Adversative Connectors 4y3q6t

Erschienen in: E. Couper-Kuhlen & B. Kortmann (eds) Cause – Condition – Concession – Contrast. Cognitive and Discourse Perspectives. (= Topics in English Linguistics 33). 235 – 256. Berlin · New York: Mouton de Gruyter 2000

Adversative Connectors on Distinct Levels of Discourse: A Re-examination of Eve Sweetser's Three-level Approach* Ewald Lang While basically acknowledging the discernability of three levels on which a sentence is interpreted (content, epistemic and speech act level), the paper tackles Sweetser's claim that the correct interpretation of interclausal connections depends not on form, but on a pragmatically motivated choice. I argue for a division of labour between grammar and pragmatics proper by (i) showing that there are more level-specific structural differences at issue; (ii) proving the notion »epistemic« to be feasible only if statements, assumptions, and inferences are kept distinct; (iii) providing an for the unavailability of content-level interpretations for adversative conjunctions.

1.

Introduction

In Chapter 4 of her exciting book From Etymology to Pragmatics (1990), Eve Sweetser offers a fresh view on the notorious diversity of interpretations we observe with causal and adversative connectors. She rejects the view that connectors like because, although or and and but are lexically polysemous in the classical sense and suggests instead to explain the extremely varied uses of these connectors in of »pragmatic ambiguity«. Drawing on Sweetser (1990:76-78) the main tenets of her approach are summarized in (I)-(III): (I)

One meaning, several uses In polysemy, a morpheme has several related semantic values; in pragmatic ambiguity, a single semantics is pragmatically applied in different ways according to pragmatic context. (1990:76)

* This paper is dedicated to Ekkehard König on the occasion of his 60 th birthday! I wish to thank the editors for useful comments on an earlier draft of this paper and Chris Wilder for improving my English.

2

(II)

Ewald Lang

Three levels of interclausal connection (»conjunction«1) Conjunction may be interpreted as applying in one of (at least) three domains [where] the choice of a »correct« interpretation depends not on form, but on a pragmatically motivated choice between viewing the coned clauses as representing content units, logical entities, or speech acts. [emphasis mine – EL]

The domains of interpretation are illustrated by clear-cut examples such as (1a-c) below, all of which have in common that the second clause asserts a proposition in the form of a statement. (1)a. John came back because he loved her. [content or fact level] b. John loved her, because he came back. [epistemic level] c. What are you doing tonight, because there's a good movie on. [speech act level] The levels are distinguished depending on what the clauses are taken to represent and hence between what sort of entities the causal relation is taken to obtain. Thus, in (1a) it is two content units between which a factual causal relation is asserted to obtain. More exactly: the content units are two propositions whose couchingin declarative sentences enables them to render assertable statements. The linking pattern for a causal relation on content level thus amounts to STATEMENT because STATEMENT. In (1b), the first clause not asserted as a statement but only as an assumption which is inferred from what is stated as fact in the second clause. The causal connector marks the second clause as sufficient evidence justifying the assumption rendered by the first clause. The linking pattern on the epistemic level is ASSUMPTION, because STATEMENT (EVIDENCE). The hypothetical status of what is asserted by the first conjunct can be made explicit by embedding it under modals – cf. (1b-i, ii) or by rephrasing it as an inference – cf. (1b-iii): (1) b-i. John must have loved her, because he came back. (1b-ii) John undoubtedly loved her, because he came back. 1

Note that Sweetser uses the term »conjunction« for the syntactic operation of clause-linking without distinguishing coordination and subordination. In order to avoid confusion with the word class »conjunctions« I will call the latter »connectors« (subdivided into »coordinators« and »subordinators«) and the constituents they link »conjuncts«; otherwise I will follow Sweetser's terminology.

Adversative connectors on distinct levels of discourse

3

(1b-iii) From the fact that he came back, I conclude that John loved her. In (1c), the first clause serves the purpose of a speech act of asking whose performance is being justified by the statement rendered by the second. The linking pattern thus is SPEECH ACT, because STATEMENT (JUSTIFICATION). The because-clause does not form part of the question but is linked instead to an understood assertion which, if explicitly inserted, may read like this: (1c-i) What are you doing tonight ? I am asking that because there's a good movie on. In the course of presenting a variety of cases that are less clear-cut than the ones in (1a-c) above, Sweetser does of course mention disambiguating features that may (but need not) be present in the grammatical structure of the conjunction. For example, she hints at the role of intonation: commaless intonation is linked with the content reading in (1a) but unavailable to the other readings in (1b) or (1c); the latter require a so-called “comma intonation” such that the clause-final intonation drop marks the first clause as an independent assertion. Sweetser also discusses the role of iconic ordering of clauses in different domains (see her Ch. 4.2; more on this in § 3 below). The overall tendency of her approach, however, is based on the following claim (Sweetser 1990:78): (III) Pervasiveness of pragmatic ambiguity Given sufficient context, we can almost always force either a contentconjunction reading or an epistemic-conjunction reading on any pair of clauses coned by because; it is just harder to find reasonable contexts for some readings than for others. <...> [in addition to because] therefore, since, so, although, and despite all show such multiple usage.

2.

Objections and amendments

While I fundamentally agree with claim (I) One meaning, several uses (I consider this to be the only promising way of dealing with connectors!), and while I am basically sympathetic to the notion of »pragmatic ambiguity«, I see some serious difficulties in accepting the claims in (II) and (III).

4

Ewald Lang

The notion of »pragmatic ambiguity« is in need of refinement and elaboration. To achieve this aim Sweetser's approach will be relativized by (IV) a. challenging the ambiguity claim advocated in (II) and (III); b. inspecting more closely the features based on which the supposed levels are distinguished; c. checking the consistency and exhaustiveness of the threelevel distinction. Pragmatics is not considered to be the sort of stop-gap device which Sweetser purports it to be; rather, what Sweetser calls “a pragmatically motivated choice of interpretation” (see (II) above) will be shown to draw heavily on structural information available from the coned clauses as well as the connectors occurring in the conjunctions at issue. Specifically, I will defend (and provide evidence for) the following counter-claims, each of which has its bearing on the three points of criticism in (IV). For ease of reference, these claims will be prefixed by “C-”. The rest of this section will concentrate on (C-1), the others will be discussed in § 3. (C-1)

There are more level-specific structural differences in the way in which clauses are being coned than Sweetser's claims (I) - (III) would it.

(C-2)

The lexical items, i.e. the connectors that are used to connect clauses on the three levels of interpretation, display more level-specific distributional restrictions than the sample of cases discussed by Sweetser would suggest.

(C-3)

The three-level distinction proposed by Sweetser has to be supplemented by another level - provisionally called »textual progression« or »discourse perspective« - which is not located within the Sweetserian hierarchy of levels but cuts across them.

(C-1) emphasizes the role of structural information in determining theinterpretation of a given conjunction. Its import can easily be demonstrated by taking up Sweetser's introductory set of examples. A closer look at (1a-c) immediately reveals that there is a whole range of structural dimensions with respect to which the clauses that

Adversative connectors on distinct levels of discourse

5

are linked together by connectors differ. Spelling out (C-1) in detail means checking the clauses of what Sweetser calls a »conjunction« for a possible display of some crucial structural features, e.g., those listed in (C1-1) to (C1-5) below. (C1-1) Are the clauses of the same or of different clause type ? (in morphosyntactic of being declarative, interrogative, or imperative sentences) It is obvious that a content-level interpretation in the sense of Sweetser's paragon example (1a) is confined to clauses which share declarative mood and thus may render STATEMENTS. Taking this as a necessary condition for content-level interpretation implies that any conjunction whose clauses differ in type is excluded from being interpreted on that level and is thus is confined to be interpreted on epistemic and/or speech act level.2 Sharing declarative mood is a necessary, though not a sufficient, condition for a conjunction of clauses to be interpretable on the content level. Consider again (1b) and a version of (1c) in which the speech act of asking is not grammatically overt but concealed in a declarative: (1b) John loved her, because he came back. (1c') I am asking you what you are doing tonight, because there's a good movie on. 2

It may be noted in ing that sameness of clause type with non-declaratives raises some problems concerning the coning of clauses on the speech act level. I will take for granted that pragmatically (i) is asymmetric, while (ii) is symmetric regarding the alternatives presented: (i) (ii)

Give me liberty or give me death! [= Sweetser's (46), p. 98] Give me a salami sandwich or a hotdog!

What needs to be clarified, however, is this: is it justified to consider (ii) as an or conjunction of two speech acts? Speech acts normally resist being presented as alternatives. Another open question is whether we are to take (ii) as a conjunction of two speech acts or rather as a single speech act of request which comprises an alternative of how the request might be complied with? Again, these questions show that a closer examination of the internal structure of the conjuncts seems to be unavoidable.

6

Ewald Lang

This is the point where further criterial features enter the picture. One of them, of course, is comma intonation. Due to its interaction with other features the indicative role of comma intonation is more far-reaching than Sweetser assumes. Below the level of sharing declarative mood there are important indicators of clause-internal differences which interact with comma intonation. Thus, in isolation, the first clause of (1b) may render a STATEMENT as does the first clause in (1a). Both behave differently, though, if we add an adjunct like presumably or a parenthetical like I guess which mark the hypotheticality or assumption status of the assertion being made: (1a') Presumably John came back because he loved her. John, I guess, came back because he loved her. (1b') Presumably John loved her, because he came back. John, I guess, loved her, because he came back. While in (1a') hypotheticality markers like presumably take scope over both clauses, their scope in (1b') is confined to the first clause. Thus the function of comma intonation to mark the first clause as an independent assertion is confirmed by its effect to mark scope boundaries. Similar observations can be made for negation and other operators. Despite looking alike on surface, the first clauses in (1a, 1a') and (1b, 1b') differ in scope determination as well as in intonation, which undoubtedly both form part of core grammar, and hence provide distinct cues for the interpretation to be assigned to the conjunction as a whole. Thus despite the fact that the first clauses in (1a) and (1b) are alike in sharing declarative mood, they also differ. The first clause in (1a) asserts a proposition in the form of a factual statement thus allowing for a content-level interpretation of the conjunction. The first clause in (1b), however, asserts a proposition in the form of an ASSUMPTION as evidenced by comma intonation and characteristic scope-marking properties. The assumption status of the first clause is crucial to epistemic-level interpretation of the conjunction. The next dimension in of level-specific structure formation as claimed by (C-1) is mentioned by Sweetser in discussing »iconic ordering« in coordinate structures. However, the point at issue can be extended to causal connectors and exploited in a more systematic way. (C1-2) Restrictions on the order of clauses with subordinators

Adversative connectors on distinct levels of discourse

7

(1a) on the content level allows both orders - John came back because he loved her and Because he loved her John came back though there is a difference that distinguishes backgrounded vs. focussed material on the level of discourse related information structure. In any case, preposing the subordinate clause does not prevent a content-level reading as shown by (2a). Coherence is maintained independently of the order of clauses. However, as shown in (2b-c'), Sweetser's examples (1b) and (1c) as well as their German equivalents do not allow of preposed because / weil clauses if they are to keep their respective readings on the epistemic or speech act level (marked * and ??, respectively): (2)a. Because he loved her John came back. Weil er sie liebte, kam John zurück. [content ] b. Because he came back, John loved her. Weil er zurückkam, liebte John sie. * [epistemic] b'. Because he came back, John presumably / I guess LOVed her. Weil er zurückkam, LIEBTE John sie vermutlich / denke ich. [epistemic] c. Because there's a good movie on, what are you doing tonight ? ?? [speech act] c'. Because there's a good movie on, I'd like to ask you: what are you doing tonight ? [speech act] Thus, unlike (1b), a conjunction like (2b) cannot be interpreted as justifying the conclusion “he loved her” from the fact that “he came back”. What is lacking here is the sort of coherence that is rendered by the original ordering. To ensure coherence in such cases requires additional lexical or prosodic focus indicators, as is illustrated in (2b'). Similarly, while clause type difference and intonation break may serve as compensatory cues, (2c) is still incoherent. An amended version of it should read like (2c'). Sweetser recognizes the role of conjunct ordering in discussing conjunctions with coordinators like and, or and but by stating that pragmatically “the primary, independent conjunct precedes the secondary, dependent conjunct” (1990:98). Yet, she simply ignores the ordering effects in conjunctions with subordinators. The facts in (2b, 2c) above show that (C1-2) has to be taken as a condition on level-specific structure formation, which in turn has serious repercussions on Sweetser's ambiguity claim expounded in (II)-(III) above.

8

Ewald Lang

First, it is a grammatical category feature of subordinators like because, since etc. to allow for clause preposing while keeping their interpretation constant - cf. (2a). If this built-in property fails, as with (2b) for example, this has to be acknowledged as a disambiguating grammatical fact, too. And it is on this basis that pragmatic interpretation starts to work, contrary to the ambiguity claim formulated in (II). Second, unlike subordinators, coordinators like and, or, but (though not for or German denn ) are grammatically unrestricted as to the order of conjuncts and hence do not provide category-based grammatical information that is taken up by pragmatic interpretation. Among other things, it is the sequential ordering of the conjuncts that is submitted to pragmatic interpretation (along the lines discussed by Sweetser). Third, in view of this, we should expect coordinate conjunctions to be less fertile, hence more ambiguous or vague, in providing grammatically determined information on which, following Sweetser, the “pragmatic choice of the »correct« interpretation” may draw. In what follows it will be shown that this assumption is by no means justified. Due to space restrictions the further pursuit of (C-1) to (C-3) will be confined to coordinate conjunctions whose clauses are linked by adversative connectors. Moreover, we will employ German data3 and present arguments which draw on results from current work being carried out in the connector project at the Institut für deutsche Sprache Mannheim (for details see HdK (=Pasch et al. (forthcoming)). 3. 3.1

Adversative connectors How ambiguous are aber / but-conjunctions ?

From the set of about 30 lexical items that may be subsumed under the label “adversative connector” in German we will select only those needed to illustrate (C-1) to (C-3). 3 The advantage of illustrating the issue at hand with German data consists in the fact that aber is an unambiguous adversative connector. It does not share the range of additional uses and readings that are covered by but, as in German these are rendered by separate lexical items like sondern, außer etc. Analysing butconstructions in comparison with aber-constructions thus will help us to pick out only those having adversative readings.

Adversative connectors on distinct levels of discourse

9

Let us first turn to coordination involving the connector aber, which (like but in English) is commonly considered as the standard adversative connector. Compared with connectors such as trotzdem ('nevertheless'), hingegen ('whereas'), or dennoch ('nonetheless', 'all the same'), aber is much less restricted regarding distributionhence to a much higher degree prone to »pragmatic ambiguity«. Nevertheless, the supposed ambiguity of aber-conjunctions is drastically reduced by structural cues which emerge from structural features of the coordinate clauses in the domain of sentence structure. In addition to C1-1 and C1-2, the relevant features include: (C1-3) Matching or independently varying syntactic clause structure (C1-4) Parallel or independently varying background-focus structure (C1-5) (Un)availability of clause-internal deletions like »gapping«, »right node raising« etc. (cf. Lang 1991). Let us now look at some data that are representative of the range of cases to be taken into . (3)a. Dein Vortrag war sehr lang, aber das soll keine Kritik sein. 'Your talk was rather lengthy, but this is not a criticism' b. Wir sind mit dem Manuskript in Verzug, aber welcher Autor hält schon Termine ein ? 'We are late with the manuscript, but what author keeps the dead-lines?' c. Willst du heute etwa ins Kino - aber ich komme nicht mit. 'Do you want to go to the movies today - but I 'm not coming along' All examples in (3) are easily recognizable as cases to be interpreted on the speech act level. Just recall the relevant indicators mentioned in (C1-1), namely »clause type difference« – cf. (3b, 3c), »intonational break« – cf. (3a), and what was said on »independent clause-internal structures« in C1-3 to C1-4 (i.e. lack of parallelism as to constituency and background-focus articulation). Just as in (1c) above, it is the conjunct-internal differences which prevent the examples in (3) from being interpretable on content level and qualifies them for speech act level interpretation. Generally

10

Ewald Lang

speaking, the reason is this: aber/but marks what is rendered by the second clause as being in contrast with inferences that might be drawn from the speech act that is performed by uttering the first clause. In short, what counts are the conjunct-internal differences in of C1-1 to C1-4. If such differences are missing, more effort is needed to determine the respective levels of interpretation. Consider the following examples, which Sweetser tends to locate at the epistemic level. (4)a. Hans hat Unmengen von Zigarren gehortet, aber er raucht gar nicht. 'Hans hoarded an obscene amount of cigars, but he doesn't even smoke' b. John keeps six boxes of pancake mix on hand, but he never eats pancakes. (cf. Sweetser (1990:100)) Sweetser comments: “But presents two conjuncts which clash with each other in some way ... At the epistemic level, the available premises may clash with an apparently necessary conclusion” (1990:100). The first part of the quote is unproblematic; but what exactly is »epistemic« about the entities connected by aber/but? (4a, b) instantiate the well-known type of »denial of expectation« interpretation of but-conjunctions (Lakoff 1971). In both cases, the clauses are mutually independent, hence semantically compatible STATEMENTS. Given this, (4a, b) qualify as coordinations of statements at the content level with the additional requirement that the interpretation involves inferring an assumption which is in contrast with the two statements being presented side by side. This analysis reduces the »epistemic« aspect of (4a, b) to the inferences to be drawn and thereby does justice to the structurally determined level-distinction. 3.2

Problems with interpreting but/aber at the content level

The awkwardness of Sweetser's use of the notion »epistemic« is furthermore revealed by the stand she takes on cases of »semantic opposition« (Lakoff 1971). She claims (1990:103): “The but in (56) [= (5a) below] does indeed indicate contrast: an epistemic contrast between two semantically opposed propositions” (italics mine - EL). From this we may conclude that (5b) lacks such an epistemic contrast, other things being equal. Again, (5b) is a clear-cut case of content-level conjunction. (5)a. John is rich but Bill is poor. [= Sweetser's (56), 1990:103 ]

Adversative connectors on distinct levels of discourse

11

b. John is rich and Bill is poor. c. John is rich, but Bill is rich, too. d. *John is rich but Bill is rich. What we are left with is the problem of where to locate (5a): it does indeed indicate a different contrast from (5b), I doubt whether this contrast can be ed for in of the »epistemic level« without rendering this notion completely void. I would contend that the alleged »epistemic« difference between (5a) and (5b) is that the interpretation of the former requires some additional inferencing which the interpretation of the latter does not. Hence, (5a) and (5b), like (5c) when compared with (5d), clearly differ as to the conditions they impose on the context in order to get interpreted properly. According to an observation first made by Bellert (1972), the interpretation of but/aber-conjunctions involves at least two semantic contrasts to be read off or inferred from the conjuncts. (5d) contains the overt contrast John - Bill but fails to provide cues to reconstruct the second contrast needed, i.e. an INFERRED ASSUMPTION which is in opposition to the STATEMENT Bill is rich. This pragmatic insufficiency has its grammatical counterpart: (5d) can be considered deviant because it does not allow a contrastive stress pattern – just try and you will see! Moreover, contrary to most work done in the field, we claim that the »semantic opposition« (more exactly, the occurrence of antonymous predicates in logically compatible STATEMENTS) in (5a) is not a crucial feature of this type of conjunction (as shown by (5c) which lacks opposition); and that this opposition is by no means the carrier of the contrast indicated by aber/but. Instead, the contrasting entity has to be inferred from outside the coordination. However, the opposition in (5a) as well as its absence in (5c) serve as structural clues to reconstructing it. We will return to this in 3.3 below. Now, if (5a) cannot be located on the epistemic level, why not consider it a content-level conjunction – very much like (5b)? Sweetser (1990:103) rejects this idea by stating: “I have not been able to unearth any indubitable content-conjunction examples with but ” and by offering “as a plausible explanation for the use of but in only two domains” [epistemic and speech act level - EL] the following consideration: “what does it mean to say that A and B »clash« or »contrast« in the real world? How can discordance or contrast exist outside of the speaker's mental concept of harmony or non-contrast ?”

12

Ewald Lang

Sure enough, the contrast induced by aber/but is not something to be looked for in the real world but something to be established by the interlocutors in assessing what, by way of coordinate coning, is presented as co-existing in the world. But this does not prevent aber/but-conjunctions from being interpretable at the content-level, that is, as coning two STATEMENTS which lay the ground for additional inferencing. Sweetser's problem with interpreting but at the content level is caused by confusing two notions which, though not unrelated, have to be kept distinct: inference (whether a proposition is asserted or inferred) and epistemic status (whether a proposition forms the content of a factual STATEMENT or an ASSUMPTION). The alternative approach suggested here tries to avoid this confusion. It draws on the fact that the interpretation of an adversative conjunction involves relating the coned clauses to an ASSUMPTION which - if not available in explicit (see (V) below) - has to be inferred from outside the conjunction. Based on these considerations it is posited that (V) (a) adversative (and probably also concessive) connectors inherently contain pointers to previous information available from the context, and due to this (b) adversative conjunctions necessarily involve some backtracking that may well go beyond the domain of sentence structure and operates on (what may be called) the level of »textual progression « or »discourse perspective«. It is this level which provides the appropriate means to interpret adversative conjunctions, and therefore we consider it a necessary supplement of Sweetser's approach. Note that the level of »discourse perspective« must not be understood as an additional level within, or on top of, the Sweetserian hierarchy, but rather as embracing the three levels discussed so far as grammatically determined patterns of discourse formation. Recall that C1-1 to C1-5 apply to sentence structure and thus clearly contribute to discourse formation and comprehension. 3.3

Adversative conjunctions and the division of labour between grammar and pragmatics

Adversative connectors on distinct levels of discourse

13

This section will provide an outline of the way in which sentence and discourse level interact in determining the interpretation of adversative conjunctions. To begin with, note that the semantic contribution of aber/but in combining two clauses into a conjunction is twofold. In a nutshell, it may be conceived of as: (VI) (a) coning semantically compatible and non-inclusive propositions that - depending on the amount of structural parallels shared by the clauses expressing them - may be bundled up or enumerated as instances of a »common integrator« (Lang 1984, 1991, in print); (b) indicating that the assertion rendered by the second clause is in contrast to an ASSUMPTION that either may be read off, or must be inferred from, previous information (Brauße 1998). (VIa) states what aber/but share with und/and, therefore (VIa) is the basis for (5a) and (5b) having identical truth conditions. This is a feature of aber/but that probably nobody will deny. (VIb) is what separates aber/but from und/and, and hence (VIb) is what causes (5a) and (5b) to have distinct conditions of use as regards semantic coherence. As such, (VIb) is perhaps not as uncontroversial as (VIa), so it may be worth commenting on. What has been stated in (VIa + b) should not be understood as amounting to “but = and + pragmatics”. This would definitely miss the target. The conditions of use of aber/but and other adversative connectors are not simply a matter of »pragmatically motivated interpretation« (as Sweetser suggests); rather they are rooted in the lexical semantics as well as in the syntactic category features of these connectors. On the other hand, I do not deny the role which contextually induced pragmatic interpretation inevitably has to play in specifying the »correct« reading which a conjunction of clauses eventually gets assigned. I differ from Sweetser, however, by claiming that in order to for (what she calls) »pragmatic ambiguity«, the division of labour between grammar and pragmatics has to be redefined. Regarding aber/but as characterized in (VIb), this division of labour may be determined as follows: That part of (VIb) which reads »indicating that the assertion rendered by the second clause is in contrast to an ASSUMPTION« has to be spelled out primarily in of grammatical structure

14

Ewald Lang

formation (including C1-1 to C1-5), while the other part, which reads »an ASSUMPTION that either may be read off, or must be iferred from, previous information«, is where pragmatics proper enters the picture. It seems reasonable to consider the interpretation of an adversative conjunction as involving a search device which is determined by a specific source-target relation. Triggered by the lexical indication that a contrast must be established, the device is determined to take the assertion rendered by the second clause as the source from which to look for an appropriate ASSUMPTION that meets the condition of contrast, i.e. the target. Though the search device is to some extent guided by grammatical structure, as claimed by (C11) to (C1-5), it is inevitably accompanied and supplemented by pragmatically based reasoning. So the part to be played by pragmatics proper is to provide supplementary means to find out what the target, i.e. the INFERRED ASSUMPTION,is. Note that by dividing up (VIb) between grammar and pragmatics assuggested afore, it is not an accident that the notion »ASSUMPTION« figures in both, and rightly so. The INFERRED ASSUMPTION forms the bridge between grammar and pragmatics in interpreting adversative conjunctions. Almost all studies of adversative connectors have attempted to for it in one way or another (for a survey see Rudolph 1996). What is new is the way to tackle it as proposed in (VIa + b). Source and target are both propositional but differ as to their epistemic status. That the target (i.e. the proposition which the second clause is put into contrast with) has to be an ASSUMPTION, not a factual STATEMENT, can be seen if, for example, (4a) is extended by a clause that renders the target in explicit , cf. the bold-faced additions in (6): (6)a. Hans hat Unmengen von Zigarren gehortet, aber er raucht gar nicht. 'Hans hoarded an obscene amount of cigars, but he doesn't (even) smoke' b. Hans hat Unmengen von Zigarren gehortet, *folglich raucht er, aber er raucht nicht. '... hence he smokes, but he doesn't smoke' c. Hans hat Unmengen von Zigarren gehortet, also könnte man vermuten, daß er raucht, aber er raucht nicht. ' ... so one could assume that he smokes,

[= 4a ]

Adversative connectors on distinct levels of discourse

15

but he doesn't smoke' If the supposed target of contrast is expressed in the form of a FACTUAL STATEMENT, the result is a contradiction - cf. (6b), but if it appears with overt markers of ASSUMPTION status as in (6c), the interpretation of the entire construction is consistent. This is what follows from the grammar part of (VI), which requires the target of contrast to be an ASSUMPTION. Let us now turn to the pragmatic aspects of (VI). If the clauses coned by aber/but both represent STATEMENTS (as is the case with the majority of adversative conjunctions), the ASSUMPTION forming the target of contrast has to be inferred. Here is where pragmatics comes into play by providing the means to infer the target assumption from previous information. Reconsidering the conjunctions in (4) and (5) above will then amount to the following analysis. In (4a), the inference is obviously drawn by resorting to world knowledge about smokers such that the STATEMENT rendered by the first clause (Hans hoarded an obscene amount of cigars) is taken as a premise which by abductive reasoning makes it plausible to assume »Hans smokes« and thus to yield a suitable target assumption. (By the way, the fact that this inference is not based on deductive but on abductive reasoning provides additional evidence of the ASSUMPTION status of the target.) In (5a-d), repeated as (7a-d) below, the inferential steps leading to the target are less clearly set out. (This is what caused the trouble for Sweetser's approach, cf. § 3.2 above). How come ? The clauses coned by aber/but in (7) are clear-cut STATEMENTS, so the target of contrast has to be inferred form outside the conjunction. Moreover, the clauses coned are absolutely parallel as to their internal structure. So they cannot but be taken as instances of a »common integrator« (Lang 1984, 1991). The common integrator, which has to be identified or inferred outside the conjunction, then serves as the location frame within which the contrast is to be placed. Thus, in comparison to (4a), the conjunctions in (7) require additional cues from the discourse level in order to be assigned an appropriate interpretation. The simplest way to illustrate these context conditions consists in showing what sort of questions the respective conjunctions in (7) might serve to answer. Suitable questions (added inside double slashes) thus reveal the distinct effects that are caused by the structural differences of the coned clauses.

16

Ewald Lang

(7)a. John is rich but Bill is poor . //As for John and Bill, are they both rich ? // b. John is rich and Bill is poor. // As for John and Bill, what about their income ? // c. John is rich, but Bill is rich, too. // As for John and Bill, they aren't both rich, are they ? // d. *John is rich but Bill is rich. // ??? // Assessing (4) and (7) once more, we may state this: in isolation these conjunctions are not really ambiguous but pragmatically underspecified regarding the cues needed to get interpreted. Finally, we will look at thosegrammatical features of adversative connectors that form the basis of the division of labour between grammar and pragmatics just outlined. The most relevant point to note is this: indicating a contrast in the way suggested in (VIb) is a semantically built-in feature of aber/but which correlates with likewise built-in syntactic properties of these connectors. In contrast to and, or etc., adversative coordinators like aber/but display the following features: First, aber /but allows for binary connection only. This is the basis for marking the assertion following aber/but as the source from which to look for an appropriate CONTRASTING ASSUMPTION, that is, the target. Second, aber /but) is intrinsically asymmetric in marking what has been called a »change of perspective « (Spooren 1989): for example, (4a) or (7a) with reversed order of conjuncts would not render the same information on the level of discourse perspective because source and target are reversed. Third, unlike und/and, which allows for scope ambiguities, aber/but is a marker of scope boundaries: negation or other scopetaking operators cannot take scope over aber/but, cf.: (9)a. Hans ist nicht dumm und faul [ambiguous as to scope of negation] 'Hans is not stupid and lazy'

b. Hans ist nicht dumm, aber faul [scope of negation unambiguous] 'Hans is not stupid but he is lazy' 4 4

(9b) points to an important difference between German aber and English but regarding scope determination under negation. Since German adversative coordinators for contrast (aber) vs. correction (sondern) are lexically distinct, their

Adversative connectors on distinct levels of discourse

17

Fourth, German aber (unlike English but ) is not confined to interclausal positions, but can also float in the second clause (this is why in Pasch et al. (forthcoming) aber has been classified as a »connective particle«). At first glance, this distributional flexibility seems to the view that aber is most prone to »pragmatic ambiguity«. It can be shown, however, that being allowed to float within the second clause is what enables aber to mark the background-focus articulation, and hence to give additional cues on the discourse-related source from which to look for the contrasting target. In the following examples, SMALL CAPS indicate main stress, accent grave contrastive stress, and square brackets focus. (10) a. [Hàns] ist [KRANK], aber [seine Fràu] [geht ARBEITEN] // Wie geht's den beiden ? // 'How are the two doing?' b. [Hàns] ist [KRANK], [seine Fràu] aber [geht ARBEITEN] // Wie geht's den beiden ? // 'How are the two doing?' c. [Hàns] ist [KRANK], [seine Fràu] hingegen [geht ARBEITEN] // Wie geht's den beiden ? // 'How are the two doing?' d. Hans ist [KRANK], aber seine Frau [geht ARBEITEN] scopal behaviour is built-in and need not be indicated by other means - cf. (i, ii), English but- constructions, however, have to render the distinction of contrast vs. correction under negation by distinct conditions on deleting non-focussed material in the second clause. Thus, but- constructions expressing contrast disallow (iii), whereas but- constructions expressing correction require (iv), the deletion of repeated material in the second conjunct (for details see Lang (1991)). (i) (ii) (iii) (iv)

Hans ist nicht dumm, aber (er ist) faul Hans ist nicht dumm, sondern (er ist) faul Hans is not stupid but *(he is) lazy Hans is not stupid but (*he is) lazy

[contrast] [correction] [contrast] [correction]

The same point can be made to explain the difference in the following cases: (v) (vii) (vi) (viii)

John ist reich, aber Bill (ist) arm. John is rich but Bill *(is) poor. John ist reich, aber Bill ____nicht /auch. John is rich but Bill *(is) not /too.

Despite the differences shown, aber and adversative but both mark scope boundaries in preventing propositional operators to take scope over them.

18

Ewald Lang

e. Hans ist [KRANK], seine Frau [geht aber ARBEITEN] f. Hans ist [KRANK], dennoch [GEHT] seine Frau [ARBEITEN] g. Obwohl Hans [KRANK ist], [geht] seine Frau [ARBEITEN] As regards form and interpretation, (10a-c) resemble »semantic opposition« cases like (6a)/(8a) as evidenced by the // questions // they would make suitable answers to. The cases (10d-f), however, due to their topic-focus articulation, are to be interpreted as concessive - very much like the overt concessive (10g). Hence, they differ from (10a-c) in the way they are linked to previous discourse. They would not form suitable answers to // Wie geht's den beiden? // 'How are the two doing?' but to something like // Was mich aufregt ist ... // 'What upsets me is ...' . The relationship between aberconjunctions and concessive constructions cannot be tackled here (but see the contributions of Thompson & Couper-Kuhlen, Barth, and Günthner in this volume). In sum, (10c-g) give a hint at the range of adversative connectors that are distributionally more restricted than aber and are hence more restrictive as to the choice of interpretation (for details see HdK). 4. Concluding remarks Sweetser's three-level approach has proved to be valuable as a heuristic means in sorting the various interpretations that conjunctions can be assigned to. However, in order to serve as a means of explanation, it needs to be modified in some respects. Regarding the level of grammatically determined sentence structure, Sweetser's level-distinction can be more precisely reconstructed in of criteria like C1-1 to C1-5, among which C1-1 (sameness or difference of clause type) is the most crucial one. Given this, we might reconstruct what should be kept of Sweetser's three-level distinction of conjunction interpretation in the following way: (VII) a. Content-level interpretation is reserved for conjunctions of declarative clauses that render statements (i.e. entities to be related to the world via truth-conditions); see (1a, 3a, 6b). b. Epistemic-level interpretation is confined to conjunctions of declarative clauses, of which at least one is overtly marked as rendering an assumption; see (1a', 1b, 1b', 3b, 3b', 7c).

Adversative connectors on distinct levels of discourse

19

c. Speech act level-interpretation is available for conjunctions of non-declaratives (with reservations about or-conjunctions like (2)) as well as for conjunctions of clauses differing as to type but with one clause being marked as rendering a speech act; see (1c, 2a, 4a-c). In addition, Sweetser's approach has to be modified in a way that would enable it to for contextual requirements imposed on the interpretation of adversative conjunctions. The level that is assumed to do this is the all-embracing level of »discourse perspective« (or »Textverlauf«). It is this level where the peculiar syntax and semantics of adversative connectors can be ed for in a promising way, and it is this level where the division of labour between grammar and pragmatics can be defined – cf. (VI a+b). Pragmatics comes in when grammar does not provide sufficient cues to get an adversative conjunction properly interpreted.

References Bellert, Irena 1972 On certain properties of the English Connectives and and but. In: Senta Plötz (ed.) Transfomationelle Analyse, 327-356. Franfurt / M.: Athenäum Brauße, Ursula 1998 Was ist Adversativität? aber oder und ? Deutsche Sprache 26:138-159 HdK [= Pasch, Renate, Ursula Brausse, and Eva Breindl] (forthcoming) Handbuch der deutschen Konnektoren. 2 vols. Ms. Mannheim: Institut für deutsche Sprache (vol. I is to appear in 2000) Kortmann, Bernd 1997 Adverbial Subordination. A Typology and History of Adverbial Subordinators Based on European Languages. (EALT 18). Berlin/New York: de Gruyter Lakoff, Robin 1971 Ifs, ands and buts about conjunction. In: Charles J. Fillmore & D. T. Langendoen (eds.) Studies in Linguistic Semantics, 114-149. New York: Holt, Rinehart & Winston Lang, Ewald 1984 The Semantics of Coordination. (SLCS 9) Amsterdam/ Philadelphia: John Benjamins

20

Ewald Lang

Lang, Ewald 1991 Koordinierende Konjunktionen. In: Semantik / Semantics, Art. 26, 597-623. (HSK 6) Berlin/New York: de Gruyter Lang, Ewald (in print) Die Wortart »Konjunktion«. In: Lexicology / Lexikologie , Art. 88. Berlin/New York: de Gruyter Martin, Jim R. 1983 Conjunction: the Logic of English Text. In: J. S. Petöfi and E. Sözer (eds.) Micro and Macro Connexity of Texts, 1-72. Hamburg: Buske Raible, Wolfgang 1992 Junktion. Eine Dimension der Sprache und ihre Realisierungsformen zwischen Aggregation und Integration. Heidelberg: Winter Rudolph, Elisabeth 1996 Contrast: adversative and concessive relations and their expressions in English, German, Spanish, Portuguese on sentence and text level. (Research in text theoy 23) Berlin/New York: de Gruyter Spooren, Willibrordus 1989 Some Aspects of the form and interpretation of global constrastive coherence relations. PhD Dissertation. University of Nijmegen Sweetser, Eve E. 1990 From Etymology to Pragmatics. Cambridge University Press

Author's address: HU Berlin, Phil. Fak. II Institut für deutsche Sprache und Linguistik Schützenstr. 21 D-10 117 Berlin email: [email protected]. de

Adversative connectors on distinct levels of discourse

21

The clauses coned by aber/but and in (7) are clear-cut so the target of contrast has to be inferred form outside the conjunction. Moreover, the clauses coned are absolutely parallel as to their internal structure. So they cannot but be taken as instances of a »common integrator« (Lang 1984, 1991). The common integrator, which has to be identified or inferred outside the conjunction, then serves as the location frame within which the contrast is to be placed. Thus, in comparison to (4a), the conjunctions in (7) require additional cues from the discourse level in order to be assigned an appropriate interpretation. STATEMENTS,

The simplest way to illustrate these context conditions consists in showing what sort of questions the respective conjunctions in (7) might serve to answer. Suitable questions (added inside double slashes) thus reveal the distinct effects that are caused by the structural differences of the coned clauses. (7)a. John is rich but Bill is poor. //As for John and Bill, are they both rich ? // b. John is rich and Bill is poor. // As for John and Bill, what about their income ? // c. John is rich, but Bill is rich, too. // As for John and Bill, they aren't both rich, are they ? // d. *John is rich but Bill is rich. // ??? // Assessing (4) and (7) once more, we may state this: in isolation and in written form, these conjunctions are not really ambiguous but pragmatically underspecified regarding the cues needed to get interpreted. n

n

Without going into details (to be found in Lang 1984, 1991), let me just point out that the derivation of the Common Integrator of a given conjunction is (a) sensitive to surface structure parallels shared by the coned clauses, and (b) can best be illustrated by means of a (multiple) question to which the given conjunction would serve as an appropriate answer. n Without going into details (to be found in Lang 1984, 1991), let me just point out that the derivation of the Common Integrator (CI) of a given conjunction (a) is sensitive to surface structure parallels shared by the coned clauses, and (b) can best be illustrated by means of a (multiple) wh-question to which the given conjunction would serve as an appropriate answer.

22

Ewald Lang

So while (ii) - (iii) share the propositional formula "someone transmits something to someone" as constituent part of their respective CIs, the specific make-up of the latter depends on the information structure of the clauses and on the (non-) coreference of the items instantiating the argument places.

(44) a. Opa vererbt den Enkelni das Hausk und Oma unsj die Hüttel b. Opa und Oma vererben den Enkelni und unsj das Hausk bzw. die Hüttel c. CI: WER vererbt

WEM

WAS?

(45) a. Opa vererbt den Enkelni das Hausk, aber Oma schenkt unsi die Hüttek b. Opa vererbt, aber Oma schenkt, [uns Enkeln]i das Hausk c. CI: WER ÜBEREIGNET WIE uns Enkeln das Haus ? (46) a. Die Sonne scheint und die Vögel singen b. ?? Die Sonne und die Vögel scheinen bzw. singen TUT WAS ? c. CI: ?? WER a. Grandpa will bequeath the housek to the grandchildreni and Grandma will (bequeath) the cottagel /*k to usj /*i b. Grandpa and Grandma will bequeath the house and the cottage to the grandchildren and to us, respectively den Enkelni das Hausk und Oma unsi / j die Hüttel b. Opa und Oma vererben den Enkelni und unsj/*i das Hausk bzw. die Hüttel c. CI: WHO will bequeath WHAT TO WHOM

WAS?

Grandpa will bequeath the housek to the grandchildren i, but Grandmawill give usi the cottagek as a gift

Adversative connectors on distinct levels of discourse

23

(45) a. Opa vererbt den Enkelni das Hausk, aber Oma schenkt unsi die Hüttek b. Opa vererbt, aber Oma schenkt, [uns Enkeln]i das Hausk c. CI: WER ÜBEREIGNET WIE uns Enkeln das Haus ? (46) a. Die Sonne scheint und die Vögel singen b. ?? Die Sonne und die Vögel scheinen bzw. singen c. CI: ?? WER TUT WAS ?