LOGICAL AND DISCURSIVE PROPERTIES OF MODALITY Fernando García-Murga University of the Basque Country (UPV/EHU)1 Abstract. The (extended) standard approach to modality claims that modal expressions are semantically characterized by three parameters: modal force, conversational background and modal time. However, these parameters cannot by themselves (a) explain the behaviour of epistemic modality with respect to temporality and propositional operators, and (b) determine the type of modality in context. Epistemic modality in non embedded sentences anchors modal time to utterance time. However, epistemic modality with a past perspective is attested, which seems to mean that epistemic modality sometimes interacts with tense. With respect to the proposition expressed, epistemic modality sometimes makes no contribution to the truth conditional content of the modalized sentence, but measures the degree of confidence the speaker shows towards the preyacent; a fact that impinges epistemic modality with a flavour of evidentiality. Another serious problem for the standard approach is that a propositional constraint of quantification is not enough for automatically deriving the type of modality. All these questions seem to militate against a uniform semantic account of modality. However, a semantically uniform analysis of modality can be maintained if the pragmatics of modality can offer an account for these problems. The hypothesis that is pursued in this paper is that modality interacts with three different contextual abstract structures: (i) The Focus of Current Situation (FCS), (ii) the Common Ground (CG) and (iii) the To-Do List (T-DL). If a modal expression interacts with CG, it receives a dynamic interpretation; if interacts with T-DL, it has a priority interpretation and if interacts with FCS, it has an epistemic interpretation. Keywords: modal time, temporal perspective, propositional operators, common ground, evidentiality. 1. Introduction2 Modality (the linguistic expression of possibility) is expressed across languages by syntactically different types of expressions. A well-known cross-linguistic semantic characteristic of modality is that modal expressions have multiple 1 E-mail address: [email protected] 2 This paper is part of the investigation Project IT769-13 of the Basque Government, UFI11/14, LIngTeDi/HiTeDi of the University of the Basque Country (UPV/EHU) and FFI2011-30074-C02-02 of the Spanish Department of Economy and Competitiveness. 1 meanings. It is not difficult to imagine contexts where the utterance of the Spanish sentence (1) receives deontic or ability readings. Similar multiplicity of meaning is observed in other languages. However, modal meanings can be blocked by different mechanisms, as examples (2)-(5) show and some of them are language- specific: (1) La presidenta puede hablar euskera DEF president can-PRES speak-INF Basque The president can speak Basque (2) La presidenta sabe hablar euskera DEF president knows-PRES speak-INF Basque The president can speak Basque (3) Dada la normativa del Congreso, la presidenta no puede hablar euskera Given DEF normative of DEF Congress, DEF president NEG can-PRES speak Basque Given the normative of the Congress, the president cannot speak Basque (4) La presidenta puede haber estado en el despacho DEF president can-PRES have-PERF been in DEF office The president can have been in the office (5) Puede llover Can-PRESS rain It can rain In sentence (2) the modal meaning of the verb ‘saber’ (to know) has only an ability reading, a fact that has to be marked in its lexical entry. In (3), it is linguistic context that forces a deontic reading. What (4) illustrates is that epistemic reading is mandatory in Spanish with composed past tense. Finally, the meteorological predicate of sentence (5) forces an epistemic reading. The conclusion is that modality interacts with multiple linguistic phenomena. At this point it is worth offering a broad classification of modality, even though any classification has the risk of unconsciously convey theoretical prejudices. Following the work of Portner (2009), we will assume that there are three main types of modality and different subtypes: 2 (6) Epistemic modality Priority modality Dynamic Modality Subjective Deontic Circumstantial Intersubjective Teleological Objective (alethic) Bouletic Ability In epistemic modality, the source of the possibilities under consideration is an agent’s epistemic state. Epistemic modality is subjective if the agent of the epistemic state is the speaker and intersubjective if the agent is a group. In the case of priority modals, possibilities come from some standard of behavior, be it a norm, a goal or a will that is in force in the circumstance of interpretation. Finally, dynamic modality comes from possibilities related to states of affairs; states of affairs that can be constrained to some set of variable circumstances, a whole range of well-established circumstances or circumstances that conform an individual’ s abilities. The standard semantic analysis of modality (Kratzer 1977, 1981 [2012]) assumes that modal expressions are uniformly interpreted following two different parameters: The modal force and the conversational background, which, in turn, is composed of the modal base and the ordering source. The modal force determines the existential or universal quantification over alternatives. The type of quantification of a modal expression is conventionally fixed in its lexical entry. The modal base is a function f that, in terms of Possible Worlds Semantics, takes as argument the circumstance of interpretation and gives a set of possible worlds3. Finally, the ordering source is a function g that takes the set 3 The circumstance of interpretation can be the ‘real’ world at hand (for priority and dynamic modality) or the world as the epistemic agent figures it (for epistemic modality). 3 of possible worlds picked up by f and puts them into an ordered structure. The ordering source fixes the order of possible worlds according to preferences (that have a doxastic or an informational nature in case of epistemic interpretation, a deontic nature in case of a deontic one, etc.) and selects the ‘better’ worlds for the domain of quantification. That is, modal base and ordering source select together the domain of quantification of modal expressions. Modality does not only involve quantification over selected alternatives. Given the metaphysical conception of time and modality in (7), modality has also an inherent temporal meaning. As the schema in (7) shows, time draws on linearly and, at any given time t , different paths open toward the future. Each opened path represents n an alternative: (7) ALT 1 w ALT 0 2 t ALT n 3 … The time with respect to which the modal base is applied constitutes the initial point of the temporal span where the alternatives are situated. This time span is called ‘temporal perspective’ and is represented by ‘[t,_)’ (Condoravdi 2002). Temporal perspective expands toward the future without limit. Given all these elements, the lexical entries for modal expressions are as follows: (8) a. POSSIBLE: λwλfλgl tl P ∃w’[w’ ˛ f(w) & g(f (w)) & AT([t,_), P(w’))] t t t b. NECESSARY: l wλfλgl tλP " w’[w’˛ f (w) & g(f (w)) fi AT([t,_), t t t P(w’))] Where w’ is any possible world into the conversational background and P is the semantic representation of the prejacent (i.e. the sentence under the modal operator), so that P(w’) means that P is true in w’. The predicate ‘AT’ makes the work of inserting eventualities in time (Kamp & Reyle 1993). The standard semantic analysis of modality constitutes a very elegant way of solving the systematic multiplicity of meanings that modal expressions show across 4 languages4. However, there are deep problems that have been pointed out in the literature. This paper will focus on two of them. The first problem is a foundational one, and was posed by Nauze (Nauze 2008). The objection basically says that even though the conversational background functions f and g select possible worlds (i.e., propositions), they cannot a priori identify those propositions as epistemic, deontic or as pertaining to any other type of modality (Nauze 2008: 154-155). The conclusion that Nauze draws from this is that conversational background fixes the content for evaluation but not the type of modality. Therefore, modal expressions that do not show strict lexical restrictions have to be, after all, ambiguous. The second problem comes from the fact that there are different linguistic phenomena that systematically take the epistemic interpretations apart from the other types of modal interpretations5. Nothing in the semantic apparatus of modality can establish a dichotomy between different types of modality. The linguistic phenomena that show such a dichotomy have to do with (i) the interaction between modality, time and aspect (Condoravdi 2002, Hacquard 2006, Demirdache & Uribe-Exebarria 2008, 2014, Ramchand 2012) and (ii) the interaction of modality with propositional operators like negation or conditional (Drubig 2001, Papafragou 2000, 2006, Portner 2007a, von Fintel and Gillies 2007a, 2007b). The guiding idea of this paper is that the logical properties of modality cannot by themselves solve the aforementioned problems. In this sense, logical properties of modality have to be put at work hand by hand with its discursive properties. The paper is organized as follows. The next section analyzes the interaction between modality, time and aspect. In section three, the discussion will be focused on the proposition expressed by epistemic modality and the allegedly evidential character of epistemic modality. Section four is devoted to the analysis of the abstract structure of discourse and to the discursive properties of modality. Given the hypothesis proposed in section four, the puzzles of the standard semantic theory 4 If modal expressions were ambiguous, it would be impossible to explain the cross- linguistic nature of the multiplicity of modal meanings. 5 The systematic distinction between epistemic modality and the other types of modality has led the scholars to lump the non-epistemic modality under one simple label; so, ‘root modality’ is the usual term for non-epistemic modality. In this paper, a distinction will be drawn between epistemic and non-epistemic modality. 5 of modality pointed out in this paper will be addressed. The paper ends with some conclusions and some proposals for future research. 2. Modality, time, and aspect All sentences display a double fold temporal dimension: The temporal location of the situation described with respect to the utterance time and the temporal structure of the very same situation. These two dimensions correspond, respectively, to the categories of Tense and Aspect. The category of Aspect divides itself into two categories: Viewpoint aspect and situation type aspect (Smith 1991). The former has to do with the visibility of the endpoints of a situation. Situation type aspect depends on the temporal extension and the telicity of the situation6. As has been said before, modalized sentences introduce an additional temporal dimension, the ‘temporal perspective’. Given that modalized sentences necessarily contain a prejacent, a temporal relation can be identified between the temporal perspective of the modal and the time when the eventuality described by the prejacent holds. This temporal relation is called ‘temporal orientation’. The crucial data with respect to temporal perspective and temporal orientation was discussed in detail by Condoravdi (Condoravdi 2002). The results can be summarized in the following table: (9) Modal Tense of the modal Temporal Temporal Interpretation verb Perspective Orientation Root Present Present Future 6 The classical aspectual categories and their features are the followings (where T represent telicity and E, temporal extension): Eventualities States (+T, -E) Eventives Activities (-T,+E) Events Accomplishments (+T,+E) Achievements (+T,-E) 6 Epistemic Present Present Present Root Past Past Future Epistemic Past Present Past The first relevant observation is that the temporal orientation of root modality is always future, but its temporal perspective depends on Tense. These data has been interpreted as follows: Root modality is such that (a) the situation designated by the preyacent is not settled when the temporal perspective comes up and (b) Tense has scope over modality. This last point is formalized as follows: (10) T(MOD(f))) In contrast, the temporal orientation of epistemic modality cannot be future and the temporal orientation (in matrix sentences) is always present. Following the generalizations, it has been claimed that: Epistemic modality is such that (a’) The situation designed by the preyacent is settled when the temporal perspective comes up (even though the speaker does not know in which ‘direction’ it has been settled) and (b’) Tense is under the scope of modality. Formally: (11) MOD (T(f))) The following examples illustrate the interpretative difference between (10) and (11): (12) The children may have won the race (13) a. In view of what now I know, the children may have won the race b. MODAL (PERFECT (the children win the race))) (14) a. In view of the circumstances in the past, the children had the opportunity of winning the race b. PERFECT (MODAL ((the children win the race))) In the epistemic interpretation (13b), the modal time [t,_) is anchored to the utterance time. The perfect shifts the time of the prejacent to the past. Therefore, the epistemic interpretation in (13a) has a present temporal perspective and a past orientation. In the non epistemic interpretation, the perfect makes the modal time 7 shift to the past. Therefore, non epistemic interpretation has a past perspective and a future temporal orientation. Since semantic interpretation goes hand by hand with syntactic configuration, it can be said that epistemic modality appears syntactically above Tense and root modality under Tense. In other words, the basic type of modality is read off from syntactic configuration (Hacquard 2006, Ramchand 2012). As for the interaction between sentence aspect and modality, the crucial point is that with present modality, only statives are compatible with an epistemic interpretation (Ramchand 2012): (15) a. The children may play the piano (root modality) b. The children may be playing the piano (epistemic modality) c. The children may be at home (epistemic or root modality) d. The children may have played the piano (epistemic or root modality) Sentence (15a), whose prejacent is eventive, has a non-epistemic interpretation. But the progressive aspect of the prejacent forces an epistemic interpretation (example (15b)). If the prejacent is stative, as in (15c), all (available) modal interpretations are acceptable. As the example in (15d) shows, with respect to aspect, no restriction applies to past modality. The interaction of modality, tense and aspect can be addressed by a system that works as follows (Condoravdi 2002, Ramchand 2012). In the case of present tense modality, if the prejacent is stative, the modalized sentence ends up with any available modal interpretation. To illustrate this case, the semantic interpretation of (16) goes as follows: (16) The child may be in the garden (=P) ∃w’[w’ ˛ f(w) & g(f (w)) & $ e [P(w’)(e) & t (e, w’) ○ [UT-T,_)] t t The overlap of the temporal spans t (e, w’) and [UT-T,_) makes the sentence compatible with the temporal orientation of any type of modality. The semantic interpretation of a present modal with an eventive prejacent differs from the interpretation of a stative because the temporal insertion of the eventuality into the modal span is an inclusive one: 8 (17) a. The child may play the piano (=Q) b. ∃w’[w’ ˛ f (w ) & g(f (w )) & $ e [Q(w’)(e) & t (e, w’) ˝ UT-T 0 UT-T 0 [UT-T_)] The eventuality time, being the preyacent an eventive in (17a) [=(15a)], has to be included in the modal span. This means that the only possible modal orientation is future. This is why epistemic reading is excluded7. No such restriction occurs when the perfect shift the modal span to the past and, so, (15d) can receive both interpretations. As has been said, epistemic modals in matrix sentences are anchored to the utterance time. In embedded sentences, the temporal perspective of epistemic modals is determined by the tense value of the main verb: (18) The parents thought that the children may have played the piano In (18), the epistemic agent is the subject of the main clause (i.e., the parents) and the epistemic state that counts for the calculation of alternatives is a past epistemic state. Even though the formal system that has been put to work seems to offer a general explanation of the main data8, there are many problems that have to be addressed. In this paper, the attention will be focused on the temporal perspective of epistemic modality. The generalization according to which epistemic modality always has a present temporal perspective is questioned by examples like (19) for English (Mari and Schweitzer 2010) and von Fintel and Gillies (2007a) and (20) for Dutch (Boogaart 2007): (19) [Context: Jack Bauer is testifying before a commission about his actions during a period when the president was kidnapped] The President could have already been dead, so I called the vice president. (20) Given what we knew at the time, John might have been the thief 7 The fact that epistemic modality cannot have future orientation can be stated in terms of a settledness presupposition: Epistemic modality presupposes that the prejacent is settled (Condoravdi 2002). However, the use in this case of the term ‘presupposition’ is misleading. 8 A point that cannot be addressed in this paper is the variability between languages with respect to the interaction between modality, tense and aspect. 9 (21) Hij moest ziek zijn He must-PAST ill be = It was very likely that he was ill „ It is very likely that he was ill These examples are taken to be cases of epistemic modality in matrix sentences with past temporal perspective. If this is correct, the modal time of epistemic modality (in non embedded sentences) would be not necessarily anchored to the utterance time and the data that has been discussed in this section would remain without explanation. 3. Modality, proposition expressed and propositional attitude In the literature on modality, phenomena related to the proposition expressed by a modalized sentence have been widely discussed (Papafragou 2000, 2006; von Fintel and Gillies 2007a, b; Portner 2007a, 2009). The basic theoretical assumption underlying the discussion is that the proposition expressed by a sentence reflects its truth conditional meaning. In this sense, any component of the proposition expressed contributes to the truth conditions of the proposition and interacts with propositional operators such as negation and implication. According to the standard analysis of modality, modal expressions uniformly contribute to the proposition expressed by the modalized sentence because all of them quantify over alternatives selected by the conversational background. However, the truth conditional behaviour of epistemic modality poses interesting challenges to the standard analysis of modality. Take the following examples: (22) a. It must be raining outside; # I see / smell the rain b. It must be raining outside; I hear the sound of water splashing in the window Example (22a) shows that the information attached to the prejacent of an epistemic modal expression cannot come from direct experience, but has to be inferred by the speaker. In the case of (22b), the sound of water serves as a premise to reach the conclusion that it is raining outside. Following this line of reasoning, the function of epistemic modality would be to establish the informational source of the prejacent and the degree of commitment toward the proposition expressed by the prejacent rather than to quantify over alternatives. That is, in epistemic modality, 10