Discourse Markers (DMs) have been a topic of research for 30 years under many different names. This paper provides researchers in the field with a coherent definition of DMs and a presentation of the syntactic and semantic properties of this functional category.
Discourse Markers (DMs) have been a topic of research for 30 years under many different names. This paper provides researchers in the field with a coherent definition of DMs and a presentation of the syntactic and semantic properties of this functional category.
Discourse Markers (DMs) have been a topic of research for 30 years under many different names. This paper provides researchers in the field with a coherent definition of DMs and a presentation of the syntactic and semantic properties of this functional category.
Discourse Markers (DMs) have been a topic of research for 30 years under many different names. This paper provides researchers in the field with a coherent definition of DMs and a presentation of the syntactic and semantic properties of this functional category.
Koninklijke Brill NV, Leiden, 2009 DOI 10.1163/187730909X12538045489818
brill.nl/irp An Account of Discourse Markers Bruce Fraser Boston University, USA bfraser@bu.edu Abstract Discourse Markers (DMs) have been a topic of research for 30 years under many dierent names. Te present paper presents an account of one view of DMs with the aim of providing researchers in the eld with a coherent denition of DMs and a presentation of the syntactic and semantic properties of this functional category that will enable them to compare their work on DMs with other researchers. In addition, an analysis of the uses of the DM but supports the claim that there is one core meaning relationship, contrast, with the interpretation of the more than 10 dierent uses of but being signalled by context and pragmatic elaboration. Keywords Discourse Markers (DMs), procedural meaning, pragmatic markers, pragmatic elaboration though prepositions and conjunctions, etc. are names well known in grammar, and the particles contained under them carefully ranked into their distinct sub- divisions; yet he who would show the right use of particles, and what signicance and force they have, must take a little more pains, enter into their own thoughts, and observe nicely the several postures of his mind in discoursingneither is it enough, for the explaining of these words, to render them, as is usual in dictionar- ies, by words of another tongue which come nearest to their signication; for what is meant by them is commonly as hard to be understood in one as another language. Tey are all marks of some action or intimation of the mind; and there- fore to understand them rightly, the several views, postures, stands, turn, limita- tions, and exceptions, and several other thoughts of the mind for which we have either none or very decient names, are diligently to be studied (John Locke, An Essay Concerning Human Understanding , 1959: 521). 1. Introduction Tis paper presents an account of Discourse Markers (DMs), lexical expres- sions such as those in italics in the following examples. 1
1 Most of the examples in this paper are constructed rather than taken from corpora. 0001112084.INDD 1 10/7/2009 4:25:43 PM 2 B. Fraser / International Review of Pragmatics 1 (2009) 128 (1) a. Jones died last night. But he had been very ill for a long time. b. I went to Boston rst and later on, went to Cape Cod. c. Te water wouldnt boil, so we couldnt make any tea. I say an account , not the account , since there is considerable variation in what might be labelled Discourse Markers. On the one hand, researchers do not agree what falls under the term Discourse Markers. For example, Schirin, motivated by her interest in the coherence of discourse, considered the term to embrace a large, imprecisely dened group of expressions, including inter- jections such as oh and now , and non-verbal expressions, whereas Fraser ( 1990 , 1999 , 2006 ), concerned with the pragmatic role played by terms expressing a semantic relationship between messages, considered Discourse Markers to be far more constrained. Blakemore ( 2002 ), while agreeing that DMs signal a semantic relationship between utterances, was interested in only those which contained procedural meaning as opposed to conceptual meaning. Te group of terms labelled as Cue Phrases by Knott and Sanders ( 1998 ) is a subset of those above plus then again and admittedlybut, not considered by the others to be DMs at all. And many researchers, interested in the properties of a specic expression such as well (e.g. Foolen, 1993 ), labelled it as a DM, even though most researchers wouldnt consider it as such. On the other hand, the labels given to the group of expressions generally considered to be DMs vary widely. For example, one nds Cue Phrases (Knott and Sanders, 1998 ), Discourse Connectives (Blakemore, 1987 , 2002 ; Hall, 2007 ), Discourse Markers (Blakemore, 2002 ; Item, 2000; Schirin, 1987 ; Fraser, 1999 , 2003, 2007; Mosegaard-Hansen, 2008 ; Lenk, 1998 ), Discourse Operators (Redeker, 1991 , 1992), Discourse Particles (Schourup, 1985; Abraham, 1991 ; Kroon, 1995; Fischer, 2000; Aijmer, 2002 ), Discourse Signalling Devices (Polanyi and Scha, 1983 ), Indicating Devices (Katriel and Dascal, 1984 , 1977), Phatic Connectives (Bazanella, 1990), Pragmatic Connectives (van Dijk, 1985), Pragmatic Expressions (Erman, 1987 ), Prag- matic Markers (Fraser, 1996 ; Brinton, 1990 ; Erman, 2001 ), Pragmatic Operators (Ariel, 1994), Pragmatic Particles (stman, 1995 ), and Semantic Conjuncts (Quirk et al., 1985 ), to name just a few. Moreover, the researchers involved may have used a common term but were interested in very dierent goals. Under the term Discourse Connectives Van Dijk (1979) was primarily interested in showing how semantic and pragmatic connectives were dierent, Schirin ( 1987 , 2005) was interested in illustrating their use in discourse coherence, Fraser ( 2006 ) was concerned with their role in pragmatic interpretation, Sweetser ( 1990 ) was concerned with their function in pragmatic ambiguity, Ducrot ( 1980 ) used them to illustrate the subtleties of argumentation, while Blakemore ( 2002 ) was interested in them for how they illustrate the conceptual/procedural meaning distinction in relevance theory. 0001112084.INDD 2 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 3 Rather than attempt to sort out these dimensions of names, denitions, and purposes in a coherent way, if that is even possible, I will present here an account of Discourse Markers which I have developed over the years, draw- ing on the research of others to augment and orient my view. In Section 2, I present a denition of DMs, a denition that provides researchers with the terms to state clearly what they have found so that their ndings may be compared with those of other researchers. In the next section, Section 3, I indicate some properties of DMs, which are associated with DMs but not critical to the denition. In Section 4 I present an account of how the mean- ing of DMs should be treated within linguistic theory, the most controversial part of this paper. Finally, in Section 5, I present some issues relevant to future research. 2. A Denition of Discourse Markers 2.1. Te Framework Pragmatic Markers (PMs) I start from the assumption that there is a functional class of lexical expressions in every language which I have called iiacxaric xaixiis (Fraser, 1996 ). Tese expressions occur as part of a discourse segment but are not part of the propositional content of the message conveyed, and they do not contribute to the meaning of the proposition, per se . However, they do signal aspects of the message the speaker wishes to convey. 2
Lexical members of this class typically have the following properties: they are free morphemes, they are proposition-initial, they signal a specic message either about or in addition to the basic message, and they are classied as pragmatic markers by virtue of their semantic/pragmatic functions. Many PMs have homophonous lexical counterparts which are classied by virtue of their syntactic function, e.g. however, clearly, allegedly, so , etc. Tere are four types of Pragmatic Markers. Te rst type, nasic iiacxaric xaixiis (BPMs) , illustrated by the italicised items in (2), signal the type of message (the Illocutionary Force cf. Bach and Harnish, 1979 ) the speaker intends to convey in the utterance of the segment. (2) a. I promise that I will be on time. b. Please , sit down. [a request but not a suggestion or an order] c. My complaint is that you are always rude.
2 Tere are also instances of syntactic and phonological pragmatic markers which I do not discuss here. See Fraser (1996) for further discussion of this point. 0001112084.INDD 3 10/7/2009 4:25:43 PM 4 B. Fraser / International Review of Pragmatics 1 (2009) 128 Te second type, coxxixrai\ iiacxaric xaixiis (CPMs) , signal a comment on the basis message. Tere are ve dierent sub-types as illustrated in (3-7). Assessment Markers (3) a. We got lost almost immediately. Fortunately , a police o cer happened by. b. Mary hurried as fast as she could, but sadly , she arrived too late for the movie. Manner-of-Speaking Markers (4) a. A: Mark, youve got to do something. B: Frankly , Harry, I dont know what to do. b. You got yourself into this mess. Bluntly speaking , how are you going to get out? Evidential Markers (5) a. A: Will he go? B: Certainly , he will go. b. I have great concerns over this. Conceivably , Tim is right. Hearsay Markers (6) a. A: Is the game still on? B: Reportedly , the game was postponed because of rain. b. I wont live in Boston. Allegedly , all the politicians are corrupt. (Non)Deference Markers (7) a. Sir , you must listen to me. b. You jerk , where do you think youre going? Te third type of Pragmatic Markers, oiscouisi xaixiis (DMs) , typically signal a relation between the discourse segment which hosts them and the prior discourse segment, perhaps produced by another speaker. Tere are three classes, illustrated in (8-10). Contrastive Discourse Markers (8) a. A: Harry is hurrying. B: But when do you think he will get here? b. Mark, a good guy. On the contrary , hes a jerk. Elaborative Discourse Markers (9) a. John cant go. And Mary cant go either. b. I dont think it will y. Anyway , lets give it a chance. Inferential Discourse Markers (10) a. A: I like him. B: So , you think youll ask him out then? b. Sue isnt here. As a result , we wont be able to see the video. 0001112084.INDD 4 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 5
3 Exclamation particles ( Wow!, Gosh!, Damn!, Yippee!) and interjections ( Hey, You there , ) are not part of a host utterance, are separate discourse segments, and are treated as pragmatic idioms. Te fourth class of Pragmatic Markers, oiscouisi sriucruii xaixiis (DSMs), signal an aspect of the organization of the ongoing discourse (Fraser, 2009 ). Tere are three subclasses, shown in (11-13). Discourse Management Markers (11) a. In summary , the economy has not ourished under the Bush administration. b. I add that he will not help you until the last minute. Topic Orientation Markers (12) a. Tats all there is to say on this for now. Returning to my previous topic , I would like to point out that b. Now, Mr. Pickard, I want to return to the questions that my now-ab- sent colleague Mr. Roemer was asking you about your communica- tions with the eld. Attention Markers (13) a. We must leave right away. Look , cant you pay attention to what Im saying. b. Hell will freeze over before thats likely to happen. Now , since you havent found anyone in London to suit your taste, what about that nice West girl? Te canonical sequence of Pragmatic Markers is the following, (14) DSM (DM (CPM (BPM (Basic Proposition)))) although there are seldom all present. In addition, in general only one token from each major type is present. 3
2.2. Te Denition of a Discourse Marker For an expression to be a DM it must be acceptable in the sequence S1-DM+S2, where S1 and S2 are discourse segments, each representing an illocutionary force, although elision may have occurred. Tere are three neces- sary and su cient conditions that a DM must meet. Condition 1: A DM is a lexical expression , for example, but , so , and in addition . 0001112084.INDD 5 10/7/2009 4:25:43 PM 6 B. Fraser / International Review of Pragmatics 1 (2009) 128
4 Schirin ( 1987 : 41) wrote I dene discourse markers at a more theoretical level as mem- bers of a functional class of verbal (and non-verbal) devices, although she never proposed a non-verbal DM.
5 I am using * to designate segments which are unacceptable, either syntactically or pragmatically. While I do not mean to suggest that DMs consist exclusively of lexical expressions, this condition explicitly excludes syntactic structures, prosodic features such as stress, pauses, and intonation, and non-verbal expressions such as a grunt or a shrug. Tis would occur, for example, when a speaker utters Its raining. Ah, go anyway with the Ah playing the same role as but . I am simply restricting my notion of DMs to lexical expressions for this paper. 4
Condition 2: In a sequence of discourse segments S1-S2, a DM must occur as a part of the second discourse segment, S2. Tis hosting by S2 occurs whether the segments are combined, as in (15a), (15) a. We were late, but no one seemed to mind. b. We were late. But no one seemed to mind. or there is a full stop, as in (15b). Often, a DM has an intonation contour which separates it prosodically from the rest of the segment, but this depends on the particular DM and the linguistic context. While every DM may occur in segment-initial position, some DMs may occur in the segment medial, and/or segment nal position, depending on the particular DM. Tis is deter- mined by the DMs syntactic analysis and what it specically signals. (16) A: Everyone started late. B 1 : However/But , we arrived on time. B 2 : We, however/*but , arrived on time. B 3 : We arrived on time, however/*but . Notice that for the (B 2 ) variation, there is the interpretation that we is being singled out as dierent from the others ( everyone ), a reading that requires a special marked intonation. Tere are adverbial DMs which occur in medial but not nal position, and vice versa as (17a-b) show, although if the segment subjects are the same, as in (17c), the DM may not occur in segment-medial position. 5
(17) a. Te trip was tiring. Despite that , he ( *despite that ) remained cheerful ( despite that ). b. Te electricity went o. Terefore , we ( therefore ) couldnt make dinner ( *therefore ). 0001112084.INDD 6 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 7 c. We know that the employees cannot make changes to contracts with- out agreemtents. *We, on the other hand , realize how di cult it is to use the law to bring these rogue employees to task. Condition 3: A DM does not contribute to the semantic meaning of the seg- ment but signals a specic semantic relationship which holds between the interpretation of the two Illocutionary Force seg- ments, S1 and S2. Tis is in contrast to other pragmatic markers such as I promise , frankly , alleg- edly , and incidentally , which like DMs, are not part of the propositional mean- ing, but make a qualication or a specic comment on S2 (Fraser, 1996 ). I shall say more below about the types of semantic relationship signalled by DMs. Te three Conditions on the denition of a DM exclude the following types of expressions, either because they do not represent a semantic relationship between adjacent Illocutionary Act segments, S1 and S2, as in (18), (18) a. Interjections ( damn , hey , wow , gosh ,) I like it here. Damn! I really like it here. b. Sentence adverbs ( certainly, surely, denitely ,) John is very nice. Denitely , we should invite him over. c. Modal particles (few in English; German: doch, ja, eben ,) She is pretty. Indeed , she is. d. Focus particles ( just, even, only ,) Everybody is ready. Even Harriet is on time. e. Evidential adverbs ( allegedly, reportedly, according to ,) People are angry. Allegedly , its because of Bush. f. Attitudinal adverbs ( frankly, stupidly, cleverly ,) Te weather is lousy. Frankly , I dont care. or the relationship is grammatical or discourse, not semantic, as in (19). (19) a. Complementizers (grammatical relations such as that, in order that, so as, for ,) I believe that John is right. He xed the door in order that the cat could get out. b. Topic Orientation Markers (discourse relations specifying relation- ships within the discourse such as rst, later, incidentally, oh that reminds me ,) Susan had to do the dishes. First , she did the glass, as she had been instructed. I have to go now. Oh, that reminds me , we were invited to Johns for dinner. 0001112084.INDD 7 10/7/2009 4:25:43 PM 8 B. Fraser / International Review of Pragmatics 1 (2009) 128 2.3. Special Case: Absent S1 and/or S2 Tough not the canonical form of DM use, one or both of the discourse seg- ments S1 or S2 may be absent, with the segment replaced by an assumption derived from the linguistic and/or situational context. Many DMs may occur without the presence of the initial S1, just in case the non-linguistic factors provide the suitable context, as the examples in (20) indicate. (20) a. Context: Joel, on seeing his bike being taken by a stranger. Joel: But thats my bike! b. Context: John, on seeing his roommate walk in smiling. John: So , you aced the exam. c. Context: Father, after a teenage boy has just left the dinner table in a hu. Father : And where do you think youre going, young man? Discourse segment S2 may be empty, with only a DM present, as illustrated in (21), with the implied S2 question in brackets. (21) a. A: Ill have another piece of cake. B: But ? [Who gave you per mission?] b. A: Well arrive late, Im afraid. B: So ? [What do you want me to do about it?] c. A: John will not take his medicine. B: And ? [What do you want me to do about it?] Tis use of a DM for the entire S2 is very restricted. And, both S1 and S2 may be empty, as the following examples illustrate. (22) Context: John, seeing someone taking his bike. John: But ! Context: John, upon suddenly encountering his girlfriend embracing his best friend. John: So ! 2.4. Classes of DMs Given the above denition, the DMs of English naturally fall into three func- tional classes: coxriasrivi xaixiis (CDMs), where a CDM signals a direct or indirect con- trast between S1 and S2 ( but, alternatively, although, contrariwise, contrary to expectations, conversely, despite ( this/that ), even so , however, in spite of ( this/that ), in comparison ( with this / that ), in contrast ( to this/that ), instead ( of this / that ), nevertheless , nonetheless , ( this/that point ), notwithstanding , on the other hand , on the contrary , rather ( than this/that ), regardless ( of this/that ), still , though , whereas , yet ) 0001112084.INDD 8 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 9 iianoiarivi xaixiis (EDMs), where an EDM signals an elaboration in S2 to the information contained in S1 ( and , above all, after all, also, alterna- tively, analogously, besides, by the same token, correspondingly, equally, for exam- ple, for instance, further ( more ) , in addition, in other words, in particular, likewise, more accurately, more importantly, more precisely, more to the point, moreover, on that basis, on top of it all, or, otherwise, rather, similarly , that is to say ). ixiiiixriai xaixiis (IDMs), where an IDM signals that S1 provides a basis for inferring S2 ( so , all things considered, as a conclusion, as a consequence ( of this/that ), as a result ( of this/that ), because ( of this/that ), consequently, for this/ that reason, hence, it follows that, accordingly, in this/that/any case, on this/that condition, on these/those grounds, then, therefore, thus ). Te rst marker in each class ( but, and, so ) is what I call the primary DM of the class and has the broadest meaning of all the DMs in a class. To date, I have not found a DM that falls into more than one class. 6
3. Incidental/Non-Dening Properties of DMs I will now present a variety of properties, more or less associated with DMs, which do not play a role in their denition. Tese might be thought of as incidental properties or non-denitional properties. First, nearly all DMs can be absent from a S1-DM-S2 sequence in which they might occur, with the relationship between the segments remaining unal- tered. For example, in the following sequences, (23) S1: Tis ight takes 5 hours. S2 1 : Teres a stopover in Paris. S2 2 : After all , theres a stopover in Paris. S2 3 : Because theres a stopover in Paris. S2 4 : So , theres a stopover in Paris. S2 5 : But theres a stop-over in Paris. S2 6 : And theres a stop-over in Paris. the DM in each can be absent with the sequence retaining the interpretation it had if the DM had been present, although the prosodic features of the sec- ond segment are often altered to signal the absent DM. Te conclusion to be drawn from this, as Schirin ( 1987 ) proposed, is that a DM does not create the relationship between two successive discourse segments, but it provides
6 Tese three classes closely parallel those of Blakemore ( 2002 ) who writes that DMs fall into three groups, 0001112084.INDD 9 10/7/2009 4:25:43 PM 10 B. Fraser / International Review of Pragmatics 1 (2009) 128 i) By allowing the derivation of a contextual implication (e.g. , so, therefore, too, also ); ii) By strengthening an existing assumption by providing better evidence for it (e.g. after all, moreover, furthermore ); iii) By contradicting an existing assumption (e.g. however, still, nevertheless, but ) and have rough parallels to the analysis in Halliday & Hasan ( 1976 ) and to Quirk et al. ( 1985 )
7 Adherents to relevance theory would reject as a result , among others, from being a DM because it contains conceptual rather than procedural meaning. I address this issue in Section 4. clues which inform the hearer of the relationship intended by speaker. To be sure, in some cases it is unlikely that a relatively implausible relationship would be recognized, absent the appropriate DM, but that doesnt bear on the de- nitional issue. Tere are, however, a few cases such as those in (24), (24) a. Fred, a gentleman? On the contrary , hes a bastard. b. Harry didnt arrive on time. In addition , the meeting was late in starting. c. We dont like Harry. On the other hand, he doesnt seem to care. d. He arrived well after the start time. As a result, the Committee can- celled the meeting. where the absence of a DM leaves an odd, if not unacceptable, sequence. Second, some researchers have proposed that the fact that a DM does not contribute to the truth conditions of the host segment should be part of its denition (e.g. Schourup, 1999 ). Tis claim is superuous. Since DMs func- tion as a relationship between two segments, not as part of the meaning of either, it follows that a DM does not contribute to the truth conditions of the host segment. Tird, in some cases the meaning of a DM is exactly the same as the expres- sion when it is used as an adverb, for example in (25). 7
(25) Te meaning of DM and a homophonous form are the same: a. DM: He didnt brush his teeth. As a result , he got cavities. b. Adverbial: Te substance suddenly hardened. Tis wasnt what we expected as a result of our work. In other cases, the meaning of the DM and its homophonous form is quite dierent. (26) Conceptual meaning of DM and a homophonous form are dierent: a. We stopped. On the other hand , there was little point in continuing. It doesnt feel right. Try it on the other hand . 0001112084.INDD 10 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 11
8 Te verbs say , look and see are not DMs according to the denition used here. b. We started late. However, we arrived on time. You should get there however you can. Finally, there are a number of DMs which have no homophonous form, for example, nevertheless , on the contrary , moreover , and conversely . To what degree this meaning dierence is discrete or represents a cline is an open question. Fraser ( 1990 : 393) is probably too strong in suggesting that it is a cline. Fourth, there are a number of DMs which have more than one meaning/use when used as a DM (e.g. but , so , instead ,). However, there appears to be no case where a sequence is ambiguous due to a DM and a homophonous non- DM form with a dierent meaning occurring in the same context. (27) a. I expect him to come. However , he will have to get here however he can. b. A: Its snowing outside. B: So I guess well have to leave early so as to get there on time. c. In addition , John was quite skilled in math, especially in addition . Fifth, DMs constitute a functional class, a heterogeneous syntactic group. Tey are drawn primarily from (28) a. cox;uxcrioxs ( and, but, or, nor, so, yet, although, whereas, unless, while ,) b . aoviiniais ( anyway, besides, consequently, furthermore, still, however ,..) c. iiiiosirioxai iuiasis ( above all, after all, as a consequence (of that), as a conclusion, as a result (of that), on the contrary, on the other hand, in other words, rather than that, regardless of that ,) and very seldom from nouns, adjectives, verbs, or prepositions. 8
Sixth, while many DMs are mono-morphemic (e.g. but , so , and thus ), there are those which are polymorphemic (e.g. furthermore, consequently, neverthe- less , and moreover ) and still others which consist of an entire phrase (e.g. as a consequence, in addition ). For DMs which take the form of prepositional phrases, there are three variations, as in (29). (29) a. DM of Fixed Form: above all, after all, as a conclusion , Mary can open the safe. After all , she knows the combination. b. DM+ this/that (where this/that refers to S1): despite this, in spite of that, in addition to that, She wasnt very pretty. Despite this , she was extremely popular. c. DM+ of this/that (where this/that refers to S1): as a result of ( doing ) this/ 0001112084.INDD 11 10/7/2009 4:25:43 PM 12 B. Fraser / International Review of Pragmatics 1 (2009) 128
9 Tere are a number of other, relatively insignicant features of DMs, stated in the negative. For example, DM cannot be modied ( *as a perfect result ), emphasized (*It was, extremely how- ever , a good idea), coordinated (* But and on the other hand , she could be the one to do it) or the focus of a cleft sentence (*It was however , that we went home). that, because of ( doing ) this/that, instead of ( doing ) this/that Canonical Form: John didnt take the letter. Instead , he left it. Inverted Form: Instead of taking the letter, he left it. Notice that the members of (c), above, are fashioned from a DM plus an ana- phoric prepositional phrase. Tese are meaning preserving variations (trans- formations) of the canonical (basic) form. In these cases, the form of the DM is altered ( instead/instead of ) but the overall interpretation of the S1-S2 sequence is not altered when the transformed DM occurs in segment-initial position of what would ordinarily be S1. Seventh, DMs are likely to take a comma pause when they begin a sepa- rate S2, as in (30), (30) a. I was tired. But/However , I went anyway. b. He was tired. So/As a result , he went home. and the primary DMs of a class ( but , and , and so ) permit an emphatic stress not typically permitted for the other members of the class. (31) a. You will have to take the chairs. BUT/*HOWEVER , dont touch those chairs over by the wall. b. He was very enthusiastic of the project. AND/*IN ADDITION , he had money to fund it. c. Te water wont boil. SO/*THUS , we cant make tea, can we? Eight, the number of speakers required diers depending on the DM. For example, when on the contrary is used in metalinguistic negation, only one speaker is possible, as illustrated in (32a). On the other hand, when but is used in the cor- rection sense (Iten, 2000 ; Hall 2007 ), two speakers must participate, as in (32b) 9
(32) a. Im not happy. On the contrary , Im ecstatic. b. A: I enjoyed meeting your sister. B: Shes not my sister but my daughter. Finally, in the view of DMs presented above, where the DM signals a seman- tic relationship between adjacent discourse segments, the scope of the marker is typically the segment before and the segment which hosts the DM. However, just as there are cases of empty S1, S2, or both, so there are cases where the DMs scope is extended. Te rst is illustrated in (33), 0001112084.INDD 12 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 13 (33) He drove the truck through the parking lot and onto the street. Ten he almost cut me o. After that, he ran a red light. However , these werent his worst oences. where the three segments referred to by these are all embraced in the scope of however . Te second case is where the scope consists of the prior segment and segment hosting the DM, but subsequent segments as well, as in (34). (34) S1: Te boss is on vacation today and everybody played. S2: So , let me guess: John stayed home; Jane went to the movies; and Harry and Susan reported to work but did nothing. Tere may be multiple arguments for both S1 and S2, but I have not found any to date. I suspect they would be rather cumbersome. Te third case is where S1 is interrupted by another segment, either spoken by the speaker of S1, as in (35a), or another speaker, as in (35b). (35) a. I dont want to go. Its such a nice day outside. However , I really do have an obligation to show up. b. A: I dont want to go. B: Well, why dont you stay home? A: But I have an obligation to be there. Lenk ( 1998 ) suggests that there is the relationship of DMs proposed to distin- guish between iocai couiiixci, relationships between adjacent segments (with the alternatives just presented above), and cionai couiiixci, relation- ships to segments mentioned earlier or intended to follow. However, her deni- tion of DMs is far broader than the one proposed here and I will not address it. 4. Te Semantic Meaning of DMs 10
4.1. Framework I now turn to perhaps the most controversial aspect of DMs, their meaning. I start from what is known as a semasiological approach, whereby one takes
10 One approach was suggested by Grice ( 1989 ), who noted that DMs such as but , moreover , and on the other hand , do not contribute to the propositional meaning of what is said, but seem to convey information about non-central or higher-level speech acts , which comment on the interpretation of the ground oor speech acts. For example, in Tree is a prime number but four is not, the function of but , what Grice calls a conventional implicature , is to signal that there is a contrast between the interpretation of the two segments. It has been pointed out that this con- trast notion of but would not cover all the uses, and that this use of the term speech act is very 0001112084.INDD 13 10/7/2009 4:25:43 PM 14 B. Fraser / International Review of Pragmatics 1 (2009) 128 dierent from the usual use. Iten ( 2000 : 203) suggests that Grice might have seen but an indi- cating the performance of an illocutionary act of contrasting, but provides no further commen- tary. I will not pursue this alterative. a specic linguistic form and investigates the range of meanings or functions the form may full. Tis is in contrast to an onomasiological approach, whereby one takes a predened set of (discourse) functions, and investigates how these functions might be expressed linguistically. Tis latter approach has been used by researchers of Cue Phrases (Knot and Sanders, 1998). In investigating the meaning of DMs, my approach has been to take a spe- cic DM, such as but , so, and, however, thus, moreover and examine what semantic relationship(s) it signals between adjacent discourse segments. In examining a particular DM, I am interested in only its use as a DM and not its use as any other expression, however they are used. Tus, I am concerned with the rst examples but not the second examples in (36). (36) a. Pi is a rational number but its not even. John is but a child. b. Te time is getting late so we shouldnt stay much longer. Te bench is about so high. c. I didnt like the food, and I absolutely detested the cool aid. Gin and tonic is a favourite summer drink. Mosegaard-Hansen ( 2008 ) proposes several types of relationships between the dierent uses of a single expression. With homonymy , for each dierent use there is a dierent lexical item , a dierent meaning, and any connection between the uses is arbitrary. Te dierent use of bill as (1) a piece of paper currency, (2) a receipt for goods, (3) the beak of a bird, and (4) a piece of leg- islation, is an example. With monosemy , there is a core meaning compatible with all uses, where pragmatic elaboration, based on the context, both linguistic and situational, makes clear the particular use on a given occasion. Tis mode of DM meaning was proposed earlier, among others by van Dijk ( 1979 ), who suggested that there were two types of connectives. He wrote: we assume that each connective has a certain (minimal) meaning which may be further specied depending on its semantic or pragmatic use. At the same time, semantic conditions may underlie conditions of pragmatic appropriateness. Tus, denoted facts may be normal conditions for the possible execution of subsequent speech actswe assume that each connective has a certain (minimal) meaning which may be further specied depending on its semantic or pragmatic use. (1979: 449) 0001112084.INDD 14 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 15
11 Relevance theory also takes the view of monosemy (though researchers dont label it as such) but with the provision that the single procedural meaning of Contradiction & Elimination can be used to account for all DM uses of, for example but . I specically reject this claim. Sweetser ( 1990 ) also took this position for her discussion of DMs, as in (37), though she didnt explain how one was to distinguish between the three uses. (37) a. Content Causation: fact of S2 caused fact of S1. He came back because he loved her. b. Epistemic Causation: belief in S2 caused conclusion of S1. He loved her because he came back. c. Speech Act Causation: fact of S2 gave reason for speech act of S1. Would you like to go out for dinner tonight, because I know youre tired. With polysemy , a single expression has more than one semantic meaning but these meanings are related in a motivated, if not fully predictable way. Tese related meanings may reect a chain, a radical category, or a network of inter- connected nodes (Mosegaard-Hansen, 2006). One challenge of this approach is to separate out the dierent uses of a DM so that it is the meaning of the DMs which is dierent, not the context. In the three approaches just dis- cussed, it is essential that the tokens under consideration be all of a single syntactic/functional category. If not, then a fourth approach, heterosemy , a type of polysemy, can be employed, where the phonetic form remains the same but the syntactic analysis changes. I am treating DMs as monosemous since most DMs have a single meaning relationship and for those which have more than one, it appears at this point that they can be dealt with by pragmatic interpretation. Te challenge for those DMs that have more than one use, for example but , so, instead, and rather , is to determine a single core meaning that can be further elaborated on by rules of interpretation, yet not be so broad as to be meaningless. In these cases, I attempt to create a path guided by linguistic context and pragmatic principles to signal which of the uses of the DMs is occurring on a given occasion. 11
In agreement with Blakemore ( 2002 ) and other relevance theory adherents, I assume that meaning involves (at least) two types: procedural and conceptual (Blakemore, 2002 ; Hall, 2007 ). However, I am in strong disagreement with Blakemore, Hall, and others who consider every expression as having either a conceptual meaning, or a procedural meaning, but not both. With regard to 0001112084.INDD 15 10/7/2009 4:25:43 PM 16 B. Fraser / International Review of Pragmatics 1 (2009) 128 DMs, they argue that DMs are only expressions such as but, however, so, nev- ertheless, and thus which allegedly are without any conceptual meaning, for example, you cant say what they mean or combine them with other expres- sions (Rouchota, 1998 ). Tey specially exclude as DMs the many expressions such as in contrast, as a result, after all, as a consequence, and furthermore which have conceptual meaning as well as procedural meaning, and in some cases, precisely that same uses as their procedural brothers. Tey conclude that because of this dierence among DMs, they do not form a meaning func- tional class, a counterintuitive conclusion to my mind (Blakemore, 2002 ). It is my view that the mutual exclusion proposed by relevance theory is too strong and misguided, and that each expression in a language may have both conceptual and procedural meaning, some having a greater emphasis on pro- cedural meaning (e.g. the past tense marker ed ), some a greater emphasis on conceptual meaning (e.g. the noun justice ). I analyze DMs as potentially hav- ing both conceptual and procedural meaning, though not in equal propor- tions ( as a result would have far more conceptual meaning than thus ). Both the DMs as a result and thus would contain procedural information to the eect that both DMs signal that S1 is the cause of S2, or that S2 was caused from the action/state of S1, but thus would have the added conceptual require- ment that the causality is assumed to follow logically. I see no way to incorpo- rate the logical requirement of thus into a procedural instruction without reference to the concept logical . Armed with both the assumption of a core meaning from which variations of use are derived through pragmatic elaboration, and the assumption that a DM can potentially contain both procedural and conceptual meaning, I now present what the semantic meaning and pragmatic interpretation of but might look like. 4.2. Te DM but I propose that the DM but signals the semantic relationship coxriasr. 12 It follows that for every use of but as a DM, there is a contrast to be found between the segments of the S1- but -S2 sequence. Of course, the segments which are compared and contrasted are not always the same. Sometime they are the explicit interpretations of S1 and S2, which I shall call direct con- trast, sometime one or both of the segments involved in the comparison are
12 Blakemore makes the same claim using the term contradiction rather than contrast. In the following discussion, I hope to make clear why I have selected contrast. 0001112084.INDD 16 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 17
13 Tese set members are not antonyms, as Lako ( 1971 ) would have it, but characterize dif- ferent descriptions, more or less complex, grouped under a general category label. an implication, which I shall call indirect contrast. For example, in (38a) the interpretation which emerges from contrasting explicit S1 and S2 is that speaker of S2 is challenging the accuracy of S1, while in (38b), the contrast between the presupposition of S1, a logical implication (Tere is a King), and the explicit S2 is an indirect contrast with the interpretation that the speaker of S2 considers S1 as defective. (38) a. A: My father is a professor. B: But your father is NOT a professor. b. A: Te King is dead. B: But there is no King. As an aide to talking about contrast, I will use the concept of Semantically Contrastive Sets (SCSs). Tese sets, characterized by a hypernym, such as sports, toys, my friends, or Boston politicians, consist of the names or descriptions for each member of the set, with some members being dierent and thus in con- trast, some not. For example, if we consider the set sports, golf and baseball would be viewed as a contrast while golf and the game of Tiger Woods would not. 13
Segments being compared for an assessment of contrast sometimes have one SCS, which means that each segment contains one member in the same syn- tactic functional location (e.g. subject, action, etc.), and sometime there are two SCSs. In all cases, the segments are typically declarative or imperative, and parallel in structure (e.g. active-active, cleft-cleft). In (39), with one SCS being that of friends, the other being sports, (39a) constitutes a meaningful contrast while (39b) does not, there being no meaningful contrast between tennis and the sport that made Billy Jean King famous. (39) a. John plays golf but Susan plays tennis. b. *John plays tennis but Susan plays the sport that make Billy Jean King famous. Tere are other cases of meaningful contrast which do not involve SCSs. Tese cases arise when one or both of the segments being compared for pos- sible contrast are relatively vague, and do not contain obvious SCSs. In (40), (40) a. We arrived late for the dinner party, but everyone seemed to ignore this. b. People cant stand latecomers to a dinner party. when S2 is compared with (40b), a possible implication (assessable assump- tion) from S1, there are no SCS members to compare yet there is a sense 0001112084.INDD 17 10/7/2009 4:25:43 PM 18 B. Fraser / International Review of Pragmatics 1 (2009) 128
14 I have included the other contrastive DM such as in contrast to reinforce what is clear to me that examples like these are really in contrast. contrast, a sense of incompatibility. I shall use this somewhat weaker notion of incompatibility in these less precise cases of comparison of implications. In the following I present many uses of but as a DM, attempt to make clear what linguistic context characterizes a specic use, and to identify what inter- pretation emerges. Te examples are fairly straightforward and more complex ones may give rise to issues I have not considered. I take no credit for discov- ering these uses and present them for others to assess and rene. Te order of presentation, having the direct contrast before the indirect contrast, is not intended to suggest any theoretical priority, nor is the order within the groups. I have labelled the explicit contrast examples EC-1, EC-2, and the implicit contrast examples IC-1, IC-2, and appended the names often given in the research literature, where relevant. 4.2.1. Explicit Contrast Tese uses of but all involve the explicit contrast of the interpretation of seg- ments S1 and S2 but in dierent linguistic contexts. EC-1 (Simple Contrast). In examples such as, (41) a. Tree is [Positive] a prime number. But ( in contrast ), four is not . b. Exterior paint is very tough but ( in comparison ), interior paint is rela- tively soft . c. John likes to dance, but/whereas I like to read . d. A: What we gain in speed we lose in sensitivity . B: But ( c onversely ), what we gain in sensitivity we lose in speed . (41a), for example, there is one speaker, the segments are declarative, are paral- lel in form (active-active), and there are two SCSs for comparison, shown in italics in the example. Te interpretation for these examples is that S1 and S2 are in contrast. 14
Abraham ( 1979 : 112) discusses sequences like, (42) a. Tere was no chicken but I got some sh. b. He doesnt have much endurance, but (to make up for that), he has long legs. which have 2 SCS each. He suggests a dierent use of but , calling it compen- satory, or negatively concessive with the but being translated by dafr in German. He further suggests that the predicate of the second clause is sig- nalled as preferred to that of the rst, and the second clause is dominant, 0001112084.INDD 18 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 19 i.e. the second clause receives the stronger accent of the two events (Abraham 1979 : 113). Tat may be the case, but I nd no evidence for it in English, nor do I nd these examples dierent than the other one-speaker contrast cases just discussed (Iten, 2000 : Ch. 5). EC-2. In the following examples, (43) a. He plays basketball but he also plays ping pong. b. Jones works as an engineer. But, in addition , he consults for the FBI. c. I like Bach, but I like Te Beatles, too . there is a single speaker and one SCS. A comparison of S1 and S2 results in a meaningful contrast (Note: the DM also does not play a role in this determi- nation) but this contrast would ordinarily result in an unacceptable sequence with but (though it is acceptable with and ), for example, (44) *He plays basketball but he plays ping pong. since but precludes positing contrasting properties of the same object. However, in these sequences there is a second, Elaborative DM, ( also, in addition, too, as well, neither ,) which signals that what was expressed in S1 is not a unique case. Here the interpretation of the but sequence, taken with the EDM, is that the segments are in contrast but are compatible. EC-3. In examples (45), (45) a. A: John is brilliant. B: But he is NOT brilliant. b. A: We are not going to move to the library. B: But we ARE going to move to the library. there are two speakers but one SCS: polarity. S2 and S1 are in contrast. S2 denies S1, with the interpretation that the speaker of S2 is posing a challenge to the message of the rst speaker. EC-4 (Correction). A fourth case is shown by (46), (46) a. A: Im going to a conference in Berlin tomorrow. B: Tat conference is not in Berlin but Boston. b. A: I see you brought your niece with you today. B: Shes not my niece but my daughter. c. A: Oh, my, Nancy fell down. B: Nancy didnt fall down, but tripped. where it is the prior contribution from speaker A who, in the view of B, makes a factual mistake. Te second speaker, in uttering S1, rejects the mistaken information, using an explicit (not incorporated) negative, and S2 provides 0001112084.INDD 19 10/7/2009 4:25:43 PM 20 B. Fraser / International Review of Pragmatics 1 (2009) 128
15 Tis use of but , which is rendered as sino rather than pero in Spanish, as sondern rather than aber in German, has been examined by Schwenter ( 2002 ), among others. Tere is an interpreta- tion of rather in all these cases, giving rise to the speculation that an alternative form would have rather in it, for example, Tat conference is not in Berlin. Rather , its in Boston.
16 Notice that except, with the exception of, apart from, aside from, excluding, save, among oth- ers, may be used in this construction without any meaning change. Teir status as DMs has not be examined to my knowledge. the correct answer in a truncated form. Here, again, S1 and S2 are in (albeit elided) contrast, with the interpretation that this use of but signals that S2 should be considered a correction of an earlier mistake. 15
EC-5. In the examples (47), (47) a. A: He likes yogurt. B: But he likes ice cream even more . b. I dislike carrots and also turnips. But what I like least is brussel sprouts. c. Tanks are due to John and Jim. But above all , I want to thank Harry. there is one SCS and one or two speakers, and but combines with a Relative Degree Form such as even more/less, what he likes best/least, and above all . Like in EC-3, there is contrast between S1 and S2 but, in addition, the degree expression signals that one segment is more/less highly valued than the other. EC-6 (Exception Use). Te sixth use is shown in the following examples, (48) a. Everyone left on time but John. b. No one said a word but me. c. Come anytime but dont come right before dinner. where there is a single speaker and two SCS (for example, in (48a), the set of all relevant people; polarity). Te member of the SCS in S1 is at the extreme value ( any, anyone, everyone, everything, nothing, no, no one, all, ), and S2 has a polarity opposite from S1 and is usually truncated (John as opposed to John did not leave on time) and has numerous variations. 16 S1 and S2 are in contrast and the interpretation of this sequence is that there is an aspect of S2 which is exceptional when compared to S1. 4.2.2. Implicit Contrast Each of the following cases has at least one implied segment for comparison rather than the explicit interpretations of S1 and S2 of the prior examples. Te contrast in these cases is often the weaker incompatibility rather than the explicit contrast of two members of a SCS. IC-1 (Contradiction and Elimination). Te rst case, illustration by (49), 0001112084.INDD 20 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 21 (49) a. We started late. [We will arrive late] but we will arrive on time. b. Its very cold in here. [Please turn up the heat] but please dont turn up the heat. c. Hes a University Professor. [University Professors are smart] but I think hes stupid. is what Lako ( 1971 ) calls Denial of Expectation and what relevance theory calls Contradiction and Elimination. In this case, S2 is compared with some implication (an assessable assumption according to Hall, 2007 ) derived from S1 (I/S1), with the result being incompatibility. Te interpretation is that the implication, I/S1, is eliminated from further consideration in the discourse and S2 is validated. It should be noted that many of the EC-1 cases of simple contrast might be interpreted as IC-1, if the situation is appropriate. For example, consider (50). (50) A: You are aware that John and Joe always drink the same thing. B: Well, today John drank beer but Joe drank wine. In this context, S2 is in contrast with the implication Joe drank beer, with the result that the implication is eliminated. I want to take a slight diversion here and consider the position of Blakemore ( 2002 : 100; 2005) and Hall ( 2007 ). For them, every DM use of but is cap- tured by some variation of contradiction and elimination, that has just been discussed, which means that there is no contrast meaning for but . Consider, however, the example in (51) taken from Saebo ( 2003 ). (51) a. Some talk. But most listen to the young woman at the piano. b. Te volcano has been simmering. But it hasnt yet erupted. c. Te driver should have seen the stop signal. But he didnt. What assumptions of S1 in these sequences could be contradicted by S2? What is to be eliminated? Hall ( 2007 ) argues that if buts core meaning is coxriasr , this meaning should work anywhere but can occur. But that argument presumes that the meaning of but doesnt change in virtue of the context in which it occurs. If, as Hall argues, the core meaning is contradiction and elimination, it should work anywhere, as well. But clearly it doesnt, as the explicit contrast uses of but , discussed above, illustrate. She says, My proposal is that, rather than activating the contradiction and elimination of an assumption, but indicates that what follows is cutting o a line of inference opened up by the previous clause. What gives rise to the dierent interpretations of but is the salience of some particular conclusion that is undermined, and this depends on the relation between the two conjuncts. (Hall, 2007 : 168) 0001112084.INDD 21 10/7/2009 4:25:43 PM 22 B. Fraser / International Review of Pragmatics 1 (2009) 128 No details to support this proposition are provided. IC-2. Now consider the three sequences in (52). (52) a1. Harry died, but this was a blessing, since he had been ill a long time. a2. Harry died, but this was a blessing. a3. Harry died, but he had been ill a long time. Te rst sequence, (a1), is a straightforward case of S2 (Tis [his death] was a blessing) contrasting with the implication of S1 (His death was not a blessing) with the interpretation that S2 is validated. Similarly with (a2) in which the reasoning for S2 is missing. Te third example poses a dierent problem. Tere is no viable contrast resulting from a comparison of S2 with S1, or S2 with I/S1. However, a rea- sonable implication to be derived from S2 is that, given the reason presented, His death was a blessing. When compared with I/S1 (His death was not a blessing), these two implied segments are in contrast, with the S2 implication being preferred, given that there is justication provided for it. IC-3. Consider the examples in (53). (53) a. A: Brighton used to be a nice city. [Its no longer nice] B: But its still nice. b. A: Take an orange. [Take some apples] B: But dont take any apples. c. A: Hes a professor [Professors are smart] B: But I think hes stupid. Here there are two speakers, and S2 contrasts with an implication of S1, much the same as in EC-3 where the rejection was of S1 itself. In this case, the inter- pretation is that the speaker of S2 is challenging the implication of S1, but not eliminating it, as in prior IC cases. IC-4. Te examples in (54) illustrate what may occur when a felicity condi- tion on the illocutionary act conveyed by S1, a logical not a contextual impli- cation, is contrasted. (54) a. Statement: statement is true. A: John is brilliant. B: But thats not true. b. Oer: object/action is desired. (John, on being oered a drink) But I dont want it; I dont drink alcohol. c. Request: hearer can perform the requested act. A: Please bring me the stool. B: But I cant do that. [why are you asking me?] In each case, S2 explicitly denies a felicity condition of S1, and a contrast exists. Te resulting interpretation is that the speaker of S2 is claiming that the act conveyed in S1 is infelicitous. Again, no implication is eliminated. 0001112084.INDD 22 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 23 IC-5. Te next use of but involves two additional logical implications. (55) Challenge to entailment of S1. A: John murdered Smith. (E=> Smith is dead) B: But Smith is not dead. A: Some of the boys left. (E=> At least two boys left) B: But only one of the boys left. A: Consider this bicycle. (E=> A bicycle has two wheels) B: But this bike has 3 wheels. (56) Challenge to presupposition of S1. A: Damn the King of France. (P=> Tere is a King) B: But there is no King of France. A: Has John stopped smoking. (P=>John smokes) B: But John doesnt smoke. A: All the boys left. (P=> At least 3 boys left) B: But only two boys left. Here, as in IC-4, S2 denies the entailment or presupposition of S1, thereby creating a contrast between the two. Te resulting interpretation is that the speaker of S2 is implying that the act conveyed by S1 is defective. Again, noth- ing is eliminated. IC-6. Te nal set of examples, in (57), are similar to IC-2 in that there are three variations. (57) a1. Tom is supposed to be here, but he isnt, since he missed his train. a2. Tom is supposed to be here, but he isnt. a3. Tom is supposed to be here, but he missed his train. S1 consists of a (usually positive) segment, which contains a verb of desiring ( wishing, wanting, hoping ), of expecting ( supposing ), or conditional modals ( would, could ), and the full S2 consists of the implied negative assertion that corresponds to S1, followed by a justication for this negative assertion. For the rst two variants of (57a1-a2), S2 (he isnt) contrasts with an implication of S1 (he is here). Te third variant must be handled dierently. S2 in the sequence (57a3), He missed his train, implies that Tom is not here. When this implication is compared with the implication of S1 (Tom is here), there is a contrast. Te interpretation is that the S2 implication is validated, given there was a justication to support it. 4.2.3. Some Other Cases I want to briey mention several other cases which dont seem to t within the contrast meaning of but . Te rst is the topic change use of but (Bell, 1998 ; label it as Discourse or Sequential but , as illustrated in (58)). 0001112084.INDD 23 10/7/2009 4:25:43 PM 24 B. Fraser / International Review of Pragmatics 1 (2009) 128 (58) a. A: I had a lovely evening last night with Harry. B: But did he repay you the money? b. Its good to see you so well, Jane. But lets talk about the real reason I came by. Te fact that there is no contrastive result from comparing S1 and S2 follows from the fact that this case is not a DM use of but . Whereas DM signals a semantic relationship holding between S1 and S2, the but in these examples is signalling a change in discourse topic, not a semantic notion. Tis but is analo- gous to pragmatic markers such as incidentally, on another topic, to return to the former topic, etc. (cf. Fraser, 2009a). Te examples in (59), (59) a. Its unbelievable, but John got married last night. b. You may not be aware of this, but Mark is a very ne pianist. c. Im reluctant to say this, but I dont like the dinner Mary has planned. pose a very interesting case. Tey have been considered by Lauerbach (1998) and Fretheim ( 2005 ), the former not considering the specic meaning of but , the latter attempting to place these examples within the relevance theory of contradiction and elimination, but with a lack of success. What is to be con- trasted is not clear to me, if this use of but is even a DM. Te nal case is illustrated in (60), a case for which I have no adequate analysis, (60) a. A: Is it done? B: But of course its done. b. A: He doesnt want to leave. B: But of course he doesnt. Would you? c. A: Can I help? B: But of course you can. where the but may or may not be functioning as a DM and it requires the of course to accompany it with this use. I list here for the sake of completeness a number of additional sequences containing but which I, at the moment, do not have acceptable analyses for, whether they be DMs or otherwise. (61) a. Te tyranny of the multitude is but a multiplied tyranny. b. He would have gone, but for the mess on the garage oor. c. She speaks either French or German, but I dont know which one. d. Who arrived at the stroke of midnight but the long lost relatives. e. He has all but / nearly clinched the championship. f. It never rains but (that) it pours. g. He is but a child. ( only, simply, just, merely ) h. Ill get you but good . 0001112084.INDD 24 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 25
17 Te sequence of but however does not occur for most people. 5. Future Areas of Research on DMs Tere are several areas where the researcher interested in DMs might make considerable headway in addition to pursuing further their meaning and how best to account for it. Te rst area is the sequencing of DMs. Recall that within the three classes of DMs, there is one term I identify as the primary DM for the class: but , and , and so , with the other members of the class being labelled secondary. In general, the primary member of a class can occur, pre- ceding a secondary member, as in (61). 17
(61) a. He didnt try to climb up but instead just sat there and sulked. b. Te rain was coming done hard so we didnt have the picnic as planned, as a result . c. I wouldnt try that and furthermore , I wouldnt encourage you brother to try it either. In many cases, the secondary DMs may occur in the S2-nal position as well, as in (61b). When there is a pair of DMs, the primary DM retreats to its core function ( but : contrast; and : elaboration; so : cause) and the secondary DM is the one that signals the intended S1-S2 relationship. What combinations of Primary-Secondary are permitted has not been studied. In addition, there are some cases of two secondary DMs from the same class, for example, (62) Well, on the one hand, Ive conded in my mother. I know she wants to help me and is worried about me. However, on the other hand I think she is angry with me. Im making her look like a bad mother. Te order is xed in these cases and only specic DMs combine. Tese, also, await study. Ten there is a combination of a Primary DM from one class with a Secondary DM from another class, as in: (63) a. Tey loaded the pallets onto the trailer, but, in addition , they strapped them down. b. He didnt move from his rocking chair and, instead , gestured to his assistant to do it. c. John went swimming. So, in addition , he wont be home for dinner. How the signals from each aect the DM interpretation is yet to be studied. 0001112084.INDD 25 10/7/2009 4:25:43 PM 26 B. Fraser / International Review of Pragmatics 1 (2009) 128 Another area is the extent to which all DMs operate in the three domains proposed by Sweetser ( 1990 ). While most DMs appear to, there are some such as for example, as a result, in contrast, that is to say , and moreover for which I cannot nd sequences in which they function in the epistemic and/or speech act domain. Finally, a third area worth looking at is the extent to which at least the pri- mary DMs ( and, but, so ) have the same uses across languages. I have con- ducted a preliminary investigation into this matter with but , using sequences which favoured the dierent interpretations of but as a DM in English, such as the examples in (64). (64) a. John is tall but Mary is short. b. I left the house late but I arrived on time. c. Shes not my sister but my mother. d. Its unbelievable but no one in the class has a middle name. e. I would kiss you but I cant. f. Jack is but a child. Te results from over 20 languages indicate that they all share the functions of but illustrated in (64a-b), most use but for (64c) although some languages have an alternative form for but (Spanish: pero , German: aber ), but the lan- guages vary considerably for the other functions which were discussed above. A more thorough, systematic study should be conducted to ascertain just how universal the functions of the Primary DMs are. To conclude, in the foregoing I have attempted to sketch out an account of DMs with su cient precision that researchers working in this area will be able to assess and compare their results with others. Te section on the meaning of but is a rst attempt to set out an alternative to relevance theory and certainly will undergo revision. Indeed, it is possible that data from other languages will cause revision of parts of my account presented here. I certainly hope not. However, I do hope this paper will be the basis for a more productive analysis of DMs than we have experienced to date. References Abraham , Werner . 1979 . But . Studia Linguistica 33 : 89 - 119 . Abraham , Werner (ed.). 1991 . Discourse Particles . Amsterdam : John Benjamins . Aijmer , Karin . 2002 . English Discourse Particles . Evidence from a Corpus . Amsterdam : John Benjamins. Ariel , Mira . 1998 . Discourse markers and form-function correlations . In A. Jucker and Y. Ziv (eds.), Discourse Markers. Descriptions and Teory , 223 - 259 . Amsterdam : John Benjamins. 0001112084.INDD 26 10/7/2009 4:25:43 PM B. Fraser / International Review of Pragmatics 1 (2009) 128 27 Bach , Kent and Robert M. Harnish. 1979 . Linguistic Communication and Speech Acts . Cambridge, MA : MIT Press . Bazzanella , Carla . 1990 . Phatic connectives as interactional cues in contemporary spoken Italian . Journal of Pragmatics 14 : 629 - 647 . Bell , David . 1998 . Cancellative discourse markers: a core/periphery approach . Pragmatics 8 : 515 - 541 . Blakemore , Diane . 1987 . Semantic Constraints on Relevance . Oxford : Blackwell . Blakemore , Diane . 1992 . Understanding Utterances . Oxford : Blackwell . Blakemore , Diane . 2002 . Relevance and Linguistic Meaning : Te Semantics and Pragmatics of Discourse Markers. Cambridge : Cambridge University Press . Brinton , Laurel . 1990 . Te development of discourse markers in English . In J. Fisiak (ed.), Historical Linguistics and Philology , 45 - 71 . Berlin and New York : Mouton de Gruyter . Dascal , Marcelo and Tamar Katriel. 1977 . Between semantics and pragmatics: the two types of but Hebrew aval and ela . Teoretical Linguistics 4 : 143 - 172 . Ducrot , Oswald . 1980 . Analyses pragmatiques . Communications 32 : 11 - 60 . Erman , Britt . 1987 . Pragmatic Expressions in English: A Study of you know , you see, and I mean in Face-to-Face Conversation . Stockholm : Almqvist and Wiksell . Erman , Britt . 2001 . Pragmatic markers revisited with a focus on you know in adult and adolescent talk . Journal of Pragmatics 33 : 1337 - 1359 . Fischer , Kerstin (ed.). 2006 . Approaches to Discourse Particles . Amsterdam : Elsevier . Foolen , Ad . 1996 . Pragmatic particles . In J. Verschueren , J.-O. stman , J. Blommaert and C. Bulcaen (eds.), Handbook of Pragmatics . Amsterdam : John Benjamins . Foolen , Ad . 1993 . Te discourse marker well : a relevance theoretical account . Journal of Pragmatics 19 : 435 - 452 . Fraser , Bruce . 1990 . An approach to discourse markers . Journal of Pragmatics 14 : 383 - 395 . Fraser , Bruce . 1996 . Pragmatic markers . Pragmatics 6 : 167 - 190 . Fraser , Bruce . 1999 . What are discourse markers? Journal of Pragmatics 31 : 931 - 953 . Fraser , Bruce . 2006 . On the conceptual-procedural distinction . Style 38 , http://ndarticles. com/p/articles/mi_m2342/is_1-2_40/ai_n17113874. Fraser , Bruce . 2006 . Towards a theory of discourse markers . In Fischer , K. (ed.), 189 - 204 . Fraser , Bruce . 2009 . Topic orientation markers . Journal of Pragmatics 41 : 892 - 898 . Fretheim , Torstein . 2005 . Excuse me, but . Working Papers ISK : 75 - 86 . Grice , Paul . 1989 . Studies in the Way of Words . Cambridge, MA : Harvard University Press . Hall , Alison . 2007 . Do discourse connectives encode concepts or procedures? Linguist 117 : 149 - 174 . Halliday , Michael and Ruqaiya Hasan. 1976 . Cohesion in English . London : Longman . Iten , Corinne . 2000 . Non-truth-conditional meaning: relevance and concessives . PhD University of London. Katriel , Tamar and Marcelo Dascal . 1984 . What do indicating devices indicate? Philosophy and Rhetoric 171 : 1 - 15 . Knott , Alistair and Ted Sanders. 1998 . Te classication of coherence relations and their linguistic markers: an exploration in two languages . Journal of Pragmatics 30 : 135 - 175 . Kroon , Caroline . 1998 . A framework for the description of Latin discourse markers . Journal of Pragmatics 30 : 205 - 223 . Lako , Robin . 1971 . Ifs, ands, and buts about conjunction . In C. Fillmore and D. Langendoen (ed.), Studies in Linguistic Semantics , 114 - 150 . New York : Holt, Rinehart and Winston . Lauerbach , Gerda . 1989 . We dont want war, but: speech act schemata and inter-schema- inference transfer . Journal of Pragmatics 13 : 25 - 51 . Lenk , Uta . 1998 . Marking Discourse Coherence : Functions of Discourse Markers in Spoken English. Tbingen : Narr . 0001112084.INDD 27 10/7/2009 4:25:43 PM 28 B. Fraser / International Review of Pragmatics 1 (2009) 128 Locke , John . 1959 . An Essay Concerning Human Understanding . London : Peter Smith . Mosegaard-Hansen , Maj-Britt . 2008 . Particles at the Semantics/Pragmatics Interface : Synchronic and Diachronic Issues. Amsterdam : Elsevier . stman , Jan-Ola . 1995 . Pragmatic particles twenty years after . Anglicana Turkuensia 14 : 95 - 108 . Polanyi , Livia and Remko Scha . 1983 . Te syntax of text and discourse . Text 3 : 261 - 270 . Pons Bordera , Salvador . 2006 . A functional approach for the study of discourse markers . In K. Fischer (ed.), 77 - 100 . Quirk , Randolph , Sidney Greenbaum, Georey Leech and Jan Svartvik. 1985 . A Comprehensive Grammar of the English Language . London : Longman . Redeker , Gisela . 1990 . Ideational and pragmatic markers of discourse structure . Journal of Pragmatics 14 : 367 - 381 . Redeker , Gisela . 1991 . Linguistic markers of discourse structure . Linguistics 29 : 1139 - 1172 . Rouchota , Villy . 1998 . Procedural meaning and parenthetical discourse markers . In A. Jucker and Y. Ziv (eds.), Discourse Markers. Descriptions and Teory , 97 - 126 . Amsterdam : John Benjamins . Saebo , Kjell . 2003 . Presupposition and contrast: German aber as a topic particle . In M. Weisgerber (ed.), Proceedings of Sinn und Bedeutung 7 : 257 - 271 . Sanders , Ted , Wilbert Spooren and Leo Noordman . 1992 . Toward a taxonomy of coherence relations . Discourse Processes 15 : 1 - 36 . Schirin , Deborah . 1987 . Discourse Markers . Cambridge : Cambridge University Press . Schourup , Lawrence . 1999 . Discourse markers . Lingua 107 : 227 - 265 . Schwenter , Scott . 2002 . Discourse markers and the PA/SN distinction . Journal of Linguistics 38 : 43 - 69 . Sweetser , Eve . 1990 . From Etymology to Pragmatics . Metaphorical and Cultural Aspects of Semantic Structure . Cambridge : Cambridge University Press . van Dijk , Teun . 1979 . Pragmatic connectives. Journal of Pragmatics 3 : 447 - 45 .