2.1 Epistemic devoir vs. conjectural future

The idea that FUT conveys future verification is solidly established in the literature and it would be an impossible enterprise to do justice to it (a.o. Schrott, Reference Schrott1997; Tasmowski and Dendale, Reference Tasmowski, Dendale and van der Auwera1998; Dendale, Reference Dendale2001; Morency, Reference Morency2010; de Saussure and Morency, Reference de Saussure and Morency2011 and references therein). The notion of ‘future verification’, however, has received many distinct interpretations across the various studies, leading to different conceptions.

Tasmowski and Dendale (Reference Tasmowski, Dendale and van der Auwera1998) and Dendale (Reference Dendale2001) are representative of one way of understanding ‘future verification’. Dendale (Reference Dendale2001) contrasts future tense and devoir and explains that devoir can be used if the speaker has evidence in the utterance context for the truth of the prejacent (p from now on), while such evidence is not needed in order to use FUT. FUT can be used, Dendale claims, even in the absence of such evidence.

Consider a scenario in which I wonder what is happening to John, who is unexpectedly tired. One of his colleagues asserts the following.

1. (4) Il doit avoir été malade, il a l’air tout pâle.

2. (5) Il aura été malade, voilà tout.

In both cases the speaker must have ‘indirect’ evidence, that is to say, he cannot have seen John ill. However, the devoir sentence requires that the speaker bases his assertion on certain pieces of evidence (e.g. John is pale), but the future sentence does not. The idea then arises that some evidence will be available in the future and that the conjecture will be verified then. In Tasmowski and Dendale's view, future sentences lack truth conditions at the time of utterance (see Tasmowski and Dendale, Reference Tasmowski, Dendale and van der Auwera1998: 327).

De Saussure and Morency (Reference de Saussure and Morency2011: 209) (S&M from now on), pick up on the idea of future verification, but include it in the semantics of the future, a path that we follow here.

Our general assumption follows the classical analysis in which FUT in French represents a future verification of a possible state of affairs in the present (or in the past, as with the future anterior).

There are some cases in which the idea of future verification works particularly well, as the following example shows. Two friends are speculating about the shape of the universe. FUT cannot be used to express a conjecture ((6)). Devoir can be used in this context ((7)).

1. (6) #L’univers sera sphérique. (S&M, 2011: 217, ex. (34))

2. (7) Il doit être sphérique.

De Saussure and Morency explain the impossibility of (6) by the fact that neither the speaker nor the hearer can directly verify whether the universe is spherical. (This is in contrast with Schrott's (Reference Schrott1997) view that the verification has to take place within the sphere of experience of the speaker. De Saussure and Morency are, arguably, less restrictive, and either the speaker or the hearer are good candidates for verifying the truth of the prejacent.)

Devoir is unproblematic in this context, as it expresses conjecture ((7)). This shows that an inference-based account does not seem to be suitable for French (pace Caudal, Reference Caudal2012).

Unlike devoir, FUT can be used when the speaker knows that the prejacent is true at the time of utterance. Again, the example is due to de Saussure and Morency (Reference de Saussure and Morency2011). A shop employee has just rearranged the shoes on the shelves. A customer enters the shop and asks where the Adidas trainers are. The employee replies (8).

1. (8) Elles seront sur le présentoir là-bas. (S&M, 2011: 219, ex. (37)Footnote ⁶ )

Here FUT does not express a conjecture. Rather, it postpones the time of verification that the shoes are on the shelf with respect to the time of utterance. Here the hearer (pace Schrott, Reference Schrott1997) will be responsible for the verification (see also Azzopardi and Bres, Reference Azzopardi and Bres2014).

There are other uses of FUT, however, for which theories holding that FUT means future verification must provide a more elaborate account.

FUT, in French, features a truly epistemic (or conjectural use). Consider a scenario in which I have been travelling to Japan and stayed at a hotel. Later, I could not find my watch. My husband utters (9).

1. (9) Tu l’auras laissée à l’hôtel.

The question immediately arises of what ‘future verification’ amounts to. Schrott (Reference Schrott1997) proposes that with the future anterior and the conjectural interpretation there is a reinterpretation at the de dicto level of the future verification at de re level that characterizes the temporal interpretation. Translating this view into other terms, we would obtain an ambiguity of the future, emerging as a type-theoretical difference: for the temporal inter- pretation FUT's denotation falls within the domain of events, and in the conjectural interpretation, in that of propositions. This view is certainly tenable and in fact has been advanced more recently by Copley (Reference Copley2002). However, before positing an ambiguity – for the sake of economy of the theory –it behoves us to attempt to provide a unified entry and explain the ambiguity in a systematic way without positing a type-theoretical distinction. The one we provide in this paper will also be able to cover cases that are not easy to categorize in either one of the two classes (de re - de dicto cases) and which clearly fall within a system of preferences. To our knowledge, this has only recently been considered in more depth (see de Saussure and Morency, Reference de Saussure and Morency2011).

Let us return to (9). In this interpretation, FUT behaves similarly to devoir (must) ((10)). In the same scenario, the following sentence is acceptable.

1. (10) Tu dois l’avoir laissée à l’hôtel.

Note also that FUT cannot be a temporal operator that forward-shifts the time of the event (or one that locates the event, at the earliest, at the time of utterance; see Kissine, Reference Kissine2008). If FUT were such an operator, a future-in-the-past reading would be obtained for (10), contrary to fact. (10) conveys that, according to what the speaker knows at the time of utterance, I left the watch at the hotel in the past (this is parallel to Italian, see Mari, Reference Mari, Calzolari and Rumshiski2009; Gianakidou and Mari, Reference Giannakidou, Mari, Aloni, Franke and Roelofsen2013, to appear-a).

De Saussure and Morency (Reference de Saussure and Morency2011) propose that FUT means future verification of a present conjecture. The delicate question arising is at what level ‘conjecture’ and ‘verifi- cation’ are coded, a question that we discuss in detail in section 5.

Let us return to the hypothesis that FUT denotes future verification and to a case that is acknowledged as difficult by de Saussure and Morency themselves. We label this the ‘present ratification’ use (this use is also called ‘futur expansif’ (Wilmet, Reference Wilmet1976), ‘futur de retrospection’ (Wagner and Pichon, Reference Wagner and Pinchon1991) or ‘futur de bilan’ (Maingueneau, Reference Maingueneau1981; Caudal, Reference Caudal2012).

1. (11) Au-delà des résultats obtenus, le grand mérite de cet expert-tacticien aura été de créer le consensus autour de sa propre personne.

   Le Monde, 12 June 1998, p. 15

2. (12) Si Buzz Aldrin a été le second homme à poser le pied sur la Lune, il aura été le premier à léviter quelques centimètres au-dessus du sol sur le mythique overboard.

   http://www.20minutes.fr/high-tech/1471391-20141030-astronaute-buzz-aldrin-teste-hoverboard

In view of these cases, a series of questions arises. Does FUT denote ‘present conjecture’? Not always. It seems to do so unproblematically only in (9). Does FUT denote future verification? Again, not always. In some cases, future verification is impossible and the sentence is true ((9)). In some other cases, the speaker knows at the time of utterance that it is true ((11)).

An adequate theory of the future, we claim, must be able to explain the facts summarized in Table 1 (see above). Like devoir, FUT seems to be able to convey present conjecture. This conjecture is grounded in indirect evidence ((9)). Unlike devoir, FUT can use direct evidence ((11)). Unlike devoir, it can denote future verification ((8)).

The diversity of these uses is puzzling. Thus, one of the goals of this paper is to propose a theory that explains these distributions in a principled way. Let us note that we do not claim that these are all and only the available uses of the French FUT (see for instance Azzopardi and Bres, Reference Azzopardi and Bres2014 for discussion of the ‘future of assertion’ and the ‘discovery’ use). We also do not consider here the rhetorical values of FUT in French (see Rossari et al., Reference Rossari, Ricci and Siminiciuc2014). To achieve our goal, we examine the evidential component of the ratificational modal, also considering the effects of this component on the system of preferences observed, to which we now turn.

2.2 Breaking the cross-linguistic patternFootnote ⁷

It is now a common observation that the epistemic/non-epistemic ambiguity of modals correlates with the aspectual properties of the complement they combine with (see most notably Sueur, Reference Sueur1979; Condoravdi, Reference Condoravdi, Beaver, Casillas, Clark and Kaufmann2002; Copley, Reference Copley2002; Malchukov, Reference Malchukov, Hogeweg, de Hoop and Malchukov2009; Boogaart and Trnavac, Reference Boogaart, Trnavac, Patard and Brisard2011; Caudal, Reference Caudal2012; Giannakidou and Mari, to appear-a). Statives (including lexical statives, resultatives, and progressives) correlate with the epistemic interpretation ((13.a)–(14.a)), while eventives ((13.b)–(14.b)) correlate with non-epistemic interpretations (including the deontic, metaphysical, . . .).

1. (13.a) He is not at school, he might be ill. (epistemic)

2. (13.b) If he does not wear his raincoat, he might become ill. (circumstantialFootnote ⁸ )

3. (14.a) The lights are on. John must be at home. (epistemic)

4. (14.b) John must go home, his mother is waiting for him. (deontic)

The same pattern is illustrated by FUT in French, most clearly by the future anterior. Simple future sentences with eventives feature the temporal interpretation ((15)), whereas future anterior sentences feature the epistemic interpretation ((16)).

1. (15)

   

2. (16)

   

As a reviewer points out to us,Footnote ⁹ in French, future anterior sentences must be treated like statives. This can be demonstrated by the fact that when a temporal adverbial (à 4 heures) is used with the future anterior, it locates the time of the result ((17.b)), and not the time of the event, unlike with simple future sentences ((17.a)) (see also Kevers, Reference Kevers2011).

1. (17.a)

   

2. (17.b)

   

Note that in (17.b), the event of completing the test can take place after the time of utterance. When the event is located in the past, a temporal adverb can locate the time of the event rather than that of the result, as in (18), given in response to the question ‘Where is my watch? I cannot find it.’

1. (18)

   

Here the adverb hier (yesterday) locates the event of leaving the watch at the hotel rather than the time of the result.

For all these cases, we endorse the view that the future anterior involves a resultative construction (see also Gosselin, Reference Gosselin1996; Vet, Reference Vet, de Saussure, Moeschler and Puskas2007) and has to be treated as a stative, most notably following a recent analysis in Boogaart and Trnavac (Reference Boogaart, Trnavac, Patard and Brisard2011). Importantly for us, this resultative construction (which we label RES(p)) provides a handle for the evidential component of the ratificational modal as we explain in detail in section 3.

Thus far, French FUT seems to align with modals across languages. The epistemic interpretation of devoir arises with statives and the root (metaphysical, deontic, temporal, . . .) interpretation arises with eventives (see Condoravdi, Reference Condoravdi, Beaver, Casillas, Clark and Kaufmann2002).

1. (19.a)

   

2. (19.b)

   

3. (19.c)

   

Most of the accounts we are aware of derive the facts by relying on the boundedness- unboundedness difference (or, alternatively, the perfective-imperfective distinction, or the eventive-stative distinction). That boundedness correlates with forward-shifting is sometimes considered a primitive of the theory (Condoravdi, Reference Condoravdi, Beaver, Casillas, Clark and Kaufmann2002). According to Boogaart and Trvanac (Reference Boogaart, Trnavac, Patard and Brisard2011) (see also Caudal, Reference Caudal2012), bounded temporal entities induce an external point of view and thus may explain futurity with respect to the main perspectival point, which is the time of utterance (when the event is not otherwise located in the past).

A closer look, however, reveals that this pattern is disrupted with statives in simple fu- ture sentences in French, breaching the systematic ambiguity observed cross-linguistically. Both sentences in (20)–(21) have a preferred temporal use, regardless of whether an eventive or a stative predicate is used.

1. (20) (. . .) en raison de la forte démographie des travailleurs hispaniques, . . . un catholique sur deux parlera espagnol aux Etats-Unis en l’an 2000.

   ‘Because of the high proportion of Hispanic workers, . . . every other Catholic will speak Spanish in the USA in 2000.’ (Le Monde, 22 September 1987, p.1)

2. (21) Sachant qu’une personne était malade au jour 1 et que chaque malade contamine deux nouvelles personnes le jour suivant, vous devez calculer à partir de quel jour toute la population de la ville sera malade.

   ‘Knowing that one person was ill on day 1 and that every sick person contaminates two new persons the subsequent day, you should calculate from the day on which the whole population of the town will be ill.’

   http://openclassrooms.com/forum/sujet/application-en-medecine-17105

The preference for the temporal interpretation with statives is to some extent unexpected. This position needs to be qualified, and revised to some extent (cf. infra). Prima facie, though, it seems to capture a clear preference, to the point that it has been claimed that the epistemic interpretation is not available for simple future sentences (Klein, Reference Klein1980).

Examples of the epistemic interpretation of simple future sentences with statives are in- deed discussed in the literature. (22) is from Baranzini and de Saussure (Reference Baranzini and de Saussure2014) (B&S, Reference Baranzini and de Saussure2014; slides presented at the Chronos conference in Pisa, 2014).

1. (22)

   

2. (23)

   

While the predominant absence of the epistemic interpretation with eventives is no surprise given the cross-linguistic pattern, there are in fact examples of epistemic readings with eventives. One famous example discussed in the literature is from Chevalier (Reference Chevalier1978: 352) – cited in Dendale (Reference Dendale2001: 4):

1. (24)

   

It is, however, to be noted that this is a stative in disguise, as the possibility of using a durational adverb with pendant (during) shows (We thank an anonymous reviewer for this observation).

De Saussure and Morency (Reference de Saussure and Morency2011) were the first to tackle the difficult question of the conditions that make the epistemic interpretation with eventives possible. There is considerable agreement that the epistemic interpretation is very difficult. Consider a scenario in which Mary is on a train, her telephone rings, but she does not answer. A friend of mine, who is waiting for Mary with me at the train station, tries to figure out why she is not answering. I reply (25). Out of the blue, the utterance is not natural.

1. (25)

   

Morency (Reference Morency2010) and de Saussure and Morency (Reference de Saussure and Morency2011), crediting Sthioul (Reference Sthioul, de Saussure, Moeschler and Puskas2007) – have claimed that naturalness is restored in non-stative sentences, in scenarios where the event of verification is overt (in (26) the verification event is that of asking Pierre).

1. (26) A: J’ai une fuite dans la salle de bain ! (S&M, 2011: 219, ex. (19)Footnote ¹⁰ )

   B: Demande à Pierre, il connaîtra un plombier !Footnote ¹¹

   A: ‘There is a water leak in my bathroom !’

   B: ‘Ask Pierre, he will know a plumber !’

These preferences are puzzling. On the one hand, there is a clear preference for the temporal interpretation of simple future sentences in French with both eventives and statives. On the other, statives are also quite readily compatible with the epistemic interpretation, and the latter can be forced with eventives, in particular when a future event of verification is overtly mentioned.

An adequate theory of FUT in French, it seems, requires that we find an explanation of the epistemic/temporal ambiguity of FUT in French which does not rely entirely on lexical aspect (i.e. Aktionsart), and can provide the following empirical coverage:

1. (i) the preference for a temporal interpretation with both eventives and statives for simple future sentences;

2. (ii) the availability of the epistemic reading with statives in simple future sentences;

3. (iii) the impossibility of the epistemic reading with eventives in the absence of the ‘rescuing strategy’ of mentioning a future event of ratification;

4. (iv) the unproblematic compatibility of the epistemic interpretation with the future anterior.

While all the work that we are aware of acknowledges these preferences, we know of no previous attempt to explain them in a principled way. The view we are about to present covers these preferences, and makes sense of the distributions observed in section 2.1, without appealing to semantic ambiguity.

3.1 Ratificational modality

Our hypothesis is that FUT in French makes a triple contribution, at different levels: temporal, modal (in the semantics) and evidential (in the pragmatics). That FUT contributes modality and temporality is now a common assumption in the literature on FUT across languages (cf. among many others, Thomason, Reference Thomason, Gabbay and Guenthner1984; Kaufmann, Reference Kaufmann2005; de Saussure and Morency, Reference de Saussure and Morency2011; Giannakidou and Mari, to appear-a). In line with the previous literature on the matter, as already mentioned, we build on the assumption that FUT forward-shifts the time of evaluation of the modal (or, in other terms, represents future verification; cf. Schrott, Reference Schrott1997; de Saussure and Morency, Reference de Saussure and Morency2011). The evidential dimension of the modal and the role it plays in the interpretations has instead been left out of account.

Research on epistemic modals (whether veridical or conjectural), has explored the hypothesis that these impose constraints on the context regarding the type of evidence available at the time of evaluation of the modality. Karttunen (Reference Karttunen and Kimball1972), Dendale (Reference Dendale2001), Nuyts (Reference Nuyts2001), von Fintel and Gillies (Reference von Fintel and Gillies2010), Giannakidou and Mari (to appear-a) – just to mention a few works from different frameworks – have explored the idea that must and its equivalent in other languages (such as French, Dutch, Greek and Italian) presupposes indirect evidence. Recall that, in order to felicitously utter (27), the speaker must not have seen the rain, but must have some indication that it is raining.

1. (27) It must be raining.

Importantly, when modality interacts with time, this constraint on the context holds at the time when the modal is evaluated. Let us provide an example independent of the French FUT. It is well known that a modal in the imperfect in French can have an epistemic meaning and can be evaluated in the past (see Boogaart and Trnavac, Reference Boogaart, Trnavac, Patard and Brisard2011; Mari, Reference Mari2015). Consider a scenario in which my husband sees that the drawer of the table is open, and asks me why it is open. I can reply (28) (discussed a.o. in Mari, Reference Mari2015).

1. (28) Je cherchais les clés, et elles pouvaient être dans le tiroir.

At the time of evaluation of the epistemic modal (which is past), the speaker had evidence (in this case, indirect, inferential evidence) that the keys were in the drawer. Indirect evidence that is available in the past is a constraint on the context, at the time of evaluation of the modal.

Our hypothesis is thus that the future contributes a temporal component which forward- shifts the time of evaluation of the modal component. The modality is ratificational and imposes an evidential constraint on the context of evaluation. As a consequence, FUT does not convey the fact that the speaker/hearer uses evidence at the time of utterance (unlike what has been argued for devoir, and in line with e.g. Dendale, Reference Dendale2001).

We now turn to the evidential constraint imposed by the modal. We propose the following principle (29) for FUT as a ratificational modal.

1. (29) Ratificational modal constraint. If a modal is ratificational, then the time of evaluation of the prejacent and the time of evaluation of the modal must coincide.Footnote ¹²

When the time of evaluation of the ratificational modal and of the prejacent do not coincide, ratification proceeds indirectly.

1. (30) Weakening of the ratificational modal constraint. If the time of evaluation of the modal and of the prejacent do not coincide, then ratification is indirect.

Our view of the evidential meaning attached to the modal component is one that considers it as a constraint on the relation between the time of evaluation of the modal and the time of evaluation of the prejacent. The notion of ‘direct evidence’ boils down to the temporal constraint that the time of evaluation of the modality (the time of acquisition of knowledge) and the time of evaluation of the prejacent coincide. This is a felicity constraint that the modal imposes on the context of its evaluation.Footnote ¹³

3.2 The ratificational modal constraint, FUT, and the effects on temporality

We have argued that FUT forward-shifts the time of evaluation of the ratificational modal. The ratificational modal is felicitously used if the judge has direct evidence at the time of evaluation of the modal. In our account, the constraint of direct evidence boils down to a temporal constraint that the time of evaluation of the modal and of the prejacent must coincide. Let us now consider how different interpretations of future sentences in French deal with this constraint. We propose to analyze simple future sentences as in (31) (see below).

We have stated that FUT represents future verification, in line with previous proposals. It is thus a complex operator that introduces temporal and modal information, and forward-shifts the time of evaluation of the modal. On the other hand, the time of evaluation of the prejacent (we assume that propositions are evaluated not only with respect to worlds, but also to times) can be either the present (in the conjectural interpretation) or the future (in the temporal interpretation). The temporal argument of the prejacent is thus not saturated by FUT. For simple future sentences, in line with Abusch (Reference Abusch, Guéron and Lecarme2004) and Giannakidou and Mari (Reference Giannakidou, Mari, Aloni, Franke and Roelofsen2013, to appear-a), we reconstruct an element NON-PAST, which locates the prejacent either at utterance time, or at a time that follows the time of utterance.Footnote ¹⁴ This leads to the intended interpretation, in a compositional framework, according to which FUT contributes future verification of a proposition whose time of evaluation is either the utterance time or a future time.

1. (31) FUT(NON-PAST(p))

   There is a time in the future of the utterance time at which a judge verifies whether p. p can be evaluated at t_u or at a t’, t’ ≻ t_u . (t’ ≻ t_u means that t’ occurs later than t_u ).

The question to be settled is thus what the time of evaluation of the prejacent is, as this can be either the time of utterance or a future time. By addressing this question, we ask under what conditions the temporal and the conjectural interpretations are obtained. Note that here, we are positing an underspecified lexical entry for both of them. What is underspecified for now is the time of evaluation of the prejacent, and this is what the evidential constraint allows us to establish.

As we have argued, the ratificational modal constraint requires that, if possible, the prejacent should be evaluated at the time of evaluation of the ratificational modal, which is in the future with respect to the time of utterance. This straightforwardly allows us to predict that, for simple future sentences, the temporal interpretation is the preferred one (we will return in section 4 to the system of preferences).

As depicted in Figure 1, in order to satisfy (29), the time of evaluation of p is fixed as being in the future with respect to utterance time and as coinciding with the time of evaluation of the modality. We use VER for ‘verification’.

Figure 1: Simple future sentences

We place a question mark next to p to indicate that it is not metaphysically determined whether p will be true in the actual world (and thus will be known to be true). Recall that we are working with metaphysical indeterminacy, endorsing the view according to which it is not determined whether the actual world to come (depicted as a line from t_u to t’) will be a p world or not. As we argue in section 5, the matter is settled only from the speaker's perspective.Footnote ¹⁵

As formulated, (31) allows us to derive the temporal interpretation of simple future sentences. The ratificational modal constraint thus has an effect on temporality, as it requires that the time of evaluation of the prejacent coincides with the time of evaluation of the modal.

As a consequence, our hypothesis immediately accounts for the fact that with the sim- ple future, the temporal interpretation is overwhelmingly obtained. The following example is the actual announcement of a TGV arrival at Avignon station.

1. (32) Le train arrivera en gare d’ Avignon dans quelques instants.

We return in section 5 to the epistemic interpretation of simple future sentences.

A more interesting example is (8), repeated here as (33). Recall that in the scenario proposed by de Saussure and Morency, the employee has herself put the shoes on the shelf. Asked by a customer where the shoes are, she replies (33):

1. (33) Elles seront sur le présentoir là-bas. (S&M, 2011: 219, ex. (37))

The customer (i.e. the ‘addressee’ here) is responsible for verifying where the shoes are.Footnote ¹⁶ FUT provides a future time of verification.

As for future anterior sentences, they are decomposed as in (34.a) (we specifically consider cases in which the event is located before the time of utterance).

As for (9) (repeated here as (34.b)), the perfect provides a proposition that describes the resulting state of the event described in the prejacent proposition p. At a future time, a judge (the addressee, here) can only witness the result of the event described in p rather than the event in p itself. Verification will thus proceed via indirect evidence (see Figure 2, below). Note that even in this case, we are arguing that FUT represents future ratification. As the time of evaluation of the prejacent and of the verificational modal cannot coincide, there is weakening and the ratification will proceed via indirect evidence (see (30)). Importantly though, future verification is still represented in the semantics.

1. (34.a) FUT(RES(p)) There is a time in the future with respect to utterance time, at which a judge verifies whether RES(p) holds.

2. (34.b) Context: I cannot find my watch any more.

   Tu auras laissé ta montre à l’hôtel.

Figure 2: Future anterior sentences

We obtain the following interpretation (see Figure 2).

As for (34.b), RES(p) is the proposition describing the state of having left the watch at the hotel. The truth of the proposition which describes leaving the watch at the hotel can no longer be verified ‘directly’, but only indirectly. The sentence thus conveys that, in the future, the judge will have indirect evidence for determining the truth of such a proposition.

Up to now, we have proposed that FUT represents future verification in all cases (including the epistemic case with the future anterior). At this point a number of issues still require an explanation. First, we must explain the system of preferences observed (see section 2.2). Second, we must show how we deal with the so-called epistemic interpretation of simple future sentences and more generally clarify the nature of the ‘conjecture’ in the epistemic interpretation (see section 5).

5.1 There is no inference of uncertainty

There is a general consensus for proposing that, if FUT conveys future verification, then at the time of utterance, the speaker is in a state of uncertainty (either about the future (in the temporal reading) or the present (in the epistemic reading).Footnote ¹⁷

Strikingly, however, future sentences can manifest Moore's paradox. Discovered by the English philosopher Moore, this paradox is the phenomenon whereby the combination (using either and or but) of two logically compatible propositions results in a contradiction. Moore's paradox features epistemic modals such as must or epistemic attitude predicates such as believe and for this reason it belongs to the class of epistemic paradoxes.

Moore's paradox has been used to show that epistemic must does not allow for uncertainty (for a discussion of this paradox in the linguistic literature, see Kissine, Reference Kissine2008; von Fintel and Gillies, Reference von Fintel and Gillies2010; Giannakidou and Mari, to appear-a): see (39) (pace Karttunen, Reference Karttunen and Kimball1972). If must allowed for uncertainty (or ‘weakness’, in Karttunen's words), the perhaps continuation would be acceptable.

1. (39) It must be raining, #but perhaps it is not raining.

The same observation can be replicated for future tense (see Kissine, Reference Kissine2008).

1. (40) Il arrivera à 4 heures, #mais peut-être il n’arrivera pas à 4 heures.

2. (41) Tu auras laissé ta montre à l’hôtel, #mais peut-être tu ne l’auras pas laissée à l’hôtel.

This shows straightforwardly that there cannot be an inference of uncertainty, as the peut-être continuation would have been allowed. In other words, future sentences convey the idea that the speaker is confident that p will be verified as being true.

This conclusion is also in line with Schrott's (Reference Schrott1997) claim that FUT involves commitment. Whereas future sentences exploit future metaphysical indeterminacy, they convey epistemic certainty.

Nonetheless, there seem to exist contradictory data to the claim that future sentences convey certainty, most notably in relation to the use of epistemic adverbs (see Tasmowski and Dendale, Reference Tasmowski, Dendale and van der Auwera1998; Dendale, Reference Dendale2001; Kissine, Reference Kissine2008). FUT is indeed compatible with such adverbs, including weak ones such as maybe.

Note that the same variety of adverbs can be used in both the epistemic and the tempo- ral readings, regardless of whether the time of evaluation of the prejacent is utterance time ((42)–(43)) or a future time ((44)–(45)) (examples accessed in February 2015).

1. (42) Mais il aura certainement oublié que ce souhait n’avait pas de chance d’être exaucé tant que le nord du pays resterait entre les mains des rebelles.

   http://www.guineeconakry.info/article/detail/duel-dentre-deux-tours-francois-hollande-plus-debatteur-que-prevu/

2. (43) Il aura peut-être voulu dénoncer, comme déjà un certain Jean-Jacques Servan-Schreiber avant lui dans les années 70, une économie de marché.

   http://www.lejournaltoulousain.fr/archives/points-de-vue/liberte-de-pensee/leffet-canto-1622

3. (44) Je tomberai sûrement toujours en marchant. Je me cognerai sûrement encore.

   http://que-des-belles-paroles.skyrock.com/2877600006-Mes-cheveux-ne-seront-jamais-parfait-ma-frange-partira-toujours-en.html

4. (45) ‘On partira peut-être pas en vacances ensemble’, précise l’ex-entraîneur de l’Olympique de Marseille.

   http://www.lexpress.fr/actualite/sport/football/didier-deschamps-ne-partira-peut-etre-pas-en-vacances-avec-laurent-blanc_1201022.html

The role of these adverbs in modal sentences would require a whole study in itself, which we leave for future investigation. We only note with Bonami and Godard (Reference Bonami, Godard, McNally and Kennedy2008) and Ernst (Reference Ernst2009) that these are adverbs that take the whole proposition in their scope.

One option is to assume that the decomposition of a sentence like (45) is as in (46).

1. (46) MAYBE(FUT(p))

It is not settled in the literature whether adverbs contribute expressive meaning (e.g. Mayol and Castrovejo, Reference Mayol and Castroviejo2013; Giannakidou and Mari, to appear-b) or whether they are syntactically complex (Kissine, Reference Kissine2008). We do not take a position on this issue.

What matters for us is that on both these views, FUT is not a quantifier of unspecified force, as Dendale (Reference Dendale2001) seems to assume (although Dendale does not overtly mention quantification for FUT, he nonetheless proposes that FUT does not always reveal certainty). If its force were unspecified, it would not manifest Moore's paradox and the sentence in (40) would be acceptable.

Overall, we do not endorse the view according to which there is an inference of uncertainty in the use of FUT. So where does the ‘conjectural interpretation’ of statives and future anterior come from? Our proposal, which we present in the next section, is that the conjectural effect does not correlate with uncertainty, but with indirect evidence.

5.2 Epistemic inference and indirect evidence

Thus far, we have argued that future sentences represent future verification (direct or indirect), and that, in the epistemic reading, the prejacent (or RES(p)) is evaluated at a future time.

We propose that the ‘conjectural interpretation’ is indeed an inference that arises when the object of the verification (p or RES(p)) already holds at the time of utterance and the speaker has no direct evidence. The speaker nonetheless believes that p or RES(p) is true at t_u , and has indirect evidence grounding his/her belief. We call this inference the ‘epistemic effect.’ If the judge acquires direct evidence in the future, at the time of utterance, s/he only has indirect evidence. Our bottom line is that there is not, semantically, a ‘conjectural’ future, but only a ‘conjectural effect’ defined as an inference. Thus we use the label ‘conjectural interpretation’ somewhat improperly to target those cases in which the ‘conjectural effect’ arises.Footnote ¹⁸

Note that having indirect evidence and basing a conjecture on indirect evidence at the time of utterance is not a requirement on the context at this point. FUT represents (semantically) future verification only. In (8), the speaker has direct evidence (the employee knows where the shoes are at the time of utterance) and the future sentence is acceptable. In this case, the ‘conjectural effect’ does not arise, and the sentence conveys future direct verification.

We can derive a parallel conclusion for future anterior sentences. FUT always represents future ratification. In the conjectural interpretation evident in examples (9)–(18), RES(p) holds at the time of utterance. At this time, the speaker has indirect evidence and the conjectural effect arises.

This epistemic inference (again, not an inference of uncertainty) can be cancelled with the future anterior as well, and this is what happens with the ‘present ratification use’ (also called (Caudal, Reference Caudal2012) the ‘futur de bilan’). In example (11), repeated here as (47), the speaker has the requisite knowledge at the time of utterance.

1. (47) Au-delà des résultats obtenus, le grand mérite de cet expert-tacticien aura été de créer le consensus autour de sa propre personne. (Le Monde, 12 June 1998, p. 15)

This case is parallel to (8), where the speaker already has knowledge at t_u and nonetheless represents future ratification.Footnote ¹⁹ The ‘present ratification’ is typically employed in contexts where notable facts are being presented. Future generations will be in charge of the verification, just as in (8), the hearer will be in charge of the verification. Recall also (12) repeated as (48).

1. (48) Si Buzz Aldrin a été le second homme à poser le pied sur la Lune, il aura été le premier à léviter quelques centimètres au-dessus du sol sur le mythique hoverboard.

   http://www.20minutes.fr/high-tech/1471391-20141030-astronaute-buzz-aldrin-teste-hoverboard

By using FUT while in possession of knowledge about the resulting state at the time of utterance, the speaker conveys the persistence of that state and its future verifiability.

Overall, from the perspective of the speaker, cases (8), (47) and (48) are parallel. FUT represents future ratification, but the epistemic inference is cancelled as the speaker has knowledge at the time of utterance. Our conclusion is thus that there is no such thing as a ‘conjectural interpretation’ coded in the semantics of the future. There is an epistemic inference, which arises when (i) p and RES(p) already hold at t_u , and (ii) the speaker has indirect evidence at t_u . This epistemic inference can be overridden, and a future sentence can be felicitously uttered when the speaker has direct evidence that p (or RES(p)) is true at t_u (as in (8), (47) and (48)).