv. lexical verbs in the present tense, (ii) over lexical verbs in the imperfect and (iii) over lexical verbs in all past tense forms which primarily function as anterior grams (+ANT in Thieroff ’s 2000 terms), including a resultative function, can be motivated by aspectual features on clause level. The common motivation rests in a [± bounded] feature: [+ bounded] favours pfv. present, [- bounded] favours the other contexts named. This feature conforms to aspectual–modal divisions, e.g. in English (and other Germanic lan-guages). However, it often proves to be at variance with the grammatical pfv.—ipfv. opposition in Slavic,19 which can largely be related to operators of viewpoint aspect (see 2.1 for a discussion).

The explanation for this apparent mismatch arises from two circumstances: First, pfv. verbs – as representatives of a limitative aspect (cf. Dahl 1985; Breu 1996, 2000, among others) – by definition denote events (understood as undivisible states of affairs) and, if telic, are associated with a focus on the final boundary of the denoted state of affairs. Given this focus, and provided the stem is telic (which is a lexical property), the assertion of having reached the respective natural endpoint of the denoted situation easily leads to the implicature of a subsequent resultative state (compare, John solved the task. ⊃ ‘The task was solved’). In turn, states (both resultative and non-resultative ones) seem to be a condition strongly favourable for epistemic readings. Second, as was already stressed in 2.1, the pf.: ipfv. distinc-tion in Slavic does not boil down to a distinction of [± bounded, ± telic] states of affairs; if it did, aspect would not be a grammatical category, but only lexically conditioned and restricted (as any sort of “mode of events/actions” or Aktionsarten in the indoeuropeanists’ tradition). Instead, pfv. and ipfv. stems also share into different time-referential oppositions which are independent from the feature of boundedness, one of them being the distinction between singulative vs. iterative (habitual, generic etc.) situations. This has been demonstrated above, and we could see that boundedness (and its implicatures) can be ruled out by such independent time-referential factors. From the analysis conducted above, we may induce that the modal distinction (deontic vs. epistemic) is not always highest in a hierarchy of functions influenced by choice of aspect in the scope of mora da.

.  For empirically based discussions of similar observations in Slavic cf. Divjak (2011), among others.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

.  Lexical vs. grammatical status of mora (da)

Up to this moment I have not taken any clearer stance as for whether the two-word form mora + da has to be understood as an analytical complex which can be under-stood as the mere sum of these two components, or whether, from a lexicographic viewpoint, it should rather be regarded as one unit consisting of two elements which, as it were, have “coalesced” semantically. The first step toward answering this question consists in pinning down grammatical restrictions concerning auxiliary complexes with mora (da) (3.1); the second step amounts to demonstrating the background of distributional facts which should be taken into consideration (3.2).

Let me stress that both steps in this analysis are strictly bottom-up and that I do not pretend to be exhaustive about all facets of this complicated issue. I am also aware that the analysis, and the results thereof, expounded below are rather “unorthodox” for many scholars of Balkan Slavic. But I hope that my considerations bring some new facets into the discussion, and that objections which may be raised against my argument will help clarify the way the data should be assessed. This, hopefully, will contribute to a metadiscussion on methodology (yet to follow at a later stage). Let me also stress that however one assesses the pros and cons of my argument regarding the categorial status of mora or mora da, the conclusions concerning the distribution of this unit in combination with different TAM-forms and the modal interpretations these combinations yield (see Section 2) remain valid.

.  Stepwise reduction of verbal features on mora

As we have seen in Section 2, mora can be used as a unit with the marking of gram-matical categories that are usual for Macedonian verbs: tense, person, and number. But we have also come across a lot of examples in which mora shows a reduction in the marking of these categories. This seems to occur in two steps: On the first step mora occurs inflected for tense other than the present (namely, the imperfect), but person+number marking (namely, 3rd person + singular) does not agree with the marking of the same inflectional categories on the lexical verb with which mora has to occur. I have called this ‘partial finiteness’. The second step results in a total loss of inflectional marking on mora, instead the form that corresponds to the bare stem or, otherwise, that looks like the 3sg.pres-form is used. This form may be considered as a default form due to its lack of agreement. This stepwise loss of inflectional catego-ries involves an asymmetric implication saying that loss of tense distinctions (present vs. imperfect) implies loss of agreement and, thus, of syntactically relevant finiteness categories (person + number), but not vice versa: even if agreement marking is lost, tense distinctions can still obtain (see Figure 1). Furthermore, the lexical verb always is marked by the inflectional categories available in Macedonian, except tense. More

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

precisely, the distinction between present and imperfect can be marked on either of the two units, mora or the lexical verb. However, if imperfect is chosen, only one of the two verbs can be marked with this tense.20 This leads to an equivalence of tense marking between the following two types of constructions:

(37) I. mora da + Vfin.imperfect II. mora.imperfect da + Vfin.present

The same applies for non-confirmative marking with the l-suffixed participle (compare (33b) and (33c)): Only one of the two verb stems can be marked for imperfect (or with the l-participle). Apart from that, the lexical verb can occur in any TAM-form under the scope of mora da. In other words: mora (da) and the lexical verb can both be marked with present tense (indicative), but if mora (da) is marked with an ending belonging to another inflectional (sub)paradigm, this form is blocked for the lexical verb.

This equivalence in terms of a past-present distinction is not necessarily paralleled by an equivalence in terms of modal subdomains. Thus, whereas construction type I (with the lexical verb marked by the imperfect) always implies epistemic reading, construction type II, while allowing for epistemic interpretations, rather prefers deontic readings (see Example 26). Moreover, although the more specific condi-tions under which in construction type II an epistemic meaning is available remain to be investigated and stated more clearly (both in terms of a rule mechanism and in frequency terms), it becomes obvious that there is an asymmetry in terms of prefer-ence for deontic vs. epistemic readings.21 This asymmetry corresponds to the findings in Hansen (this volume, 3.1–3.2) based on a systematic areal and genealogical sample of European languages. In terms of his patterns of subject-predicate agreement (see his Table 3) the loss of agreement categories on mora illustrates his type b. As shown above, this loss favours epistemic meaning, though it does not preclude deontic readings either (see (15a, 15c, 21a, 22a)). On the other hand, the pattern in which mora shows agree-ment with the lexical verb (= Hansen’s ‘complement’) corresponds to Hansen’s type c; it strongly correlates with, but is not entirely restricted to, deontic interpretations.22

.  Although this goes beyond the scope of this paper, it seems worthwhile considering whether in such complex predicates with two tensed forms present tense has to be understood as the default and imperfect as the marked choice.

.  As mentioned earlier, in order not to make things too complicated, I refrain from any comments going beyond that binary distinction, such as dynamic (= circumstantial) modality.

.  Hansen‘s classification would predict that agreeing mora (da) shows ambiguity with dynamic readings, i.e. variation within (and restricted to) root modality. This has not been tested here (see previous footnote). However, since an epistemic reading of type-c realisations of mora (da) need not be excluded, the restriction to deontic (or broader, root) modality has remained a strong tendency, but not a strict rule.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

The results of the bottom-up analysis of mora da verbal complexes presented above also confirm Hansen’s observation that “tense and subject-predicate agreement are independent of each other” (Hansen, this volume, p. 106f.); this independence explains why inflectional categories of the modal marker can be lost stepwise. What however cannot be demonstrated on mora (da) is Hansen’s claim that “modals with non- canonical subject marking tend to show a lower degree of polyfunctionality” (Hansen, this volume p. 99). The reason is simply that Macedonian does no longer have any morphological cases on nouns by which non-canonical subjects might be distinguished; likewise no prepositional substitutes are encountered. As concerns null subjects, we have to distinguish null subjects implying human beings (as with Polish trzeba ‘must, is necessary’ or można ‘can, is possible’, cf. Hansen’s Examples 10a, 14a) and null subjects caused by zero-place argument structure (e.g. with verbs like rain, snow) or in existential constructions (like there’s plenty of). Only the former are treated by Hansen, and he clearly considers them to be restricted to root modality. By con-trast, null subjects of zero-place predicates are more susceptible to epistemic meanings (see 38a–b), although dynamic usage is imaginable and encountered, too. Crucially, this distinction does not unequivocally depend on grammatical aspect (compare 38c and 38d):

(38) a. Mora da vrne / grmi. must con rain:ipfv.prs.3sg / thunder:ipfv.prs.3sg ‘It must be raining/thundering.’ b. Mora da vrnelo / grmelo. must con rain:ipfv.l-perf.n / thunder:ipfv.l-perf.n ‘It must have rained/thundered.’ c. Mora da vrne, inaku semeto nema da porasne. must con rain:ipfv.prs.3sg ‘It must (= needs to) rain, otherwise the seed will not grow.’ d. Mora da must con zavrne sneg, inaku ne ќe možeme da odime na skijanje. rain:pfv.prs.3sg snow ‘It must (= needs to) snow, otherwise we cannot go skying.’

All this said, the observed tendencies amount to a hierarchy in the relationship between morphosyntactic marking and modal (i.e. semantic) interpretation; this hier-archy is shown in Figure 1. The arrow indicates the increase in the likeliness that an epistemic reading arises (as opposed to the decrease thereof with respect to a deontic reading); it protracts from the upper-left corner with conditions for an exclusively

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

deontic reading toward the bottom-right corner at which conditions for an exclusively epistemic reading are grouped.

Deontic

(i) mora inflected for tenseand person+number (can beseen only in forms other thanthird singular)

(No epistemic meaning possible, unless lexical verb is in the imperfect.)

(ii) mora inflected only for tense23

(i.e. no marking of syntacticcategories)

Epistemic reading not preferred, but possiblydepending on other factors (see below).

(iii) mora in prs.3sg-form−−> [±inflected] cannotalways be decided upon(see ex. 13–14, 28, 30)

Epistemic24 meaning preferred or the only possibleone, if one of the following factors obtains:(a) ipfv. aspect of lexical verb,(b) anterior or simultaneous time reference,(c) lexical verb in the imperfect. Epistemic

Figure 1. Hierarchy of conditions on a cline between deontic and epistemic readings of mora da + Vfin23,24

.  As one of the reviewers persistently indicated, one could treat moraše as an equally uninflected unit (i.e. in parallel to mora as isolated from the present tense paradigm). However, in this case one would have to claim that there are petrified forms (particles?) showing traits of inflection. This sounds rather unusual. Moreover, even if analogy may be an argument in favour of this treatment, it would leave us with another problematic consequence: Suppose moraše as an uninflected unit is becoming (or has become) an epistemic modifier (with propositional scope), how can it then signal the speaker’s judgment pertaining to his/her ongoing act of speech? I am unaware of epistemic modifiers which (outside of narrative contexts, for which Reference Time and Topic Time differ) are used in a form inflected for past tense (or a marked mood). I do not know of any case where an epistemic (or evidential) particle or adverb devel-opped from a verbal form other than the present tense. Is Macedonian an exception? I think the burden of proof for this (implicit) assumption falls on my reviewer. Apart from all that, the reviewer’s proposal makes it more complicated to compare Mac. mora/moraše to a broader European background (as referred to above following Hansen, this volume). In sum, thus, there are three arguments against the reviewer’s alternative assumption, whereas my own proposal is in line with areal clines of epistemic modifiers, and it appears to be entirely compatible with the observable facts of Macedonian itself.

.  This is to include inferential readings (by implicature?), which have however been left uninvestigated. The mutual relationship between epistemic modal and inferential readings should first be clarified more thoroughly on a principled basis. Anything which proves to obtain for this relationship on a crosslinguistic basis should hold for the “Macedonian case” , too.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

Factors (a–c), favourable to an epistemic reading, are ordered among each other for their strength: (a) is the weakest, (c) the strongest factor (of all factors listed in Figure 1). In the overall hierarchy the following one-sided implications hold:

(i) inflection for person+number ⊃ (ii) inflection for tense (c) imperfect on lexical verb ⊃ epistemic meaning (overrides even (i)) (c) imperfect on lexical verb ⊃ only present tense inflection of mora.

These implications can even be considered as entailments, since they are either logi-cally true (the first one) or cannot be overridden by oher factors.

Finally, the observations made above make us inclined to conclude that non- confirmative marking (with the l-participle) takes semantic scope over any sort of modal semantics. This issue should be analysed more systematically on another occa-sion. The same holds for the future, which is marked with uninflected ќe. We have seen that ќe can occur only before mora (da), not after it (i.e. not with the lexical verb), and that this order iconically reflects semantic scope: ќe modifies the entire proposition, including the modal (see 2.2.4).

.  Other distributional facts

However, for the time being we still have to consider (a) whether mora behaves as an auxiliary or a particle, and (b) which role is fulfilled by da. As for issue (a) we should recall that an auxiliary does not have an argument structure on its own, but shares one with the lexical verb it scopes over syntactically (cf. Hansen 2001: 94f.; Kuteva 2001; Krug 2011). An auxiliary is furthermore a unit which marks some sort of rather abstract grammatical meaning as well as syntactic categories of the predicate. This brings it close to functional lexemes like prepositions or conjunc-tions, but also to particles or sentence adverbs. However, a particle, in turn, does not enter into constituent relations with its syntactic environment, neither does it form a constituent together with another element (e.g. a lexical verb). Above we have observed that in certain environments mora loses inflectional categories, in particu-lar it often “passes over” agreement and tense marking to the lexical verb, and that this behaviour is strongly associated with epistemic readings (see in particular (20b, 21b, 22b)). Therefore, as far as inflectional categories are concerned, mora does not behave like an auxiliary. But nothing precludes its analysis as a (part of a) particle or sentence adverb.

As concerns issue (b), we have to account for the fact that the necessity marker mora never occurs without da (contrary to its possibility equivalent može; see 32a–b). We might therefore infer that mora+da forms one single unit, both in functional ( morpho-syntactic) and lexicographic terms, and can no longer be analysed as a free collocation. In other words: There is ground to assume that mora da has no

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

transparent compositional meaning, i.e. cannot be read just as the sum of its compo-nents. On first sight, there seems to be one property that allows arguing against such a view: mora da can be separated in the linear unfolding of an utterance; thus, between mora and da not only manner adverbs, but even the subject-NP can be inserted.25 For subject-NPs see (22a–b) and (39b) vs. (39a):

(39) a. Marija mora da dojde. pn must con come:pfv.prs.3sg b. Mora Marija da dojde. must pn con come:pfv.prs.3sg ‘Marija must come.’ c. *Mora da Marija dojde. must con pn come:pfv.prs.3sg

For manner adverbs see (40a–b):

(40) a. Mora brzo da došol. must quickly con come.pfv.l-ptcp:sg.m ‘He must have come quickly.’ b. Mora brzo da dojde. must quickly con come.pfv:prs.3sg ‘S/He must come quickly.’ c. *Mora da brzo dojde. must con quickly come.pfv:prs.3sg

On the one hand, Examples (39a–b) show that the subject-NP of the (purported) auxiliary complex can be situated before or after mora, but it cannot follow da (see 52c). Likewise (40c) is ungrammatical, because da is separated from the lexical verb. On the other hand, we ought not to dismiss the fact that mora cannot exist without the morpheme da. How can we reconcile these seemingly contradictory facts?

To begin with, a split in linear terms does not say much about the internal coher-ence between mora and da. For da belongs among the many proclitics of Macedonian (like bi, ќe, inflected present-tense esse and others). In combination with a subsequent finite verb (= carrier of lexical meaning) this will always lead to a closer suprasegmental bond between these two elements, but not with the element preceding da. Such supra-segmental features however do not supply any strong argument in favour of internal or external constituency. Second, languages betray many rules by which components

.  This linear requirement strikingly contrasts with the behavior of object pronouns used in clitic doubling: such pronouns appear only after da (see (16, 22a–b, 25, 36)). For the present argument it is not essential that subject-NPs which are located between mora and da consti-tute an emphactic focus (as one of the reviewers has indicated to me).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

belonging to the same constituent – or even units in lexicographic terms  – can be “split”. Suffice it to think here, for instance, of the German “ Verbklammer” caused by the V2-rule for main declarative clauses with analytical predicates (e.g. the werden-future or the haben/sein-past; cf. Zifonun et al. 1997: 1500–1502). Third, linear order does not necessarily reflect semantic scope: in (40a–b) the adverb (brzo ‘quickly’) modifies only the lexical verb (došol ‘(he) has come’, dojde ‘comes’), not the whole complex with mora; in other words: it does not have propositional, but only predica-tional scope. The same applies to the subject-NP in (39a–b).26 By the same token, the scope of the adverb brzo does not comprise (the function of) da, i.e. it does not modify da došol or da dojde. If it did the utterance could be understood only as an adhortation (‘May/Should he come quickly!’; for main clause usage see below) or yield the rather nonsensical reading #‘May it quickly be(come) that he comes!’. In any case we would face difficulties in explaining, first, how the hortative meaning would be “linked“ as an input to the modal function of mora (which anyway has wider scope), and, second, how da all of a sudden changes its adherence from inflected došol to mora, or how it can become a modifier of both verb forms: Again, since mora cannot go without da, we must relate it to the modal element. At this point we may turn upside down the initial argument based on the position of the adverb (brzo): Should we infer that brzo and došol are not related in constituency, because the element da separates them in linear order? (The same would apply to the subject-NPs and the lexical verb in (42, 44) below.) As we see now, an argument that might initially be forwarded against treating mora da as a single unit turns to be a double-edged sword. Anyway, we are left with the fact that mora cannot be used without da, and that after mora no other connecting morpheme can replace da.

On the other hand, pfv. present tense forms (like dojde) cannot occur in main clauses without a marker of reality (factivity) status, such as da (see 2.2.2 and below). Consequently, da cannot leave dojde alone, otherwise the utterance becomes ungram-matical. In view of this puzzle, we have to ask which kind of analysis would allow da to simultaneously modify mora and dojde (or došol). In fact, da behaves like a janus-faced pivot on the level of nuclear juncture. Anyway, whatever behaviour this unit may show elsewhere, in the syntactic context considered so far, da cannot be regarded merely as a complementiser.

One cannot deny that in Macedonian (as well as in other South Slavic languages) da is used as a complementiser (first of all, with verbs of speech or with negated

.  The same obtains for utterances in which da really functions as a complementizer (and the subject-NP of the complement clause comes between the complement-taking predicate and da), as in (41).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

verbs of perception; see below). However, da is probably the most “versatile” unit of Macedonian as far as its syntactic behaviour and lexicographic status are concerned. It therefore would be bold and inadequate to try to ascribe to it only one function. In the literature, the following usage types of da (both in syntactic and in semantic terms) have been figured out:27

(1) complementiser (in a wide sense): (1a) “object” complements with verba dicendi and verba voluntatis.

(41) Nareduvam Marija da dojde vednaš. order:ipfv.prs.1sg pn comp come:pfv.prs.3sg immediately ‘I demand that Marija come immediately.’

(1b) “subject” complements of predicates marked with the reflexive clitic se (–> Actor deranking, (42)) or predicatives (deriving from adjectives or nouns) with evaluative meanings (arno ‘good’, nezgodno ‘inconvenient’, ubavo ‘beautiful’; sramota ‘shame’, grevota ‘sin’, etc.; see (43)):

(42) Se soobštuva da se napušti gradot. rm inform:ipfv.prs.3sg comp rm leave:pfv.3sg town.defart.m lit. ‘(It) is told that one leaves the town.’ i.e. ‘It is issued that everybody leaves the town.’

(43) Arno e da molčiš. good.n cop:prs.sg comp be_silent:ipfv.prs.2sg ‘It is good that you are keeping silent.’

(1c) adnominal (i.e. attributive) use, e.g. in nominal derivatives of rm-predicates as in (44):

(44) Soobštenieto da se napušti gradot message.defart.n comp rm leave:pfv.prs.3sg town.defart.m me iznenadi mnogu. me.acc surprise:pfv.aor.3sg much ‘The message that people have to leave the town surprised me a lot.’

.  I do not claim this list to be exhaustive. It is based mainly on Georgievski (2009), but cf. also Feleško (1974), Kramer (1986: 20–29), Topolinjska (2000: 90–104) and Mišeska Tomić (2006: 416–456, 2012: 357–377).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

(2) conjunction, i.e. as a linkage device for adverbal adjuncts: final, concessive, conditional (irreal and potential), consecutive. Compare the following example of a conditional sentence:28

(45) Da imam pari, ќe si kupam kola. conj have:ipfv.prs.1sg money fut rm.dat buy:pfv.prs.1sg car.pl ‘If/When I have money, I will buy myself a car.’ (3) with phasal verbs:29

(46) Počna ga plače. start:pfv.aor.3sg him.acc cry:ipfv.prs.3sg ‘S/He started morning because of him.’

(4) finally, with auxiliaries: može, mora etc. (structures može/mora da + Vfin). See Section 2 of this article.

Apart from that, da occurs clause-initially in non-embedded structures with a horta-tive (47) or optative function, or in suggestive questions (48):

(47) Da gi prečekate! con them.acc wait:pfv.prs.2pl ‘You should wait to welcome them!’

(48) Da ne si nešto bolen? con neg cop:prs.2sg something sick.sg.m ‘Aren’t you somewhat sick?’

It is important to stress that such main clause phenomena30 are always restricted to non-factive and non-declarative utterances. They have been labelled differently, not only with respect to Macedonian, but on a more general Balkan and South Slavic back-ground. For a survey of the classical analyses cf. Feleszko (1979). In the last decades two main positions can be opposed to one another. One position is represented by the generative-fashioned approach in Mišeska Tomić (2006: 439–444; 2012: 370). She calls the usage of da in syntactically independent clauses ‘bare subjunctives’, assuming “a

.  Since it seems to occur especially frequently in the protasis of conditional sentences, some researchers consider it a conditional marker per se.

.  Feleško (1974: 138) counts them among complements (“intensionalni rečenici”). A similar classification is found in Topolinjska (2000: 90), who aligns phasal verbs not only with modal auxiliaries, but even with volitional verbs having a clausal complement with no coreference requirement (e.g. Sakami da go vidišk ‘Ii want that youk see/meet him’).

.  Quite recently linguists have begun to treat such phenomena as ‘insubordination’ – an issue I cannot and need not go into here, since to the status and functions of mora (da) it is related only indirectly.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

covert existence of a main clause expressing a command or a wish”. The other position is taken, for instance, by Topolińska (2008b [1994]: 175f.). Originally, she treated da as “a formant of the so-called optative mood, with a broad scale of semantic variants, all of them with future temporal perspective“.31 However, lately she has stressed that dependent clauses preceded by da in Macedonian have to be subsumed under the label of ‘subjunctive’ in a strict sense (Topolinjska 2008a: 67). If we want to find a characteristic of da covering all its usage types we might say that it marks non-factivity (Topolińska 2008b [1994]: 175f.). The common denominator of both research posi-tions appears to be that both argue for regarding da as a marker of a sort of analytical subjunctive (whose rise should be understood in connection with the successive stages of loss of the infinitive in Balkan languages).

There is no need to enter now into the inherent links between the different usage types encountered with Mac. da. Let us instead return to the issue of da’s status in collocation with mora. Georgievski (2009: 9 et passim) identifies the occurrence of mora (and može etc.) with finite verbs as a verbal complex, i.e. one single predicate. This is congruent with the treatment in Hansen (this volume) and Hansen & de Haan (2009), who locate modal constructions “on a continuum between bi-clausal and mono-clausal structures” (Hansen & de Haan 2009: 526).

At this place we may subsume that the janus-faced behaviour of da can be observed not only in auxiliary complexes (as with mora), but also when da serves as a complementizer. This can be illustrated by a comparison of dependency relations, as in (39a–b) and (41), respectively. These examples are here repeated for convenience:

(39) a. Marija mora da dojde.

pn must con come:pfv.prs.3sg

b. Mora Marija da dojde.

must pn con come:pfv.prs.3sg‘Marija must come.’

(41) Nareduvam Marija da dojde vednaš.

order:ipfv.prs.1sg pn comp come:pfv.prs.3sg immediately‘I demand that Marija come immediately.’

.  Topolińska mentions this main clause usage as a third function beside da’s functions as a “basic marker of the subjunctive mood” and an “adverbial particle”. By the latter she means conjunctional use, e.g. in final clauses (Topolińska 2008b [1994]: 175f.).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

We see that irrespective of the syntactic function of da, and regardless of the level of constituency (i.e. whether the structure is mono- or biclausal), there is an intermin-gling of the sequence of two structures: between the subject-NP and the verb it belongs to (Marija dojde in both cases), on the one hand, and between da and the verbal unit to which it is related itself (the auxiliary or the complement-taking predicate), on the other hand. In other words: In any case da causes an interruption in the linear order-ing of verb-argument structure. In extreme cases, namely with pfv. verbs in the present tense, da cannot be “detached” from any of the verb forms it is related to without leaving the remaining part ungrammatical. In terms of constituency, for the biclausal structure in (41) this means that da can be treated as belonging exclusively neither to the complement-taking predicate (nareduvam) nor to the predicate of the comple-ment clause (dojde). An analogous conclusion holds for the monoclausal structure with the modal mora in (39a–b), but already “one level lower”, i.e. within one and the same constituent (lexical verb + modal modifier). On this level the scope of da is much narrower than in biclausal structures, and as a consequence it becomes impossible to analyse mora+da (+ finite lexical verb) as a combination of two distinct units; one must assume that semantic coalescence has occurred.

Now, remember that this coalescence is tightly correlated with the use of mora da in epistemic readings. And here we can even go one step further. As a semanti-cally inseparable collocation mora da becomes syntactically ambiguous in another respect: on the one hand, it can be analysed as a modal auxiliary, although this modal then occurs to be more defective than its equivalents in many other European languages (among others, Slavic ones), which have retained their “verbiness” to a fuller extent.32 On the other hand, almost nothing precludes the analysis of mora da as a particle, i.e. a single, lexicalized unit. The only issue being at variance with such an analysis is the fact that other material can be inserted between mora and da (see above).

Therefore, mora da manifests an ideal case of an intermediate stage of devel-opment from a modal construction toward an epistemic particle. Such a stage cor-responds to the transition between stages IV and VI investigated for Russ. možet byt’ ‘(i) (it) may be (+ comp) vs. (ii) maybe, probably’ in Hansen (2010). If, slightly modifying Hansen’s (2010) constructional frames, we claim the existence of the

.  As one of the reviewers remarked, the equivalent modal of Bulgarian, trjabva (da), is even more defective, since it never inflects at all when used as a modal verb (also with non-epistemic functions). However, this is true only for person-number marking, since the imperfect and the l-form are used (trjabvaše, trjabvalo). It thus corresponds to stage (ii) of Mac. mora (da) in Figure 1.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

three construction types in Figure 2, Mac. mora da illustrates a stage that, as it were, oscillates between (b) and (c):33

(a) (epistemic)[NEC34]clause1+ [daCOMP+ Vfin]clause2 biclausal (NEC = moraas an independent predicate)

(b) (epistemic) [(NEC + {da) + Vfin}]clause

modal modifier (mora da) outsideof constituent structure (= particle)

monoclausal (mora and Vfin shareinto a common nucleus united by da)

(c) (epistemic) (NEC+da) + clause

Figure 2. From biclausal to monoclausal modification and beyond34

There are only two, rather superficial, differences in comparison with Russ. možet byt‘: First, the second element of the source construction is not an infinitival copula (byt’), but it originates from a complementizer. Second, this second element does not disappear when the modal starts behaving as an epistemic sentential adverb, whereas in Russian or Polish the complementizer čto and że, respectively, can in some sense be considered as having vanished (Russ. možet byt’ > možet, Pol. może być > może). In Macedonian this second element stays where it has been, so that mora and da together are to be treated as a case of ongoing conversion (in the sense of Ramat & Ricca 1998, see below). With respect to this second issue, but not to the first one, Mac. mora da is comparable with cases like Engl. may be > maybe.35

However, the really crucial point is that mora+da at its present stage allows for an analysis as either (b) or (c). One cannot decide whether it already has been reanalysed toward stage (c) or still should be qualified as representing stage (b).

Hansen (2010) describes units that have reached stage (c) (= his stage VI) as sen-tence adverbs; equally well we might speak of (epistemic) particles. We may leave the question open whether (modal) particles and sentence adverbs can be distinguished on semantic and/or syntactic grounds. Crucial are the properties which they share with each other and which set them off from auxiliaries (or auxiliary constructions),

.  According to Hansen (2010: 83), the structure on stage V (= structure (b) in Figure (2) appears to be syntactically ambiguous because “it can either be interpreted as a[n] asyndetic complex sentence (…) or as a simple clause containing a parenthetical”. The final stage VI (= structure (c)) is reached when “we get a mono-morphemic element”.

.  NEC = Necessity operator (mora), which is per se indifferent as for the modal subdomain (deontic, dynamic, epistemic).

.  There is also an exact equivalent of Engl. maybe in Mac. možebi ‘maybe, perhaps’ (< može bi ‘can.prs.3sg + cond.ptc’).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

first of all from modal auxiliaries (as treated in Hansen 2001, 2010, this volume; Hansen & de Haan 2009). Thus, whatever applies to modal particles can, for the time being, be considered to hold for sentence adverbs as well (and vice versa). Important are the following points:

(i) Epistemic particles do not have an argument of their own, nor do they share any with a lexical verb. True enough, sometimes sentence adverbs (particles) are acknowledged as opening up a syntactic slot filled by a clause (cf. Hansen 2010: 77). However, this “valency” cannot be considered part of the constituent structure, nor is it to be regarded as a syntactic property. Rather it should be con-sidered a semantic one.

(ii) Epistemic particles often arise either through univerbation, i.e. the coalescence of two formerly morphologically independent elements, or through conversion, in both cases typically centering around a modal (cf. Hansen 2010: 75 and this volume, after Ramat & Ricca 1998: 234–239). If Mac. mora da is analysed as an epistemic particle it represents a case of univerbation (as can, e.g. Engl. maybe < may be).

.  Conclusions

The preceding discussion has surveyed distributional restrictions of the modal mora with respect to the verbal categories of Macedonian, which can be noticed in its col-location with lexical verbs. These distributions have been analysed without an account of frequency and discourse-conditioned usage patterns. Future research should dwell upon this issue with the aid of corpora. Moreover, in-depth studies are needed in order to verify (or modify) the findings formulated above (and resumed below).

Nonetheless, we can confidently state the following. The modal mora is tied to the former complementizer da, inasmuch as no other element can be chosen as an element that connects mora with a lexical verb to form a complex predicate, nor can da be left out. In this respect mora differs from its possibility equivalent može ‘can’. Moreover, the epistemic interpretation of mora shows a clear association with its petrified ( uninflected, “impersonal”) form which coincides with the prs.3sg of the verb. We have thus three properties that are strongly correlated: (i) a tight link with the clitic da, (ii) increasing isolation of the prs.3sg-form from the verbal paradigm, (iii)  epistemic interpretation (vs. deontic and, probably, dynamic). The correlation of epistemic meaning with the ipfv. aspect of the lexical verb turns out as being less strong than property (ii), together with the choice of tense of the lexical verb (e.g. imperfect ruling out practically all other factors). The hierarchy of factors that allows for a kind of algorithm able to predict the epistemic vs. deontic interpretation of mora da + lexical verb was summarized in Figure 1 (Section 3.1).

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

Furthermore, the three properties (i–iii) mentioned above condition a devel-opment of mora+da toward a lexicalized, i.e. holistic unit. There is no problem in analysing this unit as a sentential adverb (or particle), i.e. as a new lexical unit with propositional scope. In view of this, we have two alternatives: either we could speak about two constructions in which mora takes part, or we could say that there are (at least) two units mora: mora1 represents the paradigmatic, inflected verb form, while mora2 is the uninflected, paradigmatically isolated form. Whatever alternative we choose, one should add that in any case mora coalesces with da, i.e. can no longer be separated from it at least in terms of dependency relations. Thus, from a lexicographic viewpoint mora da should be treated as one unit.36

What is most interesting is that, at the present stage of the evolution of mora da, one comes across cases in which it is not possible to decide without doubt whether mora da is to be taken as a propositional modifier with adverb (particle) status or still as an auxiliary. Actually, this happens when mora does not show signs of inflection and collocates with a finite verb in forms other than prs.3sg. This potential ambiguity may cause difficulties for analysis of the syntactic (constituent) structure, but first of all this fact may show that we are observing a transitional stage in the development of a petrified modal of verbal origin toward a propositional (sentential) adverb.

Looked at from the diachronic origin of the involved morphemes, mora da dif-fers both from cases like Polish może ‘perhaps (< s/he, it can)’ in which no second morpheme has coalesced with the former auxiliary, and from cases like English maybe (< may+be) in which this second morpheme originates from a copular verb. To which extent the combination of an earlier modal auxiliary with a former complementizer is to be understood as an areal phenomenon restricted to the Balkans (or South Slavic) has to be left for further typological studies. However, what cannot be denied is that the development of mora da toward an epistemic sentential adverb (particle) has to be judged as but a particular case on the background of two more general phenomena typical of the Balkans: (a) the increase of frequency and dispersion of clause linking techniques with finite complements in both bi- and monoclausal structures, and (b) the choice of connective morphemes that suspend factivity on various levels of con-stituency, namely the following ones: (i) biclausal → complementizers, (ii) nuclear juncture (in auxiliary or phasal verb complexes), (iii) monoclausal “insubordination” (with hortative or optative function). Da is involved at all these levels.

.  Other researchers have already argued for a differentiated treatment of units shaped da (for instance, Kramer 1986: 20–65 and Čašule 1989: 94, 98f.), although with different conclusions on the facts. But, as far as I can see, nobody has attempted to distinguish different lexical units in combination with mora (or other modals). At least, Čašule (1989: 94, 101, 107f.) stressed that mora cannot occur without da and that da has become a purely connective element.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

Further research should work out the exact processes (and their chronology) which are responsible for this areally salient feature and how they interact with more universal tendencies in the development of epistemic-evidential (sc. propositional) modifiers. As for the hierarchy in Figure 1, it could easily be transformed into an algo-rithm and should be considered as a tentative heuristic to be falsified. Falsification (or verification) can be achieved only as result of more fine-grained in-depth studies which are (a) based on corpora and (b) look for more systematic variation of ( combinations of) factors and the frequencies of their cooccurence.

Abbreviations in glosses (not covered by the Leipzig Glossing Rules)

aor aoristcon connective element (in predicate complexes)defart definite articleimpf imperfectl-ptcp l-participle (in old perfect and its evidential extension)n/t-ptcp n/t-participle (in “second” perfect)n/t_indecl uninflected n/t-participle (with neuter ending)nec necessity operator

References

Abraham, Werner. 2008. On the logic of generalizations about cross-linguistic aspect-modality links. In Modality-aspect Interfaces. Implications and Typological Solutions [Typological Studies in Language 79], Werner Abraham & Elisabeth Leiss (eds), 3–13. Amsterdam: John Benjamins.

Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford: OUP.Aronson, Howard I. 1977. Interrelationships between aspect and mood in Bulgarian. Folia

Slavica 1(1): 9–32.Boye, Kasper. 2012. Epistemic Meaning: A Crosslinguistic and Functional-cognitive Study [Empir-

ical Approaches to Language Typology 43]. Berlin: De Gruyter Mouton.Breu, Walter. 1996. Komponentenmodell der Interaktion von Lexik und Aspekt. In Slavistische

Linguistik 1995 [Slavistische Beiträge 342], Wolfgang Girke (ed.), 37–74. München: Sagner.Breu, Walter. 2000. Zur Position des Slavischen in einer Typologie des Verbalaspekts: Form,

Funktion, Ebenenhierarchie und lexikalische Interaktion. In Walter Breu (ed.), Probleme der Interaktion von Lexik und Aspekt (ILA), 21–54. [Linguistische Arbeiten 412].Tübingen: Niemeyer.

Breu, Walter. 2007. Der Verbalaspekt im Spannungsfeld zwischen Grammatik und Lexik. Sprachwissenschaft 32(2): 123–166.

Čašule, Ilija. 1989. Modalnite glagoli vo makedonskiot jazik. Prilozi MANU/OLLN XIV-2: 89–117.

Dahl, Östen. 1985. Tense and Aspect Systems. Oxford: Blackwell.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Mora da as a marker of modal meanings in Macedonian

Divjak, Dagmar. 2011. Predicting aspectual choice in modal constructions: A quest for the Holy Grail? In Slavic Linguistics in a Cognitive Framework, Marcin Grygiel & Laura A. Janda (eds), 67–85. Frankfurt: Peter Lang.

Feleško, Kazimjež [=Feleszko, Kazimierz]. 1974. Od problematikata na takanarečenata kon-strukcija so da (Nekolku prašanja za diskusija). Makedonski jazik XXV: 137–144.

Feleszko, Kazimierz. 1979. Wokół południowosłowiańskiego coniunctivu. Studia z filologii polskiej i słowiańskiej 18: 185–192.

Friedman, Victor. 1999. Proverbal evidentiality: On the gnomic uses of the category of status in languages of the Balkans and the Caucasus. Mediterranean Language Review 11: 135–155.

Friedman, Victor. 2000. Confirmative/Nonconfirmative in Balkan Slavic, Balkan Romance, and Albanian, with additional observations on Turkish, Romani, Georgian, and Lak. In Eviden-tials in Turkic, Iranian and Neighbouring Languages [Empirical Approaches to Language Typology 24], Lars Johanson & Bo Utas (eds), 329–366. Berlin: Mouton de Gruyter.

Friedman, Victor. 2003. Evidentiality in the Balkans with special attention to Macedonian and Albanian. In Studies in Evidentiality [Typological Studies in Language 54], Alexandra Y. Aikhenvald & Robert M.W. Dixon (eds.), 189–218. Amsterdam: John Benjamins.

Friedman, Victor. 2004. The typology of Balkan evidentiality and areal linguistics. In Balkan Syntax and Semantics [Linguistik Aktuell/Linguistics Today 67], Olga Mišeska Tomić (ed.), 101–134. Amsterdam: John Benjamins.

Georgievski, Georgi. 2009. Da-rečenicata vo makedonskiot jazik. Skopje: Institut za makedonski jazik ‘Krste Misirkov’.

Hansen, Björn. 2001. Das slavische Modalauxiliar. Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen [ Slavolinguistica 2]. München: Sagner.

Hansen, Björn. 2010. Constructional aspects of the rise of epistemic sentence adverbs in Russian. In Diachronic slavonic syntax. Gradual Changes in Focus [Wiener Slawistischer Almanach, Sonderband 74], Björn Hansen & Jasmina Grković-Major (eds), 75–86. München: Sagner.

Hansen, Björn & de Haan, Ferdinand. 2009. Concluding chapter: Modal constructions in the languages of Europe. In Modals in the Languages of Europe. A Reference Work [Empirical Approaches to Language Typology 44], Björn Hansen & Ferdinand de Haan (eds), 511–559. Berlin: Mouton de Gruyter.

Jakobson, Roman. 1971[1967]. Shifters, verbal categories, and the Russian verb. In Selected Writ-ings II, Roman Jakobson, 130–147. The Hague: Mouton.

Johanson, Lars. 2000. Viewpoint operators in European languages. In Tense and Aspect in the Languages of Europe [Empirical Approaches to Language Typology, eurotyp 20–6], Östen Dahl (ed.), 27–188. Berlin: Mouton de Gruyter.

Klein, Wolfgang. 1994. Time in Language. London: Routledge.Klimonow, Vladimir D. & Gerda Klimonow. 2008. The connections between modality, aspec-

tuality, and temporality in Modern Russian. In Modality-aspect Interfaces. Implications and Typological Solutions [Typological Studies in Language 79], Werner Abraham & Elisabeth Leiss (eds), 147–173. Amsterdam: John Benjamins.

Koneski, Blaže. 1976. Gramatika na makedonskiot literaturen jazik. Skopje: Kultura.Kramer, Christina Elizabeth. 1986. Analytic modality in Macedonian [Slavistische Beiträge 198].

München: Sagner.Krug, Manfred. 2011. Auxiliaries and grammaticalization. In The Oxford Handbook of Gram-

maticalization, Heiko Narrog & Bernd Heine (eds) 547–558. Oxford: OUP.Kuteva, Tania. 2001. Auxiliation: An Enquiry into the Nature of Grammaticalization. Oxford: OUP.

© 2014. John Benjamins Publishing CompanyAll rights reserved

Björn Wiemer

Lehmann, Volkmar. 1999. Aspekt. In: Handbuch der sprachwissenschaftlichen Russistik und ihrer Grenzdisziplinen [Slavistische Studienbücher, Neue Folge 8], Helmut Jachnow (ed.) 214–242. Wiesbaden: Harrassowitz.

Lehmann, Volkmar. 2009. Aspekt und Tempus. In Die slavischen Sprachen: Ein internationales Handbuch zu ihrer Struktur, ihrer Geschichte und ihrer Erforschung, 1. Halbband [Hand-bücher zur Sprach- und Kommunikationswissenschaft 32(1)], Sebastian Kempgen, Tilman Berger, Karl Gutschmidt & Peter Kosta (eds), 526–556. Berlin: de Gruyter.

Mišeska Tomić, Olga. 2006. Balkan Sprachbund Morpho-syntactic Features [Studies in Natural Language and Linguistic Theory 67]. Dordrecht: Springer.

Mišeska Tomić, Olga. 2012. A Grammar of Macedonian. Bloomington IN: Slavica.Plungian, Vladimir A. 2001. The place of evidentiality within the universal grammatical space.

Journal of Pragmatics 33: 349–357.Plungian, Vladimir A. 2010. Types of verbal evidentiality marking: An overview. In Linguis-

tic realization of Evidentiality in European Languages [Empirical Approaches to Language Typology 49], Gabriele Diewald & Elena Smirnova (eds), 15–58. Berlin: Mouton de Gruyter.

Ramat, Paolo & Ricca, Davide. 1998. Sentence adverbs in the languages of Europe. In Adver-bial constructions in the languages of Europe [Empirical Approaches to Language Typology, eurotyp 20–3], Johan van der Auwera & Dónal Pádraig Ó. Baoill (eds), 187–275. Berlin: Mouton de Gruyter.

Smith, Carlotta. 1997. The Parameter of Aspect, 2nd edn [Studies in Linguistics and Philoso-phy 43]. Dordrecht: Kluwer.

Tatevosov, Sergej G. 2002. The parameter of actionality. Linguistic Typology 6: 317–401.Thieroff, Rolf. 2000. On the areal distribution of tense-aspect categories in Europe. In Tense and

Aspect in the Languages of Europe [Empirical Approaches to Language Typology, eurotyp 20–6], Östen Dahl (ed.), 265–305. Berlin: Mouton de Gruyter.

Topolinjska [=Topolińska], Zuzana. 2000. Polski—makedonski. Gramatička konfrontacija 3: Stu-dii od morfosintaksata. Skopje: MANU.

Topolinjska [=Topolińska], Zuzana. 2001. Polski—makedonski. Gramatička konfrontacija 5: Zdanie w zdaniu. Skopje: MANU.

Topolinjska [=Topolińska], Zuzana. 2008a. Polski—makedonski. Gramatička konfrontacija 8: Razvitokot na gramatičkite kategorii. Skopje: MANU.

Topolińska, Zuzanna. 2008b. Factivity as a grammatical category in Balkan Slavic and Balkan Romance. In Z Polski do Macedonii: Studia językoznawcze, problemy predykacji 1, Zuzanna Topolińska, 173–184. Kraków: Lexis. Reprint from: Slavia Meridionalis 1 (1994), 105–121.

Van der Auwera, Johan & Plungian, Vladimir A. 1998. Modality’s semantic map. Linguistic Typology 2: 79–124.

Wiemer, Björn 2011. Grammaticalization in Slavic languages. In The Oxford handbook of gram-maticalization, Heiko Narrog & Bernd Heine (eds), 740–753. Oxford: OUP.

Zifonun, Gisela, Hoffmann, Ludger & Strecker, Bruno. 1997. Grammatik der deutschen Sprache, 3 Vols. [Schriften des Instituts für Deutsche Sprache 7]. Berlin: de Gruyter Mouton.