Journal of Semantics 22: 231–280 doi:10.1093/jos/ffh025 Advance Access publication May 3, 2005
Conditional Truth and Future Reference STEFAN KAUFMANN Northwestern University
Abstract
1 INTRODUCTION Two topics in the study of semantics which have recently received much renewed attention are conditionals and expressions of modality on the one hand, and tense and temporal reference, on the other. Each of them has its own history of increasingly sophisticated formal theories. Curiously, however, despite the wide currency of early accounts of the interaction between modal and temporal elements in the semantics of certain expressions (e.g., Dowty 1977, 1979), most subsequent work in these areas addressed only one at the exclusion of the other. The standard approaches to conditionals and modality in the 1980s and 1990s dealt with sets of worlds and quantification over them, often paying little attention to the temporal and aspectual properties of the sentences involved. Likewise, theories of tense and aspect typically dealt with one world at a time, abstracting away from the modal nuances arising with differences between, for instance, past and future reference. While there are doubtless advantages to such a modular approach, there is also no question that the ultimate goal of developing a comprehensive semantic treatment of modality and temporality will The Author 2005. Published by Oxford University Press. All rights reserved.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This paper proposes a compositional model-theoretic account of the way the interpretation of indicative conditionals is determined and constrained by the temporal and modal expressions in their constituents. The main claim is that the tenses in both the antecedent and the consequent of an indicative conditional are interpreted in the same way as in isolation. This is controversial for the antecedents of predictive conditionals like ‘If he arrives tomorrow, she will leave’, whose Present tense is often claimed to differ semantically from that in their stand-alone counterparts, such as ‘He arrives tomorrow’. Under the unified analysis developed in this paper, the differences observed in pairs like these are explained by interactions between the temporal and modal dimensions of interpretation. This perspective also sheds new light on the relationship between ‘non-predictive’ and ‘epistemic’ readings of indicative conditionals.
232 Conditional Truth and Future Reference
(1) a. b.
If he submits his paper to a journal, we won’t include it in our book. If he submitted his paper to a journal, we won’t include it in our book.
Two questions are of central importance in this debate. First, assuming that conditionals are built up compositionally from the connective ‘if ’ and the two constituents, what are the expressions that form the input to this operation, especially the one that becomes the antecedent of the conditional, and how does the composition proceed? Second, how should the semantic difference between conditionals like (1a,b) be characterized? Regarding the first question, I will argue in this paper that the antecedents of (1a) and (1b) are literally (2a) and (2b), respectively.1 (2) a. He submits his paper to a journal. b. He submitted his paper to a journal. This claim, trivial though it may seem, has often been denied on the grounds that the antecedent of a (1a), unlike that of (1b), does not appear to have the same interpretation in the conditional as it does in isolation: (2a) is only felicitous under a special reading which includes 1
Some speakers, including an anonymous reviewer, find that (2a) requires the presence of a temporal adverbial like ‘tomorrow’. Others believe that the future reference time may be supplied contextually (Steedman 2002).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
require more than putting together one theory of each. Recent work on the ways in which they interact and constrain each other has brought back into focus the importance of keeping each in view when analyzing the other (Condoravdi 2002; Fernando 2003; Ippolito 2003; Kaufmann 2005, among others). The present paper contributes to this line of research. I propose an analysis of the temporal and modal properties of simple sentences and indicative conditionals, which accounts for the variety of readings observed in the latter in terms of the interplay between these dimensions. I will only be concerned with indicative conditionals, ignoring the separate problem of the morphological makeup of counterfactuals. In the remainder of this introduction, I will take a first look at the relevant data. The class of conditionals that is often treated uniformly as ‘‘indicatives’’ is not as homogeneous as the label would suggest. Pairs like (1a,b) illustrate a distinction that has been the subject of a longstanding controversy.
Stefan Kaufmann 233
(3)
a. He will submit his paper to a journal. b. ?If he will submit his paper to a journal, we won’t include it in the book.
This latter position is often accompanied by the claim that ‘will’ in the antecedent, as in (3b), results in ill-formedness unless it has a volitional reading, paraphraseable in this case as ‘If he is willing to submit his paper. . .’ (Palmer 1974; Wekker 1976; Dowty 1979; Declerck and Reed 2001). I will show that none of the above assumptions are required for an adequate semantic account. Instead, I propose an analysis that is strictly compositional, in the sense that the antecedents of all indicative conditionals receive exactly the same interpretation as they do in isolation. This proposal is uncontroversial for (1b), but the argument for extending it to (1a) requires a detailed analysis of the temporal and modal makeup of such sentences, as well as an answer to the second question, the one about the semantic difference between (1a) and (1b). Regarding this question, I will argue that there is no significant difference between these sentences. This claim, too, is at odds with much of the literature, as it appears to fly in the face of rather solid intuitions to the contrary. While indicative conditionals are generally used under uncertainty about the truth value of the antecedent, the source and nature of this uncertainty is not the same in the two cases. Funk (1985: 375–376) stated the difference lucidly: In the case of [1a] the uncertainty is largely due to the fact that the state-of-affairs described and predicated does not yet exist, i.e., is
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
an element of ‘certainty’ or ‘scheduling’ (Zandvoort 1965; Lakoff 1971; Vetter 1973; Goodman 1973; Wekker 1976; Dowty 1979; Quirk et al. 1985; Comrie 1985). This connotation is absent in the antecedent of (1a). I will refer to it as the Certainty Condition (CC), leaving open until further discussion the question of its source and theoretical status. For now, what is important is that no similar contrast is felt between (1b) and (2b). Some authors argue on the basis of such observations that the antecedent of (1a), unlike that of (1b), is not really a tensed clause (Dudman 1984, 1989; Crouch 1993). Others assume that the antecedent of (1a) is (3a) at some underlying level, and that the auxiliary ‘will’ is removed by a mechanism triggered by the ‘if ’construction (McCawley 1971; Comrie 1985; Dancygier 1998).
234 Conditional Truth and Future Reference still subject to manifestation (so that it cannot be affirmed or denied—it is unverifiable) at the moment of the sentence being uttered. In [1b], however, the state-of-affairs does exist at the time of speaking (either in the positive or negative sense—it is ‘manifested’ and could thus be verified), but the speaker has not got enough information (or is otherwise not disposed) to be sure about it and hence to affirm or deny it. Accordingly, the meaning of the conditioning frame can be said to vary from ‘if it happens that. . .’ to ‘if it is true that. . .’ (emphasis in the original)
(4) a. b.
[We’ll check what was decided about these manuscripts at yesterday’s meeting.] If he submits his paper to a journal, we won’t include it in our book. [We’ll contact potential contributors next month to see which of these manuscripts are still available then.] If he submitted his paper to a journal, we won’t include it in our book.
Now, the existence of such readings is in itself not an argument against the use of Funk’s criterion in classifying conditional sentences. We could say that (1a) and (4a) are distinct, homonymous conditionals, similarly for (1b) and (4b), and accordingly classify (1a) and (4b) together on one side, and (1b) and (4a) on the other. But we would then be dealing with
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Many authors have made similar remarks to the effect that in (1b), but not in (1a), the truth value of the antecedent is either known to someone other than the speaker, or knowable in principle, or in any case somehow ‘out there’ (Close 1980; Dancygier 1998; Declerck and Reed 2001; Garrett 2001). This distinction is real, and it constitutes an important part of the motivation for my proposal. However, I do not consider it a criterion for distinguishing between conditional sentences. Rather, it is readings of such sentences that fall in one class or the other. Consider the ‘scheduling’ reading on which (2a) is felicitous. Notice that the antecedent of (1a) can have this reading (‘If he is (now) scheduled to submit his paper. . .’), and that (1a) under this interpretation shares the property of ‘verifiability’ of a state of affairs with (1b). Similarly, the antecedent of (1b), its Past tense notwithstanding, can refer to future states of affairs that are not yet ‘verifiable’ at speech time (‘If (it turns out in the future that) he submitted his paper. . .’). Under this ‘past-in-the-future’ interpretation, the criterion would group it together with (1a). Natural contexts in which these readings become salient are given in (4a,b).
Stefan Kaufmann 235
2 PRELIMINARIES The account is based on the standard analysis of conditionals, spelled out formally in terms of quantification over possible worlds. A sentence ‘if A then C’ asserts of a set of worlds K that C is true at all, most or few worlds in K at which A is true (depending on the modal involved). Authors differ in their assumptions as to whether K represents the speaker’s knowledge (Ramsey 1929) or beliefs (Stalnaker 1968), the common ground (Stalnaker 1975) or some other conversational background (Kratzer 1979, 1991). I will focus on two kinds of conversational backgrounds: objective and doxastic (or subjective) ones. This distinction has traditionally proven useful in the analysis of conditionals and has a central place in related areas as well; cf.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
four sentences instead of two, and there had better be good arguments for postulating such a cross-cutting taxonomy. I will show that on the contrary, there are good arguments against such a move, not the least of them being that a simpler analysis is possible, as I will demonstrate. Moreover, within this analysis, the intuitive difference alluded to by Funk and others falls out as a consequence of a semantic variation whose statement makes no reference to tenses or other structural properties of the sentences involved, nor to extralinguistic notions such as the status of states of affairs. Instead, the difference involves the relation between the speech time and the time at which the constituent clauses are evaluated. The semantic effects observed by Funk and others follow as a consequence of this difference, together with certain assumptions about the interaction of time and modality that are built into the model. The details of this account have to wait until some formal preliminaries are in place. Throughout this paper, I will distinguish between predictive and non-predictive readings of conditionals. The former differ from the latter in that the antecedent (i) carries the CC when used in isolation, but not in the conditional, and (ii) refers to a state of affairs that is not yet ‘manifest’ or ‘verifiable’ at speech time. I take (i) and (ii) to be reflections of the same semantic property. Sections 2 and 3 set up the model-theoretic background and the framework of the compositional analysis. Sections 4, 5 and 6 are respectively devoted to the bare tenses, the modal ‘will’, and the conditional. Section 7 surveys the motivations and predictions of the proposed theory with regard to a variety of additional data. Section 8 raises some questions for future work.
236 Conditional Truth and Future Reference Condoravdi’s (2002) metaphysical/epistemic distinction in the treatment of modals in time. In addition to this quantificational semantic core, the account draws on the interaction of the two modalities with each other and with time. Time plays a central role inasmuch as many of the observations can be traced to the well-known asymmetry between a ‘fixed’ past and present and an ‘open’ future, discussed since Aristotle and incorporated in many tense logics, including the one I am going to use as a point of departure.
2.1 Models
Definition 1 (T 3 W-frame—Thomason 1984) A T 3 W-frame is a structure ÆW, T, <, æ, where W and T are disjoint nonempty sets; < is a transitive relation on T which is also irreflexive and linear; and is a relation in T 3 W 3 W such that (i) for all t 2 T, t is an equivalence relation; (ii) for all t, t# 2 T and w, w# 2 W, if w t w# and t# < t then w t# w#. I will spell out the proposal in terms of world-time pairs (Montagovian indices), and it will be convenient to extend the relations introduced in Definition 1 to this two-dimensional domain: Definition 2 Given a T 3 W-frame ÆW, T, <, æ, let I ¼ W 3 T. The relations < and are extended to I 3 I as follows: a. b.
Æw, tæ Æw#, t#æ if and only if w t w# and t ¼ t#; Æw, tæ < Æw#, t#æ if and only if w ¼ w# and t < t#.
I will usually refer to world-time pairs in I using the letters i, j, . . . , and write ‘i < j’ for ‘i < j or i ¼ j’. The -relation represents objective indeterminacy. It identifies for each time t those classes of worlds which are historical alternatives of each other at t. Definition 1 ensures that these historical alternatives form equivalence classes. Historical necessity—the notion that what has been cannot (now) have been otherwise, even though it could have been otherwise—is incorporated in Definition 3 as a condition on truth assignments. Only those assignments are admissible under this
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The formal framework is based on a version of Thomason’s (1970, 1984) ‘T 3 W-frames’.
Stefan Kaufmann 237
definition which respect the intuition that historical alternatives at t are indistinguishable at all times up to and including t. Definition 3 (History model) Let A be a set of propositional variables. A history model for A is a structure M ¼ ÆW, T, <, , Væ, where ÆW, T, <, æ is a T 3 W-frame and V:A1(I1f0,1g) is a truth assignment for A such that for all A 2 A and i, j 2 I, if i j then V(A)(i) ¼ V(A)(j).
Definition 4 (Historicity and lack of foreknowledge) An accessibility relation R in I 3 I is a. b.
modal if and only if Æw, tæRÆw#, t#æ implies t ¼ t#; temporal if and only if Æw, tæRÆw#, t#æ implies w ¼ w#.
A modal accessibility relation R a. b.
is historical if and only if iRj and i k jointly imply kRj; lacks foreknowledge if and only if iRj and j k jointly imply iRk.
A historical accessibility relation is one which behaves like facts, in that the set of worlds accessible through it at a given time must be constant across historical alternatives. A relation lacks foreknowledge if it does not ‘cut across’ equivalence classes of historical alternatives; in other words, if it cannot foresee the future. Speakers’ belief states are modelled by accessibility relations that resemble , but are subject to somewhat looser conditions to accommodate subjective uncertainty about the past. Whereas Definitions 1 and 3 jointly ensure that there is no objective uncertainty about the past, Definition 5 allows for an agent’s doxastic state to be consistent with alternative pasts, each ‘filling in the blanks’ in different ways. Definition 5 (Doxastic history) Given a history model ÆW, T, < , , Væ, a doxastic history is a modal relation ; in I 3 I that is historical and lacks foreknowledge, and such that a. b.
; is transitive, serial and euclidean;2 if i ; j, i# < i and j# < j, then i# ; j#.
2 Transitive: if i ; j and j ; k then i ; k. Serial: For all i there is a j such that i ; j. Euclidean: if i ; j and i ; k then j ; k. See Fagin et al. (1995); Stalnaker (2002).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The relation has a special status in that the properties of other accessibility relations are defined in terms of their interaction with it. Definition 4 fixes some useful terminology.
238 Conditional Truth and Future Reference
(5) a. The plane leaves at 4 pm. b. [According to the timetable] the plane leaves at 4 pm. I follow Kratzer in calling such bodies of information modal bases. Definition 6 (Modal base) A modal accessibility relation is a modal base if and only if it is historical. Notice that objective and doxastic relations are modal bases. Definition 6 ensures that modal bases in general are constant across historical alternatives, thus subject to historical necessity. What the timetable says about the departure time of tomorrow’s plane is objectively fixed, like the actual departure time of yesterday’s plane and unlike the actual departure time of tomorrow’s. No other
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
According to Definition 5, the agent a whose beliefs are traced through time by ;a is a rather idealized one: The monotonicity condition imposed in (b), similar to that for in Definition 1, implies that the only way to change one’s beliefs is by eliminating links from the accessibility relation. Thus it is impossible for worlds to become accessible: The agent never forgets or revises previously held beliefs. A further limitation is that all indices that are doxastically accessible from i must be cotemporal with i, which means in effect that the agent always knows what time it is. These restrictions are unrealistic, but they do no harm for my purposes in this paper. I will drop the diacritics from ; where no confusion can arise. Notice that ; is not required to be reflexive. Adding this requirement would make it an equivalence relation, implying that all of the agent’s beliefs are true. Such a relation could be properly called ‘epistemic’. I leave this option open; nothing hinges on it here. Notice also that although doxastic relations evolve along the same temporal dimension as history itself, the order in which facts become known is decoupled from the order in which they materialize, subject only to the constraint that they cannot be known in advance. Finally, I will have occasion to appeal to the fact that sentences are often implicitly relativized to various kinds of evidence or assumptions. For instance, (5a), besides making a claim about the actual departure time of the flight, could also state what the schedule says about that departure time. In the latter case, it is interpreted as implicitly prefixed with a reference to the timetable, as in (5b).
Stefan Kaufmann 239
conditions are imposed, although more can be specified in particular cases as needed.
2.2 Settledness
Remark 1 For any accessibility relation R: a. hhRu 0 hRu; b. hRhu 0 hRu. The conditions of historicity and lack of foreknowledge from Definition 4 translate into the following axioms: Remark 2 An accessibility relation R a. is historical if and only if hRu 0 hhRu; b. lacks foreknowledge if and only if hRu 0 hRhu. 3 The resulting logic depends on the properties of the accessibility relation; specifically, S5 for h and KD45 for h;.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The various accessibility relations introduced above give rise to different interpretations of ‘necessity’. Consider for simplicity the language of propositional logic, interpreted by a valuation function V which respects the condition in Definition 3 above. Let this language be closed under the operators hR for all accessibility relations R, interpreted as usual by universal quantification over the indices accessible via R from the index of evaluation. (I will use a different syntax for these and related statements below, but in this section there is no need to depart from the familiar format.) Necessity and possibility relative to temporal relations (e.g., h< and )<) correspond to Priorian tense operators. More interesting is the case of modal accessibility relations. Historical alternatives, belief states and modal bases all come with their own necessity operators.3 Necessity with respect to historical alternatives was first studied by Prior (1967) and subsequently termed settledness (Thomason and Gupta 1981; see also van Fraassen 1981). Some consequences of the interaction between different accessibility relations and are worth mentioning here. The following is an immediate consequence of the reflexivity of .
240 Conditional Truth and Future Reference In the doxastic case, (a) if an agent entertains a belief, it is settled that he does, and (b) to believe a sentence is to believe that it is settled. Only (a) is valid for modal bases in general. The distinction between truth simpliciter and settledness allows us to identify an attitude towards propositions that lies between ignorance and belief: One may believe that the question of the truth or falsehood of a sentence is settled without knowing which way, i.e., without knowing whether it is true or false. I will refer to this attitude as the presumption of decidedness.4
I offer this notion as the formal analog of the ‘knowability’ or ‘verifiability in principle’ that has often been identified in the literature as the defining property of non-predictive conditionals (cf. the quotation from Funk 1985, in the introduction). Definition 7 and the conditions on and ; jointly imply that sentences whose truth value depends on facts no later than the time of evaluation are necessarily presumed decided in any admissible belief state, whereas sentences about the future may fail to be. Furthermore, by historicity, hRu is necessarily presumed decided for any modal base R. 3 ENGLISH Turning now to the compositional analysis of English sentences, I adopt certain common assumptions about the relative scope of the modal and temporal elements involved. It will be clear, however, that the functioning of the system does not hinge on these assumptions: Retyping the elements and re-arranging the function application would require no more than a technical exercise. Four elements of the proposal will in the end jointly determine the interpretation of conditionals: modal bases, modal forces, tenses, and, in some cases, ordering sources. All four are represented by distinct semantic objects which enter the derivation at different points. Before turning to that part of the account, I will introduce some basic ingredients. 4
In Kaufmann (2002), I called this notion the ‘presumption of settledness’. An anonymous reviewer found this term confusing, since the intention is that it is the question of ‘whether u’ that is settled, not the proposition ‘that u’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Definition 7 (Presumption of decidedness) A sentence u is presumed decided by an agent a at i 2 I if and only if h;a(u/hu) is true at i.
Stefan Kaufmann 241
3.1 The setup
3.2 Sentence radicals Sentence radicals are interpreted relative to propositions (i.e., sets of indices). At this basic level, the evaluation involves existential quantification: A radical Rad is true of a set of indices whenever it is true according to the valuation function V at some member of this set. For reasons that will become clear in the discussion of temporal frame adverbials, it is useful to make the index of evaluation of the matrix clause available throughout the derivation. I will reserve the variables s, s0, s1, . . . (mnemonic for ‘speech index’) for this purpose. In the definition schema in (6), s is included for uniformity, but semantically vacuous. ‘Rad’ stands for sentence radicals, and x is a variable ranging over (characteristic functions of) sets of indices.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Semantic interpretations are built up and represented in a typed language which I will not define in full detail; it contains variables ranging over indices as well as (characteristic functions of) sets thereof and relations between them, constants denoting the relations in the model, and the usual logical connectives. Both the language and its intended interpretation will become clear as we go along. I need two atomic types, s for indices and t for truth values, out of which functional types are built recursively in the familiar way. I will refer to characteristic functions of sets of indices (i.e. functions of type Æs, tæ) as ‘propositions’, occasionally reminding the reader that they are not sets of worlds, but sets of world-time pairs. The advantage of working at this level will become evident below in dealing with predictive conditionals. The smallest linguistic units in the analysis are sentence radicals. Linguistically, they are found roughly at the VP level (assuming a VPinternal subject). Intuitively, they may be thought of as saturated but tenseless and non-modalized event descriptions. The ontological commitments of the analysis are minimal. I assume that a valuation function V assigns truth values to sentence radicals at world-time pairs without distinguishing them aspectually, thus avoiding the question of what exactly it means for a sentence of a particular aspectual class to be true at a given time. Such conditions can be added as meaning postulates. Nor do I treat events and states as objects in their own right, as has been proposed, for instance, by Condoravdi (2002) and Portner (2003). A more comprehensive theory may well need such complexity, but the present account does not depend on it.
242 Conditional Truth and Future Reference ð6Þ ½½Rad ¼ kxks:dk½xðkÞ ^ VðRadÞðkÞ Thus the denotations of sentence radicals are of type ÆÆs, tæ, Æs, tææ. As an example, the interpretation of the radical ‘he arrive’ is given in (7). The subscripts on x0 and s0 carry no special significance; they will be useful when the same example is subjected to further manipulations below, where additional variables x1, s1, . . . will be introduced for the sake of perspicuity. ð7Þ ½½he arrive ¼ kx0 ks0 :dk½x0 ðkÞ ^ Vðhe arriveÞðkÞ
3.3 Frame adverbials Aside from tenses, temporal frame adverbials are the primary sentenceinternal sources of temporal restrictions on the reference interval. A thorough analysis of their semantic properties would lead far afield, but some comments on their treatment are in order. To do at least some justice to the variety of frame adverbials, we should distinguish three major classes, which I will call (following Smith 1991) referential (those containing proper names, such as ‘on February 12, 2004’); deictic (those whose denotation depends on the speech time, such as ‘tomorrow’, ‘now’); and anaphoric (those whose denotation depends on the local reference time in embedding contexts, such as ‘then’, ‘the following day’). The examples I consider in this paper only contain adverbials of the second class. I give formal definitions for this class only; the treatment could be extended to the others, but doing so would distract from the main concerns of this paper. Deictic frame adverbials are interpreted in terms of accessibility relations: ‘Tomorrow’ denotes a relation which holds between an index i and all those indices that lie within the day following i. Anaphoric adverbials are interpreted similarly, but obtain their first argument from the reference time. Referential adverbials may be analyzed for uniformity in terms of a relation that is constant on its first argument.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In principle, this denotation may be applied to any arbitrary set of indices. As we will see, however, elements higher up in the derivation tree, such as tenses and modals, as well as the utterance time, conspire to delimit this region in various ways. In particular, the set of indices that actually gets passed to the radical level will always be a temporal interval (i.e. an uninterrupted sequence of indices with a constant world coordinate). Anticipating this result, I will call it the ‘reference interval’.
Stefan Kaufmann 243
ð8Þ tom ¼ kÆw; tækÆw; t#æ½t# lies in the day following t at w However, as we will see, to deal with conditionals, we need an intensionalized version of this. In the interpretation of ‘If he comes tomorrow. . .’ at an index s ¼ Æw, tæ, the ‘tomorrow’-relation must make accessible indices at alternative worlds which lie in the interval specified, at those worlds, as in (8) and relative to time t. The interpretation I will be using can be given in terms of (8) as in (9). ð9Þ tom ¼ kÆw; tækÆw#; t#æ:Æw#; tætomÆw#; t#æ The first argument of this denotation will generally be the speech index s. I will write ‘TOMs(i)’ for ‘TOM(s)(i)’, the statement that i is ‘tomorrow’, in the sense of (9), from the perspective of s. The denotations of adverbials are functions of type ÆÆÆs, tæ, Æs, tææ, ÆÆs, tæ, Æs, tæææ, that is, modifiers of sentence radicals, defined in the schema (10). ‘u’ is a variable over sentence radicals; ‘Adv’ and ‘ADV’ stand for words like ‘tomorrow’ and denotations like TOM, respectively.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In general, the denotations of frame adverbials cannot be defined in terms of temporal coordinates alone, but must be allowed to vary from world to world: The interval denoted by ‘after dinner’ depends on when ‘dinner’ is, which may itself be (subjectively or objectively) uncertain at evaluation time. Not all adverbials have this property — ‘February 12, 2004’ is a rigid designator, and the set of indices that are ‘tomorrow’ from the perspective of cotemporal indices does not vary across worlds — but for generality, I will assume that all adverb denotations depend on both coordinates, world and time, of their first argument. For deictic adverbials like ‘tomorrow’, it must be ensured that the time coordinate of their first argument is the speech time. For simple sentences, this is not a problem because the speech time coincides with the evaluation time; in embedded contexts, however, such as the antecedents of certain conditionals, the speech time is the evaluation time of the matrix clause, which may differ from that of the constituents. This is the reason why the speech time is made available throughout the derivation in the special variable s. The most straightforward way to represent adverbial denotations as relations between indices is extensional: The denotion of ‘tomorrow’ relates an index s ¼ Æw, tæ only with indices Æw, t#æ with the same world coordinate, where t# lies in the relevant future interval. Such relations can be specified along the following lines:
244 Conditional Truth and Future Reference ð10Þ ½½Adv ¼ kukxks:uðkj:xðjÞ ^ advs ðjÞÞðsÞ Thus the semantic contribution of adverbials consists in restricting the reference interval to ADVs. As an example, the application of ‘tomorrow’ to (7) above, detailing every step of the simplification, is given in (11).
4 THE TENSES This section deals with the interpretation of Past and Present tense; in Section 5, I will turn to the modal ‘will’. I will first summarize some assumptions I make about both of these expressions.
4.1 Past, non-past, and certainty There is little consensus on whether in English there is, in addition to Past and Present, a Future tense, and if so, how it is morphologically realized. Those who believe that there is, usually regard the auxiliary ‘will’ as its overt expression (Wekker 1976; Bennett and Partee 1978; Comrie 1982, 1985; Hornstein 1990; Kamp and Reyle 1993; Gennari 2003); alternatively, Steedman (2002) ascribes that role to Present morphology, which he assumes is ambiguous between Present and Future tense. Others reject the notion that English has a Future tense, noting that reference to future times is also realized by a variety of other grammatical means and generally inextricably intertwined with modality (Joos 1964;
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
ð11Þ ½½tomorrowðhe arriveÞ ¼ kukx1 ks1 :uðkj:x1 ðjÞ ^ toms1 ðjÞÞðs1 Þ ðkx0 ks0 :dk½x0 ðkÞ ^ Vðhe arriveÞðkÞÞ kx0 ks0 :dk½x0 ðkÞ ^ Vðhe arriveÞðkÞ ¼ kx1 ks1 ðkj:x1 ðjÞ ^ toms1 ðjÞÞðs1 Þ ks0 :dk½kj½x1 ðjÞ ^ toms1 ðjÞðkÞ ^ Vðhe arriveÞðkÞ ¼ kx1 ks1 ðs1 Þ ks0 :dk½x1 ðkÞ ^ toms1 ðkÞ ^ Vðhe arriveÞðkÞ ¼ kx1 ks1 ðs1 Þ ¼ kx1 ks1 :dk½x1 ðkÞ ^ toms1 ðkÞ ^ Vðhe arriveÞðkÞ
Stefan Kaufmann 245
(12)
He submits his paper to a journal.
As a consequence, (12) generally carries the CC. On the other hand, stative sentences do not carry the CC with a present reference time (13a), but do with a future reference time (13b). (13)
a. He is in his office (now). b. He is in his office tomorrow.
Furthermore, the CC is not carried by sentences in the Past. Thus the generalization is that it does not arise with past and present reference times, but only with future reference times (relative to the speech time). There is an obvious connection between this generalization and the motivation behind models of branching time. Leech (1971) notes that the futurate Present ‘attributes to the future the same degree of certainty that we normally accord to present and past events’ (p. 65; cf. also Quirk et al. 1985). In line with this intuition, I propose a shift of perspective which makes a unified analysis of present and futurate 5 This paper is not committed to a particular analysis of aspect (or Aktionsart); the framework should be compatible with any account of the constraints that it contributes. One candidate would be Gennari’s (2003) proposal that the Present induces a future reference time with all aspectual classes and that the co-temporal interpretation of statives is inferable via their ‘superinterval property’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Leech 1971; Palmer 1974, 1979; Dowty 1979; Quirk et al. 1985; Huddleston and Pullum 2002). I adopt the latter view. The use of ‘will’ is neither necessary for future reference (recall 2a above, repeated below as 12) nor sufficient for it (as witnessed by examples like ‘He will be here now’). Present tense, for instance, is another way of realizing future reference. But this raises another question: Is the Present morphology ambiguous, as Steedman would have it, or underspecified as ‘non-past’ (cf. Quirk et al. 1985)? The latter position would be preferable inasmuch as it allows for a simpler semantic analysis, but its viability depends on whether the CC can be explained independently. The problem is that Present-tense morphology comes with the CC only on its futurate use, when the reference time follows the speech time. The reference time is constrained by a variety of factors, including adverbials, context, and the aspectual properties of the sentence. With non-stative sentences like (12), the reference time always lies in the future, for reasons that I consider orthogonal to my present concerns.5
246 Conditional Truth and Future Reference
6 I use the term ‘epistemic’ here in its linguistic sense. In this sense, ‘epistemic’ modals are opposed to ‘root’ (e.g., deontic) modals (Jackendoff 1972; Hofmann 1976; Brennan 1993). There is some potential for confusion, since these ‘epistemic’ modals may have epistemic, doxastic or metaphysical readings in the logical sense. I should also mention at this point that I do not intend the same analysis to apply to root modals. They appear to be embedded under epistemic modals of the kind I discuss here, hence semantically in the nuclear scope of the operator I introduce in the next section. Related to this is the observation that they do not seem to scope out of conditional consequents (Frank 1996; Zvolenszky 2002). 7 I will not be dealing with ‘must’, which has been characterized as adding to the assertion the claim that it is supported by the available evidence only in some ‘indirect’ fashion (Veltman 1986; Stone 1994; Westmoreland 1995). I believe that this is basically correct; I will not discuss ‘must’ any further because the meaning just described makes it an evidential, placing it outside the scope of this paper.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
uses of the Present not only viable, but virtually inevitable: It is not the case that the CC is only carried by the futurate Present. Rather, it is part of the interpration of all (non-modalized) sentences, Past and Present; but it is only with future reference that it contributes an additional semantic component over and above the condition that the sentence be true. Specifically, I assume that all sentences in the bare tenses contain a covert epistemic necessity operator and are evaluated against, or predicated of, metaphysical or doxastic modal bases.6 Thus their truth conditions involve either settledness or belief, not merely truth simpliciter. The fact that their use with future reference tends to be infelicitous is explained by this strong interpretation, since what is not settled at speech time cannot possibly be known. Recall that this follows from the conditions on doxastic histories in Definition 5 above. The reader may wonder at this point why I choose to build this modal element into the truth conditions. This decision has a precedent in Dowty (1979:158), although he included it only for the bare Present. On the other hand, Steedman (2002) consigns it to pragmatics. Indeed, a pragmatic alternative would seem possible and tempting. I will consider it in Section 7.3.1 and show that it is ultimately untenable in view of data from predictive conditionals. Thus the bare Present combines a ‘non-past’ temporal component with a universal modal force. In this modal component, it contrasts with overt modals like ‘will’, which I will turn to in Section 5.7 Since my treatment of the tenses is partly motivated by that analysis, a brief preview is in order. It is customary to decompose modals like ‘will’ into Present tense scoping over an abstract morpheme WOLL (Abusch 1997, 1998; Condoravdi 2002). I adopt this approach below, assuming that ‘woll’ contributes a modal force (which differs somewhat from the Certainty expressed by the bare tenses) whereas the tense contributes nothing but
Stefan Kaufmann 247
the meaning of ‘non-past’. It is useful to separate these two semantic elements across the board, both for expository purposes and in the interest of a uniform analysis of the Present tense in sentences with and without overt modals. For the bare Present, I postulate an abstract modal element which I will label ‘Ø’.8 Thus a sentence like (14a) is represented as (14b). (14)
a. He comes tomorrow. b. PRES(Ø(tomorrow(he come)))
Formally, the denotation of Ø combines with the embedded sentence radical (possibly modified by frame adverbials) to form what will turn out to be in effect a generalized quantifier over indices. Its semantic type is given in (15). Recall that the type of the first argument, ÆÆs, tæ, Æs, tææ, is the type of sentence radicals. ð15Þ
ÆÆÆs; tæ; Æs; tææ; ÆÆÆs; Æs; tææ; ÆÆs; Æs; tææ; Æs; Æs; tææææææ
I introduce (for the case of Ø) a variant of the necessity operator ‘h’ familiar from modal logic. It is used here as a two-place operator, both of whose arguments are of type Æs, tæ, i.e., (characteristic functions of) sets of indices. ð16Þ
½½Ø ¼ kukTkRkiks:hðkj:iRjÞðkj:uðkk:jTkÞðsÞÞ
The two arguments of ‘h’ in (16) form its restriction and nuclear scope, respectively. The interpretation of the expression resulting from applying (16) to its arguments is as expected (I write ‘U’, ‘W’ etc. for functions of type Æs, tæ): h(U)(W) ¼ 1 iff for all indices i# such that U(i#) is true, W(i#) is true. The first argument, U, anchors this evaluation to 8 In my view, this analysis does not entail a commitment to a syntactic analysis of the bare Present which actually includes a morpheme meaning ‘Ø’. Perhaps the universal modal force is simply a default way of interpreting non-modalized sentences. On the other hand, phenomena like emphatic do-support do lend some support to the idea that there might be some syntactic motivation for this analysis.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
4.2 Modal force
248 Conditional Truth and Future Reference the perspective of a particular index (i in 16 above). As an example, consider (17), again spelled out in detail: ð17Þ ½½Øðtomorrow ðhe arriveÞÞ ¼ kukTkRki2 ks2 :hðkj:i2 RjÞðkj:uðkl:jTlÞðs2 ÞÞ ðkx1 ks1 :dk½x1 ðkÞ ^ toms1 ðkÞ ^ Vðhe arriveÞðkÞÞ ¼ kTkRki2 ks2 :hðkj:i2 RjÞ ½kx1 ks1 :dk½x1 ðkÞ ^ toms1 ðkÞ ^ Vðhe arriveÞðkÞ kj ðkl:jTlÞðs2 Þ
¼ kTkRki2 ks2 :hðkj:i2 RjÞ ks1 :dk½jTk ^ toms1 ðkÞ ^ Vðhe arriveÞðkÞ kj ðs2 Þ ¼ kTkRki2 ks2 : hðkj:i2 RjÞðkj:dk½jTk ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ
The resulting expression combines with two accessibility relations. These relations, signified by the variables T and R in (17), are temporal and modal, respectively.
4.3 Tense Since the modal force is parceled out to the abstract element Ø, the translation of the tenses can be very simple: PRES and PAST denote nonpast and past, respectively. Formally, they are accessibility relations, i.e., of type Æs, Æs, tææ.9 ð18Þ
9
a: ½½pres ¼ kikj:i < j b: ½½past ¼ kikj:j < i
One could include at this point a means for incorporating contextually given reference times. To add a reference interval rs, (18a) is changed to ‘½½PRES ¼ kikj.i < j ^ rs(j)’, where rs is a variable of type Æs, tæ that is locally free but contextually bound; similarly for (18b).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
¼ kTkRki2 ks2 :hðkj:i2 RjÞ ks1 :dk½kl½jTlðkÞ ^ toms1 ðkÞ ^ Vðhe arriveÞðkÞ kj ðs2 Þ
Stefan Kaufmann 249
The combination of (17) with Present tense is spelled out in (19). (I switch notations from ‘jTk’ to ‘T(j)(k)’ in mid-derivation for readability.) ð19Þ ½½Øðtomorrow ðhe arriveÞÞðpresÞ ¼ kTkRki2 ks2 : hðkj:i2 RjÞðkj:dk½jTk ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ ðklkm:l < mÞ ¼ kRki2 ks2 :hðkj:i2 RjÞ ðkj:dk½klkm½l < mðjÞðkÞ ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ
hðkj:i2 RjÞðkj:dk½j < k ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ
In the derivations that follow below, I will skip over simple reductions like that from ‘klkm[lRm](j)(k)’ to ‘jRk’ in (19).
4.4 Modal base Expressions like (19) require one more accessibility relation. This is the modal base, again of type Æs, Æs, tææ. It determines whether the sentence is asserted about (or evaluated against) historical alternatives or a belief state. I use the symbols ‘’ and ‘;’ for metaphysical and doxastic modal bases, respectively.10 ð20Þ
a: kikj:i j b: kikj:i ; j
The application of (19) to the objective accessibility relation is shown in (21). 10
Also like tenses, these arguments can be subject to contextually provided restrictions. This happens, for instance, in sequences of conditionals: Even if (b) is true in isolation, it is odd in the context of (a), where it tends to be interpreted as (c). a. If I win a million, I will quit my job. b. If I quit my job, I will be poor. c. If I win a million and quit my job, I will be poor. Such a restriction can be incorporated in much the same way as for the tenses, as a free variable which is then bound at the discourse level. I have no occasion to do so in this paper, however.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
¼ kRki2 ks2 :hðkj:i2 RjÞ ðkj:dk½km½j < mðkÞ ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ ¼ kRki2 ks2 :
250 Conditional Truth and Future Reference ð21Þ ½½Øðtomorrowðhe arriveÞÞðpresÞðÞ ¼ kRki2 ks2 : hðkj:i2 RjÞðkj:dk½j < k ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ
ðklkm:l mÞ ¼ ki2 ks2 : hðkj:i2 jÞðkj:dk½j < k ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ
4.5 Matrix clauses
ð22Þ kXks½XðsÞðsÞð21Þ ¼ ks:hðkj:s jÞðkj:dk½j < k ^ toms ðkÞ ^ Vðhe arriveÞðkÞÞ Figure 1 recapitulates the derivation of this example. The resulting truth conditions for some particular index s are as follows: (23) (22) (s) ¼ 1 a. iff for all j such that s j, there is a k such that j < k and toms ðkÞ ¼ 1 and V(he arrive)(k) ¼ 1; b. iff it is settled at s that he arrives on the next day. Nothing requires the modal base in (21) to be the objective accessibility relation . The application to ; instead of corresponds to a different reading for the sentence, which, however, still implies settledness. For recall that due to the lack of foreknowledge of ;, a sentence cannot be true throughout an agent’s belief state without also being settled throughout that belief state. The sentence is peculiar regardless of the modal base, since this degree of certainty about the future can rarely be attained. In the following section I discuss the limited range of circumstances in which it can.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The last step is to turn (21) into (the characteristic function of) a proposition. This is accomplished by a special operation, which applies only to matrix clauses. Its sole purposes is to ensure that deictic frame adverbials are interpreted from the perspective of the speech time. The operator is of type ÆÆs, Æs, tææ, Æs, tææ, defined as kXks.X(s)(s). (Here and below, I will use the dedicated variable ‘X’ to range over objects of type Æs, Æs, tææ that are not modal or temporal accessibility relations.) It is easy to see that the application of this operation to (21) ‘synchronizes’ i2 and s2:
Stefan Kaufmann 251
4.6 Sources of certainty Three major ways in which the Certainty Condition may be satisfied have been discussed in the literature on the bare Present. The first is certainty that the sentence is actually true, typically in cases in which its truth can be deduced from past and present facts together with natural laws that are considered deterministic. This is the case in (24) from Goodman (1973). (24) The sun sets at 8:39 tomorrow. Such certainty is also involved in speech acts other than assertions. Both (25a) and (25b), close grammatical relatives of (24), suggest that the speaker, though ignorant of the truth value of (24), assumes that the question is already settled (and thus the answer can in principle be known). This is the attitude I called the ‘presumption of decidedness’ in Section 2 above. (25)
a. Does the sun set at 8:39 tomorrow? b. When does the sun set tomorrow?
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Figure 1 Derivation of ‘He arrives tomorrow’.
252 Conditional Truth and Future Reference The presumption of decidedness need not be common belief among the interlocutors in order for a sentence like (24) to be felicitous. It is sufficient for them to agree that it may already be settled. In (26), B’s response entails the answer to A’s question (and includes some extraneous information that B can reasonably expect to be of value to A). It is perfectly felicitous, even though A’s very question shows that she entertains the possibility that the departure time may not yet be decided. (26)
A: Has there been a decision on my itinerary? B: You leave at 5:30 tomorrow.
(27)
The plane leaves at 4 pm. . . a. but I doubt that it will. b. #but I doubt that it does.
The bare present in (27) conveys certainty about the contents of the timetable, as indicated by the infelicity of the continuation in (27b), but not about the actual departure time of the plane, as shown by both the non-contradictoriness of the continuation (27a) and the fact that (27b) has no felicitous alternative interpretation as a claim about the actual departure time. (27) with its continuation in (27a) is felicitous because it typically conveys that it is settled that the plane leaves at 4 pm with respect to one modal base (the timetable) but not with respect to another (the facts). The scheduling reading offers a strategy for listeners who are confronted with the assertion of a sentence whose truth value they believe is not yet decided. The repair consists in a reinterpretation of the sentence relative to a modal base with foreknowledge.11 What that modal base is depends on the preferences and expectations of the listener. This remedy is at work in Lakoff ’s (1971) example (28): (28)
11
a. The Yankees play the Red Sox tomorrow. b. #The Yankees play well tomorrow.
For lack of space, I cannot discuss this reinterpretation in full detail. I only note that I believe the modal base (e.g., ‘According to the schedule. . .’) enters the nuclear scope of Ø, and that the whole sentence is still interpreted with respect to, for instance, the speaker’s beliefs.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The second kind of certainty likewise requires that the truth of the sentence be deducible, but from premises which for all the speaker knows might well be false. Consequently, it does not follow that the sentence is actually true, nor that the speaker believes that it is. This is the case in the typical ‘scheduling’ reading, illustrated in (27). Suppose the speaker utters the sentence after taking a look at the timetable.
Stefan Kaufmann 253
(29)
It rains tomorrow.
Ordinary inhabitants of the world must base their predictions on the evidence available at speech time. In the case of weather patterns, such evidence is generally too inconclusive to support (29). A listener who accepts the Almighty’s assertion of the sentence and updates her beliefs accordingly will end up with an inadmissible belief state under the above definitions—unless, perhaps more rationally, she reinterprets the sentence as asserting that it rains tomorrow ‘according to the Almighty’. 5 THE AUXILIARY ‘WILL’ The last section dealt with the interpretation of the bare tenses. The second element in the account is the interpretation of modals, among which I count the auxiliary ‘will’ on its relevant reading. Discussions of ‘will’ often focus on the question of whether it is primarily a tense marker or a modal. It invariably comes with modal connotations, but this observation has led different authors to opposite conclusions. Some claim that it is a modal and that its association with future reference is a side effect of the particular modal force it is usually used to express (Joos 1964; Palmer 1974, and others). Others argue that it is a marker of futurity whose modal connotations are due to the uncertainty inherent in the future (Wekker 1976). The question of whether it is one or the other loses some of its urgency if ‘will’ is treated as a hybrid modal-temporal complex ‘PRES+WOLL’, similar to ‘PRES+Ø’, which operates in both dimensions.12 12 The detailed semantic analysis of this operator is an ongoing endeavor. See Condoravdi (2003) for alternatives to some aspects of my proposal.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The difference in felicity is due to an underlying difference in the ease with which a suitable interpretation can be found under which the sentence is presumed decided. While it is natural to interpret (28a) as a statement about the published schedule, (28b) requires more effort to find a plausible reinterpretation, likely one involving a secret agreement behind the scenes. A reinterpretation along these lines makes (28b) felicitous. It does not entail then, of course, that the Yankees actually do play well on the next day. The third kind of certainty is one that Edgington (1997) attributes to the ‘Almighty’, and which is presumably available to speakers with foreknowledge. A sentence like (29) could only be used by such a being.
254 Conditional Truth and Future Reference The main uses of ‘will’ are broadly divided between the meanings of prediction or expectation on one side and volition on the other, each with further subdivisions and perhaps with a blurry boundary between them (Leech 1971; Palmer 1974; Wekker 1976; Quirk et al. 1985). I will focus exclusively on the predictive reading.
5.1 Modal force
(30)
a. The coin will come up heads (eventually). b. ?The coin comes up heads (eventually).
While (30a) makes a genuine prediction under uncertainty, (30b) can only mean that the coin is two-headed or the experiment is rigged in some other way. This contrast arises not only in artificial cases like (30), but quite generally, as in (31). (31)
a. It will rain tomorrow. b. ?It rains tomorrow.
Such facts suggest that the modal force required for ‘will’ is somehow weaker than necessity. There are various formal ways of representing such a weaker force. Here I choose one that is familiar from Kratzer’s (1981) work on graded modality.13 The idea is that in a sentence like (31a), not all historically or doxastically accessible indices are relevant for the truth of the sentence, but only those which satisfy certain default or ‘normalcy’ assumptions. Such assumptions are represented as an ordering source in Kratzer’s theory, a function from worlds to sets of propositions which, for each world w, returns the set of propositions that are ‘normally’ true at w. This is just an intermediate step, however: Ultimately, the role of the ordering source is to induce a pre-order on possible worlds which ranks 13 An earlier version of this paper offered a probabilistic treatment at this point. The change is due to space constraints and does not constitute a retraction of the earlier account.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
As a first approximation, one might consider adopting a definition along the same lines as for Ø, saying that ‘will’ involves universal quantification over alternatives. This would result in an analysis of ‘will’ that makes it equivalent to the bare Present under my analysis. In fact, however, ‘will u’ makes a weaker claim, as shown by examples like the following. Suppose a fair coin is about to be tossed some large number of times. In such a scenario (30a) is quite clearly true, whereas (30b) is odd, regardless of how many tosses are expected to be made.
Stefan Kaufmann 255
them according to their relative likelihood:14 World w$ is at least as likely as world w# from the perspective of w if and only if all the propositions associated with w by the ordering source which are true at w#, are also true at w$. Translated into the current setup, the objects which are ranked in this way are not worlds, but indices. For simplicity, I will skip the indirect definition in terms of sets of propositions and instead represent ordering sources directly as functions from indices to pre-orders between indices.
I use the variable ‘O’ to range over ordering sources, and ‘d’, ‘d#’ etc. to denote particular instances. The intended reading of ‘k di j’ is ‘k is at least as likely (from the perspective of i) as j’. Notice that di is not restricted to indices that are cotemporal with i. For now, this generality appears useless, since the domain on which the order becomes relevant has so far been defined in terms of modal accessibility relations. Below, however, in dealing with predictive conditionals, it will be necessary to compare arbitrary pairs of indices, so this property of di will be needed. The ordering source enters the interpretation of a newly defined binary operator ‘ ’, similar to ‘h’ but sensitive to relative likelihood: (U)(W) ¼ 1 iff for all i# such that U(i#), there is some i$ such that U(i$) and i$Oii# and for all i% such that U(i%) and i%Oii$, W(i$).
5.2. The modal ‘WOLL’ As I mentioned above, I assume that ‘will’ consists of the abstract element WOLL under the scope of Present tense. WOLL contributes the modal force defined in the previous section. The non-past reference interval is contributed by PRES. Thus ‘will u’ is analyzed as PRES(WOLL(u)), contrasting with PRES(Ø(u)) and differing from the latter only in its modal force. This difference in modal force explains the preference for the use of ‘will’ over the bare Present in the case of future reference: Not only is PRES(WOLL(u)) semantically weaker than PRES(Ø(u)), but the latter is rarely true at all, except in the special cases discussed in section 4.6. The interpretation of WOLL is given in (32). 14
The term ‘relative likelihood’ was applied to such pre-orders by Halpern (1997, 2003).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Definition 8 (Ordering source) An ordering source is a relation O in I 3 I 3 I such that for all i 2 I, Oi is reflexive and transitive.
256 Conditional Truth and Future Reference It combines with an ordering source O; once this ordering source is supplied, the result is of the same type as the denotation of Ø.15 ð32Þ
½½woll ¼ kOkukTkRkiks: ðkj:iRjÞðkj:uðkk:jTkÞðsÞÞ
Once the ordering source is supplied, the derivation proceeds just as with the bare Present. Consider (33), the modalized counterpart of (22). (33)
He will arrive tomorrow.
ð34Þ
½½wollð d Þðtomorrowðhe arriveÞÞ ¼ kTkRki2 ks2 : ðkj:i2 RjÞðkj:dk½jTk ^ toms2 ðkÞ ^ Vðhe arriveÞðkÞÞ
15 The temporal component attributed to WOLL in Definition (32) ignores the forward-shifting effect it has on the reference interval even without the Present tense, as observed in Past-tense ‘would’. To take it into account, we would need an explicit representation of the reference time (see Footnote 9). Part of the meaning of WOLL would then consist in an extension of this reference time into the future. The intersective meaning of the tenses would ensure that this indefinitely extended reference interval is cropped at speech time with the Past tense but left intact with the Present (see also Condoravdi 2003). I will not explore this matter further here. Definition 32 suffices for my purposes. An anonymous reviewer suggests that ‘will’ necessarily shifts the reference time forward. I am not sure if such a strong claim is warranted. The temporal behavior of ‘will’ resembles in some respects that of the bare Present. The fact that for the truth of (ia) it is not sufficient that John be crying now, is paralleled in the fact that (ia’) must have a future reference time. But like the bare Present, ‘will’ can occur with present reference times in stative sentences; cf. (ib,b’). This suggests that the reason why (ia) cannot have a present reference time has to do with the aspectual properties of the predicate, not (only) the semantics of ‘will’.
(i)
a. John will cry. b. John will be in his office (now). c. John will be in his office tomorrow.
a’. ?John cries. b’. John is in his office (now). c’. ?John is in his office tomorrow.
I do not deny the close association of ‘will’ with future reference, but neither do I conclude that ‘will’ is not compatible with present reference. It seems, rather, that it is more generally associated with lack of certainty. My truth conditions do not incorporate this (sentences with ‘will’ are entailed by their bare-Present counterparts) because I consider it a pragmatic effect: Since ‘will’ is both semantically weaker and morphologically more complex than the bare Present, it implicates the denial of the latter. Together with the assumptions built into the model, this predicts that with past and present reference times, ‘will’ generally receives a subjective reading (see section 7.5 below). These matters certainly deserve more exploration, which I however do not attempt here.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Skipping intermediate steps, the result of applying ½½woll to the ordering source d and the temporally modified sentence radical (17) is the following.
Stefan Kaufmann 257
With tense, modal base and speech time supplied, and the matrix operator applied as before, this yields (35): ð35Þ
ks: ðkj:s jÞðkj:dk½j < k ^ toms ðkÞ ^ ðhe arriveÞðkÞÞ
The derivation is as in Figure 1, except for the difference in modal force. 6 CONDITIONALS
(36)
ÆÆs, Æs, tææ, ÆÆÆs, Æs, tææ, Æs, Æs, tæææ, ÆÆÆs, Æs, tææ, Æs, Æs, tæææææ
These remarks apply to both non-predictive and predictive interpretations of conditionals. The formal difference between these
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The preceding sections dealt with the simple sentences that appear as the constituents of conditionals. In this section I will complete the picture by offering an interpretation for the conditional connective IF. I assume that IF combines with the antecedent to form a modifier for the consequent; its semantic role consists in imposing a constraint on the accessibility relation of the matrix clause. This basic idea has been entertained before (e.g. von Fintel, 1994). The contribution of this paper is the detailed compositional account of the constraints imposed on the available readings for the conditional by the interplay between the temporal and modal elements. More specifically, the complex ‘IF+antecedent’ combines with the consequent at a point in the derivation at which the latter has received its modal force and its tense, but is yet to be applied to (or predicated of) a modal base. The role of the ‘if ’-clause is to ensure that whatever that modal base is, the consequent is only evaluated at those indices at which the antecedent is true. The type of the consequent, at the point at which it is modified by the ‘if ’-clause, is ÆÆs, Æs, tææ, Æs, Æs, tæææ. This function becomes the argument of the ‘if ’-clause, and spelling out the definition requires a variable of this type. I will use the Greek letter f for this purpose. For reasons I will discuss below, I assume that the antecedent, when it combines with IF, is not of the same type as f. Instead, it receives its own modal base before it enters the conditional, thus its derivation is almost complete, except for the application of the matrix operator which synchronizes speech time and evaluation time. So the type of the antecedent is Æs, Æs, tææ. The type of IF is the following:
258 Conditional Truth and Future Reference readings is minimal, but in the interest of clarity, I will deal with each in turn.
6.1 Non-predictive conditionals A first definition of IF is given in (37). Recall that X ranges over objects of type Æs, Æs, tææ. ð37Þ
½½if ¼ kXkfkRkiks:fðklkm:lRm ^ XðmÞðsÞÞðiÞðsÞ
(38)
If he arrives tomorrow, he left yesterday.
Writing out derivations involving IF in full detail becomes a bit cumbersome. I will abbreviate the antecedent with the constant A, whose denotation is given in (21) above. The combination of IF with A is straightforward; in (39), ½½A is ‘unpacked’ in the last line. ð39Þ
½½ifðAÞ ¼ kXkfkRki3 ks3 ½fðklkm:lRm ^ XðmÞðs3 ÞÞði3 Þðs3 Þð½½AÞ ¼ kfkRki3 ks3 :fðklkm:lRm ^ ½½AðmÞðs3 ÞÞði3 Þðs3 Þ ¼ kfkRki3 ks3 : 2 !3 klkm:lRm ^ hðkj:m jÞ 6f 7 6 7 ðkÞ ^ Vðhe arriveÞðkÞÞ ðkj:dk½j < k ^ tom s 3 4 5 ði3 Þðs3 Þ
(40) gives the denotation of the consequent at the relevant stage, tensed and with modal force, but as yet without a modal base: ð40Þ
kR0 ki0 ks0 : hðkj:i0 R0 jÞðkj:dk½k < j ^ yests0 ðkÞ ^ Vðhe leaveÞðkÞÞ
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
For illustration, I will step through the derivation of (38). We will see that by using the interpretation of IF from Definition 37, we obtain a non-predictive reading.
Stefan Kaufmann 259
Feeding (40) into (39) yields (41), which simplifies as shown. I align some sub-formulas for readability, switch again from the notation ‘i0R0 j’ to ‘R0(i0)( j)’, and ‘unpack’ ½½A in the last line.
ð41Þ ½½ifðAÞðØðyesterdayðhe leaveÞÞðpastÞÞ ¼ kfkRki3 ks3 ½fðklkm:lRm ^ ½½AðmÞðs3 ÞÞði3 Þðs3 Þ kR0 ki0 ks0 :hðkj:i0 R0 jÞ ðkj:dk½k < j ^ yests0 ðkÞ ^ Vðhe leaveÞðkÞÞ
¼ kRki3 ks3 : 2 3 ki0 ks0 :hðkj:i0 Rj ^ ½½AðjÞðs3 ÞÞ 6 7 4 ðkj:dk½k < j ^ yests0 ðkÞ ^ Vðhe leaveÞðkÞÞ 5 ði3 Þðs3 Þ ¼ kRki3 ks3 :hðkj:i3 Rj ^ ½½AðjÞðs3 ÞÞ ðkj:dk½k < j ^ yests3 ðkÞ ^ Vðhe leaveÞðkÞÞ ¼ kRki3 ks3 :h i3 Rj ^ hðkj#:j j#Þ kj ðkj#:dk½j# < k ^ toms3 ðkÞ ^ Vðhe arriveÞðkÞÞ ðkj:dk½k < j ^ yests3 ðkÞ ^ Vðhe leaveÞðkÞÞ The application of (41) to an epistemic modal base results in (42). ð42Þ
½½ð41Þðklkm:l ; mÞ ¼ ki3 ks3 :hðkj:i3 ; j ^ ½½AðjÞðs3 ÞÞ ðkj:dk½k < j ^ yests3 ðkÞ ^ Vðhe leaveÞðkÞÞ
The final result, after the derivation is closed off with the matrix operator, is (43).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
¼ kRki3 ks3 : 2 3 kR0 ki0 ks0 :hðkj:i0 R0 jÞ 6 ðk :dk½k < j ^ yest ðkÞ ^ Vðhe leaveÞðkÞÞ 7 4 5 j s0 ðklkm:lRm ^ ½½AðmÞðs3 ÞÞði3 Þðs3 Þ
260 Conditional Truth and Future Reference
ð43Þ kXks½XðsÞðsÞð42Þ ¼ ks:hðkj:s ; j ^ ½½AðjÞðsÞÞ ðkj:dk½k < j ^ yests ðkÞ ^ Vðhe leaveÞðkÞÞ s ; j ^ hðkj#:j j#Þ ¼ ks:h kj ðkj#:dk½j# < k ^ toms ðkÞ ^ Vðhe arriveÞðkÞÞ ðkj:dk½k < j ^ yests ðkÞ ^ Vðhe leaveÞðkÞÞ Ultimately, the contribution of the antecedent to the overall interpretation of the sentence is limited to the restrictor of the modal operator, the underlined part of (43). A tree illustrating the derivation of this example is given in Figure 2. The truth conditions are spelled out in (44). (44)
(43) (s) ¼ 1 a. iff for all j such that s ; j and A(j) ¼ 1, there is a k such that k < j and YESTs (k) ¼ 1 and V(he arrive)(k) ¼ 1; b. iff for all j such that (i) s ; j and (ii) [for all j# such that j j#, there is a k# later than j# and within the day following s
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Figure 2 Derivation of (38).
Stefan Kaufmann 261
such that he arrives at k#], there is a k earlier than j and within the day preceding s such that he leaves at k; c. iff it is known at speech time that he left on the previous day at all worlds at which it is settled at speech time that he arrives on the following day.
6.2 Predictive conditionals The only change in moving to predictive conditionals concerns the set of indices at which the constituents are evaluated: They are cotemporal with s for non-predictive conditionals, but may lie in the future in the predictive case. I first define the notion of the ‘forward extension’ of a modal base. Definition 9 (Forward extension of modal bases) Let R be a modal accessibility relation. The forward extension of R, denoted by R, is defined as follows: iRj if and only if for some k, iRk and k < j. In other words, R is R s <, the composition of R with temporal precedence. Intuitively, from the index i of evaluation it can access indices ‘diagonally’, covering both cotemporal and subsequent moments at all worlds accessible from i via R. This new accessibility relation, being of the same type as R itself, does not necessitate any modifications to the formal setup. The generalization of IF to the predictive case merely involves the addition of the star. The role of the restriction ‘es(m)’ in (45) will be explained momentarily.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
One property of the interpretation given in (44) deserves special emphasis: The condition added by the antecedent to the restrictor is a fully tensed and modalized sentence. In the above notation, this sentence is evaluated at indices j accessible from s via ;. Since ; is a modal accessibility relation, each such j is cotemporal with s. Due to the modal force in (44b), what is checked is not whether it is true (simpliciter) at j that the person arrives at the future time in question, but whether it is settled at speech time. Thus the antecedent can be paraphrased as ‘If it is (now) settled that he comes tomorrow. . .’ For non-predictive readings, the definition in (37) is appropriate. The generalization to predictive conditionals involves a slight conceptual complication but is formally very straightforward.
262 Conditional Truth and Future Reference ð45Þ
½½if ¼ kXkfkRkiks:fðklkm:es ðmÞ ^ lR m ^ XðmÞðsÞÞðiÞðsÞ
In the derivation (38), the result of using (45) instead of the ‘unstarred’ version (37) is similarly straightforward. Only the last line is given in (46): ð46Þ ks:hðkj:es ðjÞ ^ s ; j ^ ½½AðjÞðsÞÞ
ðkj:dk½k < j ^ yests ðkÞ ^ Vðhe leaveÞðkÞÞ With the move from R to R, the antecedent in (46) can now be paraphrased as ‘If it is settled (now or at some future time) that he arrives tomorrow. . .’. Notice, however, that the Present tense in the antecedent restricts the eligible future times to ones no later than ‘tomorrow’. An important feature of this analysis, from the point of view of compositionality, is that while the antecedent is interpreted literally at each accessible index j, without any special treatment of the temporal and modal impact of its bare present tense, it is not required that there be any worlds at which the question whether he comes or not is already settled at the time of s. This is because with ;, the index j at which A is evaluated may lie in the future from the perspective of s. It also becomes clear now why it is important to make special provisions for the availability of the global index of evaluation s to TOM and YEST, the adverbials in the constituents. ‘Tomorrow’, for instance, should not end up being interpreted as TOMj, which would be the case if it were given the local evaluation time of the antecedent. This would wrongly result in the interpretation ‘If it is settled (now or later) that he arrives the next day. . .’.16 Likewise for ‘yesterday’ in the consequent. In principle, the relation R may reach arbitrarily far into the future. However, the tenses of both antecedent and consequent, their reference times, temporal frame adverbials, and the fact that the antecedent is most typically not settled ahead of time, all conspire to 16 There may well be languages in which the word for ‘tomorrow’ has such a logophoric use, but English is not one of them.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
ðkj:dk½k < j ^ yests ðkÞ ^ Vðhe leaveÞðkÞÞ 31 0 2 es ðjÞ ^ s ; j ^ hðkj#:j j#Þ 7C B 6 ¼ ks:h@ kj4 ðkj#:dk½j# < k ^ toms ðkÞ 5A ^Vðhe arriveÞðkÞÞ
Stefan Kaufmann 263
impose more or less tight restrictions on this range, in ways which I will explore in some detail below. In addition, the unbound variable es of type Æs, tæ (mnemonic for ‘evaluation time’17) is a placeholder for a contextually given parameter imposing further restrictions on the range of R. That such contextual restrictions exist is shown by sentences like (47). Here the relevant time of evaluation is restricted by the context to fall between ‘today’ and ‘tomorrow’. (47) [Let’s wait for today’s decision regarding his travel arrangements.] Then, If he arrives tomorrow, we’ll book his room.
6.3 Relative likelihood Turning once again to the modalized case, we can now cash in on the definition of d as returning, for each index i, an order on indices that is not restricted to those that are cotemporal with i. No modification in the definition of the operator is required in order for it to function as desired with the relation R. Consider the sentence in (48), whose interpretation is given in (49). (48)
If he leaves today, he will arrive tomorrow.
es ðjÞ ^ s ; j ^ hðkj#:j j#Þ ð49Þ ks: kj ðkj#:dk½j# < k ^ tods ðkÞ ^ Vðhe leaveÞðkÞÞ ðkj:dk½j < k ^ toms ðkÞ ^ Vðhe arriveÞðkÞÞ The truth conditions of this formula can be paraphrased as follows: (50)
(49) (s) ¼ 1 iff for all j s.t. s ; j and it is settled at j that he leaves no earlier than j and within the day containing s, there is a j# s.t. s ; j#, j# d j, it is settled at j# that he leaves no earlier than j# and within the day containing s, and for all j$ s.t. s ; j$, j$ d j# and it is settled at j$ that he leaves no earlier than j$ and within the day containing s, there is a k no earlier than j$ and within the day following s such that he arrives at k.
17 Or, if one prefers Garrett’s (2001) term, ‘time of enlightenment’. It also resembles Crouch’s (1993) ‘temporal deictic center’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This sentence can be paraphrased as ‘If it is settled (later today) that he arrives tomorrow. . .’ It clearly shows that the CC is part of the interpretation of the antecedent even in predictive conditionals.
264 Conditional Truth and Future Reference
(51)
?If he leaves today, he arrives tomorrow.
Since the possibility of such falsifying circumstances usually cannot be ruled out, the strong claim made by this sentence is generally not warranted, hence the sentence sounds odd. 7 DISCUSSION The development of the formal setup is now complete. In this section I will discuss how it accounts for the various interpretive possibilities of conditionals. I will largely spare the reader the tedium of going through formally explicit truth conditions, only highlighting those aspects which are relevant to the examples at hand.
7.1 Future perspective The reader may ask why the forward shift in evaluation time—the modification of R to R—is introduced here by the interpretation of IF, rather than the modal in the consequent. The latter approach would align my treatment for predictive conditionals with extended-now theories of ‘will’, such as Abusch’s (1997; 1998) ‘n-expanding’ account. Abusch attributes the well-formedness of Past and Present tense with reference to future events, exemplified in (52), to the presence of ‘will’ in the matrix clause.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Now suppose there are indices at doxastically accessible worlds and in the relevant time range (no earlier than speech time but within the same day) at which the subject leaves and later runs into some disaster, such as a heavy storm or a car breakdown, which delays his arrival. The conditional is still true if each such index is outranked in normalcy by one at which he leaves and arrives as expected. Suppose again that the trip takes more than a full day, so that the subject cannot make it to his destination if he leaves too late in the afternoon. Again, the conditional is still true if he is expected to leave earlier, i.e. if each of the late indices is outranked in normalcy by an earlier one (not necessarily at the same world) at which he leaves and arrives as planned. Notice that for this to follow, it is essential that the ordering source compare indices across different times. In sum, (48) is true as long as it is expected that if the subject leaves today, he most likely does so on time and without encountering any unforeseeable difficulties along the way. In contrast, either of the above falsifying circumstances renders (51) false.
Stefan Kaufmann 265
(52)
On March 1, we will discuss the abstracts which were/are submitted by email.
The present paper is silent on this issue, although one way to tackle it suggests itself within the current approach, namely by endowing WOLL, like IF, with the ability to forward-extend the modal base (cf. also footnote 15). However, this modification alone would not generalize to a full treatment of predictive conditionals. The forward shift cannot be due to the presence of ‘will’ in the consequent, because it arises with other forms as well. The following examples (53 is cited here from Crouch 1993) illustrate this.18
(55)
If I smile when I get out, the interview went well. Take a small ball of dough and drop it in a glass of cold water. If it floats, you did it right. [NYT 01/11/95] Where a 45-win season would have seemed like a terrific accomplishment before the season began, now there’s a feeling that if they don’t win 55 games, something went wrong. [NYT 01/24/95]
In these sentences, the consequent refers to a future time, but the Past tense is licensed nevertheless because it is evaluated from the perspective of a still more distant future. In these examples, it is not clear what, if not IF itself, would trigger the forward shift of the evaluation time. The future evaluation time also provides the perspective for antecedent tenses other than the Present. Consider (56), which in an appropriate context may have a reading under which both the (hypothetical) arriving and the leaving of the subject occur in the future from the perspective of the speech time (‘If it turns out that she arrived in the morning. . .’). (56)
If she arrived in the morning, she left the night before.
These examples show that the forward shift in temporal perspective must be attributed to the conditional construction itself, rather than the modal ‘will’ in the consequent.
18 Here and below, the labels NYT and WSJ indicate the corpora in which the examples are attested (New York Times and Wall Street Journal, respectively).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(53) (54)
266 Conditional Truth and Future Reference
7.2 Accessibility relations The interpretation I gave for IF does not require the antecedent and consequent of a conditional to share the same accessibility relation. The antecedent enters the derivation fully equipped with its own modal base. This is not forced on me by the formalism, but it seems to be the right approach in view of examples in which the modal bases for antecedent and consequent are clearly different, such as (57). (57)
If Paul will be alone on Christmas Day, he will let us know.
(58)
a. b.
If my wife deceives me, I won’t believe it. If Reagan works for the KGB, I’ll never believe it.
Both of (58a,b) may be true, even though they would be patently contradictory if the antecedent could only be interpreted as ‘If I learn that. . .’ Thus the modality with respect to which the antecedent is
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Although (57) does have an interpretation which evaluates both constituents with respect to the speaker’s epistemic state (‘I expect that if I come to expect Paul to be alone on Christmas Day. . .’), this is not the only interpretation, and not even a prominent one. On its most natural reading, the sentence asserts that the speaker expects Paul to get in touch if (and when) Paul thinks he will be alone on Christmas Day. To capture this reading, the constituents should not be required by definition to share the same modal base. It is also worth pointing out in this connection that the formal separation between objective and subjective accessibility relations is crucial for the proper analysis of certain conditionals that are problematic for simpler accounts. Consider a conceivable alternative according to which all conditionals would be interpreted uniformly with respect to the speaker’s epistemic state, as is the case, for instance, in Data Semantics (Veltman 1985, 1986; Crouch 1993). The analog of Data Semantics in the present framework would correspond to the assumption that all sentences are evaluated uniformly with respect to ;. However, this approach is unable to account for sentences like (58a), a variant of an example attributed to Richmond Thomason by van Fraassen (1980), and (58b) from Lewis (1986).
Stefan Kaufmann 267
evaluated must be independent of that of the consequent, and it must be possible for it to be objective.19
7.3 Futurate antecedents Regarding conditionals whose antecedent refers to a time later than the consequent, I would like to address two issues. The first concerns the modal bases with respect to which the antecedent can be interpreted, the second the case of antecedents with the modal ‘will’. I will discuss these in turn.
19
(i)
The example of Thomason’s on which (58a) is based is (ia): a. b. c.
If my wife is deceiving me, I will believe that she is not. If I find out that my wife is deceiving me, I will believe that she is not. If my wife is deceiving me (currently), I believe that she is not.
This sentence arguably has a separate reading, brought out in (ib), according to which the speaker resolves to turn a blind eye on his wife’s transgressions, if any; however, this involves a different reading of ‘will’ which I am not concerned with in this paper. (ia) does not pose a problem on its predictive reading, despite the fact that the conditions on doxastic states deprive the speaker of the ability to conceive of the possibility that a sentence he believes to be false may be objectively true. For (ia) is a conditional prediction about future beliefs which the speaker may not yet have at present. (ic), on the other hand, is predicted by my account to be necessarily either false or infelicitous (due to vacuity). Whether this prediction is right or not, I believe that it is peripheral to the topic of this paper. 20 As opposed to the ‘Peircean’ interpretation relative to and the doxastic one relative to ;; see Burgess (1979) for a discussion of this terminology.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
7.3.1 A blind Ockhamist alley. In the discussion of the operator Ø included in the semantics of the bare Present, I limited the modal accessibility relations which may provide the restriction of the quantification to ones which lack foreknowledge (i.e., doxastic or metaphysical ones). This rules out a third possibility which would appear plausible at first, and on which I based an earlier proposal concerning predictive conditionals (Kaufmann 2002) which I hereby retract. This alternative approach, recast in the current framework, would use the identity relation on indices as a modal base for the interpretation of the Present. Since necessity with respect to this relation corresponds to truth simpliciter, rather than settledness, I dub it the ‘Ockhamist’ interpretation.20 Even though it is ultimately untenable, it is superficially attractive, therefore I will devote some space to its rebuttal. The Ockhamist interpretation is attractive because in combination with reasonable pragmatic assumptions, it seems to offer an elegant explanation of some of the facts about the CC observed earlier. Briefly
268 Conditional Truth and Future Reference put, the story goes as follows: The truth of sentences with bare tenses (Present and Past) is Ockhamist truth, thus (59a) is already true (at speech time) at just those indices at which he submits his paper at the future time in question. No reference to settledness is involved. (59)
a. He submits his paper to a journal. b. If he submits his paper to a journal, we won’t include it in our book.
(60)
a. If I come out smiling, the interview went well. b. If I come out smiling, the interview goes well. c. If I come out smiling, the interview will go well.
The worlds at which the speaker comes out smiling at the future time in question are ones at which it is already (Ockham-) true at speech time that he does. Suppose (60a) is true, i.e. at all those worlds the consequent is true as well. Then the Ockhamist approach affords no explanation of the fact that the same assertion could not also be made by using (60b) or (60c). Moreover, the felicity of the Past tense in the consequent of (60a) could not be explained without further stipulations, whose nature and motivation is unclear. Even worse, the fact that among the three sentences, only (60a) has a reading which is not paraphrased as ‘If it is scheduled/planned that I come out smiling. . .’,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The fact that (59a) is infelicitous, even though it may be (Ockham-) true, unless it receives a ‘settledness’ interpretation, is explained in terms of pragmatic assertability conditions, which are motivated by the interaction between doxastic and metaphysical accessibility relations. For consider the result of eliminating from the listener’s belief state all those (links to) indices at which the sentence is (Ockham-) false. If this update would result in a state with foreknowledge, one which ‘cuts across’ classes of historical alternatives, the sentence is infelicitous. In order to avoid such an outcome, the listener will accommodate the presumption of decidedness if possible (i.e. if she believes that the question may be settled). Antecedents of conditionals like (59b), on this account, are exempt from this complication because they are not asserted. This story is compelling as far as it goes. It must be dismissed, however, because a number of facts about conditionals would otherwise remain unaccounted for. To see this, consider again Crouch’s example (53) from Section 7.1, repeated here as (60a).
Stefan Kaufmann 269
(61)
If he arrives at noon, he will take the morning train.
If (61) is to be asserted about one and the same trip, its antecedent can only have a scheduling reading. 7.3.2 ‘Will’. The antecedent of (61) must receive a scheduling reading. Antecedents expressing genuine predictions, on the other hand, would be expected to be modalized with ‘will’, like their unembedded counterparts. This topic deserves some brief background discussion. As I mentioned in the introduction, the existence of conditionals whose antecedents contain ‘will’ with a predictive, rather than volitional interpretation, has sometimes been denied, possibly due to the influence of Jespersen’s writings (Close 1980). It is now generally accepted, however, that conditionals with predictive ‘will’ in the antecedent are in fact plentiful. Leech (1971) and Close (1980) discussed a number of now-standard examples (see also Dancygier 1998; Garrett 2001, and others). The following add to this list (emphasis added). (62)
If your nanny will need money in each of the next four years, she should predict her annual cash flow and invest so that money will be available as required to pay tuition bills and other costs. [NYT 08/07/94]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
whereas (60b,c) both have only that reading, remains mysterious under the Ockhamist account. On the current account, all of these facts fall out from the assumptions about the interaction between settledness and time that are built into the model. We can assume that the truth of the antecedent will not be predetermined, so that it only becomes settled at the relevant time after the interview. The consequent therefore cannot be true at an earlier time, which explains the use of the Past tense. In (60b,c), on the other hand, the Present tense in the consequent requires its evaluation time, and therefore that of the antecedent as well, to be no later than the interview. Such an interpretation is not impossible, but it automatically gives rise to a ‘scheduling’ reading of the antecedent, since the reference time is now later than the evaluation time. This is predicted by the nonOckhamist truth conditions of the Present tense. In general, the Ockhamist account offers no explanation for the fact that predictive conditionals whose antecedent refers to a future time later than that of the consequent always receive a ‘scheduling’ interpretation. Another example to the same effect is (61).
270 Conditional Truth and Future Reference (63) (64)
(65)
(67) (68)
The antecedents in all of (62) through (68) can be paraphrased as ‘If it is (or becomes) predictable that. . .’ That ‘will’ implies a somewhat weaker notion of predictability than the bare present is nicely illustrated by the two conditionals in (65): The customer may cancel the order as soon as the delay is foreseeable. An automatic cancellation, on the other hand, occurs after the 30 days have in fact passed (unless the delay is ‘scheduled’). The above examples also exhibit what appears to be a general pattern which to my knowledge has not yet been explored in sufficient detail, and for which I am not prepared to offer an explanation either: The antecedents of (62) through (65) have stative predictates; for them ‘will’ is perfectly acceptable. In contrast, (66) through (68) have nonstative predicates and seem to require the progressive; without it, their antecedents would be much less acceptable, unless they receive a volitional interpretation: (66#) If the worker will run errands, ask to see driver’s license and proof of insurance. (67#) If you will install the hardware, remember that the system will come unmonitored. . . (68#) If you will teach a course in summer session 2002 or next academic year in which such a service might be useful, remember to contact me. . .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(66)
If it will take 18 months to develop a product—a software package or high-tech device, for example—the carrying costs would be too high to use plastic, Scullin said. [NYT 09/20/94] ‘‘I really said to God: ‘I am willing to have an abortion. I don’t think I’ll ever get over it, but if I won’t be a great parent for a kid to be born to under current circumstances maybe it’s better if You recycle this one.’’ [NYT 11/30/94] If the delay will be up to 30 days, the purchaser can cancel the order and if it is more than 30 days, the order is automatically canceled unless the consumer agrees to a longer delay. [WSJ 10/12/89] If the worker will be running errands, ask to see driver’s license and proof of insurance. [NYT 08/08/94] If you will be installing the hardware, remember that the system will come unmonitored; hooking it up to a monitoring station can be difficult. [NYT 08/17/94] If you will be teaching a course in summer session 2002 or next academic year in which such a service might be useful, you are most welcome to contact me (as the administrator) to be enrolled. [university service announcement]
Stefan Kaufmann 271
In discussing the combination of ‘will’ with the Progressive, Leech (1971) made the following suggestion, which I believe holds the key to understanding these facts (see also Palmer 1979):21 It is tempting to speculate that this usage has grown up through the need to have a way of referring to the future uncontaminated by factors of volition, plan, and intention which enter into the future meanings of ‘will/shall’ + Infinitive, the Present Progressive, and ‘be going to’ + Infinitive. (62–63)
(69)
a.
If he will be alone on Christmas Day, he will let us know now.
21
Wekker (1976), who claims that non-volitional ‘will’ does not exist in conditional antecedents because he found no instances in his 600,000-word corpus, adopts Leech’s opinion about ‘will/shall + Progressive’ (p. 118) for other contexts. 22 The non-veridicality of the context may be related to this phenomenon. Palmer (1979; 150) noted similar effects in questions: (ia,b) are volitional and deontic, respectively, but these connotations are absent in (ic,d). (i)
a. Will you come to the party? b. Shall I come to the party?
c. Will you be coming to the party? d. Shall I be coming to the party?
Similarly, it seems, for negation: (ii) a. I/you/she won’t come to the party. b. I/you/she won’t be coming to the party. 23 (69) is a variant of an example from Leech (1971) which was also used, apparently independently, by Close (1980). I changed the consequent of (69), as well as (70) below, because my analysis does not cover imperatives. Leech’s version is (b), Close’s is (a): (i)
a. If you will be alone on Christmas Day, let us know about it. b. If you will be alone on Christmas Day, let us know now.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
These insightful but purely descriptive remarks raise some interesting questions: First, what determines whether or to what extent a verb, when used in the infinitive with ‘will’, is subject to ‘‘contamination’’ by factors of volition etc.? Above I noted that the predicates of the examples in which ‘will + be V-ing’ is used are non-stative, but some other, at best loosely related property—such as Agent proto-roles in their thematic structure (Dowty 1991)—might be the real culprit. Second, why does the Progressive eliminate this reading? And finally, what about the conditional antecedent is such that if it can have a volitional interpretation, it does have it there, even when this reading is not as prominent outside of that context?22 I do not at this point have an answer to these questions. The interpretation of such conditionals under the current proposal is illustrated with (69a), which we already encountered in (57) above. The truth conditions are paraphrased informally in (69c); I do not include the contextual variable es here.23
272 Conditional Truth and Future Reference 0 2 31 s ;a j ^ ðkj#:j ;b j#Þ b. ks: @ kj4 ðkj#:dk½j#
ðkj:dm½ j < m ^ nows ðmÞ ^ Vðhe let us knowÞðmÞÞ ‘I expect that he will let us know now if he (now) comes to expect to be alone on Christmas Day’.24
(70)
a. b.
c.
If he will be alone on Christmas Day, he will let us know. 0 2 31 s ;a j ^ ðkj#:j ;b j#Þ ks: @ kj4 ðkj#:dk½j#< k ^ cdayðkÞ^ 5A Vðhe be aloneÞðkÞÞ ðkj:dm½ j < m ^ Vðhe let us knowÞðmÞÞ ‘I expect that he will let us know if he (at some point) comes to expect to be alone on Christmas Day.’
Here the evaluation time of the antecedent may fall anywhere between speech time and Christmas Day. Notice, by the way, that this sentence comes out true even in case the subject informs the speaker after Christmas. This may appear counterintuitive at first, but it should be kept in mind that the notification would then be ‘I expected to be alone on Christmas Day’, rather than ‘I was alone on Christmas Day’.
7.4 Non-predictive conditionals In the framework developed above, the distinction between predictive and non-predictive conditionals can now be characterized in terms of 24 Recall that just as nothing requires the modal bases of antecedent and consequent to be the same, nothing requires them to be different either. Here they are given as ;a and ;b, respectively, because this is the most natural interpretation. In principle, the consequent could have the speaker’s modal base as well, although it requires some effort to think of a scenario in which the conditional would then be true. ‘His’ letting ‘us’ know would still require his expecting in addition to the speaker’s. The superscripts on and indicate that these ordering sources reflect the (possibly different) expectations of the respective subjects.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Notice that the truth conditions imply that the index of evaluation j lies in the interval denoted by now (the consequent states that j is no later than m and m is in NOW). This seems to be correct for this example. This restriction is correctly predicted to disappear if ‘now’ is removed from the consequent, as in (70).
Stefan Kaufmann 273
temporal relations, as announced in the introduction. The semantic difference corresponds to that between the use of R and R, respectively: The hallmark of predictive readings is that the evaluation time of their constituents is allowed to range over both present and future indices. The definition is stated in semi-formal terms in Definition 10. Definition 10 (Non-predictive reading) A conditional is interpreted non-predictively if and only if all indices at which its antecedent is evaluated are accessible from the speech index by a modal accessibility relation.
7.5 Epistemic conditionals Finally, I return once again to the question of taxonomy. Throughout this paper I distinguished between predictive and non-predictive interpretations, exemplified by (1a,b) (repeated here as 71a,b) on their
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Recall that modal relations only access indices that are cotemporal with the speech index. Thus the definition captures the temporal difference that is at the bottom of the distinction. The definition does not directly make reference to the formal difference between R and R. I deliberately avoid such reference in order to leave open two options for explaining how non-predictive readings come about. Such an explanation may be either semantic or pragmatic. The semantic analysis would maintain that the difference lies literally in the presence or absence of the star on the accessibility relation. The only way to account for such variability within my analysis would be to say that IF is ambiguous, its two readings corresponding to the definitions in (45) and (37) above. The pragmatic alternative treats IF uniformly as star-inducing, attributing the restriction to speech time to the contextually given parameter e that we saw in the formulas in Section 6.2. Non-predictive readings arise, under this account, whenever e restricts the domain of quantification to cotemporal indices, i.e., ones that are accessible via a modal relation. Clearly the difference in interpretation is the same under both hypotheses. I see no argument for one choice or the other, except a theory-internal one: In the spirit of Grice’s ‘Modified Occam’s Razor’, ambiguity should not be postulated if the observed variation can be accounted for in other ways. I favor the pragmatic alternative with a uniform interpretation of IF as in (45).
274 Conditional Truth and Future Reference most prominent readings. There is a wide and at times confusing variety of alternative terminological proposals for this same distinction.25 Here I would like to comment on Kaufmann (2005), where (71a,b) are labelled predictive and epistemic, respectively. (71)
a.
If he submits his paper to a journal, we won’t include it in our book. b. If he submitted his paper to a journal, we won’t include it in our book.
25 These include open vs. closed (Funk 1985); conditional vs. hypothetical (Dudman 1989); hypothetical vs. interactional (Garrett 2001); doesn’t-will vs. didn’t-did (Bennett 1995; Edgington 1995); and sometimes forward-looking vs. backward-looking. 26 The latter case motivates Edgington’s (1995) observation that there is ‘no ideal, objective thing to think’ about a conditional, the falsehood of whose antecedent is settled.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The term ‘epistemic’ corresponds to ‘doxastic’ for present purposes. The only difference between epistemic and doxastic accessibility relations is that the former are required to be reflexive (since knowledge, unlike belief, is veridical), but this difference is irrelevant to the fact that both interpretations involve subjective beliefs. I already argued that the distinction should not be drawn between conditionals sentences, but between their readings. With this in mind, the question arises of how close a relationship there actually is between the temporal property of ‘non-predictiveness’ I defined above and epistemic interpretations. The first thing to note is that nothing in my proposal implies that (71a) cannot also be interpreted with respect to an epistemic state. Indeed, such an interpretation is the most prominent one in this case, largely due to the first-person pronoun in the consequent. Thus it is not the case that predictive conditionals are necessarily non-epistemic. It is the case, however, that non-predictive conditionals tend to receive epistemic interpretations (Gibbard 1981). This fact finds a natural explanation in the present framework. Suppose (71b), on its natural non-predictive reading, is applied at some index s to the metaphysical accessibility relation . Since the interpretation is non-predictive, all accessible indices are cotemporal with s. By historical necessity, the truth value of the antecedent is constant across all these indices. As a result, if the antecedent is true, the conditional is equivalent to its consequent, and if the antecedent is false, it is either vacuously true or its truth value is undefined (depending on what is assumed about the consequences of quantifying over an empty domain).26 Due to the inclusion of the CC in the truth conditions, this is predicted not only for the Past tense, but for the futurate Present as well.
Stefan Kaufmann 275
8 FUTURE DIRECTIONS I have proposed a compositional treatment of simple and conditional sentences which accounts for the observed range of available readings in terms of the interplay between modal and temporal semantic ingredients. One of the main claims of the paper is that the Present tense in the antecedents of indicative conditionals receives the same interpretation that it does in isolation. I rejected earlier claims that this Present is either semantically vacuous or derived from some other underlying form. Among the proponents of this latter view is Dancygier (1998), who calls the transformation ‘if-backshift’ (p. 39) and relates it to what she calls ‘hypothetical backshift’, the well-documented presence of an apparently extraneous layer of Past morphology in counterfactuals. The problem of 27 This is part of a larger pattern in the interaction between modal and temporal dimensions in the interpretation of modals. For instance, Condoravdi (2002) explains the observation that (ia,b) can only have epistemic interpretations, whereas (ic) is not so constrained, in terms of the same interaction between epistemic and metaphysical modal bases.
(i)
a. b. c.
He may be sick (now). He may have been sick (last week). He may get sick.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Thus the combination of a non-predictive reading with a metaphysical modal base necessarily yields an interpretation that is somewhat degenerate. Although there is no reason to ban such readings by semantic stipulations in the truth conditions, it is good pragmatic practice to reserve non-predictive interpretations for epistemic (or doxastic) modal bases, for there the analog of the above degeneracy does not arise unless the truth value of the antecedent is known.27 Based on this, we can say that particular conditional sentences are associated with epistemic interpretations if, or to the extent that, their prominent interpretation is non-predictive. This is the case with (71b), whose predictive interpretation involves a ‘past-in-the-future’ reading for the antecedent that requires some context to become salient. In contrast, the predictive interpretation is the more natural one for (71a), whose non-predictive interpretation involves a ‘settledness’ reading for the antecedent that is less prominent out of context. Thus the above argument goes some way towards explaining the prima facie plausibility of the term ‘epistemic’ as a label for non-predictive conditionals, even though the preference for epistemic readings is really a modal side effect of their temporal interpretation.
276 Conditional Truth and Future Reference
Acknowledgments I am grateful above all to Cleo Condoravdi, to two anonymous reviewers and the audiences at CLS 38, the 2002 Stanford Workshop on Mood and Modality, and the 2003 ESSLLI Workshop on Conditional and Unconditional Modality, for comments on earlier versions of this paper which led to great improvements. All remaining errors are my own. Part of this work was supported by the Japan Society for the Promotion of Science (JSPS, Grant #P01063).
STEFAN KAUFMANN Department of Linguistics Northwestern University 2016 Sheridan Road Evanston, IL 60208-4090 USA e-mail:
[email protected]
Received: 13.10.03 Final version received: 12.08.04 Advance Access publication: 03.05.05
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
the extra layer of Past in counterfactuals is an interesting one in the present context. While some authors have sought to incoporate its temporal meaning into the semantics of counterfactuals (Tedeschi 1981; Dahl 1997, and many philosophers), others, including Dancygier, take it to be devoid of temporal significance (see also Palmer 1974; Heim 1992; Iatridou 2000; Ippolito 2003). Dancygier suggests that both types of backshift involve a non-temporal reading of otherwise temporal expressions, specifying this reading along the lines of James (1982) and Fleischman (1989). I disagree with this treatment of indicatives, but have said nothing about counterfactuals. If my proposal has succeeded in accounting for ‘if-backshift’ while preserving the temporal interpretation of the Present, the question arises what it would take to extend it to counterfactuals. I leave this question for future research. Another open question concerns modalities other than those I dealt with in this paper. I mentioned in Footnote 6 that both doxastic and objective modals are subsumed under the linguistic category of ‘epistemic’ modality, as opposed to ‘root’ modality. There is a structural difference between these two classes that is as-yet not well understood, but it appears that root modals tend to be more deeply buried in the derivation tree (hence the label can be a bit misleading). A hypothesis I consider worth exploring is that root modals are generally embedded in the nuclear scope of an outer epistemic modal of the kind I have discussed here. I leave this question, along with the task of exploring its implications for the analysis of conditionals, for future work.
Stefan Kaufmann 277
REFERENCES out from the 14th Amsterdam Colloquium. Crouch, R. (1993) The Temporal Properties of English Conditionals and Modals. Ph.D. thesis, University of Cambridge. Dahl, O. (1997) ‘The relation between past time reference and counterfactuality: A new look’. In Athanasiadou and Dirven 97–112. Dancygier, B. (1998) Conditionals and Prediction. Cambridge University Press. Declerck, R. & Reed, S. (2001) Conditionals: A Comprehensive Empirical Analysis, volume 37 of Topics in English Linguistics. Mouton de Gruyter. Dowty, D. (1977) ‘Towards a semantic analysis of verb aspect and the English ‘‘imperfective’’ progressive’. Linguistics and Philosophy 1:45–77. Dowty, D. (1979) Word Meaning and Montague Grammar, volume 7 of Studies in Linguistics and Philosophy. Reidel. Dowty, D. (1991) ‘Thematic proto-roles and argument selection’. Language 67:547–619. Dudman, V. H. (1984) ‘Parsing ‘If ’sentences’. Analysis 44:145–153. Dudman, V. H. (1989) ‘Vive la Re´volution!’ Mind 98:591–603. Edgington, D. (1995) ‘On conditionals’. Mind 104:235–329. Edgington, D. (1997) ‘Commentary’. In Woods Clarendon Press, 97–137. Fagin, R., Halpern, J. Y., Moses, Y. & Vardi, M. Y. (1995) Reasoning about Knowledge. MIT Press. Fernando, T. (2003) Reichenbach‘s E, R, and S in a finite-state setting. http://www.cs.tcd.ie/Tim.Fernando/
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Abusch, D. (1997) ‘Sequence of tense and temporal de re’. Linguistics and Philosophy 20:1–50. Abusch, D. (1998) ‘Generalizing tense semantics for future contexts.’ In S. Rothstein, (ed.), Events and Grammar, volume 70 of Studies in Linguistics and Philosophy. Kluwer, 13–33. Athanasiadou, A. & R. Dirven (eds). (1997) On Conditionals Again, volume 143 of Amsterdam Studies in the Theory and History of Linguistic Science. John Benjamins. Bennett, J. (1995) ‘Classifying conditionals: The traditional way is right’. Mind 104:332–354. Bennett, M. & Partee, B. (1978) ‘Toward the logic of tense and aspect in English’. Reproduced by the Indiana University Linguistics Club. Brennan, V. (1993) Root and Epistemic Modal Auxiliary Verbs. Ph.D. thesis, University of Massachusetts, Amherst. Burgess, J. P. (1979). ‘Logic and time’. The Journal of Symbolic Logic 44:566– 582. Close, R. (1980) ‘Will in if-clauses’. In S. Greenbaum, G. Leech, & J. Svartvik, (eds), Studies in English Linguistics for Randolph Quirk. Longman, 100–109. Comrie, B. (1982) ‘Future time reference in the conditional protasis’. Australian Journal of Linguistics, 143– 152. Comrie, B. (1985) Tense. Cambridge University Press. Condoravdi, C. (2002) ‘Temporal interpretation of modals: Modals for the present and modals for the past’. In D. I. Beaver, L. Casillas, B. Clark, & S. Kaufmann (eds), The Construction of Meaning. CSLI Publications, 59–88. Condoravdi, C. (2003) ‘Moods and modalities for will and would’. Hand-
278 Conditional Truth and Future Reference Harper, W. L., Stalnaker, R. & Pearce, G. (eds) (1981) Ifs: Conditionals, Belief, Decision, Chance, and Time. Reidel. Heim, I. (1992) ‘Presupposition projection and the semantics of attitude verbs’. Journal of Semantics 9:183–221. Hofmann, T. R. (1976) ‘Past tense replacement and the modal system’. In J. D. McCawley, (ed), Notes from the Linguistic Underground, volume 7 of Syntax and Semantics, Academic Press, 85–100. Hornstein, N. (1990) As Time Goes By: Tense and Universal Grammar. MIT Press. Huddleston, R. & Pullum, G. K. (2002) The Cambridge Grammar of the English Language. Cambridge University Press. Iatridou, S. (2000) ‘The grammatical ingredients of counterfactuality’. Linguistic Inquiry 31:231–270. Ippolito, M. (2003) ‘Presuppositions and implicatures in couterfactuals’. Journal of Semantics 11:145–186. Jackendoff, R. (1972) Semantic Interpretation in Generative Grammar. MIT Press. James, D. (1982) ‘Past tense and the hypothetical: A cross-linguistic study’. Studies in Language 6:375–403. Joos, M. (1964) The English Verb. University of Wisconsin Press. Kamp, H. & Reyle, U. (1993) From Discourse to Logic. Kluwer. Kaufmann, S. (2002) ‘The presumption of settledness’. In M. Andronis, E. Debenport, A. Pycha, & K. Yoshimura (eds), Proceedings of CLS 38, 313–328. Kaufmann, S. (2005) ‘Conditional predictions: A probabilistic account’. Linguistics and Philosophy 28:181–231. Kratzer, A. (1979) ‘Conditional necessity and possibility’. In U. Egli, R. Bau¨erle, & A. von Stechow (eds)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
wien.pdf. (Last checked: January, 2004.) von Fintel, K. (1994) Restrictions on Quantifier Domains. Ph.D. thesis, University of Massachusetts. Fleischman, S. (1989) ‘Temporal distance: A basic linguistic metaphor’. Studies in Language 13:1–50. van Fraassen, B. C. (1980) ‘Review of Brian Ellis, ‘‘Rational Belief Systems’’’. Canadian Journal of Philosophy, 10:457–511. van Fraassen, B. C. (1981) ‘A temporal framework for conditionals and chance. In Harper et al. 323–340. Frank, A. (1996) Context Dependence in Modal Constructions. Ph.D. thesis, Institut fu¨r maschinelle Sprachverarbeitung (Stuttgart). Funk, W.-P. (1985) ‘On a semantic typology of conditional sentences’. Folia Linguistica, 19:365–414. Gabbay, D. & Guenthner, F. (eds). (1984) Extensions of Classical Logic, volume 2 of Handbook of Philosophical Logic. D. Reidel. Garrett, E. J. (2001) Evidentiality and Assertion in Tibetan. Ph.D. thesis, University of California. Gennari, S. (2003) ‘Tense meanings and temporal interpretation’. Journal of Semantics, 20:35–71. Gibbard, A. (1981) ‘Two recent theories of conditionals’. In Harper et al. 211– 247. Goodman, F. (1973) ‘On the semantics of futurate sentences’. In Mostly Syntax and Semantics, volume 16 of Working Papers in Linguistics, Department of Linguistics, The Ohio State University, 76–89. Halpern, J. Y. (1997) ‘Defining relative likelihood in partially-ordered preferential structures’. Journal of Artificial Intelligence Research, 7:1–27. Halpern, J. Y. (2003) Reasoning about Uncertainty. MIT Press.
Stefan Kaufmann 279 Smith, C. S. (1991) The Parameter of Aspect, volume 43 of Studies in Linguistics and Philosophy. Kluwer. Stalnaker, R. (1968) ‘A theory of conditionals’. In Studies in Logical Theory, American Philosophical Quarterly, Monograph: 2. Blackwell, 98–112. Stalnaker, R. (1975) ‘Indicative conditionals’. Philosophia, 5:269–286. Stalnaker, R. (2002) ‘Common ground’. Linguistics and Philosophy, 25:701– 721. Steedman, M. (2002) ‘The productions of time’. Unpublished MS, Version 4.1. Stone, M. (1994) ‘The reference argument of epistemic Must.’ In Proceedings of IWCS 1, 181–190. Tedeschi, P. (1981) ‘Some evidence for a branching-futures semantic model’. In P. Tedeschi & A. Zaenen (eds), Tense and Aspect, volume 14 of Syntax and Semantics. Academic Press, 239– 270. Thomason, R. H. (1970) ‘Indeterminist time and truth value gaps’. Theoria 36:264–281. Thomason, R. H. (1984) ‘Combinations of tense and modality’. In Gabbay and Guenthner 135–165. Thomason, R. H. & Gupta. A. (1981) ‘A theory of conditionals in the context of branching time’. In Harper et al. 299–322. Traugott, E. C., ter Meulen, A. Snitzer Reilly, J. & Ferguson, C. A. (eds), (1986) On Conditionals. Cambridge University Press. Veltman, F. (1985) Logics for Conditionals. Ph.D. thesis, University of Amsterdam. Veltman, F. (1986) ‘Data semantics and the pragmatics of indicative conditionals’. In Traugott et al. 147–168. Vetter, D. (1973) ‘Someone solves this problem tomorrow’. Linguistic Inquiry 4:104–108.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Semantics from Different Points of View. Springer, 117–147. Kratzer, A. (1981) ‘The notional category of modality’. In J. Eikmeyer, & H. Rieser (eds), Words, Worlds, and Contexts. Walter de Gruyter, 38–74. Kratzer, A. (1991) ‘Modality’. In A. von Stechow, & D. Wunderlich (eds) Semantik: Ein internationales Handbuch der zeitgeno¨ssischen Forschung. [¼Semantics]. de Gruyter, 639–650. Lakoff, G. (1971) ‘Presupposition and relative well-formedness’. In D. Steinberg, & L. Jakobovits (eds), Semantics. Cambridge University Press, 329–340. Leech, G. (1971) Meaning and the English Verb. Longman. Lewis, D. (1986) ‘Postscript to ‘‘Probabilities of conditionals and conditional probabilities’’ ’. In Philosophical Papers, volume 2. Oxford University Press, 152–156. McCawley, J. D. (1971) ‘Tense and time reference in English’. In C. Fillmore & D. Langedoen (eds), Studies in Linguistic Semantics. Holt. Mellor, D. H. (ed.) (1990) Philosophical Papers: F. P. Ramsey. Cambridge University Press. Palmer, F. R. (1974) The English Verb. Longman. Palmer, F. R. (1979) Modality and the English Modals. Longman. Portner, P. (2003) ‘The (temporal) semantics and (modal) pragmatics of the perfect’. Linguistics and Philosophy, 26:459–510. Prior, A. (1967) Past, Present and Future. Oxford University Press. Quirk, R., Greenbaum, S. Leech, G. & Svartvik, J.(1985) A Comprehensive Grammar of the English Language. Longman. Ramsey, F. P. (1929) ‘General propositions and causality’. Printed in Mellor (1990) 145–163.
280 Conditional Truth and Future Reference Wekker, H. C. (1976) The Expression of Future Time in Contemporary British English. Number 28 in North-Holland Linguistic Series. North-Holland. Westmoreland, R. R. (1995) ‘Epistemic must as evidential’. In P. Dekker & M. Stokhof (eds), Proceedings of the 10th Amsterdam Colloquium 3:683–702. Amsterdam. ILLC.
Woods, M. (1997) Conditionals. Clarendon Press. Zandvoort, R. W. (1965) A Handbook of English Grammar. Longmans. Zvolenszky, Z. (2002) ‘Is a possibleworlds semantics of modality possible? A problem for Kratzer’s semantics’. In B. Jackson (ed.), Proceedings of SALT XII.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Journal of Semantics 22: 281–305 doi:10.1093/jos/ffh026 Advance Access publication May 3, 2005
A Modal Analysis of Presupposition and Modal Subordination ROBERT VAN ROOIJ Institute for Logic, Language and Computation, University of Amsterdam
Abstract
1 INTRODUCTION Consider the following examples: (1) a. A thief might break into the house. He might take the silver. b. It is possible that John used to smoke and possible that he just stopped doing so. c. It is possible that Mary will come and it is possible that Sue will come too. The sequence (1a) was discussed by Roberts (1989) as a problematic example for standard discourse representation theory and dynamic semantics (Kamp 1981; Heim 1982; Groenendijk & Stokhof 1991). The sequence (1b) was already given by Gazdar (1979) as a counterexample to the satisfaction theory of presupposition defended in the seventies by, among others, Karttunen and Stalnaker. This satisfaction theory predicted that if A presupposes P, the sentence It is possible that A presupposes P as well. The sequence (1b) was taken to be a counterexample to this theory, because it is typically used in a situation in which it is not presupposed that John used to smoke. The closely related (1c) is an even more serious problem for proponents of the satisfaction theory who claim that a trigger like too does not entail what it presupposes, in this case The Author 2005. Published by Oxford University Press. All rights reserved.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In this paper I will give a modal two-dimensional analysis of presupposition and modal subordination. I will think of presupposition as a non-veridical propositional attitude. This allows me to evaluate what is presupposed and what is asserted at different dimensions without getting into the binding problem. What is presupposed will be represented by an accessibility relation between possible worlds. The major part of the paper consists of a proposal to account for the dependence of the interpretation of modal expressions, i.e. modal subordination, in terms of an accessibility relation as well. Moreover, I show how such an analysis can be extended from the propositional to the predicate logical level.
282 A Modal Analysis of Presupposition and Modal Subordination
2 PRESUPPOSITION
2.1 The representation of presuppositions The notion of presupposition plays a crucial role in dynamic semantics. A context is taken to represent what is presupposed. Stalnaker (1974, 1998, 2002) has always argued that presupposition should be thought of as a propositional attitude and thus represented in a similar way: by means of an accessibility relation. But what do agents presuppose? The standard answer is: what is common ground between the participants of the conversation. According to discourse representation theory, what is common ground is what is explicitly represented in a discourse representation structure, a DRS. This DRS, in turn, represents what has been explicitly agreed upon by the conversational participants. This suggests that presupposition should by default be fully introspective: what is presupposed is also presupposed to be presupposed, and what is not
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
that (it is possible that) Mary will come. The reason is that according to such a two-dimensional analysis of presuppositions it is predicted that (1c) can be true and appropriate without there being a possible world in which both Mary and Sue are coming, i.e. the binding problem. In this paper I will formulate a two-dimensional theory of presupposition satisfaction in which the binding problem does not arise. I will do this by taking seriously the proposal of Stalnaker that presupposition should be thought of as a propositional attitude and represented by an accessibility relation between possible worlds. In the most substantial part of the paper, I will show how such a modal analysis can help to account for the phenomenon that the appropriate interpretation of one embedded sentence can depend on that of another: a phenomenon normally discussed under the heading of ‘modal subordination’. This modal analysis of modal subordination will be rather different from the more representational analyses proposed by, among others, Roberts (1989) and Geurts (1995). This paper will be organized as follows. First, I will briefly motivate and formalize a two-dimensional analysis of presupposition satisfaction. In section 3, I will discuss the phenomenon of modal subordination and propose a modal analysis in terms of a changing accessibility relation. This analysis will be developed further in the remaining sections to account for disjunctions, conditionals, belief and desire attributions, and the subjunctive mood. Until then I limit myself to the propositional level. In the final substantial section of this paper I will briefly indicate how the analysis can be extended such that anaphoric dependencies across modals can be taken care of as well. I will end with some conclusions.
Robert van Rooij 283
2.2 Formalization When a speaker presupposes something, he presupposes it in a world or a possibility. A possibility will be represented by a pointed model, ÆR, wæ,4 where w is a distinguished world representing the actual world and should be thought of as a valuation function from atomic propositions to truth values, and where R is the presuppositional accessibility relation that is (by default) serial, transitive and Euclidean.5 I will take R(v) to be the worlds accessible from v: fu 2 W : vRug. As a result, it will be the case that what is presupposed is introspective: "v, w: if v 2 R(w), then R(v) ¼ R(w), although it need not be veridical, i.e., it might be that w ; R(w). To determine in possibility ÆR, wæ whether P is presupposed, we have to check what is presupposed in this possibility, R(w). The two-dimensional (or four-valued) analysis of presupposition that was popular in the seventies treats the logic of truth and that of presupposition at separate dimensions. 1
See Fernando (1995) for an analysis of context where full introspection is assumed as well. Stalnaker (2002) suggests what is presupposed by an agent is what she believes is commonly believed by the discourse participants. This has as a result, however, that the attitude of presupposition does not obey negative introspection, because more things can be taken to be commonly believed than what is explicitly agreed upon. 3 Updating through the elimination of arrows instead of worlds has been used, among others, by Landman (1986a) and Veltman (1996). Its limitations for multi-agent settings are discussed in Gerbrandy (1999). 4 If we think of a world as representing everything that is the case, including some modal facts, a pointed model should be thought of as such a world. 5 A relation R is serial if "x: dy: xRy; transitive if "x; y; z : ðxRy & yRzÞ/xRz; and Euclidean if "x; y; z : ðxRy & xRzÞ/yRz: 2
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
presupposed is also presupposed not to be presupposed.1 I will represent what is presupposed by a primitive accessibility relation R, i.e., one that cannot be reduced to what the participants of the conversation know or believe.2 Although the presuppositional accessibility relation should be fully introspective, the relation should not be based on the assumption that what is presupposed also has to be true: discourse can be based on an assumption that later turns out to be false. So, presupposition should be represented by an accessibility relation that need not be reflexive. The non-veridicality of what is presupposed suggests that we should treat the valuation of truth separately from context change—distinguish content from force. In this section I will show how we can systematically account for presupposition satisfaction without giving up the possibility of determining the content of a sentence separately from the way it changes the context. For context change, I will rely mainly on work in dynamic epistemic semantics, where updates are defined in terms of eliminating arrows instead of eliminating worlds.3
284 A Modal Analysis of Presupposition and Modal Subordination
½½AR;w ¼ Æ1=0; +æ;
6
iff
wðAÞ ¼ 1=0; if A is atomic ðthen always definedÞ
A presuppositional analysis can be ‘two-dimensional’ in two different ways. One is through the assumption that the presupposition and the assertion of the sentence can and must be determined relatively independently of each other. The analysis of what Karttunen & Peters (1979) call ‘conventional implicatures’ and the presuppositional analyses of Herzberger (1973), Gazdar (1979), Soames (1979), and Van der Sandt (1988) are all two-dimensional in this sense. But there is also a less radical way in which an analysis of presuppositions might be called ‘twodimensional’. On this reading it means only that at least some sentences in which a presupposition trigger occurs can be true even though the triggered presupposition is actually false. I believe that Stalnaker’s analysis of presuppositions, just as my own analysis, should be thought of as twodimensional in this more liberal sense. 7 Although I use a four-dimensional logic, I am not explicit about when a sentence is true or false, although its presupposition is not satisfied. But this is needed if we want to allow Even John was there to be true although it is not presupposed that John’s being there was unlikely (thanks to Kai von Fintel for reminding me of this). However, there is no problem in principle of distinguishing those cases as well.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This is appealing because sometimes a sentence can—at least according to Karttunen & Peters (1979), for instance—be true although its presupposition is false. They claim that a sentence like Even John was there, for example, can be true without it actually being unlikely that John was there.6 Standard dynamic semantics treats conjunction in an asymmetric way: the second conjunct should be interpreted with respect to the initial context updated with the first conjunct. This is a desirable feature of a framework to account for the asymmetric behavior of presuppositions in conjunctive sentences. In this section I will combine the desirable features of both the two-dimensional and the dynamic analysis of presuppositions. Thinking of presupposition as a non-veridical propositional attitude, we can account for the dynamic aspects of presupposition satisfaction without giving up the idea behind a twodimensional analysis. That is, we will predict that conjunction behaves asymmetrically with respect to presupposition satisfaction because what is presupposed when the second conjunct is interpreted is predicted to be the result of updating what is presupposed when the first conjunct is interpreted with the first conjunct itself. Still, ‘and’ will be treated truthconditionally in a symmetric way. The reason is that truth and presupposition satisfaction are defined separately from the update function (although they will be defined simultaneously). Making use of Beaver’s (1995) presupposition operator, I will represent an atomic sentence A that presupposes P as follows: @P ^ A. For the time being, I will concentrate only on the truth-conditional connectives. I will assume that a sentence has two values: (i) a sentence is true or false, i.e. 1 or 0; (ii) a sentence has no presupposition failure or it has one, i.e. + or . The combined truth and presupposition satisfaction conditions of sentences are given below (where ‘’ is a placeholder):7
Robert van Rooij 285
½½:AR;w ¼ Æ1=0; +æ iff
½½AR;w ¼ Æ0=1; +æ;
½½A ^ BR;w ¼ Æ1; +æ iff
½½@AR;w ¼ Æ1; +æ iff " v 2 RðwÞ : ½½AR;v ¼ Æ1; +æ ¼ Æ; æ otherwise
Æ; æ otherwise
½½AR;w ¼ Æ1; +æ and ½½BUpdðA;RÞ;w ¼ Æ1; +æ ¼ Æ; æ iff ½½AR;w ¼ Æ; æ or ½½BUpdðA;RÞ;w ¼ Æ; æ ¼ Æ; +æ otherwise
UpdðA; RÞ ¼ fÆu; væ 2 R j ½½AR;v ¼ Æ1; +æg:
Notice that this update function is eliminative, but instead of eliminating worlds in R(w) it eliminate tuples, or arrows, in R. It eliminates all arrows in R that point to a non-A-world. This has the effect that after the update of R with A, not only all worlds v accessible from w verify A, but all worlds u accessible from v make A true as well. Thus, after the update with A it is not only presupposed that A, but it is also presupposed to be presupposed that A. Moreover, on the assumption that R is fully introspective, Upd(A, R) will be fully introspective as well. Also after the update, everything that is not presupposed is also presupposed to be not presupposed. Our analysis is very similar to standard dynamic semantics. If we were to say that ½½)AR;w ¼ Æ1; æ iff dv 2 RðwÞ : ½½AR;v ¼ Æ1; æ and assume that possibility statements don’t have any dynamic effect,8 8
Though we will give a somewhat different analysis of possibility statements later.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Observe again that the presupposition value of a conjunction is determined in a symmetric way. That is, if either A or B is false or has a presupposition failure, the conjunction A ^ B will be false or have a presupposition failure as well. However, to determine the presupposition value of a conjunction of the form A ^ B in possibility ÆR, wæ, we look at the presupposition value of B in possibility ÆUpd(A, R), wæ— the update function is relevant here. This is the point at which we take over the insights of dynamic semantics. The update Upd(A, R) is defined as follows:
286 A Modal Analysis of Presupposition and Modal Subordination
9
Soames (1989) observed already that this is problematic for the standard way of accounting for presupposition satisfaction in dynamic semantics. 10 Throughout the paper I will assume the same for an aspectual verb like stop.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
we predict just like Veltman (1996) an asymmetry between )A ^ :A and :A ^ )A; the former is okay, the latter is not. However, this contrast in acceptability is explained in a somewhat different way: Veltman’s explanation appeals to acceptability of update, while we explain the contrast in terms of truth. We predict that the former sequence can be true, but the latter cannot. If we assume that sentence A presupposes P iff "ÆR, wæ : if [[A]]R,w ¼ Æ, +æ, then "v 2 R(w) : [[P]]R,v ¼ Æ1, +æ, the above implementation gives rise to the same presuppositional predictions as the standard implementation of the satisfaction account. In particular, on the assumption that John stopped smoking gives rise to the presupposition that John used to smoke, this implementation predicts that sentences like John didn’t stop smoking and John stopped smoking and Mary is sick will also gives rise to this presupposition, but John used to smoke and he stopped doing so will never give rise to presupposition failure. Although the predictions of the above implementation of the satisfaction approach are similar to the predictions of the standard approach, there are still some important differences. First, by treating presupposition as a propositional attitude, we can evaluate in a distributive way whether a presupposition associated with a sentence is satisfied by what the speaker presupposes. This is possible, of course, because we have represented in a single possibility all the information that is normally represented only in a whole context/information state. Second, and related, we can now account for the dominant view in the seventies that presupposition satisfaction and truth should be evaluated at different dimensions. According to Karttunen & Peters (1979) and others, a sentence like Even Bill likes Mary presupposes something that it does not entail. Thus, the sentence can be true without it actually being unlikely that Bill likes Mary, because what is presupposed need not be true.9 Notice that we can now account for this intuition without assuming with Karttunen & Peters (1979) that we should thus represent presuppositions separately from assertions. On the other hand, we can also account for the intuition that a factive verb both presupposes and entails that its complement is true.10 To analyse Sam realizes that P we add the following construction to the language: if P is a sentence, Real(s, P) is a sentence too. To interpret the formula, we add a reflexive
Robert van Rooij 287
accessibility relation to the model, Ks, modeling what Sam realizes.11 The formula is then interpreted as follows: ½½Realðs; PÞR;w ¼ Æ1; +æ iff "v 2 Ks ðwÞ : ½½PR;v ¼ Æ1; +æ ¼ Æ0; +æ iff dv 2 Ks ðwÞ : ½½PR;v ¼ Æ0; +æ; ¼ Æ; æ otherwise
3 MODAL SUBORDINATION
3.1 Previous analyses According to standard dynamic semantics (Veltman 1996), the embedded sentence of ‘possibly A’ should be interpreted with respect to the same context as the whole sentence. This gives rise to the prediction that ‘possibly A’ triggers the same presupposition as A itself. However, if it has already been established that it is possible that John used to smoke, i.e. after (2a) has been asserted, (2b) need not presuppose that John used to smoke. (2) a. It is possible that John used to smoke, b. and it is possible that he just stopped doing so. The phenomenon that a modal expression depends for its interpretation on another modal, as illustrated by (1a) and (2a–b), is known as ‘modal subordination’. Consider Roberts’ (1a) from the introduction again: (1a) A thief might break into the house. He might take the silver. In both (1a) and (2a–b), we get the intuitively correct reading if we assume that the modal in the first clause takes scope over the whole sequence. Aside from issues of compositionality, however, Roberts 11
Our simple update function has limitations here: if we attributed to Sam attitudes about what the discourse participants presuppose, things would go wrong. I will ignore such attributions in this paper. See, among others, Gerbrandy (1999) for an analysis in which this problem is overcome.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Notice that because Ks is reflexive, according to this analysis the formula entails, but does not presuppose, that P. To account for the presupposition, we represent the sentence Sam realizes that P by the formula @P ^ Real(s, P), which both presupposes and entails that P. If we now represent Sam does not realize that P by :(@P ^ Real(s, P)), this sentence presupposes that P, but can still be true in case P is false (in case w ; R(w)).
288 A Modal Analysis of Presupposition and Modal Subordination (1989) showed already that such an analysis would be on the wrong track. It would not be able to account for a slightly different sequence like (3a), where we have a necessity instead of a possibility operator in the second sentence. (3) a. A thief might break into the house. He would take the silver. b. If a thief broke into the house, he would take the silver.
12 Roberts’ (1989) constraints are the following: (i) modal subordination ‘requires non-factual mood’ (p. 701); (ii) ‘it must be plausible that the modally subordinate utterance has a hypothetical common ground suggested by the immediately preceding context’ (p. 701); and (iii) modal subordination may not make antecedents available to anaphoric expressions that have no explicit representation in the given DRS (p. 705). For a critical discussion of these constraints, see Kibble (1994) and Geurts (1995). 13 Geurts (1995) claims that a modal presupposes its domain and assumes an exclusively anaphoric account of presupposition (satisfaction) for these cases. Given the important role that (non-global) accommodation plays in Geurts’ (1995) analysis of presuppositions, this restriction is somewhat surprising.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Intuitively, the second sentence of (3a) means something like (3b). To account for this, Roberts takes up Kratzer’s (1981) idea that the domain of a modal is context dependent, and extends it by proposing that the actual selection proceedss via ‘accommodation’ of the material that has been mentioned explicitly in an earlier modal statement. For (1), (2a–b), and (3a), for instance, this means that the embedded clauses of their first sentences will be accommodated to function as the antecedents of the modals might, possible, and would, respectively. In this way she makes correct predictions for all of (1a), (2b) and (3a). Although Roberts’ analysis reflects what intuitively goes on in the examples illustrated above, the exact mechanism that she uses has been rightly criticized by Kibble (1994), Geurts (1995), and others. Not only is her use of accommodation rather ad hoc and non-compositional, it also seems to be much too powerful a device, even with the constraints on accommodation that she proposes.12 Kibble (1994) and Geurts (1995) propose that instead of selecting the domain of a modal by means of (antecedent) accommodation, we should assume that the domain is picked up anaphorically.13 Moreover, they suggest that modal statements also make such domains, or subordinated contexts, anaphorically available for later modals: they introduce propositional discourse variables to the discourse and these are mapped by assignment functions to the set of possibilities that verify their embedded clauses. Thus, they can account for (3a) by making sure that the first sentence introduces a propositional discourse referent that is assigned to a set of possibilities that verify the embedded clause. The
Robert van Rooij 289
(4) a. John might be at home reading a bookx b. Actually, he’s still at the office. c. Itx’ll be War and Peace. The reason is that in Geurts’ account of modal subordination, nonmodal sentences that talk about the actual world, such as (4b), are not taken to eliminate any of the elements in the set of world-assignment pairs introduced by the embedded sentence of (4a). Thus, Geurts’ analysis of modal subordination does not allow for enough interaction between the main context and the subordinated ones.15 Kibble (1994) allows for such interaction, but his analysis cannot account for interpretational dependence of clauses in counterfactual or subjunctive mood. This is so because in order to interpret a clause (with free variables) embedded under a modal operator, each possibility in the introduced subordinated context has to be compatible in his formalism with the ‘main’ context. Geurts’ (1995) analysis doesn’t have this limitation, because each possibility of the ‘main’ context carries the information contained in the subordinated contexts. However, by making use of standard set theory, Geurts does not allow for the situation that if Æw, f æ is a possibility that satisfies the main context (DRS) and also assigns a set of possibilities to newly introduced propositional discourse marker p, that there is a world v such that Æv, f æ 2 f(p). A somewhat ad hoc analysis is given to assure that this won’t 14
Kaufmann (1997) discusses a similar example involving a conditional: (i) If John bought a book, he’ll be home reading it by now. (ii) John works at a gas station. (iii) It’ll be a murder mystery. 15 As shown in Frank (1997), however, this problem can be overcome.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
second sentence of (3a) is then analysed as true in case all those possibilities verify the embedded clause of this latter sentence. Although these anaphoric analyses of modal subordination are more constrained and appealing than Roberts’, they are not unproblematic. For one thing, they are still too unrestricted: modal statements just introduce and are allowed to pick up propositional discourse markers. This makes it possible that a clause embedded under one kind of modality can figure as the antecedent of a modal expression of a completely different kind. A second problem is that Geurts’ analysis doesn’t allow for enough interaction between main and subordinated context, while Kibble demands too strong such an interaction. Consider first an example (discussed in Kibble, 1994) that is problematic for Geurts (1995). He falsely predict that (4c) is an appropriate continuation of (4a–b):14
290 A Modal Analysis of Presupposition and Modal Subordination
3.2 A modal analysis The basic idea is technically very simple: possibility statements introduce an ordering on the worlds. Because we assume that what is presupposed is a propositional attitude and should be represented by an 16 See, however, Fernando (1996) and Frank (1997) for less straightforward implementations of this idea within standard set theory.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
happen. In particular, a somewhat arbitrary distinction is made between embedded and unembedded information: although a propositional discourse marker can be mapped to what is presupposed to be possible, there is no propositional discourse marker that captures what is presupposed in the discourse as a whole. Not only do I believe that this is undesirable for conceptual reasons, it also has an unfortunate empirical consequence. It is predicted that modal statements cannot anaphorically pick up what is presupposed in the entire discourse. Because a modal can, intuitively, use this kind of information as its domain of quantification, a somewhat artificial distinction has to be made between anaphoric and non-anaphoric dependent modals. One way to overcome this problem is to introduce a distinguished propositional discourse marker that represents with respect to each world-assignment pair what is presupposed in that possibility. A straightforward implementation of this idea, however, requires the use of non-wellfounded set theory.16 In this paper, instead, I will propose a more traditional way to account for presuppositions and modal expressions: in terms of accessibility relations. In contrast to Roberts’ analysis of modal subordination in terms of accommodation, the anaphoric analyses described above keep the subordinated contexts made use of in the interpretation of previous sentences ‘in memory’ by adding propositional discourse markers to the discourse. Another way to store previously used subordinated contexts was proposed by Kaufmann (1997, 2000). Instead of representing a context just by a set of possibilities that verify everything established until now, he represents it by a stack of such sets (see Zeevat (1992) as well), where a set ‘below’ the top-level represents a subordinated context. I have no principled objection to such an analysis. Still, I would like to see a less ‘representational’ approach towards modal subordination where what is presupposed at a particular moment in time can simply be represented by a single set of possibilities. But now the challenge is how to account for the introduction of subordinated contexts without giving up the assumptions that we represent what is presupposed at a particular moment in time in terms of a single accessibility relation.
Robert van Rooij 291
accessibility relation, we can implement this idea in an appealing way. Following Veltman’s (1996) analysis of normally, I will assume that the dynamic effect of a possibility statement is to eliminate arrows from worlds that make the embedded clause true to worlds that don’t, thus ensuring that the former are ‘preferred’ over the latter.17 Updð)A; RÞ ¼ fÆu; væ 2 R j if ½½AR;u ¼ Æ1; +æ; then ½½AR;v ¼ Æ1; +æg
½½)AR;w ¼ Æ1; +æ iff ¼ Æ0; +æ iff
dv 2 RðwÞ : ½½AR;v ¼ Æ1; +æ dv 2 RðwÞ : ½½AR;v ¼ Æ; +æ and " v 2 RðwÞ : if ½½AR;v ¼ Æ; +æ; then ½½AR;v ¼ Æ0; +æ; ¼ Æ; æ otherwise
According to this rule it holds that if A presupposes P, )A can be used appropriately only if it is assumed to be possible that P is presupposed. Because out of context (or so we assumed) it holds that " v 2 R(w) : R(v) ¼ R(w), under normal circumstances )A presupposes the same as A itself. However, the rule can also account for the 17
That the proposed analysis is technically similar to Veltman’s (1996) analysis of normally doesn’t mean that the ordering relation represents the same kind of information. 18 This update rule is defined on the assumption that either w ; R(w) or w is not an A-world, because otherwise we would falsely predict that after the use of the possibility statement only other A-worlds would be accessible from w. In general we cannot make this assumption, of course. Fortunately, there are several ways to solve this problem. One way is to assume that w is the distinguished actual world, and that we change the update rule for possibility statements as follows: Upd()A, R) ¼ fÆu, væ 2 Rj (if [[A]]R,u ¼ Æ1, +æ, then [[A]]R,v ¼ Æ1, +æ) or u ¼ wg. In this paper I leave the update rule unchanged, and adopt a more tricky solution. Assume that if w 2 R(w) and w is an A-world, we don’t go to the new pointed model ÆR#, wæ, but rather to the pointed model ÆR#, wæ with a new world w. This new world is exactly like w of the original pointed model, except that w;R(w) although w 2 R(w). Because our technical problem has a simple solution, I will ignore this complication in the main text. (Thanks to Frank Veltman and Henk Zeevat for discussion on this point.)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
According to the update function, possibility statements disconnect Aworlds from :A-worlds, although A-worlds can still be seen from :Aworlds and from the actual world w. Suppose that before the update R ¼ fÆw, væ, Æw, uæ, Æv, væ, Æv, uæ, Æu, uæ, Æu, væg, where v is an A-world and w and u are :A-worlds. Then R is introspective: R(w) ¼ R(v) ¼ R(u) ¼ fv, ug. After the update with )A, however, the new accessibility relation Upd()A, R) won’t be introspective anymore: the tuple Æv, uæ will be eliminated, which means that Upd()A, R)(w) 6¼ Upd()A, R)(v) ¼ fvg 6¼ Upd()A, R)(u).18 Thus, if R was Euclidean before the update with )A, it won’t be Euclidean anymore afterwards. Possibility statements will be interpreted as follows:
292 A Modal Analysis of Presupposition and Modal Subordination
(1c) It is possible that Mary will come and it is possible the Sue will come too Our analysis does not give rise to this false prediction. The reason is that, although we have defined update separately from truth, the definitions of truth and appropriateness are closely related: for a possibility statement to be true and appropriate, there has to be an accessible world in which the embedded sentence is both true and appropriate. Thus, the logics of truth and appropriateness (or presupposition satisfaction) are not as independent of each other as proposed in Herzberger (1973) and Karttunen & Peters (1979). Notice that if we take hA to be an abbreviation of :):A, we predict that A has to be interpreted only in possibilities that satisfy the presupposition of A: hA has value Æ1, +æ in ÆR, wæ iff dv 2 R(w) : [[A]]R,v ¼ Æ, +æ and "v 2 R(w) : if [[A]]R,v ¼ Æ, +æ, then [[A]]R,v ¼ Æ1, +æ. But this means that h(@P ^ A) can only be true in ÆR, wæ iff either P itself is presupposed and A is true in all accessible worlds, or it is presupposed that P is possible and A is true in all accessible P-worlds.19 19 Notice that this analysis predicts that if A doesn’t give rise to a presupposition, the sentence Would A can only be true if A is true in all accessible worlds. But the following sequence suggests that this is a wrong prediction: It might be rainy next week. The strawberries would rot in the garden. Here we seem to have modal subordination even though the embedded clause of the second sentence doesn’t trigger a presupposition. One way in which this problem might be solved is to assume that the embedded clauses of would-sentences should be interpreted only with respect to accessible worlds v such that "u 2 R(v) : R(u) ¼ R(v). In the above example, the embedded clause would now only be interpreted in worlds where it rains next week.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
sequence (2a–b), where the presupposition of the embedded clause of (2b) is not a presupposition of its embedding sentence as a whole. The reason is that after the interpretation/update of (2a) there is a world v consistent with what is presupposed in the actual world w in which John used to smoke and in which it is presupposed that John used to smoke. Thus, because from such a world v only worlds in which John used to smoke are accessible, the embedded sentence of (2b) can be interpreted appropriately as well. The concrete accessibility relation R discussed above illustrates what it means that after the update of R with )A, the A-worlds are the preferred ones: although in each v 2 R(w) it was the case that both )A and ):A were true, only )A is true at all worlds v 2 Upd()A, R)(w). In the introduction we noted that Karttunen & Peters’ (1979) twodimensional analysis gives rise to the binding problem: if it is assumed that Sue will come too presupposes, but does not entail, that somebody different from Sue will come, it falsely predicts that (1c) can be true although there is no possible world in which anyone besides Sue will come.
Robert van Rooij 293
4 EXTENDING THE ANALYSIS
4.1 Modal splitting and questions Now consider (5). (5) Either John stopped smoking, or he just started doing so.
½½¤AR;w ¼ Æ; æ iff dAi : ½½)Ai R;w ¼ Æ; æ ¼ Æ1; +æ iff ½½¤AR;w 6¼ Æ; æ and Ai dAi : ½½Ai R ;w ¼ Æ1; +æ ¼ Æ0; +æ otherwise Updð¤A; RÞ ¼ \j ðUpdð)Aj ; [i UpdðAi ; RÞÞÞ
To illustrate the update function, suppose that before the update R ¼ fÆw, væ, Æw, uæ, Æw, xæ, Æv, væ, Æv, uæ, Æv, xæ, Æu, uæ, Æu, væ, Æu, xæ, Æx, xæ, Æx, væ, Æx, uæg, where v is an A-world, u a B-world, and x makes neither A nor B true. To calculate Upd(A _ B, R), we first calculate Upd(A, R) [ Upd(B, R). It is easy to see that Upd(A, R) [ Upd(B, R) ¼ R ¼ fÆw, væ, Æw, uæ, Æv, væ, Æv, uæ, Æu, uæ, Æu, væg. Now we determine Upd()A, R) \ Upd()B, R). This new accessibility relation, R, is fÆw, væ, Æw, uæ, Æv, væ, Æu, uæg, a relation that is not Euclidean. Making use of the last step, I propose that A _ B requires both )A and )B to be appropriate. As for the case of possibility statements, this means that normally all the presuppositions of the disjuncts are presuppositions of the whole disjunction as well. However, when the context is split, this doesn’t have to be the case. This accounts for the problematic (5), if we assume that the context was split (perhaps after accommodation of the disjunctive presupposition) between, on the one hand, worlds where John smoked before, and, on the other, worlds where he did not. Questions give rise to modal subordination too:20 (6) Did John used to smoke? and did he stop smoking? Although it is standardly assumed that a polar question gives rise to a partition, with respect to modal subordination there seems to be 20
For a different analysis of modal subordination with questions, see van Rooy (1998).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Landman (1985) proposed to account for such examples by assuming that the disjuncts should be interpreted with respect to two mutually exclusive subordinated contexts, possibly created by an earlier use of a disjunctive sentence that has split the context. Our analysis of possibility statements suggests a straightforward analysis (where A is a set of formulas and RA is fÆu, væ 2 R : [[A]]R,v ¼ Æ, +æg):
294 A Modal Analysis of Presupposition and Modal Subordination a difference between the positive and the negative answer: only the positive answer can be picked up to figure as the domain of a later modal. To account for this, we can simply assume that the update of R with the yes/no question A? is the same as the update of R with )A: Upd(A? ,R) ¼ Upd()A,R). And this gives rise to the following correct predictions: after question (7a), both (7b) and (7c) are appropriate and do not give rise to presuppositional readings: (7) a. Did anyone solve the problem? b. It is possible that it was John who solved the problem. c. Either it was John who solved the problem, or the problem was too difficult.
4.2 Indicative conditionals Conditionals show modal subordination behavior as well. Example (8a) shows that the antecedent of a conditional might depend on an earlier epistemic modal; (8b) shows that the interpretation of an epistemic modal may depend on a conditional sentence used earlier, while (8c) shows that the interpretation of one conditional can depend on the interpretation of another. (8) a. I might have been wrong. If I realize that I was wrong, I will tell everybody. b. If John feels bad, he will start smoking. His girlfriend might make him to stop. c. If Mary comes, we’ll have a quorum. If Susan comes too, we’ll have a majority. I will follow Stalnaker (1976) in assuming that not only subjunctive, but also indicative conditionals should be analysed in terms of selection functions/similarity relations. I will assume that the selected worlds are such that they satisfy the presupposition of the antecedent. Out of context, the same will be presupposed in the selected worlds, or possibilities, as in the actual possibility. However, the selected
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The reason is that after the update with (7a), the most preferred worlds are ones where somebody solved the problem. As a result, if the question was non-redundant, the presupposition of the embedded sentence of (7b) is satisfied in at least one accessible world. Thus, the sequence (7a–b), represented abstractly as A? ^ )(@A ^ B), is correctly predicted to be appropriate, without giving rise to the presupposition that A. A similar consideration shows that the same holds for the sequence (7a–c).
Robert van Rooij 295
possibilities might also depend on an earlier introduced subordinated context. To account for this, I make use of a similarity relation between worlds, v <w u, meaning that v is at least as close to w as u is. To make the selected antecedent-worlds explicitly depend on what is presupposed, I will define a new relation
v
ðiÞ v <w u; or ðiiÞ v w u and RðuÞ RðvÞ4RðwÞ; or ðiiiÞ v w u and RðwÞ4RðvÞ RðuÞ:
fÆR;wæ ðAÞ ¼ fv 2 W j ½½AR;v ¼ Æ1; +æ & :du 2 W : ½½AR;u ¼ Æ1; +æ & u <ÆR;wæ vg
A conditional sentence of the form if A then B is then counted as true in ÆR, wæ iff B is verified in all pointed models consisting of a closest A-world to ÆR, wæ and accessibility relation Upd(A, R):
½½A>BR;w ¼ Æ1; +æ
iff
fÆR;wæ ðAÞ4fv 2 W j ½½BUpdðA;RÞ;v ¼ Æ1; +æg
Following Stalnaker’s (1976) suggestion that the antecedent of an indicative conditional selects, if possible, worlds compatible with what is presupposed, I will assume that the use of an indicative conditional demands there to be an accessible world in which its antecedent is true and appropriate. That is, if A > B represents an indicative conditional, it can only be appropriate in ÆR, wæ iff [[)A]]R,w ¼ Æ1, +æ. Notice that this enables us already to account for sequence (8a). This is because after the update of R with possibility statement )A?, the appropriateness condition that an indicative conditional of the form (@A ^ B) > C comes with, i.e., that A is presupposed in at least one world in Upd()A, R)(w), is guaranteed. To account for sequences (8b) and (8c), we have to make sure that conditional sentences themselves make subordinated contexts accessible. In these subordinated contexts, both antecedent and consequent
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Thus, v is closer to w than u with respect to R iff either v is closer to w than u, or they are equally close, but fewer additional presuppositions (in addition to those made in w) are made in v than in u, or fewer presuppositions (among those made in w) are suspended in v than in u. The set of closest A-worlds to ÆR, wæ is the following set:
296 A Modal Analysis of Presupposition and Modal Subordination should be true and presupposed. This suggests that we should define the update rule for (indicative) conditionals as follows: UpdðA>B; RÞ ¼ fÆu; væ 2 R : ½½A>BR;v ¼ Æ1; +æ and if ½½A ^ BR;u ¼ Æ1; +æ; then ½½A ^ BR;v ¼ Æ1; +æg With the help of this update rule we can account for (8b) and (8c) as well. The reason is that according to this rule, all accessible worlds make the conditional true, and worlds in which both the antecedent and consequent are true only ‘see’ other such worlds. But this means that a later modal or conditional statement that presupposes what is entailed by the antecedent and/or consequent, as do the second sentences of (8b) and (8c), can now also be interpreted appropriately.
According to Karttunen (1974), Stalnaker (1988), Heim (1992) and Zeevat (1992), a belief attribution like (9a) presupposes (9b):21 (9) a. John believes that Mary stopped smoking. b. John believes that Mary used to smoke. How can we account for this in our two-dimensional approach? We have assumed that what is presupposed can be represented by a primitive accessibility relation in the model. Suppose that our model contains the accessibility relation Bj which represents what John believes as well. If R(w) represents what is presupposed in w, then Rj(w) ¼ [fBj(v) : v 2 R(w)g represents what is presupposed in w about what John believes. This suggests the following definition of an indexed accessibility relation: Rwj ¼ fÆu; væ 2 W 2 j dx 2 RðwÞ : v 2 Bj ðxÞg: This definition has the effect that for all worlds v, Rwj ðvÞ ¼ Rj ðwÞ: Notice that Rwj is introspective: "v"u 2 Rwj ðvÞ : Rwj ðuÞ ¼ Rwj ðvÞ: Using this indexed accessibility relation, we can adopt the following combined truth and appropriateness conditions: w
½½Belðj; AÞR;w ¼ Æ1; +æ iff "v 2 Bj ðwÞ : ½½ARjw;v ¼ Æ1; +æ iff dv 2 Bj ðwÞ : ½½ARj ;v ¼ Æ0; +æ ¼ Æ; æ otherwise This immediately accounts for the non-presuppositional reading of the sequence (9b)–(9a). In particular—and similarly as for Stalnaker (1988)—we don’t have to introduce a new update function for belief attributions: any update of R automatically updates Rwj as well.
21
For disagreement, see Geurts (1998).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
4.3 Belief and desire
Robert van Rooij 297
According to Heim (1992), desire attributions should be interpreted with respect to what is (presupposed to be) believed as well. The following discourse seems perfectly acceptable: (10) a. John believes that Mary used to smoke, b. but he hopes that she stopped doing so. But as noted by Asher (1987), Heim (1992), and later by Geurts (1998), desire attributions can be conditionally dependent on other desire attributions as well: (11) a. John wants Mary to come, b. and he wants Bill to know that Mary will come.
½½Desðj; AR;w ¼ Æ; æ iff ½½)ARj ;w ¼ Æ; æ ¼ Æ1; +æ iff ½½)ARj ;w ¼ Æ1; +æ and fv 2 Bj ðwÞ : ½½ARj ;v ¼ Æ1; +æg > fu 2 Bj ðwÞ : ½½ARj ;u ¼ Æ0; +æg ¼ Æ0; +æ otherwise
UpdðDesðj; AÞ; RÞ ¼ fÆu; væ 2 R j ½½Desðj; AR;v ¼ Æ1; +æg
UpdðDesðj; AÞ; Rj Þ ¼ fÆu; væ 2 Rj j if ½½ARj ;u ¼ Æ1; +æ; then ½½ARj ;v ¼ Æ1; +æg
The truth condition basically says that the embedded sentence should be interpreted with respect to all worlds in Bj(w) in which this sentence can be interpreted appropriately. It also demands that if A 22
For a discussion of some alternative analyses of desire attributions, see Van Rooy (1999).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Intuitively, this suggests that desire attributions can introduce subordinated contexts and can be interpreted with respect to such contexts as well. How can we account for this? Just as we earlier took R to be a primitive accessibility relation that cannot be reduced to what the participants of the conversation know or believe, I think that now we have to assume the existence of a primitive accessibility relation Rj that represents what has been explicitly established about what John believes. Thus, in the beginning of the conversation Rj ¼ W 3 W and UpdðBelðj; AÞ; Rj Þ ¼ fÆu; væ 2 Rj : ½½ARj ;v ¼ Æ1; +æg: I will assume for simplicity the following analysis of desire attributions: Des(j, A) is true in w iff all the A-worlds in Bj(w) are preferred to the :A-worlds in Bj(w).22 Taking presupposition satisfaction into account as well, we define the interpretation rule as follows (where X > Y iff "x 2 X: "y 2 Y: x > y):
298 A Modal Analysis of Presupposition and Modal Subordination presupposes P, the desire attribution is predicted to be appropriate if either P is presupposed to be believed by the agent, or P is presupposed to be desired. The update rules are similar to what we have discussed above: the first one just demands that the desire attribution has to be true, while the second turns the A-worlds in Rj(w) into the preferred ones.
4.4 Subjunctive mood and negative sentences
(3a) A thief might break into the house. He would take the silver. But now consider the following example: (12) I don’t smoke. I wouldn’t be able to stop. This example is more problematic than (3a) because now there is no modal operator in the first sentence of the sequence that has a non-global effect on the accessibility relation. To account for this example one might propose, again, to accommodate the presupposition locally within the scope of the subjunctive modal. I would like to suggest, however, that, once we represent what is presupposed by an accessibility relation, no need for local presupposition accommodation arises. The first idea that comes to mind to account for examples like (12) is to propose that negative sentences have an effect similar to that of modals. But it is not straightforward to work out this suggestion, because the worlds that verify the negated clause are not accessible anymore after the interpretation of the first sentence. In this section I would like to suggest tentatively two alternative solutions for such examples. A first proposal would be to assume that the modal in the second sentence is interpreted with respect to an accessibility relation that is obtained by subtracting from the original accessibility relation (in which the whole of (12) is interpreted) the relation resulting from the interpretation of the first sentence of the sequence. Let t be the moment in time at which the second sentence should be interpreted. Then we can define Rt as fÆu, væ 2 Rt1ju,v;Range(Rt)g. Thus, Rt consists of the arrows between worlds in the previous information state,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In section 3 we saw already that certain examples of modal subordination involving subjunctives are unproblematic for our account. Our analysis of possibility operators and their duals in section 3.2 immediately gives the desired reading for a sequence such as (3a):
Robert van Rooij 299
Rt1, that were eliminated by the last assertion. In terms of this accessibility relation, we can analyze the subjunctive mood as follows:
t
t
½½Would AR ;w ¼ Æ1; +æ iff dv 2 Rt ðwÞ : ½½AR ;v ¼ Æ; +æ and t " v 2 Rt ðwÞ : if ½½AR ;v ¼ Æ; +æ; t then ½½AR ;v ¼ Æ1; +æ
23 So, with Kaufmann (2000) I agree that we need to keep previous contexts ‘in memory’. Accepting this—according to one reviewer—we give up the assumption that we can account for modal subordination in terms of a single accessibility relation. But I don’t agree, for we still might think of R as a single accessibility relation for each time point. In systems that add to possible worlds a temporal dimension it is very normal to assume perfect memory: at each point in time you recall what you knew, believed, or presupposed before. To implement this assumption, we don’t have to add to the model some extra accessibility relations, although the accessibility relation now has to represent more information. I won’t go into the details of such a system here.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Notice that according to this interpretation rule of a subjunctive modal, the second conjunct of a formula like :A ^ Would (@A ^ B) is predicted to be true with respect to accessibility relation R if all accessible Aworlds are B-worlds as well. In particular, we can now account for the sequence (12) without making use of local accommodation. Still, I don’t think this proposal is unproblematic. The most problematic aspect, I believe, is the fact that the suggested analysis can’t explain the contrast between positive and negative sentences: why can subjunctive modals only ‘pick up’ negated sentences? As discussed by Horn (1989), among others, there exists a crucial distinction between the contexts in which positive and in which negative sentences can be used appropriately: in contrast to their positive counterparts, negative sentences require a context in which the truth of the positive sentence is expected, or at least very salient. One way of being salient is to be the topic of conversation: a question to be addressed. Geurts (1995) suggested tentatively that this might be the reason why sequences like (12) are appropriate. I think that this is indeed a good suggestion. We have seen in section 4.1 what the dynamic effect of a (positive polar) question is: the worlds in which the positive sentence is true come to be preferred to worlds where it is false. A subsequently used modal expression can then be interpreted with respect to these most preferred worlds. This can’t be the whole story, of course, because after the first sentence of (12) these most preferred worlds are eliminated. So, if t is the moment in time at which the second sentence should be interpreted, the relevant accessibility relation should not be Rt, but rather Rt1, i.e. the context of interpretation for the first sentence of (12).23 I think that this suggestion is a natural one, especially given the fact that would is a past-tense modal. If this first sentence presupposes that the
300 A Modal Analysis of Presupposition and Modal Subordination topic of conversation is whether I smoke, the most preferred worlds in Rt1(w) are all worlds where I smoke, and the second sentence of (12) can be interpreted appropriately. 5 INDEFINITES AND PRONOUNS
Updðdxr A; ÆR; aæÞ ¼ ÆfÆi; jæ 2 R½x=r : ½½AR½x=jðrÞ;j ¼ Æ1; +æg; a½x=aðrÞæ Thus, in the actual possibility, the speaker’s reference of the indefinite is introduced (although this need not be an individual that
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Although we have discussed Karttunen & Peters’ (1979) binding problem already with respect to possibility statements, the most famous problematic example involves indefinites: their false prediction that the individual that satisfies the presupposition of a sentence like Someone managed to succeed George V on the throne of England need not be the one who actually succeeded George V. In this section I will show that this problem will not arise in our framework if we extend it to the predicate-logical case. However, the main goal of this section is to indicate how we can account for modal subordination phenomena that involve anaphoric dependencies across the sentential boundary. To take indefinites and pronouns into account, we have to make our accessibility relation one between more fine-grained possibilities. In contrast to standard dynamic semantics, I assume that pronouns are (normally) used referentially, referring back to the speaker’s reference of its antecedent indefinite. Such a speaker’s reference of (an occurrence of ) an indefinite depends on the referential intentions the speaker has. This kind of information should be represented already in a possibility (actual or non-actual). But this means that these possibilities have to contain more information than the world-assignment pairs standardly assumed in dynamic semantics. A clause with an occurrence of an indefinite is represented by dxrA, where r is a reference function. Let us assume that the set of possibilities I is a set of functions from (i) n-ary predicates to their interpretations; (ii) variables to individuals; and (iii) reference functions to individuals. If dxrA is interpreted in possibility i, then i(r) is the speaker’s reference of the occurrence of the indefinite in i, and the dynamic effect will be that from now on x will be assigned to i(r) in i, i.e. i(x) ¼ i(r). Let us define R[x/r] as fÆi[x/i(r)], j[x/j(r)]æ : Æi, jæ 2 Rg, and R[x/d] as fÆi[x/d], j[x/d]æ : Æi, jæ 2 Rg. Now we define the update of ÆR, aæ with dxrA as follows (where a is the actual possibility):
Robert van Rooij 301
makes the sentence true). Also in each possibility that is compatible with what is presupposed, the speaker’s reference of the indefinite in that possibility is introduced, although this introduced individual has to verify the sentence in that possibility. The (rigid) truth and presupposition satisfaction conditions of the new clauses are given below (where ~ x is an n-ary sequence)):
½½P~ xR;i ¼ Æ1=ð0Þ; +æ iff
½½dxr AR;i ¼ Æ1; +æ iff
ið~ xÞ 2 iðPÞ ðor ið~ xÞ;iðPÞÞ and ið~ xÞ is defined x x ½½AR½ =iðrÞ ;i½ =iðrÞ ¼ Æ1; +æ
R; a~+ A
iff
da# a : ½½AR;a# ¼ Æ1; +æ x
x
Now it follows that R; a~+ dxr Px iff dd 2 D : ½½PxR½ =d ;a½ =d ¼ Æ1; +æ: But this means that the sentence is true in a just in case there is an individual that has property P in this world, just as expected. In van Rooy (2001), I argued that the notion of speaker’s referent is important for at least two reasons. First, to account for the phenomenon of pronominal contradiction: although speaker’s reference has no truth conditional effect on the interpretation of indefinites, it does for the interpretation of pronouns. Second, to understand what a discourse referent used in dynamic semantics really represents: a discourse referent represents what is presupposed about the actual speaker’s referent. I will not discuss these arguments any further here. However, in one sense the present implementation of the second intuition is much more appealing than the one I have given in van
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Notice that the above rules say that dxrP(x) is rigidly true in ÆR, aæ if and only if the speaker’s referent of the indefinite in a has property P in this world/possibility. Existential sentences, however, don’t seem to have such strong truth conditions. As argued for in van Rooy (2001), although speaker’s reference is crucial for the analysis of pronouns, it doesn’t seem to influence the truth or falsity of the clause in which the indefinite occurs. To account for this, I will follow the same procedure as I proposed in van Rooy (2001), and define the semantic notion of truth as an abstraction of the more pragmatic notion of rigid truth where speaker’s reference is crucial for the interpretation of indefinites. Let us say that j i iff j is a possibility just like i, except that j might assign different individuals to reference functions than i does. Now I define the notion of truth (and presuppositional appropriateness) of sentence A in possibility ÆR, aæ in terms of this notion as follows:
302 A Modal Analysis of Presupposition and Modal Subordination
(13) a. A wolf may come in. It would eat you first. b. )dxr[Wx ^ Cx] ^ hEx Notice that [[)dxr[Wx ^ Cx] ^ h Ex]]R,a ¼ Æ1, +æ iff [[)dxr[Wx ^ Cx]]]R,a ¼ Æ1, +æ and [[h Ex]]R#,a# ¼ Æ1, +æ, where ÆR#, a#æ ¼ Upd()dxr[Wx ^ Cx], ÆR, aæ). The first conjunct is true iff di 2 R(a) : i(r) 2 i(W) \ i(C). The update of ÆR, aæ with the first conjunct results in fÆj, iæ 2 R[x/r] : i(r) 2 i(W) \ i(C)g [ fÆj, iæ 2 R : :dd 2 D : d 2 i(W) \ i(C)g.24 If we then only look at possibilities i 2 R#(a) where i(x) is defined, the second conjunct is predicted to be true iff "i 2 R(a): "d 2 D : if d 2 i(W) \ i(C), then d 2 i(E). Thus the second conjunct says that in every possibility where there is a wolf who comes in, it eats you first. I believe that this is the correct reading for the second sentence of (13a). In section 4.1 we followed Landman’s modal splitting analysis of disjunctive sentences. Intuitively, such an analysis must be able to predict that (14a) really means (14b). (14) a. Call this number. The phone will be answered by either a doctor or a secretary. The doctor can tell you right away what’s a matter with you, or the secretary can make an appointment for you.
24
x
Based on the assumption that [[dxrA]]R,i ¼ Æ0, +æ only if :dj i : ½½AR½
=iðrÞ ;j½x =jðrÞ
¼ Æ1; +æ.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Rooy (2001): it’s accounted for now in terms of standard set theory making use of a standard way to model propositional attitudes, i.e. an accessibility relation. The binding problem of Karttunen & Peters’ (1979), involving indefinites, was due to the fact that they represented presupposition and assertion separately. Our analysis, instead, only interprets them at different dimensions. We represent their problematic sentence abstractly as follows: dxr [@Px ^ Qx]. An easy calculation shows that this formula is predicted to be true and appropriate in ÆR, aæ, R, xa ~x+ dxr [@Px ^ Qx], just in case dd 2 D : " i 2 RðaÞ : ½½PxR½ =d ;i½ =d ¼ Æ1; +æ & x x ½½QxR½ =d ;a½ =d ¼ Æ1; +æ: Thus, it is required that the same individual has to satisfy both the presuppositional part and the assertive part: the binding problem does not occur. This prediction is independent of our assumption that indefinites come with speaker#s referents. How does our analysis of indefinites and pronouns account for anaphoric dependencies across modal statements? Consider the following classical sequence:
Robert van Rooij 303
b. Either a doctor will answer and he can tell you what is wrong, or a secretary will answer and she can make an appointment. Let us assume that Upd(A _ B, ÆR, aæ ) ¼ Upd(A, ÆR, aæ) kUpd(B, ÆR, aæ), where ÆR, aæ k ÆR#, a#æ ¼ ÆR [ R#, a [ a#æ.25 If we now represent (14a) as something like (dxrA _ dysB) ^ (Px _ Qy), this is indeed what we predict. The new possibility will be ÆUpd(A, R[x/r]) [ Upd(B, R[y/s]), a[x/a(r), y/a(s)]. Because of the definedness condition, we correctly predict that the first disjunct of the second disjunction will be regarded as a continuation of the first disjunct of the first disjunction, and similarly for the second disjuncts.
In this paper I have proposed a modal two-dimensional analysis of presupposition and modal subordination. For the analysis of presupposition I combined the strong points of the standard dynamic analysis and the two-dimensional one: what is presupposed by a sentence follows from the interpretation rules, and presupposition satisfaction is determined (almost) independently from truth. For the analysis of modal subordination I proposed that the embedded clauses of modal statements should be interpreted with respect to possibilities that verify what is presupposed. Roberts (1989, 1996) discusses many more phenomena under the heading of ‘modal subordination’ than I do in this paper. Some of her additional examples, e.g. bathroom sentences, I would definitely interpret in terms of descriptive pronouns instead of as involving modal subordination (cf. van Rooy, 2001). Other examples, however, in particular those where the interpretation of one quantified phrase depends on that of another, are closer to the ones discussed in this paper. It remains to be seen whether, and if so how, we should extend our analysis to cover these examples as well. Acknowledgements The research for this paper has been made possible by a fellowship of the Royal Netherlands Academy of Arts and Sciences (KNAW), which is gratefully acknowledged. I am grateful to Cleo Condoravdi and Stefan Kaufmann for inviting me to submit a paper to the special issue. I would like to thank an anonymous referee, 25 Notice that because we think of worlds as a combined function from (i) propositional variables to truth values; (ii) discourse referents to individuals; and (iii) variables to individuals, the union of two worlds is well-defined. Notice also that after the update of a world with new information, this can have an effect only on the individuals it assigns to variables.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
6 CONCLUSION
304 A Modal Analysis of Presupposition and Modal Subordination Frank Veltman, Henk Zeevat, and the editors of this special issue for valuable comments and discussion.
ROBERT VAN ROOIJ Institute for Logic, Language and Computation (ILLC) University of Amsterdam Nieuwe Doelenstraat 15 1012 CP Amsterdam The Netherlands. e-mail:
[email protected]
Received: 15.03.03 Final version received: 30.03.04 Advance Access publication: 03.05.05
Asher, N. (1987) ‘A typology of attitude verbs and their anaphoric properties’. Linguistics and Philosophy 10:127–189. Beaver, D. (1995) Presupposition and Assertion in Dynamic Semantics. Ph.D. dissertation. CCS, Edinburgh. Fernando, T. (1995) ‘Are context change potentials functions?’ In H. Kamp and B. Partee (eds), Contexts in the Analysis of Linguistic Meaning. Palgrave MacMillan, Stuttgart/Prague. Frank, A. (1997) Context Dependence in Modal Constructions. Ph.D. dissertation, University of Stuttgart. Gazdar, G. (1979) Pragmatics. Academic Press. London. Gerbrandy, J. (1999) Bisimulations on Planet Kripke. Ph.D. Dissertation. University of Amsterdam. Geurts, B. (1995) Presupposing. Ph.D. Dissertation. University of Stuttgart. Geurts, B. (1998) ‘Presuppositions and anaphors in attitude contexts’. Linguistics and Philosophy 21:545–601. Groenendijk, J. and M. Stokhof (1991) ‘Dynamic predicate logic’. Linguistics and Philosophy 14:39–100. Heim, I. (1982) The Semantics of Definite and Indefinite Noun Phrases. Ph.D. dissertation. University of Massachusetts, Amherst.
Heim, I. (1983) ‘On the projection problem for presuppositions’. In M. Barlow et al. (eds), West Coast Conference on Formal Linguistics. Stanford University, 114–123. Heim, I. (1992) ‘Presupposition projection and the semantics of attitude verbs’. Journal of Semantics 9:183–221. Herzberger, H. (1973) ‘Dimensions of truth’. Journal of Philosophical Logic. 2:535–556. Horn, L. (1989) A Natural History of Negation. University of Chicago Press. Chicago. Kamp, H. (1981) ‘A theory of truth and semantic representation’. In J. Groenendijk et al. (eds), Formal Methods in the Study of Language. Mathematical Center, Amsterdam, 277–322. Karttunen, L. (1974) ‘Presuppositions and linguistic context’. Theoretical Linguistics. 1:181–194. Karttunen, L. and S. Peters (1979) ‘Conventional implicature’. In C.-K. Oh & D. Dinneen (eds), Syntax and Semantics, Presupposition. Vol. 11. Academic Press. New York, 1–56. Kaufmann, S. (1997) ‘What’s at stack in discourse’. In P. Dekker et al. (eds), Proceedings of the the 11th Amsterdam
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
REFERENCES
Robert van Rooij 305 Tilburg University Press. Tilburg, 534–548. Rooy, R. van (2001) ‘Exhaustivity in dynamic semantics; Referential and Descriptive Pronouns’. Linguistics and Philosophy 24:621–657. Sandt, R. van der (1988) Context and Presupposition. Croom Helm. London. Sandt, R. van der (1992) ‘Presupposition projection as anaphora resolution’. Journal of Semantics 9:223–267. Soames, S. (1979) ‘A projection problem for speaker presupposition’. Linguistic Inquiry 10:623–666. Soames, S. (1989) ‘Presupposition’. In D. Gabbay & F. Guenthner (eds), Handbook of Philosophical Logic. Volume IV, 553–616. D. Reidel. Dordrecht. Stalnaker, R. (1974) ‘Pragmatic presupposition’. In Munitz & Unger (eds), Semantics and Philosophy. NYP, 197– 213. Stalnaker, R. (1976) ‘Indicative conditionals’. In A. Kasher (ed.), Language in Focus. Reidel. Dordrecht, 179–196. Stalnaker, R. (1988) ‘Belief attribution and context’. In R. Grimm & D. Merrill (eds), Contents of Thought. Tuscon. University of Arizona Press, 140–156. Stalnaker, R. (2002) ‘Common ground’. Linguistics and Philosophy 25:701–721. Veltman, F. (1996) ‘Defaults in Update semantics’. Journal of Philosophical Logic 25:221–261. Zeevat, H. (1992) ‘Presupposition and accommodation in update semantics’. Journal of Semantics 9:379–412.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Colloquium. ILLC. Amsterdam, 187– 192. Kaufmann, S. (2000) ‘Dynamic discourse management’. In M. Faller et al. (eds), Formalizing the Dynamics of Information. CSLI Publications. Stanford, 171–188. Kibble, R. (1994) ‘Dynamics of epistemic modality and anaphora’. In H. Bunt et al. (eds), International Workshop on Computational Semantics, ITK, Tilburg, 121–130. Kratzer, A. (1981) ‘The notional category of modality’, In J. Eikmeyer & H. Rieser (eds), Words, Worlds, and Contexts. W. de Gruyter, Berlin, 38– 74. Landman, A. (1986a) Towards a Theory of Information. Ph.D. dissertation. University of Amsterdam. Landman, A. (1986b) ‘Conflicting presuppositions and modal subordination’. In Papers from the 22nd Regional Meeting, Chicago Linguistic Society, Chicago, 195–207. Roberts, C. (1989) ‘Modal subordination and pronominal anaphora in discourse’ Linguistics and Philosophy 12:683–721. Roberts, C. (1996) ‘Anaphora in intensional contexts’. In S. Lappin (ed.), Handbook of Contemporary Semantics. Blackwell, Oxford, 215–246. Rooy, R. van (1998) ‘Modal subordination in questions’. In J. Hulstijn & A. Nijholt (eds), Twendial #98, 237– 248. Rooy, R. van (1999) ‘Some analyses of pro-attitudes’. In H. de Swart (ed.), Logic, Game Theory and Social choice.
Journal of Semantics 22: 307–337 doi:10.1093/jos/ffh021 Advance Access publication June 20, 2005
Causality and Tense—Two Temporal Structure Builders LEONOOR OVERSTEEGEN Universiteit van Tilburg
Abstract
1 BACKGROUND Models for temporal representation generally focus on expository or narrative texts. Numerous linguists describe what tense tells us about the location of an eventuality in relation to the moment of speech, or about the relation between eventualities in successive sentences (for example, Reichenbach 1947; Kamp & Rohrer 1983; Hinrichs 1986; Moens & Steedman 1988; Webber 1988; Lascarides & Asher 1993). The relative ordering of eventualities expressed in successive sentences is especially complex—I simplify it here to suit my purpose. In general, the ordering between the eventualities expressed by two sentences in the simple past depends on the types of eventuality.1 Usually, two states overlap whereas two events are interpreted as successive. For two successive 1
Ever since Vendler (1957), classifications of types of events have been proposed, e.g. by Bach (1986) and Moens & Steedman (1988). In this paper, the basic distinction between states and events is respected but more refined distinctions are left aside. The Author 2005. Published by Oxford University Press. All rights reserved.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
By force of causes precede effects, causality contributes to the temporal meaning of discourse. In case of semantic causal relations, this contribution is straightforward, but in case of epistemic causal relations, it is not. In order to gain insight into the semantics of epistemic causal relations, paradoxical cases are analyzed of text fragments in which temporal and causal meaning seem to be irreconcilable. A solution is proposed consisting of several parts: First, an analysis of epistemic causal coherence relations, in which the effect sentence is argued to be a (covert) belief sentence. Secondly, a method for deriving (epistemic) causal coherence relations. And thirdly, a format for representing temporal meaning in which contributions of tense and causality can be reconciled. Finally, it is argued that understanding epistemically causally connected sequences (or: argumentation) is not equal to being persuaded to believing them. The difference is shown to be expressible in the representation format proposed.
308 Causality and Tense—Two Temporal Structure Builders
(1) Causes precede effects h causeðe1 ; e2 Þ / :e2 <e1 The law in (1) can be phrased as follows: it is necessarily the case that an eventuality e1 which causes another eventuality e2 is not preceded by that eventuality e2. In an interval semantics approach, one could also use the positive variant of (1), requiring that the initial point of the interval representing e1 has to precede the initial point of the interval representing e2. This requirement would result in different relations between the intervals, depending on the event types expressed (cf. Hinrichs 1986; Moens & Steedman 1988; Webber 1988).3 These two sources should not supply conflicting temporal information. ‘A text is temporally incoherent if the natural interpretation of text is such that there are no inferrable relations between the events. A text is temporally misleading, or as we shall say, unreliable if the natural interpretation of the text is such that the inferred relations between the events differ from their actual relations in the world.’ (Lascarides and Oberlander, 1993: 6). In most (temporally) coherent
2
This default, a consequence of the anaphoric property of tense, can be overruled by several (context-)factors as in: (The party went on all night.) But John had left early because his ex-girlfriend showed up (anonymous reviewer). The complex nature of the pluperfect is the main topic of Oversteegen & Bekker (2002). 3 If one wants to allow for causal relations between eventualities starting simultaneously, the relation between the initial points of the corresponding intervals should be < instead of <.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
sentences, one of which is in the simple past and the other in the pluperfect, the default situation is a precedence relation: the eventuality in the pluperfect will precede the eventuality in the simple past.2 This default is independent of the sequence of tenses in the sentences (simple past before pluperfect, or pluperfect before simple past). A second source of temporal information is causality. Causal relations are relations between events (von Kutschera 1993), or rather between eventualities. A causal coherence relation between the eventualities expressed in successive sentences may be marked by a causal connective, as it is in John fell because Max pushed him. Alternatively, it may be inferrable, as in John fell. Max (had) pushed him. In both cases, the causal relation implies a temporal relation. From ancient times on and in all disciplines, the temporal property of causality has been recognized. Lascarides and Asher (1993) refer to this regularity as an indefeasible law, the Causal Law:
Leonoor Oversteegen 309
texts, the two sources of temporal information collaborate nicely, as they do in the examples in (2): (2a)
John went to the hospital. He had twisted his ankle on a patch of ice. (Webber 1988) (2b) John arrived late for work. He had totalled his car. (Lascarides & Asher 1993) (2c) Then everybody applauded. Keith had announced his promotion. (Caenepeel 1995)
(3)
Connections when changing tense (CCT) if two sequential sentences, a and b, must be related by a coherence relation, and a carries the simple past tense, and b carries the pluperfect, then b is (part of ) an elaboration or explanation of a, or b and a form parallel or contrast, with the event in b preceding that described in a. (Lascarides & Asher, 1993: 471)
The only discourse relation allowed by (3) which is a causal relation is explanation.4 Because of the last part of the definition, we expect the latter sentence to give an explanation of the former in a causally related sequence of the intended type. Indeed, all sentence pairs in (2) are related by explanation.
4 The definition of explanation given by Lascarides & Asher (1993) is: Æs, a, bæ ^ cause(me(b), me(a)) > Explanation (a,b) where me(a) stands for ‘main eventuality described in a’.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In all of these cases, both the tense combination of the two sentences and the causal relation inferrable between the eventualities expressed in them put the eventuality in the latter sentence before the eventuality in the former sentence. If causality implies temporal precedence, we should expect to find mutual restrictions between tense and causality. Inter-sentential causal relations, or causal coherence relations, should occur only between sentences with certain tense combinations, and not with others. In other words, the combination of tenses in two sequential clauses would restrict the set of options of possible coherence relations expressed. Indeed, we find a rule in Lascarides and Asher (1993) which narrows down the set of possible (causal) coherence relations between a specific combination of tensed sentences. A (somewhat simplified) description of the rule is given in (3).
310 Causality and Tense—Two Temporal Structure Builders 2 A PROBLEM WITH EPISTEMIC CAUSAL RELATIONS
(4a) John came back because he loved her. (4b) John loved her, because he came back. In (4a), real-world causality connects the two clauses. Let us call this a semantic causal relation. With respect to (4b), Sweetser claims that the speaker’s knowledge of John’s return causes the conclusion that John loved her. Whereas the causal relation in (4a) indeed connects two eventualities ‘in the world’, the causal relation in (4b) concerns a realworld eventuality and a conclusion or a claim or a belief about a real world eventuality. Examples like (4b) are labeled epistemic causals.5
5 Knott & Mellish (1996) attempt to understand what happens with regard to causal relations in argumentation as well. However, they understand the epistemic causal relation as being symmetrical. See Knott and Mellish, 1996: 154 (15):
(15) The footprints are deep. It follows that/so the thief was a heavy man. They define an epistemic relation as a relation between the intended effect of each individual utterance. Why would this be undesirable? Argumentation is not based on symmetrical relations between data and conclusions. It is exactly because the speaker supposes that his data are easily adopted—or even belong to the shared knowledge—that he hopes to make the hearer believe his conclusion or beliefs. Therefore, I adopt Sweetser’s view that an epistemic causal relation defines a relation between an eventuality and a belief about another eventuality.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
It is clear how the two sources of temporal information may collaborate. A causal relation may be lexicalized, e.g. by a causal connective, or inferred, with the help of real-world knowledge. In both cases, we use the indefeasible law in (1), causes precede effects, to determine the sequence of eventualities. This sequence usually parallels the sequence dictated by the combination of tenses. Causal relations between sentences are not interesting for their temporal property only; the discourse theoretical literature focuses on other aspects of causal coherence relations. For the entire set of coherence relations, of which causal relations are a subset, a dichotomy is recognized. This dichotomy has been described as semantic versus pragmatic (Van Dijk 1979; Sanders et al. 1992); subject matter versus presentational (Mann & Thompson 1987); ideational versus rhetorical (Redeker 1990); and ideational versus interpersonal (Systemic Functional Grammar in general). In order to explain the nature of the dichotomy, I shall use Sweetser (1990). Sweetser (1990) distinguishes different types of domains of interpretation: coherence relations can be interpreted in each of them. The following causal examples are taken from Sweetser (1990: 77):
Leonoor Oversteegen 311
Eve’s husband (presumably) had hit her (e1) because she immediately came out again (e2). (6) Zeker een derde deel van de kamer werd door het zwarte instrument in beslag genomen (e1). Gek was ze geweest, gek dat ze zoveel geld had uitgegeven voor dat ding, dat apparaat, dat monster (e2). (At least one third of the room was occupied by the black instrument. She had been crazy, crazy to have spent so much money on that thing, that machine, that monster.) (5)
In example (5), a connective is present, defining a ‘backward’ causal relation with the eventuality in the second sentence causing the first one. In example (6), a ‘forward’ causal relation will be intended: the attribution of craziness to ‘her’ is the consequence of the negative evaluation of the presence of the big, black instrument, a grand piano. In Table 1, an overview is given of the temporal relations in (5) and (6) as implied by different sources. The columns 2 and 4 of Table 1 contain paradoxical conclusions. The temporal relations derived from different sources form a contradiction for both (5) and (6).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
One might expect that epistemic causals are not subject to the Causal Law in the way that semantic causals are. An epistemic causal relation ‘reflects the process of reasoning from simple perceptions in order to assess the probability of a state of affairs from the speaker’s point of view’ ( Jaszczolt 2003). It does not relate two eventualities and, at first glance, does not need to reflect iconicity. Indeed, we find examples of epistemic causals in which the cause appears to follow the effect, as in He presumably finished his paper, because I see him leave. Does this imply that ‘in epistemic temporal relations the overt temporal marking is irrelevant, whereas for semantic causals it must cohere with the Causal Law’ (anonymous reviewer)? That would be a sad conclusion for a linguist believing in compositional semantics, and one, I will argue, we need not draw. By analysing some more cases in which epistemic causality and tense seem to contradict each other, I will explain how tense and causality can maintain their generally accepted interpretation. Their contribution to the semantic representation structure is not a different one in the epistemic cases. The representation structure appears to be more complex but it is in line with general ideas about epistemic modality. Below are two examples of such paradoxical cases, the one constructed, the other coming from a corpus of literary texts:
312 Causality and Tense—Two Temporal Structure Builders Table 1
(5) (6)
Tense combination
Temporal relation
Causal relation
Temporal relation
PPf – sP sP – PPf
e1 < e 2 e2 < e1
Cause (e2, e1) Cause (e1, e2)
e2 < e1 e1 < e2
3 COMPONENTS OF A SOLUTION
A model for temporal representation that reflects the information provided by tense. A correct, formally supported analysis of causal relations. A description of the contribution of causal relations to the model.
I describe these three components in sections 3.1, 3.2, and 3.3, respectively. Furthermore, in order to explain the paradoxical clashes between causality and tense as in (5) and (6), the specific nature of epistemic meaning has to be addressed and it has to be incorporated in the representation model. This is done in section 4.
3.1 The TTT framework—the temporal part of the story To represent temporal information, I make use of the two track model (TTT) advocated in several publications (Oversteegen 1988, 1989, 1993; Oversteegen & Bekker 2002). It is not claimed that only TTT is suitable for the purpose. Other representation formats can do the job just as well, if they are characterized by two properties: 1. For all temporal categories, independent contributions to the representation structure are defined.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
It is not difficult to see which error is made in the analysis in Table 1: The causal relations in (5) and (6) do not hold immediately between the eventualities e1 and e2. What is caused in both cases is not the eventuality expressed, but a mental state: a belief held by the speaker about the eventuality. In other words, the first sentence in (5) and the second sentence in (6) are not simply assertions. Claims following from an argument are not supposed (by the speaker) to be purely informative: in my opinion, they should be analysed as (covert) belief sentences. What we need for the analysis of sequences like (5) and (6) is:
Leonoor Oversteegen 313
2. A certain amount of vagueness in the relations between semantic entities in the representation structure is allowed.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The motivation for choosing the type of representations used in TTT comes from McTaggart (1927), who argues that time can be thought of as something dynamic, and constantly advancing, and that it can also be considered from a God-like perspective, covering the history of all eventualities. Oversteegen (1989) proposes a linear, twotrack representation of temporal meaning in which each of these conceptualizations of time can be expressed. For each sentence, in this representation, called TTS for two track structure, both the dynamic and the static temporal meaning aspects can be reflected. In the former view, an eventuality is seen from a particular point in time and the predicates PAST, PRESENT, and FUTURE are relevant. Some expressions of time have a deictic nature, and the truth of a tensed statement depends on the temporal point of view taken. Natural language elements expressing this dynamic and deictic temporal meaning are represented on the S-track. Here, S stands for ‘speaker’or ‘point of speech’. The S-track has a point structure and harbours points of speech and points of evaluation. Points of temporal perspective in general (including points of reference) are represented on the S-track. McTaggart referred to this type of temporal ordering as the ‘A-series’. The static conceptualization of time is independent of a point of view. Eventualities are temporally ordered by the precedence relation but not with respect to the point of speech or with deixis. McTaggart referred to this type of temporal ordering as the ‘B-series’. It is referred to as static because, once an eventuality e1 is ordered before some other eventuality e2, this ordering is permanent and will not change over time. In TTT, natural language elements establishing this static and non-deictic temporal ordering are represented on the E-track, where E stands for ‘eventuality’. By means of a TTS contribution, temporal expressions are represented according to both these views, and the representation contributions are allowed to interact. Since tense is a deictic category characterised by the predicates PAST, PRESENT, and FUTURE, it is primarily represented on the S-track. Tense also establishes a relation between an eventuality on the E-track and evaluation points on the S-track; we refer to this as a ‘bridge’ between the two tracks. This point on the S-track may be S itself but it may also be an S#-point, depending on the tense used, since S# only plays a part in past tenses; the present relevance quality of present tenses is reflected by the fact that S itself functions as the relevant evaluation point.
314 Causality and Tense—Two Temporal Structure Builders
(7) present
(8) past
In TTT, the perfect is not treated as an autonomous tense but as a temporal component which combines with a basic tense form to result in a ‘full’ tense such as the present perfect. For example, the value of the relevant evaluation point will be S if the perfect is combined with a present tense, and it will be S# if the perfect is combined with a past tense.6 When the perfect combines with one of the basic tense forms, its contribution to the final representation is the constraint that the eventuality must precede the E-track element that corresponds to the relevant evaluation point. The evaluation points for the eventuality are provided by the basic tense form. Combining the perfect with the present and the past results in the derived forms shown in (9) and (10) below. (9) present perfect
(10)
past perfect
6 Something similar holds for the future. However, the future is not solely a temporal category and I will return to it in the section on modality (4.1).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Event type is associated primarily with the E-track because it is independent of temporal location with respect to the point of speech; a process located in the future will be just as much a process when it is located in the present or the past. The perfect is not dependent on the point of speech either: it can combine with past, present, or future and it is, therefore, primarily considered an E-track notion. Temporal adverbials can be deictic (cf. yesterday) and, consequently, defined on the S-track, or can express relative position (cf. an hour earlier) and, consequently, be defined on the E-track. For each clause, the linguistic categories expressing temporal meaning contribute independently to a TTS. Sequences of clauses (or larger portions of discourse) can be represented by means of a (complex) TTS. The contributions of the present tense (7) and the past tense (8) to the TTS of the entire sentence are given in the following configurations:
Leonoor Oversteegen 315
3.2 Deriving causal coherence relations It has been noted in the linguistic literature, at least since the seventies (Lakoff 1971), that the semantics of causal coherence relations is richer than the semantics of additive coherence relations. Apart from stating the truth of the propositions expressed in both clauses, there is a ‘relational proposition’ connecting them. There is no consensus about the status of this relational proposition: some have referred to it as a topos (Moeschler 1990), as an inference (Kehler 1994), as a conventional implicature (Noordman & Vonk 1992), or as a background assumption (Sidiropoulou 1992). I think it is uncontroversial, however, to state that the examples in (11a–c), containing semantic causal relations, make sense only if the speaker holds true the regularities in (12a) to (12c) and expects the hearer to do the same: (11a) The car stopped because the engine broke down. (q because p) (11b) The car stopped because it was out of petrol. (11c) The girl looked radiant because she was in love. (12a) If the engine of a car breaks down, the car will stop. (p / q) (12b) If a car is out of petrol, the car will stop. (12c) If a girl is in love, she will look radiant. 7 For definitions of temporal categories in this TTS model, I refer to Oversteegen 1986, 1989, 1993, and Oversteegen & Bekker 2002. 8 At least, this is the way in which assertions are treated. This treatment conforms to Stalnaker’s treatment of assertive propositions as functions from possible worlds to truth values.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In sum, the components of a TTS relevant to our purpose are e (the interval representing the eventuality), S (point of speech and evaluation point), and S# (imaginary point of speech and evaluation point). A TTS is built up in a compositional way: each tense, event type, temporal adverbial, and temporal connective has its own characteristic contribution.7 For an individual sentence or set of sentences, the sum of these independent contributions plus the interactions between them results in the TTS of the complete temporal meaning of a (set of ) sentence(s). The tense of a sentence contributes to its TTS in terms of the order of points on the S-track, and both tense and event type may contribute to specifying which eventualities correspond to which points on the S-track. In order for an assertive sentence to be true in a world w, these ordering and correspondence relations must be compatible with the temporal ordering of eventualities and times in w, and the eventualities must have occurred at the times indicated by the correspondences shown in the TTS, i.e., the eventualities must be ‘evaluated’ at these time points.8
316 Causality and Tense—Two Temporal Structure Builders
(11aa) The car stopped. The engine broke down. (11bb) The car stopped. It was out of petrol. (11cc) The girl looked radiant. She was in love. Suppose that causal coherence relations, like the marked one in (11b), or the unmarked one in (11b), carry a presupposition, like the one in (12b): (11b) The car stopped because it was out of petrol. (12b) If a car is out of petrol, the car will stop.
(q because p) ( p / q)
For the speaker, this means ‘to take it for granted, or at least to act as if one takes it for granted, as background information—as common ground among the participants in the conversation’ (Stalnaker 2002: 701). Therefore, if the speaker did not suppose that (12b) is an appropriate regularity, he would not utter (11b). However, whether (12b) is an appropriate regularity, depends at least partly on the context. According to Stalnaker ‘what the context provides is the domain of possible worlds that propositions distinguish between . . . context is not just information that mediates between utterance and proposition; it is the material out of which propositions are constructed.’ (Stalnaker 1999: 156). Suppose mention is made in the context of a car which has both a petrol tank and a gas tank. In such a context (12b) is not an appropriate regularity—and (11b) is not acceptable either. Therefore, taking (12b) to be the presupposition for (11b) is a false simplification. We can only phrase the appropriate presupposition if we consider the context of the utterance. Mackie (1965) introduces a notion that gives us the required context dependence for the conditional relations underlying causal
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Lagerwerf & Oversteegen (1994) argue that the regularities in (12) behave as presuppositions for the sentences in (11). The reason for preferring this option to other options, like treating the regularities as inferences or implicatures, is that they function as prerequisites for the causally connected clauses. If one does not believe in a relation like the one expressed in (12a), one will not accept (11a) either. On the other hand, if one does not have the knowledge expressed in (12a), one can still accept (11a), but only if (12a) is accommodated, or added to one’s knowledge base. This strongly resembles the way presuppositions behave. Usually, presuppositions are ascribed to a certain lexical item or a certain construction. It is interesting to note that it is not the causal connective, however, that triggers the presupposition, but the causal coherence relation. The same presuppositions as in (11) should be attributed to the juxtaposed sentences below:
Leonoor Oversteegen 317
coherence relations. In Mackie’s view, for every causal relation, there is a corresponding underlying conditional relation. The cause in the causal relation then corresponds to a so-called INUS condition (with IN for INsufficient and US for UnneceSsary): (13)
A is an INUS condition of a result P if and only if, for some X and for some Y, (AX or Y) is a necessary and sufficient condition of P, but A is not a sufficient condition of P and X is not a sufficient condition of P.
(12b*)
If a car runs on petrol and on gas (Z) and it is out of petrol (A), the car will stop (P). (12b#) If a car runs on petrol only (X) and it is out of petrol (A), the car will stop (P). Now, suppose we have the additional information X that the car can only take one type of fuel. We have to interpret AX and P in Mackie’s definition for (12b) as in (12b#). If the context supplies us with the information X (the car runs on petrol only), the regularity in (12b) applies. If the context supplies us with the information Z (the car runs on petrol and on gas), it does not. Henceforth, I will refer to a conditional like (12b#) in relation to a causal relation like (11b), as ‘realization of the INUS scheme’. The utility of the INUS condition consists in its being an insufficient condition: only given certain additional information do we accept the regularity of which it is a part. This additional information comes with the context and with the speech situation. Consequently, the regularity may be acceptable in one situation or 9
In the remaining part of the paper, I will refer to the complete conditional (AX _ Y / P) as the INUS scheme. I will systematically neglect the disjunct Y: it is of no importance here to find the ‘entire’ set of causes in the causal relations (cf. Mackie 1965).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The antecedent A in definition (13) is made unnecessary by force of the disjunct ‘or Y’. The disjunct is added to ‘AX’ in order to make it clear that there may be alternative causes; AX is not considered to be a complete cause. Mackie defines the antecedent A as insufficient. That is to say, in combination with some additional information X, the antecedent may be sufficient, but in combination with alternative additional information, for example Z, it may not be sufficient, or even false.9 Applying definition (13) to the conditional in (12b), we can get the following situations. Z may be: the car has both a petrol tank and a gas tank. Therefore, if it is out of petrol, it can still switch to gas. The additional information makes the conclusion in (12b) invalid for such a car (cf. (12b*)).
318 Causality and Tense—Two Temporal Structure Builders context, but unacceptable in another situation or context.10 This implies that we do not need to assume an unlimited set of ‘causal laws’ (as in Lascarides and Asher, 1992, 1993). The conditionals underlying the causal relations may be formed ad hoc, incorporating occasional information about context and situation.11 I adopt Mackie’s view that an (underlying) INUS condition, as a part of an INUS scheme, is the essence of a causal relation. A complete realization of the INUS scheme is presupposed by a causal coherence relation, be it semantic or epistemic (as I shall argue in section 4.2).
The examples in (11), contain two occurrences of the simple past (see (8)) and a causal relation between two eventualities resulting, by force of the Causal Law in (1), in a temporal relation. In general, the precise temporal relation between two eventualities in simple past depends on the types of eventuality involved. If both eventualities are terminative (the real events), then the intervals representing them are ordered by precedence. If one or both are states (including Vendler’s activities), the intervals representing them may overlap. For details of the underlying mechanism determining the sequence of events and states, I refer to Hinrichs (1986), Webber (1988) and Oversteegen (1993). (14a) (15a)
Carl patted Joy on the head. Joy kissed Carl. Carl was in love with Joy. She was in love with him.
The anaphoric nature of tense warrants a default interpretation for (14a) and (15a): In (14a) the default is that Joy responded to Carl’s tender gesture with a kiss. But Webber (1988) argues that this default can easily be overruled. If the first clause had been Carl pushed Joy away, it would have been more natural to interpret a reversed sequence of events (cf. Lascarides & Asher 1991, 1993). In (15a), the default is that being in love was mutual, or at least coincided at a certain period. 10
Therefore, the relation between A and P in (12b) should be represented by the symbol > for defeasible implication (Lascarides & Asher, 1991, 1993, and Lascarides & Oberlander, 1993). The symbol indicates that the conditional relation functions in a non-monotonous logic (Asher & Morreau 1991). Henceforth, this symbol will be used. 11 In argumentation theory we find a concept strikingly similar to the INUS scheme. Gerritsen & Snoeck Henkemans (1990), working within the framework of Pragma-dialectic argumentation theory developed by Van Eemeren and Grootendorst (1982), describe a procedure for the reconstruction of the intended link between arguments and claims, as the pragmatic optimum. Both the INUS scheme and the pragmatic optimum have the form of a conditional, and both have to be reconstructed from two causally connected text parts amplified with textual and situational context information.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
3.3 The contribution of semantic causal relations to the representation structure
Leonoor Oversteegen 319
Given the anaphoric potential of tense, what contribution does an explicit causal relation make to the temporal structure? Compare the following sentences: (14a) Carl patted Joy on the head. Joy kissed Carl. (14b) Joy kissed Carl because Carl patted her on the head. (15a) Carl was in love with Joy. She was in love with him. (15b) Joy was in love with Carl because he was in love with her.
(11a) The car stopped because the engine broke down. (11#a)
(11b) The car stopped because it was out of petrol. (11#b)
(11c) The girl looked radiant because she was in love. (11#c)
The representations should be read as: (11#a): e2 < e1; (11#b): e1 e2, and e2 starts before e1; (11#c): e1 s e2, and e2 starts before e1. 12 This difference is due to the ‘indivisibility’ of events, as it has often been called in the aspectual literature. Consequently, in case of an event causing a state, I will assume that the event entirely precedes the state. However, assuming that only the start of the event precedes the start of the state, and, consequently, allowing the event and the state to overlap, would not change the analysis presented in the paper in any significant way.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The presence of because appears to exclude certain interpretations. In (14b) Joy’s kiss can only be seen as a response to Carls’ gesture. And in (15b), there is no question anymore who fell in love first. In other words: the start of the eventuality expressed in the main clause necessarily follows the start of the eventuality expressed in the becauseclause, in line with the Causal Law. In case of two events, as in (14b) and (11a), this generally results in two separate events immediately following each other, whereas two eventualities of which at least one is a state, as in (15b), (11b), and (11c), still have the possibility to overlap.12
320 Causality and Tense—Two Temporal Structure Builders
4 EPISTEMIC CAUSAL RELATIONS The paradoxical examples (5) and (6) are nothing more, of course, than instruments for refining our understanding of epistemic causal relation. They force us to distinguish the interacting meaning aspects. In this section, the components of a solution proposed in the sections on semantic causal relations (3.2 and 3.3) will be re-evaluated and adapted for epistemic causals (in 4.2 and 4.3). But previous to this, an extension of TTT for the representation of epistemic meaning is required. Section 4.1 presents this extension.
4.1 Epistemic modality in the representation structure Epistemic modality in general is described as speaker-oriented modality. It reflects the speaker’s ‘degree of commitment to the expressed proposition or the likelihood of a state of affairs’ (Nuyts 2001). For example, the future tense may express epistemic modality. The certainty of the claim that an eventuality takes place at some point on the E-track is relativized by the use of the future tense. The E-track reflects the type and order of eventualities asserted by the speaker. If the speaker feels obliged to relativize the truth-status of the eventuality, it would be incorrect to represent such an eventuality on the E-track without representing the reservation. The solution for epistemically modified expressions in TTT is adopted from Van Fraassen (1969). He introduces supervaluations to account for the expressed measure of certainty for a certain eventuality to take place. In TTT this idea is realized by a branching of the E-track. From the point in time at which the speaker sees two (or more) possible extensions of the E-track, or futures, it branches and,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
At this point, the three components claimed in section 3 to be the prerequisites for the representation of epistemic causal relations have been described. As a model for the representation of temporal information, TTT was introduced. An analysis of causal relations, making use of Mackie’s INUS condition, was provided. And, in conclusion, the meaning contribution of the semantic causal relation was defined. It appeared to consist of two different aspects. One aspect of the contribution is the determination of the precedence relation between the (starting points of) e-intervals in the TTS (this section). The other aspect is of a propositional nature: It concerns the accommodation of a presupposition reflecting a general regularity (section 3.2).
Leonoor Oversteegen 321
consequently, makes it possible for the speaker to express a lesser degree of certainty (Oversteegen 1989). (16)
present future
(17)
past future
4.2 Deriving epistemic causal coherence relations The epistemic causal relations can be treated in parallel to the semantic causals (Lagerwerf & Oversteegen 1994). A sentence like (4b) carries a presupposition as well, but in this case, the presupposition does not hold immediately between the two propositions connected. For (4b), it would be if a boy comes back for a girl, there is reason to believe that he loves her. More examples of epistemic causals, counterparts of the examples of semantic causal relations in (11), are given in (18). The INUS conditions associated with them are given in (19). (18a) (18b) (18c) (19a) (19b) (19c)
The engine (presumably) broke down, because the car stopped. (q because p) The car (presumably) was out of petrol, because it stopped. The girl (presumably) was in love, because she looked radiant. If a car stops, there is reason to believe that its engine has broken down. ( p > Bq) If a car stops, there is reason to believe that it is out of petrol. If a girl looks radiant, there is reason to believe that she is in love.
As was the case for semantic causal coherence relations, context plays an important role in ‘licensing’ the derivation of epistemic causal coherence relations. The conditions required to derive them are similar to those for semantic causals: we should be able to reconstruct a (contextually inspired) realization of the INUS scheme, since such a realization functions as a presupposition to the epistemic causal
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The TTSs do not distinguish between the E-track extensions: the speaker conveys the belief that the course of events from the evaluation point S (in the present) or S# (in the past) on, may well contain the relevant eventuality. Likewise, adverbials like perhaps or presumably indicate a belief held by the speaker instead of a simple assertion. Similarly, they cause a branching of the E-track.
322 Causality and Tense—Two Temporal Structure Builders
4.3 The contribution of epistemic causal relations to the representation structure I can now turn to the concrete proposal for representing temporal aspects of epistemic causal relations. It follows directly if we combine the TTS for a semantic causal relation, like (11#a), and our knowledge of the presupposition for epistemic causal relations, as in (20). (11a) The car stopped (e1) because the engine broke down (e2). (11#a)
(20)
If a car stops, there is reason to believe that its engine has broken down (p > Bq)
The modification of (11#a), required for an epistemic causal relation, can be read from the presupposition: it is not the eventuality e1 which stands in a causal relation to e2, but the belief state that e1 occurs 13 An epistemic causal does not necessarily reflect a reasoning from result to cause. An epistemiccausally connected sequence of sentences like: There must be puddles on the streets. It has been raining for hours, expresses a reasoning from a cause to an (uncertain) result (cf. Noordman & De Blijzer 2000).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
relation. The difference between semantic and epistemic causals consists in the status of the consequence in the INUS scheme. In case of a semantic causal relation, it is an assertive proposition, whereas it is a belief in case of an epistemic causal relation. Section 5.2 gives examples of the derivation of epistemic causals. It would be convenient if epistemic causal relations could easily be recognized. That is, if the belief clauses were marked as such, for example, by the use of a verb expressing a cognitive state. Sweetser argues that epistemically connected pairs in written English can be recognized by the use of a comma—in Dutch, even this signal is not necessarily present. The cases I am concerned with here can be recognized by reconstructing the conditional regularity underlying the causal relation. In the case of an epistemically connected pair of clauses, the causal relation is generally a reasoning from result to cause, or, in other words, an abduction (Hobbs et al. 1993; Lagerwerf & Oversteegen 1994).13 However, in section 3.2 it was argued that semantic causal relations need not be linguistically marked either, and, consequently, must often be derived.
Leonoor Oversteegen 323
(at some time). Consequently: e2 < Be1 (but see note 12). This belief state Be1 is reported in the simple past and it is this belief state, not the eventuality which it refers to, that is evaluated at the S# introduced by the simple past. This is reflected in (20#a). (20a)
The engine (presumably) broke down (e1), because the car stopped (e2).
(20#a)
14
Introducing a separate realm for representing attitudinal or perspectivized meaning has been proposed in other frameworks as well, for example, in DRT as ‘articulated DRSs’ (Kamp 1990), in SDRT (Asher, 1993), and in Mental Space theory (Fauconnier, 1994; Fauconnier & Sweetser, 1996).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The epistemic causal relation tells us that e2 precedes Be1. What does (20a) tell us about the relation between e2 and e1? Tense does not give us further indications about the location of e1: both clauses are in the simple past tense. However, if we infer, on the grounds of real-world knowledge, a causal relation between the engine breaking down and the car coming to a halt, we do have information about the temporal relation between those two eventualities. It is well known from the literature that evidence relations reason from effect to cause, and, consequently, it is uncontroversial to apply the Causal Law to e1 and e2 (cf. Hobbs 1993). Given that the Causal Law says that e1 should be located before e2, where should e1 be located? Not on the E-track side by side with the eventualities the speaker claims to occur at some time. In the epistemic case, the speaker believes that e1 occurred—he does not assert it. Kamp (1990: 46) says the following about beliefs like these: ‘Its identity as a belief, i.e. as an attitude with that particular mode, is given as an inseparable aspect of the attitude itself.’ Therefore, the content of the belief, in this case, the occurrence of an eventuality, should be represented separately. In section 4.1 it was argued how this can be done in TTT.14 Therefore, in a TTS, the eventualities claimed (by the speaker) to belong to the real world are represented in a different realm than the eventualities occurring in his beliefs. A TTS contains several branches if and when the speaker conveys less certainty about the factual state of an eventuality. An eventuality he believes to occur or have occurred is located on a different branch than the asserted events are. In such a TTS, the branching E-tracks share the same time structure. Separate
324 Causality and Tense—Two Temporal Structure Builders parts of a discourse representation ‘may share the same discourse referents’ (Kamp, 1990: 46). Points in time are such discourse referents. In (20#) all individual contributions to the TTS of (20a) are presented at once—I will illustrate the way such a TTS comes about step by step in section 5. The TTS should be interpreted as telling us: 1. That all eventualities are located in the past, 2. That the full stop of the car (e2) precedes the belief that the engine broke down (Be1), 3. The engine breaking down (e1) is believed to be the cause of the full stop of the car (e2), and as such precedes this stop (e1 < e2).
(20#)
At this point, the specific nature of an epistemic causal relation has been described in terms of TTT. The contribution to the TTS made by tense, by the causal relation between an eventuality and a belief about another eventuality, and by the relation between the two eventualities, have been specified. In short, the temporal/causal aspects of an epistemic causal coherence relation are representable in TTT.15 Returning to the problem which arises when tense and causality seem to provide paradoxical information in epistemic causal sequences, we are now in the position to put the analysis to the test: does it provide a way out of the paradox? 15
Temporal information in causal coherence relations is implicit. Freksa & Barkowski (1999) argue that a spatial representation, which a TTS is, is especially suitable when: 1. The knowledge to be represented consists of a large number of interrelated relations and many of these relations can be captured simultaneously by the arrangement of a comparatively small number of objects (p. 202); 2. Implicit knowledge is represented (p. 205); and 3. It is not necessary to make all the knowledge represented explicit: inferences can be made (p. 205).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This last piece of information forces the E-track to start branching before e2. Whereas it may be clear that e1 cannot be located on the same Ebranch as e2 and Be1, one might argue that e2 and Be1 could be represented at both branches: assertions (of beliefs) may be argued to express beliefs as well. I have no strong opinion on this matter. Several counterexamples to the claim that every assertion expresses a belief have been given in the literature (Siebel 2003).
Leonoor Oversteegen 325
5 SOLVING THE PARADOXES The components of my explanation of the clashes of tense and causality in an argumentative piece of discourse have been discussed separately. Now, I bring them together in an explanation of the paradoxical cases (5) and (6). First, I demonstrate how the concept of INUS conditions guides (or ‘licences’) the derivation of the epistemic causal relation between the juxtaposed sentences in (5) and (6). Then, I give the TTS representation of these sentences in such a way, that both tense and causal coherence relation have their independent contributions to the TTS without causing incompatibility.
As may be expected, the analysis of causal relations by means of INUS conditions shows that it is not the presumed hitting itself that is caused in (5), repeated below for convenience. If we take the A in (13), the INUS condition, it symbolizes she immediately came out again. The speaker would not want us to accommodate a presupposition to the effect that ‘if she leaves a building at a certain speed or within a certain time, she will be hit’ (5)
Eve’s husband (presumably) had hit her (e1) because she immediately came out again (e2)
Realizing an INUS scheme for (5) implies demonstrating that a complement X for the INUS condition A can be constructed in such a way that the conclusion P is a reasonable consequence of the INUS condition and the complement together. Let us ignore other possible causes represented by Y, and focus on the sufficient condition AX. What contextually determined X implies, in conjunction with A, that the belief that Eve’s husband had hit her is justified? The situation may be that a marital fight has taken place and that one of the partners, the wife, has walked out. If the speaker needs no more evidence for his conclusion, then probably the husband has been violent before and this fact belongs to the common ground which the speaker is assuming. However, it is also possible that a sudden sound is mentioned in the context that is interpreted as a blow by the speaker, at the moment he sees the wife coming out quickly. In conclusion, as a complement X for the INUS condition A in (5), either (5b) or (5c) would be possible. The point is that a legitimate choice for one of them, or yet another possibility, can be made in context only. (5b)
A: Eve immediately came out again X: Eve’s husband was mad at Eve & Eve’s husband is a violent man
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
5.1 Licensing epistemic causal relations in the examples
326 Causality and Tense—Two Temporal Structure Builders (5c) A: Eve immediately came out again X: Eve’s husband was mad at Eve & a sudden, loud sound had just been heard
(6) Zeker een derde deel van de kamer werd door het zwarte instrument in beslag genomen (e1). Gek was ze geweest, gek dat ze zoveel geld had uitgegeven voor dat ding, dat apparaat, dat monster (e2). At least one third of the room was occupied by the black instrument. She had been crazy, crazy to have spent so much money on that thing, that machine, that monster. Again, we ignore other possible causes and focus on AX. What could X be? Again, the unknown X represents contextually supplied information which, together with A, should lead to P, the conclusion that it had been a mistake to spend so much money on the grand piano. The context does give us the relevant information: it appears that the subject used to play professionally but that she stopped playing altogether at some time before the period of time referred to in (6). After a few dramatic events, the subject has not been able to play the piano anymore. This contextual information makes it possible to construct an acceptable complement X for the INUS condition: (6) A: at least one third of the room was occupied by the black instrument X: she does not play the piano anymore and the instrument is a constant reminder of a broken career In this case, we do not have to construct the complement X out of the blue. The context provides us with the information required to complete the realization of the INUS scheme. Consequently, the context dependence Stalnaker (1999) refers to (see section 3.3 of this paper) is realized in the approach to epistemic causal relations presented. We accept this type of argumentative relations only if we are able to construct an
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In terms of TTT, we now have established the relation between e1 (¼ A) and Be2 (¼ P). We need to do the same for (6), repeated below for convenience, by demonstrating that a complement X for the INUS condition A, in this case at least one third of the room was occupied by the black instrument, can be constructed in such a way that the conclusion P is a reasonable consequence of A and X together. Again, licensing a semantic causal relation would not work: the speaker of (6) would not expect us to have or to accommodate the presupposition that big black instruments make one crazy.
Leonoor Oversteegen 327
appropriate conditional relation, a realization of the INUS scheme, based on an INUS condition and completed by contextual, linguistic or extralinguistic, information. And, again, in terms of TTT, we have argued for the relation between e2 (¼ A) and Be1 (¼ P).
5.2 The temporal paradox
(5)
Eve’s husband (presumably) had hit her because she immediately came out again (5a) contribution of PPF (5b) contribution of sP to TTS (clause 1): to TTS (clause 2):
In section 1 and section 3.3, the anaphoric nature of tense was mentioned. It results, as a default, in the location of an eventuality in the pluperfect before the location of an eventuality in the past (see table 1). This result was first obtained by Reichenbach’s principle: the permanence of the reference point. In terms of TTT, it means that combining the TTS contributions of the simple past in (5a) and the pluperfect in (5b), the evaluation points S#1 are mapped onto each other. The resulting TTS contribution of the combination of tenses is given in (5c). (5c)
The contribution of the epistemic causal relation to a TTS is complex, as was argued in section 4.3. It introduces a belief state Be1, which is interpreted as being caused by the eventuality e2. Consequently, the e2 must precede Be1 (see also (20#)). This part of its contribution is given in (5d). Furthermore, it tells us that the eventuality e1, the content of
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
For the paradoxical cases (5) and (6), the temporal representation follows the lines of the epistemic causals discussed in section 4.2. In these cases, however, the temporal relation between e1 and e2 is explicitly defined by the tenses used. The paradox is caused by the seemingly controversial information contributed by the tenses and by the causal coherence relation. In (5a) and (5b), the TTS contributions of the tenses of both sentences are given.
328 Causality and Tense—Two Temporal Structure Builders the belief state Be1, is not asserted by the speaker but believed to be true. Therefore, e1 can not be located on the same branch of the E-track as e2. Note that the relation between e1 and e2 is not defined by the epistemic causal relation. The contribution in (5e) should be interpreted as leaving the temporal relation between e1 and e2 undetermined. (5d)
contribution 1 of epistemic (5e) causal relation to TTS:
contribution 2 of epistemic causal relation to TTS:
(5#)
total TTS:
There are several independent observations confirming the adequacy of (5#) as a representation for (5). First, the interpretation of the belief state at the imaginary speech point S#1 should be: at S#1, the speaker could have said truthfully I believe that Eve’s husband has hit her. Note that, indeed, the S#1e1 configuration can be mapped to the contribution of a present perfect to a TTS (see (9)). Secondly, substituting the pluperfect in the first clause of (5) by a simple past should result in ambiguity. Since the relative position of e1 and e2 is due to the contribution of the tense combination, it would be unclear, in case of two simple past tenses, what sequence of events is intended. (21)
Eve’s husband (presumably) hit her because she immediately came out again.
Indeed, there are two interpretations for (21): the first one is similar to the interpretation of (5). The second one presents the eventuality in the first clause as a result (or reaction) of the eventuality in the second
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We obtain the final TTS for (5) by combining (5c), (5d) and (5e)—and this synthesis is not paradoxical anymore (as were the columns 2 and 4 in Table 1). Besides, the vagueness in the relation between e1 and e2 disappears as a result of the contribution of (5c): A precedence relation is defined
Leonoor Oversteegen 329
(6)
Zeker een derde deel van de kamer werd door het zwarte instrument in beslag genomen. Gek was ze geweest, gek dat ze zoveel geld had uitgegeven voor dat ding, dat apparaat, dat monster.
At least one third of the room was occupied by the black instrument. She had been crazy, crazy to have spent so much money on that thing, that machine, that monster. (6a) contribution of sP to TTS (6b) contribution of PPF to TTS (clause 1): (clause 2):
As in (5c), the combination of simple past and pluperfect determines (by default) the temporal relation between e1 and e2 (see (6c)). (6c)
Naturally, the contribution of the epistemic causal coherence relation in (6) is similar to the one in (5). The only difference consists in the 16
The relation between Be1 and e1 remains the same.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
clause. In other words, it implies the inversed precedence relation between e1 and e2.16 Thirdly, as was argued in section 4.2, epistemic causal relations usually reason from result to cause. This implies that an inversed semantic causal relation can be found ‘backing up’ the epistemic causal relation (cf. the inversion of the clauses of the semantic causal relations in (11) in the epistemic causal relations in (18)). For (5), this semantic causal relation would be: Eve immediately came out again because her husband (had) hit her. This semantic causal unambiguously places the hitting (e1) before the coming out (e2). Consequently, the relation of logical compatibility between the two causal sequences is reflected in the TTS for (5), the epistemic counterpart. The sequential relation between the eventualities as defined by the semantic counterpart is a subpart of (5#), the TTS for (5). The second paradoxical case can be treated in a similar way, as will be demonstrated now.
330 Causality and Tense—Two Temporal Structure Builders relation between e1 and Be2. In this case, e1 is a state as well, and states may overlap (see (11#c)). And for (6), this should be the case: the belief that she had been crazy to spend money on the piano does not put an end to the instrument crowding the room. Again, the two parts of the contribution are presented separately (see (6d) and (6e)). And again, in (6e), the temporal relation between e1 and e2 is unspecified. contribution 1 of epistemic (6e) causal relation to TTS:
(6#)
total TTs:
contribution 2 of epistemic causal relation to TTS:
The TTSs (5#) and (6#) represent the paradoxical information reconstructed in Table 1 in a coherent way. These TTSs do justice to our intuition about the type of causal relation employed in the sentences (5) and (6), but they also incorporate the ‘hard’ semantic temporal information supplied by tense and the Causal Law stating that causes precede effects.17 6 THE INTENTION OF THE SPEAKER AND UNDERSTANDING VERSUS ADOPTING In the logic-semantic literature, overt belief sentences have been investigated extensively (e.g. by Kamp1990, Singh & Asher 1993, Stalnaker 1999). There are similarities between overt belief sentences and the type of sentences discussed here. In both cases, the commitment of a subject to a propositional content is expressed and, in both cases, there is no expectation that the content of the proposition is considered to be uncontroversial. However, only in the ‘covert’ belief sentences discussed here, is the believer necessarily identical to the speaker. This makes the intention of the speaker interesting. In uttering the sentence, 17 The rule CCT, presented here as (3) in section 1, tells us that the second clause of (6) must function as an explanation for the first clause. Being crazy when she bought the piano would be presented as an explanation for her creating the undesirable situation in the room described in the first clause.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(6d)
Leonoor Oversteegen 331
To accept a proposition is to treat it as true for some reason. One ignores, at least temporarily, and perhaps in a limited context, the possibility that it is false. Belief is the most basic acceptance concept: the simplest reason to treat a proposition as true is to believe that it is true. But there may be various reasons to ignore the possible situations in which some proposition is false even when one realizes that one of those possible situations may be the actual one (. . .), one may grant something for the purpose of an argument. In cases where communication is facilitated by accepting propositions that one or the other of the participants don’t believe, we need a notion of common ground based on a notion of acceptance that may diverge from belief. (Stalnaker 2002: 716) This notion is described as the belief about what is accepted. Thus a presupposition still reflects the speaker’s belief about the common ground—but not necessarily his belief about common belief. The branching TTS form enables us to distinguish between understanding an epistemic causal relation (or argumentation) and
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
the speaker conveys a belief which is the conclusion of an argument. The argument is supposed by the speaker to be common ground because one part is explicitly present in the discourse and the other part is presupposed. Presuppositions are considered to be speakers’ beliefs about the common ground (Stalnaker 2002). However, the belief sentence presenting the conclusion of the argument is not phrased as a personal attitude. As a consequence, an epistemic causal relation is often interpreted as an attempt to make the hearer adopt the conclusion. In other words, the speaker has the intention of turning his personal belief into common belief, shared with the hearer. The hearer will be prepared to share this belief only if he can reconstruct the missing argument: the presupposition built from the INUS condition, its (contextually determined) complement(s), and the conclusion. However, reconstruction of the presupposition is not a sufficient condition for sharing the belief of the speaker. Understanding an argument is not equal to believing it to be true. Stalnaker (2002) makes the same distinction. Treating presuppositions as speakers’ beliefs about common ground, he argues that common ground equals common beliefs in many cases. However, it is also possible that a speaker treats a proposition as common ground, i.e. as a presupposition, whereas he does not believe the proposition to be believed by the recipient(s). In these cases, common ground is not identical to common beliefs.
332 Causality and Tense—Two Temporal Structure Builders
(5$)
(6$)
The collapsing of tracks can be interpreted as a transfer of eventualities from one domain (of expressed beliefs) to another one (of assertions). This change of state becomes more plausible if one considers chain arguments. An example of a chain argument is given in (22). (22)
It is going to rain (1). Everybody is carrying an umbrella (2). We should postpone the match (3).
The speaker starts with a belief (1) and supports it with an assertion (2). The final claim or belief (3) is based, in turn, on the belief (1), which may be accepted by the reader as a fact because of the support in (2). This is a minimal example of a chain argument: a clause that is
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
adopting it (or being persuaded). If we merely understand the relation between event and belief state in an epistemic causal relation, we are able to reconstruct the presupposition, but we do not add it to our knowledge base, or accommodate it. The branching TTS structure remains intact. The representation structures (5#) and (6#) model this situation. Adopting the epistemic causal relation, however, implies either believing already or accommodating the regularity expressed in the presupposition. In that case, the branching of the E-track can be collapsed into a non-branching TTS. The results of (5) and (6) are given in (5$) and (6$), respectively. The eventualities Be1 (in (5#)) and Be2 (in (6#)) have disappeared in (5$) and (6$). The presence of e1 and e2 at an earlier time makes them redundant and strange. As a result of the procedure, the causal link between e2 and believing e1 (in ((5)) or between e1 and believing e2 (in (6)) is present in the accommodated presupposition only.
Leonoor Oversteegen 333
originally presented as a conclusion functions as an argument for another conclusion. We get the relations presented in (22a). (22a)
e2 causes Be1 so e2 < Be1 e1 causes Be3 so e1 < Be3
7 CONCLUSION AND DISCUSSION In this paper, I argued that we can gain insight into the temporal aspects of argumentative parts of discourse. On the one hand, this is done by analysing the properties of epistemic causal relations in comparison with semantic causal relations and, on the other hand, by recognizing the specific nature of the related eventualities in epistemic causal relations. The analysis consists of distinguishing ‘hard’ semantic aspects of these causal relations (i.e. the precedence relation between cause and result, the existence of a presupposition) and ‘soft’, defeasible aspects of these relations (e.g., the precise content of the INUS scheme), and explaining their relation. Moreover, these aspects of causality are combined in a straightforward way with the ‘hard’ temporal information contributed by the tenses used in the causally connected sentences. Originally, TTS representations were meant to model assertive sentences. The extended TTS appears to be capable of representing a combination of interacting properties: causal, temporal, and epistemic properties. The status and function of presupposition in the analysis presented may seem problematic. The presupposition, in the form of a conditional relation (the INUS scheme), is also assumed necessary for obtaining the correct (epistemic) causal relation in the absence of a presupposition trigger (like because)! However, it is argued that bridging, the resolution
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A TTS for this text would contain branching E-tracks: e2 is on the same branch as Be1, but e1 would be on another branch. Similarly, e1 would be on the same branch as Be3, but e3 would be on another branch again. This increasing branching may seem artificial, but it reflects the growing ‘instability’ of the claims: the more ‘deeply-branched’ eventualities are represented, the further they are from the asserted facts. When is it possible to use such chain arguments with a reasonable chance of success? According to the approach advocated in this paper, the epistemic causal relation is accepted if an adequate realization of the INUS scheme, functioning as a presupposition, is accommodated. If this happens and the belief state concerning an event is adopted by the hearer, the ‘e-claim’-branch can be collapsed onto the E-track. The difference in status, fact or claim, disappears for the hearer.
334 Causality and Tense—Two Temporal Structure Builders
18 Asher & Lascarides (1998) argue that ‘Full understanding takes the full [discourse representation] structure and integrates it with the beliefs of the interpreter, and as such comes after discourse interpretation’ (Asher & Lascaridis 1998: 85). In fact, this order is suggested by the analysis presented in this paper as well.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
of anaphora, and the interpretation of presuppositions are ‘a byproduct of computing the discourse structure of a discourse, which we view as a necessary precondition for discourse interpretation . . .’ (Asher & Lascarides 1998: 85). If presuppositions can be envisaged as a byproduct of computing the discourse structure, which is, in turn, the precondition for discourse interpretation, then why would we need a trigger for them apart from the general demand that the discourse must be coherent, and, consequently, coherence relations, as constituents of discourse structure, must be computable? Taking the INUS scheme to be an inference instead of a presupposition for a causal relation, would be neglecting its essential role in causality (see section 3.2). The paper also addressed different effects of epistemic causal coherence relations on hearers. Whereas the effect in the epistemic causal relation, the conclusion of the argument, is not supposed by the speaker to be common ground already, he does treat the presupposition for the epistemic causal relation as common ground. In the words of Stalnaker (2002: 716), the speaker believes that it is common belief that the content of the presupposition is accepted. If it really is, the hearer may adopt the epistemic causal relation. In this case, we may say that he adopts the belief expressed. Alternatively, he may simply understand what the speaker means, without adopting his belief. Focusing on the hearer, I argued that we must have two different representation structures available: one for the hearer who has properly understood the epistemic causal relation (or the argument), and the other for the hearer who believes the causal link to hold and is convinced by the argument. The question whether they both represent the meaning of the sentence pairs should probably be answered in the negative.18 Even if it is the intention of the speaker to convince the hearer of the truth of the conclusion, should this intended effect be seen as part of the meaning or is it a consequence of the linguistic meaning of the sentences, i.e. its perlocution? Leaving the answer to this question open to further research, I may have provided a little more transparency in the relations between epistemic causal relations, temporal meaning and information status. And as a profitable consequence, a general difference between the use of causality in narrative texts and in persuasive texts.
Leonoor Oversteegen 335
Acknowledgements This paper is dedicated to Henk Verkuyl, my first tutor in semantics. The comments made by an anonymous reviewer were of great benefit to the paper. LEONOOR OVERSTEEGEN
Received: 27.11.02 Final version received: 23.12.04 Advance Access publication: 20.06.05
REFERENCES Fraassen, B. C. van (1969) ‘Presuppositions, supervaluation and free logic’. In K. Lambert (ed.), The Logical Way of Doing Things. Yale. New Haven, 67–91. Freksa, Ch. & Barkowski T. (1999) ‘On the duality and on the integration of propositional and spatial representations’. In Rickheit and Habel (eds), Mental Models in Discourse Processing and Reasoning. Elsevier Science BV, 195–212. Gerritsen, S. & Snoeck Henkemans, A. F. (1990) ‘Verzwegen argumenten als probleem bij het schrijven van betogende teksten’. Tijdschrift voor Taalbeheersing 12-4:299–308. Hinrichs, E. (1986) ‘Temporal anaphora in discourses of English’. Linguistics and Philosophy 9:63–82. Hobbs, J. R., Stickel, M. E., Appelt, D. E. & Martin, P. (1993) ‘Interpretation as abduction’. Artificial Intelligence 63:69–142. Jaszczolt, K. M. (2003) Book review: ‘Epistemic modality, language and conceptualization: A cognitive-pragmatic perspective’. Journal of Pragmatics 35: 657–663. Kamp, H. (1981) ‘A theory of truth and semantic representation’. In J. A. G. Groenendijk, T. M. V. Janssen & M. B. J. Stokhof (eds), Formal Methods in the Study of Language, 136:277–322. Mathematical Centre Tracts. Amsterdam.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Asher, N. (1993) Reference to Abstract Objects in Discourse. Kluwer Academic Publishers. Dordrecht/ Boston/London. Asher, N. & Lascarides, A. (1998) ‘Bridging’. Journal of Semantics 15:83–113. Asher, N. & Morreau, M. P. (1991) ‘Commonsense entailment: A modal theory of nonmonotonic reasoning’. In Proceedings of the 12th International Joint Conference on Artificial Intelligence. Sydney, Australia. Bach, E. (1986) ‘The algebra of events’. Linguistics and Philosophy 9:5–16. Caenepeel, M. (1995) ‘Aspect and text structure’. Linguistics 33:213–254. Dijk, T. van (1979) ‘Pragmatic connectives’. Journal of Pragmatics 3:447–456. Eemeren, F. H. van & Grootendorst, R. (1982) Regels voor redelijke discussies. Een bijdrage tot de theoretische analyse van argumentatie ter oplossing van geschillen. Ph.D., Universiteit van Amsterdam. Eemeren, F. H. van & Grootendorst, R. (1992) Argumentation, Communication and Fallacies. Lawrence Erlbaum. Hillsdale, N.J. Fauconnier, G. (1994) Mental Spaces. Cambridge University Press. Cambridge. Fauconnier, G. & Sweetser, E. (eds) (1996) Spaces, Worlds and Grammars. University of Chicago Press. Chicago.
336 Causality and Tense—Two Temporal Structure Builders Proceedings of the 4th International Workshop on Semantics of Time, Space and Movement and Spatio-Temporal Reasoning, University of Toulouse, 250–259. Lascarides, A. & Asher, N. (1993) ‘Temporal interpretation, discourse relations and commonsense entailment’. Linguistics and Philosophy 16-5:437–494. Lascarides, A. & Oberlander, J. (1993) ‘Temporal coherence and defeasible knowledge’. Theoretical Linguistics, 19-1:1–37. Mackie, J. (1965) ‘Causes and conditions’. American Philosophical Quarterly 2-4:245–255/261–264. Reprinted in Sosa, E. (1975) Causation and Conditionals. Oxford University Press. Mann, W. C. & Thompson, S. A. (1987) Rhetorical Structure Theory: A Theory of Text Organization. Technical report RS-87-190, USC/Information Sciences Institute. McTaggart, J. M. E. (1927) ‘The unreality of time’. Mind 17:457–481. Moens, M. & Steedman, M. (1988) ‘Temporal ontology and temporal reference’. Computational Linguistics, 14:15–28. Moeschler, J. (1990) ‘Topoi and inferences’. Journal of Integrated Study of Artificial Intelligence 3:249–264. Noordman, L. G. M. & Blijzer, F. De (2000) ‘On the processing of causal relations’. In E. Couper-Kuhlen & B. Kortmann (eds), Cause, Condition, Concession and Contrast: Cognitive and Discourse Perspectives. Mouton de Gruyter. Berlin/New York, 35–56. Noordman, L. G. M. & Vonk, W. (1992) ‘Reader’s knowledge and the control of inferences in reading’. Language and Cognitive Processes 7(3/4):373–391. Nuyts, J. (2001) Epistemic Modality, Language and Conceptualization: A Cognitive-Pragmatic perspective. Human
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Kamp, H. (1990) ‘Prolegomena to a structural account of belief and other attitudes’. In Anderson, C. A. & J. Owens (eds), Propositional Attitudes: The Role of Content in Logic, Language and Mind. CSLI. Kamp, H. & Rohrer, Chr. (1983) ‘Tense in texts’. In R. Ba¨uerle, C. Schwarze & A. von Stechow (eds), Meaning, Use and Interpretation of Language. De Gruyter. Berlin/New York, 250–269 Kehler, A. S. (1994) Common topics and coherent situations: Interpreting ellipsis in the context of discourse inference. In Proceedings of the 32nd Meeting of the Association of Computational Linguistics. ACL94. Las Cruces, New Mexico. Knott, A. & Mellish, Ch. (1996) ‘A feature-based account of the relations signalled by sentence and clause connectives’. Language and Speech 39 (2–3):143–183. Kutschera, F. von (1993) ‘Causation’. Journal of Philosophical Logic 22: 563–588. Lagerwerf, L. & Oversteegen, E. (1994) ‘Inferring coherence relations through presuppositions’. In P. Bosch & R. van der Sandt (eds), Focus and Natural Language Processing, Vol. 3. Proceedings of a Conference in celebration of the 10th anniversary of the Journal of Semantics. IBM Working Papers of the Institute for Logic and Linguistics. Heidelberg, 503–512. Lakoff, R.T. (1971) ‘Ifs, ands and buts about conjunction’. In J. Fillmore & D. Langendoen (eds), Studies in Linguistic Semantics. Holt, Rinehart & Winston. New York, 114–149. Lascarides, A. & Asher, N. (1991) ‘Discourse relations and defeasible knowledge’. Proceedings of the 29th Annual Meeting of the Association of Computational Linguistics, 55–62. Lascarides, A. & Asher, N. (1992) ‘The pluperfect in narrative discourse’.
Leonoor Oversteegen 337 Sidiropoulou, M. (1992) ‘On the connective ‘‘although’’ ’. Journal of Pragmatics 17:201–221. Siebel, M. (2003) ‘Illocutioniary acts and attitude expression’. Linguistics and Philosophy 26:351–366. Singh, M. P. & Asher, N. (1993). ‘A logic of intentions and beliefs’. Journal of Philosophical Logic 22(5):513–544. Stalnaker, R. C. (1999) Context and Content; Essays on Intentionality in Speech and Thought. Oxford Cognitive Science series. Oxford University Press. Oxford. Stalnaker, R. C. (2002) ‘Common ground’. Linguistics and Philosophy 25:701–721. Sweetser, E. (1990). From Etymology to pragmatics, metaphorical and cultural aspects of semantic structure. Cambridge University Press. Cambridge, U.K. Vendler, Z. (1957) ‘Verbs and times’. The Philosophical Review 66:143–160. Webber, B. L. (1988) ‘Tense as a discourse anaphor’. Computational Linguistics 14:61–73. Arlington, Virginia.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Cognitive Processing series, vol. 5. John Benjamins. Amsterdam. Oversteegen, E. (1988) ‘Temporal adverbials in the two track theory of time’. In V. Ehrich, & H. Vater (eds), Temporalsemantik. Max Niemeyer Verlag. Tu¨bingen. Oversteegen, E. (1989) Tracking Time: A Proposal for the Representation of Temporal Expressions in Dutch. Dissertation. University of Utrecht. Oversteegen, E. (1993) Dutch Temporal Connectives. Technical report. European Union Basic Research Project EP6665, Tilburg, Holland. Oversteegen, E. & Bekker, B. (2002) ‘Computing perspective: The pluperfect in Dutch’. Linguistics 40-1: 111–161. Redeker, G. (1990) ‘Ideational and pragmatic markers of discourse structure’. Journal of Pragmatics 14:367–381. Reichenbach, H. (1947) Elements of Symbolic Logic. Macmillan. London. Sanders, T., Spooren, W. & Noordman, L. (1992) ‘Toward a taxonomy of discourse relations’. Discourse Processes 15:1–35.
