Correction-but: A relevance-theoretic reappraisal

Language & Communication 45 (2015) 27–45

Contents lists available at ScienceDirect

Language & Communication journal homepage: www.elsevier.com/locate/langcom

Correction-but: A relevance-theoretic reappraisal Valandis Bardzokas* Language Academy, Livesey House, University of Central Lancashire, Preston, PR1 7EA, United Kingdom

a b s t r a c t Keywords: Correction ‘but’ Relevance theory Procedural/conceptual meaning Descriptive/metarepresentational negation

The current paper unveils the problems undermining the relevance-theoretic procedural view of correction-but, as it stands, in face of the fact that the specific application appears to defy inclusion in the procedural definition. On this view, the uniformity of a general account of but seems to be at risk. The paper aims to propose a more plausible relevancedriven account of correction that does not compromise the cause of a univocal treatment of but in procedural terms. To achieve this aim, it introduces a distinction in correction interpretation that has so far eluded attention in the relevant literature: descriptive and metarepresentational. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Pragmatic theory crucially evolved from a constant concern with the identification of the interface between semantic and pragmatic interpretation. Central to this issue has been the perpetual challenge of ascertaining which side of the divide a certain class of words falls on, specifically the class of discourse markers. Among them figures prominently the marker but, a case of contention between semanticists and pragmatists since Grice’s work (1989). In the Gricean spirit, the classification of a marker as either semantic or pragmatic served the higher cause of safeguarding the autonomy of a minimal truth-functional semantics. Against this ulterior motive, the determination of the meaning of but strictly by appeal to the semantics/pragmatics distinction takes precedence over a more detailed description of its meaning. In fact, however, a host of context-sensitive interpretations of but have been identified and discussed in subsequent research; semantic opposition,1 denial of expectation, objection and correction interpretations are cases in point (R. Lakoff, 1971; Anscombre and Ducrot, 1977; Rouchota, 1998). But in face of this range of meanings, the prospect of a minimal and, thus, uniform account of the connective under discussion looks more remote and the analyst’s attention may be deflected from the original question in pragmatic research: which side of communication the meaning of but operates on. In this sense, if the plan for a uniform account is to be realistically sustained, the most suitable course of action to take is a single treatment of all varieties of but (Blakemore, 2002; Hall, 2007; Iten, 2000, 2005), rather than dwelling on the delineation of a single meaning.

* Liberty Court, Brook Street, Preston, PR1 7EA, United Kingdom. Tel.: þ44 306944145405. E-mail addresses: [email protected], [email protected]. 1 In Blakemore’s (1989) terms, the case of semantic opposition is discussed in terms of contrast in reaction to the problems besetting R. Lakoff’s (1971) suggestion. However, the focus of more recent relevance-theoretic discussions (Hall, 2007: 152; Iten, 2005: 112) on the overall (procedural) behavior of but appears to deprioritize this point. Whatever the motivation for their decision, the current paper will continue this practice, employing, henceforth, the term of semantic opposition as a general term. http://dx.doi.org/10.1016/j.langcom.2015.08.004 0271-5309/Ó 2015 Elsevier Ltd. All rights reserved.

28

V. Bardzokas / Language & Communication 45 (2015) 27–45

But are all varieties ultimately amenable to a single treatment? Indeed, there is an interpretation associated with but that has been accorded improper regard until fairly recently, that is in the more context-sensitive framework of relevance theory (Blakemore, 2002; Iten, 2000, 2005). In Iten’s (2005) terms, it is the case of correction, a variety originally introduced in Anscombre and Ducrot’s (1977) discussion of (the corresponding use of) French mais. To exemplify, the utterance in (1) (1) That’s not my sister but my mother, [Iten, 2005: 149] is construed as an instance of correction, in that the utterance is intended to rectify the hearer’s misconception as to the identity of the speaker’s relative. Iten’s comment on the particular variety of but previews the case of its potentially differential behavior: While it is at least conceivable that ‘semantic opposition’ or ‘contrast’ but may be reduced to ‘denial of expectation’ but, there is another use of but [correction use] which doesn’t seem amenable to such a treatment (Iten, 2005:113) Despite this acknowledgment, it is interesting that relevance-theoretic investigations have typically favored a uniform meaning description to the benefit of a parsimonious treatment. More specifically, Blakemore (2002) and Iten (2000, 2005) propose a singular procedural account of but as equally applicable to the use of correction. But, if Iten’s earlier point on the peculiarity of correction use is justified, how does it affect the suggestion for an overall procedural application? In other words, the issue that seems of relevance, here, is whether or not the standard procedural treatment of but can be preserved in handling the variety of correction. If it cannot, the course of action to take looks straightforward. Either the procedural approach to correction is abandoned altogether (the future directions of the research being uncertain) or it is maintained, though not exactly as it stands. Both of the two alternatives would necessarily involve implications for the overall procedural account of but, though the implications of the former would sound more far-reaching. To catch but a glimpse of the terminal point of our discussion, we will basically argue that a general, procedural scheme of analysis, as this is standardly couched in the relevance-theoretic literature on but, can be employed to accommodate effectively the case of correction, though on certain additional (contextual and, also, theoretical) stipulations. More specifically, our discussion will point out a number of problems that undermine the current relevance-theoretic practice of stretching the procedural view of but to correction, a rather variegated, as it turns out, linguistic phenomenon. We will argue that a more accurate meaning description is required; one that involves sharper contextual observations of correction illuminating the peculiarities of this category of meaning. This paper, then, sets out to offer a more rigorous and unifying account of correction-but, implementing an apparatus of meaning analysis that proved reliable and versatile in past research on connectives (Bardzokas, 2012, 2013, 2014), i.e. that of relevance theory (Sperber and Wilson, 1995). In this line of investigation, it follows that the current work also measures the extent to which relevance theory can rise to the occasion of this new challenge. More concretely, our reappraisal of the variety of correction draws on the interface between levels of meaning that have fundamentally shaped the relevance-theoretic view of discourse markers as linguistic constraints on communication i.e. the interface between procedural encoding (Blakemore, 1987, 2002) and explicit/implicitly communicated meaning (Wilson and Sperber, 1993; Carston, 2002). Through the options opened up by these cross-cutting distinctions, we will navigate towards the view of correction as encoding a linguistic constraint on implicit import. Section 2, our point of departure, offers an introduction to the varieties of but. Section 3 offers a review of the relevance-theoretic account of but in procedural terms, while Section 4 exposes the problems besetting the typical relevance-theoretic procedural treatment of correction. Section 5 examines the plausibility of an alternative view of correction based on the suggestion of explicit, rather than implicitly communicated content, particularly of the higher order. Whereas Section 6 keeps up our considerations of explicit content, it entertains the possibility of expounding the phenomenon analyzed in connection with the basic (propositional) order of explicitness. By a process of elimination, Section 7 arrives at the conclusion that the phenomenon of correction can be treated in the standard procedural terms, on two conditions; firstly, that the interpretation of correction is analyzed more rigorously as including two subcategories of meaning, one of which being readily amenable to the traditional procedural analysis of but; secondly, that the range of contextual effects involved in accounting for but is marginally broadened, so that the new account takes care of data that fall into the second of these two sub-categories of meaning. Finally, Section 8 discusses the prospective implications of our position. Let us embark on our enquiry by way of introducing the varieties of but.

2. Basic varieties of but There appear to be two types of interpretation that figure prominently in the literature of but: a) denial of expectation and b) semantic opposition (R. Lakoff, 1971; Abraham, 1979; Kitis, 1982, 2000), the latter also discussed as adversative-but (König, 1985). By way of illustration, the time-worn, humorous example in (2) and the one in (3) typify the two varieties, respectively. (2) John is a Republican but he is honest. (3) John is short but Mary is tall.

V. Bardzokas / Language & Communication 45 (2015) 27–45

29

In the sense of denial of expectation, the realization of clause p, i.e. John is a Republican is viewed as generating the implicit assumption that John is dishonest on the evidential grounds that Republicans are likely to be dishonest. The presence of the but-introduced clause, i.e. but q, then, is especially justified as denying the particular implicit assumption. On the other hand, the utterance in (3) is roughly understood as a separate case, i.e. a case of contrast (R. Lakoff, 1971) intended to conjoin semantically opposed (or antonymous) attributes, for instance being short and being tall. A distinctive feature of semantic opposition, according to R. Lakoff (1971) is that the referents of the two subjects of the conjoined clauses are not identical. Moreover, subsequent research revealed the potential for further categorization in the meaning of but. Consider, for instance the contextual situation in (4): (4) A: John is in Paris at the moment. B: But I’ve just seen him in Oxford Street. [Rouchota 1998: 25] As illustrated in (4) (B), the but-introduced clause occurs in utterance-initial position, in which place it can be construed in terms of the speaker raising an objection to the proposition expressed in the preceding utterance. The clausal sequence in objection contexts is typically shared by two speakers, i.e. (4) A and (4) B, speaker B denying indirectly speaker A’s suggestion, i.e. that John is in Paris at the moment. Turning now to the interpretation that the current work sets out to investigate, Anscombre and Ducrot (1977) contribute the but-variety reflected in (5) (B), from a cross-linguistic perspective. Although they talk about French mais, the particular connective is the translation counterpart of English but. Moreover, the variety that they add seems equally possible in English. (5) A: Your sister looks a lot like you. B: That’s not my sister but my mother. [example in English taken from Iten, 2005: 149] Iten (2005: 114) dubs this use of but ‘correction’, to the extent that an utterance of the sort in (5) (B) makes sense vis-à-vis B’s intention to correct A’s mistaken impression of the identity of B’s relative. Now, the variety of correction bears the following hallmarks (Anscombre and Ducrot, 1977: 25–5): a) p takes the form of unincorporated, i.e. uninflected or monolexemic, negation (e.g. not p), b) both p and q are uttered by the same speaker, c) q is presented as the reason for rejecting p and d) p and q characterize the same kind of fact (in ways which are considered incompatible with each other). On this view, q is seen as refuting p directly, in which case it is capable of replacing p. The qualification in (d) seems accountable for the reduction in the linguistic material (or ellipsis) in q. That is, the type of material that p and q have in common, i.e. in our case the verb be, is omitted.2 On the basis of these structural facts, we can safely draw the following conclusion pertinent to the concerns of the current investigation: the structure of correction is not identical with that of the remaining but-varieties. While both p and q conjoined in (2)–(4) reflect fully sentential clauses, the same is not true of the correction case in (5) (B), where the logical form in q is subsentential. In fact, it is also irreversibly subsentential, in the sense that the reduced form of q does not afford a fully sentential version (that does not result in a denial-of-expectation interpretation). However, while q assumes a subsentential form, it is also interpretable in terms of its fully sentential form, especially in the sort of linguistic environment that is free of the structural constraint of correction but (plus ellipsis), as in That’s not my sister. That’s my mother. Having discussed the varieties of but, we can now turn to our initial issue, namely to determine the extent to which these varieties can merge into a single category of procedural encoding, so that we can secure the much cherished uniform treatment of the connective. To take but a preview of the imminent challenge, given its distributional uniqueness, correction may or may not prove analyzable on a par with the remaining varieties of but. To pursue this discussion, we first need to outline the current procedural view of the overall meaning of but in the relevance-theoretic framework. 3. A procedural account of the meaning of but Utterance comprehension, the point of departure in linguistic communication, depends on two disparate processes undertaken by the hearer, i.e. decoding the linguistic material of a sentence that the speaker utters and using it as evidence to infer the speaker’s intention in uttering the sentence. In view of the well-attested gap between linguistically and inferentially determined elements of utterance interpretation, lexical items are standardly partitioned into two types of encoding in the relevance-theoretic framework: conceptual and procedural (Wilson, 2011). Whereas both types of encoding are coherently provided for by recourse to the relevance-theoretic principles regulating the mechanics of comprehension (Sperber and Wilson, 1995),3 each of them has been assigned a distinct function in deriving

2 For a detailed critique of the validity of Anscombre and Ducrot’s (1977) defining properties of correction, the reader is referred to Iten (2005: 117–119). Our aim here is to expose the basic distributional differentiations of correction underlying the prospect of an ensuing differentiating treatment. 3 Here we mainly refer to the cognitive and the communicative principles of relevance, although we should add the definition of relevance, as well as the relevance-theoretic comprehension heuristics (Wilson and Sperber, 2004).

30

V. Bardzokas / Language & Communication 45 (2015) 27–45

the interpretation of an utterance. Roughly speaking, conceptual encoding has been held to correlate with the structure of lexical concepts formulating a sentence. On the other hand, procedural lexical items act as inferential constraints required for the hearer to bridge the gap between the linguistically determined aspects of meaning and the speaker’s intention in uttering them. In this sense, a procedural discourse connective serves the communicative purpose of guiding the hearer’s inferential process in the direction of the speaker’s intended utterance meaning at the least processing cost possible (Wilson and Sperber, 1993; Blakemore, 2002). Now, relevance-oriented accounts of discourse markers have traditionally referred to the distinction between conceptual and procedural meaning (Blakemore, 1987, 2002). More specifically, discourse connectives are commonly described in procedural terms, due to their facilitative operation in exposing to the hearer the ultimate interpretation of the utterance containing them. That is, to achieve relevance, the conjunctive encoding blends in with the hearer’s existing contextual assumptions to yield contextual effects, i.e. strengthening a current assumption, weakening an existing assumption or generating a contextual implication. In the process of this contextualization, the hearer constructs an interpretive assumption as to the speaker’s intended meaning: an assumption that fulfills the hearer’s expectations of relevance (sufficient cognitive effects with minimum processing cost). To elaborate, the use of a procedural connective is intended to lead the hearer to the above-mentioned contextual effects by constraining aspects of the deductive process, for instance either its premises or its conclusion-derivation phase. In this respect, the application of therefore, for example, in an utterance like Ben knows the combination; therefore, he can open the safe, aims to guide the hearer inferentially to the ultimate and relevant interpretation of he can open the safe as a conclusion. In a similar vein, in a case like Ben can open the safe; after all he knows the combination, the use of after all is taken to assist the hearer in establishing the relevance of after all, he knows the combination as representing one of the premises of the deductive inference (Blakemore, 2002). Similarly, the use of but in (6) achieves relevance by force of but q, i.e. but leave some for tomorrow, encoding the contextual effect of contradicting and eliminating the possible assumption conveyed implicitly by p, namely that the hearer could have all the pizza (Blakemore, 2002). (6) There’s pizza in the fridge, but leave some for tomorrow. In recent pragmatic approaches to meaning, there is a stark contrast between conceptual and procedural interpretation. Thus, on the whole, the ultimate interpretation of an utterance, as this is intended by the speaker, is aligned with its conceptual representation. This conceptual representation derives directly from decoding the lexical material of an utterance.4 On the other hand, as has been pointed out, procedural meaning is encoded to reduce the amount of inferential work required on the part of the hearer for the delivery of the ultimate representation of an utterance. As such, it does not end up being part of this representation (Sperber and Wilson, 1995; Wilson and Sperber, 1993; Blakemore, 2002). The type of meaning to be assigned to a lexical item can be determined against hard-and-fast criteria proposed by Wilson and Sperber (1993). For instance, as opposed to procedural encoding, conceptual encoding can act as input to logical inference rules, it enters contradiction relations and is thought of as bearing truth-conditional meaning.5 As such, in contrast to conceptual meaning, procedural meaning is not so accessible to the interpreter’s consciousness or amenable to explanation in terms of synonymy (Sperber and Wilson, 1995). As will be explained below, the advantage of the procedural account is that it displays wide applicability at no extra cost. In fact, it seems that a procedural account of but would straightforwardly assimilate the cases of denial (2), semantic opposition (3) and objection (4) (B). To begin with, in (2) the proposition following the occurrence of but in he is honest is taken to eliminate the assumption that John is dishonest, which is communicated implicitly on the contextual basis that Republicans are dishonest. In the same vein, the case of semantic opposition (3) can be reanalyzed in procedural terms subject to contextual enrichment, such as in (7),6 where the interpretation of but Mary is tall in (7) (B) seems contingent, not only on the use of John is short within the same sentence, but essentially on the preceding utterance (7) (A). (7) A: John and Mary are both quite short, aren’t they? B: Actually, John is short but Mary is tall. [example adapted from Iten, 2005: 112] Contrary to the suggestion that (3) represents a reading of semantic opposition that is impervious to pragmatic operations (R. Lakoff, 1971), the current conversational setting seems to adduce the contextual premises for the implicated conclusion that

4

This generalization allows for the exception of lexical adjustment and free enrichment (see Carston, 2002). However, the suggestion that truth-conditional and conceptual meaning are not necessarily coextensive is also discussed in detail (Wilson and Sperber, 1993; Ifantidou, 1994, 2001). 6 The integrity of R. Lakoff’s (1971) distinction between semantic opposition and denial of expectation has been questioned by a number of theorists (Abraham, 1979: 106; Kitis, 1982: 82; Foolen, 1991: 84; Winter and Rimon, 1994: 373). In fact, Kitis (1982: 82) and Foolen (1991: 85) attribute the interpretation of (3) as one of semantic opposition to the decontextualization the but-conjunction. 5

V. Bardzokas / Language & Communication 45 (2015) 27–45

31

Mary, too, is short. The use of the but-introduced clause to implicitly eliminate the assumption that Mary is short seems to render (7) (B) or (3) readily treatable by reference to a procedural account. Accordingly, in (4) (B) objection realizations seem treatable by appeal to the procedural mechanism of elimination (Rouchota, 1998; Iten, 2000, 2005). Thus, speaker B’s utteranceinitial occurrence of but in (4) (B), But I’ve just seen him in Oxford Street, is viewed as indirectly canceling the proposition in (4) (A), John is in Paris at the moment on the contextual grounds that one cannot be concurrently present in two different locations. Considering the segmented or conversational structure of (4), the use of the but-clause here is more appropriately taken to hinge on prior discourse or wider context, rather than some preceding clause (in a but-conjoined utterance). Up to the present point we have concluded that the set of but-varieties discussed above can be grouped together uniformly along the lines of procedural interpretation. On the procedural view, the but-introduced clause in each of the examples (2)– (4) (B) is justified as a means of eliminating an assumption that has been implicitly conveyed as a matter of p (within the bounds of the conjunction), or as a matter of the discourse in progress prior to p. But what about the structurally peculiar case of correction? The answer to our question is far from trivial. If the procedural analysis proves to apply equally to cases of correction as well, then, it can be said to secure an all-encompassing account of the meaning of but as a whole. In what follows, we will address the challenge of analyzing the special case of correction in procedural terms. 4. Problems connected to a procedural account of correction7 Relevance-theoretic analyses of the meaning of but (Blakemore, 2002; Iten, 2005; Hall, 2007), have focused on an exclusively procedural treatment, allowing no exception for the interpretation of correction. However, does the higher cause of a uniform account entitle them to a generalization of this sort? Let us begin to investigate this matter. As has been suggested above, a satisfactory procedural account involves the fulfillment of the following two-pronged criterion: a) to identify an implicitly communicated assumption (that undergoes elimination) and b) to attribute the elimination to the use of but. Let us consider the contextual situation of correction in (5) (B) (repeated below as (8) (B)) coupled with that in (9), in connection to this criterion. (8) A: Your sister looks a lot like you. B: That’s not my sister but my mother. (9) John isn’t going to Paris but to Berlin.8 [example taken from Jasinskaja, 2012: 1907] Here, the challenge of a procedural account resides in identifying the inferentially derived assumption that is eliminated by use of the phrase in but my mother (8) (B) or but to Berlin (9). What is denied in the utterance is the proposition that the woman referred to (henceforth C) is B’s sister (8) (B), or that John is going to Paris (9). But is the elimination of this proposition effected inferentially? Moreover, is this elimination the work of the but-introduced conjunct? Unfortunately, a set of positive answers would involve an unfavorable implication; to claim that the proposition that C is the speaker’s sister (or that John is going to Paris) is eliminated inferentially by way of the but-clause is to render the operation of the negation in p, i.e. -p (That’s not my sister/John is not going to Paris) vacuous or irrelevant to the denial of the proposition.9 But this is clearly an implausible claim. Suffice it to consider two facts: a) the negation of the proposition at hand remains intact even in the absence of but q, i.e. That’s not my sister/John isn’t going to Paris as well as b) the logical, rather than inferential character of the negation operator. How can we, then, keep up a procedural account of correction under these interpretive circumstances? Given that the negation in -p is specified linguistically (and, therefore, explicitly), one way to salvage the procedural account would be to identify an assumption eliminated inferentially, in addition to the one denied explicitly by means of the negation in -p. In this line of reasoning, two instances of denial in an example like (8) (B) or (9) would become available, i.e. communicated explicitly and implicitly, and the implicitly denied (or eliminated) assumption would be attributed to the realization of the but-introduced segment. In this way, the procedural account would be warranted. Where do we, then, look for this additional implicitly denied assumption? There are two directions in which to search. To start with, contrary to Blakemore’s (2000) construal of procedural but, we could extend the search space beyond the bounds of the conjunction, as, for instance, in cases of objection. Indeed, Iten (2005: 145) appears to take notice of the role of broader context. Interestingly, however, she draws a line to this possibility in view of the case of correction:

7 In this paper, the term ‘correction’ is only used tentatively in acknowledgment of the need for terminological consistency with the relevant literature. In fact, the claim made here is that the case of correction reduces to procedural encoding; as such, it is not translatable in notional, that is, conceptual terms. In this connection, we will argue that the ‘correction’ function of but is as procedural as the rest of the connective’s functions, e.g. denial of expectation. Thus, within the relevance-theoretic context of meaning analysis, but is treated as a discourse relation, in broad terms, used in various contexts to communicate procedural meaning. 8 Although this example is taken from a relevance theory-external source (Jasinskaja, 2012), we have included it in our critique of Iten’s (2005) view of correction for the sake of a lucid portrayal of the phenomenon. 9 The explicit character of negation in correction uses been pinpointed independently of relevance-theoretic investigations (Jasinskaja, 2012).

32

V. Bardzokas / Language & Communication 45 (2015) 27–45

[.] Blakemore’s (2000) claim that the [denied] assumption is ‘derived from the first [clause]’ is clearly too restrictive (it doesn’t apply to cases where there is no first clause). This could easily be remedied by saying that the but clause denies an assumption that is derived (or likely to be derived) in the context. However, there are a number of examples that are prima facie problematic for this account. The most obvious problem is posed by examples of correction but (Iten, 2005: 145).10 Iten’s exception for the variety of correction makes sense, considering a case like (8) (B) or (9), whereby the interpretation of correction (correcting speaker A’s mistaken assumption as to the identity of C, or correcting the speaker’s misconception as to John’s destination point) is obviously achieved within the co-textual bounds of -p, without recourse to further contextual resources. This observation effectively rules out the option of exploiting the parameter of the broader context or prior discourse in determining the implicitly denied assumption needed to accommodate correction procedurally. It seems, then, that we are necessarily left with the remaining alternative: to seek the implicit denial within the co-textual bounds of -p in the utterance in (8) (B) or (9). But what would the implicitly denied piece of information consist in? Looking for an answer within the restrictive bounds of (8) (B), Iten (2005) considers the prospect of a single clause, i.e. That’s not my sister in -p, receiving distinct interpretations in parallel: a) the negation explicitly expressed in -p That’s not my sister, and b) the affirmative counterpart of -p, i.e. That is my sister, communicated implicitly by virtue of the negation itself. In this way, once again, an instance of implicitly communicated denial is secured (the denial of That is my sister in (8) (B) or that of John is going to Paris in (9)) that can be attached to the operation of but. But the question that arises is how a single occasion of utterance processing can yield both of these interpretations at once, i.e. an affirmative interpretation and its negative counterpart, without resulting in contradiction. To solve this problem, Iten (2005) appears to pursue her proposal by drawing a distinction between the notions of manifestness and accessibility. “It is widely accepted that utterances expressing negative propositions make accessible their positive counterparts. Thus, That’s not my sister might not make manifest the assumption that the woman in question is the speaker’s sister, but it certainly makes it accessible” (Iten, 2005: 145). Thus, on Iten’s view, it can be argued that the logical falsification of the assumption that C is the speaker’s sister makes manifest the negation of the assumption, but it provides simultaneous access to its affirmative equivalent. But how exactly does the distinction between manifestness and accessibility serve Iten’s procedural considerations of correction? In relevance-theoretic terms, an assumption that is communicated is an assumption that becomes mutually manifest to the conversationalists, who can accept it as true or probably true (Sperber and Wilson, 1995). Presumably, in drawing this distinction in communicated content, i.e. between manifestness and accessibility, what Iten (2005: 145) has in mind is an attempt to circumvent the contradiction involved in her claim, i.e. that both the positive and the negative version of a single assumption can be communicated in parallel and, thus, made equally manifest. In this sense, unlike a manifest assumption, an accessible assumption may be entertainable without necessarily being represented or accepted as true or false. Therefore, Iten’s argument seems to suggest that the denied assumption communicated, namely that C is not speaker B’s sister or that John is not going to Paris, does not preclude access to the assumption that C is the speaker’s sister or, correspondingly, that John is going to Paris. In theory, then, Iten’s way secures a chunk of information (That is my sister/John is going to Paris) whose elimination is made accessible by virtue of the but q. In this sense, the but-q operates independently of the explicit negation in -p. Iten’s claim seems theoretically interesting, but the question that emerges from the previous discussion is whether it is tenable. As it stands, the particular argument seems to transpose the problem of tracking down an assumption denied implicitly from context-level to co-text level analysis. The problem is more accurately phrased in the following question: how can the opposite assumption of the assumption that has been explicitly, and thus, strongly denied continue to arise inferentially (or to gain accessibility) within the same utterance? The paradox in Iten’s suggestion seems to be that while a proposition is logically negated, it is also accessed inferentially in its affirmative version, only to be doubly denied, this time by the operation of but. Apart from its unnecessary perplexity, this argument lacks psychological plausibility. Why should the hearer bother to access the exact opposite assumption of the one asserted? In her critique of the specific suggestion, Hall (2007) appears to share the same reservation. In fact, under the strain of this counter-evidence, Iten’s (2005) second thoughts about dropping the requirement for accessibility is understandable. However, to abandon this argument is to abandon the final resort in our endeavor to distinguish two denied assumptions and link the implicitly derived one to the application of but. In the end, we basically return to the original problem of identifying the inferred assumption that the use of correction but attaches to, in the pervasive presence of negation. Having dropped the suggestion of accessibility, Iten (2005: 149–150) (also Blakemore, 2002: 111–112) seems to make a final, though rather fruitless attempt to secure the availability of an inferentially communicated assumption denied by the use

10

Ducrot’s (1984) work seems relevant to Iten’s comments.

V. Bardzokas / Language & Communication 45 (2015) 27–45

33

of but. The way the argument runs, it seems that the proposition in q contributes to the contradiction and elimination of the (already negated) proposition in p, i.e. That is my sister/John is going to Berlin, by providing further evidence for the explicitly eliminated proposition. It is, thus, claimed that It thus makes sense for B [in (8) B] to use a but clause to contradict and eliminate the assumption the first clause hasn’t quite succeeded in eliminating. The assertion that C is B’s mother is likely to succeed in eliminating A’s mistaken assumption: after all, it’s rare (if not entirely impossible) for one and the same person to be someone’s mother and sister and thus the assumption that C is B’s mother is good evidence for her not being B’s sister. This shows that, even in correction cases, but is used if the speaker thinks there is a danger that an ‘undesirable’ assumption could be made manifest to the hearer (Iten, 2005: 149–150). Nevertheless, this argument, too, seems to contain flaws. Firstly, it is based on the suggestion that the negation employed in p, for some reason does not succeed in eliminating the assumption that C is B’s sister. Indeed, the hearer may still entertain ‘undesirable’ assumptions following the utterance of p and, in this sense, may need to be convinced of speaker B’s honesty. But does the hearer remain unconvinced because the negation failed to achieve elimination? Obviously not. In this light, then, Iten’s account appears to assume that a reoccurrence of the elimination is required, this time by means of employing the butintroduced conjunct. However, if repetition made a difference in convincing the audience of their misidentification, then a reiteration of -p alone, i.e. That’s not my mother.That’s not my mother, might achieve the desired effect. Indeed, we might embrace the suggestion that but q provides further evidence for the explicitly eliminated proposition i.e. That is my sister/John is going to Berlin. But does this further evidence amount to further contradiction or elimination? Is the effect of elimination gradable, in the first place? If it were, then, a more accurate account of the contextual effect involved in correction would seem to be sketched in terms of but q being used to strengthen the elimination effected (explicitly) in -p, rather than repeating the elimination of the same assumption. In face of the predicament that these questions present, we surmise that it is not the contradiction/elimination that but q contributes to. What does it contribute to, then? If Iten’s issue of persuasion has an actual part to play in correction analysis, it is reasonable to argue that, the hearer is convinced by use of elaboration, rather than double denial. In other words, but q is seen as elaborating on the denial of p, i.e. That is my sister/John is going to Paris. And it is precisely this sense of elaboration that the material in q, i.e., my mother, to Berlin, seems to work towards. To this effect, it may be recalled that the denial of p is proved by the linguistic fact of negation, and, as such, does not require evidence. Has, then, our discussion of a procedural approach to correction reached an impasse? We need to examine one last claim in reaching our conclusion. Blakemore (2002) offers what seems to be a partial solution to the problem of a procedural analysis. She appears to adopt a rather holistic attitude to an utterance of correction, so as to overcome the recalcitrant issue of explicit elimination in -p. To be more precise, Blakemore argues that both -p and q are equally involved in the process of the same inferential elimination. In the example that she offers (10), for instance, (10) He’s not clever but hardworking. [Blakemore 2002: 110] the assumption that gets eliminated (He is clever) is (implicitly) communicated by the joint presence of both conjuncts (He’s not clever and but hardworking), considering that being hardworking is not incompatible with being clever (Blakemore, 2002: 110–111) and, thus, both qualities can describe the same person. Hence, In making this interpretation [He is clever] salient for the interpretation of the segment it introduces, but is also making it salient for the first segment. The effect is to make [He is clever] manifest to the hearer for the interpretation of both segments. In other words, what distinguishes utterances such as [(10)] from other utterances involving but is that the interpretation of each of its segments involves an inferential procedure that results in the contradiction and elimination of the same assumption (Blakemore, 2002: 111–112). However interdependent the operations of the two segments may be, the particular argument continues to be vulnerable to the same shortcoming. In face of the negation encoded in -p, we are still left wondering as to how He is clever is made manifest or salient and what the ‘inferential procedure’ resulting in this assumption consists in. Despite her reassurances to the contrary, Since the first segment of [(10)] is denying the truth of [He is clever], either [He is clever] cannot be understood as an assumption eliminated by the but segment, which it clearly is, or we must re-analyze but so that the inferential procedure it activates does not necessarily have to result in the elimination of a manifest assumption (Blakemore, 2002: 111). it is unclear how the interpretation of [He is clever] comes through inferentially in the linguistic context sustaining the exact opposite interpretation. To make matters worse, Blakemore’s view of correction seems to have limited application. To be more precise, her account of (10) seems based on the suggestion that the two competing attributes i.e. clever and hardworking, are compatible. However, consider the following example:

34

V. Bardzokas / Language & Communication 45 (2015) 27–45

(11) He’s not clever but dumb.11 The possibility of -p, i.e. the negation of being clever, and q jointly communicating, or making salient, or making manifest the assumption that the person referred to is clever would run counter to our intuitions, to say the least. These examples (10)– (11) will be discussed further in the process of advancing our proposal. To conclude our discussion,12 we have so far attempted to demonstrate the implausibility of a procedural account of correctionbut in concordance with the traditional relevance-theoretic prescriptions. We have attempted to show that the criterion of the availability of inferentially delivered elimination does not obtain. In this connection, previous attempts to solve the problem of correction on regular procedural stipulations result in implausible, strained arguments lacking in descriptive rigor. In the present section we ruled out the option of correction-but operating procedurally by virtue of eliminating information that is communicated implicitly in uttering an example like (8) (B) or (9). In this line of evidence, we are left with two choices: either to reconsider the relevance-theoretic stipulations of the procedural operation of correction-but or turn from the procedural approach. At this point we wish to put our conclusion into perspective; to rule out the elimination-oriented procedural account is not to exclude the procedural approach to correction altogether. In fact, an effective description of all previous examples of correction (8) (B)-(11) could lie in Blakemore’s prophetic, as it were, mention of a prospective alternative in the final lines of the quotation above, i.e. “[..] or we must re-analyze but so that the inferential procedure it activates does not necessarily have to result in the elimination of a manifest assumption” (Blakemore, 2002: 111). Regrettably, Blakemore remains reticent about the specifics of this alternative, which makes her comment open to interpretation. We will return to this alternative after we have considered the robustness of further procedural options that the relevancetheoretic toolkit provides us with. The alternative offered below addresses the explicit, rather than the implicit side of the divide between implicit/explicit content. 5. A procedural constraint on higher-order explicatures Up to now we have examined the plausibility of a procedural analysis of correction in terms of the impact that the use of but has on implicitly communicated content. More generally, however, a thorough relevance-theoretic examination of procedural discourse markers seems to make reference not only to linguistic constraints on implicatures, but, also, to the distinction between explicit and implicit meaning, as has been documented in the relevant literature (Wilson and Sperber, 1993; Blakemore, 2002). Thus, whereas procedural discourse markers have typically been described in terms of encoding a constraint on implicit meaning, i.e. implicated conclusions or implicated premises within the deductive operation,13 (as in the case of but, therefore, after all), the possibility of a procedural constraint on explicit content has also been showcased. For example, in (12), (12) Peter is a genius, huh, the presence of the particle huh is intended to reverse the literal meaning of the proposition ‘Peter is a genius’, by constraining the speaker’s dissociative (ironic) attitude to the proposition. In this sense, the particle exposes a contextuallyderived higher level of explicit meaning, in other words a higher-order explicature (rather than an implicature), which engulfs the literal, propositional content of the utterance. In addition to the case of discourse particles, it has been shown that discourse connectives, too, and, in particular, causal subordinators such as because, can be analyzed procedurally (Kitis, 2006; Bardzokas, 2012, 2013) in view of their effect on the inferential derivation of higher-order representations (Iten, 1998) in given, non-propositional contexts, i.e. epistemic, speechact (Sweetser, 1990), or metacommunicative connections (Kalokerinos, 2004).14 Other interesting works in causal meaning are traceable to Zufferey (2012), Zufferey and Cartoni (2012), In this line of evidence, the next question to pose seems fairly evident; could the case of correction-but be classified as a procedural constraint on explicit, rather than implicitly communicated content, and, more specifically, on explicatures of the higher order? But a higher-order-explicature-oriented procedural analysis of correction suggests the relevance of higherorder explicitness to our discussion, in the first place. Is this level of interpretation of relevance, after all? In answer to this question, our primary concern is to identify the proposition expressed in the conjunction that is assimilated in the higher-order explicature carried by the use of correction but. Can it be suggested that, for instance, That’s

11 The specific example is taken as an emphatic denial of the referent’s possession of cleverness, rather than an ironic/sarcastic comment, although the latter reading is not improbable in given contexts. 12 In a bid to propose a more sustainable procedural account of but, Hall (2007) pursues a reconceptualization of the connective. Nonetheless, she mainly concentrates on varieties other than correction (Hall, 2007: 171–172). In this respect, suffice it to say that the plausibility of her argumentation notwithstanding, we will not take up this issue in the current work. 13 The reader is referred to Sperber and Wilson, 1995 and Blakemore, 2002 for a more detailed account of the operation of the deductive device. 14 In a nutshell, John loved her, because he came back or What are you doing tonight, because there is a good movie on are taken to instantiate nonpropositional causal connections in contradistinction to an example like John came back because he loved her expressing a ‘real-world’ (Sweetser, 1990), cause-and-effect or propositionally explicit causal connection. In the non-propositional cases, the subordinate clause is meant to reveal a higher-order explicature, e.g. He thinks that John loved her because he came back, She’s asking him what he’s doing tonight because there is a good movie on, restoring the underspecified causal content of the connections.

V. Bardzokas / Language & Communication 45 (2015) 27–45

35

not B’s mother but B’s sister, as a whole, is interpretable in terms of a speech-act description, e.g. correction or clarification? Consider, for example, integrating (8) (B) into the following higher-order assumption: Speaker A clarified that C was not A’s sister but A’s mother. On the surface, an integration of this sort seems possible, since it is supported by the syntax of correction, say in (8) (B). That is, the elliptical, subsentential syntax of an otherwise bi-clausal utterance seems to license, even urge the integration of (8) (B), as a whole, into a compound formulation, i.e. a single higher-order assumption schema (in this case The speaker clarified/explained that.), as is the case with single clausal units. However, a syntactic regularity of this sort does not compel an overall representational conflation between -p and q. As has been widely acknowledged (Carston, 2002), a linguistic expression is drastically underdetermining of the speaker’s meaning, the determination of which necessitates the propositional development of the linguistic expression. As has already been pointed out, the interpretation of the utterance in (8) (B) comprises a fully propositional link conjoining the clauses in -p, i.e. C is not A’s sister, and q, i.e. C is A’s mother. Along these lines, the interpretation of (8) (B) is more accurately accounted for as involving, not a representational conflation of -p and q into a single higher-order assumption, but the integration of each one of these propositions into a separate higher-order explicature, for instance denying or protesting, on the one hand, and elaborating or clarifying, on the other, among a number of interpretive choices. Presumably, the suggestion of two distinguishable higher-order representations (attached to -p and q) tallies with Jasinskaja’s (2012) work on correction applications of but. Her view of correction originates with an argumentation-theoretic approach to adversative markers (Jasinskaja and Zeevat, 2008, 2009), along the lines of Winterstein (2010), but also develops further around a reaction to its shortfalls. At any rate, the core argument seems of relevance. In short, discourse markers like but are, in principle, thought of as being used to convey information ‘about the discourse topics addressed by the clauses they connect, where discourse topics can take the form of questions [.] along the lines of Roberts (1996) and Büring (2003)’ (Jasinskaja, 2012: 1906). In acknowledgment of the deficiencies of the argumentative approach with specific reference to correction cases, Jasinskaja seems to incorporate a question-based component into it, so as to accommodate examples like (9), i.e. John isn’t going to Paris, but to Berlin in a unified framework of analysis. In corrections such as (9)],15 the question is [.] a simple wh-yes/no [.]: where “whether” John is going [.]. The first conjunct addresses the subquestion Where isn’t John going? – he isn’t going to Paris; the second conjunct addresses the subquestion Where is John going? – he is going to Berlin [.]. The requirement that the instantiations of the yes/novariable be distinct ensures that one conjunct is positive, while the other contains syntactic negation (Anscombre and Ducrot, 1997) [.]. On this account,16 the (implicit) question-answer pattern proposed suggests the occurrence of discreet, fully fledged responses encoded in the conjuncts. These responses, in turn, can be claimed to yield distinct speech-act interpretations, or, alternatively, distinct higher-order explicatures. For instance, the first conjunct is primarily interpreted as denying the assumption that John is going to Paris, as opposed to the accompanying conjunct, which opportunely justifies or elaborates on the denial. Up to now we have argued that the suggestion of a representational conflation of -p and q under a single higher-order explicature would effectively distort our view of correction. In this light, what options do we have left of analyzing correction in terms of higher-order explicit content? To the extent that -p and q can be interpreted individually as attaching to a separate higher-order explicature, the use of but can be said to constrain either one of them. To espouse this argument, however, it is essential that we justify the occurrence of but in relation to contributing to a higher-order explicature that is not independently deliverable by the use of -p or q in an -p. q version of correction. Unfortunately, providing a justification of this sort is no foregone conclusion, considering the connectiveless type of conjunction involved in this case. We are, thus, inclined to exclude the notion of higher-order explicatures from our considerations of correction. Still, our reflections of higher-order explicatures in this section do not complete our discussion of explicit content and its potential role in the procedural description of correction but. Actually, there is one more level of explicit representation that comes into view, in this connection: explicatures of the basic order.

6. A constraint on base-order explicatures This section follows up on our discussion of explicitness. However, on this occasion, it considers the possibility of a procedural treatment of correction in terms of explicit meaning of the basic order. In fact, a thorough consideration of explicit representation includes the propositional content of the utterance, that is, its base-order explicature, apart from its higher-order explicature. The rationale underlying this comprehensive view of explicitness is that, in a cognitively plausible line of thought, pragmatic inference affects the determination of both levels of explicit content (Sperber and Wilson, 1995; Carston, 2002).

15

The numbering has been adjusted to the requirements of the current presentation of data. The current paper cannot do justice to the arguments presented in Jasinskaja (2012). For a detailed account of correction in her question-based system, the reader is referred to the original work. 16

36

V. Bardzokas / Language & Communication 45 (2015) 27–45

The first question that seems to arise is whether it is even possible to link this level of explicitness with procedural encoding. Pursuing an answer to this question suggests a prior exploration of the extent to which the basic level of explicit meaning is applicable to the analysis of correction. We begin this exploration below. To elucidate the pragmatic impact on propositional (truth-evaluable) content, we may recall cases of referential indeterminacy and semantic disambiguation. Leaving aside these widely acknowledged pragmatic phenomena, the range of inferential operations on propositional content seems to be substantially wider. More specifically, Carston (2002, 2004, 2009) discusses pragmatic determinants of propositional content in the terms of base-order explicatures operating under the labels of ‘free pragmatic enrichment’ and ‘ad-hoc concept construction’. The following examples typify these two types of baseorder explicature, respectively. (13) It’s the same [as what?] (14) Ann wants to meet a bachelor. [Carston 2002: 22] The illustration in (13) achieves full propositional interpretation, subject to the derivation of the base-order explicature reflected in the bracketed question, i.e. A is the same as B. Although the pragmatic process involved in free enrichment does not seem mandated by the logical form of the sentence, the same is not true of ad hoc concept construction cases, where a lexical concept undergoes inferential adjustment to a broader (looser) or narrower (more specific) concept. Example (14) reflects a case of lexical adjustment. Consider a contextual scenario where Ann is interested in marriage. The determination (and ensuing truth evaluation) of the proposition expressed by the utterance appears to hinge on the contextual narrowing of the encoded denotation of ‘bachelor’ into the more specific concept of ‘not any bachelor’, but one who is considered eligible for marriage. So far we have illustrated the role of base-order explicatures in establishing explicit meaning. But, in connection to our current discussion, can this level of explicature relate to a meaning description of correction (setting aside procedural considerations at this point of our analysis)? We might embrace this prospect in light of concrete evidence originating with research on discursive meaning. Causal studies (Bardzokas, 2012, 2014) have shown that the realization of because-introduced subordinations may be justified as a constraint on the retrieval of inferential elements of the propositional content in the main clause. To take a brief detour for the sake of illustration, contrast the following examples (15)–(16). (15) John came back because he loved her. [Sweetser, 1990: 74] (16) The boss demoted Mickey because he was needed. (17) The boss [simply17] demoted Mickey because he was needed. [Bach, 1994: 129] The causal interpretation in (15) seems to contrast with that in (16), in the sense that, unlike (16), it seems to receive the interpretation of a propositionally explicit cause-and-effect relation between John’s emotional state and his return. In this sense, the corresponding propositions are encoded with a view to expressing a cause-and-effect (or ‘real-world’, in Sweetser’s (1990) terms) interpretation of the sort ‘the reason why John returned is that he loved her’. By contrast, in the case in (16), Mickey cannot, under ordinary interpretive circumstances, have been demoted on account of the fact that he was needed. In this case, rather than involving the proposition of the main clause in a cause-and-effect relation, the conjunction is here deemed especially justified relative to revealing a base-order explicature in the main clause, i.e. the unarticulated constituent simply, as in (16) (Bardzokas, 2012, 2014).18 Having elucidated the potential effect of a discourse marker on base-order explicatures, we can now turn to the next issue that calls our attention. Is it possible to take this level of representation on board in accounting for correction cases? Consider, once again, examples (8) (B) and (9), repeated below. (8) A: Your sister looks a lot like you. B: That’s not my sister but my mother. (9) He’s not going to Paris but to Berlin. For starters, setting aside the case of reference fixation, any attempt to explain the relevance of but vis-à-vis the derivation of a base-order explicature within the bounds of -p seems futile in either example. For instance, whatever unarticulated elements in the proposition That’s not my sister/He’s not going to Paris are said to be recoverable, they are not attributable to but

17 18

By simply we mean went only as far as to just demote him. See also Blakemore (1987) for acausal interpretation of therefore.

V. Bardzokas / Language & Communication 45 (2015) 27–45

37

q. Moreover, all lexical concepts seem to retain their encoded sense. Could, perhaps, the derivation of a base-order explicature be sought within the bounds of the but-introduced conjunct? At this point it is worth recalling that the particular conjunct contains ellipsed material that is inferentially recovered. Is it, then, possible to describe the propositional development of the ellipsed material in q in terms of base-order explicature derivation, owing to the realization of but q? Unfortunately, ellipsis cannot be explained in these terms. By way of reminder, it is in the absence, rather than the presence of the marker at hand that the propositional expansion can occur, if the interpretation of correction is to be maintained. Otherwise, a fully articulated version of the conjunction that preserved the use of but would suspend the interpretation of correction, resulting in a denial of expectation meaning: That isn’t my sister but it is my mother, John is not going to Paris but he is going to Berlin, as has already been pointed out. Following this discussion, it transpires that correction analysis does not access explicit content of the basic order. Nevertheless, this conclusion might be contested on account of the choice of examples under discussion. Indeed, so far we have continued the tradition (in the literature) of scrutinizing timeworn examples, rather than presenting richer data. This type of research practice, however, may ignore contextual parameters instrumental to a more refined description of correction. We will look further into this matter in what follows, but mainly in due course (Section 7). To embark on our discussion of the necessity for further contextual considerations, let us recall Blakemore’s (2002) example in (10), i.e. He is not clever but hardworking. Here, the utterance makes good sense in a context whereby, for instance, the conversationalists discuss an average student’s unsuccessful application for a position in a postgraduate program. Conversationalist B explains to A the reasons for the student’s success at different exams in the past. (18) A: He passed all his previous exams, didn’t he? He must be clever. B: He is not clever but hardworking. Now contrast the case in (18) (B) with that in (19) (B). The above-mentioned student’s application has been successful. The student’s demonstration of commendable qualities is undisputed. (19) A: He is clever. B: He is not [only] clever but brilliant. In the context in (19), the interpretation of speaker B’s utterance as a whole relies on the derivation of the explicature, i.e. only.19 Indeed, the reading of -p would be deprived of this propositional enrichment, in the event of leaving out the butintroduced conjunct. Unlike the situation in (19) (B), the reading of the utterance in (18) (B) does not appear to involve such a pragmatic task. This new piece of evidence constricts our previous generalization with respect to the involvement of base-order explicatures in our discussion of correction. As it turns out, correction may, at least partly, be discussed in connection to free enrichment. But how is the occasional, rather than constant presence of a base-order explicature in -p to be more methodically accounted for? Firstly, it seems that the distinction in propositional interpretation between (18) (B) and (19) (B) ultimately reflects a distinction in the scope of the negation in -p. That is, while the negation in -p includes a base-order explicature (e.g. only) in its scope, in (19) (B), rather than just the adjective clever, it comprises no such element of meaning in (18) (B). Also, examples such as (19) illustrate that the sort of correction that is amenable to a base-order-explicature driven analysis contains gradable adjectives of the same family. Compare, for instance, an added illustration, as in (20): (20) A: She is happy. B: She isn’t [only] happy but ecstatic. Here, the material that seems to undergo negation in (20) (B) is ‘only happy’, rather than just ‘happy’. Once again, this interpretation reflects the case of a free enrichment emerging from the use of but ecstatic. It appears, then, that the interpretation of correction is more responsive to contextual factors than originally envisaged. In fact, this point has not yet been stressed enough, as will become clear in due course. So far we have only shown that, depending on the context, the interpretation of correction occasionally involves the communication of a base-order explicature within the scope of the negation in -p. We will elaborate on the role of negation in correction analysis below (Section 7). At this point, suffice it to say that, in view of the aforementioned distinction in the scope of negation, it seems that we may group the case in (18) (B) together with (8) (B), (9) (B), and (11) (B), while distinguishing this group of cases from that in (19) (B). Having established all this, we can now resume our initial issue: does the operation of the base-order explicature in, say (19) (B) or (20) (B) suggest a (partial) treatment of correction on procedural grounds?

19 The explicature in the but-introduced conjunct, i.e. also, is basically seen as being derivative from the interpretation of the first conjunct, i.e. He is not only clever. In this connection, we will mainly attend to the explicature that appears in the first conjunct, i.e. only.

38

V. Bardzokas / Language & Communication 45 (2015) 27–45

However the reply to this question may be, cases like (19) (B) or (20) (B) represent a rather limiting class of correction applications. In this connection, rather than rushing into the description of a prospective procedural link between these examples and base-order explicit content, we find it purposeful to first explore the extent to which they can be handled in more general and unifying procedural terms at no extra cost. Thus, if it turns out that these cases bear procedural meaning, it is likely that they do not do so by virtue of the operation of base-order explicatures per se, but, rather, because of procedural operations that apply more generally to the entire range of correction cases. We will take up this issue below (Section 7), whereby we offer a reconceptualization of the contextual effect of elimination as a source of procedural meaning (with respect to definite applications of but). To recapitulate the results of our procedural enquiry of correction up to this point, we have offered a number of suggestions exploiting the distinction made between implicitly and explicitly communicated meaning. These suggestions have turned out to be untenable. In the first place, no procedural account can be pursued by allusion to linguistic constraints on the inferential process of elimination. Also, we have ruled out the prospect of an analysis (procedural or otherwise) in terms of the marker affecting explicit content. This point applies, as a whole, to the case of higher-order explicatures and also, in part, to the case of base-order explicatures. Lastly, our discussion of the relevance of base-order explicatures has revealed the role that the (scope of) negation encoded in p assumes in refining our account of correction. Let us turn to this last point in resolving our issues of correction. Having ruled out all these options of a correction analysis, is it not high time that we abandoned the hope of a procedural application, along with the suggestion of a univocal account of but? In fact, we wish offer an alternative that restores this hope and that has, interestingly enough, gone unnoticed in traditional procedural accounts of correction (Iten, 2000, 2005; Blakemore, 2002). 7. The procedural account revisited: preliminary remarks on negation20 In previous sections we illustrated a rather shortsighted perspective taken by past relevance-theoretic accounts of correction, which prevents us from observing interesting contextual aspects of the specific variety of interpretation. Let us attempt to shed some light on them. For instance, as has been observed, the component of negation is integral to the correction construction, not only in structural terms, that is as a distinctive syntactic feature of correction, but, also, as the semantic means of eliminating a misconception on the part of the hearer. In this connection, the common tendency (in correction analyses) to pay scant regard for the identity of an obvious linguistic fact is rather surprising, at first sight. It starts, however, to make better sense considering the analysts’ devotion to the cause of accentuating the application of inferential elimination in procedural accounts of but, so that the interpretation of correction is uniformly accommodated. A biased perspective of this sort, however, can be said to obscure the potential contribution of negation, a linguistic type of contribution that resists the eliminationbased, procedural treatment. Our current observation will attempt to take remedial course of action, on this score, taking account of the role of the negation encoded in -p and, specifically the role of a distinction that figures prominently in the literature on the pragmatics of negation (Carston, 2002), i.e. descriptive vs. metarepresentational21 (see also Moeschler, 2013; Kaup and Zwaan, 2003). The distinction is illustrated below: (21) We didn’t see the hippopotamuses. But we did see the rhinoceroses. (22) We didn’t see the hippopotamuses. We saw the hippopotami. [Carston, 2002: 267] The utterance in (21) represents the case of descriptive negation, in the sense that the predicate included in the scope of negation is used literally to describe some aspect of the world. On the contrary, a descriptive reading of the example in (22) would result in a contradiction of the sort ‘We didn’t see p; we saw p’. Therefore, the utterance is more meaningfully understood as an objection to some non-descriptive aspect of the material falling within the scope of the negation. The point of objection here is the plural suffix for the noun hippopotamus. (Carston, 2002: 267–268).22 So far, so good but how is this distinction in negation relevant to our study of correction? We will address this issue below.

20 The reader may form the impression that our discussion of negation and its role in resolving the issue of correction but is presented rather late in the paper. However, there is a rationale behind this decision. The paper proposes negation as a way out of the problems besetting previous accounts of correction in the current literature. Since these accounts largely ignore the contribution of negation to the meaning of correction, a review of them can only reflect this problem, i.e. the absence of the role of negation, rather than hide it. 21 The term ‘metarepresentational’ is employed by Carston (2002) to do general duty for Horn’s (1985, 1989) suggestion of metalinguistic negation as well as her suggestion of ‘metaconceptual’ negation, the latter case describing instances of negation as a response to a thought rather than a verbalized utterance. 22 In Carston’s terms, this distinction in meaning ultimately merges into a single semantics under an echoic interpretation that involves a pragmatic enrichment of the truth-functional meaning of the negative operator. Be that as it may, the distinction continues to be potentially purposeful in correction analysis.

V. Bardzokas / Language & Communication 45 (2015) 27–45

39

7.1. A procedural account of metarepresentational correction As has been shown, correction conjunctions integrate a negation component in -p. Can it, then, be argued that the distinction commonly drawn in the scope of negation, i.e. descriptive vs. metarepresentational, transfers to the analysis of correction? After all, examples of negation, e.g. in (21)–(22) can be worded in the co-textual terms of correction, with the proviso of a minor lexical reformulation using but (and ellipsis); we didn’t see the hippopotamuses but the rhinoceroses, we didn’t see the hippopotamuses but the hippopotami. Indeed, it seems that the proposed distinction can be retained in correction interpretation. Examples (23)–(27) reflect instances of descriptive correction, in contrast to (28)–(32) depicting metarepresentational correction. The metarepresentational examples in (28)–(31) have been taken from Carston (2002: 295), though appropriately adjusted to the cotextual environment of correction. To these examples, we add the utterance in (32), which seems to be readily accommodated on metarepresentational grounds, on account of the non-descriptive aspect of material, i.e. base-order explicature, that the scope of -p includes. Illustrations (23)–(26) are reiterated for the sake of immediate contrast. (23) That’s not my sister but my mother. (24) John is not going to Paris but to Berlin. (25) A: He passed all his previous exams, didn’t he? He must be clever. B: He is not clever but hardworking. (26) A: He flunked the test. I always thought he was clever. B: He is not clever but dumb. (27) The horse is not black but white. (28) Jane doesn’t eat tom[eɪDouz] but tom[a:touz]. (29) We didn’t see hippopotamuses but hippopotami. (30) I won’t deprive you of my lecture on negation but spare you it. (31) Poor old Mr. Dean’s not a bachelor but an unmarried man. (32) Mary’s not [only] happy but ecstatic. Once the distinction in correction interpretation is observed, our account opens up new vistas in standard (eliminationbased) procedural analysis. We will first focus on the latter selection of examples (28)–(32), deferring our discussion of (23)– (27) to the following section (subsection 7.2). A careful examination of the data above indicates that the set of metarepresentational examples (28)–(32) contrasts with the set of descriptive examples (23)–(27) in an important respect. Let us explore this issue further. Can the latter set of examples (28)–(32) receive a procedural treatment? And if it can, can we speak of an eliminationbased treatment, along the standard procedural lines of analyzing but, in general? It is very likely that we can. To assess the plausibility of this clam, let us first recall the two requirements that need to obtain for a procedural qualification of but. The first requirement concerns the availability of an inferentially eliminated (or denied) existing assumption, while the second one predicts that the operation of elimination is attributed to the use of the but-introduced conjunct, rather than the negation effected in the -p conjunct. Starting from the first criterion, the interpretation in (28)–(32) is actually taken to rely on the elimination of an inferentially triggered existing assumption. To take the example in (28), apart from decoding the negated proposition, i.e. Jane doesn’t eat tomatoes, an exhaustive interpretation of the first conjunct can be said to include the derivation of a further, underlying output assumption or contextual implication (about Jane’s dietary proclivities), i.e. that Jane prefers other vegetables to tomatoes, or that tomatoes do not agree with Jane, or that she has an allergic reaction to tomatoes. Obviously, by our suggestion of an underlying assumption, we do not refer to the positive counterpart of the negated proposition, à la Iten (2005). In fact, Iten’s approach, in this respect, would leave us in the dark as to the rationale of making manifest the same assumption, i.e. Jane eats tomatoes, in both conjuncts of a but conjunction,23 particularly in face of the delivery of a metarepresentational interpretation. At this point, our argumentation might benefit from elucidating the constraining function of but more concretely in the course of deductive processing. The proposition contained in the first conjunct of (28), acts as input to the deductive device deployed in processing the utterance Jane doesn’t eat tomatoes. This proposition combines with existing contextual assumptions. This contextualization process can be said to generate a contextual implication or output assumption that is

23

Put differently, why do we need to both decode and infer the same assumption, especially in a correction conjunction.

40

V. Bardzokas / Language & Communication 45 (2015) 27–45

relevant to our optimal understanding of the original input assumption, e.g. Jane prefers other vegetables, Jane is allergic to tomatoes, to repeat but a few instances. Having supplied this contextual information, the hearer expects that the information following -p (in but q) will serve to keep up the descriptive or literal understanding of the denial in -p, e.g. ‘Jane doesn’t eat tomatoes but cucumbers’. However, processing the remaining conjunction in but q reveals a metarepresentational (in this case phonetic) clue that eliminates the hearer’s expectation of a continued descriptive reading, along with the inferred assumption that Jane prefers other vegetables. From Jasinskaja’s (2012) point of view, an exhaustive interpretation of a descriptively interpreted case, such as Jane doesn’t eat tomatoes but cucumbers, would presumably involve addressing the general implicit question of What (vegetables) “whether” Jane eats, the subquestions addressed by -p and but q being What doesn’t Jane eat and What does Jane eat?, respectively. The metarepresentational case in (28), however, topples the hearer’s expectation for a regular descriptive reading of -p but q (on which the general question of What (vegetables) “whether” Jane eats would be addressed). In this sense, contrary to the hearer’s expectations, the but q conjunct does not serve to contribute an answer to the subquestion What does Jane eat? To ground this argument to our terms of analysis, the hearer’s expectation for addressing the topic-question of What (vegetables) “whether” Jane eats or, more specifically, of the subquestion What does Jane eat? can be said to emerge from the underlying contextual assumption (communicated by the use of -p) that Jane prefers different vegetables/food. Thus, in eliminating this assumption, the use of but q also topples the expectation of answering the foregoing question. Jasinskaja makes the following relevant note of correction cases, although with reference to descriptive (24), rather than metarepresentational cases: [.] the negative conjunct in corrections like (24) often [.] contains linguistic devices that bear a presupposition that John is going to some place (Jasinskaja, 2012: 1905).24 To this comment she also adds that: This suggests that the proposition that John is going somewhere is normally a settled fact among the speaker and not under dispute in such cases (Jasinskaja, 2012: 1905). Jasinskaja does not specify whether the notion of presupposition is employed here in its technical sense. However this may be, we will opt for the broader relevance-theoretic notion of contextual assumptions, if anything, to include a wider range of relevant contextual assumptions, even weaker implicatures, derivable in using the first conjunct, say in (28): she had (other) vegetables, she is allergic to tomatoes, and so forth. Turning to the second criterion of a procedural account of but, since it is the negation contained in -p that contributes to the communication of the underlying assumption, it cannot undertake the conflicting, simultaneous task of eliminating the same assumption. In fact, here the function of elimination is secured by means of the proposition expressed in the but-introduced conjunct, i.e. Jane eats tomatoes. On this stipulation, the contradiction between the proposition that Jane does not eat tomatoes and the proposition in q that Jane eats tomatoes, indicates the metarepresentational level of interpretation in (28), in this case a phonological element of the previous utterance undergoing negation. On this account, the proposition in q reserves a specialized role for engendering the elimination of the underlying implicit assumption, i.e. Jane prefers other vegetables.25 Accordingly all the remaining instances in the set of examples under discussion (29)–(32) seem to fall readily under the procedural definition. That is, the metarepresentational clue in but q seems to topple (or support the elimination of) an implicit assumption that seems to amount to the hearer’s expectation that but q will keep up the descriptive (non-metarepresentational) reading of -p: that we saw all (other) sorts of animals in the zoo (29), that I will lecture you on negation (30), that poor old Mr. Dean is married (31), and, lastly, that Mary is sad (32). Up to this point, we have distinguished two categories of correction interpretation in correlation with the distinction drawn in discussions of negation: descriptive and metarepresentational. This differentiation seems to promise a more manageable handling of the phenomenon at hand. In this line of interpretation, we have expounded the case of metarepresentational correction in standard procedural terms. But our discussion of the descriptive cases is still pending.

7.2. A procedural account of descriptive correction As we have already witnessed (in Section 4), descriptive cases, e.g. (23)–(25), seem to defy inclusion in the procedural definition, as a matter of the unavailability of the contextual effect of elimination. Moreover, we ruled out the options of a

24 In this sense, Jasinskaja (2012: 195) rejects the possibility that p in (24) could raise the assumption that John isn’t going anywhere, an assumption that would have served the requirement of an elimination-based, procedural account of but for tracking and eliminating an inferentially derived assumption. 25 In light of the contribution that the but-introduced conjunct makes to the metarepresentational interpretation of the conjunction as a whole, it may be argued here that we can take a different path in explaining the operation of procedural elimination. In particular, it can be said that the (contextually delivered) metarepresentational interpretation of the negation (in the first conjunct) operates procedurally in the direction of eliminating the assumption negated, i.e. Jane didn’t eat tomatoes, thus also restoring its positive counterpart, i.e. Jane eats tomatoes. In our opinion, this path leads to an impasse. For starters, we need to consider the plausibility of this claim, given that it involves the elimination of a linguistically encoded negation by inferential means. More importantly, however, the act of correction ultimately concerns a phonetic, rather than a propositional aspect of the previous utterance. In the end, we are not informed as to the truth of Jane eating tomatoes. Thus, it cannot be the negation in Jane does not eat tomatoes that is susceptible to procedural suppression.

V. Bardzokas / Language & Communication 45 (2015) 27–45

41

procedural meaning description in terms of explicit content. Should we abandon the standard procedural view of (23)–(27), then? Not necessarily. Given the impediments presented in the original relevance-theoretic explanation of (descriptive) correction, how is the standard procedural application to be extended to descriptive cases? It seems that an extended procedural application seems to require a minor, yet critical refinement in the mechanics of procedural interpretation. Keeping in mind the key requirement of pragmatic-effect derivation in procedural interpretation, the answer to this query seems to lie, once more, in identifying the type of constraint that correction but in (23)–(27) imposes on the deductive operation governing utterance comprehension. If it is not elimination that but is used to constrain, what kind of constraining function does it serve? The possibility of an alternative constraining function is now reminiscent of Blakemore’s perceptive comment on (the then cruder conception of) the variety of correction, i.e. “[..] or we must re-analyze but so that the inferential procedure it activates does not necessarily have to result in the elimination of a manifest assumption” (Blakemore, 2002: 111). We are now ready to turn to this issue. In keeping with our metarepresentational investigation, the point of our departure here involves a thorough observation of examples (23)–(27) with respect to the deductive operation, starting with the recalcitrant case in (23), i.e. That’s not my sister but my mother. The proposition captured in the first clause of this conjunction, i.e. That’s not my sister, acts as input to the deductive device in the course of processing the corresponding utterance. The specific proposition is likely to contextualize with current assumptions like ‘C seems acquainted with the speaker’, or ‘C resembles the speaker’. On this postulation, the upshot of the contextualization process may be said to consist in the general or weakly manifest contextual implication that C is (another) relative or that C is a friend of the speaker’s. In Jasinskaja’s (2012) terms, the derivation of this implication can be claimed to motivate the questions ‘Who isn’t she’ and ‘Who is she?’. These questions seem to arise as a result of the hearer’s need to arrive at optimal utterance processing. In response to this implicit request for further information, the speaker offers the but-conjunct to supply the replacive proposition that C is the speaker’s mother, thus also confirming the hearer’s initial speculative assumption, i.e. ‘C is a relative’ or ‘C is a friend’ of the speaker’s. How does the information in q, i.e. ‘C is the speaker’s mother’, interact with the contextual assumption that C may be another relative or friend? Acting as input to the contextual premises, where the probability that C may be a relative or friend represents a general or weakly manifest assumption, the but-conjunct serves to strengthen this assumption. In this sense, by way of elaboration or instantiation, the material in but my mother confirms the hearer’s general impression of C’s identity. Likewise, but to Berlin in (24) is used to achieve the strengthening of the assumption that John is probably going elsewhere. Also, to repeat the case in (18) (B) or (25) (B), assuming a contextual situation where an average student’s application for a postgraduate program has been unsuccessful and where conversationalist B explains to A the reasons for the student’s success at different exams in the past, the utterance in question receives a descriptive interpretation. Here, the conjunct in but hardworking is meant to strengthen the assumption generated in the direction of building on the topic of the student’s previous achievements, i.e. that the student may possess offsetting qualities. Example (26) (B) seems treatable on an equal footing. To alter the previous scenario minimally, speaker B, who is not surprised by the student’s results, denies emphatically the student’s ability to make it to the postgraduate program, confirming speaker A’s expectation (rather than explaining to speaker A the student’s previous academic achievements). In this situation, given the unsurprising result of the students’ failure, the comment in He is not clever (in the form of negation) can be deemed understated, raising the contextual assumption that the student’s level of genius deserves a more emphatic description than that described by not clever, i.e. he is completely dumb. In this line of interpretation, the but-conjunct may be said to contribute to the verification of this assumption. Finally, in (27), in the process of a discussion about a horse to be purchased, -p generates the contextual assumption that the horse referred to is of a different color, to the purchaser’s satisfaction or lack of it. This is the assumption that undergoes contextual strengthening in virtue of uttering the but-conjunct. At this point, it might be objected that an explanation of descriptive correction is not impossible in terms of procedural elimination given the appropriate context. Consider Blakemore’s example in (25) (B) in a slightly different contextual scenario. A prospective student has applied for a postgraduate program but does not seem to live up to the admission standards. In this sense, the use of -p is viewed as carrying the implicit assumption that the student should not be recommended, as opposed to q, which is intended to impart the implicature that the student should be recommended. In this context, it appears that the latter assumption contradicts the former. On this approach, we might consider dropping the suggestion of contextual strengthening (or the underlying requirement for descriptive correction, for that matter), in defense of an exclusively procedural, elimination-based, account. While the elimination-based account seems applicable here, it seems vulnerable to a few objections. Firstly, to the extent that the conflicting assumptions involved in this sort of elimination process are both derived contextually, they are both made weakly manifest, as compared to more typical cases of procedural elimination, e.g. (2)–(3),(6) (let alone the cases of metarepresentational correction), where the eliminating assumption is expressed linguistically. In this case, the relation between the two inferential assumptions may more accurately be described as one of undecided conflict, rather than in terms of elimination. A similar type of balanced suggestion can be said to derive from the situation in (27) or other denial-ofexpectation-resembling cases like the following. (33) She’s not experienced [so you should not hire her] but a fast learner [so you should].

42

V. Bardzokas / Language & Communication 45 (2015) 27–45

Also, there seem to be limitations to the suggestion of a potentially general (elimination-based) application of the type of interpretation proposed for the situation in (25) (or (33)). The specific interpretation does not seem to receive the same contextual support everywhere, as was explained in our discussion of (25) (B), where the evaluation of the student’s application is a past, rather than future event. Finally, as has been argued, a number of descriptive readings of correction (23)– (24), (26) seem to fall outside the elimination-based generalization. As a consequence, there are no solid grounds for proceeding with a more profligate, tripartite distinction in correction interpretation. Apart from the case of metarepresentational elimination, this distinction would also involve dividing the descriptive category of meaning into the sub-categories of elimination-based and strengthening-based interpretation.26 As a final remark, it is worth pointing out that discussions of negation commonly refer to distinctions other than descriptive and metarepresentational, i.e. presupposition-canceling and presupposition preserving. While this is true, the latter distinction does not appear to play a central part in the current meaning description of but. Consider, for instance, the example The king of France is bald. In order to cancel the existential presupposition that there is a king of France, an explicit justification is required of the sort: the king of France is not bald because there is no king of France. But this sort of justification requires the realization of a because-introduced clause, and, thus, any discussion of correction is effectively ruled out. What can restore this discussion is the option of preserving, rather than canceling, the existential presupposition, as in a case like The king of France is not bald but long-haired or The king of France is not [simply] bald but bald as a coot. In this sense, however, the use of but falls, once again, into either the descriptive or the metarepresentational class of correction. In summation, in analyzing the overall phenomenon under scrutiny, we have basically postulated the requirement of splitting correction data into two categories of interpretation: metarepresentational and descriptive. In this light of enquiry, we have found out that the former category of data is readily amenable to a procedural treatment in the traditional terms of elimination. Accordingly, the latter cases are equally treatable on purely procedural grounds, albeit, on this occasion, in terms of an alternative contextual effect: contextual strengthening. But where does our suggestion of a contextual strengtheningdriven analysis of (descriptive) correction leave us with regard to the standard procedural analysis of but? In order to assess the advantages of our account, a careful examination is required of its prospective implications in a unifying semantic/pragmatic system of meaning analysis.

8. Implications On the one hand, our account of the metarepresentational aspect of correction is aligned with the traditional, eliminationbased relevance-theoretic account of but. On the other hand, however, our construal of descriptive correction-but introduces a subtle distinction in the procedural meaning of but. Here, the employment of the connective highlights the effect of contextual strengthening, rather than the effect of elimination. In this sense, our account may raise issues of uniformity. Is it the uniformity of the procedural account of but that is at stake? We trust that it is not. An account that underscores the contextual effect of strengthening as an alternative to the effect of elimination does not depart from the standard definition of procedural encoding. On the contrary, it continues to exploit the constraining operation of a discourse marker on implicit meaning, and, more specifically, on one of the contextual effects involved in the constraining operation. In this respect, our account basically suggests broadening the range of contextual effects involved in the procedural interpretation of but, so that correction, in general, and descriptive correction, in particular, is described parsimoniously in procedural terms. On this view, definite contexts license a reassignment of the procedural application to the operation of a contextual effect other than that of elimination. However, moving to a different level of uniformity, it may still be claimed that our account of descriptive correction appears to blur the transparency of a formerly direct link between the effect of elimination and the overall interpretation of but. While this is a valid claim, our reappraisal of correction proposes a minor (though crucial) adjustment in the description of an otherwise operative procedural device. Put differently, it is a small price to pay to uphold an essentially univocal and robust approach to but that does not compromise the equally crucial requirement of descriptive rigor.27 But does it not sound implausible or counter-intuitive to associate the meaning of but with the function of contextual strengthening, even if by way of exception, given that the effect of strengthening is basically associated with the employment of after all (Blakemore, 2002)? There seems to be no compelling evidence corroborating this suggestion. In fact, it is

26 Even if our conclusion were to be refuted, the requirement for a descriptive (contextual strengthening-based) vs. metarepresentational (eliminationbased) distinction in corrective interpretation would be left unaffected. 27 A univocal account of but has been the holy grail, not only in Gricean and relevance-theoretic approaches to discourse, but also in argumentationtheory-inclined accounts (Jasinskaja, 2012). More specifically, in Jasinskaja’s view, both denial-of-expectation cases and correction cases (the latter of which she does not categorize further) are explicable univocally in terms of a contrast between something positive and something negative. In the former cases, the contrast, roughly speaking, is one between an argument and a counter-argument, while the latter is one of a syntactically positive and a syntactically negative sentence. Although the concerns of the present work are relevance-theoretic and, as such, do not provide the appropriate place for a proper critique of Jasinskaja’s view of a unitary approach to but, we surmise that, despite the unifying terms of the polarity between something ‘positive’ and ‘negative’, her account basically preserves the distinction between an interpretation of but involving an argumentative (in other words, inferential) dimension and one that does not. But how merging these interpretations is ultimately achieved against the semantic/pragmatic distinction is no foregone conclusion. If our reservations raise a valid point, then our account here presents an advantage over Jasinskaja’s in face of the requirement of uniformity.

V. Bardzokas / Language & Communication 45 (2015) 27–45

43

conceivable that a narrow perspective on correction-but in the strict terms of elimination is justified by considerations of analogy, i.e. a treatment of correction along typical procedural (denial-of-expectation) cases, rather than by argumentation. Let us, then, make amends, in this respect. The operation of ‘elimination’ is, more often than not, placed under the microscope for the purpose of seeking the inferential proposition that undergoes elimination. This practice can be held accountable for defocusing the analyst from a more distinct view of the precise role that the propositional content of the but-introduced conjunct fulfills in support of the function of elimination. In this respect, a more comprehensive observation of elimination would heed this level of interpretation, too. An observation of this sort would also benefit a more accurate analysis of correction. We will begin this observation against the backdrop of typical cases of procedural but-conjunction, with a view to measuring the extent to which the proposition in q is justified consistently across contextual uses of but. Let us repeat the utterance in (2) in example (34) below. (34) John is a Republican but he is honest. In performing the task of procedural elimination, the proposition in q, i.e. John is honest, is, also conceived of as providing replacive material28 in substitution for the contextually derived and eliminated proposition that John is dishonest. In this respect, the replacive material of the but-introduced conjunct is justified in terms of supporting the function of elimination, in the sense that the material in q seems to supply the evidential grounds for the elimination of the assumption that John is dishonest. Accordingly, with regard to (descriptive) correction, Jasinskaja brings the replacive aspect of correction to our attention, discussing the case in (24), i.e. John isn’t going to Paris but to Berlin. Thus, she notes that ‘‘Paris’ is replaced by ‘Berlin’’ (Jasinskaja, 2012: 1901). It appears, then, that, on the corrective interpretation, the replacive material in q substitutes a linguistic element in -p. And by way of this substitution the material in q instantiates the probabilistic assumption triggered by -p: ‘Berlin’ instantiating ‘some other place’ or ‘mother’ instantiating ‘some other relative’, thus leading to the effect of strengthening. Overall, then, it can be argued that the replacive aspect of but q has a contributive part to fulfill in effecting elimination, whether the elimination is effected explicitly or implicitly. From this perspective, highlighting the distinct contribution of the proposition encoded in q to the meaning of but, there is no reason to suppose that the replacive aspect that q provides is strictly reserved for the derivation of the contextual effect of elimination (involved in interpreting traditional procedural cases as well as metarepresentational cases of correction). After all, the application of but is common to typical procedural cases but also to cases of correction with either a descriptive or a metarepresentational sense. If it makes sense that the aspect of replacement can be described separately from the effect of elimination, and can, thus be detached from it, rather than being intertwined with it, it is likely that replacement is not incompatible with alternative contextual effects, i.e. the effect of contextual strengthening, as has been implied above. Conversely, the suggestion of compatibility implies that, if the meaning aspect of replacement is not inseparable from the effect of elimination, neither is it from the effect of strengthening.29 For instance, the aspect of replacement seems incongruous with the use of after all, even though the interpretation of the particular marker involves the effect of strengthening. Apparently, as a prototypical marker of contextual strengthening, after all is basically made available in analogous contexts, rather than contexts of correction. On our approach to replacement as a pervasive meaning feature in applications of but, the replacive conjunct is ultimately taken to adduce its propositional material in substitution for the eliminated proposition, whether the elimination is effected inferentially (as in typical procedural cases as well as metarepresentational correction cases) or linguistically (as in descriptive cases of correction). If elimination is realized inferentially, then the replacive material is supplied in q in the direction of the foregoing contextual effect. If elimination is realized by linguistic means, i.e. negation, and, thus, independently of the occurrence of but q, then q contributes its propositional content to the derivation of a separate contextual effect, that of contextual strengthening.

9. Conclusion Correction uses are a hot potato in procedural discussions of but. They have resisted a classification in uniform procedural terms along denial-of-expectation uses to date, owing to the fact that the operation of elimination (involved in correction interpretation) is achieved verbally, that is, in the form of negation, and is, thus, communicated strongly rather than inferentially, or weakly. The current project attempts to restore the assumption of a treatment of correction in a uniform account of but. In doing so, it takes on board a number of alternative methods of procedural analysis that draw on the interface between procedural and

28

In line with Jasinskaja (2012: 1901), we do not employ the term ‘replacive’ in a technical sense. Here the but q conjunct is viewed in isolation from p only in the interest of observing the compatibility of its meaning with the meaning of p, in other words, that the suggestion of replacement is compatible with either contextual effect. Thus, if we only have p (without but q), the weak implicatures are still communicated but they stand no chance of elimination or strengthening, so, in this case, there is, of course, no point in discussing the phenomenon of correction. 29

44

V. Bardzokas / Language & Communication 45 (2015) 27–45

explicit meaning. By a process of elimination, we arrive at the conclusion that correction is best understood in terms of the standard procedural approach, though on certain additional assumptions. As it turns out, correction is a context-sensitive, multi-faceted phenomenon that is sensitive to a distinction in meaning figuring prominently in discussions of negation: metarepresentational and descriptive negation. This distinction generates analogous categories of correction interpretation, i.e. metarepresentational and descriptive, the former being unobserved in the research on correction to date. We have argued in some detail that the utterances falling into the metarepresentational category can readily be accommodated by recourse to the traditional approach to procedural but. On this account, the metarepresentational element of these utterances seems to reveal an assumption that is suppressed by the use of the butintroduced conjunct. On the other hand, the procedural approach seems equally applicable to cases of descriptive correction, albeit on the stipulation of heeding the implication of a contextual effect other than elimination: that of strengthening an existing assumption. More specifically, a general or weakly manifest contextual assumption (derived inferentially by use of -p) is strengthened by the more concrete (replacive) information presented in but q. Regarding the objective of a uniform account of but, the implication of our account of correction is far from trivial: the meaning of the connective proves procedural across contexts. Admittedly, in the proposed line of thought, our account involves a re-identification of the types of contextual effect encoded in but. In other words, it involves broadening the range of contextual effects involved in the procedural operation. On this view, however, rather than detracting from the uniformity of a standard procedural account, our account necessarily keeps sight of the standard guidelines employed in the analysis of procedurally described discourse markers. Thus, but turns out to encode a semantic constraint on implicitly recovered content, while the procedural mechanism employed to this end may vary according to the contextual effect achieved on differing contextual occasions. Acknowledgments The current discussion benefited profoundly from the detailed review of the paper. I am especially grateful to the anonymous reviewer who invested personal time in the paper’s best interest. References Abraham, Wernre, 1979. But. Studia Linguist. 33, 89–119. Anscombre, Jean-Claude, Ducrot, Oswald, 1977. Deux mais en français? Lingua 43, 23–40. Bach, Kent, 1994. Conversational impliciture. Mind Lang. 9, 124–162. Bardzokas, Valandis, 2012. Causality and Connectives: From Grice to Relevance. John Benjamins, Amsterdam/Philadelphia. Bardzokas, Valandis, 2013. The semantics and pragmatics of causal connectives: conceptual and procedural aspects of modern Greek giati and epeidi. In: Kecskes, I., Romero-Trillo, J. (Eds.), Research Trends in Intercultural Pragmatics, pp. 29–54. Bardzokas, Valandis, 2014. Linguistic constraints on causal content: the case of Modern Greek markers. J. Pragmat 60, 160–174. Blakemore, Diane, 1987. Semantic Constraints on Relevance. Blackwell, Oxford. Blakemore, Diane, 1989. Denial and contrast: a relevance theoretic analysis of but. Linguist. Philos 12, 15–37. Blakemore, Diane, 2000. Procedures and indicators: nevertheless and but. J. Linguist. 36, 463–486. Blakemore, Diane, 2002. Relevance and Linguistic Meaning: the Semantics and Pragmatics of Discourse Markers. Cambridge University Press, Cambridge. Büring, Daniel, 2003. On D-trees, beans and B-accents. Linguist. Philos. 26, 511–545. Carston, Robyn, 2002. Thought and Utterances: the Pragmatics of Explicit Communication. Blackwell, Oxford. Carston, Robyn, 2004. Explicature and semantics. In: Davis, S., Gillon, B. (Eds.), Semantics: a Reader. Oxford University Press, Oxford, pp. 1–44. Carston, Robyn, 2009. The explicit/implicit distinction in pragmatics and the limits of explicit communication. Int. Rev. Pragmat. 1 (1), 35–62. Ducrot, Oswald, 1984. Le Dire et le dit. Minuit, Paris. Foolen, Ad, 1991. Polyfunctionality and the semantics of adversative conjunctions. Multilingua 10, 79–92. Grice, Paul, 1989. Studies in the Way of Words. Harvard University Press, Cambridge, Massachusetts. Hall, Alison, 2007. Do discourse connectives encode concepts or procedures? Lingua 117 (1), 149–174. Horn, Laurence, 1985. Metalinguistic negation and pragmatic ambiguity. Language 61, 121–174. Horn, Laurence, 1989. A Natural History of Negation. Center for the Study of Language and Information, Leland Stanford Junior University. Ifantidou, Elly, 1994. Evidentials and Relevance. PhD Thesis. University College, London. Ifantidou, Elly, 2001. Evidentials and Relevance. John Benjamins, Amsterdam/Philadelphia. Iten, Corinne, 1998. Because and although: a case of duality? In: Rouchota, Villy, Jucker, Andreas (Eds.), Current Issues in Relevance Theory. John Benjamins, Amsterdam/Philadelphia, pp. 1–24. Iten, Corinne, 2000. Non-truth-conditional Meaning, Relevance and Concessives. University of London. PhD thesis. Iten, Corinne, 2005. Linguistic Meaning, Truth Conditions and Relevance: the Case of Concessives. Macmillan, Palgrave. Jasinskaja, Katja, 2012. Correction by adversative and additive markers. Lingua 122, 1899–1918. Jasinskaja, Katja, Zeevat, Henk, 2008. Explaining additive, adversative and contrast marking in Russian and English. Revue Sémant. Pragmat. 24, 65–91. Jasinskaja, Katja, Zeevat, Henk, 2009. Explaining conjunction systems: Russian, English, German. In: Riester, A., Solstad, T. (Eds.), Proceedings of Sinn und Bedeutung 13. Vol. 5 of SinSpeC. University of Stuttgart (Working Papers of the SFB 732). Kalokerinos, Alexis, 2004. The justifying connection, mostly in Greek. J. Greek Linguist. 5, 27–80. Kaup, Barbara, Zwaan, Rolf, 2003. Effects of negation and situational presence on the accessibility of text information. J. Exp. Psychol. Learn. Mem. Cogn. 29, 439–446. Kitis, Eliza, 1982. Problems Connected to the Notion of Implicature. University of Warrick PhD Thesis. Kitis, Eliza, 2000. Connectives and frame theory; the case of hypotextual, antinomial ‘and’. Pragmat. Cogn. 8 (2), 357–409. Kitis, Eliza, 2006. Causality and subjectivity: the causal connectives of modern Greek. In: Pishwa, H. (Ed.), Language and Memory: Aspects of Knowledge Representation. Series: Trends in Linguistics. Mouton de Gruyter, Berlin/New York, pp. 223–267. König, Ekkehard, 1985. On the history of concessive connectives in English. Diachronic and synchronic evidence. Lingua 66, 1–19. Lakoff, Robin, 1971. If’s, and’s and but’s about conjunction. In: Fillmore, Charles, Langendoen, Terence (Eds.), pp. 114–149. Moeschler, Jacques, 2013. How ‘logical’ are logical words? Negation and its descriptive vs. metalinguistic uses. In: Taboada, M., Trnavac, R. (Eds.), Nonveridicality, Evaluation and Coherence Relations. Brill, Leiden, pp. 76–110.

V. Bardzokas / Language & Communication 45 (2015) 27–45

45

Roberts, Craige, 1996. Information structure in discourse: towards an integrated formal theory of pragmatics. OSU Work. Papers Linguist. 49, 91–136. Rouchota, Villy, 1998. Connectives, coherence and relevance. In: Rouchota, Villy, Jucker, Andreas (Eds.), Current Issues in Relevance Theory. John Benjamins, Amsterdam/Philadelphia, pp. 11–59. Sperber, Dan, Wilson, Deirdre, 1995. Relevance: Communication and Cognition, second ed. Blackwell, Oxford. Sweetser, Eve, 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge University Press. Wilson, Deirdre, 2011. The conceptual-procedural distinction: past, present and future. In: In Escandell-Vidal, V., Leonetti, M., Ahern, A. (Eds.), Procedural Meaning: Problems and Perspectives (Current Research in the Semantics/Pragmatics Interface). Emerald Group Publishing, pp. 3–31. Wilson, Deirdre, Sperber, Dan, 1993. Linguistic form and relevance. Lingua 90, 1–25. Wilson, Deirdre, Sperber, Dan, 2004. Relevance theory. In: Gregory, Ward, Horn, Larry (Eds.). Blackwell, Oxford, pp. 607–632. Winter, Yoad, Rimon, Mori, 1994. Contrast and implication in natural language. J. Semant. 11, 367–406. Winterstein, Grégoire, 2010. La dimension probabiliste des marqueur de discours: Nouvelles perspectives sur l’argumentation dans la langue. Ph.D Thesis. Université Paris Diderot, Paris. Zufferey, Sandrine, 2012. ‘Car’, ‘parce que’, ‘puisque’ revisited: three empirical studies on French causal connectives. J. Pragmat. 44 (2), 138–153. Zufferey, Sandrine, Cartoni, Bruno, 2012. English and French causal connectives in contrast. Lang. Contrast 12 (2), 232–250. Valandis Bardzokas teaches EAP at the University of Central Lancashire, UK. He holds a PhD in Linguistics from Aristotle University of Thessaloniki. As of late his research interest has mainly focused on the application of discourse markers on the distinction between semantic/pragmatic meaning. His book on causal connectives was published by John Benjamins in 2012.