Presumptive mood , factivity and epistemic indefinites in Romanian

This paper investigates the properties of the Romanian presumptive mood and the role it plays in the distribution of the epistemic indefinite vreun. We focus on the morphologically complex future-based paradigm, which includes forms based on the literary and colloquial variants of the future auxiliary. The colloquial forms are shown to have lost the ability to express purely temporal meanings, being used exclusively with modal-evidential readings. We further argue that the future-based presumptive is closely related to epistemic modals, with which it shares the ability to express indirect inferential evidentiality. We capitalize on their contrasting behavior in factive settings to explain the interaction with the epistemic determiner vreun. Specifically, we argue that the key property that makes the (colloquial) future-based presumptive a suitable licensor for vreun is its incompatibility with contexts where the modalized proposition is established to hold. We discuss a similar pattern in some American Spanish varieties, where the so-called epistemic future has lost its ability to determine temporal (future) orientation. Keywords. Romanian; presumptive; epistemic modality; factivity; epistemic future; epistemic indefinites 1. Introduction This paper investigates the relation between mood and modality, by discussing the properties of the presumptive mood in Romanian. The starting point comes from the distributional restrictions exhibited by the epistemic indefinite vreun (Fălăuş 2009, 2014). The determiner vreun (vreo in the feminine) conveys speaker’s ignorance with respect to the referent of the nominal phrase. It is ruled out from episodic (1) and deontic modal contexts (2a), but licensed under epistemic modals and verbs (2a, 3a), as long as they are not factive (3b): (1) *Monica s-a întâlnit cu vreun prieten. Monica refl-has met with vreun friend ‘Monica met some friend.’ (2) a. Epistemic modal Trebuie să se fi întâlnit cu vreun prieten. Must subj refl.3sg be met with vreun friend ‘She must have met some friend.’ b. Deontic modal *Trebuie să mă întâlnesc cu vreun prieten. must subj refl.1sg meet with vreun friend ‘I must meet with some friend.’ *	   For thoughtful comments and suggestions, I would like to thank the editor and two anonymous reviewers, as well as Brenda Laca and Dora Mihoc. All errors and omissions are mine. I gratefully acknowledge funding from the Spanish Ministry of Science and Innovation (FFI2011-29218), the Basque Government (IT769-13) and the University of the Basque Country UPV/EHU (UFI11/14).


Introduction
This paper investigates the relation between mood and modality, by discussing the properties of the presumptive mood in Romanian.The starting point comes from the distributional restrictions exhibited by the epistemic indefinite vreun (Fălăuş 2009(Fălăuş , 2014)).The determiner vreun (vreo in the feminine) conveys speaker's ignorance with respect to the referent of the nominal phrase.It is ruled out from episodic (1) and deontic modal contexts (2a), but licensed under epistemic modals and verbs (2a, 3a), as long as they are not factive (3b): (1) *Monica s-a întâlnit cu vreun prieten.Monica refl-has met with vreun friend 'Monica met some friend.'(2) a. Epistemic modal Trebuie să se fi întâlnit cu vreun prieten.Must subj refl.3sgbe met with vreun friend 'She must have met some friend.' b.Deontic modal *Trebuie să mă întâlnesc cu vreun prieten.must subj refl.1sgmeet with vreun friend 'I must meet with some friend.' (3) a. Cred/ bănuiesc că s-a întâlnit cu vreun prieten.think.1sgsuspect.1sg that refl.3sg-hasmet with vreun friend 'I think/suspect (s)he has met some friend.' b. *Știu că s-a întâlnit cu vreun prieten.know.1sg that refl.3sg-hasmet with vreun friend 'I know (s)he has met some friend.'With this background in mind, we will focus on the fact that a verbal form in the so-called presumptive mood systematically licenses vreun: (4) S-o fi întâlnit cu vreun prieten.refl.3sg-aux.3sgbe met with vreun friend '(S)he might/must have met some friend.' These examples suggest a connection between presumptive mood and epistemic (non-factive) operators.The goal of this paper is to understand and flesh out this connection.It should be mentioned from the outset that my primary focus here is empirical.As such, the paper seeks to provide an accurate description of the core properties of the paradigm used in (4), which helps us understand its ability to license vreun.Although I outline some of the potentially relevant theoretical issues and offer some cross-linguistic remarks, a comprehensive account of the Romanian presumptive, as well as a detailed comparison with similar constructions across languages must be left for another occasion.
As will be discussed in detail in section 2, the term 'presumptive' is used to refer to an irrealis mood that subsumes several morphological paradigms, with different distributions and interpretations.My primary focus will be the future-based presumptive, which is in many respects similar to the so-called epistemic or probability future in languages such as Italian, Greek or Spanish (e.g.Squartini 2004;Giannakidou & Mari 2013).Drawing on insights in recent literature (e.g.Reinheimer-Rîpeanu 1994, 2000;Zafiu 2009;Irimia 2010;Mihoc 2013Mihoc , 2014)), future-based presumptive forms are shown to share an (indirect) evidentiality meaning component with epistemic modals, which plays a crucial role in satisfying the licensing conditions of the epistemic determiner vreun.We further explore the connections with epistemic modality by examining the quantificational force of presumptive forms and by comparing their ability to occur in settings where the uncertainty component typically associated with indirect evidentiality is cancelled.Section 5 summarizes our results and lays out the main issues in need of further research, focusing on the relation between presumptive forms and epistemic future in other Romance languages.

Background
The presumptive mood is, together with the conditional and the subjunctive, traditionally described as a non-indicative, irrealis mood (e.g.GALR The Romanian Academy Grammar 2008), primarily used to convey some form of uncertainty with respect to the claim being made.Its usage is generally characterized as belonging to spoken language.Morphologically, it presents different paradigms, but the number and the nature of these paradigms remain controversial (see Friedman 1997 andZafiu 2009 for an overview of the studies dealing with the Romanian presumptive).On its most inclusive acceptance (e.g.Irimia 2010), the label 'presumptive mood' designates three distinct morphological constructions: a future-based one, a conditional-based one and a subjunctive-based one. 1 The future-based form comes in two varieties, depending on whether it uses the literary or the colloquial form of the future auxiliary.The presumptive simple is formed in combination with an infinitive form (future and conditional auxiliaries) or with the subjunctive form of the verb (subjunctive paradigm).Furthermore, each of the three markers can combine with the auxiliary BE plus a past or a present participle (presumptive perfect or progressive).The following  The simple and perfect presumptive are homonymous with the simple and perfect forms of the future, conditional and subjunctive constructions.The following examples are thus ambiguous, as attested by their paraphrases: (5) Ar fi avut o problemă.cond.3sgbe had a problem a. CONDITIONAL READING: (S)he would have had a problem.b.PRESUMPTIVE READING: (S)he reportedly had a problem.(6) Va fi avut rezultatele.fut.3sg be had results-the a. FUTURE PERFECT READING: (S)he will have had the results (e.g.tomorrow afternoon).b.PRESUMPTIVE READING: (S)he probably had the results (e.g.yesterday).
This has raised the question of the existence of a distinct presumptive mood: are we dealing with extended (modal-evidential) uses of the future, conditional and subjunctive forms (e.g.Dimitriu 1979) or are there distinctive properties that set the presumptive apart?One argument that has been adduced in favor of a separate paradigm is the ability to combine with the present participle to form a progressive, a pattern unattested elsewhere in the Romanian grammar.What is interesting about this form is that in the case of the conditional paradigm, for example, the use of the progressive form leads to the unavailability of the conditional, counterfactual reading, leaving the presumptive interpretation as the only one available: (7) Ar fi având o problemă.cond.3sgbe having a problem a. *CONDITIONAL READING: (S)he would have a problem.b.PRESUMPTIVE READING: (S)he reportedly is having a problem.
One other argument that has been used to argue for a differentiation of presumptive and its homonymous modal/temporal forms is the meaning difference.As already illustrated above, the common function of presumptive forms is to express hypotheses or inferences concerning a certain state of affairs.More specifically, it has been argued that presumptive forms are evidentiality markers (e.g.Irimia 2010), i.e. they represent grammaticalized expressions of a source of information, e.g.inference, report, direct or indirect evidence (cf.de Haan 2001;Squartini 2004).As such, presumptive forms have distributional and interpretive properties that set them apart from the homonymous future/conditional or subjunctive forms (see Irimia 2010, Mihoc 2013 and section 3 below).Note however that the evidentiality component cannot be the sole factor responsible for the differences between presumptive and morphologically related paradigms.Both Zafiu (2002Zafiu ( , 2009) ) and Mihoc (2013) show that future-based presumptive is the only form in the paradigm that is able to express evidentiality by itself (as discussed in more detail in section 4).Conditional and subjunctive-based forms typically require the presence of additional evidentiality markers, e.g.verba dicendi, verbs such as seem, interrogative markers, or lexical items such as reportedly, approximately, etc.Although one can find examples where no such markers are present (e.g. ( 5) above), typically in headlines (de Haan 2001, 14;Mihoc 2013, 27), the presumptive reading is not fully grammaticalized and tends to be dispreferred, indicating that this is not a core meaning component.These differences are discussed at length in the above-mentioned studies and need not concern us here.In the following sections, we will examine more closely the relation between evidentiality, modality and the future-based presumptive.For present purposes, the only relevant point is the lack of homogeneity among the presumptive forms concerning their ability to express evidentiality.
These introductory remarks indicate that the presumptive is a complex paradigm, with intricate morpho-syntactic and semantic properties, a situation that explains the large number of controversies surrounding it.In the following, we focus exclusively on the future-based form, which contemporary Romanian grammars (e.g.Avram 1997;GALR The Romanian Academy Grammar 2008) analyze as the uncontroversial representative of the presumptive mood.Examining the properties of this construction, the goal is to (i) assess its similarity with epistemic modality, and as such, (ii) get a better understanding of its ability to license the epistemic indefinite vreun. 33 Fălăuş (2009Fălăuş ( , 2014) ) notes that vreun can also be licensed by conditional-and subjunctive-based forms of the presumptive, but this use is much less frequent and may be subject to dialectal variation: (i) Cică ar fi apărut vreun nou virus extrem de periculos.evid.advcond.3sgbe appeared vreun new virus extremely of dangerous '(I hear/They say) A new, extremely dangerous virus has appeared.'Furthermore, the occurrence of vreun seems to depend on the presence of another evidentiality marker (e.g. the hearsay adverb cică), suggesting that non-future presumptive forms cannot (straightforwardly) act as licensors.

Future-based presumptive: forms and usage
This section delves into the properties of the future-based presumptive.We begin with an overview of the existing paradigms and their usage in contemporary Romanian.The complete paradigm of the future-based presumptive is given in As noted earlier, there is disagreement in the literature concerning the inclusion of all three paradigms (present, perfect and progressive) within the presumptive.Crucially however, all the forms included in Table 2 had and in many cases, still have presumptive, modal-evidential readings.For the presumptive progressive, this is the only possible interpretation, having lost its ability to carry any temporal or aspectual value.In contrast to this, the presumptive simple and perfect forms are identical to the (simple and perfect) future of the indicative.The future and presumptive paradigms differ however in their temporal properties: whereas 'real' future tense requires a future time frame of reference, presumptive forms are compatible with present, future or past time frames of reference. 4To see this, consider their interaction with temporal adverbials: (8) FUTURE PERFECT, FUTURE ADVERBIALS Can we meet for dinner tomorrow?Sigur, până mâine seară voi fi terminat de corectat.sure by tomorrow evening fut.1sg be finished of grading 'Sure, by tomorrow evening, I will have finished grading.' (9) a. PRESUMPTIVE PERFECT, PAST ADVERBIALS I wonder when our grades will be posted.Cu puţin noroc, profesorul o fi terminat de corectat ieri with little luck professor-the fut.3sg be finished of grading yesterday şi sunt deja afişate.and are already posted 'With some luck, the professor will have finished grading yesterday and they are already posted.'b.PRESUMPTIVE SIMPLE OR PROGRESSIVE, PRESENT ADVERBIALS I heard that your best friend moved away.
Ţi-o fi (fiind) greu zilele astea you-fut.3sgbe being hard days-the these 'It must be hard for you these days.' The sentences in ( 9) convey inferences about the present or the past and it is clear that they do not make reference to any future event.In contrast to this, the context makes salient a future perfect reading for the verbal form in (8).While presumptive forms are typically used with a present or past reference time, Mihoc (2013Mihoc ( , 2014) ) also provides examples with a future time frame of reference: (10) PRESUMPTIVE SIMPLE; FUTURE ADVERBIALS Context: A keeps making up excuses to avoid doing a chore.B complains: Azi e ocupat, mâine-o fi bolnav şi tot aşa! today is busy tomorrow-fut.3SGbe sick and forever thus 'Today he's busy, tomorrow he'll be sick, and so on and so forth (= and he is basically never able to do it)!'These examples illustrate how adverbials might help solve the ambiguity problems raised by the future-presumptive homonymy.This raises the question of what happens in the absence of adverbials.Are we dealing with constant ambiguity, especially with a future time frame of reference, compatible with both future and presumptive readings?The answer is no, not anymore.In contemporary Romanian, the literary future is only used to convey future-indexical meanings, whereas the colloquial future form lost its ability to express purely temporal meanings and is used exclusively as a presumptive. 5,6As a result, the colloquial version cannot be used in contexts where the temporal value is clearly the one at stake: (11) a. Avionul va/*o decola la ora 7.
✓LIT.FUT./ *COLL.FUT.plane-the fut.3sg take-off at hour 7 'The plane will take off at 7 o'clock.' b.Vremea va/*o fi predominant frumoasă şi caldă pentru această perioadă.weather-the fut.3sg be predominantly beautiful and warm for this period 'The weather will be mostly beautiful and warm for this time of the year.'Conversely, replacing the colloquial form with the literary one leads to plain ungrammaticality (12a) or change in meaning (12b-c), as attested by the following examples, taken from Reinheimer Rîpeanu (1994: 5): (i) It's ringing at the door.
Va fi poştaşul!fut.3sg be mailman-the 'It must be the postman./It'sprobably the postman.' The Romanian speakers I consulted disagree with this judgment: for them, the future form can only acquire a temporal reading.Since this is strongly disfavored by the context in (i), the sentence is judged degraded or unacceptable.Reinheimer Rîpeanu (1994) and Zafiu (2009) offer detailed discussion of the evolution of the various future constructions and the specialization of the literary and colloquial forms.Further research is necessary in order to determine the source of the difference in judgments with Mihoc's work. 6Exceptions can be found in certain fixed expressions, e.g.Om trăi şi om vedea 'We will live and we will see' or Ce-o fi o fi 'Whatever will be will be'.The use of a presumptive with an indexical value can still be found in older Romanian texts and might subsist in some regions (see Reinheimer Rîpeanu 1994 for details).
(12) a. ✓COLL.11)-( 12) illustrate, the literary and colloquial future forms of the presumptive present paradigm are not interchangeable. 7he presumptive perfect follows a similar pattern, although it is admittedly less clear-cut.More precisely, once again, the indexical (future perfect) value is disallowed by the colloquial form, being restricted to the literary form (see ( 8)).However, in formal language, the literary future can also occasionally acquire nonindexical readings, as illustrated by the following example from a newspaper (due to Zafiu 2009: 18) (13) Probabil că mulţi dintre dumneavoastră se vor fi mirat când au auzit că un fost membru al conducerii BANCOREX a fost împuşcat.(Academia Caţavencu, 41, 1999, 4) 'Probably many of you have been surprised (literary future) to hear that a former board member at BANCOREX has been shot.' Taking stock, the future-based presumptive comes in a variety of forms and paradigms.Several diachronic studies indicate that all of these forms initially had temporal-aspectual values, even the progressive construction (see Reinheimer Rîpeanu 1998Rîpeanu , 2000;;Zafiu 2009 and references within).Subsequently, they all gradually acquired modal-evidential meanings.Starting with the second half of the 18 th century for the progressive presumptive, and the 19 th century for the presumptive simple and perfect, the potential ambiguities raised by this abundance of forms led to a division of labor.As a result, the progressive paradigm (literary and colloquial future alike) and the colloquial forms no longer have purely temporal-aspectual uses.In contrast to this, indexical uses are prevalent with the literary future forms of the presumptive simple and perfect (although the latter seems to retain some modalevidential readings).Our discussion shows that while the tension between indexical and modal readings has not been fully resolved in contemporary Romanian, there are forms within the presumptive paradigm that have an unambiguously modal value.Since it is the modal-evidential value that is most relevant for the licensing of epistemic indefinites, it is on these forms that I will concentrate in the remainder of this paper.The next sections examine more closely the relation of (non-indexical) presumptive forms with evidentiality and modality.

Future-based presumptive: semantic properties
In the previous sections, we loosely referred to presumptive meanings as modalevidential.It is now time to probe into the semantic properties of the (future-based) presumptive, which has been argued to act as evidentiality marker (e.g.Irimia 2010) and/or epistemic modal (e.g.Mihoc 2013Mihoc , 2014)).

Presumptive, evidentiality and epistemic modality
The future-based presumptive is the only form in the paradigm able to express evidentiality all by itself (Zafiu 2009;Irimia 2010;Mihoc 2013Mihoc , 2014)).Adopting the classification of evidential markers in Willett (1988), which draws a distinction between direct (i.e.perceptual) and indirect (inferential, reported) source of evidence, Mihoc (2013) shows that the future-based presumptive specializes in indirectinferential evidentiality (see also Irimia 2010).This means that it encodes inferences based on indirect or even direct but possibly insufficient information.To see this, consider the examples in ( 14 The evidence underlying the inferential process can be of various kinds, what is crucial is that it does not settle the truth of the modalized proposition.It should however be noted that future-based presumptive forms doesn't seem able to signal reported information.This is arguably due to the fact that in Romanian, it is the conditional presumptive that constitutes the default marker of reportatives, which explains the marginal status of the future-based presumptive in (15): (15) Se zice că ✓ar fi plecat/ ?? o fi plecat în Spania.SE says that cond.3sgbe gone fut.3sg be gone in Spain 'They say she has probably left for Spain.' The contexts in ( 14) illustrate the primary, inferential use of the future-based presumptive: the speaker makes a claim based on evidence available to her, without committing herself to the truth of this claim.To use Mihoc's words, the speaker expresses her 'best guess' given the evidence.The common feature of these examples is their incompatibility with contexts that leave no doubt with respect to the truth of the statement, which we can informally refer to as 'factive'.One could not utter the sentences in ( 14) looking outside and seeing pouring rain or seeing the baby crying.As is well-known, this property characterizes not only plain evidentials, but also epistemic modals (see de Haan 2001;Squartini 2004;von Fintel & Gillies 2010, among many others).Both in Romanian and in English, epistemic modals could be used in the contexts in ( 14)-( 15), but would be ruled out in a setting where the proposition to which the modal applies is undoubtedly true, as in ( 16): (16) Watching by the window and seeing pouring rain: #Trebuie/poate că iar plouă.must maybe that again rains 'It must/may be raining again.' There is an ongoing discussion about whether the restriction in ( 16), especially in the case of necessity epistemic modals, is related to strength of the modal, evidence source, speaker commitment, or amount of speaker's knowledge (for recent discussion see e.g.von Fintel & Gillies 2010; Matthewson 2010;Kratzer 2012;Giannakidou & Mari 2013).While this is an important issue, most relevant for our discussion is the similar behavior between epistemic modals and presumptive forms.This resemblance led Mihoc (2013Mihoc ( , 2014) ) to argue that the future presumptive is best analyzed as an epistemic modal. 8She shows that this account nicely captures its temporal properties, as well as its tight relation with indirect evidentiality.
There are however distributional differences between epistemic modals and presumptive forms.First, in inferential settings that rule out uncertainty, as in the following example, modeled from von Fintel & Gillies ( 2010 While epistemic modals are allowed in contexts where no reasonable doubt subsists concerning the truth of the modalized proposition, the infelicitousness of (17c) indicates that the presumptive disallows this option.Further support for this 8 More precisely, building on Condoravdi's (2003) account of will, Mihoc (2014) analyzes the futurebase presumptive as a necessity epistemic modal.The 'best guess' component is derived from a mechanism of domain restriction via the ordering source (along the lines of Kratzer 2012).The reader is referred to Mihoc (2013Mihoc ( , 2014) ) for details. 9 There might be speaker variability concerning this judgment.According to Mihoc (2014), the sentence in (17c) is acceptable in this context.Out of my 10 Romanian informants, one accepted the presumptive (saying however that the use the presumptive implies some (however unlikely) uncertainty).The source of these conflicting judgments must remain a matter for future investigation.contrasting behavior comes from embedding differences: epistemic modals can be embedded under factive verbs (18a), whereas presumptive forms cannot (18b): 10 (18) a. Ştiu că Ion trebuie să fi fost acasă în momentul crimei.know.1sg that Ion must subj be been home in time-the murder.gen'I know that Ion must have been home at the time of the murder.' b. *Ştiu că Ion o fi fost acasă în momentul crimei.know.1sg that Ion fut.3sg be been home in time-the murder.genAs before, the presupposition that the embedded proposition holds rules out the use of the future-based presumptive.The contrasts in ( 17) and ( 18) suggest that presumptive mood cannot be fully equated with epistemic modality, despite the tantalizing similarity observed in evidential settings.
A further difficulty in analyzing the future presumptive as an epistemic modal relates to quantificational force.Mihoc (2013Mihoc ( , 2014) ) provides evidence that presumptive forms occur in contexts where both necessity and possibility modals can be used.As the sentences in ( 14) and their English paraphrases illustrate, the presumptive can be used in contexts where necessity epistemics are acceptable, i.e. there is strong enough evidence to support a possibly high degree of confidence concerning the modal claim.However, the unacceptability of the presumptive in (17) shows that the identity between presumptive forms and necessity epistemic modals is unwarranted.There are other facts suggesting that the quantificational force of the presumptive may be weaker than (or at any rate, different from) that of a necessity modal.First, like possibility modals, and unlike necessity modals, it passes the contradiction test, as confirmed by the acceptability of the sentences in ( 19 may subj be.3sg and may subj not be.3g is too early subj tell.1pl'It may be and it may not be, it's too early to tell.' c. #Trebuie că este şi trebuie că nu este, e prea devreme să spunem.
must that is and must that not is is too early subj tell.1pl'It must be and it must not be, it's too early to tell.' These examples show that a presumptive form can be conjoined with its negated version, without giving rise to a contradiction, an option that does not seem to be available for necessity modals.Second, when we examine the inferences possibility and necessity modals give rise to, we once again observe the presumptive patterns with the weaker modality: (20) a.He may be home.In fact, he must be home (he never goes out on Sunday).
b. #He must be home.In fact, he may be home.c. O fi acasă.De fapt, trebuie/#poate să fie acasă.fut.3sg be home in fact must may subj be.3sg home 'He is probably home.In fact, he must/may be home.'On the other hand, just like necessity epistemics, the presumptive gives rise to Moore's paradox: (21) a. #He must have been home at the time of the murder, but I don't believe it.
b.He might have been home at the time of the murder, but I don't believe it.c. # O fi fost acasă în momentul crimei, dar nu cred.fut.3sg be been home in time-the murder.genbut not believe.1sg We thus seem to have conflicting evidence concerning the force of the modality conveyed by the presumptive: contexts such as ( 14) show that just like necessity modals, the presumptive can be used for statements made with a high degree of confidence (where a possibility modal would not be appropriate).On the other hand, we observe similarities with both possibility and necessity modals, suggesting that the quantificational force of the presumptive cannot be entirely assimilated to either a necessity or possibility modal.
To furthermore reinforce the variable quantificational force of the presumptive, consider its compatibility with probability adverbs.The examples in ( 22) show that presumptive forms can felicitously co-occur with both high and low probability adverbs: (22) Sigur /Precis /Probabil /Poate o fi plecat din oraş.for-sure certainly probably perhaps fut.3sg be gone from town '(S)he certainly/undoubtedly/probably/possibly is out of town.' The conclusion that suggests itself in view of the data is that presumptive forms can sit anywhere on a probability scale.The point has been most clearly made in Mihoc (2013: 14): "[the presumptive] may range from mere possibility to high probability (the more obvious sense), or, otherwise put, from sheer speculation through educated deduction or inference to a statement that is not understood as being modalized."Descriptively speaking, this makes the presumptive very similar to variable force epistemic modals (e.g.discussed for example for St'át'imcets in Rullman et al 2008; Matthewson 2010), or so-called 'variable upper-end degree epistemic modality' (Kratzer 2012: 46).In section 5, we discuss in more detail the similarities between the Romanian future-based presumptive and the so-called epistemic future in other Romance languages.
Taking stock, the properties discussed in this section point out a close connection between future-based presumptives and epistemic modals. 11However, the two categories are not identical.First, unlike epistemic modals, the presumptive is excluded from inferential contexts that force a 'factive' reading of the proposition to which it applies (as in ( 17)-( 18)).Second, the variability in the quantificational force of the presumptive sets it apart from both possibility and necessity epistemic modals.The more general lesson to be drawn is that future-based presumptive cannot be straightforwardly equated with an epistemic modal of the may/must variety.Or alternatively, that an account that takes the presumptive to be an epistemic modal will need to explain the source of the conflicting evidence concerning quantificational force.
With these properties in mind, we are now ready to go back to the licensing of the epistemic determiner vreun and explain why the future-based presumptive is its most frequent licensor.This will confirm the conclusions reached above on the diverging behavior of presumptive forms and epistemic modals.Fălăuş (2009Fălăuş ( , 2014) ) has argued that the determiner vreun is licensed in (i) negative polarity and (ii) non-factive epistemic contexts.The restriction to epistemic contexts (repeated from (2) above) sets vreun apart from other epistemic indefinites studied to date (see e.g.Aloni & Port 2010, Alonso-Ovalle & Menéndez-Benito 2013, Chierchia 2013, Giannakidou & Quer 2013), which seem acceptable regardless of the flavor of the modal involved, as illustrated for algún in ( 24 Maria has comp finish ALGUN article for tomorrow 'Maria has to finish some paper for tomorrow.'Interestingly enough, the presence of an epistemic modal is not sufficient to license vreun.The context in (25) establishes that the ace must be with one of the players (although the speaker does not know which one).The epistemic modal can be used, but vreun is infelicitous (taken from Fălăuş 2014:133):

Back to epistemic indefinites
(25) We are playing cards.In order to win, I still need the ace of spades.However, by now all cards have been played and it is clear that the ace is not in the pile of cards on the table.So the only option is that one of the other players has it: Asul trebuie să fie la *vreun/✓un jucător (şi poate fi oricare din ei).12 ace-the must subj be.3sg at vreun a player and could be any of them 'The ace must be with some/a player (and it could be any of them).' A similar behavior can be observed in the scope of attitude predicates.Vreun requires an epistemic predicate, but is ruled out from the scope of factive predicates, e.g.know, discover, realize (as in 26c): (26) a. *M-a obligat/sfătuit să contactez vreo agenție de voiaj.
me-has forced advised subj contact vreun agency of travel '(S)he forced/advised me to get in touch with some travel agency.' b.Cred/ bănuiesc că s-a întâlnit cu vreun prieten.think.1sgsuspect.1sg that refl.3sg-hasmet with vreun friend 'I think/suspect (s)he has met some friend.'c. *Știu că s-a întâlnit cu vreun prieten.know.1sg that refl.3sg-hasmet with vreun friend 'I know (s)he has met some friend.' Once we put the restriction to (non-factive) epistemic modals together with the unacceptability under factive attitude verbs, it becomes clear why the (future-based) presumptive is such a frequent licensor for vreun.On the one hand, we have a determiner that (in non-negative polarity contexts) requires a non-factive epistemic operator.On the other hand, we have a modal(ized) paradigm within the verbal system with precisely the semantics of a non-factive epistemic.
Like in many other languages, the necessity and possibility modal auxiliaries in Romanian are used to express a wide range of modal meanings.To a large extent, it is the context that determines whether the modal acquires an epistemic reading or not.In other words, the modal is ambiguous.Furthermore, we have seen that epistemic modals are compatible with contexts where there is no uncertainty involved (as in ( 17)-( 18) and ( 25)).In contrast to this, (future-based) presumptive forms are unambiguously interpreted as epistemic operators and are systematically unacceptable in factive contexts and settings where the proposition is viewed as certain.The picture we obtain is the one summarized in table 3 This explains why the presumptive is the most frequent licenser of the epistemic determiner vreun: its semantics has all the necessary ingredients (and only those).No possible ambiguity with respect to the kind of modality involved and systematic incompatibility with factive contexts.Precisely the licensing factors relevant for the distribution of vreun. 13

Cross-linguistic remarks and open issues
The main goal of this essentially descriptive paper was to understand the properties of the future-based presumptive that are responsible for the licensing of vreun.Our discussion revealed a tight connection between presumptive mood and epistemic modality, which I have argued to be relevant in our understanding of the licensing conditions of epistemic indefinites.There are several empirical and theoretical issues raised by facts described above, which we have set aside in this paper, but are essential to account for the Romanian presumptive and more generally, the properties of morphologically future constructions.An important area of investigation concerns the rest of the presumptive forms mentioned in section 2. Crucial to establishing the existence of a presumptive mood is determining the relation with tense and aspect (both lexical and grammatical).We have argued that the future-based presumptive does not impose any restrictions on temporal orientation, being compatible with past, present and future frames of reference.It remains to be seen how the conditional and subjunctive-based forms behave in this respect.Related to this, the use of the progressive (which is unattested elsewhere in the grammar) needs to be accounted for (see Mihoc 2014 for an attempt in this sense).
In addition to these language-internal aspects, an explanatory analysis of the Romanian presumptive will have to take into account relevant cross-linguistic facts. 14 The presumptive forms described in this paper closely resemble so-called epistemic or conjectural future in Romance.According to Nueva gramática de la lengua española NGLE (2009NGLE ( : 1771NGLE ( -1773)), the morphologically future form in (27) has a modal value, just like the presumptive form, it conveys speaker's present hypothesis concerning a certain state of affairs: (27) a. Estará enfadada conmigo.
be.fut3sg upset with-me 'She is probably upset with me.' b.Sabrás que ya no vivo aquí.know.fut2sg that anymore not live.1sghere 'You probably know that I don't live here anymore.' 13 Fălăuş (2014) derives the restricted distribution of vreun from the assumption that it is an alternativeactivating indefinite (in the sense of Chierchia 2013), which is furthermore incompatible with situations where all elements in the quantificational domain count as possible values (as in ( 25)).The unacceptability in episodic and non-epistemic contexts is argued to follow from independent properties of the licensing operators.The reader is referred to Fălăuş (2014) for details. 14 In line with the focus of the present journal issue, we limit our remarks to Romance languages.But it should be borne in mind that epistemic future is not restricted to Romance (e.g.Giannakidou & Mari 2013) and presumptive/conjectural/assumptive mood is also attested in Uzbek, Tajik, or Hindi (e.g.Masica 1991).In addition, the Romanian presumptive mood shares several properties with the subjunctive mood in St'át'imcets (Lillooet Salish), carefully discussed in Matthewson (2010).A detailed cross-linguistic comparison is clearly beyond the scope of this paper.
Similar facts can be found in French (e.g.Bellahsène 2007, de Saussure & Morency 2012) or Italian (e.g.Bertinetto 1979, Mari 2009).An important difference between (contemporary) Romanian and these other Romance languages is the lack of temporal readings with the colloquial form of the future presumptive (see section 3).In contrast to this, as documented in the aforementioned studies, epistemic futures in Romance seem to admit a purely temporal interpretation: (28) a. Se convencerá de que somos listos.refl convince.fut3sg of that be.1pl smart '(S)he'll realize we are smart.'b.Il negozio chiuderà alle 4 del pomeriggio.
the shop close.fut3sgat 4 of-the afternoon 'The shop will close at 4 in the afternoon.' The split in Romanian between literary and colloquial future (illustrated in section 2), both in terms of their morphology and their usage, can be viewed as the disambiguation of the corresponding future forms in the rest of the Romance languages.This provides a useful empirical basis to test the predictions made by current accounts concerning the connections between the temporal and the modal uses of the future.
Interestingly, Romanian is not the only Romance language in which a morphologically future form has lost its ability to determine temporal orientation.In the following, I briefly discuss a similar pattern in American Spanish varieties, i.e.Spanish spoken in Rio de la Plata -Buenos Aires et Montevideo, drawing on the data and analysis developed in Fălăuş & Laca (in press), to which the interested reader is referred for details.More concretely, in these varieties of Spanish, future morphology cannot be used in predictions or statements of intentions, such as ( 29)-( 30), where the prospective periphrasis ir+a+infinitive 'go to+infinitive' needs to be used instead: Just like the Romanian presumptive, the morphologically future form in these varieties of Spanish only has epistemic uses, lacking any influence on the temporal orientation.This is in contrast with European Spanish, where epistemic uses co-exist with temporal, forward-shifting ones (e.g. the sentences in (29a) and (30a) above are felicitous in European Spanish).Another potentially interesting difference between European and American Spanish is the fact that the latter allows the morphological future to combine with the prospective periphrasis ir+a+infinitive 'go+to+infinitive', as illustrated in (31) (due to Laca 2014): (31) No irán a comer esa porquería.not go.fut3pl to eat this rubbish 'I hope you're not going to eat this rubbish' The study of these dialectal differences is still in its early stages, but these examples should suffice to convey the benefits of studying the micro-and macrovariation we find in the area of future forms, both in terms of their morpho-syntactic and semantic properties.
The semantic behavior of Spanish epistemic future and Romanian future-based presumptive is remarkably similar.Just like we have observed for Romanian (cf.section 4), epistemic future seems to have variable quantificational force, can felicitously co-occur with adverbs expressing various degrees of probability (32), combines often with other expressions of uncertainty (33) and is infelicitous in factive contexts (34): (32) Seguramente/Probablemente/Posiblemente se habrá asesorado con surely probably possibly refl have.fut3sgadvised with algún amigo.some friend 'It is likely/probable/possible that (s)he got advice from some friend.'(33) A: La boleta de la luz?
'A : The electricity bill ?I don't know, who knows, it may be in the drawer of the commode.' (34) #Juan sabe dónde estará la llave.
'Juan knows where the key might be.' Fălăuş & Laca (in press) offer a more detailed comparison between Romanian future-based presumptive and epistemic future in American Spanish.The two forms are shown to display a wide array of distributional and interpretive similarities (as already illustrated above), suggesting a possible unified account. 15However, there are also some intriguing differences, which call for further investigation.Most noticeably, there seem to be different constraints concerning the epistemic agent responsible for the modal claim, i.e. the agent who needs to be uncertain with respect to the truth of the epistemic statement.In Romanian, the agent must be (or include) the speaker -if the context makes it clear that the speaker is not in a state of uncertainty, then the presumptive is infelicitous (35): 15 More concretely, Fălăuş & Laca argue against an analysis of the presumptive and the epistemic future in terms of epistemic modality.They show that unlike epistemic modals, the future forms they investigate take obligatorily wide scope with respect to a variety of operators (e.g.negation, conditional, because).On their proposal, the semantic contribution of the presumptive and epistemic future (in the American Spanish varieties) amounts to the expression of the speaker's comment, whereby she conveys that the relevant epistemic agent is attributing to the prejacent a degree of probability that is lower than the contextual threshold for assertions (in the sense of Davis, Potts & Speas 2007).The reader is referred to the original paper for further details.
(35) In a quiz context, where the speaker knows where the treasure is # Ce spuneți, copii, unde o fi ascunsă comoara?what say.2pl children where 3fut.sgbe hidden treasure-the 'What do you think, children, where may the treasure be?' In Spanish on the other hand, one can find the epistemic future used even in examples where the speaker knows the truth of the modalized proposition, as illustrated in (36).Unlike in Romanian, where the use of the presumptive necessarily conveys that the speaker does not (or pretends not to) know the answer to the question, the epistemic future is acceptable in contexts where the speaker is clearly in a state of knowledge with respect to the question under discussion: (36) During an oral exam, the professor asks: ¿Por qué es importante la paleografía para el historiador?[Silence] Vamos, muchacho, piense: ¿por qué será importante la paleografía en la carrera de historia?'Why is paleography important for the historian?[Silence] Come on, boy, why could paleography be important in the study of history ?'A possibly related difference between Romanian and Spanish comes from the use of the future in degree exclamatives, which is only possible in Spanish: (37) ¡Serás imbécil!be.fut3sg stupid 'You are so stupid !' There is clearly no uncertainty component involved in exclamatives, raising the question of the connection with the other contexts of usage.Fălăuş & Laca (in press) suggest that this kind of occurrence relies on a complex rhetorical strategy, which allows the speaker to exploit obvious facts to convey irony.More cross-linguistic investigation is needed on this matter, but the examples above suggest that the ban of the presumptive in situations where the speaker is not in any possible state of uncertainty may explain why the presumptive is ruled out from exclamatives (which lack the necessary uncertainty component).
Before concluding, I should mention that there are several theoretical issues currently under debate concerning the distinction between epistemic future and (necessity) epistemic modals (for recent discussion, see in particular Matthewson 2010, Giannakidou & Mari 2013).How do the temporal and the modal value relate?Do epistemic future and epistemic modal auxiliaries differ in terms of quantificational force, ordering source, speaker commitment/confidence?Related to this, is the difference located at the truth-conditional level of meaning or is there a separate (pragmatic, evaluative) dimension that needs to be considered?A close examination of embedding properties, as well as of the systematic use of presumptive/epistemic future in concessive constructions (38) and in (root and embedded) interrogatives (38) may pave the way to answering some of these questions: (38) a.I have just met Maria and found her to be arrogant and rude.Although our starting point was the licensing of the epistemic indefinite vreun, we have tried to show that the Romanian presumptive mood is a rich and interesting area of investigation.We have further shown that there are intriguing similarities between the presumptive and the epistemic future in Spanish, particularly in American Spanish varieties where the morphologically future form seems to lack temporal value.We hope that the properties discussed here provide useful insights, which can foster crosslinguistic comparison and can contribute to the ongoing debates concerning the relation between mood, tense and (epistemic) modality.

(
29) a. # Nació en 1960.En el 2015 tendrá 55 años.was-born in 1960 in the 2015 have.fut.3sg55 years b.Nació en 1960.En 2015 va a tener 55 años.was-born in 1960 in 2015 go.3sg.to have 55 years '(S)he was born in 1960.In 2015 (s)he will be 55.' (30) a. #Te llamaré todos los jueves.you call.fut2sgall the Thursdays b.Te voy a llamar todos los jueves.you go.1sg to call all the Thursdays 'I'll call you every Thursday.' table illustrates the different possibilities for the third person forms of the verb a avea 'to have': 2

Table 2 .
Future-based presumptive FUT./ *LIT.FUT Cum de n-o/*nu va obosi s-o ia mereu de la capăt!how of not-fut.3sgget-tired subj-it take.3sgalways from start 'How come (s)he doesn't get tired of always starting over!' b.COLLOQUIAL FUTURE, PRESUMPTIVE READING Mă întrebam ce-o fi cu dumneata, dacă nu cumva oi fi supărat.me wonder.impfwhat fut.3sg be with you if not somehow fut.2sg be upset 'I was wondering what might be with you, if you might not be upset.'c.LITERARY FUTURE, INDEXICAL READING Mă întrebam ce va fi cu dumneata, dacă nu cumva vei fi supărat.me wonder.impfwhat fut.3sg be with you, if not somehow fut.2sg be upset 'I was wondering what will happen with you (in the future), if you would not be upset.'As the sentences in ( ):(14) a. Inferences based on indirect evidence: I am waiting for Marc at the train station.The train arrives, but he is not there.Knowing that he tends to be late in whatever he does, I say: O fi ratat trenul.fut.3sg be missed train-the 'He probably missed the train./Hemust have missed the train.' b.Inference based on sensory, but insufficient evidence: i) It had been raining lately, but it was sunny this morning.At lunch, I see people coming in with wet umbrellas: Poate să fie şi poate să nu fie, e prea devreme să spunem.

Table 3 .
Distribution in epistemic settings People have always told me she is very smart: O fi ea (aşa) deşteaptă (cum se zice), dar e tare nesuferită.fut.3sg be she so smart as SE says but is very insufferable 'She might be (as) smart (as they say), but she is really insufferable.' b.Le parecerá una tonteria, pero aquello me salvo la vida.cl.3sg seem.fut3sg a stupid-thing but that cl.1sg saved the life 'You may think it's silly, but that thing saved my life'