Annotating and sharing language paradigms online

Working papers Working papers isk isk 1/2004 3/2006 Innovation and Creativity Department of Language Department of Language and Communication Studies and Communication Studies Working Papers isk 3/2006 Department of Language and Communication Studies Working papers isk Department of Language and Communication Studies Norwegian University of Science and Technology 7491 Trondheim Norway ISSN 1503-9390 © The authors and the Department of Language and Communication Studies Websites: Department: http://www.hf.ntnu.no/isk_english/ Working Papers isk: http://mime.hf.ntnu.no/hf/isk_english/Research/Publications Contact: isk@hf.ntnu.no thorstein.fretheim@hf.ntnu.no olaf.husby@hf.ntnu.no Working Papers 3/2006 isk Contents: Metaphor. How an analysis of metaphors can shed light on the relationship 5 between implicature and explicature Randi Waade On the functional independence of explicatures and implicatures 15 Thorstein Fretheim Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first'): A 27 pragmatic analysis based on a univocal lexical meaning Thorstein Fretheim The metarepresentational use of main clause phenomena in embedded clauses 41 Thorstein Fretheim A relevance-theoretic analysis of UNLESS 59 Thorstein Fretheim Explaining connections in Akan discourse: the role of discourse markers 85 Nana Aba Appiah Amfo ‘Abroad’ and semantically related terms in some European languages and in 107 Akan (Ghana) Thorstein Fretheim and Nana Aba Appiah Amfo A constructional approach to syntax and the treatment of passive in 125 Norwegian Petter Haugereid Diagnostic use of nonword repetition for detection of language impairment 139 among Turkish speaking minority children in Norway Olaf Husby Annotating and sharing language paradigms online 151 Dorothee Beermann and Atle Prange WORKING PAPERS ISK 3/2006 5-14 ISK, NTNU Metaphor. How an analysis of metaphors can shed light on the relationship between implicature and explicature Randi Waade 1. Introduction The traditional relevance theoretic understanding of metaphor is that metaphoric meaning is implicated meaning. However, Carston (2002) has argued that metaphoric meaning is part of the explicated meaning of an utterance. Starting out with Sperber and Wilson’s (1986) own term ’loose use’, she presents a convincing argument for including metaphor meaning in the explicature of an utterance. In the same way that ‘narrowing’ is known as part of the explicated meaning, so should ‘broadening’, Carston argues. This paper examines the pros and cons of her analysis, and proposes to modify it so that it might be capable of making stronger claims than the original analysis presented by Carston. My proposal rests heavily on Carston’s analysis, but makes certain stronger claims regarding the role of the encyclopedia and the logical content of the word that is used metaphorically. I will investigate different aspects of this analysis, and objections that it may come up against. First I am going to present an outline of Carston’s metaphor analysis. 2. The Analysis Problems relating to narrowing and broadening are closely connected to semantics, and to how we represent our semantic knowledge of words. In this article I will assume the picture of semantic knowledge presented by Carston (2002), without addressing any other possible models, as that would fall outside the intended scope of this article. According to Carston, our semantic knowledge of a word contains its logical content, our encyclopedic knowledge about the word, and its lexical properties. This can be represented in the following manner: 1 1 Carston herself does not use this way of illustrating our knowledge of words, but it is heavily based on what she writes (’According to the relevance-theoretic view, an atomic concept consists of an address or node in memory which may make available three kinds of information: logical content, encyclopaedic or general knowledge, and lexical properties’ (2002:321)). Any misinterpretations are purely my own. One might obviously claim that the logical entry is part of the lexical properties, however, this does not have any consequences for my analysis. One way of solving this problem is simply to place the logical content as a smaller circle within the circle representing the lexical properties. In this article I choose to keep them apart, since this in my opinion gives a clearer model. 6 Waade (1) Logical Encyclo- Lexical Content pedic Properties knowledge The lexical properties consist of the phonological and syntactic properties of the word; in the logical entry we find the basic meaning of the word, while in the encyclopedia we find a large number of individually different assumptions about the word. A semantic analysis of the word ‘bachelor’ in this system might thus look like this: (2) Unmarried XYZÆ [bætʃələr] man ØÅABC MNOQ Noun RSTU The letters in the circle representing encyclopedic knowledge symbolises the variation in the assumptions triggered by the word ‘bachelor’. This might include different people who are bachelors, like the Pope, an older man living isolated, or a young, handsome man who frequents bars or discotheques. In addition one might entertain a large number of different and even contrary assumptions about bachelors: as previously mentioned, they could be young or old, they might be desperately looking for a wife, or they might be reluctant, or even averse, to marrying. One might have ideas of an untidy and filthy home and a somewhat higher level of alcohol consumption than average. Assumptions of this kind are stored in the encyclopedia of the word. Since the information stored under ‘bachelor’ is very diverse, and even contains mutually exclusive knowledge, one will have to limit the amount of information which will be relevant in the context at hand. This is exemplified in Carston (2002:324): (3) I would like to meet some bachelors2 By using this sentence, it is improbable that the speaker means that she wishes to meet the Pope, men who are reluctant to marry, older men, etc. Somewhat depending on context we might assume that the speaker is looking for a younger man who wishes to settle down. Thus the search for relevance will mean a narrowing of the encyclopedic knowledge, while the basic meaning (the logical content) and the lexical properties are retained: 2 All examples are taken from Carston (2002) Metaphor 7 (4) Unmarried XYZÆ [bætʃələr] man ØÅABC MNOQ Noun RSTU The representation of the encyclopedic knowledge in (4) means that part of the general knowledge we have of bachelors is disregarded, as it is irrelevant in this specific context. However, according to Carston (2002), there is no great difference between what happens in cases of narrowing and what happens in cases of broadening, as we see for example in metaphoric expressions. Another example from Carston (2002:328): (5) Ken’s a (real) bachelor [where Ken is technically married] In this case we see that the logical content cannot hold. Ken is no unmarried man, he is married. Nor can we use all of our encyclopedic knowledge. In the same way as in the example above, some properties are irrelevant. In most contexts it would be improbable that Ken is compared to the Pope or a man eager to marry etc. It is more probable that the conveyed message is that Ken is a man with a high consumption of alcohol, a man who is very keen on going out with buddies, or a man who is unusually messy. The representation of this is as given in (6). The logical content is set aside, and a selection of the encyclopedic knowledge is retained. (6) Unmarried XYZÆ man ØÅAB [bætʃələr] CMNO QRSTU Noun We thus see the difference between narrowing and broadening, or literal and metaphoric meaning. In narrowing we keep the logical content, and make a selection among the available assumptions in the encyclopedic entry. In broadening, on the other hand, we set aside the logical content. However, we have to make a selection in the encyclopedic entry, in the same way as we had to choose among the available assumptions in the encyclopedia in cases of narrowing. This is my subjective interpretation of the model presented by Carston (2002), and it differs from hers in some areas. An important difference between my analysis and hers is that she does not emphasize that all logical properties are set aside in metaphoric expressions, whereas this is the cornerstone of my analysis. If only some logical properties are lost, it will be harder to see the difference between metaphoric and literal speech. This is how Carston describes the difference between narrowing and loosening: 8 Waade The only difference between them is that, in the case of narrowing, all the logical properties are retained, while in the case of loosening, some of them are dropped (Carston 2002:334, my italics). At a later point, she describes the difference as follows: The use is a literal one if the logical/definitional properties of the linguistic encoding are preserved; it is non-literal if they are not (Carston 2002:340) In the first quote it seems that only some logical properties are lost in metaphoric language, in the second quote it seems that all are lost. There is a certain vagueness in Carston’s account which makes it hard to grasp what she really means. The problem with the definition rests primarily in the domain of semantics, however, and depends on what counts as belonging to the logical content. If language users have a clear intuitive feel of when an expression is used metaphorically, this is explained by saying that in metaphoric expressions, the logical content is completely set aside. As we saw above, the logical content of the word ‘bachelor’ will be ‘unmarried man’, but in the so-called metaphoric utterance ‘Ken’s a (real) bachelor [where Ken is technically married]’, it will obviously only be the component ‘unmarried’ which is disregarded. Ken is a man, regardless of whether the utterance is interpreted metaphorically or not. This may be the reason why some people might have a problem with seeing the word ‘bachelor’ in the sentence ‘Ken is a (real) bachelor’ as metaphoric use. In this case, not all the logical properties are ignored (‘unmarried’ is ignored, but the entailed concept ‘man’ is retained). Perhaps the utterance would be more likely to be accepted as a metaphor if both ‘man’ and ‘unmarried’ are deleted, for example if one refers to a lonely pet as a ‘bachelor’. We might therefore end up with a third category between literal use, where all logical properties are retained and metaphoric use where no logical properties are retained. This category will be the border-line metaphor, where some logical properties are retained, as seen above. There might be individual differences as to whether one accepts these border-line metaphors as genuine metaphors or not. Finding waterproof criteria for determining whether a certain property of a word is part of its logical properties or its encyclopedic properties may be a problem. It would be circular to claim that if a notion, for example ‘bachelor’, is used metaphorically, it is a prerequisite that all the properties of the logical content are ignored in the pragmatic interpretation. However, it is possible that by using language users’ intuition about whether an utterance is metaphoric or not, we can come to grips with the distinction between logical properties and encyclopedic properties of words by constraining the concept ‘metaphor’ in the way proposed here. 3. Metaphoric statements as explicature The sentence ’Ken is a real bachelor’ was discussed in the previous section. This is an expression where the word ’bachelor’ is used in what Sperber and Wilson (1986) would call a ’loose’ way (’loose use’). However, there might be disagreement among language users as to whether this should actually be called a metaphor. I now want to look at more typical and more advanced metaphors, taken both from everyday talk and from poetry. Metaphor 9 Doing so I wish to show the connection between ‘loose use’ and metaphors, at the same time showing that the analysis can also be fruitfully applied to more prototypical types of metaphor. Lakoff and Johnson (1980) in their book ‘Metaphors we live by’ describe how our understanding of the world is structured around metaphors. The first metaphor of this kind which is mentioned is ‘discussion is war’. Three sentences where this metaphor is used are: (7) a. Your claims are indefensible. b. He attacked every weak point in my argumentation. c. I’ve never won an argument with him. If we assume that the logical content of ’defend’, ’attack’ and ’win’ concerns actions of war, we see that these no longer hold in the metaphoric sentences. The encyclopedic content, we assume, contains information on what ‘defense’, ‘attacking’ and ‘winning’ amounts to in different contexts, and it is this information which is used to describe a discussion between opponents. The claims in their analysis of conceptual metaphors become vague mainly because of an uncertainty as to what different people have stored as their semantic representation of different notions. Without an already defined content – both logical and encyclopedic – it will be difficult to assess the validity of the theory. My youngest sister sometimes refers to other people as ’sheep’, as in ‘you’re such a sheep!’ as a strong way of asserting that someone has done something silly. The logical content of ‘sheep’ is set aside in this metaphor when the word is used metaphorically to characterize a person. In addition it seems to attribute some of the properties found in the encyclopedic entry for ‘sheep’, such as laziness, stupidity, indolence etc., to the person described. There are many metaphoric expressions of this type, where the properties of an animal are used to describe a person. In all such cases we see that the logical properties of the word denoting a kind of animal are not supposed to count in the comprehension process, whereas parts of our knowledge stored in the encyclopedia are activated in the process of recovering the explicature. Poetic metaphors are special because of their indeterminateness and vagueness. With this sort of creative metaphor it is often precisely ambiguity and vagueness which is the objective. But why would someone choose to express a proposition with the help of a metaphor? The answer is that certain thoughts cannot easily be expressed literally. Therefore, the untranslatability of metaphors cannot as such be used as an argument for saying that the meaning is not communicated explicitly, as an explicature. The encyclopedia contains an incredibly large number of heterogeneous assumptions. Perhaps it might also contain notions of images, smell, emotion and sound, which make the meaning difficult to express verbally. Despite the difficulties found in explaining poetic metaphors, it is possible to see certain similarities between these and the previous ones, and we find that here too it seems that the logical content is set aside: (8) a. skrivemaskinenes stormvær i rummet (‘the typewriters’ stormy weather in the room’) b. telefonkablenes nervefibre (‘The nerve fibers of the telephone cables’) 10 Waade c. Her er ditt vern mot vold, her er ditt sverd: Troen på livet vårt, menneskets verd (This is your shelter against violence, this is your sword: The belief in our life, the human value) 3 Roy Jacobsen’s use of the humane ‘nerve fiber’ to describe telephone cables, the use of ‘stormy weather’ to describe the sounds from the typewriters, and Nordahl Grieg’s use of ‘sword’ to describe human faith and hope, are all metaphoric. They use the encyclopedic properties of certain words to describe the qualities of other words. At the same time the logical content is no longer communicated. Both ‘stormy weather’, ‘nerve fibers’ and ‘sword’ lose their original basic meaning as a type of weather, human fibers, and metallic weapons of war when used metaphorically. 4. The Encyclopedia The account of metaphors proposed here depends on the assumption that our semantic representations contain both the logical properties of words and their encyclopedia. The encyclopedia obviously contains a lot of information, but in the interpretation of a metaphor, the communicator relies on the hearer’s capability to see the intended meaning. 4 How is the encyclopedic content organized in order to achieve this? This entry is internally structured in terms of the degree of accessibility of its constituent parts to various processing systems such as the utterance comprehension system. Accessibility is an organizing principle that entails more or less constant rearrangement of the internal structure of the entry; differences in both the frequency and recency of use in processing of specific items of information affect their accessibility and, at any moment, newly impinging information may effect changes in the accessibility hierarchy of the entry, or of specific subsections within it. (Carston 2002:321) Accessibility, then, is a key concept for the encyclopedia. The encyclopedic knowledge activated in the hearer’s mind in each case depends on context. In addition there is a constant reorganization of the encyclopedia. What is most available will depend on the context one finds oneself in, and the experiences one has recently had. A person who has recently seen the Norwegian children’s program ‘Portveien 2’ will have certain properties of a giraffe easily available, and these will be different from the most easily available properties of a giraffe for a person who has recently seen a documentary on giraffes. In other words, the saliency of different encyclopedic properties of lexical items is rather unstable. How much information is stored in the encyclopedia is unclear. To assume that there is too much information in the encyclopedia will have consequences for the notion of implicature. A normal analysis of ‘It is cold in here’ will in a certain context have the implicature ‘Close the window!’. If one could claim that ‘cold’ contained a lot of information in its encyclopedia, for example where the cold usually comes from, what 3 From Rolf Jacobsen’s ’En matt rute’ (a), Roy Jacobsen’s ’Byens metafysikk’ (b), and Nordahl Griegs ’Til Ungdommen’ (c). All translations are my own. 4 It is obviously not only for metaphors that the problem of organising the encyclopedia is relevant. In all utterances the speaker is dependent on the hearer’s capability to see the intended meaning. Metaphor 11 you might do to make it warmer etc, which is then activated in the right context, a consequence might be that ‘It is cold in here’ could be claimed to have the explicature ‘Close the window’, or rather ‘It’s so cold in here that you should close the window’. This is a claim I do not want to make, and it illustrates the importance of a clear idea about what the encyclopedia might contain. Not all words have logical, lexical and encyclopedic content. Only words and phrases that encode concepts can be associated with encyclopedic information. Words that encode a procedure are not associated with any encyclopedic properties. There are also certain words which are argued not to have logical content, such as proper names. In addition there might be concepts without lexical content, such as ‘best childhood friend’, which to an individual obviously might have both logical content (‘best childhood friend’) and encyclopedic properties (memories and emotions relating to this person), without having any lexical content, at least in English (Carston 2002:322). Which consequences might this knowledge have for one’s account of metaphors? As we have seen, the encyclopedic content has a special meaning for interpreting metaphors, and the question then becomes how words without an encyclopedia can be used metaphorically? ’something’ in ’something’s happened’ do[es] not obviously give access to a rich set of encyclopedic assumptions from which the more specific intended concept is built. These enrichments are effected in some other way, relying on inferences that follow more directly from the presumption of relevance itself: (…) that something out of the ordinary has happened (Carston 2002:335) If it is possible for words without an encyclopedia to be used metaphorically, this will be strong counter evidence for the theory that metaphor meaning is an explicature, because of the role the encyclopedia plays for this theory. However, I have not been able to find any such metaphors (for example using ‘something’), even though this in itself obviously does not exclude the possibility that there might be some. Typical words without encyclopedic properties are pronouns, articles and conjunctions, or more generally words that encode a procedure rather than a concept. Usually it is members of the open word classes nouns, verbs, adjectives, and whole predicate phrases that have become lexicalized, which are used metaphorically, but I do not want to exclude the possibility that there might also be other types. If words without an encyclopedia cannot be used metaphorically, the explicature analysis of metaphors would seem to be supported. While metaphoric meaning is dependent on the contents of the encyclopedia, expressions involving irony, where the meaning is an implicature of the utterance, is independent of it. This is an important criterion for explicatures: they build on the semantic knowledge we have of the words in the utterance. Therefore the implicature of an utterance can be almost anything, dependent on context. Saying ‘it is cold in here’ could have the implicated meaning that speaker wants somebody to close the window, if it is cold outdoors, or that speaker wants someone to open the window, if he for example finds himself in a cold basement on a warm summer’s day. Metaphors do not give the same opportunity to convey messages that are contrary to what is linguistically encoded for making this kind of opposing messages. ‘Peter is a lion’ cannot mean either ‘Peter is a coward’ or ‘Peter is a tough guy’, depending on context. One could obviously make an ironic remark, so that ‘Peter is a lion’ would mean ‘Peter is a coward’, but then it will not be the metaphor which 12 Waade expresses the concept ‘coward’ – the metaphor still tells us that ‘Peter is a tough guy’, but because the utterance is used ironically, it expresses the idea that ‘Peter is a coward’. 5. Objections to the analysis: Elisabeth Camp There are scholars who still claim that metaphor meaning is a part of the implicature of the utterance, and not its explicature. In this section I wish to take a closer look at Elisabeth Camp’s contribution to the debate as expressed in the article ’Contextualism, Metaphor and What is said’, showing why her objections do not constitute any real threat to the belief that metaphor meaning is a part of the explicature of an utterance. Camp opens the discussion by looking at the importance of context: Metaphor is a deeply context-sensitive linguistic phenomenon (…). The particular assumptions she [speaker] intends her hearers to employ in determining her metaphorical content can vary considerably across different conversational contexts, producing a wide variety of possible meanings. (Camp 2006:1) Camp is obviously correct in that context is important in deciding the meaning of an utterance. This, however, goes for any utterance, and is not peculiar to metaphors. In addition I want to claim that metaphor meaning generally in itself is very constrained and fixed to the semantics of the word, as I have argued above. Camp uses the metaphor ‘Bill is a mouse’ as an example. The way I see it, this metaphor can only have a meaning which is closely connected to the meaning of the word ‘mouse’: it can relate to the size of the mouse, its fear of larger animals, its fear of dangers we see as irrelevant etc. Even with my contextual deficit in regard to how the sentence is to be understood, I will be capable of figuring out more or less what it means. What it might implicate, on the other hand, is obviously something completely different, and also something I will have difficulty understanding without more knowledge of the context. Whether the speaker by uttering ‘Bill is a mouse’ means ‘I don’t think you should become Bill’s girlfriend’, the teasing ‘Bill, you daren’t even do that!’, or almost any other implicature, is completely unavailable to me without knowledge of the context. In this way, Camp is obviously correct in that we are dependent on context to decide the meaning of an utterance. This would however also have been the case in a literal utterance, for example ‘Bill is a sissy’ or something similar. When it comes to the metaphor analysis itself, Camp (2006) claims that: [speaker] knowingly says something which, if she meant it, would commit her to the claim that Bill is a small rodent (…). [Speaker] won’t intend to be taken as committing herself to such an absurdly false claim; her hearers realize this, and interpret her metaphorically instead (…). This broadly Gricean model is both intuitively plausible and theoretically satisfying. (Camp 2006:1) It is difficult to take this claim seriously. Grice’s model the way it is sketched here is in no way intuitively plausible or theoretically satisfying. It gives no explanation for why the speaker would think of saying something that she won’t ‘be taken as committing herself to’, or for how it is possible for the hearer to find out what the speaker might have meant by saying ‘Bill is a mouse’. How is the search for meaning constrained on her account? The account gives us very little relevant information on how we understand metaphors. Metaphor 13 Camp also describes the poetic metaphors as a problem area, holding them up as evidence for seeing metaphoric content as implicated: It is often not appropriate to report the speaker of a highly poetic metaphor as having said her metaphorical content. For instance, having considered (11) The hourglass whispers to the lion’s paw, in its original context of utterance (by W.H. Auden in ‘Our Bias’, translating a poem by Alex Sitnitsky), I am willing to hazard the poet means something like that every source of activity and forcefulness is ultimately undone by the passage of time. I’m not at all certain of this. But even if I were, I’d still be very hesitant to report the poet as having said this, precisely because his metaphor is so clearly intended to be elusive and allusive (Camp 2006:8, my italicization) In order to further discuss this problem, it is essential to have clear definitions on what is meant by ‘saying’, and also the concepts ‘implicature’ and ‘explicature’. Although I agree with Camp in that it is difficult to give a literal interpretation of the metaphor above, I disagree with her on how this difficulty should be accounted for. The allusiveness of the metaphor should, in my opinion, be accounted for in terms of how our mental lexicon is organized, not by claiming that it must be an implicature. However, in poetry, the explicated meaning of the metaphor may give rise to a number of implicatures. In this case, where the exact meaning of the metaphor is difficult to retrieve, the two (explicature and implicature) may be intertwined, and confusion as to what is what may arise. 6. Conclusion In this article I have investigated the claim made by Robyn Carston (2002) that the meaning of metaphors is part of the explicature of an utterance. I have proposed to modify her analysis with respect to the role of the logical content of a word when it is used metaphorically. As I have claimed above, it seems that the logical content of the word is simply deleted, or ignored, when the word is used metaphorically. We draw on a context accessible to us in our selection of encyclopedic properties which are relevant for appreciation of the meaning conveyed by the metaphor, so that only a small subset of all the properties that we associate with a certain word is brought to bear in the interpretation. This, however, is not peculiar to metaphors, but pertains to all language use. Rarely, if ever, will the entire encyclopedic knowledge we have of a word be employed in a given context. The difference between metaphorical and literal language is thus a difference in whether the logical content does or does not play a role in the process of understanding what concept a given word used metaphorically represents. Any pragmatic theory wishing to address the relationship between implicature and explicature – or ‘what is said’ and ‘what is implicated’ – must include metaphors as an important part of this discussion. The way I see it, there are strong arguments in favor of claiming that the meaning of metaphors is part of the explicature of an utterance. Still, there are areas that could need further clarification, such as how to define the contents of the logical entry and the encyclopedia, in order to obtain clarity in this central relevance- theoretic problem. 14 Waade References Camp, Elisabeth (2006) ‘Contextualism, Metaphor, and What is Said’ in (forthcoming) Mind and Language 21, 2006 Carston, Robyn (2002) Thoughts and utterances: The Pragmatics of Explicit Communication, UK: Blackwell Publishing Lakoff, George and Mark Johnson (1980) Metaphors we live by, Chicago: University of Chicago Press Sperber, Dan, and Deirdre Wilson (1986) ‘Loose Talk’ in Davis, Steven (ed.) Pragmatics: A Reader 1991, 540-9, New York: Oxford University Press Author’s e-mail address: randiwaade@yahoo.com WORKING PAPERS ISK 3/2006 15-26 ISK, NTNU On the functional independence of explicatures and implicatures Thorstein Fretheim Robyn Carston has proposed that, within Relevance Theory the proposition expressed by an utterance U (its explicature) must play a communicative role that is independent of the role played by any implicature of U. Thus no implicature can entail the explicature of U, and whenever it looks as if that sort of situation obtains, the presumed implicature should be redefined as the explicature. This paper shows that the assumption that an explicature may not be entailed by an implicature cannot be maintained, but the fact that explicatures and implicatures are not always functionally independent of one another is argued not to be a problem for Relevance Theory. 1. Pragmatic processes at the explicit and the implicit level of communication There is a general agreement that context and context-sensitive pragmatic processes contribute to the truth conditions of explicitly communicated propositions, but opinions diverge as soon as one starts discussing the extent to which pragmatically derived information should be allowed to intrude into the proposition expressed by an utterance. Adherents of relevance theory (Sperber and Wilson 1986, 1995; Carston 2002) claim that relevance-driven pragmatic inference plays just as important a role in the recovery of the proposition expressed as in the derivation of implicatures, and that the set of truth conditions encoded by a given sentence hardly ever exhausts the set of truth conditions for the proposition that a speaker communicates. Carston says, If all pragmatically derived elements are treated as implicatures we are left with no candidate for what is said, or the explicit utterance content, other than the logical form of the linguistic expression used, which is standardly subpropositional. But we do not communicate logical forms (though we do communicate via logical forms), and we do not retain logical forms in memory; we communicate and remember assumptions or thoughts or opinions, which are fully propositional. So a distinction has to be made between pragmatic inference that contributes to the recovery of the explicitly communicated content and pragmatic inference that eventuates in implicated assumptions. (Carston 2002: 107) The view expressed in the final sentence of this quotation is expressed even more strongly, and with some added implications, by Recanati (1989, 1993, 2004), because Recanati makes a distinction between ‘primary pragmatic processes’, which are pre- propositional and outside the conscious awareness of the interpreter, and ‘secondary 16 Fretheim pragmatic processes’, which are said to be post-propositional, and which the interpreter is consciously aware of. The former are local processes that contribute to the hearer’s understanding of what is said and are not considered to be inferences in the strict sense; the latter are believed to be true inferences of a global sort. What Recanati labels the Availability Principle is based on the assumption that two qualitatively different types of pragmatic process – primary and secondary – are involved in utterance interpretation. Processing that is required for the determination of the truth conditions of the proposition expressed (‘what is said’) happens at a sub-conscious level. The output of such processes is consciously available but the processes themselves are not. True inferential processes on the other hand are consciously available and are the ones involved in the derivation of conversational implicatures, at least the particularized ones (PCI’s) whose existence depends on inferences that are not of a default sort. Recanati’s position on the difference between two kinds of pragmatic process is generally rejected by leading relevance theorists, who think that all pragmatic processes are inferential processes of an unreflective nature, and that the processing of explicatures and implicatures happens in parallel, so that the latter kind of inferential process does not depend on the hearer’s mental representation of a pragmatically fully developed proposition, an explicature. Recanati himself emphasizes in his most recent work that, even if what is said is logically prior to the working out of implicatures, it is not temporally prior in the sense that recovery of what is said takes place before any implicatures can be computed (Recanati 2004: 47), but he still finds reasons to maintain his differentiation between primary and secondary pragmatic processes. Carston (1988) is a very influential paper that deals with, and defends, the functional autonomy of explicatures and implicatures in the pragmatic processing of utterances. Let us look at one of her examples (Carston 1988: 155). (1) A: How is Jane feeling after her first year at university? B: She didn’t get enough units and can’t continue. Carston notes that there are certain words in B’s answer that are in need of disambiguation. This is true of the verb get and the noun unit, she says (though today she would presumably prefer to say that both get and unit are conceptually underspecified words that are subject to ad hoc development in context, and that no lexical ambiguity is involved here). No one would deny that on the implicit side of the pragmatic analysis of B’s answer is the derived assumption that Jane regrets the situation she finds herself in, but that implicature is based on an interpretation of the phrase can’t continue as ‘cannot continue with her university studies’ and the phrase enough units as ‘enough university course units to qualify for admission to second year study’. What criteria do we possess, which enable us to decide whether these pragmatic developments of the grammatical phrases enough units and can’t continue belong to the proposition expressed or are conceptual constituents of an implicature? As part of the answer to that question Carston suggested a criterion of functional independence of the explicature of an utterance and any implicatures of the same utterance: the Functional Independence Principle. Adopting this principle means that any alleged implicature that is seen to subsume all contextual effects that could be derived from the proposition expressed by an utterance cannot maintain its status as implicature, rather it should be redefined as the explicature, a hybrid mental representation made up of conceptual structure provided via the logical form on the one hand and context-sensitive inferences on the other. However, Carston hastens to add that “while it is instructive to consider such criteria, they might well be seen as rather On the functional independence of explicatures and implicature 17 superficial, descriptive principles, if not ad hoc”, because, she says, “they follow from a single principle directing utterance interpretation, the principle of relevance, which is itself embedded in a general theory of human cognition and communication, the relevance theory of Sperber & Wilson (1986).” (Carston 1988: 156). As an example of what she means by saying that the explicature of a given utterance must play a communicative role that is independent of the role of any implicature derived via pragmatic processing, and vice versa, Carston considers the exchange between A and B in (2). (2) A: Have you read Susan’s book? B: I don’t read autobiographies. implicated premise: Susan’s book is an autobiography. implicated conclusion: B has not read Susan’s book. (Carston 1988: 157) In (2) the truth conditions pertaining to the implicatures ‘Susan’s book is an autobiography’ and ‘B has not read Susan’s book’ are independent of the truth conditions of the explicature ‘B does not read autobiographies’. There is no overlap in content between the explicature and the implicatures, which function independently of the explicature as premise and conclusion in a process of non-demonstrative inference (cf. Sperber and Wilson 1986). Suppose the functional independence criterion is found not to hold water. Would a disclosure of the empirical shakiness of the criterion have consequences for the relevance-theoretic view of the semantics/pragmatics distinction and the distinction between explicit and implicit communication? I am going to argue that Carston’s Functional Independence Principle should not be accorded any role in a cognitively based theory of utterance interpretation like relevance theory. There is rather strong empirical evidence that this principle is not tenable; on the other hand, the viability of relevance theory does not in any way hinge on it. 2. An explicature can entail an implicature Carston has later observed that there can be no incompatibility between the fact that an explicated proposition p entails a proposition q and the assumption that the entailed proposition q is an implicature derived through the hearer’s processing of the explicature p in a specific context (see Carston 2002: 137-41, 189-91). The existence of situations where an implicature of an utterance is an entailment of the explicature of the same utterance would appear to be a counterexample to the Functional Independence Principle. If the implicature is entailed by the explicature of the utterance, the two communicated assumptions cannot both be autonomous, they are not independent of one another. Carston does not draw the conclusion that the existence of implicatures that are entailed by the proposition expressed – as in (4) B below – is bad news for the relevance- theoretic view of the contrast between explicated and implicated propositions. In contrast to (3), the corresponding answer in (4), adapted from Carston (2002), implicates that B did buy vegetables, because B bought a squash and an eggplant. In (4) the implicated conclusion is entailed by B’s explicated premise. 18 Fretheim (3) A: Did you buy any vegetables? B: I bought some apples and pears. (implicature: B did not buy any vegetables) (4) A: Did you buy any vegetables? B: I bought a squash and an eggplant. (implicature: B bought some vegetables) Carston concludes, “In my view, the concept of ‘entailment’ and the concept of ‘implicature’ belong to different explanatory levels or different sorts of theory, the one a static semantic theory, the other a cognitive processing pragmatic theory.” (Carston 2002: 141).5 3. An implicature can entail an explicature I am now going to show that we also have to allow for the existence of situations in which an implicature entails the explicature of an utterance. Carston’s independence principle was meant to give pragmatists a criterion that would enable them to tell explicatures and implicatures apart in cases of doubt. Her argument was that, if the meaning of the explicature is fully included in the meaning of an implicature derived on the basis of the explicature and a set of contextual assumptions, then the explicature will be redundant, as no contextual effect can be derived from the explicature which is not at the same time derivable from the entailing implicature. The thought represented by the explicature is duplicated in a truth-conditionally stronger implicature. The Functional Independence Principle supports Carston’s own proposal that all Gricean generalized conversational implicatures be eliminated, including the reasoning from ‘p and q’ to ‘p and then q’ (‘conjunction buttressing’) and the reasoning from ‘if p, then q’ to ‘if and only if p, then q’ (‘conditional perfection’), where the stronger readings are said to be due to generalized conversational implicatures, Levinson’s GCI’s (Levinson 2000), which are triggered by the presence of particular linguistic elements. For relevance theorists all legitimate implicatures are of the particularized sort (PCI’s), the outcome of ‘one-off inferences’. The claim that GCI’s should be dispensed with is based on the fact that relevance theorists embrace a view of explicit communication as much more context-dependent and inference-driven than what had been assumed by Grice and the neo-Griceans. An explicature is defined as an ostensively communicated assumption which is inferentially developed from a truth-conditionally incomplete encoded logical form (Sperber and Wilson 1986; Carston 2002). All standard examples of GCI, which a neo-Gricean like Levinson (2000) regards as a preferred or normal interpretation based on default inference, are redefined as explicatures in the framework of relevance theory. Like any other explicature, the ones that Grice defined as GCI’s are conceived by relevance theorists as the result of a union of a grammar-dependent logical 5 Burton-Roberts (2005) strongly disagrees with Carston that an entailment of the explicature of an utterance can be implicated. When you explicate a proposition p, his argument goes, then you explicate the truth-conditional content of p, which includes all entailments of p. Hence if an entailment of p is implicated, then p is implicated. To deny this would be to construe ‘entailment’ in a novel, idiosyncratic and unprecedented way. On the functional independence of explicatures and implicature 19 form and context-based inference supported by the communicative principle of relevance.6 The reasoning that underpinned Grice’s argument that a coordinating connective like English and gives rise to what he called a generalized conversational implicature was prompted by his intention to prove that natural language operators like conjunction and disjunction connectives and conditional connectives work exactly like the corresponding operators of propositional logic. The fact that we often understand a speaker’s use of the connective and as if she had actually said and then is due to factors that are extraneous to whatever information the language code can give us. Such inferences about what the communicator means by what she says were claimed by Grice to be due to principles of conversation, his Co-operative Principle and his categories of conversational maxims (Grice 1975, 1989). A given English conjunction of clauses or verb phrases conforming to the general formula ‘p and q’ may implicate ‘p and then q’ in one context, ‘p and therefore q’ in a different context, etc. More strongly than any other publication, Carston’s 1988 paper offered a very clearly expressed alternative to the standard implicature analysis of the temporal and causal implications of the use of and- conjunction. If it turns out that we have to allow for the existence of situations where an implicature entails the proposition expressed, the usefulness of the Functional Independence Principle as a heuristic will be seriously diminished. Carston herself gives vent to the view that its importance should in fact not be overrated, saying, However, it is not clear to me that ‘functional independence’ is worth any kind of vigorous defence; it was in fact intended as only a useful heuristic and should probably never have been elevated by the label ‘principle’ at all. I was (and am still) of the view that the communicative principle of relevance itself or, more particularly, the comprehension strategy that follows from it, effects a sorting of pragmatic inferences into contributions to the proposition expressed (explicature) and implicatures, and so subsumes whatever correct predictions ‘functional independence’ might make.7 (Carston 2002: 191) On the other hand, empirical evidence that might cause the demise of the Functional Independence Principle has been scarce. Having addressed an argument that Recanati (1989) presented against her Functional Independence Principle, Carston (2002: 191) retorted, “A more compelling counterexample to the principle would involve a communicated assumption which (…) is clearly an implicature, but which is an implicated premise (rather than an implicated conclusion) and entails what is said; I have not come across such a case.” (her emphasis). She did not explain why she feels an implicated premise that entails what is said would be more damaging to the RT view of the saying-implicating distinction than an implicated conclusion that entails what is said. I have chosen not to cite Recanati’s fairly complicated counterexample in the present paper. However, Burton-Roberts (2005), which is a review of Carston (2002), presents data that may be a problem for Carston’s proposal about the functional 6 The communicative principle of relevance says that, ”Every act of ostensive communication communicates a presumption of its own optimal relevance” (Sperber and Wilson 1995: 260). 7 The relevance-theoretic comprehension strategy enjoins addressees to construct interpretation in order of accessibility and to stop when their expectation of relevance is satisfied. 20 Fretheim autonomy of explicatures and implicatures. The kind of example on which his argument rests is given in (5).8 (5) A: There’s no milk. B: The milkman’s ill. B’s answer implicates, by way of a PCI, that there is no milk because the milkman is ill. In other words, what B implicates subsumes what B says, or explicates, which is no more than that the milkman is ill. PCI’s are a type of implicature whose existence has never been denied by relevance theorists, yet this is an implicature that entails the explicature of B’s utterance. On Carston’s account the thought ‘There is no milk in the house because the milkman is ill’ should preferably be explicated by B’s utterance of The milkman’s ill. However, the explicated assumption that he is ill seems to be functionally independent of the explicature of A’s utterance in (5) and independent of any inferred causal relation between B’s explicature and A’s explicature. Still, I am not sure what would stop a relevance theorist from saying that the logical form of B’s sentence can be developed into the explicature ‘There is no milk in the house because the milkman is ill’. It seems to me that Burton-Roberts’ milkman example leads to a paradox for relevance theory, although Burton-Roberts himself did not intend his counterargument to the Functional Independence Principle to be interpreted that way. He insists (p.c.) that the causal relation between the premise p and the conclusion q in (5) cannot be explicated even for Carston. Now consider the minimally different question-answer pairs in (6) and (7).9 (6) A: What happened when that bug flew into your mouth? B: I swallowed it. (7) A: What happened when that bug flew into your mouth? B: I swallowed. The linguistic intuition of native speakers of English is such that they understand the proposition expressed by the utterance of (7) B differently than the proposition expressed by the utterance of (6) B. The version in (6) with an overt direct object argument is a direct way of saying that B swallowed the bug; the version in (7) with no complement after the verb is an indirect way of conveying that information. Non-native speakers of English generally feel the same about this pair, because they draw on their native language competence and their own language shows exactly the same lexical difference between an intransitive verb and a phonologically identical transitive verb that may both be glossed as ‘swallow’. I am going to argue that there is no process of enrichment that takes you from the encoded logical form of B’s answer in (7) to an explicature that equals the explicature communicated by B’s answer in (6). We understand the answer in (7) to convey an explicit premise that opens for the implicated conclusion that the bug was passed down into B’s stomach when B swallowed. The addressee of (7) B is aware of what is said and is capable of working out the inferential connection between what is said (‘I swallowed’) 8 I am grateful to Noel Burton-Roberts for pointing footnote 8 of his review article out to me (Burton- Roberts 2005: 397). 9 Anne Bezuidenhout’s paper ‘The Semantics/Pragmatics boundary’ (2005) aroused my interest in the verb swallow and its correspondents in other languages. On the functional independence of explicatures and implicature 21 and what is implied by what is said (‘My swallowing caused me to swallow the bug’). This is a ‘post-propositional’, ‘secondary’ pragmatic process in Recanati’s sense. It meets his availability condition, unlike the ‘primary’ pragmatic processes of mandatory saturation like the resolution of the reference of the pronoun it in (6). Bach (1994) introduced the notion of ‘impliciture’, standing for what is implicit in what is said, as what he considered to be a viable alternative to the relevance-theoretic ‘explicature’.10 He makes a terminological distinction between ‘conceptual incompleteness’ and ‘semantic incompleteness’. Conceptual completion, or what Bach calls ‘expansion’, is sometimes required to arrive at the ‘impliciture’ of an utterance. This is for him a case of free enrichment. Semantic completion on the other hand is when the context supplies an argument for an implicit argument role. In such cases Bach’s ‘impliciture’ would be arrived at via saturation, or what Bach calls ‘completion’. When someone says I’ve finished, there may be an implicit argument role there for the activity that was finished, and context will help in the recovery of the event which occupies the implicit argument role. The same is presumably true of the null-complement of the verb notice in (8). (8) A: William was very silent today. B: Yes, I noticed. A process of mandatory saturation is required for the hearer’s mental representation of what B noticed. Though there is no overt complement after the verb notice, this is nevertheless a transitive verb whose argument structure includes an argument that refers to the object of the act of noticing. The verb eat is similarly conceptually transitive even when it has no overt object argument. Could the verb swallow be analyzed in the same way as finish, notice and eat? Does the inferential processing of B’s answer in (7) involve mandatory saturation, Bach’s ‘completion’? Hardly, because the intransitive verb swallow encodes a different concept than its transitive counterpart. There is no hidden or overt argument referring to that which is swallowed when swallow is used as an intransitive verb. Nor is it possible to let a sentence with the intransitive verb swallow undergo ‘free enrichment’ that results in an explicature that represents an unmentioned affected object. If you wish to explicate the information that a particular solid object or liquid stuff was swallowed, you have to fill in the direct object slot, even if the linguistic filler is just a pronoun that represents a contextually very salient discourse entity, as in (6). In order to convince you that intransitive and transitive uses of swallow are semantically distinct, I suggest we consider the pair of (9)-(10). (9) Mark swallowed twice. (10) #Mark swallowed it/something twice. An utterance of (9) causes us to infer that Mark swallowed without swallowing anything. There may have been nothing in Mark’s mouth that could possibly be swallowed. (9) is meaningful but (10) will not normally be, because when something has been swallowed it is hard to see how it could be swallowed once more. The American Heritage Dictionary of the English Language lists two basic, non-figurative meanings of the verb swallow, 10 What I’ve written in this paragraph is in large part due to personal communication with Anne Bezuidenhout. 22 Fretheim one of which corresponds to its meaning when it is a transitive verb and the other one to its meaning when it is used intransitively. According to this dictionary, the transitive verb swallow means “to cause (food, for example) to pass from the mouth via the throat and the esophagus into the stomach by muscular action”, and “ingest” is suggested as another verb with that meaning. Under the same dictionary entry we also find an intransitive use of the verb swallow, whose meaning is defined as “to perform the act of swallowing”. The Oxford Paperback Dictionary & Thesaurus similarly tells us that the verb swallow has two meanings, one for its transitive and one for its intransitive use. The former is “to cause (food, drink, etc.) to pass down the throat”, and the latter “to move the throat muscles as if doing this, especially through fear”, where demonstrative this refers to the action denoted by the transitive verb. Finally, Shorter Oxford Dictionary distinguishes between the lexical meaning of the transitive verb swallow, which is said to be “to take into the stomach through the throat and gullet, as food or drink”, and the intransitive verb, whose meaning is entered as “to perform the act of deglutition, as in an effort to suppress emotion” (“deglutition” is a scientific word for “swallowing”). The muscular action involves shutting of the epiglottis so that the entrance to the trachea is shut off, and the intransitive verb swallow encodes no more information than the phrase ‘perform an act of swallowing’ does. In contrast, the transitive verb swallow encodes the complex deglutition concept, which necessarily includes the physical mechanism encoded by intransitive swallow as an instrument. Suppose we change A’s utterance in the above pair (6)-(7) slightly and replace past tense swallowed by the present perfect tense form have swallowed. (11) A: So what happened to the bug that flew into your mouth? B: I’ve swallowed it. (12) A: So what happened to the bug that flew into your mouth? B: #I’ve swallowed. If we accept the hypothesis that it is possible to let an utterance of the sentence I’ve swallowed undergo free enrichment so that the contextually derived explicature will be something like ‘I have swallowed the bug that flew into my mouth’, then we have no explanation for our feeling that B’s utterance I’ve swallowed is incoherent in the context provided by A’s question in (12), a question about the fate of the poor bug. To the extent that (12) B makes us think of a deglutition process of swallowing the bug, it does so in spite of the syntactic form of the sentence used by B, which is not an acceptable way of communicating what was said in (11). While the explicature of B’s answer in (7) above leads us straightforwardly to the truth-conditionally more constrained implicated conclusion that B swallowed the bug, B’s answer in (12) tells us that B performed an act of swallowing and leaves us with a feeling that the answer is either not relevant or produced by someone whose English proficiency is rather poor.11 Observe that my example with the verb swallow involves no paradox similar to what we experience with Burton-Robert’s example. It is not so that the thought ‘B swallowed the bug’ arising from (7) B is implicated according to one criterion and explicated according to a different criterion. My argument was that there is no reason to 11 Sperber and Wilson’s ’presumption of optimal relevance’ (Sperber and Wilson 1995: 270) includes the following statement: The ostensive stimulus is the most relevant one compatible with the communicator’s abilities and preferences (emphasis mine). On the functional independence of explicatures and implicature 23 even consider an explicature analysis to be a possible candidate in (7), because native speakers of English do understand the proposition expressed by B in (7) to be ostensively distinct from the proposition expressed by B in (6), and the noted difference in acceptability between (11) B and (12) B corroborates the evidence presented earlier. Why do I insist that there is a genuine difference between my own example of an implicature that entails the explicature of an utterance and Burton-Roberts’ mentioned example of the same phenomenon? Let us repeat his data for convenience. (5) A: There’s no milk. B: The milkman’s ill. In Burton-Roberts’ illustration rendered in (5), the causal relation between B’s proposition (reason) and A’s proposition (consequence) is arguably expressed in a direct manner. B could have given the same answer by using the more complex sentence structure ‘There’s no milk because the milkman’s ill’, which requires more decoding but less context-dependent inference. What B did instead was to suppress repetition of information that would just be an echo of what A said in (5). B could also have used the more clearly elliptical structure ‘Because the milkman’s ill’ to convey exactly the same as in (5) where the causal connective was left out: (5’) A: There’s no milk. B: Because the milkman’s ill.12 The pragmatic processing of what B says in (5) arguably involves a pragmatic completion process, a relevance-theoretic enrichment. At the same time there are also strong arguments for an analysis of B’s utterance as one which gives rise to a particularized conversational implicature that subsumes the explicature. My example (7), repeated here, is different. B’s ‘when’-clause in (7’), an echo of A’s identical clause, does not change our interpretation. What we see in (7’) is the intransitive verb swallow, grammatically and conceptually, even if the given information expressed in the temporal clause of A’s question is echoed in B’s answer. The assumption that B swallowed the bug is no more explicated in (7’) than in (7). (7) A: What happened when that bug flew into your mouth? B: I swallowed. (7’) A: What happened when that bug flew into your mouth? B: I swallowed when that bug flew into my mouth. Moreover, it is impossible to preface the causal connective because to B’s answer in (7): (7”) A: What happened when that bug flew into your mouth? B: #Because I swallowed. There is a suppressed proposition in (5) B whose recoverability depends on the content of interlocutor A’s utterance. In (7) B there is no suppressed proposition that is brought into 12 I admit that B’s answer with the prefaced connective because would be more natural, or more likely, if A had said There’s no milk. Why?, but (5’) B is by no means impossible. 24 Fretheim focus by the content of A’s preceding utterance. I conclude that the assumption that speaker B swallowed the bug is conveyed in a truly indirect way in (7), while the assumption that there is no milk because the milkman is ill is conveyed in a less obviously indirect way in Burton-Roberts’ example (5). Neither Carston nor any other relevance theorist has thus far offered us necessary and sufficient conditions that enable us to decide where free enrichment at the explicit level of communication stops and particularized conversational implicatures take over.13 A relevance theorist could, if so inclined, argue that B’s answer in (5) directly explicates the information that there is no milk because the milkman is ill, in which case (5) would be no threat to the Functional Independence Principle. In contrast, the explicature of B’s answer I swallowed in the exchange between A and B in (7) can be paraphrased as ‘I performed an act of swallowing’, and this is a self-contained communicated thought. B did not mean to swallow but it unfortunately happened, and the (implicated) consequence was the bug’s being passed down into B’s stomach. By saying I swallowed, B did not explicate either ‘I swallowed, so I swallowed the bug that was in my mouth’ or the truncated version ‘I swallowed the bug that was in my mouth’ which disregards the semantic difference between intransitive and transitive swallow. 4. Conclusion My example involving the verb swallow, and the argument that rests on it, showed that there are situations where an implicature of an utterance entails its explicature, but the fact that an explicature can be entailed by an implicature is not a problem for relevance theory, it affects none of its fundamentals. We still have what Carston (2002: 191) refers to as embedding tests, of which the most famous one is Recanati’s Scope Principle (Recanati 1989). The Scope Principle says that a pragmatically determined aspect of meaning is part of what is said (thus no conversational implicature) if it falls within the scope of logical operators such as negation or a conditional connective. When applied to conjunction data, this test reveals that the strengthened interpretations involving a temporal sequence or a causal relation between the conjuncts come out as explicatures, or what is said in Recanati’s sense. I would like to add that, even if the explicature of B’s utterance in my example (7) is entailed by an implicature communicated by means of that utterance, this fact does not necessarily mean that the explicature is made redundant (because it would yield no contextual effects that are independent of the contextual effects of the entailing implicature). It cannot be literally true that speaker B’s utterance of the sentence I swallowed in (7) yields no contextual effect apart from what is attributable to the implicature ‘B swallowed the bug that had flown into B’s open mouth’. The explicature of B’s utterance in (7) possibly makes it easier for the hearer to infer that B swallowed1 (let me use subscript 1 for the concept encoded by the intransitive verb swallow) by mistake. He should not have swallowed1 at that particular time, because by so doing he could not help swallowing2 (let me use subscript 2 for the concept encoded by the transitive verb swallow) the bug that had flown into his mouth. (6) B – I swallowed it – is compatible with the assumption that B deliberately swallowed2 the bug. (7) B – I swallowed – strongly suggests that it happened by accident. 13 Lack of clarification concerning the limits of processes of free enrichment relative to derivation of implicatures may at the end of the day turn out to be a nagging problem for relevance theory, but that issue would take us beyond the topic of the present paper. On the functional independence of explicatures and implicature 25 Still, this possible pragmatic difference between (6) B and (7) B is not the main reason why I think the kind of data that I have focused on in the present paper should not worry relevance theorists, even though I do think the mentioned semantic properties of transitive and intransitive swallow show that the Functional Independence Principle should be abandoned. In my view, the relevance of B’s utterance in (7) is not dependent on the inferred assumption that the swallowing1 happened by mistake. There is absolutely no reason to doubt that the explicature of the utterance of (7) B is sufficiently relevant in spite of the fact that it is entailed by the strongly communicated implicature that B swallowed the bug. After all, Carston herself has insisted that the concept of ‘implicature’ belongs to a cognitive theory of on-line processing of information, while the concept of ‘entailment’ belongs to a static theory of semantics. One major contextual effect of B’s answer in (7) is the explicated information that B swallowed1. I would go as far as to claim, presumably against the opinion of Carston and other relevance theorists, that for the addressee A the mental representation of the communicated assumption that B swallowed (when the bug had flown into his mouth) is temporally prior to A’s mental representation of the implicated conclusion that B swallowed2 the bug when he swallowed1. In my view the temporal sequence is no less crucial here than in an imagined situation where B actually says the following, in a discourse where there has been no prior mention of a bug: I swallowed – so I swallowed a bug, with a temporal break, for rhetorical effect, between the two clauses. Acknowledgement I am indebted to Anne Bezuidenhout and Noel Burton-Roberts for very useful discussions. Any weak points in the argumentation are my responsibility alone. References Bach, Kent, 1994. Conversational impliciture. Mind and Language 9: 124-62. Bezuidenhout, Anne, 2005. The Semantics/Pragmatics boundary. In 2nd edition of The Encyclopedia of Language and Linguistics. Elsevier Publishers. Editor-in-Chief: Keith Brown. Burton-Roberts, Noel, 2005. Robyn Carston on semantics, pragmatics and ‘encoding’. Journal of Linguistics 41: 389-408. Carston, Robyn, 1988. Implicature, explicature, and truth-theoretic semantics. In: Ruth M. Kempson (ed.), Mental Representations: The Interface between Language and Reality. Cambridge: Cambridge University Press, 155-82. Carston, Robyn, 2002. Thoughts and Utterance: The Pragmatics of Explicit Communication. Oxford: Blackwell Publishing. Grice, H. Paul, 1975. Logic and conversation. In: Peter Cole and Jerry L. Morgan (eds.), Syntax and Semantics Vol. 3. New York: Academic Press, 41-58. Grice, H. Paul, 1989. Studies in the Way of Words. Cambridge MA: Harvard University Press. Levinson, Stephen C., 2000. Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge MA: The MIT Press. Recanati, François, 1989. The pragmatics of what is said. Mind and Language 4: 295- 329. 26 Fretheim Recanati, François, 1993. Direct Reference: From Language to Thought. Oxford: Blackwell Publishing. Recanati, François, 2004. Literal Meaning. Cambridge: Cambridge University Press. Sperber, Dan and Deirdre Wilson, 1986. Relevance: Communication and Cognition. Oxford: Blackwell Publishing. Sperber, Dan and Deirdre Wilson, 1995. Relevance: Communication and Cognition. With a Postface. Oxford: Blackwell Publishing. Author’s e-mail address: thorstein.fretheim@hf.ntnu.no WORKING PAPERS ISK 3/2006 27-39 ISK, NTNU Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first'): A pragmatic analysis based on a univocal lexical meaning Thorstein Fretheim 1. What this paper is not about There is more than one temporal adverb først ('first') in Norwegian. One lexical item that will not concern me is the word that translates as erst ('first') in German but as only in English (not as first or at first) and as seulement ('only') in French (not as d'abord). This is an inherently negative focus marker which does not contribute to the proposition expressed but instead directs the addressee's attention to a higher-level explicature in the relevance-theoretic sense,14 a communicated expression of the speaker's attitude to the information offered, like the near-synonymous expression ikke før with the meaning of 'not until'. One example will suffice; it is taken from the Oslo Multilingual Corpus (OMC), a bidirectional translation corpus: (1) Norwegian source text: Først da han kommer opp på kvisten om kvelden, husker han at han glemte å nevne noe om firmanavnet. (BHH1)  English target text: Only when he enters his attic room in the evening does he remember that he forgot to say anything about the name of the firm. 2. Anaphoric and non-anaphoric first and først The topic of this paper is on the one hand what I shall refer to as the 'non-anaphoric' item først which corresponds to German zuerst, English first or at first and French d'abord, as in først P, siden Q ('first P, later Q'), and on the other hand, the use of først, and first, which reveals, upon close examination, that it is a discourse anaphor that contributes to the truth-conditional content of the utterance and must be pragmatically enriched ('saturated'; see Recanati 2004) through association with an antecedent located in the preceding discourse. The example in (2) illustrates the non-anaphoric use of først. (2) Norwegian source: Da jeg takket henne etter vår siste kraftprøve, så hun først alvorlig på meg, så begynte hun plutselig å fnise med hånden over munnen. (BHH1)  English target: When I thanked her after our last test of endurance, she at first looked seriously at me, then began to giggle with her hand over her mouth. 14 A higher-level explicature is a proposition embedded under a propositional attitude description or a speech-act description. See for instance Blakemore (1992: ch.6). 28 Fretheim. The co-occurrence of the two Norwegian adverbs først and så in the two consecutive main clauses of (2) reveals that it is not just their order that reflects the order of events in the scenario described, as the chain of events might also be said to be encoded by means of the pair of temporal adverbs først and så, but as I argued in a recent paper (Fretheim 2006), Norwegian så does not represent a concept. Unlike Danish så, it is not an anaphor whose truth-conditional interpretation depends on its being linked, in the hearer's pragmatic processing of the discourse, to an antecedent with a conceptual content. Rather, its function is to offer the hearer the procedural information (Blakemore 1987, 2002; Wilson & Sperber 1993) that the inferential processing of the verbal stimulus should take into account the fact that the woman's looking seriously at the 1st person narrator and the same woman's sudden giggling were two events that happened sequentially, not simultaneously or in the opposite order of what is signalled iconically by the syntactic order of the clauses. I believe it is possible to justify a similar view of the lexical meaning of the word først in (2).15 There is no accessible antecedent proposition P which would enable the hearer to saturate the adverb først in (2) as 'before P'. 'Before she started to giggle with her hand over her mouth' would be a truthful enrichment of først but it is not a candidate referent here. That information is not accessible - not 'manifest' to the hearer as a relevance theorist would say (Sperber and Wilson 1986) - at the point in (2) where først appears, as it is asserted only in the next sentence. An anaphor presupposes the existence of the entity referred to, and the anaphor først presupposes the existence of an event presented as the later of two temporally related events. Moreover, it is plainly true that the Norwegian author could have chosen a different syntactic form to convey the same message. (2') consists of a juxtaposition of one declarative followed by a conjunction of declaratives, and it will be processed and comprehended in the same way as the original in (2). (2') Jeg takket henne etter vår siste kraftprøve. Hun så alvorlig på meg og begynte plutselig å fnise med hånden over munnen. 'I thanked her after our last test of endurance. She looked seriously at me and began to giggle with her hand over her mouth.' What is described in the coordination of clauses is the woman's first reaction to the narrator's having thanked her, which was soon succeeded by a phase in which she was trying to suppress a strong inclination to laugh at him. There are numerous ways of expressing this in Norwegian, and quite a few of them would involve no use of adverbs of temporal succession or any other linguistic device which encodes semantically that the two things happened in a particular order. I conclude that først is an optional element in (2), and a word that does not contribute to the proposition expressed. Now compare the illustration in (2) to a use of først which is anaphoric and consequently truth-conditional. (3) is also from the OMC. The concessive construction is an interpretation of Brita's thought; she is the perceiver. (3) Plutselig var Hildeguns blitt større enn hennes, enda Brita hadde fått bryster først. (BV1) 15 Notice that, unlike first, at first which was the English translation opted for in (2) is never an anaphor. Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first') 29 'Suddenly Hildegun's (i.e. Hildegun's breasts) had become bigger than hers (i.e. bigger than Brita's breasts), although Brita had developed breasts first (i.e. before Hildegun).'  Translation: Brita noticed her breasts, which had suddenly become bigger than hers, although Brita had developed breasts first. You cannot leave out først in (3) and still communicate what (3) means. Først represents, by way of an antecedent-anaphor relation, the only new information in the asserted concessive clause of (3). Thus (3') would be an irrelevant utterance in any imaginable context. (3') ??Plutselig var Hildeguns (bryster) blitt større enn Britas (bryster), enda Brita hadde fått bryster. 'Suddenly Hildegun's (breasts) had become bigger than Brita's (breasts), although Brita had developed breasts.' The Norwegian adverb først and the English adverb first are essential elements of the encoded semantic meaning of (3), without which the proposition communicated by an utterance of (3) would be inaccessible to the addressee. Any competent reader of (3) will read først/first as 'before Hildegun developed her breasts', using information encoded in the main clause appearing ahead of the concessive clause to 'saturate' the conceptually quite impoverished temporal adverb. This is truly different from the way in which først and at first will be processed by someone reading (2). What the occurrences of først in (2) and (3) have in common, however, is their lexically defined procedural meaning (see section 3 for details): a token of først will strongly encourage the addressee to impose a temporal order upon two or more events described, so that the proposition modified by først represents the earliest event in the sequence. Sometimes, as in (2), the addressee will not be able to identify the proposition representing the temporally later event at the point in the discourse where the word først appears, because the subsequent event is only described in a later utterance, very often in the immediately following one, and an overt sign like så ('then') or siden ('later') in the next clause is a strong signal that the addressee should not try to locate an antecedent proposition anywhere in the preceding discourse, because there is none. At other times, most typically when først is not the initial but the final item in its clause, the adverb must be conceptually saturated in context. This is what we witness in (3), where we learn that Brita developed her breasts prior to the period when Hildegun developed hers, but Hildegun's breasts developed at a speed which resulted in the situation described prior to the concessive clause in (3). The reader of (3) will use information provided in the first clause to inferentially compute the reference of the anaphoric item først at the end of the next clause. Occasionally it would seem that a given occurrence of først may represent a borderline case, neither convincingly anaphoric nor undisputably non-anaphoric. Compare the talk exchange between A and B in (4) and (5). 30 Fretheim. (4) A: Hva har du tenkt å gjøre nå? 'What are you going to do now?' B: Først skal jeg lese dette. 'First I'm going to read this.' (5) A: Hva har du tenkt å gjøre nå? 'What are you going to do now?' B: Jeg skal lese dette først. 'I'm going to read this first.' Processing B's utterance in (4), the interlocutor A will understand that B is going to give priority to a reading of the text referred to, so maybe one could argue that først in (4) B will be enriched as 'before I do anything else'? A's previous question does not contain conceptual material that is repeated, by way of a resumptive anaphor, in B's answer, so one might be a bit reluctant to analyse først in (4) in the same way that we analysed først in (3). On the other hand, the encoded semantics of B's declarative in (4) is definitely in need of a pragmatic process of enrichment that concerns the time at which the speaker is going to read what the demonstrative dette ('this') refers to. It is presumably going to happen right away, but saying that looks like another way of conveying the same as the suggested paraphrase 'before I do anything else'. Still, (5) B illustrates something qualitatively different in regard to the pragmatic processing needed to recover the explicature16 of the utterance. The speaker's placing først in the sentence-final position is an important cue that the hearer will exploit in his pragmatic processing. Reading (5) we are inclined to imagine a context of interpretation in which først is an anaphoric echo of something said earlier in the same discourse. A may have told B to perform a specific task, and when B makes no sign to comply with A's request, A asks him 'What exactly are you going to do now?', thus implicating that B should do what they had agreed on as quickly as possible. If we assume that B had promised to try to repair a broken doorknob and sits down reading an article in a newspaper instead, then B can utter (5) B with the intention that the adverb først be saturated as 'before I start repairing that doorknob'. It is certainly possible to imagine a similar context for the pragmatic interpretation of (4) but there are less procedural cues there that would support a pragmatic treatment of først as anaphor. The following example found in the OMC is another illustration of the sentence-final use of først as an anaphoric adverb. (6) "Jeg vet at du har sett meg, men jeg så deg først." (THA1)  "I know you've seen me, but I saw you first." 16 The 'ground-floor explicature', which is the proposition arrived at through inferential development of the encoded logical form of the sentence used (Sperber and Wilson 1986). Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first') 31 The word først licenses use of the adversative connective men ('but') in (6). Without it there would be no communicated contrast between the two conjuncts, and the use of men ('but') would hardly be justified. If the writer were to drop først in (6) and write Jeg vet at du har sett meg, men jeg så deg, the reader would presumably fail to grasp the writer's communicative intention or question the suitability of the adversative coordinator. In fact, any occurrence of først inside a conjunct starting with men serves an anaphoric function. Men points to a contrast, not just a parallelism as in the assumption 'You saw me and I saw you' entailed by (6), and it is therefore natural for the hearer to use the preceding conjunct as antecedent in the pragmatic process of saturating først. I may have given you the impression that I believe an occurrence of først to be anaphoric and in need of enrichment just in case the overt presence of this word in a given sentence is mandated. This condition may be too strong, though. Consider the example from the OMC presented in (7), where the Norwegian source text and the English translation both contain a token of først and first, respectively, but where the German and French translations contain no word that corresponds to those temporal adverbs. (7) Fadese nr. 1: jeg hadde glemt å si fra at jeg kom i egen bil, så de fire herrene fra menighetsrådet hadde først gjort seg en forgjeves tur på jernbanestasjonen. (BHH1)  Faux pas No. 1: I had forgotten to give notice that I was arriving in my own car, so that the four gentlemen from the parish council had first made a fruitless trip to the railway station. Patzer Nummer eins: Ich hatte nicht mitgeteilt, dass ich im eigenen Wagen kommen würde, deshalb hatten die vier Herren vom Gemeinderat vergeblich am Bahnhof gewartet. Bévue no 1: Comme j'avais oublié de dire que j'arriverais dans ma propre voiture, les quatre messieurs du conseil paroissial sont vainement allés m'attendre à la gare. The antecedent of the anaphora først and first cannot be extracted directly from the semantics of the first sentence, whose meaning is 'I had forgotten to give notice that I was arriving in my own car'. Rather, these anaphoric words are 'indirect anaphors' (Erkü & Gundel 1987) in (7), words whose antecedent is the previously unexpressed thought that may be paraphrased as 'before the four gentlemen found me'. Something like that is arguably part of the explicature retrieved on the basis of cues encoded in the German and French texts as well. If we admit that the antecedent of først may sometimes have to be inferred from the context with which the addressee operates, then a possible objection against analysing først as an anaphor in our earlier example (4) may lose some of its initial attraction. Først seems to be omissible, with no loss of relevance, in (8), too. (9), without først, will automatically be interpreted in the same way as (8). 32 Fretheim. (8) Studiet av slike parkeringsforbud er så tidkrevende at man først må ha funnet en parkeringsplass om man vil lese og forstå dem. (HME3) 'The study of such no-parking signs is so time-consuming that one will have to have found a parking space first, if one wishes to read and understand them.' (9) Studiet av slike parkeringsforbud er så tidkrevende at man må ha funnet en parkeringsplass om man vil lese og forstå dem. The perfective expression må ha funnet ('must have found') indicates the relative order of events in (8) as well as (9), but the former does so redundantly with the word først. Can we conclude that there is an expressed antecedent-anaphor relationship in (8)? (8'), without the conditional clause, makes perfect sense, if the communicator's intention is to convey that the study of such no-parking signs is so time-consuming that one has to park the car before one can read and understand them. (8') Studiet av slike parkeringsforbud er så tidkrevende at man først må ha funnet en parkeringsplass. 'The study of such no-parking signs is so time-consuming that one must have found a parking space first.' This sentence provides information ('the study of such no-parking signs') which will arguably be used to enrich først, but the same information was offered at the beginning of (8). If we say there is an antecedent-anaphor relation between the nominalization studiet av slike parkeringsforbud ('the study of such no-parking signs') and først (= 'before one can study such no-parking signs') in (8'), then we should analyse (8) in the same way. There is an indisputable resemblance in content between what I have labelled the non-anaphoric use of først and its anaphoric use, yet there is also an undeniable difference between the two uses, as the non-anaphor, illustrated by an example such as (2), implies that the word does not constrain the set of truth conditions defining the proposition expressed, while the anaphor først, illustrated in examples such as (3) or (6), does contribute to truth-conditional content just like those anaphors (e.g. personal pronouns, demonstratives and miscellaneous indexicals) whose function has been thoroughly documented. It does so by virtue of the conceptual meaning of the linguistic antecedent that the addressee has to match with the anaphoric adverb in the inferential phase of the utterance comprehension process. On the other hand, I have not glossed over the fact that there may be borderline cases, like (4) and (8), where it is not obvious that an antecedent candidate is available at the point where først appears. Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first') 33 3. The procedural lexical meaning of the temporal adverb først Først functions like a pointer to those context-dependent inferences the hearer should draw in order to achieve cognitive effects as intended by the communicator. This approach to its lexical meaning rests on the differentiation made within Relevance Theory between conceptual semantics and procedural semantics originating with Blakemore (1987). (10) The procedural lexical meaning of the temporal adverb først: Place the respective states of affairs represented by two discourse propositions P and Q on a temporal scale, so that the state of affairs which is represented by P and is described in the linguistic structure modified by først precedes the state of affairs which is represented by Q. I am not saying that (10) exhausts the semantics of først. The word may have a conceptual meaning, too, possibly one which is neutral between a local and a temporal interpretation and which it shares with a connective like English before (either 'in front of' or 'prior to') and Norwegian før. There is also an atemporal use of først/first in listings (enumerations), and there is an adjectival use which is frequently atemporal (e.g. 'the first letter of the alphabet'). The identity of the proposition Q referred to in (10) always has to be determined through inference, but procedural indicators such as så ('then') or siden ('later') may make the inferential task easier for the addressee and the linguistic stimulus more relevant. Først triggers the assumption that the sentence so modified describes the first of two or more successive events, in accordance with the lexical definition offered in (10), which I claim to be valid for the anaphoric as well as the non-anaphoric use of først. Så, as in P og så Q og så R ('P, and then Q, and then R') has a 'defective' distributional pattern. It can occur in Paul Diderichsen's (1946) 'fundament' ('front field') and in his 'nexus field' ('middle field'), but not in the so-called 'content field' ('end field'), which is reserved for adverbs with a truth-evaluable import. Først, on the other hand, can occupy any one of the three positions in which a Norwegian adverb can be found, but it is noteworthy that it can occur in the content field just in case it serves the function of a truth-conditional anaphor in the post-semantic (pragmatic) phase of the comprehension process. Thus, placement of først in the content field yields procedural information that encourages the addressee to look for an accessible antecedent proposition that will enable him to enrich the word as intended. A sentence-initial først, however, is indeterminate with respect to the anaphor/non-anaphor distinction, and one must resort to contextual evidence to resolve the issue. To conclude, the difference between the anaphor først and the non-anaphor først is argued to be a wholly pragmatic matter, and not an ambiguity that resides in the linguistic code. Linguistically, først has the univocal lexical meaning stated in procedural terms in (10), but the structure of the discourse, the presence of one or more procedural cues like the syntactic position of først, the availability (or not) of a likely candidate antecedent, and, as I am going to demonstrate in section 4, the intonation 34 Fretheim. pattern imposed on the syntactic form, all these things taken together give addressees important contextualizing information that crucially affects the way they will behave in pursuit of the communicator's meaning. 4. Intonation, først, and information structure When there is contextual evidence that a given token of først is to be treated pragmatically as an anaphor to be conceptually saturated through association with an antecedent, the intonational form mapped onto the syntactic form will make the intended context for interpretation more manifest and will consequently place further constraints on the proposition expressed in a linguistic structure that contains this adverb. Consider the alternative dialogues in (11) and (12) between two friends A and B who are both under age, A being slightly older than B. (11) A: Er det greit for foreldrene dine at du er med meg til København? 'Is it ok for your parents that you come with me to Copenhagen?' B: Jeg må snakke med dem først. 'I must speak to them first.' (i.e. before I can answer your question) (12) A: Er det greit at jeg spør foreldrene dine om du kan være med meg til København? 'Is it ok if I ask your parents if you can come with me to Copenhagen?' B: Jeg må snakke med dem først. 'I must speak to them first.' (i.e. before you see them about it) The highlighted (bold-faced) parts of B's respective responses in (11) and (12) indicate the scope of først. Those phrases, an infinitival complement in (11) and a subject pronoun in (12), contain information that contrasts with the information given in the antecedent that allows the interlocutor to enrich the meaning of the sentence-final occurrence of først and obtain a truth-evaluable proposition. In spite of the fact that B's sentence is the same in both dialogues, the proposition expressed (the explicature) is not the same, because først must be treated pragmatically as an anaphor in these structures and the associated antecedent cannot be the same in (11) and (12). Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first') 35 The highlighted subject pronoun jeg ('I') in B's declarative in (13) below is not the information focus there, although jeg is the focus of the corresponding answer in (12). The 1st person pronoun referring to speaker B is presented as a contrastive topic in (13). B makes the point that the rules that apply in B's home are not the same as the rules that A's utterance has caused B to believe to apply in A's home. The scope of først in (13) B is the adjacent infinitival complement snakke med dem ('speak to them'), which contrasts with the attitude expressed by the interlocutor A. (13) A: Vi trenger ikke å be foreldrene våre om lov til å dra alene til København. 'We don't need our parents' permission to go alone (without them) to Copenhagen.' B: Jeg må snakke med dem først. 'I must speak to them first.' (i.e. before I can decide to go) 220 The focus of information in the utterance of (13) B is the speaker's contradiction of one particular entailment of the proposition expressed by A, namely that B does not have to 170 ask for his parents' permission to go to Copenhagen unaccompanied by an adult person. F0 (Hz) Those who, like me, believe that linguistic semantics underdetermines truth- conditions also believe that a change in background assumptions (= context) can change truth-conditions. The primary function of intonation is to indicate important 80 0 1.4152 aspects of the context in which an utterance is to be processed and interpreted. Figures 1, 2 and 3 are three likely intonation contours that an East Norwegian speaker of the 220 respective utterances of (11) B, (12) B, and (13) B could have chosen. They express different information structures, which means that they contextualize differently and so F0 (Hz) 170 give the hearer different sorts of input to the task of determining what truth-conditions apply. 80 0 1.4152 220 170 F0 (Hz) 80 0 1.5643 Figure 1 L H¯ H* L H¯ L% [[jeg [1MÅ]AU ]FP [[2SNAKKE-med-dem-først]AU]FP]IU 220 170 F0 (Hz) 36 Fretheim. 80 0 1.4152 220 170 F0 (Hz) 80 0 1.4152 Figure 2 L H¯ H* L H L L% [[[1JEG-må]AU]FP [2snakke-med-dem]AU [1først]AU ]IU 220 170 F0 (Hz) 80 0 1.4152 Figure 3 L H¯ H* L H L% [[[1JEG-må]AU ]FP [2snakke-med-dem-først]AU ]IU The fundamental frequency (F0) contour displayed in Figure 1 is a bi-focal intonation, an intonation pattern that includes two occurrences of a F0 peak in a single Intonation Unit (IU). This F0 peak highlights the lexically accented grammatical constituent heading the Accent Unit (AU) in which the peak appears as a right-edge acoustic phenomenon. I refer to the peak as a focal tone (H ), which terminates the Focal Phrase (FP), a constituent intermediate between the IU and the AU in the intonational hierarchy.17,18 Normally the information-structural implication of a bi-focal intonation like the one in Figure 1 is that one of the FPs delimited by a right-edge H will contain contextually given (topical) information, while the other FP will contain contextually irretrievable information. The contour in Figure 1 suggests that speaker B presupposes that A has activated the assumption that the consent of B's parents will have to be solicited, so the assumption that A's parents must be informed before A and B can 17 The point of FP termination is indicated by solid vertical bars in the three F0 contours. A dotted vertical line, appearing once in Figure 1 and once in Figure 2, indicates an AU boundary which is not also an FP boundary. The early one in Figure 1 marks the point of transition from the unstressed anacrusis jeg to the accented syllable må constituting the first AU. 18 I take the High tone (H*) of East Norwegian Accent 2 to be the only specified lexical tone. All tones that lack an asterisk in my notations are post-lexical tones, including the default tone L of East Norwegian Accent 1 forms. These phonological assumptions have no bearing on the issues discussed in this paper. Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first') 37 proceed with their travel arrangements is heard to be confirmed through the focal tone aligned with the modal auxiliary må ('must') in (11) B/Figure 1. An action in accordance with the complement of må is claimed to be necessary, and the speaker's insistence is highlighted by the focal peak displayed at the end of the syllable må in Figure 1. Thus the focally accented modal verb is the information focus and the assumption that B's parents must get acquainted with A's plan is presented as information that both A and B can be expected to activate and bring to bear on the issue raised by A, even if it is not mentioned explicitly in B's answer. Use of an unstressed form må followed by a focally accented infinitival phrase would have changed the information structure. That intonation would be a way of presenting the phrase snakke med dem ('speak to them') as irretrievable information rather than presupposed information. There appears to be just a very minor difference between the two contours of Figure 2 and Figure 3, but there is a phonological difference there which the ears of native speakers are attuned to. The relatively lower pitch on the final word først in Figure 2 tells the hearer that there is a word-accent on that utterance-final item, while the relatively higher pitch on the final syllable først in Figure 3 is a post-lexical H which terminates any East Norwegian AU but which is phonologically distinct from the FP-final H . A word-accent on først, as in Figure 2, is a cue that helps the hearer to identify the scope of først as a non-adjacent syntactic phrase. When items that appear together in the syntactic string are not meant to be parsed as a combination of elements that should be perceived as a unit for the purpose of efficient pragmatic processing, a word-accent like the one on først in Figure 2 may be required to avoid incorrect parsing on the part of the hearer. An unaccented først, as in Figure 1 and Figure 3, indicates that the scope of først is the contiguous syntactic constituent.19 The contiguity of the infinitival complement and først then mirrors the fact that the contrast is between 'involving and consulting one's parents before making a decision' and 'not involving one's parents before making a decision'. I asked ten East Norwegian informants to assess the acceptability of the intonations visualized in the contours of Figures 1, 2 and 3 spoken in response to A's stimuli in (11), (12) and (13). The combinations of (11) A - Figure 1, (12) A - Figure 2 and (13) A - Figure 3 were the only ones that were accepted by every listener. The patterns of Figures 2 and 3 were generally rejected as impossible in an answer to (11) A, as was the use of the Figure 1 pattern employed in the context set up by the utterance of (12) A. Three out of ten informants accepted the Figure 1 pattern in the context of (13) A. The most exciting contrast for me was the one between the apparently very similar contours of Figure 2 and Figure 3. How many would say that they could be used interchangeably in response to (12) A and (13) A? It turned out that no one accepted the combination of (13) A and (12) B/Figure 2, while four accepted the combination of (12) A and (13) B/Figure 3. This result indicates that, in their processing for relevance, native speakers are sensitive to the phonological difference between an accented først and three AUs (Figure 2) and an unaccented først and just two AUs (Figure 3). Even a slight phonetic difference like the one between Figure 2 and Figure 3 has an impact on the way people try to match først with potential antecedents in A's utterances, and therefore on their comprehension of the truth-conditional content of B's replies. 19 The story to be told is slightly different when the temporal adjunct først is a separate FP, but that story cannot be told here. 38 Fretheim. 5. Conclusion There are two distinguishable uses of the originally superlative temporal adverb first and its Norwegian cognate and semantic counterpart først, but the lexical meaning proposed in (10) does not discriminate between the two uses. However, two distinct pragmatic strategies will be available to the hearer, and I have resorted to the labels 'anaphoric' and 'non-anaphoric' in my reference to the different pragmatic handlings of the adverb. The lexical entry for først (and for its English counterpart first) gives us information (cf. (10)) that triggers the addressee's search for the proposition Q, which may or may not be truth-conditionally equivalent to a temporal clause of the 'before'- type. When først is pragmatically equivalent to a temporal clause, it is a discourse anaphor whose referent must be determined on the basis of context-dependent information, often, but not exclusively, in the form of a conceptual representation mentally transferred from a discourse antecedent. When no candidate antecedent is accessible, the hearer will process a given token of først as a non-truth-conditional item, and identification of Q will be suspended until the hearer has had a chance to process the next sentence, as in discourse structures conforming to the formula 'First P and then Q'. Though the token frequency of the non-anaphoric use of first and først may exceed that of the anaphoric use in certain written and spoken genres (and especially in narratives), I have been more concerned with the latter use in my paper, because I feel it deserves it more. To the best of my knowledge, no one has paid attention to the anaphoric properties of English first before, or the lexical semantics and the pragmatic implications of corresponding adverbs in other languages. I take the procedural semantic definition offered in (10) to be the necessary point of departure for inferential processing of any occurrence of the temporal adverb first or Norwegian først, regardless of syntactic position and the speaker's choice of intonation. Its position in the syntactic string and the speaker's intonational phrasing are procedural cues that constrain the hearer's selection of a context, thereby affecting his determination whether or not to handle the word as an anaphor to be saturated in context. In the absence of a highly accessible antecedent, the hearer will go for a non-anaphoric interpretation of the adverb and expect information about the later of the two events (represented by Q) to be revealed in the continuation of the discourse. The syntactic pattern 'First P and then Q' reflects the order of extra-linguistic events in an iconic manner and is therefore suitable when a non-anaphoric interpretation of the temporal adverb is intended. Acknowledgement I am grateful to Francis Cornish for his valuable comments on the content and style of the next-to-final version of this paper. Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first') 39 References Blakemore, Diane (1987). Semantic Constraints on Relevance. Oxford: Blackwell. Blakemore, Diane (1992). Understanding Utterances: An Introduction to Pragmatics. Oxford: Blackwell. Blakemore, Diane (2002). Relevance and Linguistic Meaning: The Semantics and Pragmatics of Discourse Markers. Cambridge: Cambridge University Press. Diderichsen, Paul (1946). Elementær Dansk Grammatik. Copenhagen: Gyldendal. Erkü, Feride and Jeanette Gundel (1987). The pragmatics of indirect anaphors. In: Verschueren J. & Bertuccelli-Papi, M. (eds.), The Pragmatic Perspective: Selected Papers from the 1985 International Pragmatics Conference, 533-46. Amsterdam: John Benjamins. Fretheim, Thorstein (2006). English then and Norwegian da/så compared: A Relevance-theoretic account. Nordic Journal of Linguistics 29: 45-94. Recanati, François (2004). Literal Meaning. Cambridge: Cambridge University Press. Sperber, Dan and Deirdre Wilson (1986). Relevance: Communication and Cognition. Oxford: Blackwell. Wilson, Deirdre and Dan Sperber (1993). Linguistic form and relevance. Lingua 90: 1-2 Author’s e-mail address: thorstein.fretheim@hf.ntnu.no 40 Fretheim. WORKINg PAPERS ISK 3/2006 41-57 ISK, NTNU The metarepresentational use of main clause phenomena in embedded clauses Thorstein Fretheim 1. Introduction 1.1 Metarepresentation Consider the direct report in (1) and the indirect report in (2). (1) “Mick is just as scared as me,” she thought. (2) She realized that Mick was as scared as her. An utterance of (1) metarepresents (cf. Sperber 2000; Noh 2000) a thought attributed to the woman referred to. What looks like a quotation, a case of direct speech, is just an imitation of something that could have been uttered – an interpretation, or metarepresentation, of something the woman could have truthfully communicated to someone at the time when the thought reported in (1) entered her mind, but which was never articulated. The theoretical framework on which the present paper rests is relevance theory, and metarepresentation in relevance theory is the use of one representation to represent another in virtue of some resemblance between them, whether in content or form or both (Sperber and Wilson 1986; Noh 2000). A first-order metarepresentation is an interpretation of a thought that the speaker has at the time of the utterance. As this description applies to all use of language, it may be redundant to refer to first-order metarepresentation as ‘metarepresentation’. Second-order metarepresentation is when the interpreted thought is one that the speaker attributes to some other person, or one that the speaker used to entertain at an earlier time, and in most cases the term ‘metarepresentation’ is reserved for second-order metarepresentation. Third-order metarepresentation implies that the speaker interprets how some other person has interpreted the thought of a third person. Not only syntactic phenomena and lexical choices but also pitch-based vocal gestures can be used to mimic structural features that could have appeared in a linguistic expression, imagined or real, of the thought metarepresented by the reporting speaker. Intonation employed for the purpose of metarepresenting a specific thought may echo certain intonational features associated primarily with use of a different type of syntactic construction, and a report may use prosodic features that are an integral part of the information structure of the imitated utterance type. Fretheim (1998) describes the use of a particular intonation pattern in Norwegian negative interrogatives, which mimics, i.e. metarepresents, an intonation pattern typical of affirmative declaratives. This is a way of signalling the speaker’s doubt about the truth of the negative proposition. The intonation is reminiscent of an intonation that could not have been used in a negative declarative, only in its positive counterpart. 42 Fretheim What I have already described as an ‘indirect report’ presented in (2) has a rather more open speech act potential than the direct quotation in (1). It could be used to perform an indirect report of the woman’s thought, conveying much the same as in (1), but it could also report something the woman had actually uttered (a true case of ‘indirect speech’), or it could be used to affirm the speaker’s belief that the complement clause expresses a true proposition and that the woman had acknowledged that fact. In the latter situation the speaker of (2) confirms the main clause proposition, or confirms just the complement proposition and volunteers the main clause information about the woman’s attitude to the embedded proposition. The different uses of a sentence such as the one shown in (2) will typically occur with different, context-dependent intonation patterns when spoken. A speech act that confirms the truth of the complement proposition or the main clause proposition represented in (2) will normally be produced with a pitch accent on the verb form realized. The main clause verb conveys new information and the following complement contains topical information when the complement proposition is confirmed, while the affirmative polarity of the main clause proposition is the only new information when the speech act confirms that the woman realized what is described in the complement. However, if (2) is produced with the intention to convey the same as (1), there will typically be a pitch accent on at least one item in the complement clause, for instance one on Mick and maybe also one on her, or one on the first token of as and one on her. These are prosodic features that mimic certain accentual aspects of an imagined utterance performed by the female referent. When two strong accents are located inside the complement, that part of the complex sentence has its own topic-comment structure with the subject Mick as topic. It is a prosodic structure indicative of free indirect speech, or what Banfield (1993) refers to as ‘represented speech or represented thought’, though the echoic elements are intonational features, and not the standard morphosyntactic features typical of free indirect speech, as illustrated for instance by (3), from Banfield (1993: 342). (3) Oh, she was so pleased to see him – delighted! The day was so charming, didn’t he agree? (Katherine Mansfield, “Miss Brill”, in Stories, 1956) In the free indirect speech of (3) there is a shift from the direct speech use of the 1st person pronoun I to 3rd person she, a shift from the direct speech use of 2nd person you to 3rd person him in the first declarative and he in the tag question, and a shift from present tense is and don’t in direct speech to past tense was and didn’t. We may compare the indirect speech of (2) to the slightly different alternative (2’), where Michael Burns is the full name of the man referred to as Mick in (2), and ask how this difference affects our understanding of how much of the utterance serves as a metarepresentation of the female referent’s thought. (2’) She realized that her good friend Michael Burns was as scared as her. The description her good friend Michael Burns is apparently the narrator’s, not the woman’s. She would not think of Mick as ‘my good friend Michael Burns’, though she might obviously use that descriptive phrase to refer to him in a conversation with someone unfamiliar with Mick. The only part of a spoken utterance of (2’) which might include echoic elements indicative of free indirect thought is the predicate phrase. We can The metarepresentational use of main clause phenomena in embedded clauses 43 imagine that not only the final word her but also the first token of as or the adjective scared have a pitch accent: ‘… that her good friend Michael Burns was AS scared as HER.’ or ‘ … that her good friend Michael Burns was as SCARED as HER.’ Still, the meaning derived from the ‘that’-complement of (2’) is a metarepresentation of an assumption which the speaker attributes to the female referent, regardless of how the speaker of (2’) chose to describe Mick. 1.2 Main clause phenomena Main clause phenomena have been observed to occur in complement clauses (in the complements of especially verbs of saying and thinking), and in certain types of adverbial clause, like causal clauses and concessive clauses as well as adversative ‘whereas’- clauses. The use of main clause word order in an embedded clause (cf. the seminal paper by Hooper and Thompson 1973) is a good example of a main clause phenomenon used in the metarepresentation of the content of something that originally was, or could have been, communicated by means of a main clause. Some languages, for example German and the Scandinavian languages, display a systematic difference between main clause and subordinate clause word order, and the former may be used in an embedded clause A’ when the speaker’s intention is to mimic certain formal aspects of some previously produced main clause A whose communicated assumption the utterance of A’ purports to metarepresent. In addition to syntagmatic contrasts in word order, there are various kinds of non- truth-conditional items whose function relates to the illocutionary force of the utterance or the speaker’s attitude to the proposition expressed, and which are therefore restricted to main clauses, or rather to main clauses plus embedded clauses that are perceived to share certain formal characteristics of main clauses because they are interpretations of thoughts attributed to someone, either to the speaker herself or to some contextually salient 3rd person. The topic of this paper is the interaction of a specific intonational main clause phenomenon and a set of syntactic main clause phenomena in spoken Norwegian, all of which are prominent linguistic devices used for the expression of metarepresented thoughts. Section 2 introduces the intonational phenomenon that interests me; its function in embedded clauses and its interaction with word order is illustrated in section 3; section 4 deals with its interaction with the peculiarly Scandinavian (in fact more typically Norwegian than Danish or Swedish) use of a right-detached pronominal copy of the subject nominal (the in situ subject phrase may be either a full lexical phrase or a pronoun); section 5 introduces yet another main clause phenomenon: modal particles, and relates the distribution and use of one such modal particle in Norwegian to the three main clause phenomena right dislocation, word order and the special intonational phrasing used in embedded clauses that function as second-order metarepresentations. 2. The Norwegian double-peak intonation What is an intonational main clause phenomenon in Norwegian speech? I tried to give an answer to that question in a paper I wrote many years ago (Fretheim 1981). My most important finding at the time was that whenever some segment of a spoken utterance in Norwegian contains what I called two peaks of prominence, rather than just one, that 44 Fretheim segment is a syntactic main clause. I paid special attention to complex and compound sentence structures that contained a first clause and a second clause, where the second clause could in principle be parsed syntactically either as a main clause coordinated with the first clause or as a subordinate clause embedded as a constituent in a complex sentence.20 A corresponding English contrast of this nature is seen in the syntactic ambiguity of I’ll make coffee, so we won’t fall asleep (cf. the coordinating so in I’ll make coffee, so we won’t fall asleep, I can assure you vs. the subordinating so that in I’ll make coffee, so that we won’t fall asleep.) My study included one listening comprehension test. Thirteen East Norwegian subjects were asked to listen to six prosodically distinct versions of the three clauses in (4), the final one with a null subject, which were presented to them in exactly the written form shown here, with no punctuation marks at the end of line 1 or line 2. (4) Herman hørte en stønnende lyd som om noen hadde skadet seg men forsøkte å late som ingenting. ‘Herman heard a moaning sound as if someone had hurt himself but tried to pretend it was nothing’ (literally: but tried to pretend like nothing) The final line means either that ‘the moaning agent tried to subdue the pain and emit as little sound as possible’, or that ‘Herman pretended he had heard nothing, i.e. that he had not heard any disturbing sounds’. My listeners were asked to decide whether the null subject of the final clause referred to Herman or to the unseen and (therefore) unknown person who ‘tried to pretend it was nothing’. (5) below is a notation of two of the six intonationally distinct stimuli used in the comprehension test. Parts (a) and (b) were identical in those two stimuli but the conjunct after the connective men (‘but’) was either pronounced as indicated in (ci) or as indicated in (cii). IU = Intonation Unit (which includes either one or two FPs); FP = Focal Phrase (which is bounded to the right by a focal tone, a fundamental frequency peak that lends prominence to the syntactic constituent whose final item is the accented word heading the FP-final AU); AU = Accent Unit (which is headed by a pitch-accented word form – Norwegian Accent 1 or Accent 2 – optionally followed by one or more unaccented word forms). Small caps are used for the accented part of syntactic constituents highlighted by a focal tone. (5) (a) [ [ [1Herman]AU [2hørte-en]AU [2stønnende]AU [1LYD]AU ]FP ]IU (b) [ [som om [2noen-hadde]AU [2SKADET-seg]AU ]FP ]IU (ci) [ [men for[1søkte-å] [2late-som]AU [2INGENTING]AU ]FP ]IU (cii) [ [men for[1søkte-å] [2LATE-som]AU ]FP [ [2INGENTING]AU ]FP ]IU When (ci) followed the sequence (a)-(b), the final IU was invariably processed as the second of two conjuncts at the embedded clause level, which means that the null subject was understood to be bound by the existential quantifier noen (‘someone’) which is the subject in (b), but when the final IU was produced with the double-peak pattern of (cii), a majority of my informants conjoined the ‘but’-conjunct in (c) at the upper-clause level and judged Herman to be the referent of the null subject. 20 No attention was paid to the use of main clause intonation in embedded clauses in the paper from 1981. The metarepresentational use of main clause phenomena in embedded clauses 45 My hypothesis was that the presence of two focal tones in a clause was a prosodic main clause phenomenon in spoken Norwegian. I predicted that the listeners would pay attention to the double-peak pattern in the ‘but’-conjunct of (cii), and they did. The intonation employed in (cii) cues a processing of ‘X tried to pretend it was nothing’ as an assertoric illocutionary act, which implies that the referential value of ‘X’ cannot be the unidentified individual referred to in the non-indicative ‘as if’-clause. Due to the prosodic highlighting of both the verb late som (‘pretend’) and its complement ingenting (‘nothing’), (cii) tends to be perceived as the linguistic vehicle of what relevance theorists call an explicature (Sperber and Wilson 1986). It is presumably more costly, ceteris paribus, to parse the string (c) in such a way that it is not linked to the immediately following clause as the second conjunct at the subordinate clause level but rather to the complex structure consisting of a main clause and an embedded comparison clause. However, the double-peak pattern in (cii) alerts the hearer to the fact that the initially less obvious attachment at the main clause level is the correct resolution of the syntactic ambiguity in (4). No pauses were ever made between the IUs in any of the six stimuli that the test subjects were given, so there was no physical temporal break there which could be perceived as a juncture sign; the IU boundaries, specifically the one between (b) and (c) in (5), were indicated exclusively by means of pitch movements. It is reasonable to believe that the focal tone on the verb in (cii) was a garden-pathing element that forced most listeners to adopt a different parsing strategy than they did when they listened to (ci), though this is hardly more than a conjecture, as my 1981 experiment included no measuring of reaction time. Today my inclination is to analyze the Intonation Unit of (5cii) as an echoic metarepresentation of the thought expressed by an utterance of (6), where the 1st person subject pronoun refers to the protagonist Herman in (4). (6) [[jeg [2LATER-som]AU ]FP [ [2INGENTING]AU ]FP ]IU ‘I’m going to pretend that it’s nothing.’ As Herman obviously never said anything in the situation described in (4), the double- peak pattern in (6) cannot literally be said to mimic an utterance attributed to Herman, but it is possible to use a main clause phenomenon like the Norwegian intonational phrasing in (cii) to metarepresent a thought that the speaker believes that Herman could have truthfully articulated on that spot where he suddenly heard the sounds that he interpreted as moaning. The important point, however, is not whether or not (6) would have been a likely utterance for Herman to produce if an interlocutor had been present, what is important is the fact that the intonational phrasing employed in (5cii) is the same as the double-peak pattern appearing in an independent declarative like (6), an intonational form which is impossible unless it occurs in a segment of speech that is being used either to perform a separate illocutionary act, or else to metarepresent the thought that could trigger such an act. Two focal tones imposed on a given syntactic string means that the IU contains two FPs; one will be associated with contextually given information and one with new information (Fretheim 1991, 1992, 2002; Nilsen 1992), both when the string is a syntactically independent sentence and when it is a sentence constituent that metarepresents a thought that the speaker attributes to someone. We are experiencing the narrative in (4) through Herman’s senses and Herman’s mind, no matter whether we let the connective men (‘but’) link two conjuncts at the main clause level or at the embedded clause level. The discourse entity Herman has the highest 46 Fretheim cognitive status (Gundel, Hedberg & Zacharski 1993) in the narrative fragment (4) and is the referent that we empathize with (cf. Kuno 2004). The ‘deictic origo’ in (4) is transposed from the speaker to the protagonist Herman (Levinson 2004). (5cii) is a second-order interpretation, or metarepresentation: the narrator metarepresents a thought attributed to someone else, to Herman. (5ci), however, may be labelled a third-order metarepresentation: the narrator metarepresents a thought of Herman’s, which is in turn a metarepresentation of what the moaning stranger was thinking and feeling at the moment. Since the speaker of (5ci) does not metarepresent a thought attributed to Herman but rather a thought that the speaker believes Herman to have attributed to the unknown person, we, as potential addressees, are not able to empathize with the unknown person the way that we empathize with the protagonist Herman. It would therefore not be optimal for the speaker of (4) to use an intonation pattern like the one in (5cii) in order to metarepresent a thought that Herman attributed to that moaning stranger. The referent of the ‘big PRO’ subjects of the verbs forsøkte (‘tried’) and late som (‘pretend’) in (5cii) should be Herman, while the corresponding referent in (5ci) is more likely to be the moaning stranger, possibly due to parsing strategies such as the ones proposed by Frazier (1979; Frazier and Clifton 1996). 3. Main clause intonation in complement clauses Certain main clauses with a clausal direct object do not contribute to the proposition expressed. These are most typically main clauses with a 1st person subject and the present tense of a verb that denotes an attitude to the proposition expressed in the complement clause, or a performative verb of the Austinian sort. The explicated proposition in (7) is not about the speaker’s acknowledging an obligation to tell the hearer that the weather has changed; rather, the speaker uses this grammatical construction to assert that the weather has changed. (7) Jeg må fortelle deg at været har forandret seg. I must tell you that the.weather has changed Refl ‘I must tell you that the weather has changed.’ Imagine that (7) is part of a telephone conversation, and that the addressee does not know what the weather is like in the area where the speaker is calling from. The ground-floor explicature seems to be confined to the complement – ‘the weather has changed here’ – and the information in the main clause reveals the speaker’s propositional attitude, conveying a so-called higher-level explicature (Blakemore 1992; Wilson and Sperber 1993). As the explicature of an utterance of (7) is expressed in the complement clause, the speaker could have omitted the main clause altogether, saying (8) instead. The complement of (7) may contain any intonation pattern that could be imposed on the shorter version in (8). (8) Været har forandret seg. ‘The weather has changed.’ The mere fact that the speaker actually called the hearer to tell him about the change in the weather is likely to give rise to the same contextual assumption as the longer version The metarepresentational use of main clause phenomena in embedded clauses 47 of (7) would: the speaker feels an immediate urge to impart this information, because no matter whether there has been a change to the worse or to the better, the change may be radical enough to call for a reconsideration of a certain planned event that would involve the hearer’s participation. (9) describes something that the speaker intends to do in the future. The intonation employed in producing an utterance of this sentence could be the neutral intonation in (9’) where a single FP exhausts the IU, or the intonation in (9”) where the ‘that’-complement contains a prosodically highlighted topic (‘the weather’) and a prosodically highlighted focus (‘has changed’). (9) Jeg har bestemt meg for å fortelle dem at været har forandret seg. ‘I’ve decided to tell them that the weather has changed.’ (9’) [[jeg har be[1stemt-meg-for-å-for]AU [1telle-dem-at]AU [1været-har-for]AU [[1ANDRET-seg]AU ]FP ]IU (9”) [[jeg har be[1stemt-meg-for-å-for]AU [1telle-dem-at]AU [1VÆRET-har-for]AU ]FP [[1ANDRET-seg]AU ]FP ]IU The double-peak pattern in the complement of (9”) imitates, and anticipates, an intonation that it would be natural for the speaker to choose when producing an utterance of (8) in the future, as announced in (9). In (10)-(10’), the finite verb is placed before the negation operator. This is a Norwegian main clause phenomenon that interacts with the main clause intonation at the embedded clause level. The explicature is restricted to the clause where the two focal tones are. This is indicated by the combination of main clause intonation, syntactic topicalization and subject-verb inversion in the complement of (10), while the complement is turned into a main clause in (10’) and the non-propositional part of the utterance is an utterance-final parenthetical. (10) [[du skjønner at [1DET]AU ]FP [[1kan-ikke]AU [1DU-avgjøre]AU ]FP ]IU you understand that THAT can not you decide ‘You see, YOU can’t determine THAT.’ i.e. ‘It’s not up to you to make that decision.’ (10’) [[[1DET]AU ]FP [[1kan-ikke]AU [1DU-avgjøre]AU ]FP skjønner du]IU that can not you decide understand you ‘YOU can’t determine THAT, you see.’ The clause Det kan ikke du avgjøre (literally: that can not you decide) metarepresents the thought that must have triggered the utterances of (10) and (10’) alike. The straight order SVO in the complement of (11) means that the adverb ikke (‘not’) can follow the finite verb kan (‘can’) as in a main clause, or precede the finite verb as in a subordinate clause, but the latter order, shown in (12), is not compatible with the main clause intonation, consequently the combination of intonational phrasing and syntactic form in (12) leads to grammatical illformedness. 48 Fretheim (11) [[du skjønner at [1DU]AU ]FP [[1kan-ikke]AU [2avgjøre]AU [1DET]AU ]FP ]IU you understand that you can not decide that ‘You see, YOU can’t determine THAT.’ (12) *[[du skjønner at [1DU]AU ]FP [[2ikke-kan]AU [2avgjøre]AU [1DET]AU ]FP ]IU you understand that you not can decide that The word order ikke kan (‘not can’) in the complement is only acceptable if we move the earlier of the two focal tones to the main clause, as shown in (13) where the combination of intonation and syntax reveals that the main clause contributes to the propositional content. An utterance of (13) will be interpreted as a biased question. It is a request for information because there is a 2nd person subject, and it is biased because the syntax is declarative: ‘You understand that you can’t make that decision, don’t you?’ The second FP in (13) contains a contextually given proposition, one whose truth is obvious to the speaker but apparently not to the hearer, and the first FP highlights the positive polarity of the proposition expressed (polarity focus). (13) [[du [1SKJØNNER-at]AU ]FP [[1du]AU [2ikke-kan]AU [2avgjøre]AU [1DET]AU ]FP ]IU you understand that you not can decide that ‘You understand that you can’t make that decision, don’t you?’ Following Sperber and Wilson (1986) we can say that the question in (13) is an interpretation of a thought that is desirable to the speaker at the time of utterance, and true if the addressee realizes that the complement expresses something true. While a parenthetical clause construction could be used as an alternative to (11) above, no such option is available as a structural alternative to (13), because no part of the propositional content of an utterance may be put into a parenthetical clause such as skjønner du (‘you see’) in (10’). Use of main clause word order in the complement is not permissible when the intonational phrasing reveals that the main clause contains conceptual material that contributes to the proposition expressed, so (14) is just as bad as (12). (14) *[[du [1SKJØNNER-at]AU ]FP [[1du]AU [1kan-ikke]AU [2avgjøre]AU [1DET]AU ]FP ]IU you understand that you can not decide that The main clause word order kan ikke (can not) in (14) offers the hearer the procedural information that the proposition expressed does not include any concepts encoded in the main clause, but the focal tone on the finite verb skjønner offers contrary procedural information, it indicates that the proposition expressed does include the concepts encoded in the main clause. Mens (‘while’) in (15) is not the temporal connective but the homonymous adversative/contrastive connective that may be glossed as ‘whereas’/‘while’. The double- peak intonation in the well-formed alternative (a) indicates that the ‘while’-clause expresses a separate proposition, not a constraint on the preceding main clause proposition. (15) [[[1GEIR]AU ]FP [[1sov-i]AU [1fem]AU [2TIMER]AU ]FP ]IU Geir slept in five hours The metarepresentational use of main clause phenomena in embedded clauses 49 a [[mens [1ODD]AU ]FP [[2ikke]AU [1sov-i-det]AU [2HELE-tatt]AU ]FP ]IU while Odd not slept in the whole taken b *[[mens [1ODD]AU ]FP [[1sov-ikke-i-det]AU [2HELE-tatt]AU ]FP ]IU while Odd slept not in the whole taken ‘Geir slept for five hours, while Odd didn’t sleep at all.’ Although the intonation pattern in the clause introduced by mens is a main clause intonation, the word order ikke sov (not slept) in alternative (a) is that of a subordinate clause, contrasting with the illicit main clause order sov ikke (slept not) in alternative (b). There is no conflict here, though. Syntactic main clause phenomena in Norwegian are subject to heavier constraints than the intonational main clause phenomenon described in this paper. This is why version (b) is an ungrammatical combination of syntactic and intonational form. Syntactic criteria reveal that a Norwegian adversative ‘while’-clause is a subordinate clause. However, the intonation used in (15a) is appropriate because the ‘while’-clause is an interpretation of a thought attributed to Odd and presumably an echo of something that Odd told the speaker of (15). Use of two focal tones is even possible within a restrictive relative clause. The intonation indicated in (16) offers the hearer the procedural information that the proposition of the relative clause metarepresents a thought attributed to the man and echoes his intonation, although aldri har (never has) is the subordinate clause order that could not occur in direct speech. (16) [[jeg har [1nettopp]AU [2snakket-med-en]AU [1mann-som]AU [2ALDRI-har]AU ]FP [[1sett]AU [1JERV-her]AU ]FP ]IU ‘I have just spoken to a man who has NEVER seen any WOLVERINES here.’ 4. The interaction of intonation, word order, and right- dislocated pronouns This section takes a look at the interaction of double-peak (‘bi-focal’) intonation, the difference between main clause and subordinate clause word order, and right-dislocated pronouns in Norwegian. In a sentence like (17) the joint procedural effect of the pronoun han (‘he’), which is a dislocated copy of the subject NP Ola in the complement, and the main clause order spiser ikke (eats not) where the verb precedes the negator, causes the addressee to process the main clause in the same way as the corresponding part of (18) where the dislocated pronoun is followed by a parenthetical ‘you know’. (17) Du vet at Olai spiser ikke kjøtt, hani. you know that Ola eats not meat he ‘You know Ola (he) doesn’t eat meat.’ (18) Olai spiser ikke kjøtt, hani vet du. Ola eats not meat he know you ‘Ola (he) doesn’t eat meat, you know.’ 50 Fretheim (19), with a bi-focal topic-focus structure inside the complement, is a very likely intonation structure for an utterance of (17), and so is the similar intonation imposed on the main clause in (20). (19) [[du vet at [2OLAi]AU ]FP [[1spiser-ikke]AU [1KJØTT]AU ]FP hani]IU (20) [[[2OLAi]AU ]FP [[1spiser-ikke]AU [1KJØTT]AU ]FP hani vet du]IU The speaker of (19) or (20) metarepresents a thought attributed to Ola and uses ‘you know’ to indicate her belief that the hearer is aware of Ola’s dietetic regimen. These utterances would probably be relevant as premises in a process of non-demonstrative inference. The expected conclusion might appear in the speaker’s next utterance, or it could be left to the addressee to derive as an implicature. The right-dislocated han (‘he’) in the above structures is a pronominal copy of the subject phrase Ola, so it belongs to the complement clause in (19). There are three main clause phenomena in the embedded clause of (19): (a) the right-dislocated pronoun, (b) the main clause order spiser ikke as opposed to the subordinate clause order ikke spiser, and (c) the speaker’s intonational phrasing, one focal tone on the topic phrase Ola and one on the focus phrase spiser ikke kjøtt (eats not meat). The word order spiser ikke forces an intonation pattern that is consistent with what the word order indicates, namely that the proposition expressed does not include the main clause of (19). Moreover, right dislocation is a main clause phenomenon that requires use of main clause word order in the clause bounded by the two coreferential NPs. Any information that is encoded in a string bounded by a left-edge referring expression and a right-edge coreferential copy of it is truth-conditional information, and in (19) no information encoded anywhere outside the area encompassed by the coreferential items Ola and han contributes to the proposition expressed. In (21) the first of the two focal tones is on the propositional attitude verb in the main clause. This intonation is a sign that the main clause does contribute to the propositional content, and it cannot be reconciled with either the right dislocation at the complement level or the word order in the complement. (21) is ungrammatical. (21) *[[du [1VET-at]AU ]FP [[2Ola]AU [1spiser-ikke]AU [1KJØTT]AU ]FP hani]IU It is possible to dislocate a copy of the main clause subject pronoun du, but that requires use of the embedded clause word order in the complement clause: (22) is grammatical, but (23) is not. (23) is bad because the main clause word order in the embedded clause indicates that the truth-conditional content does not include information conveyed in the main clause, and this procedural information is contradicted by the dislocation construction which here spans the entire complex sentence. (22) Du vet at Ola ikke spiser kjøtt, du. you know that Ola not eats meat you ‘You know that Ola doesn’t eat meat.’ (‘I know you know that!’) (23) *Du vet at Ola spiser ikke kjøtt, du. you know that Ola eats not meat you The metarepresentational use of main clause phenomena in embedded clauses 51 The speaker’s intonational phrasing in (24), with the same word order as in (22), contradicts the use of right dislocation at the upper clause level. (24) *[[dui vet at [2OLA]AU ]FP [[2ikke-spiser]AU [1KJØTT]AU ]FP dui]IU The intonation of (24) tells us that there is a complete topic-focus structure in the string covering the complement clause. I have previously argued (Fretheim 2001) that a grammatical constraint is needed, which rules out the presence of two focal tones in the string between a right-dislocated pronoun and the nominal phrase that binds it. (25) may be produced with the intonation structures of (26) or (26’) where there is a focus but no topic in the domain between the coreferential items Ola and han, but (26”) is impossible because both focal tones are in the intervening area, between Ola and han, first one on the finite verb and then one on the pronoun det which is here a focally accented demonstrative. (25) Olai vet det, hani. Ola knows it/that he ‘Ola knows.’ (26) [[[2OLAi]AU ]FP [[1VET-det]AU ]FP hani]IU OLA KNOWS it he ‘OLA knows.’ (26’) [[[2Olai]AU [1VET-det]AU ]FP [[1HANi]AU ]FP ]IU Ola KNOWS it HE ‘OLA knows.’ (26”) *[[[2Olai]AU [1VET]AU ]FP [[1DET]AU ]FP hani]IU ‘Ola knows that.’ The constraint (Fretheim 2001) that bars the presence of two focal tones in the string between Ola and han in (26”) is also violated by (24) above, but due to the mismatch between use of right dislocation at the upper clause level and main clause intonation at the complement clause level, (24) would be bad even without this constraint. An utterance of (27) metarepresents the assumption that Ola is aware of the situation that the pronoun det (‘it’) represents; the main clause conveys non-propositional, attitudinal information. An utterance of (28), however, presents the proposition expressed in the complement as topical, while the asserted information, in the main clause, is that it is incontestably true that Ola knows. (29) combines the intonation in (28) and the right dislocation in (27), which results in a contradiction and an unacceptable combination of syntax and intonation. (29) is ungrammatical for the same reason as (21) above. (27) [[det er klart at [2OLAi]AU ]FP [[1VET-det]AU ]FP hani]IU it is clear that Ola knows it he ‘It’s evident that Ola knows.’ (28) [[det er [1KLART-at]AU ]FP [[2Ola]AU [1VET-det]AU ]FP ]IU it is clear that Ola knows it ‘It’s evident that Ola knows.’ 52 Fretheim (29) *[[det er [1KLART-at]AU ]FP [[2Olai]AU [1VET-det]AU ]FP hani]IU it is clear that Ola knows it he The focal tone on the adjective klart (‘clear’, ‘evident’) in (28) signals that this predicate belongs to the speaker’s explicature, but that assumption is contradicted by the procedural information encoded by means of the right-dislocation construction in (29). The intonation used in (29) tells us that the explicature includes the main clause predicate, as in (28), but the use of right dislocation tells us that the explicature does not include the main clause predicate; the linguistic material that contributes to the explicature must be in the area that is surrounded by the coreferential items Ola and han. 5. Modal particles in the middle-field position and the right-detached position A modal particle is another type of main clause phenomenon, naturally, because modal particles encode a range of attitudes to the proposition expressed and/or to the speech act performed. Modal particles in Norwegian appear either in the syntactic ‘middle field’ – after the finite verb in a main clause and before the finite verb in a subordinate clause – or in a right-detached position (in ‘extraposition’), optionally in cooccurrence with other right-detached items like right-dislocated pronouns, parenthetical clauses, or vocatives. Middle field particles and extraposed particles have slightly different functions. Extraposed modal particles always convey something about the speaker’s attitude to the proposition expressed by the utterance, while a modal particle in the middle field of an embedded complement clause can metarepresent an attitude to the proposition expressed in the complement. In my illustrations in the present section, the individual to whom the speaker attributes the thought metarepresented in the complement is invariably the person referred to by the subject NP of the preceding main clause. In the fundamental frequency tracking rendered in Figure 1, the utterance-final modal particle vel – basically an indicator of epistemic uncertainty – is a particle located outside the FP-final AU with its right-edge focal tone, but inside the FP. The location of the focal tone is here symbolized as H¯. Vel is aligned with the high boundary tone H% which cues a processing of vel as not only a marker of uncertainty but also a sign that the speaker is using the utterance to perform a request for confirmation. In Figure 2, vel is inside the final AU and the two phonological phenomena H¯ (focal tone) and H% (high boundary tone) coincide on the temporal scale. Phonetically they are both located in the syllable that contains the utterance-final particle vel. An occurrence of vel inside the AU headed by the second, stressed syllable of the past tense form for|svant of the intransitive verb meaning ‘disappear’ is consistent with a parsing of vel as a particle whose syntactic position is inside the complement clause. This opens for a processing of the complement as the vehicle of a metarepresented thought, but Figure 2 could also be interpreted in the same way as Figure 1, depending on the context accessed by the hearer.21 21 H* is the East Norwegian tone for Word-Accent 2, L is the post-lexically inserted default low tone for Word-Accent 1, and H is a post-lexical marker of the end of an AU which is not FP-final. The metarepresentational use of main clause phenomena in embedded clauses 53 Figure 1 Hun tenkte at hodepinen forsvant vel 230 F0 (Hz) 160 80 0 2.4 230 H* L H H* L H L H¯ H% 230 [[hun [2tenkte-at]AU [2hodepinen-for]AU [1SVANT]AU vel ]FP ]IU ‘She thought that the headache would disappear, didn’t she?’ 160 F0 (Hz) 160 F0 (Hz) 80 80 00 2.4 2.4 Figure 2 Hun tenkte at hodepinen forsvant vel 230 230 (Hz) 160 160 F0 (Hz) 80 80 00 2.4 2.4 H* L H H* L H L H¯/H% 230 230 [[hun [2tenkte-at]AU [2hodepinen-for]AU [1SVANT-vel]AU ]FP ]IU ‘She thought that the headache would disappear, wouldn’t it?’, or ‘She thought that the headache would disappear, didn’t she?’22 F0 (Hz) 160 160 F0 (Hz) The prosodic boundary between the verb forsvant and the particle vel in Figure 1 is a metarepresentational reflection of the syntactic boundary between the verb and the extra- clausal particle. Here the metarepresentation is based exclusively on a resemblance in linguistic80 structure between the metarepresenting prosodic device and the 800 2.4 metarepresented 0 syntactic attachment of the particle. 2.4 The most conspicuous prosodic feature displayed in Figure 3 is that the particle vel is accented, 230 and even furnished with a focal tone. 22 Observe that the interpretive differences matching the syntactic and prosodic differences between Figure 1, 160 Figure 2 and Figure 3 can be captured by means of an English difference in the choice of tag F0 (Hz) question, didn’t she? for a particle vel which is extraposed at the main clause level and wouldn’t it? for a particle vel which is inside the complement clause. 80 0 2.4 160 F0 (Hz) 54 Fretheim 80 0 2.4 Figure 3 Hun tenkte at hodepinen forsvant vel 230 F0 (Hz) 160 80 0 2.4 H* L H H* L H L H¯ L H¯/H% 230 [[hun [2tenkte-at]AU [2hodepinen]AU for [1SVANT]AU ]FP [[1VEL]AU ]FP ]IU ‘She thought that the headache would disappear, wouldn’t it?’ All extraposed 160 particles are unaccented; a middle field particle is either unaccented or F0 (Hz) accented. This means that an accent on a Norwegian modal particle offers the procedural information that the particle is inside a clause, in the middle field position. Thus the accented particle vel in Figure 3 belongs to the embedded complement and the speaker metarepresents the woman’s wondering whether her headache would disappear. This is an 80 unambiguous 0 case of ‘free indirect thought’ 2.4 What happens when the two main clause phenomena right-dislocated pronoun and modal particle are co-present in a Norwegian utterance? All the illustrations presented in the remainder of this paper make use of the same logical form that is encoded by the sentences appearing in Figures 1-3, but in addition the utterance types include a right- dislocated pronoun. As always, the speaker’s intonational phrasing contributes crucially both to the hearer’s decisions about how to parse the string and to the pragmatic interpretation of the utterance. The intonation gives procedural information which enables the hearer to resolve issues such as whether the particle and the dislocated pronoun interact and operate at the same syntactic level, or whether the combination of the particle and the dislocated pronoun either cues an interpretation of the complement proposition as a metarepresented thought attributed to the female referent, or else an interpretation which relates the modal particle to the speaker’s uncertainty. The order of linguistic elements that cooccur in an extraposed cluster is such that a right-dislocated pronoun obligatorily precedes a modal particle. Whenever the reverse order is used in a given utterance, the particle and the pronoun do not form a cluster at the same syntactic level but belong to different clause levels. When the pronoun precedes the particle, the hearer’s inferential processing may be consistent with an analysis of the two items as an extraposed cluster, but only if the speaker’s intonational phrasing supports that analysis. The pronoun hun (‘she’) and modal particle vel are adjacent words both in (30a-b) and in (31a-b), but their linear order in (30) is reversed in (31), and that change in the word order results in ungrammaticality and a linguistic stimulus whose relevance is radically decreased. (30) a [[huni [[2tenkte-at]AU [2hodepinen-for]AU [1SVANT]AU ]FP huni vel ]IU b [[huni [[2tenkte-at]AU [2hodepinen-for]AU [1SVANT]AU ]FP [[1HUNi-vel ]AU ]FP ]IU The metarepresentational use of main clause phenomena in embedded clauses 55 (31) a *[[huni [[2tenkte-at]AU [2hodepinen-for]AU [1SVANT-vel]AU ]FP huni ]IU b *[[huni [[2tenkte-at]AU [2hodepinen-for]AU [1SVANT-vel]AU ]FP [[1HUNi ]AU ]FP ]IU The dislocated pronoun hun in (30) is a copy of the main clause subject, so there both the pronoun and the following particle vel must be right-detached at the upper clause level. This means that the particle in (30) modifies the speech act, which can only be understood as a request for confirmation of the proposition of the complex host sentence: ‘Can you confirm my belief that the woman thought that her headache would disappear?’ (31) is bad because the two main clause phenomena do not converge; the dislocated pronoun hun gives procedural information that is not compatible with the procedural information that is due to the modal particle vel and the fact that the verb precedes vel as in a main clause. Because the particle precedes the dislocated pronoun, we can conclude that it belongs to the complement clause. Could the complement clause contain the speaker’s metarepresentation of a thought attributed to the female referent, and if so, would the particle vel indicate this woman’s uncertainty as to how long her headache might last? Possibly, but the presumption that the complement clause in (31a-b) might metarepresent a thought that the speaker attributes to the woman is contradicted by the speaker’s use of pronominal dislocation. When the speech act is not modified by any modal particle as in (30), copying of the main clause subject hun tells us that the explicature of the utterance includes everything encoded in the domain between the two coreferential pronouns. If we delete the modal particle from (30a-b), we are left with an assertion about the woman’s thoughts. One cannot at the same time both assert that the woman thought her headache would (soon) disappear and metarepresent her uncertainty concerning that desirable state of affairs. This situation contrasts with what is communicated by (32a-b), where the dislocated pronoun is a coreferential copy of the lexical phrase hodepinen (‘the headache’), the subject NP of the complement clause. (32) a [[hun [[2tenkte-at]AU [2hodepineni-for]AU [1SVANT-vel]AU ]FP deni ]IU b [[huni [[2tenkte-at]AU [2hodepineni-for]AU [1SVANT-vel]AU ]FP [[1DENi]AU ]FP ]IU Den, whether unstressed as in (32a) or stressed as in (32b), belongs to the metarepresentation. The producer of these utterances mimics the utterance Hodepinen forsvinner vel, den (headache-Def disappears Part, Pron; ‘I guess the headache will disappear’), which was never uttered by the woman referred to but whose meaning closely resembles the thought that the speaker attributes to her. In (33) there are two tokens of the particle vel, one before and one after the dislocated pronoun den. The only accessible interpretation of (33a-b) is that the second token of vel is extraposed at the embedded clause level, so the complement is here a metarepresentation of a question that the woman was asking herself. (33) a [[hun [[2tenkte-at]AU [2hodepineni-for]AU [1SVANT-vel]AU ]FP deni vel]IU b [[huni [[2tenkte-at]AU [2hodepineni-for]AU [1SVANT-vel]AU ]FP [[1DENi-vel]AU ]FP ]IU (34a-b) are utterance types that make native Norwegians laugh. There is a totally unexpected switch to hun after what sounds like a metarepresentation of the woman’s thought by the point at which the complement clause is closed by the particle vel. (34)a *[[hun [[2tenkte-at]AU [2hodepineni-for]AU [1SVANT-vel]AU ]FP huni vel]IU 56 Fretheim b *[[huni [[2tenkte-at]AU [2hodepineni-for]AU [1SVANT-vel]AU ]FP [[1HUNi-vel]AU ]FP ]IU (34a) and (34b) fail for the same reason as (31a-b), but they are even more glaringly bad because the first token and the second token of vel are felt to serve different and mutually exclusive functions, the first one being a sign of the woman’s uncertainty and the next one a question marker. 6. Conclusion One type of intonational phrasing in Norwegian – the double-peak pattern – is a main clause phenomenon that interacts with, and either supports or contradicts, use of certain syntactic main clause phenomena that are frequently found to occur in embedded clauses: (i) word order, (ii) pronominal right dislocation, and (iii) modal particles. These are devices that mark the embedded clause as a metarepresentation of a thought that the speaker attributes to someone. In order for the double-peak pattern to interact fruitfully with a syntactic main clause feature, the two peaks (focal tones) must be located at the same clause level as the syntactic main clause feature. For example, right dislocation of a pronominal copy of the main clause subject is only compatible with subordinate clause word order and no use of main clause intonation or a modal particle in the clausal complement. Conversely, pronominal right dislocation at the complement level means that if any of the other three main clause phenomena examined is co-present, it must be located inside the complement. Acknowledgements Many thanks to the audience at the SPRIK conference in Oslo, June 2006, and to Rein Ove Sikveland for the F0 contours. References Banfield, Ann, 1993. Where epistemology, style and grammar meet literary history: the development of represented speech and thought. In: Lucy, John A. (ed.), Reflexive Language: Reported Speech and Metapragmatics. Cambridge: Cambridge University Press, 339-64. Blakemore, Diane, 1992. Understanding Utterances. Oxford: Blackwell. Frazier, Lyn, 1979. On Comprehending Sentences : Syntactic Parsing Strategies. IULC. Frazier, Lyn and Charles Clifton, Jr., 1996. Construal, Cambridge MA : MIT Press. Fretheim, Thorstein, 1981. Intonational phrasing in Norwegian. Nordic Journal of Linguistics 4 : 111-37. Fretheim, Thorstein, 1991. Intonational phrases and syntactic focus domains. In: Verschueren, Jef (ed.), Levels of Linguistic Adaptation : Selected Papers of the International Pragmatics Conference, Antwerp 1987, Volume II, Amsterdam: John Benjamins, 81-112. Fretheim, Thorstein, 1992. Themehood, rhemehood and Norwegian focus structure. Folia Linguistica XXVI/1-2 : 111-150. The metarepresentational use of main clause phenomena in embedded clauses 57 Fretheim, Thorstein, 1998. Intonation and the procedural encoding of attributed thoughts : The case of Norwegian negative interrogatives. In: Rouchota, Villy & Jucker, Andreas H. (eds.), Current Issues in Relevance Theory. Amsterdam : John Benjamins, 205-36. Fretheim, Thorstein, 2001. The interaction of right-dislocated pronominals and intonational phrasing in Norwegian. In: van Dommelen, Wim A. & Fretheim, Thorstein (eds.), Nordic prosody: Proceedings of the VIIth Conference. Frankfurt am Main: Peter Lang, 61-75. Fretheim, Thorstein, 2002. Intonation as a constraint on inferential processing. In: Bel, Bernard & Marlien, Isabelle (eds.), Speech Prosody 2002. Proceedings of the 1st International Conference on Speech Prosody, CNRS Laboratoire Parole et Langage, Université de Provence, SProSIG, 59-65. Gundel, Jeanette K. & Thorstein Fretheim, 2004. Topic and Focus. In: Horn, Laurence R. & Ward, Gregory (eds.), The Handbook of Pragmatics. Oxford: Blackwell Publishing, 175-96. Gundel, Jeanette K., Nancy Hedberg & Ron Zacharski, 1993. Cognitive status and the form of referring expressions in discourse. Language 69: 274-307. Hooper, Joan B. & Sandra A. Thompson, 1973. On the applicability of root transformations. Linguistic Inquiry IV : 465-98. Nilsen, Randi Alice, 1992. Intonasjon i interaksjon: Sentrale spørsmål i norsk intonologi. (Intonation in interaction : Central issues in Norwegian intonology). Doctoral dissertation, University of Trondheim. Noh, Eun-Ju, 2000. Metarepresentation. Amsterdam: John Benjamins. Sperber, Dan (ed.), 2000. Metarepresentations: An Interdisciplinary Perspective. Oxford: Oxford University Press. Sperber, Dan & Deirdre Wilson, 1986. Relevance: Communication and Cognition. Oxford: Blackwell. Wilson, Deirdre & Dan Sperber, 1993. Linguistic form and relevance. Lingua 90: 1-25. Author’s e-mail address: thorstein.fretheim@hf.ntnu.no 58 WORKING PAPERS ISK 3/2006 59-87 ISK, NTNU A relevance-theoretic analysis of UNLESS Thorstein Fretheim 1. The allegedly ‘exceptive’ proposition P of Q, unless P This paper questions the adequacy of what I shall refer to as the standard approach to the meaning and use of the connective unless. All writers who are representatives of the standard analysis of unless owe a lot to Mike Geis (1973), who in a seminal paper on if and unless pointed out a number of constraints on the distribution of unless compared to if (…) not and argued for a difference in encoded meaning between the two expressions which was based on observed differences in their grammatical behavior. While I agree with Geis and several authors of more recent publications that unless and if not do not mean the same, the difference that I see is not the one of the standard analysis, whose most significant strand is that unless-clauses express an ‘exceptive’ case, implying that the proposition expressed by the main clause (the apodosis) is false just in case the positive ‘unless’-clause proposition is true, and that the truth of P is the exception rather than the norm. I am going to argue that unless is truth-conditionally identical to if not; the lexical component of unless that distinguishes it from if not is a non-truth- conditional encoded aspect of meaning. Declerck and Reed (2001: 449) see the very common interpretation of ‘unless P’ as ‘if and only if ~P’ as a case of conditional perfection, or what they call a “necessity implicature”. The semantics of unless makes the truth of the negative conditional clause proposition (the protasis) in (1)-(3) a sufficient condition for the truth of the apodosis of the main clause. The hearer’s strengthening of ‘if’ to ‘iff’ is due to context-dependent inference: a Quantity-based implicature on the (neo-)Gricean account and a pragmatic process of enrichment of the encoded logical form on the relevance-theoretic account. Looking at (1)-(3) it is fairly easy to appreciate the fact that the ‘if and only if’ interpretation is defeasible. These are Declerck and Reed’s own examples. (1) Unless you pay your debts you get into trouble. (2) Unless you drive carefully, you’ll have an accident. (3) (On this course,) unless you work very hard you don’t get top grades. A producer of an utterance of (1)-(3) does not guarantee that you won’t get into trouble if you do pay your debts, 23 that you won’t have an accident if you are aware that you must drive carefully (and act upon that insight), or that you will succeed in getting top grades if you work very hard. In (3), the strongest inference licensed by the ‘unless’- conditional is that, if you work very hard, there is a chance that you will get top grades, depending ultimately on a number of other variables, like your intellectual grasp of the subject matter, your general aptitude for the kind of academic work you are engaged in, 23 It may be argued that the concept communicated by the noun trouble in (1) is an enriched ad hoc concept which differs from the concept encoded by this noun. The communicated concept would then presumably be ‘the kind of trouble stemming from failure to pay one’s debts’. 60 Fretheim and so on, but if you do not work really hard, the chance to obtain a top grade is reduced to nil. One’s inability or unwillingness to work hard is expressed as a sufficiency condition pertaining to the material conditional of (3). However, if the condition expressed in the positive proposition P in the complement of unless is true, you may at least maintain a hope to obtain top grades. Thus far there is no difference between (1)- (3) and the corresponding conditionals with if not instead of unless. Other ‘unless’-conditionals appear to tell us more straightforwardly that the apodosis Q may be true only if the negative protasis ~P is true, hence false only if the positive proposition P, derived from the complement of unless, is true. Declerck and Reed (2001: 450) offer (4) as an example; the proposition expressed by an utterance of (4) entails the proposition expressed by (5), they say, but implicates the propositions of (6) and (7) through strengthening of the conditional to a biconditional. (4) Unless you pay me now I will sue you. (5) ‘If you don’t pay me now I will sue you.’ (6) ‘Only if you don’t pay me now will I sue you.’ (7) ‘Only if you pay me now will I not sue you.’ The pragmatic strengthening of the encoded meaning of (4) seems to be less dependent on background assumptions than what we witnessed in the corresponding ‘unless’- conditionals in (1)-(3). It would probably even take some extra effort to suppress a biconditional interpretation of an utterance of (4). It is well known that a regular ‘if’-conditional is also susceptible to being strengthened to a biconditional in the absence of a contextual assumption which suggests that no such strengthening is intended. By adopting the analysis of unless to be presented in this paper I am able to account for why ‘unless’-conditionals seem to automatically turn the hearer’s attention to the stronger, biconditional reading. Unless seems to support what has traditionally been referred to as ‘conditional perfection’ (Geis and Zwicky, 1971) even more strongly than if (not) does. Echoing Geis (1973), Declerck and Reed (2001) insist, surprisingly, that unless means the same as except if, and this claim is repeated by Dancygier and Sweetser (2005; see also von Fintel, 1991; Traugott, 1997). Normally a statement of this sort will be interpreted as a claim that the lexical meaning of unless equals the lexical meaning of except if. The prepositional part except of the complex connective except if explicitly singles out P as an exception, or maybe even the unique exception, to the set of conditions that would make the main clause apodosis true. Declerck and Reed’s position is surprising because it is evidently not consistent with their correct observation that utterances like (1)-(3) above will not always be interpreted as biconditionals. Nowhere have Declerck and Reed actually demonstrated, or even tried to demonstrate, that except if is like unless in that the biconditional interpretation of a given conditional construction with except if is not due to direct semantic encoding but to pragmatic strengthening. Doesn’t (3’), where except if appears in place of unless, encode the assumption that top grades is the consequence of hard work? Six native English informants told me that for them, there is intuitively a difference between (3) and (3’), as the latter, as opposed to the former, is not compatible with the possibility that the referent works very hard and still fails to get top grades. (3’) (On this course,) except if you work very hard you don’t get top grades. A relevance-theoretic analysis of UNLESS 61 On the other hand, another four native English informants detected no semantic difference at all between (3) and (3’), reading both utterance types as biconditionals. Their interpretations, however, are consistent with a well-known claim of adherents of contextualism,24 namely that we have intuitions about pragmatically enriched propositional forms but not about the encoded logical form of a sentence or the inferences that fill the gap between the logical form and the proposition expressed (Recanati, 1993, 2004; Gibbs and Moise, 1997). Dancygier (1998: 169) insisted that “we need an explanation for the negation involved in the interpretation of a sentence with unless, if (as seems to be the case) this negative interpretation is not in the semantics of the conjunction.” So if the negation is not in the word unless, where is it? What is negated is not P, Dancygier and Sweetser (2005) say, but Q. ~Q is true in an ‘exceptive’ context, or ‘mental space’, in which P is true and its negated counterpart false. Operating with a default syntactic order where the main clause precedes the ‘unless’-clause, Dancygier and Sweetser (2005: 184) claim that, “… the speaker first asserts Q without explicitly identifying the space to which she is adding Q”, i.e. without having made it manifest to the hearer by linguistic means what contextual conditions must be satisfied in order for either Q or ~Q to be true, but then an ‘unless’-clause may be added as an afterthought or reservation, a way of encoding the unique contextual condition that would fail to make the main clause proposition true. For them the main clause of an ‘unless’-construction “is asserted, with the reservation that in the case of P, it will not hold.” (Dancygier and Sweetser, 2005: 183). They also suggest that the function of the ‘unless’-clause will occasionally be “to start a new line of exceptive reasoning (in a new sentence) which simply did not occur to the speaker at the time when Q was being presented.” (ibid: 184). A true Q is the default case for them; a true ~Q is a consequence of a true P, but P is said to represent the exception and unless is supposed to direct the hearer’s attention to this fact. Reading Dancygier and Sweetser (2005) I got the impression that the authors do not pay enough attention to the distinction between the explicitly communicated truth- conditional content of ‘unless’-conditionals and their total communicated content. They consider the proposition Q of ‘Q, unless P’ to represent the normal state of affairs but they fail to appreciate and acknowledge the implications of the fact that a true Q in ‘unless’-conditionals depends on the truth of the negative proposition ~P. The protasis of a conditional conforming to the type ‘Q, unless P’ / ‘Unless P, Q’ is a negated P, whose truth is a sufficient condition for a true apodosis Q. Denying that ~P is the explicit protasis in constructions of the type ‘Q, unless P’ is to deny that an ‘unless’- construction expresses a conditional, a denial which is also consistent with their claim that Q is asserted. Even so they never say that a declarative followed by an ‘unless’- clause should not be identified as the apodosis part of a conditional, and granted that an ‘unless’-clause can also precede the main clause, an analysis that does not recognize the conditional meaning of ‘unless’-constructions is not commendable. Although I reject the belief that unless means except if at the lexical level, I concede that there is a close semantic resemblance between what unless encodes and what except encodes and that there are situations, maybe even a considerable number of situations, in which unless and except if would be interchangeable. There is an undeniable cognitive similarity between the way we understand ‘unless’-conditionals and the way we understand the communication of exceptions. It is therefore not 24 ‘Contextualists’ argue that sentence meaning vastly underdetermines not just what is communicated by an utterance of a given sentence but even ‘what is said’. 62 Fretheim surprising that there are languages which use a single connective that may be glossed sometimes as ‘unless’ and sometimes as ‘except’. Paul Newman, the Hausa specialist, notes (Newman, 2000) that there is an intimate connection between the way you express ‘if not’/‘unless’ and the way you express ‘except’ in Hausa. The Hausa word sai has a number of functions, including meanings that may be glossed as ‘(not) until’, ‘except’, ‘only if’, ‘unless’, as in Bà zân iyà hawan wannàn ginì ba sai an sakà tsanì, which means something like ‘I won’t be able to climb this wall unless a ladder is put up’. This is not a negative assertion modified by information about a particular exception to it; rather, it is an act of advising the addressee that a ladder is needed before the wall can be climbed successfully (see some remarks on advisory illocutionary acts modified by an ‘unless’-clause in the next section of this paper). Nevertheless, it is apparently true that the most natural gloss for Hausa sai is in many cases ‘except (if)’. It is also worth observing that the meaning of the German preposition außer is very close to the meaning of the English preposition except, and außer is also a frequent translational correspondent of both except for, except if, and unless in the Oslo Multilingual Corpus (OMC), a bidirectional translation corpus. As I am not denying the fact that a clause introduced with unless often informs the addressee of an exception to a general statement, I find it natural that certain languages use one word for what would be either except (if) or unless in English. All I am saying about the relationship between unless and except (if) is that ‘unless’-clauses do not encode an exception by virtue of the lexical meaning of unless, a position I am going to defend in this paper. I have yet to see a convincing argument that attempts to refute the intuitively quite appealing lexical analysis of unless as truth-conditionally identical to if not, and I am going to use this paper to demonstrate that it is not at the level of truth-conditional content that unless differs from if not. Truth-conditionally, unless is exactly like if not. The observed restrictions on the use of unless that appear to favor the opposite conclusion will instead be shown to depend upon a non-truth-conditional component of the lexical meaning of unless. Whenever an ‘unless’-conditional is felt to be bad and the corresponding conditional with if not is pragmatically in order, the unacceptability of the former is due to a pragmatic mismatch between the truth-conditional meaning of the negative protasis and the special non-truth-conditional meaning encoded by the connective unless. Although my analysis goes against standard views on the semantics and discourse function of unless, it is in some respects a more conservative analysis because it accounts for not only all the cases where unless and if not work differently but also all occurrences of unless that could have been replaced by if not without changing the relevance of the stimulus, except maybe in very subtle, non-truth-conditional ways. The standard approach to unless has been promoted by linguists who were keen to demonstrate that this connective differs significantly from if not, but one unhappy consequence of this is that they disregarded the important semantic and pragmatic similarities between unless and if not when either expression is acceptable. 2. Non-exceptive unless-clauses As observed by Declerck and Reed (2000, 2001), the proposition P expressed in the complement of the connective unless does not always represent an exceptional state of affairs, and this is particularly true of ‘unless’-constructions where the main clause precedes the ‘unless’-clause and there is no written comma between the clauses, or A relevance-theoretic analysis of UNLESS 63 where there is no ‘comma intonation’ in a spoken utterance. In the present section I am going to show that ‘unless’-clauses are frequently not meant to express an exception, not even at the inferred pragmatic level of interpretation. By the same token, the main clause proposition Q is not asserted, as Dancygier and Sweetser (2005) claim, except maybe in certain cases where the ‘unless’-clause is not only an afterthought added to a preceding main clause but actually an instance of self-repair, an act of nullifying the assertion of the previous utterance. The data in (8)-(12), from the corpus OMC mentioned earlier, cannot be reconciled with a view of P in ‘unless P’ as a proposition that represents a situation the addressee should regard as exceptional. (8) is a conditional clause which is unaccompanied by a main clause. It is, true enough, added to the sentence preceding it as an afterthought condition, the most typical context for the use of unless according to Dancygier and Sweetser, and yet the reader is not to interpret the proposition P – ‘That person is made quite differently than I am’ – as the exception that would make ~Q true, the exception to the normal kind of situation in which Q is true. (8) Norwegian source text: Om han da ikke25 er innrettet ganske annerledes enn jeg? if he Part not is constructed quite differently than I (KF1)  English translation: Unless he’s made quite differently than I am. French translation: A moins qu’il ne soit dans des dispositions tout à fait différentes des miennes? German translation: Vielleicht ist er aber ganz anders veranlagt als ich? maybe is he however quite differently put-together than I Notice what the German translator did here. The German target text in (8) is not formed as an afterthought condition; it is a self-directed question with the epistemic modal adverb vielleicht (‘maybe’, ‘perhaps’), the meaning being ‘Perhaps he is put together quite differently than me’. This looks like a fairly free paraphrase of the source text, but in my view it is a faithful interpretation of the original Norwegian text. To the extent that it captures the informative intention of the author of the source text, the German target text reveals that the narrator’s thought cannot be interpreted as representing an exception to anything stated earlier. The narrator is trying to find a coherent explanation for the male referent’s behavior as described in the preceding discourse, and like the thought expressed in the self-directed question in the German translation in (8), the conditional clauses in the Norwegian source text and the English and French target texts communicate the narrator’s failure to understand the behavior of the man referred to. Those clauses do not express an exception to an earlier generalization, they express the communicator’s doubt about the validity of a thought expressed immediately before the point in the discourse where (8) appears. In example (9) the ‘unless’-clause follows the main clause of the conditional, but there is no linguistic indication of a break suggestive of an afterthought. The communicative point is not that the man called Ty could not bring himself to sleep with 25 The Norwegian modal particle da in conjunction with the negation operator in a conditional clause has much the same pragmatic function as unless; see Fretheim (2000a). 64 Fretheim the narrator (due to the birth of their third child) and that the narrator adds an exception to what is presented as the default situation. On the contrary, Ty is apparently seriously worried that there will be no sexual intercourse as a consequence of his failing to convince his wife that they now have to take prophylactic measures. (9) English source text: After the third one, in the summer of ’76, Ty said he couldn’t bring himself to sleep with me unless we were using birth control.  Norwegian translation: Etter den tredje, sommeren 76, sa Ty at han ikke klarte å after the third summer-Def 76 said T. that he not managed to ligge med meg med mindre vi brukte prevensjon. lie with me unless we used contraception Putting a comma in front of unless in the English version of (9), or in front of the connective med mindre (‘unless’) in the Norwegian translation, would not have been in accordance with the author’s intention. Although proposition P does represent Ty’s imperative, the point of unless and med mindre is not to inform the reader that the embedded clause after the main clause describes the exceptional situation which would falsify the previous statement. Ty’s decision about sexual abstention is upheld only if there is no sign of a change in his wife’s attitude to birth control. In his opinion they cannot risk another pregnancy after her third birth in 1976. Ty’s own preference, however, is definitely not to give up sleeping with her. What is communicated by means of the unless-conditional in (10) is that Uncle is prepared to do whatever it takes to make him happy. We are not being told that this man is basically not a happy person and that the complement of unless informs the reader of an exception, or maybe the only exception, to this generalization. (10) English source text: “Highly unlikely, but Uncle is never happy unless every stone is explored and every avenue thoroughly upturned.” (PDJ3)  Norw. translation: “Nei, men du kjenner Gammern, han er ikke fornøyd før hver no but you know the Old one he is not satisfied until every gate er endevendt og hver stein undersøkt.” street is upturned and every stone explored The ‘unless’-clause in the English source text of (10) tells us how the man will act in pursuit of happiness. This point is well taken by the Norwegian translator who decided to paraphrase the direct speech in (10) as ‘You know the Old one, he will not be satisfied until every avenue is upturned and every stone explored’ (emphasis mine), implying that this man is always prepared to do the utmost in pursuit of his own feeling of satisfaction, or happiness. A comma before the ‘unless’-clause would have been as A relevance-theoretic analysis of UNLESS 65 wrong here as in (9), because the ‘unless’-clause does not express a reservation and it is not asserted that the man is never happy. The narrator’s claim in (11) is that the subject referent always (a hyperbolic generalization, surely) sings in order to get things done, so his singing is no exceptive case. (11) English source text: He can’t do anything unless he makes it a song. (OS1)  Norwegian: Han kan ikke gjøre noe som helst med mindre han gjør det til en sang. he can not do anything at all unless he makes it into a song German: Er kann nichts tun, ohne ein Lied daraus zu machen. he can nothing do without a song of-it to make As argued for in the subsequent sections of this paper, the encoded meaning of unless that differentiates this connective from if not is the communicator’s advising the addressee that the positive protasis P may be true, and in (11) the main clause creates a context which makes us understand that P (‘He makes any act a song’) is in fact true beyond any doubt. This is consistent with the analysis of the lexical meaning of unless to be presented in section 3, according to which unless is a connective that encourages the addressee to imagine a true P and to use that as an input assumption to compute contextual effects. ‘Unless’-clauses are fairly frequent modifying accompaniments of imperatives, as in Don’t answer unless you’re sure you have the right answer, which can be an act of recommending the hearer not to give an answer prematurely, or impulsively. A possible context for the performance of an utterance of negative imperatives of this sort could be one in which an answer is in fact required, but the speaker’s point is that the hearer ought to take his time before answering. Sometimes an implicit advisory speech act can be derived from the ‘unless’-clause, where the complement of unless expresses the positive proposition whose fulfillment leads to a felicitous situation. (12) is another comma-free ‘unless’-construction from the OMC, but its pragmatic interpretation would be no different if the order of clauses were interchanged. (12) The species won’t survive unless people like you reproduce themselves. (MW1) The relevance of (12) depends on the hearer’s recognition of the recommendation which is communicated implicitly and which is represented linguistically in the complement of unless. The invented example (13) is another advisory speech act, the advice being derived as the implicated conclusion (ii), based on the implicated premise (i), which is made more accessible to the hearer when unless is used than if the connective is if and its complement is a clause that contains the negation operator not. The formal absence of not in the complement of unless (because unless is inherently negative) may be described as an echoic metarepresentation of the speaker’s recommendation. Since there is no marker of negation in the complement of unless in (13), it is presumably easier for the hearer to associate that non-negative substring you take two of them with the speaker’s positive attitude to P and negative attitude to the explicit protasis ~P, than if the utterance had been Those pills have no effect if you do not take two of them. 66 Fretheim (13) Those pills have no effect unless you take two of them. (i) +> Those pills do have a positive effect if you take two of them rather than just one. (ii) +> I advise you to take two pills. The addressee of an utterance of (13) is advised to take two pills (at a time), i.e. to behave in such a way that the positive proposition P expressed in the complement of unless becomes true, because a situation in which P is true is necessary (and sufficient) to obtain the expected effect of the pills. It is not the communicator’s intention to tell the addressee that the pills have no effect and that the situation represented by P is the only exception that would render the negative main clause proposition false. The focus of information is the content of the ‘unless’-clause, which is an integral part of a conditional planned as a conditional, not afterthought information that makes an apodosis out of something not initially planned as the apodosis of a conditional. My next example is an interesting ‘unless’-construction that I found in Cappelen and Lepore’s book Insensitive Semantics (2005: 13). (14) The distinction between various versions of MC and RC will seem unimportant (since these questions don’t even arise unless one makes certain false assumptions). The philosophers Cappelen and Lepore point out that there are Moderate Contextualists (MC) and Radical Contextualists (RC) galore in the linguistic as well as the philosophical camp, and these scholars make the very common, allegedly false assumption that there is a significant difference between what so-called MC proponents and so-called RC proponents stand for. The authors’ point is that what they regard as misguided questions arise again and again among contextualists inspired by the late Wittgenstein, Searle, Recanati, Sperber & Wilson, etc. It is not as if the questions referred to are not addressed by semanticists and pragmatists. On the contrary, Cappelen and Lepore deplore the fact that so many of the practitioners in the field of semantic and pragmatic research are guilty of having made a certain false assumption, namely that the difference between embracing MC arguments and embracing RC arguments is a genuine and important difference. Finally, one of Dancygier and Sweetser’s own illustrations of how unless becomes an instrument in “exceptive space building” actually provides evidence against their analysis of unless. There is nothing exceptive about proposition P (enriched pragmatically as ‘Old Potter was there to supervise the use of the film camera’) in their attested example (15) (Dancygier and Sweetser, 2005: 184). (15) “I take good pictures: you’ll see. Bloody old Potter would never let me shoot an inch of film unless he was there.” The speaker claims to be a good photographer and we understand he has acquired the skill by learning from a man named Potter. From this we will certainly not infer that Potter did not let the speaker use the film camera, with one exception specified in the complement of unless. (15) causes us to form a mental picture of two complementary types of situation, one in which the speaker is shooting film and Potter is present and A relevance-theoretic analysis of UNLESS 67 one in which Potter is not present and the speaker is therefore not shooting film. An utterance of (15) would not be relevant if the speaker had intended to communicate that the former kind of situation was relatively rare, or exceptional. Indeed, (15) seems to implicate that old Potter saw to it that he himself was present whenever the speaker was shooting film during his apprenticeship period. As in our previous illustrations, the author of the text fragment in (15) used no comma before the ‘unless’-clause, a graphical device that Dancygier and Sweetser see as a characteristic cue to their favored exceptive circumstance reading of ‘unless’- clauses. I venture to propose the generalization that, whenever the main clause precedes the conditional clause and no comma is written to separate the clauses, the ‘unless’- clause does not present P as an exception that would falsify the (asserted) proposition of the preceding main clause. Now consider the following illustrations (from Bayiga 2005: 34-35) of the use of the inherently negative conditional connective okujjako nga (literally: to remove + complementizer; i.e. ‘unless’) in the Bantu language Luganda, the major indigenous language of Uganda. Bayiga notes that the linguistic structures in Luganda which correspond to English ‘unless’-clauses are either placed after the main clause, as in (16) and (17) below, or before it, as in (18). When the order is as shown in (16) and (17), there is, the author says, not necessarily a break between the clauses which would indicate that the conditional clause is added as an afterthought. (16) Ojja kusigala nga okola ekibonerezo ekyo okujjako nga okyusizza mu mpisa. you-will to-stay compl you-do punishment demonstr unless you-change in behavior ‘You will continue to serve that punishment unless you change your behavior.’ (17) Peter tajja kwoza bintu okujjako nga oyozezza engoye. P. not-he-will to-wash dishes unless you-have-washed clothes ‘Peter will not wash the dishes unless you do the laundry.’ (18) Okujjako nga osasula sente zange, ennyumba yo eja kukumibwako omuliro. unless you-pay money mine house yours it-will to-be-set-on fire ‘Unless you pay what you owe me, your house will be set on fire.’ In (16), the addressee is advised to change his behavior in order to avoid the situation described in the preceding apodosis; the utterance of (17) is another act of advice, which, if followed up by the addressee, will prevent the situation described in the apodosis; finally, an utterance of (18), where the Luganda clause introduced by the connective okujjako nga (‘unless’) precedes the main clause, is a request to the addressee to make the positive proposition in the complement of the connective come true, so that the disastrous situation depicted in the following main clause will not arise. I have demonstrated in this section that the positive complement of unless represents a state of affairs which has the speaker’s approbation if the main clause represents a disadvantageous state of affairs and the speaker’s disapproval if the main clause represents a beneficial state of affairs. This generalization, however, presupposes an analysis of the explicit propositional content of the ‘unless’-clause as negated, just as when the conditional clause contains the words if and not. 68 Fretheim 3. The non-truth-conditional meaning of unless stated in procedural terms The belief that the domain of linguistic semantics should be defined in such a way as to incorporate the encoding of procedures for the addressee to follow has long been regarded as one of the cornerstones of Relevance Theory (RT). RT makes a distinction between conceptual semantics, which concerns the encoding of meaning that enters into the representational content of an utterance, and procedural semantics, which constrains the inferential phase of the hearer’s comprehension (Blakemore, 1987, 2002; Blass, 1990; Wilson and Sperber, 1993; Andersen and Fretheim, 2000; Fretheim, 2000b, 2001, 2006; Fretheim, Boateng and Vaskó, 2002; Bezuidenhout, 2003; Amfo, 2005; Ifantidou, 2001; Unger, 1996). A linguistic expression with an encoded conceptual meaning typically contributes to the truth-conditional content of an utterance containing that expression, but in addition there are linguistic items with a conceptual meaning which have no impact on truth-conditional content but which contribute to what relevance theorists call higher-level explicatures (Blakemore, 1992; Wilson and Sperber, 1993), like the attitudinal adverb unfortunately in (19) below. Comparing (19a) and (19b) we appreciate that there is no difference in truth-conditional content (‘ground-floor’ explicature) between them, provided the time of utterance and the boat referred to are the same. However, the longer version in (19a) communicates something about the attitude of the speaker to the explicated proposition: it is unfortunate that the number of men in the boat is two instead of three. The occurrence of unfortunately in (19a) cold mean a greater processing load than what the addressee of (19b) is faced with, but there may be a compensatory gain in terms of contextual effects. If the speaker intends to inform the hearer that she feels it is regrettable that there are only two men in the boat, she can produce an utterance of (19b) accompanied by a sad facial expression, or she can add the word unfortunately that encodes her feeling toward the proposition expressed, as in (19a). Either way the hearer will obtain the intended contextual effects only if he grasps the speaker’s proposition as well as her attitude to it (the higher-level explicature). (19) a. There are unfortunately not three, but two men in the boat. b. There are not three, but two men in the boat. The higher-level explicature associated with an utterance of (19a) can in turn direct the hearer to implicated consequences of the fact that there are just two men, or to the reason why the speaker feels it is unfortunate that they are not three. Maybe there were originally three men in the boat and one of them is now missing, etc. Other linguistic items – so-called ‘function words’ in particular – encode a procedure, information which is intended to facilitate the hearer’s derivation of contextual effects from linguistic stimuli by advising the hearer on how to manipulate the more or less composite concepts encoded by so-called content words belonging to open lexical classes. Procedural encoders constrain the hearers’ search for a relevant interpretation. Specifically, they reduce their mental effort in the pragmatic phase of the comprehension process, help them recognize logical relations between consecutive parts of the discourse, identify explicatures as intended, and choose inferential paths leading to implicated conclusions. A relevance-theoretic analysis of UNLESS 69 Connectives of various sorts are representatives of the group of linguistic items which encode a procedure rather than a concept, though some of them are made up partly, or wholly, of lexical elements with an undeniable conceptual meaning, like the preposition in and the noun case in the conditional connective in case, where the noun actually alludes to the occasion-sensitivity (Bezuidenhout, 2002) of information put into a conditional clause. Other connectives, like if, do not bring any particular concept to mind. Rather, if is an operator that instructs the hearer to construe the proposition in its scope as a contextual premise whose fulfillment affects our interpretation of the rest of the conditional construction, often directing us to mental representations of causal relations between the propositions that serve as protasis and apodosis, respectively. Even linguistic items made up of one or more components that we recognize immediately as words with a specific conceptual meaning when they appear in other grammatical environments tend to be disassociated from those concepts when they are a proper part of a structurally complex connective. Occasionally an item will go through a historical process in which an innovative use of a word results in the emergence of a new lexical item with a procedural semantics. The English noun while, a cognate of the Latin noun quiēs (‘rest’, ‘quiet’), is the historical source of the English temporal connective while. The temporal connective while tells us that the proposition in its scope coincides temporally with the space of time – the ‘while’ – during which the proposition of the accompanying main clause is true. And then a later lexical split has led to the existence of two procedural encoders while in present-day English, the concessive connective meaning ‘whereas’ in addition to the older, less abstract temporal connective (cf. Traugott and Dasher, 2005). Unless is a conditional connective with a procedural semantics that constrains the hearer’s inferential comprehension both at the explicit and at the implicit level of communication. On the one hand unless serves the same function as the combination of the connective if and the negation marker not, which means that the accompanying main clause is to be logically related to the conditional clause with unless as apodosis and protasis, respectively, at the explicit level of truth-conditional content; furthermore, unless tells the hearer to contextualize in a specific manner, instructing him to conduct his pragmatic handling of the utterance in accordance with the lexically determined contextual constraint stated in (20): (20) The procedural meaning of unless: Associate the complement of unless with a proposition P that contradicts the negative protasis of the conditional, and let your inferential processing be guided by the assumption that this proposition may be true. My claim is that the non-truth-conditional semantic meaning of unless presented in (20) takes care of all known restrictions on the use of this connective (see section 4) and all of its potential contributions to the overall relevance of utterances in which it is present, including truth-conditional strengthenings of ‘Q if ~P’ to ‘Q if ~P & ~Q if P’. Dancygier and Sweetser (2005: 202) cite the ‘unless’-conditional rendered in (21). It would appear that my analysis of unless falls short of explaining the fact that an utterance of (21) would normally implicate that the addressee has called the speaker at a time when he is not “in the utmost need”. It is natural to believe that speaker and hearer are in the midst of a call of the sort that the speaker deems to be unnecessary. (21) “I told you that you were not to call me unless you were in the utmost need …” 70 Fretheim (Andrew Lang, Red Fairy Book, U-Va Electronic Text Center) As the relevance of (21) would normally depend on the speaker’s judgment that the condition specified in the complement of unless is not satisfied on the present occasion, it is pertinent to ask how (21) can be an acceptable alternative to (22) with if …not? (22) “I told you that you were not to call me if you were not in the utmost need …” The answer is that (21) makes the hearer focus on the positive proposition P derived from the complement of unless in a way that (22) does not. An utterance of (21) makes it more manifest to both parties in the conversation that the speaker admits that P could be true in a conceivable alternative context, and so an utterance of (21) implicates more strongly than (22) that the speaker would accept a call from the hearer in the event that P was true. Even if P (‘You are in the utmost need of help’) is not true at the time of utterance, the speaker of (21) implicitly concedes that there could be contexts in which P would be true. On the other hand an utterance of (22) can be said to implicate, probably more strongly than (21), that the speaker believes it to be true that the hearer is not in the utmost need. (23) is an English translation of a Norwegian source text found in the OMC. The final utterance consists of no more than the word unless followed by three dots which suggest a startling interruption of the train of thoughts running through the head of this woman sitting by her father’s bedside. (23) His voice sounded strangely remote and alien and faded away in a long sigh as he released her hand. A moment later his head fell back on the pillow and his eyes closed, and when the daughter bent over him she couldn’t decide whether he had fallen asleep or had merely sunk into one of his many moments of preoccupation. Unless … What follows (23) in the novel that this excerpt belongs to confirms our suspicion that the man in the bed actually died at the point when his eyes closed. The three dots after unless are indicative of his daughter’s confusion and maybe unwillingness to accept the thought that she had just experienced the moment of her father’s death. (23) does not tell us explicitly that the old man died, we have to infer it on the basis of the descriptions of the man offered in the preceding sentences, in conjunction with the non- truth-conditional procedural meaning of the connective unless at the end of the quote. This meaning, stated in (20), helps us to derive the information that the man passed away. Unless is logically ‘if not’, so the preceding descriptive information can be true only if the subdued proposition that we associate with the zero complement of unless is false. However, unless also advises us to compute contextual effects that would follow from the truth of the proposition we derive from the complement of unless. There is no such complement in (23), so there is not even an incomplete conceptual meaning in the fragment Unless … that would constitute a logical form, the basis for any inferential development at the explicit level of content. Nevertheless, the fragment is relevant to us because we feel intuitively that it provides the reader with a different explanation for the man’s behavior than his daughter’s assumption that he might have fallen asleep or that he might have sunk into one of his moments of preoccupation. The most accessible thought is clearly that she had been witness to his death. This is communicated as an implicature of the single-word utterance Unless … in (23). ‘Her father died’ is an A relevance-theoretic analysis of UNLESS 71 implicated conclusion for the reader of the excerpt of (23) but at the same time it is an implicated premise P in an ‘unless’-conditional of the type ‘~Q if P’: if her father died when his eyes closed, then it is false that he had fallen asleep and it is false that he had sunk into another of his frequent moments of preoccupation. The zero complement of unless must represent an alternative to the suppositions expressed in the preceding ‘whether’-complement. If not … could conceivably have been used instead of Unless … in (23), but a grammatically interrupted utterance If not … – and even the not so radically stripped sentence fragment If he hadn’t … – would not have offered the reader the same cue to the correct interpretation of (23) as unless. The sequence of words if not does not encode a procedure that causes the reader of the book to assess the assumption that the real reason why the man’s head fell back on the pillow and his eyes closed is one that it was extremely difficult for his daughter to accept as true, and which had therefore first escaped her imagination. The word unless may be said to trigger a ‘causal disposition’ (Bezuidenhout, 2003) to process a negative conditional in one particular way. It makes one important contextual assumption especially manifest to the addressee and should, when used correctly, reduce the addressee’s search for a pragmatic interpretation that makes the stimulus optimally relevant to him. The encoded procedure highlights the proposition associated with the complement26 of unless, which may be true for all the speaker knows, even though it contradicts the negative protasis of the explicated conditional. While Bezuidenhout sides with relevance theorists in her belief that certain decoding processes do not access concepts but trigger procedures, those procedures, she says, “are not strictly part of the language” (2003: 128). I agree with Bezuidenhout that the procedures triggered are outside the language system but the actual trigger is inside. A function word like if does not encode a concept; it informs us how to construe the logical relationship between the syntactically dependent ‘if’-clause and the clause of which it is a constituent. All natural language systems seem to have at least one linguistic item with a function analogous to that of if. Unless is a connective with a function similar to if, but it is lexically more complex, both because it is inherently negative and because it has a non-truth-conditional lexical component that makes its syntactic distribution more restricted than the distribution of if … not. The knowledge that enables us to judge whether a given occurrence of unless makes the syntactic structure well-formed or ill-formed (e.g. counterfactive conditionals with unless or polar interrogatives modified by an ‘unless’-clause; see section 4) is, I would claim, part of our linguistic competence. 26 The reader may wonder why I keep referring to the syntactic complement of unless. Why not ‘the scope of unless’? The connective if is a natural language operator with a certain scope, a negation operator like not also has a scope, which is indicated by conventional means in some cases and determined via inference in other cases. Unless embodies both the meaning of if and the meaning of not. I would be truly reluctant to talk about the scope (singular) of if not, whether those two items are syntactically adjacent or are discontinuous in the syntactic structure, and I am equally reluctant to talk about the scope of unless, as unless and if not are truth-conditionally equivalent expressions. 72 Fretheim 4. Restrictions on unless 4.1 Unless is impossible if the apodosis expresses a consequence of a true protasis An ‘unless’-clause is presented to the addressee as a piece of contextual information which embodies two contrary points of view with regard to one’s epistemic attitude to propositions that serve the function of antecedent (protasis) and consequent (apodosis) in a conditional. On the one hand the main clause of the conditional construction expresses the apodosis Q and the protasis is the negative proposition ~P of the ‘unless’- clause. On the other hand the connective unless advises the addressee to consider consequences of a potentially true proposition P instead of ~P. This sort of ‘polyphony’,27 the signaled conflict between two conditionals, one of them explicated (~P Q)28 and the other one implicated (P ~Q),29 is made impossible whenever ~P is claimed to be true, or stipulated to be true, or imagined to be true for the sake of argumentation, more generally, whenever the conclusion in the main clause of a conditional depends upon ~P as a true premise. If the speaker’s intention is for the hearer to process the main clause in the context of a negative protasis whose truth is either stipulated or presupposed, then if … not is the speaker’s only option; unless will then be impossible, because by using unless the speaker grants the hearer that, as far as the speaker is concerned, P may well be true, so one should take into consideration what would follow from a true P as well as what is explicitly said to follow from a true ~P. Declerck and Reed (2001: 451) use a pair like (24)-(25) below to illustrate what I believe to be the effect of the specific constraint on unless stated in (20) in the previous section, a non-truth-conditional constraint that does not affect conditionals with if … not. This is a lexically based constraint that directs the hearer to a particular non-truth-conditional belief, i.e. to the recognition that the speaker does not exclude the possibility of a true P. There is nothing analogous to this in Declerck and Reed’s, or in Dancygier and Sweetser’s, account of the meaning of unless. Since (24) is a conditional promise which is dependent on the assumption that the hearer is not going to let the speaker down, (24) cannot be replaced by (25), which Declerck and Reed somewhat surprisingly furnish with an asterisk. (24) If you don’t let me down, I’ll give you £50. (25) *Unless you let me down, I’ll give you £50. Notice, though, that unless you let me down in (25) is all right if we switch the order of the clauses and add the ‘unless’-clause as an afterthought. In (26), the declarative sentence I’ll give you £50 was not originally intended to be processed as part of a 27 Current work on ‘polyphony’ is inspired by Bakhtinian dialogism and by Ducrot’s argumentation theory (e.g. 1982, 1984) and various research traditions influenced by his work (e.g. the Scandinavian Theory of Linguistic Polyphony – ScaPoLine – outlined in Nølke, 2006; see also Fløttum, Dahl and Kinn, 2006). 28 An ‘explicated proposition’, or ‘explicature’, is the relevance-theoretic equivalent of a pragmatically enriched concept of ‘what is said’ (Recanati, 1993, 2004). For Sperber and Wilson (1986) and Carston (2002) an explicature is a proposition which is inferentially developed from a grammatically encoded ‘logical form’. 29 According to Relevance Theory (Sperber and Wilson, 1986; Wilson and Sperber, 2004), an ‘implicated proposition’, or ‘implicature’, is a communicated thought which the addressee recovers solely by means of (global or local) inferential processes. A relevance-theoretic analysis of UNLESS 73 conditional, as opposed to the situation in (25) where the position of the ‘unless’-clause in front of the main clause forces us to read the latter as a planned apodosis. (26) I’ll give you £50 – unless you let me down. Notice also that (27) is a lot more acceptable than (25), because here except if has replaced the connective unless, and except if does not encode the same procedural information as unless. (27) Except if you let me down, I’ll give you £50. My conclusions based on the data of (24)-(27) are the following: unless is wrong in (25) because the speaker’s plan is to obtain the hearer’s pledge that she will not be let down, and the speaker’s promise of a sum of £50 is meant to make it easier for the hearer to guarantee that no such thing is going to happen. Unless would have told the hearer that possible consequences of a false protasis should be considered as well, and that would be counterproductive in a conditional where the transfer of £50 mentioned in the main clause depends on the truth of ~P. (27) is acceptable, because it presents the apodosis as the normal outcome and the thought that the hearer will let the speaker down as the (only) exception. (26) is also acceptable, because in uttering the declarative of (26) the speaker first asserts that the hearer is going to receive £50 from her, but then she adds the ‘unless’-clause because she feels it is reasonable to inform the hearer that she cannot rule out the possibility that he might be tempted to let her down, which will automatically lead to nullification of the speaker’s promise. Declerck and Reed (2001) have no explanation for why if … not and unless are not interchangeable when the communicated assumption is that the apodosis Q results from the truth of the negative protasis ~P. Consider again their pair (24)-(25), compared to another pair of theirs, (28)-(29). (28) I will be surprised if that book doesn’t sell well. (29) *I will be surprised unless that book sells well. For Declerck and Reed, (25) and (29) are presented to the reader as if they were equally bad, and bad for the same reason. In fact, there is a true difference between them (observation due to Peter Svenonius, p.c.). (25) is just pragmatically odd (and not even that for some native speakers), and we have seen that, if the clause order is interchanged and the ‘unless’-clause is added as an afterthought (cf. (26)), the result is a perfectly acceptable linguistic structure. (29), on the other hand, does have the clause order employed in (26), and will benefit neither from a change in the order of clauses, illustrated in (30), nor from the presentation of the ‘unless’-clause as afterthought information, illustrated in (31). (30) *Unless that book sells well, I will be surprised. (31) *I will be surprised – unless that book sells well. Hearing something like (30) or (31) one gets the impression that the speaker’s contingent surprise reaction is directed to something other than the state of affairs represented in the conditional clause, but on that kind of interpretation the speaker will normally be surprised at something that already exists as a fact, which calls for the 74 Fretheim simple present tense rather than the modal auxiliary will appearing in (30) and (31), for example something like ‘Unless that book sells well, I’m surprised that she spent a whole year writing it.’ On the intended interpretation that the speaker’s surprise reaction would follow from the book’s not selling well, unless cannot replace if not. (32)-(33) is another pair cited by Declerck and Reed (2001: 451), in which unless cannot substitute for if … not.30 (32) This table is going to be difficult to move. I hope John comes. We’ll have a problem if he doesn’t come. (33) This table is going to be difficult to move. I hope John comes. ??We’ll have a problem unless he comes. The speaker of (32) stipulates the truth of ~P in the conditional clause and states what its undesirable consequence will be; in the preceding utterance she informs us that she is hoping for P to come true. Given the fact that the speaker of (32)-(33) has already told us about her positive attitude to P, her consequent negative attitude to ~P and her reasons for preferring P, it is truly weird for the speaker of (33) to encourage the hearer to work out possible consequences of P. The speaker’s expression of a hope that John comes is compatible with the continuation in (32), however, where the focus is on what they risk, if it turns out that the condition specified in the protasis holds. The fact that the main clause and the following ‘unless’-clause were not separated by a comma in the final sentence of (33), or by some other indication that the ‘unless’-clause is to be processed as an afterthought, affects the relevance of the utterance negatively. Nothing is wrong in (33’) and (33”), though, where the middle utterance of (32)-(33) has been eliminated, where a comma is added in (33”) which was not in (33), and where we understand there to be a communicated cause-consequence relation between the first and the second utterance. (33’) This table is going to be difficult to move. Unless John comes, we’ll have a problem. (33”) This table is going to be difficult to move. We’ll have a problem, unless John comes. In (34), the speaker’s reference to ~P is an echo of someone’s interpretation of the addressee’s preference. Generally speaking, indicative conditionals, where the protasis echoes a thought attributed to someone (see Noh, 2000), require the immediately more transparent negative conditional expression with if and not. (35), with unless, sounds very strange compared to the perfectly normal and easily interpretable (34). (34) You don’t have to be there if you don’t want to be there. (35) #You don’t have to be there unless you want to be there. The negative conditional clause in (34) establishes the context in which the speech act of the main clause is to be processed and interpreted, but an utterance of (35) would 8 The authors gave no explanation for the fact that they put two (raised) question marks in front of the final declarative in (33), instead of an asterisk, or more appropriately, the double hatch sign # indicating irrelevance or incoherence. A relevance-theoretic analysis of UNLESS 75 seem to convey the following extremely unlikely message: ‘I appreciate that you may in fact wish to be there, but if that is indeed what you desire, then it is obligatory for you to be there. You may choose whether or not to be there only if you do not want to be there’. The function of the afterthought conditional clause in (36) is to establish a mutually manifest context which is intended to make the preceding wh-question relevant. It is relevant just in case the speaker has not already arrived in Trondheim. (36) is an attested example from an email message, where the authentic sign ‘–’ underscores the intended afterthought function of the conditional clause. (36) When do you arrive in Trondheim – if you aren’t already here? In contrast to the negative ‘if’-clause in (36), the afterthought ‘unless’-clause in (37) would encourage the hearer to consider an alternative context, namely one in which the hearer is already in Trondheim, but that context contradicts a thought which is presupposed by the wh-question, i.e. the speaker’s own assumption that the hearer has not arrived in Trondheim yet. (37) When do you arrive in Trondheim – # unless you’re already here? I admit that it is possible to say something like When do you arrive in Trondheim – or maybe you’re already here?, which does not appear to be so different from (37), but the disjunctive marker or is seen to explicitly offer the addressee a chance to give an affirmative answer to exactly one of two mutually exclusive questions. 4.2 There are no negative polarity items in an unless-clause Dancygier (1985) noted that the negative polarity item (NPI) anything in (38) must be replaced by the positive polarity item something in the ‘unless’-construction in (39). This difference between (38) and (39) falls out of my analysis of unless. The speaker of (38) assumes the truth of ~P, which rules out unless; the speaker of (39), however, opens for the possibility that P may be true, and a potentially true positive proposition P requires the form something rather than anything. (38) You’d better keep silent if you haven’t anything to say. (39) You’d better keep silent unless you have something to say. An NPI in the complement of unless in (39) would make it harder to undertake the kind of parallel processing that unless requires, i.e. a consideration of the pragmatic implications of a true P as an alternative to the explicit protasis ~P. While (38) could be uttered in a conversational setting where the interlocutor has already admitted that he does not have anything to say, an utterance of (39) would not be relevant in that kind of setting, as it would be a way of admitting that the hearer may have something to say. Example (40) is from the OMC. (40) “Also,” he said, “if you don’t see any point to life, I can’t figure why a rainstorm would make you nervous.” (AT1) 76 Fretheim Substituting unless for if not in (40) is grammatically illicit. (41) “Also,” he said, *“unless you see any point to life, I can’t figure why a rainstorm would make you nervous.” The speaker responsible for the direct quotation in (40) is apparently echoing a sentiment of the interlocutor’s in the negative conditional clause: ‘You’re telling me that you don’t see any point to life, but why, then, this anxiety when you are faced with something as trivial as a rainstorm?’ Unless encourages the addressee to consider possible cognitive effects stemming from the assumption that P is true instead of ~P. In (41) this instruction is obviously irrelevant, as it has to be matched against the information that the speaker cannot figure why a rainstorm would make the hearer nervous. But the ‘unless’-clause in (41) is not only irrelevant, it is also ungrammatical. I take the grammar-determined fact that unless blocks the appearance of an NPI like anything in the same sentence to be a natural consequence of what I have defined as the procedural meaning of unless. There is a pragmatic reason for the emergence of this constraint on the legitimate use of NPIs in the grammar of English. The earliest reference to the prohibition of the co-occurrence of an NPI with the connective unless is due to Geis (1973), who gave us a pair such as (42)-(43) to speculate on. (42) If John doesn’t care a bit for Mary, he shouldn’t marry her. (43) *Unless John cares a bit for Mary, he shouldn’t marry her. If we accept Dancygier and Sweetser’s (2005) argumentation and say that an ‘unless’- clause does not express a negative proposition, then we surely do have an explanation for the ban on NPIs in ‘unless’-clauses. However, it still holds that John should not marry Mary if he does NOT care for her; that is the explicitly conveyed conditional in (42) and it would also be the explicitly conveyed conditional in (43) if we drop the problematic NPI a bit. I would analyze (42) as an indicative conditional where the negative protasis represents the speaker’s interpretation of the situation for the couple referred to. This contrasts with the misplaced ‘polyphonic’ character of the alternative version with unless in (43). The NPI a bit is disallowed in (43), but the reason for this is not that the ‘unless’-clause does not express a negative proposition. While the explicit protasis is negative just as in (42), unless is a conditional connective that instructs the hearer to pay attention to the potential truth of the contradictory positive proposition in its complement. The ethical judgment that John shouldn’t marry Mary is void in a context in which P is true instead of ~P, i.e. a context in which John does care for Mary. This alone is not, however, sufficient to render (43) ungrammatical, because, as I said, the string is made grammatical if the NPI a bit is left out. The presence of a bit in (43) causes the addressee to concentrate exclusively on the consequence of a true negative proposition ~P, whereas the connective unless invites the addressee to decide whether it is ~P or P that meshes with the addressee’s view of the world, and to carry out the inferential processing of the stimulus accordingly. As the complement of unless in (43) includes an NPI that would be illicit if it occurred in a syntactically independent declarative (*John cares a bit for Mary), the grammatical ban on NPIs in the complement of unless seems well motivated. The speaker’s instructing the hearer to pay attention to P as well as ~P in the pragmatic interpretation of the utterance will lose its A relevance-theoretic analysis of UNLESS 77 effect if the complement contains an NPI, because its presence would draw the hearer’s attention away from the possibility of a true non-negative proposition P.31 4.3 An unless-clause cannot modify a ‘yes/no’ interrogative By adopting the position that the special meaning encoded by the connective unless should be defined in terms of the speaker’s attitude to the negative protasis, I can account for the fact that unless is never used in a conditional clause adjoined to a ‘yes/no’ interrogative. While (44) is grammatically well-formed and easy to interpret, (45) is just nonsense. (44) Would you say that the sentence sounds better if I do not use the word ‘beautify’ twice? (45) *Would you say that the sentence sounds better unless I use the word ‘beautify’ twice? The speaker’s polar question in (44) must be processed and interpreted in a context that is consistent with the overtly expressed condition of the negative ‘if’-clause. We interpret (44) as a question whether the interlocutor feels that the sentence referred to is stylistically better if the word beautify is used just once, than if it is used twice in the same sentence. The written draft at hand apparently contains two tokens of the verb beautify. Given that contextual premise, it would be strange for the speaker to advise the hearer to imagine what implications a double occurrence of that rather special verb might have, while at the same time asking if a single occurrence would sound better. The condition specified in the conditional clause of (44) is an integral part of the propositional content that the speaker requests the hearer to confirm or deny, and it would therefore be counterproductive to use the connective unless, which tells the hearer to take even the contrary condition into account in his pragmatic processing. 4.4 An unless-clause cannot co-occur with a declarative which is modified by a modal operator like maybe/perhaps While (46) may impress one as being a rather callous remark, it is a timely one in the context of the implicated premise that the hearer is pressing his body so tightly against the chicken that it is squeezed. What is claimed to be a potential consequence of a true protasis in (46) is that the chicken has a better chance to survive if the interlocutor stops squeezing the poor creature. (46) Maybe the chicken will survive if you do not squeeze it. By virtue of its lexically encoded procedural meaning the connective unless in (47) directs the addressee’s attention to the uncertainty of the thought expressed in the negative protasis, but as several researchers have observed (see section 1 of this paper), an ‘unless’-clause is typically intended to be processed as an exception when it is added 31 Ivan Garcia-Alvarez has informed me that certain NPIs are acceptable in ‘unless’-clauses, giving me ever as one example, as in Unless you ever met my brother, you’ve no idea what he might decide to do. My native English informants are not too enthusiastic about that sentence, and I have no explanation for why ever should work here. 78 Fretheim as an afterthought. There is no point in stating an exception to a statement that is already softened by a modal operator like maybe. (47) *Maybe the chicken will survive, unless you squeeze it. The ungrammatical string of (47) contrasts with the grammatically well-formed structures of (46) on the one hand and (48) on the other. Observe the significance of the comma-free form of (46) and my deliberate use of a comma in (48). Maybe in (46) suggests that the conditional clause is not added as an afterthought, but the acceptability of the ‘unless’-conditional in (48) depends upon our ability to process the conditional clause there as afterthought information. The speaker of (48) is warning the hearer of what is likely to happen if the hearer starts to squeeze the chicken. (48) The chicken will survive, unless you squeeze it. Maybe encodes an epistemic uncertainty similar to the signaled uncertainty of a person performing a polar question, like the speaker of (44) above. A conditional with an ‘unless’-clause does two things for us: it instructs us to consider implications of a true P, and it tells us what follows if ~P is true, as in (48). (It tells us more than what possibly follows if ~P holds.) The reason for a speaker’s adding an ‘unless’-clause as an afterthought, as in (48), is precisely that the preceding statement is felt to be somewhat too strong. It cannot be too strong if it is modified by a modal operator like maybe or perhaps. 4.5 Unless is disallowed when a pro-form in the main clause is meant to be coreferential with the proposition of the conditional clause In (49) the subject pronoun it is a cataphoric pronoun whose referent is the proposition expressed in the following conditional clause. Because of the parallel pragmatic processing that unless invites the addressee to engage in, an ‘unless’-conditional cannot ever contain a main clause pronoun whose antecedent is to be located in the conditional clause. (49) It’s all right with me, if you do not open the window. (50) #It’s all right with me, unless you open the window. The speaker of (49) is telling the interlocutor that the apodosis of the conditional is true in a context in which the protasis ~P is true. The addressee’s focus must be on the negative protasis, and not on its contradictory counterpart P, because the issue is whether the speaker is comfortable with the situation represented by ~P. Consequently the conditional clause must be of the indicative sort, as it is in the ‘if not’-conditional of (49). It cannot introduce the uncertainty that is an encoded part of any ‘unless’-clause. From an utterance of (49) we understand that the window referred to is currently closed, and the question is whether it should remain closed or be opened. Unless in (50), however, instructs the addressee to focus simultaneously on consequences of ~P and consequences of P in the pragmatic processing of the utterance. Due to the fact that the producer of an ‘unless’-clause admits that she is uncertain about the truth value of the protasis, neither P nor ~P can serve as antecedent for the pronoun it in (50). This token A relevance-theoretic analysis of UNLESS 79 of it cannot be processed as a cataphor, but it could conceivably be processed as an anaphor whose intended antecedent is located somewhere in the immediately preceding discourse. For example, an utterance of (50) would be in order if the speaker wishes to communicate that it is all right for her to sit close to the window, on condition that the window is not opened. A pair like (51)-(52) illustrates the same constraint. The speaker of (51) uses her negative conditional clause to create a context which provides a propositional referent for the demonstrative pronoun that in the main clause. That is a resumptive copy of the descriptive content of the clause that falls within the scope of the conditional operator if. In (52) the demonstrative that cannot corefer with the protasis of the ‘unless’-clause or with its contradiction; it must corefer with something mentioned in the discourse that occurred prior to (52), an exact parallel to the problem of resolving the reference of the pronoun it in (50). (51) [If you don’t know Brussels well]i, thati is no problem. (52) [Unless you know Brussels well]i, thatj/*i is no problem. An utterance of (52) gives us the impression that it would be disadvantageous if the hearer does know Brussels well. Intimate knowledge of the city of Brussels is said to create a problem pertaining to the referent of that, whose antecedent cannot be located within the conditional construction. If the antecedent-anaphor relation is as indicated by the subscripts in (51), it makes no sense to instruct the hearer to focus on the contradiction of the protasis ~P in the pragmatic processing for relevance. Our understanding of the main clause predication depends on the stipulated truth of ~P, which is why (51) works but (52) does not. The inability of an ‘unless’-clause to serve as the antecedent of a pronoun in the main clause of the conditional construction is a constraint that has been ignored in previous research on the semantic meaning and pragmatic implications of unless. (54), with unless, compared to (53), with if … not, is bad (on the reading ‘She never told me she did not want to go to the sauna’), and the reason is the same as for the pair of (28)-(29), repeated here, whose illformedness is mainly due to the fact that there is no comma or similar graphical device in front of the ‘unless’-clause to mark its discourse status as an afterthought that states an exception to the preceding generalization. (53) If she did not want to go to the sauna, she never told me. (54) #Unless she wanted to go to the sauna, she never told me. (28) I will be surprised if that book doesn’t sell well. (29) *I will be surprised unless that book sells well. (53) is an indicative conditional. The speaker seems to accept as true the thought that the woman did not want to go to the sauna, but the speaker also avoids committing herself to that thought. The direct object argument of the ditransitive verb tell in (53) is ostensively the proposition expressed in the preceding conditional clause: ‘If we assume that it is true that she did not want to go to the sauna, I can inform you that she never told me that she did not want to go to the sauna’. The information offered in the main clause of (53) is relevant on condition that the protasis, the negative proposition ~P, is taken to be true. If its contradictory counterpart P is true, the speech act is void. There is 80 Fretheim a communicated coreference relation between ~P and the linguistically omitted direct object complement of told in the main clause, in other words the direct object of told in (53) is a variable bound by the negative proposition of the preceding conditional clause. This intended coreference relation is made completely inaccessible by the ‘unless’- construction in (54). In order for an utterance of (54) to be interpretable, we have to imagine that there is something else, something not referred to in (54), which the woman never told the speaker. Analogously, the object of the speaker’s surprise in the passive construction in (28) is manifested as a variable bound by the following conditional clause proposition. I admit that the fact that the ‘unless’-clauses in (50), (52), (54) and (29) cannot serve as antecedents of an anaphor in the main clause will automatically be accounted for if we assume the analysis of Dancygier and Sweetser (2005), who do not locate the negation in the protasis of the ‘unless’-conditional but in the apodosis: ‘P ~Q’. We need a negative protasis, as in (49), (51), (53) and (29), in order to interpret the main clause variable as being bound by that protasis. If no negative proposition is expressed in an ‘unless’-clause, it goes without saying that there can be no bound variable in the main clause which represents the propositional content of the ‘unless’-clause. What their analysis does not account for, however, is why even a positive proposition cannot bind an overt pronominal anaphor or a zero pronominal anaphor (‘hidden indexical’) in the main clause of an ‘unless’-construction. Why is it that we cannot interpret the main clause of (54) as shown in (54’), while (53’) is evidently the correct truth-conditional interpretation of (53)? (53’) ‘If she did not want to go to the sauna, she never told me that she did not want to go to the sauna.’ (54’) * ‘Unless she wanted to go to the sauna, she never told me that she wanted to go to the sauna.’ (53’) is an indicative conditional, but an ‘unless’-conditional is never indicative. You cannot stipulate the truth of ~P and simultaneously encourage your interlocutor to work out what would follow from a true contradictory counterpart P (cf. the procedural constraint on appropriate use of the connective unless stated in (20) in section 3). There are good reasons why Dancygier and Sweetser’s analysis of ‘unless’- conditionals must be rejected, even if their account seems to predict, correctly, that no negative proposition in the conditional clause of an ‘unless’-construction can function as the antecedent of an anaphoric argument in the accompanying main clause (because there is no such negative proposition there). An utterance of (53) means (53’) because the negative protasis P invites the hearer to process the following apodosis in a context in which it is factually true that the woman did not want to go to the sauna. The ‘unless’-clause of (54’) expresses exactly the same premise, in fact, because unless lexically incorporates both ‘if’ and ‘not’. Observe that (54’) is anomalous just like *If she did not want to go to the sauna, she never told me that she wanted to go to the sauna. On the other hand, (55) is anomalous because of the encoded instruction “Let your inferential processing be guided by the assumption that P may be true” stated in (20). * (55) Unless she wanted to go to the sauna, she never told me that she did not want to go to the sauna. A relevance-theoretic analysis of UNLESS 81 The encoded non-truth-conditional constraint on unless, expressed in (20), rules out the possibility that (54) could be interpreted just like (53), and likewise that (52) could be interpreted just like (51). An utterance of Unless you know Brussels well, you need a map of the city is truth-conditionally equivalent to If you do not know Brussels well, you need a map of the city. Either conditional is compatible with the speaker’s belief that the addressee may be familiar enough with Brussels. However, the procedural meaning of unless blocks a pragmatic interpretation of the main clause of (52) as ‘it is not a problem that you do not know Brussels well’, which happens to be the correct interpretation of the main clause of (51). 4.6 There are no unless-clauses in counterfactive conditionals Geis (1973) observed that unless is impossible in counterfactive conditionals. He illustrated this grammatical constraint with the pair of (56)-(57). (56) If you hadn’t helped me, I would never have been able to finish on time. (57) *Unless you’d helped me, I would never have been able to finish on time. My explanation for the ungrammaticality of (57) is that a counterfactive conditional like the one in (56) presupposes the falsity of the protasis, that is, an utterance of (56) presupposes that the speaker’s interlocutor helped the speaker, and the communicated result of this is that the speaker managed to complete some work on time. While the producer of a counterfactive conditional with a negative protasis ~P commits herself to its contradiction P, the linguistic function that distinguishes an ‘unless’-clause from an ‘if not’-clause is the encoded information that the addressee should attend to the possibility that P is true is his pragmatic processing of the conditional. This conflict between the meaning of a counterfactive conditional and the meaning of unless, I would claim, is why there is no place for unless in a counterfactive conditional. Notice that the ‘unless’-clause in (58) is in order. In spite of a syntactic similarity and an apparent semantic similarity between (57) and (58), the syntactic form of the apodosis in the latter reveals that this is not a counterfactive conditional. (58) I don’t think I’ll be able to finish on time – unless you’d helped me, of course. The conditional clause introduced with the connective unless in (58) communicates the speaker’s hope that the interlocutor might consider giving her a helping hand. Unless causes the addressee to summon up an alternative to the factual situation, one that the speaker finds desirable. The use of the past perfect had helped must here be explained as a mitigating device that makes the request more indirect than if the form of the afterthought conditional clause had been unless you help me, with the present tense of the verb. 5. Unless-clauses with a negative complement Thus far I have assumed that the syntactic complement of the connective unless expresses a positive proposition, not a negative one, and previous analyses of ‘unless’- conditionals have taken it for granted that this is a true assumption. Sure enough, a 82 Fretheim negative complement of unless means double negation and can therefore be assumed to be rare for performance reasons, but such structures do occasionally occur. Take a look at (59). (59) Unless you don’t remember anything from that night, you have nothing to worry about. Just tell them exactly what happened. The complement of unless in (59) contains the NPI anything, but there is also an overt negation operator there, and the latter is what licenses the appearance of the NPI in the ‘unless’-clause. I am defending an analysis of unless as a word that instructs the addressee to consider the possibility (or even plausibility) that its syntactic complement expresses a true proposition. My analysis explains why (59) and (60) have the same truth-conditional content but differ semantically with respect to the speaker’s attitude to the conditional clause proposition. (60) If it is not so that you don’t remember anything from that night, you have nothing to worry about. Just tell them exactly what happened. An utterance of (60) could be an echoic metarepresentation (Sperber, 2000; Wilson, 2000; Noh, 2000) of someone’s denial that he did not remember anything from that night. In contrast, the pragmatic function of the conditional clause with unless in (59) cannot be to echo the interlocutor’s claim that it is false that the referent did not remember anything, because unless is never an option when the speaker uses a conditional clause to metarepresent someone’s commitment to ~P. There is nothing about (60) that prevents us from interpreting it in exactly the same way as (59), if there is contextual support for that interpretation; conversely, due to the non-truth-conditional lexical meaning of unless it is impossible to interpret the conditional expressed in (59) as an indicative conditional with an echoic conditional clause. The double negation (~ ~P P) in (59) implies that the interlocutor has nothing to worry about if he does remember what happened that night (and makes no secret of what happened), so in (59) it is the truth of the negative proposition ~P derived from the complement of unless, which warrants the conclusion that the apodosis may unfortunately not hold. The fact that there may be a negative proposition in the complement of unless means that the definition of the procedural meaning of unless proposed in (20) should be modified slightly, but what is needed is just a tiny simplification, as shown in (20’): (20’) The procedural meaning of unless: Associate the complement of unless with a proposition that contradicts the negative protasis of the conditional, and let your inferential processing be guided by the assumption that this proposition may be true. 6. Unless, but not except if, is possible in ‘speech-act conditionals’ This section offers additional evidence that unless works differently than except if. This time the grammar-based restriction affects clauses introduced by except if but not clauses introduced by unless. The negative protasis of an ‘unless’-clause does not have A relevance-theoretic analysis of UNLESS 83 to be logically connected to the preceding apodosis, but the proposition expressed in an ‘except if’-clause must be so connected. So-called ‘speech-act conditionals’ (the term is due to Sweetser, 1990) are “cases where the ‘if’-clause appears to conditionally modify not the contents of the main clause, but the speech act which the main clause carries out” (Dancygier and Sweetser, 2005: 113). The speaker of (61) uses the afterthought ‘unless’-clause to help the hearer activate a contextual assumption (P), which, if true, would make the information in the preceding declarative irrelevant. (61) There’s a seat here, unless you prefer to stand. The ‘unless’-clause in (61) does not express an exception to anything conveyed in the preceding main clause. In fact, (62) with except if deviates markedly from normal English usage, because it appears to tell us that no seat is available if the hearer prefers to stand, a thought which contradicts most people’s experience with causes and consequences. (62) #There’s a seat here, except if you prefer to stand. The trouble with an utterance like (62) is that an embedded clause whose connective is except if forces us to construe the conditional as a biconditional: the truth of the apodosis depends on the truth of the protasis and the falsity of the apodosis depends on the falsity of the protasis. Hearing (61), with unless, we have no inclination to adopt a biconditional interpretation, because (61) is a pseudo-conditional at the level of propositional content. The truth of its main clause proposition in no way depends on the truth of the following conditional clause proposition. An utterance of (61) gains relevance on condition that the utterance is interpreted as a speech-act conditional which allows the hearer to pragmatically derive the meaning ‘Unless you prefer to stand (~P), I feel I should inform you that there is a free seat right here’. The implicature evoked by the non-truth-conditional meaning of unless is ‘If you prefer to stand (P), please ignore my information about the free seat’.32 The function of the unless-clause in (61), then, is to direct the hearer’s attention to a context which includes the assumption that the hearer has no need for a seat, because he prefers to stand. (63) is an analogous example, and again the corresponding ‘except if’-construction in (64) is bad because it does not permit the requisite ‘speech- act’ interpretation of the conditional. (63) There are some leftovers in the fridge, unless you prefer meat. (64) #There are some leftovers in the fridge, except if you prefer meat. 32 Speech-act conditionals, like (61), are also commonly referred to as ‘Austinian conditionals’ (cf. Austin, 1961), and Geis and Lycan (2001) call them ‘nonconditional conditionals’. They offer the following interesting example of a speech-act conditional, produced by actress Deborah Kerr in the movie From Here to Eternity, which the interlocutor, played by Burt Lancaster, pretended to comprehend as if it had been a regular material conditional, and that move is followed up in her reaction to his question: Deborah Kerr: If you’re looking for the captain, he isn’t here. Burt Lancaster: And if I’m not looking for the captain? Deborah Kerr: He still isn’t here. Observe that it would have been impossible for the man to say *And unless I’m looking for the captain?, and the reason for this is the general ban on unless in polar questions discussed in §4.3. 84 Fretheim There is a similar difference in acceptability in French between (65), whose conditional clause starts with the connective à moins que (‘unless’), and (66), where the conditional connective is sauf si (‘except if’).33 (65) Il y a un restant dans le frigo, à moins que tu préfères de la viande. (66) #Il y a un restant dans le frigo, sauf si tu préfères de la viande. The ‘speech-act’ conditional with unless in (63) and the one with à moins que in (65) will normally evoke the implicature that the leftovers are not meat, so the idea – as always in ‘speech-act’ conditionals – is that the speaker acknowledges that the information offered in the preceding sentence may be irrelevant to the interlocutor. French sauf si does not work in a fashion similar to à moins que, because just like except if, it causes the addressee to interpret the conditional as a biconditional. The verb prefer actually sounds a bit misplaced when there is syntactic evidence that it is included in the scope of negation. She preferred not to stand upright for a long time, where the negator not is located in the infinitival complement, sounds more natural than She didn’t prefer to stand upright for a long time, the latter sounding like ‘echoic’ negation (Carston, 2002), also known as ‘metalinguistic’ negation (Horn, 1989), because the speaker’s negation seems to target the interlocutor’s previous use of the verb prefer, which is echoed in the denial. (63) is one of those extremely rare case where unless is acceptable but if not deviates from normal usage. ? (63’) There are some leftovers in the fridge, if you don’t prefer meat. The contextualizing ‘if’-clause (or ‘unless’-clause) in a speech-act conditional does not contribute to the proposition expressed, which in the case of (63) or (65) covers no more than the main clause information that there are some leftovers in the refrigerator. An ‘except if’-clause, however, is always a truth-conditional modification of the explicit content of the conditional construction. 7. Conclusion Contrary to the dominant view of unless as a word that establishes an ‘exceptive’ context, scenario, or mental space, I maintain that the truth-conditional meaning of unless is just like the meaning of if not, but unless encodes information of a non-truth- conditional sort, which can be captured in terms of the relevance-theoretic notion of ‘procedural semantics’. Unless instructs the hearer to engage in a parallel processing of the conditional, in which attention is to be paid not only to the explicit negative protasis ~P but also to pragmatic consequences of an alternative context in which the contradictory counterpart P is true. A pragmatic processing which enjoins the hearer to look for not only implications of a true ~P but also implications of a true P might be believed to increase the hearer’s processing effort, leading in the worst case to gratuitous processing that reduces the relevance of the utterance, but as the producer of an ostensively communicated message guarantees the optimal relevance of her utterance 33 The French intuitions are Jonathan Brindle’s, and the translations of (63) and (64) into French are his. A relevance-theoretic analysis of UNLESS 85 (Sperber and Wilson, 1995), the tacit assumption is that the hearer will be rewarded – through added contextual effects – for the added mental effort. Quite often the presence of unless in a conditional construction will cause the hearer to strengthen the explicit truth-conditional meaning to a biconditional, and it is presumably the frequency of such cases that has made the ‘exceptive’ analysis of unless so popular. However, an explicature based on a biconditional interpretation is not always intended, and in speech-act conditionals, exemplified by (61), (63) and (65), the hearer will definitely not be tempted to read ‘if’ as ‘if and only if’ because of the presence of unless. Similarly, an utterance of (21) in section 3 does not suggest that we are to derive the information that the addressee is obliged to call the speaker if he happens to be in the utmost need; rather, we understand that he is not forbidden to call the speaker if that condition is satisfied. The non-truth-conditional lexical meaning of unless proposed in (20)/(20’) correctly filters out all known types of conditionals in which unless may not be used, and it gives a correct account of the semantic contribution of unless to the pragmatic interpretation of those ‘unless’-conditionals which are relevant in a context. To the best of my knowledge, no alternative account of the meaning and use of unless tackles the whole range of restrictions discussed in section 4. Acknowledgements A version of this paper was presented at the annual meeting of the LAGB in Newcastle, August-September 2006, and an earlier version was presented at the MONS meeting in Bergen, November 2005. I would like to thank the audiences in Newcastle – especially Billy Clark, Noel Burton-Roberts, Ivan García-Alvarez and Kasia Jaszczolt – and Bergen – especially Bergljot Behrens, Cathrine Fabricius Hansen and Peter Svenonius – for their constructive feedback. Thanks also to Randi Alice Nilsen, Lena Kildal Skaalbones and Jan Krogstad in Trondheim. References Amfo, Nana Aba Appiah, 2004. Modal marking in Akan: The case of anka. Journal of Pragmatics 37: 997-1013. Andersen, Gisle and Thorstein Fretheim (eds.), 2000. Pragmatics Markers and Propositional Attitude. Amsterdam: John Benjamins. Austin, J.L., 1961. Ifs and cans. In J.O. Urmson & G.J. Warnock (eds.), Philosophical Papers. Oxford: Oxford University Press, 153-80. Bayiga, Florence T., 2005. Communicating causal relations in Luganda. Unpublished MPhil thesis, Norwegian University of Science and Technology. Bezuidenhout, Anne, 2002. Truth-conditional pragmatics. Philosophical Perspectives, 16, Language and Mind, 105-34. 86 Fretheim Bezuidenhout, Anne, 2003. Procedural meaning and the semantics/pragmatics interface. In Claudia Bianchi (ed.), The Semantics/Pragmatics Distinction. Stanford: CSLI Publications. Blakemore, Diane, 1987. Semantic Constraints on Relevance. Oxford: Blackwell. Blakemore, Diane, 1992. Understanding Utterances: An Introduction to Pragmatics. Oxford: Blackwell. Blakemore, Diane, 2002. Relevance and Linguistic Meaning: The Semantics and Pragmatics of Discourse Connectives. Cambridge: Cambridge University Press. Blass, Regina, 1990. Relevance Relations in Discourse: A Study with Special Reference to Sissala. Cambridge: Cambridge University Press. Cappelen, Herman and Ernie Lepore, 2005. Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism. Oxford: Blackwell. Carston, Robyn, 1996. Metalinguistic negation and echoic use. Journal of Pragmatics 25: 309-30. Carston, Robyn, 2002. Thoughts and Utterances: The Pragmatics of Explicit Communication. Oxford: Blackwell. Dancygier, Barbara, 1985. If, unless, and their Polish equivalents. Papers and Studies in Contrastive Linguistics 20: 65-72. Dancygier, Barbara, 1998. Conditionals and Prediction: Time, Knowledge and Causaion in Conditional Construction. Cambridge: Cambridge University Press. Dancygier, Barbara and Eve Sweetser, 2005. Mental Spaces in Grammar: Conditional Constructions. Cambridge: Cambridge University Press. Declerck, Renaat and Susan Reed, 2000. The semantics and pragmatics of unless. English Language and Linguistics 4 (2): 205-41. Declerck, Renaat and Susan Reed, 2001. Conditionals: A Comprehensive Empirical Analysis. Berlin: Mouton de Gruyter. Ducrot, Oswald, 1982. La notion de sujet parlant. Recherches sur la philosophie et le langage. Université de Grenoble, 65-93. Ducrot, Oswald, 1984. Le dire et le dit. Paris: Minuit. Fintel, Kai von, 1991. Exceptive conditionals: The meaning of unless. NELS 22: 135- 48. Fløttum, Kjersti, Trine Dahl and Torodd Kinn, 2006. Academic Voices. Amsterdam: John Benjamins. Fretheim, Thorstein, 2000a. Procedural encoding of propositional attitude in Norwegian conditional clauses. In Gisle Andersen and Thorstein Fretheim, 2000, 53-84. Fretheim, Thorstein, 2000b. Constraining explicit and implicit content by means of a Norwegian scalar particle. Nordic Journal of Linguistics 23: 115-62. Fretheim, Thorstein, 2001. In defence of monosemy. In T. Énikő Németh and Károly Bibok (eds.), Pragmatics and the Flexibility of Word Meaning. Amsterdam: Elsevier Science, 79-115. Fretheim, 2006. English then and Norwegian da/så compared: a Relevance-theoretic account. Nordic Journal of Linguistics 29: 45-93. Fretheim, Thorstein, Stella A. Boateng and Ildikó Vaskó, 2002. Then - adverbial pro- form or inference particle? A comparative study of English, Ewe, Hungarian and Norwegian. In K.M. Jaszczolt and Ken Turner, Meaning Through Language Contrast. Volume 2. Amsterdam: John Benjamins, 51-74. Geis, Michael L., 1973. ‘If’ and ‘unless’. In Braj Kachru, Robert B. Lees, Yakov Malkiel, Angelina Pietrangeli and Sol Saporta (eds.), Issues in Linguistics: A relevance-theoretic analysis of UNLESS 87 Papers in Honor of Henry and Renée Kahane. Chicago: Chicago University Press, 231-53. Geis, Michael L. and Arnold Zwicky, 1971. On invited inferences. Linguistic Inquiry 2: 561-6. Geis, Michael L. and William G. Lycan, 2001. Nonconditional conditionals. Appendix to William G. Lycan, Real Conditionals. Oxford: Clarendon Press, 184-205. Gibbs, Raymond W. and Jessica Moise, 1997. Pragmatics in understanding what is said. Cognition 62: 51-74. Horn, Laurence R., 1989. A Natural History of Negation. Chicago: Chicago University Press. Newman, Paul, 2000. The Hausa Language: An Encyclopedic Reference Grammar. New Haven: Yale University Press. Ifantidou, Elly, 2001. Evidentials and Relevance. Amsterdam: John Benjamins. Nølke, Henning, 2006. The semantics of polyphony (and the pragmatics of realization). Unpublished paper, University of Aarhus. Noh, Eun-Ju, 2000. Metarepresentation: A Relevance-theoretic Approach. Amsterdam: John Benjamins. Recanati, François, 1993. Direct Reference: From Language to Thought. Oxford: Blackwell. Recanati, François, 2004. Literal Meaning. Cambridge: Cambridge University Press. Sperber, Dan, 2000. Metarepresentations: A Multidisciplinary Perspective. Oxford: Oxford University Press. Sperber, Dan and Deirdre Wilson, 1986. Relevance: Communication and Cognition. Oxford: Blackwell. Sperber, Dan and Deirdre Wilson, 1995. Relevance: Communication and Cognition. Oxford: Blackwell. 2nd edition with a Postface. Sweetser, Eve, 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Traugott, Elizabeth C., 1997. Unless and but conditionals: A historical perspective. In Angelika Athanasiadou and René Dirven (eds.), On Conditionals Again. Amsterdam: John Benjamins, 169-90. Traugott, Elizabeth C. and Richard B. Dasher, 2005. Regularity in Semantic Change. Cambridge: Cambridge University Press. Unger, Christoph, 1996. The scope of discourse connectives: implications for discourse organization. Journal of Linguistics 32: 403-38. Wilson, Deirdre, 2000. Metarepresentation in linguistic communication. In Dan Sperber (ed.), 2000, 411-448. Wilson, Deirdre & Dan Sperber, 1993. Linguistic form and relevance. Lingua 90: 1-25. Wilson, Deirdre & Dan Sperber, 2004. Relevance theory. In Laurence R. Horn & Gregory Ward (eds.), The Handbook of Pragmatics. Oxford: Blackwell, 607-32. Author’s e-mail address: thorstein.fretheim@hf.ntnu.no 88 WORKING PAPERS ISK 3/2006 89-103 ISK, NTNU Explaining connections in Akan discourse: the role of discourse markers Nana Aba Appiah Amfo 1. Introduction Relevance theory (Sperber and Wilson 1986, 1995) views utterance interpretation as crucially involving the recovery of intended cognitive, or contextual, effects. These contextual effects could come in one of three forms. First, they could strengthen an existing assumption. Second, a communicated assumption could combine with an already existing one to yield a contextual implication. And finally an existing assumption could be contradicted, which leads to elimination of the weaker assumption. Consider the Akan sentences in (1) to (3). (1) A$be@na@ no$a@-a$ a$du$a$ne@. ç$-si$-i$ n$ne@E@ma@ n@so@. 1 Abena cook-COMPL food. she-wash-COMPL things also “Abena cooked. She did the laundry as well.” (AS) (2) Me$ di$di@-i@ n$tE@m@ na$ me$ pu@e@-e$E@. I eat-COMPL quickly CONJ I go.out-COMPL “I ate quickly and I went out.” (AS) (3) E$kç@m@ de$ me$ na$n@so@ me$-m$-pE@ sE@ me$ di@di@. Hunger have me but I-NEG-like COMP I eat “I am hungry but I don’t want to eat.” (AS) The propositions expressed in (1) are that ‘Abena cooked. Abena did the laundry’. The inclusion of nso in the second sentence suggests that that sentence is to be in interpreted in a similar context as the preceding one. Assuming the addressee derives the assumption that Abena is domestically inclined from the first utterance, that assumption will be strengthened by the time he has finished interpreting the second utterance. This is because nso encourages the addressee to interpret the second utterance within a similar context as the first. The connective na in (2) combined with other contextual information leads the addressee to the inferred information that the speaker ate before 1 The examples in this paper are from the three major dialects, Asante Twi, Akuapem Twi and Fante. The Akuapem Twi and Fante examples are attested forms from published Akan texts. The examples are marked as AS (Asante), AK (Akuapem) and FA (Fante) respectively at the end of the English translations. They are then followed by the initials of the writer (in lower case) if they are taken from a text, and the page number. Tones have been marked for the examples: $ for low tone, @ for high tone, and ! @ for downstepped high. The abbreviations used are as follows: CM=Conditional marker; COMP=Complementizer; COMPL=Completive aspect; CONJ=Coordinative connective; CTM=Contrastive/Emphatic topic marker; DCM=Dependent clause marker; DEF=Definite article; HAB=Habitual aspect; NEG=Negation affix; PERF=Perfect aspect; PL=Plural affix; POSS=Possessive morpheme; PRES=Present tense; REL=Relative clause marker. For ease of presentation, I refer to the speaker as she and the addressee as he. 90 Amfo she left the contextually determined geographical domain. According to both relevance theory (Carston 2002) and contextualist philosophers of language (Recanati 2004), the information about the temporal order of the events is not implicated but is included in the proposition expressed. The usual expectation is that anyone who is hungry would like to eat, but in (3) the use of nanso suggests that such an expectation ought to be dropped in favor of the information provided in the nanso-clause – that the speaker does not want to eat nevertheless. The derivation of such contextual effects, which are crucial in the interpretation process, are obtained within an appropriate context. Context, as defined within the relevance-theoretic framework, is not confined to the co-text. It is a psychological construct involving a subset of the addressee’s assumptions about the world. As illustrated in examples (1)-(3), a particular utterance may contain certain linguistic items which help in making connections between different parts of a discourse or of a single utterance, by triggering certain contextual assumptions which form part of the context within which an utterance is interpreted. This paper focuses on the use of three discourse markers in Akan, a Niger Congo (Kwa branch) language. They are (n)so, na and nanso, illustrated in (1), (2) and (3) respectively. (N)so is one of a number of focus markers in Akan. Na is the mundane clausal coordinating connective in Akan and nanso is the contrastive connective. The latter’s use suggests that an earlier held assumption ought to be abandoned and further eliminated in favor of the one which has been introduced by nanso. These markers make distinct contributions to the interpretation process by highlighting certain connections which need to be made between parts of an utterance or the discourse as a whole. Although this paper is primarily about the selected Akan markers, some comparative references will be made to their close English equivalents. The paper demonstrates that in spite of language specific grammatical and cultural features which may consequently lead to certain idiosyncratic pragmatic implications, a general cognitively based theory of communication such as Relevance theory is sufficient to account for the use of discourse markers cross-linguistically. At the same time, it draws attention to certain language internal details which ought not to be lost sight of in the pursuit of a general theory of utterance interpretation. Section 2 gives a brief relevance-theoretic account of how an addressee links various utterances and parts of a discourse. Section 3 considers the discourse function of the connective (n)so. It is noted that a (n)so-utterance shares an entailment with another utterance in the immediately preceding discourse. Thus the (n)so-utterance, like an English too/also-utterance, introduces additional evidence in favor of an earlier assumption, which has been derived from one of the immediately preceding utterances. In spite of this similarity of function between English too/also and Akan (n)so, it is observed that the context within which (n)so may be felicitously used is much wider than that of too/also. Section 4 considers the various inferential relations signaled by the use of the mundane Akan clausal coordinating connective na. It highlights the fact that an explanatory inferential relation, where the second conjunct is the explanation, is permissible. The theoretical consequences of this observation are considered against the backdrop that such a relation is prohibited with regard to and, the English equivalent of na. Section 5 concentrates on the contrastive connective nanso, which can aptly be considered as the Akan counterpart of English but. It gives an indication that the proposition expressed in the utterance which follows nanso is contrary to what is to be expected. Its role as a contrastive marker is compared with na, the ordinary clausal Explaining connections in Akan discourse: the role of discourse markers .91 connective when the use of the latter induces a contrastive interpretation between the conjuncts. Section 6 is the conclusion. 2. Explaining Connections in Discourse Blakemore (1992) suggests that connections between parts of a discourse can be made either by intonation or by the use of discourse connectives or markers. The use of these markers practically ensures the correct context selection, which subsequently leads to retrieval of the intended contextual effects. In other words, their use facilitates the achievement of maximum contextual effects for minimum processing effort thereby making the utterance optimally relevant. Blakemore (1992, 2002) examines the role of discourse markers as constraints on the interpretation process by exemplifying almost exclusively from the English language. She suggests that discourse connectives can be classified into three categories, parallel to the kind of contextual effects that can be achieved by processing an ostensive verbal stimulus. 2 Connectives such as after all, besides, moreover, furthermore and utterance initial also introduce additional evidence in a bid to strengthen an earlier held assumption by the addressee. Others such as too and also interact with the phenomenon of focus to direct the addressee to derivation of contextual effects which are parallel to some already derived contextual effects, possibly derived from the immediately preceding discourse. Connectives which direct the addressee to drop an earlier held assumption in favor of the one which is just about to be introduced include however, nevertheless and but. Akan speakers, like English speakers, utilize discourse markers as a way of drawing the addressee’s attention to certain linkages he ought to make in the process of utterance interpretation. The relevance-theoretic account of utterance interpretation involves two distinct processes: decoding and inference. The outcome of the decoding process is an input to the inferential process. With regard to the latter process, the addressee constructs conceptual representations which enter into inferential computations. One claim made within Relevance theory is that human languages not only possess linguistic forms which provide inputs to conceptual representations, but they also contain words and expressions which provide information as to how, or in which direction the computation of the conceptual representations are expected to proceed. The latter group of words and expressions are said to encode procedural information, rather than conceptual information. This conceptual versus procedural distinction has been championed by Blakemore (1987, 1992, 2002) and supported by several other studies (Blass 1990, Jucker 1993, Clark 1993, Unger 1996, inter alia). The Akan markers which are of concern in this paper fall into the category of linguistic expressions which encode procedural meaning. That is, they provide information as to what contextual assumptions should guide the addressee in his inferential manipulation of the decoded logical form of an utterance and consequent computation of contextual effects. The procedural meaning they encode can be characterized as follows: (N)so: process the (n)so-utterance within a parallel context provided by the immediately preceding utterance(s). 2 Blakemore (1992) admits that this classification is not exhaustive and can be added on. 92 Amfo Na: look out for some form of contextually determined inferential relation between the conjunct propositions. Nanso: drop an earlier held assumption in favor of the following one. Equipped with such information, an Akan addressee is able make the necessary connections that ought to be made between particular utterances and the larger discourse. Let’s turn our attention to the individual markers. I begin with (n)so. 3. (N)so: Additive Focus Marker3 (N)so is one of two inclusive focus markers in Akan, which has been classified as an additive focus marker following König (1991). 4 The presence of the additive focus marker in an utterance enjoins the addressee to look for parallelisms of context within the immediately preceding discourse. In (4), the use of so signals that there is some other game that the subject referents Araba and AkyerE participate in regularly. Indeed, in a few utterances prior to (4), we are told that the girls play together and they play the game of ampe. 5 By the inclusion of so, the speaker expects the addressee to interpret (4) within the general context that the girls play together and the specific context that they play ampe together. By so doing, the speaker encourages the addressee to take (4) as further evidence for the fact that Araba and AkyerE are regular playmates. Thus, the so- utterance provides additional evidence to support the earlier held assumption that the girls are playmates. (4) A@ra@ba@ na A$kye$rE@ to$w$ m$-ba@ so@. 6 Araba CONJ AkyerE throw.HAB PL-seed also “Araba and AkyerE throw marble as well.” (FA/skc 9) With regard to (5), the assumption introduced at the beginning of the paragraph (see Appendix A) is that Maame Esi is a neat person. The writer goes on to give specific instances which support the assertion made in the opening sentence, including the information in (5). (5) Me@me@n@da@ bi@a@ra@ so@ ç@-ho$r$ m$-bo$fr@a@ no@ hç$n@ n@-ta@r@ na$ ç@-to$w$ do@. Saturday every also she-wash.HAB PL-child DEF POSS PL-dress CONJ she-iron top “Every Saturday, in addition, she washes the children’s clothes and irons them.” (FA/skc 8) In the final sentence of the paragraph, the writer suggests, by the use of so, that that particular sentence ought to be interpreted within a similar context as the immediately preceding ones. Pertaining to the cleanliness of the children, there are other things that 3 Nso is represented as so in the Fante dialect. Since most of my examples in this section are from the Fante dialect, I will be using these two forms interchangeably, or will write (n)so. 4 See Amfo (2006) and the references therein for a detailed discussion of (n)so and other markers of information structure in Akan. 5 Ampe is a game in which children jump and clap their hands simultaneously and throw their legs. It used to be the preserve of girls but that is fast changing. 6 Tow mba (to throw marble) is used here in reference to a particular game for children. Explaining connections in Akan discourse: the role of discourse markers .93 Maame Esi (who is the referent of the pronominal prefix ç-) does regularly, on the other days of the week. She, for instance, gives them daily baths. And on Saturdays, in addition to what has already been mentioned, she makes sure their clothes are clean as well. Indeed, in example (6), which forms part of the preceding discourse, it is stated that Maame Esi bathes the children. (6) Ma$a$me@ E@si@ gu$a@r@ Ko$dwo@ na$ A@ra@ba@. Maame Esi bath.HAB Kodwo CONJ Araba “Maame Esi bathes Kodwo and Araba” (FA/skc8) Even though it is not information encoded in this particular sentence or any other in the paragraph that contains (5) and (6), it can be deduced from the co-text and other contextual information that she does so every day. Thus washing and ironing the children’s clothes on Saturdays is something she does in addition to the daily baths she gives them and other daily obligations towards them, including brushing their teeth. So indicates that there is some aspect of the context of the previous utterance which needs to be carried over in the interpretation process of the so-utterance. So, like all other focus markers in Akan, is placed after the constituent it takes scope over. When it occurs in utterance final position, its scope is underdetermined, though it will invariably include the final syntactic constituent. It could either be over a phrasal constituent (for instance the grammatical object of the sentence) or over the whole utterance. In (5), the scope of so is over the adverbial phrase Memenda biara (every Saturday). Even though it is not stated explicitly that Maame Esi bathes her children and brushes their teeth every day, this can be easily deduced, especially since such personal cleanliness tasks are daily routines. Once so has been used, the addressee turns to the immediately preceding set of utterances to look for a similar context within which to interpret the so-utterance. In (5), what the writer communicates by modifying ‘every Saturday’ with so, is that Maame Esi goes through this type of daily routine with her children every day of the week, and then in addition on one particular day of the week, Saturday, there are other duties that she performs for their benefit, namely washing and ironing their clothes. Indeed the modification of Memenda biara (every Saturday) by so aids in deriving the contextual information that the personal cleanliness routines are performed daily. Since the reader expects that the writer aims at optimal relevance by his use of so, and since no other day of the week has been explicitly mentioned as a the day on which Maame Esi performs those duties for her children, the reader, justifiably, and with the aid of other contextual information derives the assumption that on every day of the week Maame Esi is busy seeing to it that her children are clean. The parallel context justifying the use of so in (5) is expressed in (7). (7) Maame Esi makes sure her children appear clean every day. Indeed, in Akan, the so-modified constituent does not have to be the information focus of the utterance, in order for the addressee to be guided to the parallel context required for the interpretation of the utterance in question. The situation described above can be compared with the one presented in the exchange in (8). 94 Amfo (8) A: Da@bE$n$ na$ i@-bo@-tu@m@ a@-kç@ gya$ me$ wç$ N$kwa$n$ta@? When FM you-FUT-be.able.to CONS-go leave me at Nkwanta “When can you come along with me to Nkwanta?” B: Me@ kç$ sku@u$l$ da@bi@a@ra@. Me@me@n@da@ so@, me@-kç$ gu@a@m$. I go school every.day. Saturday also, I-go market “I go to school every day. In addition, on Saturdays, I go to the market. ” (FA) In interpreting B’s utterance, A will enrich dabiara (every day) as every week day. The assumption derived from the first sentence of B’s utterance is that ‘she is busy (on weekdays)’. By modifying Memenda (Saturday) with so, B expects to draw A’s attention to the previously derived assumption that she (A) is busy on weekdays. In other words, she is not available on weekdays and not on Saturdays either. Unlike the situation with English too or also, the context of a (n)so-utterance and that of the immediately preceding utterance do not have to be identical. Some similarity of contexts is sufficient. The kind of parallel context which permits the use of English too/also is more constrained than that of a (n)so-utterance. Consider examples (9) and (10) which are consecutive sentences from an Akan text. (9) Ko$dwo@ e$-dzi$ m$-fe@ du@-e@si$a@. O$-e@-wi@e@ sku@u$l$. Kodwo PERF-eat PL-year ten-six. He-PERF-finish school “Kodwo is sixteen years old. He has completed school.” (FA/skc 12) (10) A@ra@ba@ so@ e@-dzi$ m$-fe@ du@-e@bi$a@sa@. ç$-a@-kç$ m$fi$kyi@r@. Araba also PERF-eat PL-year ten-three. she-PERF-go backyard “Araba, likewise, is thirteen years old. She has menstruated.” (FA/skc 12) The first sentence in the paragraph which contains (9) and (10) states that Kodwo and Araba have grown (see Appendix B). The reader had been introduced to them in an earlier chapter as small children. The writer then goes on to mention Kodwo’s age and what stage he is in his life. Even though Araba is not the same age as Kodwo and is not at exactly the same stage in life as Kodwo, the fact that (10) is an update about Araba’s life, just as (9) is about Kodwo’s life is enough of a parallel context to warrant the use of so. Example (10) is to be taken as additional evidence for the assumption provided in the sentence preceding (9) – that Kodwo and Araba have grown; they are not the little children they used to be. English also/too cannot be used felicitously in such a context. ‘Araba, likewise, is thirteen years old’ coming after the English translation in (9) is strictly speaking unidiomatic, but it is probably the only way to translate the Akan sentence while preserving the communicative function of so. So in (10) directs the addressee to an implicature derived on the basis of (9) and (10), that Kodwo and Araba are both in their teenage years now. Kodwo is sixteen years and has completed basic school, thus he has begun a new stage in his life. Even though Araba is three years younger and is still in school, she has menstruated, and that can be seen as the beginning of her adult life. This is in contrast to when the reader was first introduced to them as small children. Explaining connections in Akan discourse: the role of discourse markers .95 Example (11) is a slightly modified form of (1). In (11), nso takes scope over the second conjunct clause çsii nneEma (she did the laundry), rather than just the constituent referring to the laundry. (11) A$be@na@ no$a@-a$ a$du$a$ne@ na$ ç$-s$i-$i n$ne@Em @ a@ n@so@. Abena cook-COMPL food CONJ she-wash-COMPL things also “Abena cooked and did the laundry as well.” (AS) The proposition expressed in the second conjunct of (11) is to be interpreted against the background that the referent of the third person pronoun, Abena, has already performed some domestic chores, namely cooking. The reason why a speaker uses nso to encourage the kind of parallelism indicated by nso, to be taken into account in the addressee’s inferential processing, will be contextually determined. It may be the speaker’s intention, for instance, to encourage the addressee to derive an implicature such as ‘Abena is tired and cannot perform any more domestic chores’. Notice that the shared context for the conjuncts in (11) is more constrained than the situation we found in (10). The shared context for the two conjuncts in (11) is (12); the subject referent for both conjuncts is the same and the activities she engaged in are confined to the home, and are most likely to have been performed on the same day. On the other hand, (9) and (10) have different subject referents, who are of different gender and ages and at different specific stages in their lives. (12) Abena did house chores. Notice that even though a so-utterance may be considered as additional evidence in favor of an earlier held assumption, it cannot, like English also, “ introduce another argument, in addition to that given in the preceding text, for the same conclusion” König (1991: 65). It is not difficult to imagine a context in which the proposition expressed in (13) is expected to be taken as another argument in favor of why some wife left her husband. (13) Also, he is the domineering type. The argumentative function performed by also in (13) cannot be performed by so, rather it is ebio, one of two recurrence markers in Akan (cf. Amfo 2005), which can be used to introduce another argument in addition to one (or some) which has (have) already been mentioned. So can be used in such a context only when it combines with the temporal adverb afei. Thus (14) is an acceptable Akan translation of (13), whereas (15) is not. 7 (14) A@fe@i@ so@ ç$-yE$ n@hyE@. Now also he-be.PRES domineering (15) So@ ç$-yE$ n@hyE@. Also he-be.PRES domineering 7 (15) is a grammatical Akan sentence, however so could only be interpreted as a contracted form of nanso/naaso. The preceding conjunct proposition will as such be contextually determined. 96 Amfo The communicative function of so, illustrated in (16) is distinct from the one discussed above. The use of so in (16) does not give access to any parallel context. What it does is to communicate a speaker attitude of disdain or disapproval. The speaker of (16) in addition to communicating the proposition that Kofi is pompous, certainly succeeds in communicating the fact that she is disgusted at such an attitude displayed by Kofi. (16) Ko$fi@ so@ ç$-kye$rE@ no@ho@ pa@pa@. Kofi also he-show REFL well “Kofi is indeed pompous.” (FA) As suggested in Amfo 2006, this use of so may be considered as a case of interpretive use (cf. Wilson and Sperber 1988, Noh 2000, Blakemore 1992, Blass 1990) where the speaker metarepresents her or somebody else’s thoughts while communicating her own attitude to that thought. This function of so can be applied to whole quoted utterances. Example (17) from Amfo (op. cit.), originally taken from a text in Akuapem Twi is an illustration of the above. (17) “Te$na$ hç@ na$ wu@-be@-hu@” n@so@ de@, E$-n$-yE@ a$sE$m$pa@. “Sit there CONJ you-FUT-see” also CTM, it-NEG-be good.case ““You wait and you will see” is not a good statement.” (AK/a 3) The use of nso in (17) is an indication that the quoted phrase as used in its original context was not intended to be complementary. The speaker suggests that such as statement is not to be taken pleasantly. Indeed, this is complemented by the explicit statement “it is not a good statement’. This particular function of Akan so is unparalleled in English, even though English too may sometimes be used marginally, in colloquial or even child speech, to communicate a speaker attitude of disdain, possibly also protest as in (18). (18) I did too! 4. Na: coordinating connective Na is the ordinary clausal coordinating connective in Akan, equivalent to English and. As noted in Amfo (in press), its use in an utterance indicates to the addressee to look out for certain kinds of inferential relations between the conjuncts. The encoded meaning of na combined with certain specific grammatical features as well as other contextual information, retrieved from a variety of cognitive and perceptual sources, result in the identification of the specific inferential relation involved. These relations could be temporal, causal, parallel, contrastive or even explanatory as illustrated in examples (19) to (23) below. (19) Me$ wi@e@-e$ m’-a$dwu@ma@ no@, na$ me@ pu@e$-E@. I finish-COMPL my’-work DCM, CONJ I go.out-COMPL Literally: “When I finished my work, and then I went out.” (I went out after I had finished my work) (AS) Explaining connections in Akan discourse: the role of discourse markers .97 (20) A$dwo@a@ bç$-ç$ fa@m@ wç$ su$ku@u$ na$ o$-nya@-a$ ku@ro@. Adjoa hit-COMPL ground at school, CONJ she-get-COMPL sore. “Adjoa fell at school and (as a result) got a wound” (AS) (21) ç$-n$-ni@ a$bo$ta@re@ na$ ç$-yE$ dwE$E@. He-NEG-has patience CONJ he-be arrogant “He is impatient and he is arrogant.” (AS) (22) Na@ me$-m$-pE@ n’-a@sE@m@, na$ se$e$se@i@ de@E@ me$-a$-hu@ sE@ ç$-yE$ ni$pa$ pa@pa@. Then I-NEG-like his’-matter, CONJ now CTM I-PERF-see COMP he-be person good “I used to dislike him but now I have realized he is a good person.” (AS) (23) To$ wo$ bo@ na$ a$ha@ yE$ to$ro$. Throw your chest CONJ here be slippery “Take your time because it is slippery over here.” (AS) The presence of the connective na combines with contextual information to give the addressee access to contextual information which aids in arriving at the intended interpretation. In (19), na combines with the dependent (temporal) clause marker no to indicate a sequential ordering of events; the event described in the first conjunct temporally precedes the one described in the second conjunct. As a result of the nature of events described in the conjunct clauses in (20), the encoded meaning of na is enriched such that it communicates a cause-consequence relation between the conjuncts. In (21), the two conjuncts çnni abotare (he is impatient) and çyE dwEE (he is arrogant) may be considered as equal premises in the derivation of a single conclusion. For instance, (21) may be given as the reason why the referent of ç- is constantly at loggerheads with his work colleagues. In (22), the presence of the nasal negation morpheme in the first conjunct, combined with its absence in the second conjunct, is an overt grammatical feature which facilitates the contrastive interpretation of na in that utterance. The assertion in the second conjunct of example (23), that the place referred to by deictic use of aha (here) is slippery, is presented as an explanation or justification of why the addressee needs to comply with the advisory directive given in the first conjunct. The fact that a given token of a na-conjunction can give rise to an explanatory inferential relation between the individual conjuncts, as in (23), and specifically that the second conjunct is considered as an explanation of the proposition expressed in the first conjunct, contradicts Carston’s (2002), Blakemore’s (1992) and Blakemore and Carston’s (2005) claim that “one conjunct cannot function as an explanation for the state of affairs described in the other” (Carston 2002: 245), and especially not in such a way that the explanation appears in the final conjunct. Their claims follow from the semantic constraint that the second conjunct in an and-conjunction cannot function as an explanatory comment on the first. They support this claim with the fact that an and- conjunction is supposed to be processed as a single unit and as such it is inconceivable for the second conjunct in an and-conjunction to function as an elaboration of some information given in the first conjunct as illustrated in (24) below, taken from Carston (2002: 247). The second clause in (24a) is interpreted as an elaboration of the proposition expressed in the first clause; an interpretation which is blocked when the two utterances are conjoined with and, as in (24b), rather than juxtaposed as in (24a). 98 Amfo (24) a. I ate somewhere nice last week; I ate at MacDonald’s. b. I ate somewhere nice last week and I ate at MacDonald’s. However, as can be seen from the Akan data in (23) above, explanatory and elaboratory relations are permissible between na-conjuncts. In Akan it is irrelevant whether the explanation is the second conjunct or the first. In (25), where the first conjunct explains why the directive in the second conjunct must be complied with, the pragmatic implications are the same as in (23) where the explanation is given in the second conjunct. (25) Ah$ a@ yE$ to$ro$ na$ to$ wo$ bo@. Here be slippery CONJ throw your chest “It is slippery over here so take your time.” (AS) Do these facts about an Akan na-conjunction suggest that it is not processed as a single unit for relevance? Of course it is. In (23) for instance, the speaker believes that the addressee will readily comply with the directive to wo bo (take your time), when he is provided with an explanation of why this directive needs to be followed up. In other words, the utterance achieves optimal relevance only when the two conjuncts are interpreted as a single processing unit. 8 To wo bo (take your time) in (23) and (25) is an imperative sentence, and the grammatical constraint that for one of the conjuncts to be interpreted as an explanatory comment on the other, one of them has to be an imperative and the other a declarative needs to be taken seriously. Indeed, it is this very constraint that allows a directive- justification (or the reverse) interpretation of a na-conjunction. The grammar of Akan allows the sentence type combinations outlined in (26) to be treated as a single processing unit. (26) a. IMP na DCL b. DCL na IMP It is clear that the grammar of English does not allow such an interpretation with and- conjuncts, not even under constrained grammatical conditions. However this should not be taken as a general cognitive principle underlying the interpretation of natural language mundane coordinating connectives. It is a language internal fact which ought to be treated as such. 5. Nanso: Contrastive connective9 Nanso can be argued to consist of two morphemes, na and nso, which have possibly lexicalized over time and is now thought of as a single word rather than a sequence of two words, an assumption corroborated by the way the word is written. Na, the 8 According to Sperber and Wilson (1995), an utterance is optimally relevant when there is the best possible balance of effort against effect within an appropriate context, in other words, minimal processing effort results in maximum contextual effects. 9 The Fante equivalent of nanso is naaso. Explaining connections in Akan discourse: the role of discourse markers .99 mundane coordinating connective in Akan, can be used in a conjunction where a contrastive relation exists between the conjuncts (cf. Amfo (in press) and section 4 above). However, the underdetermined na can be used in certain contexts, where a contrastive interpretation is intended. Also, the kind of contrastive relation which may be suggested by a na-conjunction is quite distinct from that of a nanso-utterance, as I will demonstrate presently. Nso, the other component in the morphological make up of nanso is a contrastive marker, functionally distinct from the additive focus marker discussed in section 3. This use of nso is quite colloquial and it introduces an utterance which expresses a contrary proposition to one which is mutually manifest to the interactants, either through the immediately preceding discourse or via their encyclopedic memory. In (27), taken from Amfo (in press), the speaker indicates that the first conjunct proposition ‘if the man heard’ is contrary to the information which has already been reported, that the man claims he has no knowledge of the issue under discussion. (27) SE@ n@so@ pa$pa@ no@ a@-te$ na$ ç$-se@ ç$-n$-te@-e$E$ a@, ç$no$ n@so@ CM but man DEF PERF-hear CONJ he-say he-NEG-hear-COMPL CM, he also a@-bç$ m@mç@de@n@. PERF-hit effort “If however the man has heard and he is saying he has not, it is up to him.” (AS/rtr) Finally, nso may be used in colloquial speech to conjoin clauses, functioning, more or less, as a contracted from of nanso. Now, let us return to the function of the contrastive connective nanso. Its use indicates to the addressee that what follows is contrary to what may be the usual expectation. Such an expectation is usually triggered by the immediately preceding utterance. This is illustrated in (28), where the speaker expected that due to her early arrival she will meet the referents of the third person plural object pronoun, wçn. (28) Me$-du$ru$-u$ su$ku@u$ hç@ n@tE@m@, na@n@so@ m-a$-n$-kç$ to@ wç$n$. I-arrive- COMPL school there early, but I-COMPL-NEG-go reach them. “I arrived at the school in good time but I didn’t meet them.” (AS) Notice that even though na may be used in communicating some form of contrast, its use in such contexts, as in (28), does not involve contradicting and consequently dropping an earlier held assumption. 10 For instance, even though the utterance in (29) communicates a contrast due to the meaning of the lexical items tuntum (black) and kçkçç (red) combining with the encoded meaning of na, it does not communicate the assumption that Afrakoma is expected to be light skinned because Nana Ama is. In order words, what is communicated is a mere semantic opposition between being light skinned and dark. This is the same situation in English, but in English, but is equally suitable for this purpose. However the use of nanso in (30) does indicate that the expectation that Afrakoma will also be light skinned (for instance because she is Nana Ama’s twin sister) has to be dropped in favor of the assertion that she is dark. 10 My suspicion is that in certain marked contexts na may be used to signal ‘denial of expectation’, but this is rare and of course not surprising given the versatility of the marker na. 100 Amfo (29) Na$na@ A@m !a@ yE$ kç$kç$ç@ na$ A$fr$a$ko@ma@ yE@ tu$n$tu$m@. Nana Ama be red CONJ Afrakoma be black “Nana Ama is light skinned and Afrakoma is dark.” (30) Na$na@ A@m !a@ yE$ kç$kç$ç@ na$n@so@ A$fr$a$ko@ma@ yE$ tu$n$tu$m@. Nana Ama be red but Afrakoma be black “Nana Ama is light skinned but Afrakoma is dark.” The two conjuncts in (29) and (30) semantically encode the same information, however the choice between na or nanso lead to different contextual assumptions, as stated above. The communicative role of nanso is in line with that of English but, as suggested by Blakemore (2002). Nanso, like but, is relevant to the extent that it activates “an inference which results in the elimination of an assumption” (Blakemore 2002: 100). The assumption which is eliminated is most likely to have been derived from the processing of the immediately preceding utterance, even though it may also be retrieved from the short or long term memory. The appearance of the supposedly human creature described in the first utterance in (31) is expected to ignite fear in anyone who sets eyes on it. The introduction of the second utterance with nanso is an indication that as far as Kwadwo was concerned such an expectation was not met. The addressee is thus expected to drop the contextual assumption in (32) which is likely to be derived as a result of processing the first complex sentence in (31). (31) Wo@-hwE$ o$ni@pa@ a@ w-a$-fu$w$ m$-mE$n$ na@ ne$ ho@ a@-bo@n@ na@ ne$ You-see human.being REL he-PERF-grow PL.horn CONJ POSS skin PERF-smell CONJ POSS hwe@ne@ mu$ gya@ a@, wu@-se$ yi@w@, a$nwo$nwa$de@ wç$ w@ia@se@. Na$n@so@ Kwa$dwo@ a$-n$-su$ro@. nose inside fire CM, you-say yes, wonders be.at world. But Kwadwo COMPL-NEG-fear “When you see a human being who has grown horns, and who is smelling and has fire coming out of his nose, you would realize that indeed there are wonders in the world. However Kwadwo was not afraid.” (AK/a 12) (32) Anyone who sees such a fearful creature will be terribly frightened. What is communicated by the first clause in (33) is that the referents of the subject pronoun wç (that is, the parents of one of the children referred to in the nanso-clause) talked a lot about the issue referred to there, specifically the recalcitrant behavior of these children. Contrary to expectation, the children did not show any signs of giving up that behavior. Nanso (here naaso) gives an indication that the expected results of these counseling sessions were not achieved. (33) Wç$-ka@-a$ ho@ a@sE@m@ pi$i$, na$a@so@ n@na$ m$br$E$ m$bo$fr$a$ no@ se@ a@ra@ na$ wç@-se@. They-talk-COMPL skin matter a.lot, but then what children DEF say just FM they-say “They talked a lot about it but (however) the children persisted in their (mis)deeds.” (FA/skc 20) Explaining connections in Akan discourse: the role of discourse markers . 101 6. Conclusion By focusing on some discourse markers/connectives in Akan, this paper has demonstrated that one way of signaling relevance relations in discourse cross- linguistically, is by the use of certain discourse markers or connectives as suggested by Blakemore (1992). They provide procedural information by indicating the direction in which a particular utterance has to be interpreted. Akan (n)so indicates that the utterance in which it is contained ought to be interpreted within a context similar to that of the immediately preceding utterance. However, unlike the situation with English too/also which requires a context almost identical to that of the immediately preceding utterance, a broader (or more general) context is sufficient to license the use of (n)so. The entailment condition required for the use of (n)so is less constrained. Na, the ordinary coordinating connective equivalent to English and, combines with specific grammatical features of the conjunction and other contextual information to communicate a number of inferential relations between conjuncts. These relations are temporal, causal, parallel, contrastive and explanatory. It has been noted that the fact that explanatory relations can be communicated by the use of a na-conjunction (contra Carston 2002) is a language internal feature permitted by the collocation of the clause types outlined in (26). Nanso, the Akan contrastive connective has a similar function to that of English but, as analyzed by Blakemore (2002). It indicates that an earlier held assumption needs to be dropped in favor of the following utterance. This study has shown that coherence relations within single utterances and between parts of a discourse can generally be achieved by the use of discourse markers since their use provides procedural information leading to the identification of the kind of contextual information the addressee needs to access, in his bid to arrive at the intended interpretation. However in pursuit of a general cognitive theory of utterance interpretation we need to proceed cautiously recognizing certain language-internal facts, as illustrated for instance by the use of na in (23), where the inferred logical relation between the conjunct propositions is one that is alien to and-conjunctions in European languages. References Amfo, N. A. A. 2005. Recurrence Marking in Akan. Pragmatics, 15, (2/3): 151-168. Amfo, N. A. A. 2006. “Lexical Signaling of Information Structure in Akan” Paper presented at the Linguistics Department Seminar Series, University of Ghana, September 2006. Amfo, N. A. A. in press. Clausal Conjunction in Akan, Lingua. Anonymous 1961. Ananse Akuamoa 1. Bureau of Ghana Languages Blakemore, D. 1992. Understanding Utterances: an introduction to pragmatics, Oxford: Blackwell Publishing. Blakemore, D. 2002. Relevance and Linguistic meaning: the semantics and pragmatics of discourse markers. Cambridge: Cambridge University Press. [Cambridge Studies in Linguistics 99] Blakemore, D. and Carston, R. 2005. The Pragmatics of Sentential Coordination with and. Lingua, 115:(4): 569-590. 102 Amfo Blass, R. 1990. Relevance relations in discourse: a study with special reference to Sissala. Cambridge: Cambridge University Press. Carston, R. 2002. Thoughts and Utterances: the pragmatics of explicit communication. Oxford: Blackwell Publishing. Clark, W. 1993. Relevance and Pseudo-imperatives. Linguistics and Philosophy, 16: 79- 121. Coleman, S. K. 1995. Woana Nye Araba Nyinsenii. Sam-Woode Limited Jucker, A. H. 1993. The Discourse Marker well: a relevance-theoretic account. Journal of Pragmatics, 19: 435-52. König, E. 1991. The Meaning of Focus Particles: a comparative perspective. London: Routledge. Noh E.-J. 2000. Metarepresentation: a relevance-theory approach. Amsterdam/Philadelphia: John Benjamins Publishing Company. Recanati F. 2004. Literal Meaning. Cambridge: Cambridge University Press. Sperber, D. and Wilson, D. 1995 Relevance: communication and cognition, Oxford: Blackwell Publishing, 1st edition 1986. Unger, C. 1996. The scope of Discourse Connectives: implications for discourse organization. Journal of Linguistics 32: 403-38. Wilson, D. and Sperber, D. 1988. “Representation and relevance”. In Mental Representations, R. M. Kempson (ed.), 133-53. Cambridge: Cambridge University Press. Appendix A Maame Esi mmpE fi. Maame Esi nyim dE fi dze yarba ba. Maame Esi ma Kodwo na Araba sawee ma wçwe. çdze sawee no twutwuw mbofra no hçn se ho. Otwutwuw hçn se ho ma hçn se ho yE fEw. Maame Esi guar Kodwo na Esi. Oguar mbofra no ma hçnho tsew na wçkç skuul. Memenda biara so çhor mbofra no hçn ntar na çtow do. Translation Maame Esi doesn’t like dirt. Maame Esi knows that dirt brings about diseases. Maame Esi gives Kodwo and Araba chewing sponge. She uses the chewing sponge to clean the children’s teeth. She cleans their teeth nicely. Maame Esi bathes Kodwo and Esi. She bathes them till they are clean, before they go to school. Every Saturday, she washes the children’s clothes and irons them. (FA/skc 8) Appendix B Kodwo na Araba ayeyE mpanyimfo. Kodwo edzi mfe duesia. Oewie skuul. Araba so edzi mfe duebiasa. çakç mfikyir. Maame Esi ehu dE Araba akç mfikyir. Araba da kç skuul. Aka mfe ebien ama oewie. Araba wie a, çbçkç Nsçwdo Skuul. Asem a Araba taa Explaining connections in Akan discourse: the role of discourse markers . 103 ka nye dE owie n’adze sua a, çbçkç akEyE edwuma wç asopitsi. çse, da bi çbEyE nEEse na çaboa ayarfo. Translation Kodwo and Araba have grown. Kodwo is sixteen years old. He has completed (basic) school. Araba, likewise, is thirteen years old. She has menstruated. Maame Esi has noticed that Araba has menstruated. Araba still attends (basic) school. She has two more years to go. When she is through, she will attend secondary school. What Araba usually says is that when she completes her studies, she will go and work in the hospital. She says one day she will be a nurse and help sick people. 104 WORKING PAPERS, ISK 3/2006 105-122 ISK, NTNU ‘Abroad’ and semantically related terms in some European languages and in Akan (Ghana) Thorstein Fretheim and Nana Aba Appiah Amfo 1. Introduction This paper takes a look at a semantic field that comprises on the one hand linguistic expressions which encode a movement away from a spatial deictic center, or origo (Bühler 1934; Levinson 2004) defined as a country or a nation (a homeland), and on the other hand the state of being outside the deictic origo or the locating of an event somewhere outside the deictic origo. The central lexical term with this meaning in English is abroad. We are concerned with the denotative meaning of abroad and corresponding expressions in some other European languages, among which Norwegian will be given the most attention. What referent is designated by a given token of abroad, or by its closest semantic relatives in other languages? There are two strategies that are available to the English communicator, and correspondingly to the addressee trying to determine the reference of tokens of abroad in a discourse: the spatial deictic origo may be identified as the communicator’s homeland, or it may be defined as the homeland of some salient discourse referent. The latter includes what Bühler (1934) used to call Deixis am Phantasma (“deixis in the imagination”), which counts on the addressee’s ability and willingness to transpose himself to an imagined deictic origo and to identify the reference of abroad relative to that domestic location, whether fictional or factual. Our use of abroad and similar terms in European languages is compared to the way that the closest lexical correspondents of abroad are used and understood in Akan, a Kwa language of the Niger-Congo family and the dominant indigenous language in Ghana. Two points of difference between the European situation and the Akan situation will receive special attention. Users of Akan rely almost exclusively on a speaker-oriented strategy that causes the hearer to identify the deictic origo as Africa, or more narrowly as Ghana, or even more narrowly as what is domestic to the Akan people. Moreover, while the English term abroad has very little connotative meaning for the average speaker, the corresponding terms in Akan are characterized by a mixture of denotative and connotative meaning. Some other terms belonging to the same semantic field as abroad do have a connotative meaning component, like their lexical counterparts in Akan. What is foreign to us is unfamiliar and may even impress us as being odd, and we believe this particular connotative meaning to be responsible for the tendency of such terms to have developed signs of lexical ambiguity (polysemy) over time, maybe via metonymic associations. Words in this category include adjectives like English foreign (“strange” in addition to “non-domestic”) and French étranger and étrange, and nouns like English foreigner and French étranger. The lexical base of foreign and foreigner is that of the Latin adverbs forīs, meaning “outside”, especially “outside the home”, and forās, the directional partner of forīs, whose meaning is “out”, especially “out of a house”. The stem for- of these adverbs is derived from the Latin word for “door”, foris, which is primarily a device whose function 106 Fretheim and Amfo is to shut those people out who have no business on the other side of the door. Domi (“at home”) is the Latin antonym of the adverb forīs. It is derived from domus (“house”), a noun that has also given rise to the Latin adjective domesticus, a polysemous word meaning “pertaining to the house” but also “homely”, “private”, “personal” and “national” (cf. the modern English adjective domestic). The reference of words like foreign and especially foreigner is strongly context- dependent, like the reference of abroad, but in addition their denotative and connotative meanings are intertwined in such a way that it is difficult to describe the former independently of the latter. Within this particular subfield of the lexical field examined in this paper we find just as much interdependence of denotative and connotative meaning in the European languages discussed as in Akan. While our paper is largely descriptive in nature, our relevance-theoretic allegiance (Sperber and Wilson 1986, 1995; Carston 2002) will be seen to shine through. We regard utterance comprehension as an enterprise that relies on different perceptual and cognitive inputs, one of which is the encoded logical form of the linguistic stimulus. Linguistic semantics underdetermines not only what the communicator means but also what she explicitly says. Processing an utterance that contains the adverb abroad involves resolution of the reference of this word, without which no proposition can be construed. We conclude that the assignment of reference to a given token of abroad is often a highly context- sensitive task. Insomuch as it is possible to identify one or more words in Akan which fulfill a similar communicative function, we conclude that assigning reference to those Akan expressions is not context-dependent to quite the same extent. 2. The European situation 2.1 “Abroad” The original meaning of the place adverb abroad was “over a wide (or, if you like, broad) area”. It was unknown before the Middle English period but has been conventionally used with a number of related, yet conceptually distinct lexical meanings since then, like “out of one’s own country”, “away from one’s place of residence”, “out of doors”, “on the move”, “at large” and “broadly” (or “widely”). The meaning that we are going to focus on is the salient present-day concept “away from someone’s home country”, where one important issue to be resolved by the addressee is the reference of this “someone”. It appears that the individual whose nationality determined the reference of a token of abroad in an English utterance was originally the communicator, so that the word referred to any country that was foreign to the communicator. This usage involving resolution of the reference of abroad in relation to a deictic origo which is the country of the speaker herself is still very much alive to this day, of course. Someone uttering (1) presupposes that the interlocutor is able to determine the reference of abroad pragmatically, without any grammatical cuing. Even if the language in (1) is English, the reference is to anywhere outside Finland if the speaker is a Finn, anywhere outside Switzerland if the speaker is Swiss, and so forth. (1) These products tend to be cheaper abroad. Abroad’ and semantically related terms in some European languages and in Akan 107 It is important to remember, though, that the reference here is not really to all those countries in the world that count as foreign countries for the speaker; abroad in (1) may be said to refer loosely (Sperber and Wilson 1986a, 1986b; Carston 2002) to certain countries whose price level, for the kind of product referred to, the speaker happens to be familiar with. Linguistic data such as (1) represent a major challenge for any ‘minimal semantics’ approach to truth-conditional content, which ascribes truth conditions to compositional properties of the linguistic semantics of sentences (cf. Borg 2004; Cappelen and Lepore 2005) rather than to utterances of sentences processed in a context (cf. Carston 2002; Recanati 2004). Faced with an utterance of (2), an addressee could be said to determine the reference of abroad by the same pragmatic strategy as in (1), but one could also argue that the reference of abroad in (2) is determined by the reference of the grammatical subject, the 1st person pronoun. The speaker and the subject referent coincide here. (2) I was abroad for two weeks. The bottom-up, linguistically driven, reflexive comprehension of what abroad refers to reflects a tendency to let the reference of a number of conceptually related adverbial lexical entries be determined by the reference of the grammatical subject, like the adverb home in They were home, or Come home! or out in How often did you go out?, in the sense of “How often did you leave home?” Thus, abroad, like home, is in effect an indexical expression. It is a characteristic of all indexicals that their reference may be determined ‘anaphorically’ through local discourse-driven resolution, or else ‘deictically’ through purely extra- linguistically driven resolution. Suppose that an utterance of (1) above was produced by an Englishman by the name of Martin. A Canadian could report on Martin’s statement in (1) by uttering (3). (3) Martin believes those products to be cheaper abroad. The meaning of (3) could conceivably be that the Canadian speaker claims that Martin holds the belief that those products are cheaper outside Canada than in Canada, but that would not only be a misrepresentation of what Martin actually said in (1), it would also be an interpretation which the hearer is extremely unlikely to access even in the absence of adverse contextual evidence. The proposition expressed in the complement of the matrix clause verb form believes in (3) will normally be understood to be the speaker’s reported interpretation of Martin’s thought, and as the term abroad will be understood to belong to the speaker’s interpretation, or metarepresentation (Sperber 2000; Noh 2000; Ifantidou 2001) of a thought attributed to Martin, its reference will be determined relative to Martin’s homeland rather than the speaker’s. If the hearer knows that Martin is English, that knowledge will determine his resolution of the reference of abroad. If the hearer is unaware of Martin’s nationality, he will still not opt for the alternative speaker-oriented interpretation, unless there is rather strong contextual evidence that the two different perspectives on how to process the word abroad coincide. 108 Fretheim and Amfo Observe, though, that what we have here described as the grammatically reflexive resolution of the reference of abroad does not have to be due to a linguistic item that ‘binds’ the reference of this word, so that abroad would be judged to take a referential value that contrasts with and excludes the nationality of the matrix clause subject referent. The second utterance in (4) is an instance of ‘free indirect speech’, or ‘free indirect thought’, as the case may be (Sperber and Wilson 1986a; Lucy 1993). It metarepresents something Martin said, which accounts for his decision reported in the preceding utterance. (4) Martin decided not to buy a large quantity of vitamin C tablets in the local pharmacy. Those products tended to be cheaper abroad. Past tense tended reveals that the second sentence in (4) is neither a direct speech report nor a descriptive statement issued by the speaker of (4). If this “local pharmacy” is in, say, Saffron Walden, abroad refers loosely to countries on the other side of the British Channel. That recognition depends just as much on pragmatic inference as the alternative processing which takes the speaker’s homeland to be the deictic origo rather than Martin’s homeland. Occasionally a conflict may arise between contextual, top-down resolution and grammatically local, bottom-up resolution of the intended reference of abroad. Suppose a Norwegian informs an Englishman as shown in (5). (5) Ross and Linda went abroad on the 1st of September. Suppose further that the hearer knows that the referents Ross and Linda are US citizens who settled in Norway very many years ago but who were supposed to spend part of the season back in the USA; the hearer may still have insufficient contextual evidence for drawing either the conclusion that the couple left Norway, i.e. went to America, on the 1st of September, or the conclusion that they left America (where they had already been for some time) on that date. On the former interpretation the reference of abroad would presumably be interpreted relative to the home country of the speaker; on the latter interpretation the reference would be conceived relative to Ross and Linda’s home country, bound by the reference of the subject nominal Ross and Linda. This looks like a reference resolution task that involves just bottom-up processing, yet obviously there is just as much top-down processing in this case, because the utterance presupposes the hearer’s knowledge of what would count as abroad for Ross and Linda. This is a contrived example, and it is probably very seldom that a communicator would choose the descriptive indexical adverb abroad in a situation where it may create problems for the hearer’s inferential processing because he or she cannot readily decide whether a grammatically reflexive interpretation or a speaker-oriented interpretation of the reference of abroad is intended. It is somehow misleading to say that the couple Ross and Linda in (5) went abroad if they are still US citizens who went to their original homeland, and it would presumably be just as confusing to be told that they went abroad if the intention of the speaker of (5) was to convey that they returned to Norway on the 1st of September. The problems pertaining to the proper usage of the term abroad in a context like the one considered in the preceding paragraph are not so different from the problems that speakers of Akan in Ghana may encounter in their non-native use of English abroad in less Abroad’ and semantically related terms in some European languages and in Akan 109 artificial contexts than the one stipulated for the utterance of (5), or in their selection of the appropriate Akan term when they use their native Akan to refer to the whereabouts of people who are not Akans, and not even Africans (see section 3 of this paper). As noted, abroad is an indexical adverb, although not in the same sense as the deictic place adverbs here and there, for example, because the reference of here or there is never reflexively bound by the subject nominal of the sentence, and an even more crucial difference is that the lexical meaning of abroad is defined negatively: the word can denote anything except the home country of some salient discourse referent, like the communicator or someone whose thought is metarepresented by the communicator1. In other European languages, the concept that corresponds most closely to the concept encoded by the English adverb abroad is typically expressed by means of a prepositional phrase in which the concept “foreign” is taken care of with the help of a (nominal) prepositional object, while the difference between staying there, as in She was living abroad, and going there, as in She went abroad as often as she could, is typically expressed by a difference in the choice of preposition, as when the German neuter gender noun Ausland (lit.: outland) is preceded by im (= in dem) for duration abroad and ins (= in das) for movement to somewhere abroad. Expression of movement from abroad requires use of a prepositional phrase even in English – from abroad – corresponding to aus dem Ausland (lit.: out of the outland) in German. In French the noun corresponding to German Ausland is étranger, which happens to be the French word for “stranger” or “foreign person” as well, a word which is conceptually different from the noun étranger corresponding to English abroad and German Ausland, as in the prepositional phrase à l’étranger (“abroad”). The latter word does not encode the information that the country is strange in the sense of unfamiliar, exotic, even mysterious, it simply denotes whatever country is not the home country of the person whose nationality negatively determines its reference. In addition there is even an adjective étranger/étrangère with the two meanings of “foreign”, as in “foreign language”, exemplified by M. Durand connaît deux langues étrangères (“Mr. Durand knows two foreign languages”) or as in “unfamiliar”, exemplified by Je suis étranger à cette affaire (“I’m unfamiliar with this matter”). This is probably not a case of lexical vagueness but rather an ambiguity at the lexical level, because if you speak and write a foreign language, you are clearly not unfamiliar with it. In Norwegian and the other Scandinavian languages the definite article is suffixed to the nominal stem, so the indefinite form utland (lit.: outland) of the nominal part of the Norwegian expression that equals English abroad appears as a bare indefinite noun only in set phrases like (gjester) fra inn- og utland ((guests) from in- and outland, that is, “(guests) from Norway and from abroad”). The definite form utlandet (outland + definite article) combines with a preposition meaning “to the place”: til utlandet, or “at the place”: i utlandet, or “from the place”: fra utlandet. In addition there is an adverb utenlands, with an added genitive -s which works just like English abroad in that it is used without a preposition both for residence in a foreign country and for going to a foreign country, while 1 The deictic adverb here can also refer loosely to some newly presented information in a discourse, as in Here we see a good example of how ruthless scientists can be. 110 Fretheim and Amfo the expression of “coming from abroad” is fra utlandet, never just #utenlands or the ungrammatical *fra utenlands. Old Norse had the same form útland, but the extension of the Old Norse word was considerably more restricted, as fara út (“go out”) was typically used with reference to a voyage westwards to places like Iceland, the Faroes, the Orkneys, Shetland. To this day the Norwegian phrase reise ut (“go out”) is associated mainly with travelling out of Norway, to foreign countries; it does not simply mean “to depart”. There is a conceptual difference between the compounds utreise (lit.: out-travel) and avreise (lit.: off-travel), the former being used mainly with reference to air flights and the latter with reference to the start of a journey on ground, like a train departure. In (6) the sentence itself encodes information which activates a context that enables the reader to identify the Norwegian nation as the deictic origo. There is an antonymous pair of expressions in (6). Hjemme (“at home”) contrasts with ute (lit.: out), and here the most suitable gloss for ute is “abroad”. The illustration in (6) is from the Oslo Multilingual Corpus, henceforth abbreviated as OMC (http://www.hf.uio/german/sprik/english/corpus.shtml). The indication “(UD1)” is the code used in the OMC to identify the author, and the arrow pointing to the right, →, shows that what follows is a translation of the source text preceding the arrow. (6) Av politiske fanger og motstandsfolk døde 658 hjemme, 1.433 ute. of political prisoners and resistance-people died 658 at-home 1,433 out (UD1) → Among political prisoners and members of the underground, 658 died at home and 1,433 abroad. Though it looks very much like a truism, we wish to stress at this point that it is totally unproblematic to translate a token of the Norwegian phrase i utlandet (meaning “abroad”) into English as abroad and still preserve the idea that abroad in the target text can refer to anywhere except Norway, and it is equally unproblematic to translate in the other direction. The reason we are saying this will become clear when we later turn to a consideration of the correspondents of “abroad” and related concepts in Akan. Consider the illustration in (7), which we have also taken from the OMC. The pronoun they in the English target text refers to a Norwegian couple. (7) Dessuten hadde de bodd i utlandet i flere år, moreover had they lived in outland-Def in more years, nærmere bestemt i Sverige. more-closely determined in Sweden) → They had even lived abroad for several years, in Sweden to be precise. The subject pronoun they in the translation refers to Norwegians, and it is their national identity that determines the reference of the later indexical adverb abroad in the same clause. We can easily switch the perspective needed to interpret the reference of such indexicals as abroad, depending on our knowledge of the nationality of the subject referent. Abroad’ and semantically related terms in some European languages and in Akan 111 Translated texts will include instances of ‘Deixis am Phantasma’ which involve deictic transpositions of even a higher order than the original text that the translation is based on. The reference of an occurrence of abroad is arguably more context-dependent when the subject of the clause does not bind its reference, which is true of the following example from the OMC, again a translation from Norwegian into English. In (8), the reference of abroad depends on our identification of those who made the investments, and unlike example (4) this is not free indirect speech. (8) Inntektene fra investeringer i utlandet ble halvert. Income-Def-Pl from investments in outland-Def became halved → Revenue from investments abroad was halved. The English prepositional object investments abroad in the English translation refers to “Norway’s investments outside Norway”. This interpretation is made available through top- down processing which includes the strongly manifest assumption that this is a text whose topic is the state of Norway’s economy. A similar illustration is given in (9), where the English target text happens to include a bit more information than the corresponding Norwegian source text due to differences in the segmentation of the text into periods. Here the phrasal conjunctions of i [Norge og utlandet] and [in Norway and abroad] are also seen to cue the reader’s interpretation of utlandet in the source text and abroad in the target text by making it clear that the two conjuncts are complementary. (9) Firmaet har investert 20 milliarder kroner i Norge og utlandet over de siste fem år uten å øke sin gjeld. lit.: The firm [i.e. Norsk Hydro] has invested 20 billion kroner in Norway and abroad over the last five years without to increase its debt → In December 1985, Norsk Hydro was 80 years old and it had invested 20 billion kroner in Norway and abroad over the past five years, without increasing its debt. Finally, agentless passives do not contain a grammatical subject that can bind the reference of a later occurrence of abroad, so we understand abroad in (10), from the OMC, to refer to what is abroad for the buyer of the goods referred to by the pronoun they (maybe as opposed to the narrator who is responsible for the hedge “doubtless”), and the referential identity of the buyer must again be retrieved through top-down, context-sensitive processing. (10) They had doubtless been bought abroad and were more than likely grossly expensive. Language vastly underdetermines not only what a speaker intends to communicate by means of an utterance but also its explicitly communicated truth-conditional content (Sperber and Wilson 1986a; Carston 2002; Recanati 2004). Pragmatists who recognize the normality of semantic underdetermination at the level of explicit propositional content will not be worried by the absence of a linguistic item that binds the form abroad referentially. 112 Fretheim and Amfo We have no reason to postulate a covert syntactic element which is supposed to ‘compensate’ for the lack of overt reference to Norway in the English translation in (8) above. The NP investments is enriched in context as “Norway’s investments”, so that abroad can be enriched as “outside Norway”. As “going abroad” means crossing a border between one’s home country and a foreign country, it is not surprising that the Norwegian directional adverbial phrase til utlandet in the source text of (11) is translated into English as shown here. (11) Men til utlandet kommer hun seg nå ihvertfall ikke, … (KF1) but to outland-Def comes she Refl now at-least not → She won’t be able to cross the border, though, … The phrase cross the border in the English translation will be conceptually enriched (Carston 2002; Recanati 2004) through context-driven inference as a crossing of the border between Norway and one of the three countries that border on Norway: Sweden, Finland, or Russia. The set of contextual assumptions accessible to the reader will determine whether or not the author’s intention is for the reader to narrow the selection of countries down to just a single one of those three; it could be that the intended reference is indiscriminately and disjunctively to any one of those three countries. In one respect the English translation cross the border may be said to constrain the reader’s interpretation more than til utlandet in the Norwegian source text does. You can go abroad by crossing an ocean as a passenger in an airplane but if you say that you cross the border between two countries, you normally imply that there is no ocean between the country you leave and the country you arrive in. On the other hand, the phrase til utlandet, which contains an indexical, utlandet, must be saturated in context by being associated with an ‘antecedent’ individual whose homeland serves the role of deictic origo and fixes the identity of the country that the agent leaves. The phrase cross the border does not do that, because it contains no indexical whose referential resolution depends on a process of antecedent-based enrichment. However, either concept “go abroad” or “cross the border” must be embedded in a specific context that determines how they are to be manipulated pragmatically in the inferential phase of the reader’s comprehension task. In the German OMC excerpt presented in (12), the term abroad is applied with reference to a rather special situation: the City of Berlin, surrounded by East Germany (DDR) on all sides, was divided into East Berlin, which belonged to the German Democratic Republic (DDR), and West Berlin, which belonged to the Bundesrepublik, i.e. West Germany, but even the population of West Berlin were subjected to rules and regulations which meant that West Germany was to all intents and purposes “abroad” to the citizens, another country. (12) Denn nicht nur servierte man ihnen ein plattes Nein auf die Frage, for not only served one them a straight No on the question ob über Schritte zur deutschen Einheit gesprochen werden könnte, if over steps to the German unity talked be could Abroad’ and semantically related terms in some European languages and in Akan 113 Chruschtschow selbst gab ihnen auf den Weg, Khrushchev himself gave them on the way die Bundesrepublik solle für West-Berlin Ausland werden. the Federal Republic should for West Berlin abroad be (WBR1) “For not only were they given a straight ‘No’ to the question if it would be possible to take steps toward a united Germany, Khrushchev himself gave them the message on the way, that the Federal Republic of West Germany (BRD) should be ‘abroad’ (outland) for West Berlin.” We have shown in the present section that the addressee must frequently have recourse to certain information presented in an earlier part of the discourse in order to resolve the reference of abroad. We used the term ‘anaphoric’ about those occurrences of abroad whose reference must be determined locally through a search for some ‘antecedent’ in the immediately preceding discourse, but we have also seen that there is not always an antecedent-like expression in the immediately preceding discourse which will provide the information necessary to saturate the indexical abroad, that is, to fill it with that conceptual information about the homeland of some salient person or institution, which is required to resolve the reference of this very special element of spatial deixis. An illustration like (8) above shows that the information about the homeland which is complementary to the geographical area denoted by abroad may be accessible only if the addressee takes a global perspective on the discourse topic. It would not be consonant with accepted conventions to call the token of abroad in (8) an instance of ‘anaphora’ (see De Mulder 1998 on different definitions of this term in linguistic literature). 2.2 “Foreign” The adjective foreign requires the same context-dependent recognition of the intended deictic origo as abroad. Occasionally there is nothing in the sentence itself or in the preceding sentences belonging to the same discourse segment which cues the reader’s assignment of reference to the attributive adjective foreign. Look at the following excerpt from the OMC. (13) He talks about Kenya, the country’s economy, foreign aid, the contrast between town and country, what do I think of Nairobi? (TB1) Someone reading and processing (13) in the discourse in which this sequence of metarepresented predications and a final question occurs must bear in mind that he or she is reading an English translation of a Norwegian novel, and not an English source text. The 1st person narrator and the male referent of the pronoun he are both Norwegians, the country referred to is Kenya, but the indefinite NP foreign aid refers to Norway’s aid to developing countries in general and, in the present context, to Kenya in particular. The adjective foreign refers to what is outside Norway, but the complex phrase foreign aid refers to Norwegian aid bestowed upon countries in the Third World, and not aid given by foreign states (cf. the term Foreign Office). 114 Fretheim and Amfo While English foreign and French étranger refer either to what is outside some contextually determinate home country or to what is unfamiliar or strange, these terms translate into Norwegian in one of two ways: as utenlandsk (literally ‘outlandish’, but without the special meaning and connotations of that English adjective)22 relating to things abroad, or as fremmed, relating to what is unfamiliar. 2.3 “Foreigners” The Norwegian noun utlending (“foreigner”) is evidently a close semantic relative of i utlandet (“abroad”) and utenlandsk (“foreign”) but while there are no negative connotations attached to i utlandet/abroad and few, for most citizens, to utenlandsk/foreign, the nouns utlending and foreigner have a certain affective meaning, which may be strengthened in some contexts and weakened in others. One reason for this lexical difference is probably that utlending and foreigner are terms that refer to human beings, to (groups of) individuals who may lose face and who may be insulted, while utenlandsk/foreign is not restricted to animates and i utlandet/abroad does not refer to anything animate. A few instances of utlending, or plural utlendinger, in Norwegian source text excerpts from the OMC are translated as foreign visitor(s), and the definite singular form utlendingen (“the foreigner”) was even translated as the newcomer once. That translation may actually be interpreted as a kind of ‘euphemistic’ alternative to use of the English term foreigner, which is undoubtedly the most straightforward, unmarked translation of Norwegian utlending. When we refer to certain people as foreigners, we classify them as foreigners not because of where they are in relation to a deictic origo (as is the case when we interpret the adverb abroad), but because we attribute certain inherent properties to them, and with some language users, some of those properties may be not altogether favorable. Does the observed tendency to avoid use of the term foreigner in English translations of the noun utlending in the OMC mean that the English term is more stigmatized than its Norwegian counterpart? That would be too rash a claim. Norwegian utlending is also stigmatized in certain contexts, probably more so today than a couple of generations ago, when Norway had much fewer immigrants from outside Europe. We asked ten informants, five males and five females, all of them holding a university degree, what kind of person they would immediately think of as the referent of the nominal phrase en utlending (“a foreigner”) in (14), which is part of an imagined conversation between two Norwegians. The informants were asked whether they interpreted the person described as a foreigner to be a non- Norwegian residing in some foreign country or an immigrant to Norway who is of foreign extraction, like a Pakistani or a Turk. 2 The English adjective outlandish grew out of Old English ūtlendisc, which is derived from the noun ūtland (“foreign land”), the same form as present-day Norwegian and Swedish utland, which is the nominal base of the prepositional phrase i utlandet (literally “in the outland”, i.e. “abroad”). Chambers Dictionary of Etymology (1988: 741) states that, “The extended sense of unfamiliar, strange, odd, bizarre, is first recorded in 1596.” This semantic narrowing (rather than ‘extension’, we would claim) testifies to the widespread cognitive link between what is foreign/alien and what is strange and therefore sometimes incomprehensible. The “foreign”/“strange” polysemy was lost in the adjective outlandish, leaving us with just the latter meaning Abroad’ and semantically related terms in some European languages and in Akan 115 (14) Jeg søkte på den jobben, men det var en utlending som fikk den. I sought on that job-Def but it was a foreigner who got it “I applied for that job, but a foreigner got it.” Everyone’s most immediate thought was that the job had been offered to a person who was neither native to Norway nor someone of non-Norwegian lineage who had settled in the country. When asked why they associated the term utlending with a foreigner residing in a foreign country rather than with an immigrant who had possibly become a Norwegian citizen, no one hinted that their subjective interpretation was colored by the fact that, after all, first generation immigrants in Norway usually have a hard time finding themselves a job that matches their qualifications. All ten informants said they would be hesitant to describe an immigrant who masters the Norwegian language reasonably well and is accustomed to Norwegian culture as en utlending (“a foreigner”). Two of them added that the person who got the job and was described as en utlending could not be someone from one of the other Scandinavian countries either. One informant said that he would not refer to someone as utlending if he knew what country was his or her home country; then he would rather describe the person as a German, an Italian, and so forth. Our informants were also asked how they would interpret the speaker’s use of the term utlending if the job referred to was a job outside Norway, for example in France. Given that changed context, they all agreed that the description en utlending was not appropriate, whether or not the one who got the job was French, because as they said, in that situation even the speaker of (14) is a foreigner in the sense of not being French. The deictic origo is France, the country where the job is. However, the whole group of informants agreed that (15), where en utlending is replaced by en annen utlending (“another foreigner”), is considerably more natural than (14), if the job is not in Norway. (15) Jeg søkte på den jobben, men det var en annen utlending som fikk den. I sought on that job-Def but it was an other foreigner who got it “I applied for that job, but another foreigner got it.” Our conclusion, based on the reactions we obtained from our informants, is that utlending is at least a mildly stigmatized term for speakers of Norwegian, and for that very reason it is less of a problem to use it with an indirect reference to one’s own self, as in (15) where en annen utlending (“another foreigner”) implies that the Norwegian speaker places himself in the category of foreigners, relative to France, where the job is. Doing that cannot be a discriminatory act; nor would it be discriminatory to refer to the one who got the job as utlending, if that person is completely unknown to the speaker, though the description certainly implies that the person is not French. It is totally impossible for a Norwegian to refer to himself as a foreigner because he is not French and to another person as a foreigner because that person is not Norwegian, and to group the two individuals linguistically together in one set, as in (15). In order for the description en annen utlending in (15) to be appropriate, the person denoted by the speaker of (15) cannot belong to the French nation, nor will this person be identified as another compatriot, a Norwegian. If both persons are Norwegians, the only relevant description would be en annen nordmann (“another 116 Fretheim and Amfo Norwegian”) without an indexical expression, which focuses on what the two applicants have in common. 2.4 The antonymic “at home” The reference of the indexical lexical phrase at home is usually determined by a grammatical subject that binds it, or identified through association with antecedent information in the discourse, but at home can also refer to what is home for the communicator, even when there is no explicit 1st person reference anywhere in the discourse. Hjemme (“at home”) is a Norwegian place adverb whose denotative properties are similar to those of English at home. These expressions may be classified as a kind of antonym of abroad/i utlandet, but they obviously do not encode information that restricts the deictic origo to someone’s homeland. As with Latin domi and domesticus mentioned in the Introduction, the deictic origo can be a family, a town, a region, or a nation, depending on various sorts of contextual clues. The unmarked interpretation of the juxtaposition of predications in the Latin expression forīs bella, domi sēditiōnes (“abroad wars, at home conflicts”) is that the communicator refers to conflicts in the home country and wars outside the home country but this is not the only permissible interpretation. Norwegian i hjemlandet (lit.: “in the homeland”) may also be classified as an antonym of i utlandet, and one that encodes the information that the deictic origo is a country, but this expression is hardly ever used with reference to the speaker’s/narrator’s homeland, it is almost invariably used ‘reflexively’ with reference to the homeland of a 3rd person subject referent, i.e. to someone else’s homeland. In the OMC it is seen to correspond to English phrases like at home, in their homeland, in the motherland, but also in her country, which creates a psychological distance, showing that the narrator does not empathize with the 3rd person referent. The most popular German correspondent of Norwegian i hjemlandet is the prepositional phrase in der Heimat (lit.: “in the home”) but the prepositional phrase im Ursprungsland (lit.: “in the origin land”) is also sometimes used as a translation of Norwegian i hjemlandet. The German noun Heimat is used both with reference to a home with the meaning of dwelling and to a home country (Heimatland), the same familiar vagueness that characterizes English at home and its Norwegian equivalent hjemme. While a phrase like at home includes reference to someone’s residence as well as to larger geographical areas like a specific region of a country or one’s home country, abroad displays no comparable vagueness in its potential to refer, unlike Latin forīs (“abroad”). Abroad invariably refers to some geographical area which is at least a whole country and possibly the full complementary set of countries outside the deictic origo. In Norwegian, however, the place adverb ute (“out(side)”) is occasionally seen or heard to alternate with the phrase i utlandet. This is found in the set phrase hjemme og ute (“at home and abroad”) and even in collocations like reise ute (lit.: travel outside) which is normally used with the meaning of “travel (around) abroad” (see also example (6) in §2.1). Abroad’ and semantically related terms in some European languages and in Akan 117 3. The Akan situation Akan belongs to the Kwa branch of the Niger-Congo family of sub-Saharan languages and is spoken mainly by the Akans of Ghana. The English word abroad has no exact equivalent in Akan but it may get an approximate translation by means of one of the two words amannɔne and aburokyiri. Both words appear to be compounds, though the individual components of the former are not very transparent. What can be identified as one segment of that word is aman, which means “countries”. Both words refer to foreign land; however, the referential extension of aburokyiri is much more narrowly constrained than that of amannɔne. On the other hand, aburokyiri is the default lexical choice when an Akan makes a general reference to foreign countries. Amannɔne refers to any place outside of the Akan’s home country. The word is quite formal and not as widely used in colloquial Akan as its near-synonym aburokyiri. However, in formal contexts like newscasting, amannɔne is often heard to be used in such a way that its lexical properties approach those of English abroad, so that it may even be hard to distinguish amannɔne and abroad semantically. On the other hand, when there is a need to refer to the fact that someone has gone abroad in everyday conversation, the Akan speaker tends to use the more mundane term aburokyiri or the phrasal verb tu kwan meaning “to travel”, which is emotively neutral. Interestingly, even the Oslo Multilingual Corpus contains a few correspondences in which a text in one European language refers simply to travel, while the corresponding fragment in the language translated from, or translated into, contains a reference to the more narrowly specified concept of traveling abroad, as shown in (16), in which the translator’s introduction of the indexical word abroad in the English version is indicative of a pragmatic enrichment of the concept of traveling or “journey-making”, which is constrained in context as “traveling outside the narrator’s home country” (for a discussion of conceptual narrowing, see Carston 2002; Sperber and Wilson 1998). (16) Hun svarte med å rekke meg hånden, og jeg tolket det som et løfte. she answered with to reach me hand-Def and I interpreted it as a promise Det innebar imidlertid at jeg på et senere tidspunkt måtte gi henne min adresse, that implied however that I on a later time-point must give her my address noe jeg alltid har vært svært forsiktig med når jeg er ute og reiser. something I always have been very cautious with when I am out and travel (KF1) → Her response was to shake my hand, which I construed as a promise. This meant I’d have to give her my address later on, a thing I’ve always been wary about doing when abroad. The mundane Akan word aburokyiri can be roughly translated as “abroad”. As observed by Osam (1997), etymologically, the word can be said to consist of two forms, oburoni and nkyi. Oburoni is the Akan word for “white man” and nkyi refers to 118 Fretheim and Amfo “homeland”, so aburokyiri is the white man’s homeland.3 Aburokyiri in its initial use referred to Great Britain, the land of the colonial masters. As the Akan got exposed to the fact that there are other white people than the British, the concept was extended to cover the continents of Europe, America and Australia. While expressions like abroad, Norwegian i utlandet, German im Ausland or French à l’étranger are practically devoid of any conventionalized emotive meaning, Akan aburokyiri has retained much the same connotative meaning components as the noun oburoni (“white man”) has in the language. Aburokyiri is frequently translated as overseas. This translation may be due to the fact that one had to cross the ocean to get into the white man’s land. Whether or not the continent of Asia is referred to as aburokyiri depends on the informant’s general view of the world. For most people with little education and/or western exposure, an oburoni is anyone who is light skinned and has relatively straight hair vis-à-vis the dark skinned and kinky curly hair of the African. For such people Chinese, Japanese and Indians are aburofo (“white people”) and subsequently their countries are referred to as aburokyiri. Some Akans, especially those who are educated or are aware of racial and other ethnic distinctions, are hesitant to refer to countries in Asia and the Middle East as aburokyiri, but apart from that, the consensus, by and large, is that aburokyiri refers to any geographical area outside of Africa. When the reference is to a country within Africa (or Africa and Asia, for some speakers), the Akan speaker will prefer to be specific and mention the name of the country rather than use a general term with loose reference, such as aburokyiri, or even amannɔne. What Akans refer to as aburokyiri was at least originally a conceptually fixed geographical area outside of Africa. It was the continent of Africa that was conceived as the deictic origo, not Ghana. After all, national borders were not fixed the way they are today. Ekua E. Appiah kindly assisted our work by questioning Akan-speaking Ghanaians about their understanding of the denotative value of the noun aburokyiri. She asked eleven medical doctors in a hospital in Accra whether they would, or could say that a person was at aburokyiri if he or she had gone from Ghana to Togo. Just one out of eleven said it would be in order to use that term under the circumstances, and the others present unanimously objected that this informant’s intuition was influenced by the way he would use English abroad. It is significant that the location referred to as aburokyiri does not depend on one’s present whereabouts or one’s current home base. For instance, a native of Ghana based in Norway could produce an utterance of (17), where aburokyiri refers to where she is presently staying, but if she is going to spend a couple of weeks abroad in Germany, she will not say that she is going to aburokyiri or staying in aburokyiri, since she is already in Europe, which is aburokyiri for her. 3 Compounds which use oburoni as a basis are not unusual in Akan. Christaller (1933) records a few, such as aburobua (“clay pipe”), aburogua (“armchair”), burokuruwa (“Western jar”, “mug”, “cup”), buronya (“Christmas”) and burongo (“olive oil”). Though some of these words may have largely gone out of use, having been replaced by their nativized European names such as kɔɔpoo for cup, others such as buronya (“Christmas”) and aburokyiri (“abroad”) are still found in the everyday repertoire of Akan speakers. Another compound involving the root oburoni which has not become obsolete is aburofo nkate (“almonds”, lit.: white men’s groundnuts). Abroad’ and semantically related terms in some European languages and in Akan 119 (17) Me wɔ aburokyiri. I be-at abroad “I am abroad.” Consequently, using aburokyiri is not an option when the reference of English abroad would have to be determined in the context of a deictic transposition (‘Deixis am Phantasma’). This state of affairs differs markedly from the English use of abroad described above, and it is actually reflected in the Akan’s use of the English word abroad. Akans are often hesitant to use the word abroad in English conversation, unless the context is such that its use matches their concept of aburokyiri. Thus, when expressing themselves in English they would generally prefer to mention the country they are going to and would avoid using the formulation I’m going abroad. Even the term amannɔne is used cautiously in contexts analogous to what was described above. Nevertheless, unlike the situation with aburokyiri, the reference of amannɔne could be derived inferentially from one’s knowledge of the subject referent. For instance, it is appropriate to use amannɔne, but less so aburokyiri, as a near-equivalent of the concept encoded by the English word abroad in (18) with reference to an Englishman by the name of Tony who found a girlfriend outside Europe. (18) Tony wɔ mpena wɔ amannɔne baabi, na mmom Ɛ-n-yƐ T. has girlfriend be-at foreign-land somewhere Conj however it-Neg-be obi a ɔ-te Ghana ha sei. someone Rel she-stay G. here Dem “Tony has a girlfriend somewhere abroad (i.e. outside England), but it is not someone who lives here in Ghana.” Use of amannɔne does not exclude taking a particular non-African person’s perspective and using that person’s homeland as the deictic origo that determines the reference of this term. Substitution of aburokyiri for amannɔne in (18) would make the utterance extremely awkward, no matter whether Tony found his girlfriend somewhere in Africa or in a different part of the world. You would not normally use aburokyiri to refer to a foreign country perceived from the point of view of an Englishman. As English is the only official language of Ghana, there has traditionally been no great demand for translation of texts, fiction or non-fiction, from a major European language into Akan. Finding a suitable Akan translation of a word with the conceptual meaning of English abroad and with no connotative overtones is by no means trivial. A normal strategy that would seem to work most of the time is to try to be referentially more specific, that is, to refer to what the original text calls abroad by means of an appropriate geographical name, a proper name, instead of using a referentially imprecise term that is supposed to render the meaning of the English indexical abroad in as literal a way as possible. It would not be absolutely impossible to approximate a literal translation into Akan of Norwegian i utlandet or English abroad, as the example in (18) was meant to demonstrate, but if a translator were to do so, comprehension of the linguistic result would usually require a bit of make-believe on the part of the Akan reader. 120 Fretheim and Amfo It should be added at this juncture, that new coinages like Amerika aburokyiman mu (lit.: America abroad-land inside, i.e. “in America”) and by the same token Pakistan aburokyiman mu, India aburokyiman mu, etc., are gaining ground in news reports on something that has happened in a country outside of Africa, so that the deictic origo is to be construed as that foreign country. It is quite impossible, however, to refer to an African country by using this kind of nominal construction, and with the word aburokyiri, which is after all the most frequently used correspondent of abroad, no analogous shift of perspective would be possible, even though the reference of aburokyiri is no longer confined to the land of the colonial masters. We claimed that the English [+human] noun foreigner and its equivalents in other European languages have a connotative meaning, or implicated meaning as some neo- Griceans might say, which is comparable to the connotative meaning of Akan aburokyiri. It is far from surprising, then, that the most natural Akan translations of foreigner, the words ɔhɔhoɔ and ɔmanfrani, are also laden with certain negative connotations. The Akan word ɔhɔhoɔ refers to someone who does not belong to a given geographical domain, a visitor or outsider, but the deictic origo is never transferred to a territory outside Africa. The geographical domain could range from a house to an institution to a town/city or a country. It is applicable not only to foreigners from outside Ghana but even to people who transcend a border inside the country of Ghana. Thus the concept ɔhɔhoɔ covers what is conceived as “visitor” as well as “foreigner” in a language such as English. A morphologically and semantically related term is ɔhɔhomani (lit.: visitor + citizen) which refers to someone who was originally an outsider in the community but who has become completely assimilated to Akan culture and social conventions. An ɔmanfrani, on the other hand, was originally someone who had been captured as a slave and brought to a different nation, where he had later made his home. Such a person was invariably thought of as an alien and would not ever be regarded as native to the land where he was living, but he would typically have adapted fully to the conditions of his present place of abode. Morphologically, the word consists of the root ɔman (“country”) plus the verb fra (“to mix”) and the human singular suffix ni also occurring in Ghanani (“a person from Ghana”) and in the word ɔhɔhomani mentioned in the preceding paragraph. In present-day Akan, the word ɔmanfrani is offensive and is used only in contexts where provocation is intended. It still has the negative connotations pertaining to someone who was originally bought as a slave from outside of the geographical domain in question, but who has acclimatized much to the dislike of some natives. Our informants judge its use to be strongly pejorative. 4. Conclusion English abroad and corresponding expressions in other European languages are a type of spatial deixis words whose reference can only be resolved through extra-linguistic inference. This is true even when the speaker’s intended interpretation is such that abroad refers to countries that are complementary to the homeland of the referent of the grammatical subject of the clause in which the word appears. For any given occurrence of abroad, the addressee has to draw on contextual assumptions to determine whether its reference is intended to be ‘bound’ by a local subject nominal, or to be computed on the basis of global assumptions about the subject matter of the discourse, or else to be Abroad’ and semantically related terms in some European languages and in Akan 121 determined by the nationality of the communicator, whether the voice is that of a speaker in a conversational dialogue or of a narrator in a novel. The Akan terms corresponding most directly to abroad were shown to have very different denotative properties than the English term. The English language has a very strong position in Ghana, and in regard to certain types of literature it may be said to have what approaches a monopoly position. Not much fictional or non-fictional prose is published in the form of Akan translations from English or other European languages. The Ghanaian consumer of international literature cannot expect to find what he or she is looking for in a text translated into Akan. In our opinion this is one reason why a shift of the deictic center away from the Akan society, away from Ghana and away from Africa in attempts to render the concepts we associate with the concepts signified by words such as abroad and foreigner has not really become part of the linguistic conventions governing either written or spoken Akan. The speaker of Akan is not accustomed to using Akan, at the expense of English, if a situation arises where some sort of ‘Deixis am Phantasma’ is required. It is simply not Akan one would use in a situation where a European perspective on matters outside Europe is required. Acknowledgement We wish to express our gratitude to our informants in Ghana and Norway. References Borg, Emma (2004) Minimal Semantics. Oxford: Oxford University Press. Bühler, Karl (1934) “The deictic field of language and deictic words.” Reprinted in Speech, Place and Action: Studies of Deixis and Related Topics, R. Jarvella and W. Klein (eds), 1982, 9-30. New York: John Wiley. Cappelen, Herman and Ernie Lepore (2005) Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism. Oxford: Blackwell. Carston, Robyn (2002) Thoughts and Utterances: The Pragmatics of Explicit Communication. Oxford: Blackwell. Chambers Dictionary of Etymology. (1988) R. K. Barnhart (ed). Edinburgh: Chambers Harrap Publishers Ltd. Christaller, J. G. (1933) A dictionary of the Asante and Fante language called Tshi (Chwee, Twi). Basel: The Basel Evangelical Missionary Society. De Mulder, Walter (1998) “Anaphora”. In Handbook of Pragmatics, J. Verschueren, J.-O. Östman, J. Blommaert and C. Bulcaen (eds), 1-19. Amsterdam and Philadelphia: John Benjamins. Ifantidou, Elly (2001) Evidentials and Relevance. Amsterdam and Philadelphia: John Benjamins. Levinson, Stephen C. (2004) “Deixis”. In The Handbook of Pragmatics, L. R. Horn and G. Ward (eds), 97-121. Oxford: Blackwell. 122 Fretheim and Amfo Lucy, John (ed) (1993) Reflexive Language: Reported Speech and Metapragmatics. Cambridge: Cambridge University Press. Noh, Eun-Ju (2000) Metarepresentation. Amsterdam and Philadelphia: John Benjamins. Osam, E. Kweku (1997) “The origin of racial labels in Akan.” Paper presented at the annual conference of the Linguistics Association of Ghana, University of Ghana. Recanati, François (2004) Literal Meaning. Cambridge: Cambridge University Press. Sperber, Dan (ed) (2000) Metarepresentations: A Multidisciplinary Perspective. Oxford: Oxford University Press. Sperber, Dan and Deirdre Wilson (1986a) Relevance: Communication and Cognition. Oxford: Blackwell. Sperber, Dan and Deirdre Wilson (1986b) Loose talk. Proceedings of the Aristotelian Society LXXXVI: 153-71. Sperber, Dan and Deirdre Wilson (1998) “The mapping between the mental and the public lexicon”. In Language and Thought: Interdisciplinary Themes, P. Carruthers and J. Boucher (eds), 184-200. Cambridge: Cambridge University Press. Authors’ e-mail addresses: thorstein.fretheim@hf.ntnu.no nana.amfo@hf.ntnu.no WORKING PAPERS, ISK 3/2006 123-138 ISK, NTNU A constructional approach to syntax and the treatment of passive in Norwegian Petter Haugereid Abstract A syntax-based approach to argument structure for HPSG is presented, where it is assumed that argument structure frames are syntactic constructs and that passive is a syntactic element. This makes it possible to avoid stipulating detailed information about argument structure in open class lexical entries. 1. Introduction Different phenomena in the discussion around argument structure such as unaccusativity and unergativity, valence alternations and voice are in frameworks such as LFG, HPSG and to a great extent also GB, accounted for by means of subcat stipulations in lexical entries. In the approach suggested here, five syntactic argument roles called arg1-role, arg2-role, arg3-role, arg4-role and arg5-role are assumed in order to account for the creation of argument frames. The arg1-role corresponds to the external argument role in GB/Minimalism. The arg2-role corresponds to the deep direct object role, the arg3-role corresponds to the deep indirect object role, the arg4-role corresponds to so-called delimiters, that is, resultatives and end-of- paths. The arg5-role corresponds to other PP complements that serve as preconditions for the event. The five argument roles are realized by syntactic entities such as phrase structure rules, inflections and function words. The approach has much of the formal inventory of an HPSG grammar (see e.g. [PS94]). All the constituents in the grammar (lexical and syntactic) are represented as typed feature structures. But the idea that argument structure is a syntactic construct is taken from Minimalism (See [Bor05a], [Bor05b] and [Åfa03]). With such a syntactic approach to argument structure, a verb is by default compatible with all possible argument frames. This is different from other HPSG approaches where verbs are listed with particular argument frames in the lexicon. The ideas are implemented in an HPSG grammar (Norsyg) for Norwegian, which covers all argument frames for verbs listed in lexical resources for Norwegian such as Troll and NorKompLeks.4 The paper opens with a presentation of the five argument roles in Section 2. Then the basic syntactic machinery is presented in Section 3, and finally, an analysis of passive for Norwegian is given in Section 4.5 4 A short description of the grammar and download instructions are given at http://www.hf.ntnu.no/hf/isk/Ansatte/petter.haugereid/norsyg.html. 5 A mechanism for arriving at the correct word order is implemented in Norsyg, but it will not be discussed in this paper. 124 Haugereid 2. Argument roles As mentioned in the introduction, five argument roles are assumed in this approach: The arg1-role (external argument), the arg2-role (deep direct object), The arg3-role (deep indirect object), the arg4-role (delimiter) and the arg5-role (argument that precedes the action). 2.1 Arg1-role The arg1-role corresponds to the external argument in GB/Minimalism, and can be seen as the initiator of the event. The arg1-role can be connected to the realization of an NP subject like in (1a). It can also be connected to the passive auxiliary as shown in (1b) or with the infinitival marker as shown in (1c). In (1c) the infinitival marker is connected to the arg1-role of sleep. (1) a. John sleeps. b. John is admired. c. John tries to sleep. 2.2 Arg2-role The arg2-role corresponds to the deep direct object. It can be connected to the realization of the direct object like in (2a). In (2a) the argument is an NP, but it can also be a subordinate clause (see 2b) or an infinitival clause (see (2c)). The arg2-role can be connected to the subject if the clause is unaccusative (see (2d)), or if the clause has passive voice (see (2e)). Like the arg1-role, the arg2-role can also be connected to an infinitival marker (see (2f)). (2) a. John admires Mary. b. John claims that he didn't sleep. c. John tries to smile. d. The man came. e. John is admired. f. John tries to come. 2.3 Arg3-role The arg3-role corresponds to the deep indirect object. It can be connected to the realization of the indirect object, like in (3a). It can be connected to the subject if the clause has passive voice (see (3b)). It can also be connected to the infinitival marker if the infinitival clause has passive voice (see (3c)). (3) a. John gave Mary flowers. b. Mary was given flowers. c. Mary hoped to be given flowers. 2.4 Arg4-role The arg4-role corresponds to so-called delimiters, that is, arguments that typically tell where the arg2-role ends up. The arg4-role can be connected to a resultative like in (4a) and (4b) or a end-of-path like in (4c). A constructional approach to syntax and the treatment of passive in Norwegian 125 (4) a. The lecture bored the students to death. b. John drank the glass empty. c. John threw the stone into the water. 2.5 Arg5-role The arg5-role corresponds to PP arguments that serve as a precondition to the action. It can be a source as in (5a) or an instrument as in (5b). The distinction between arguments that precede the action (what here is referred to as the arg5-role) and arguments that follow the action (the arg4-role) can also be found in [Cro91, 183-240]. (5) a. Water dripped from the roof. b. John painted the wall with a big brush. 2.6 Constellations of argument roles Argument frames are assumed to be constellations of argument roles. The argument frame of an unergative intransitive verb like sleep consists of only one argument arg1, and it is called an arg1-frame (see 1a). The argument frame of a transitive verb like admire consists of an arg1-role and an arg2-role, and is called an arg12-frame (see 2a). An unaccusative verb like come has only an arg2-role, and so its argument frame is an arg2-frame (see 2d). A ditransitive frame is called an arg123-frame (see 3a), and a transitive frame with a delimiter is called an arg124-frame (see e.g. 4a). It is also possible to have no argument roles, as illustrated with rain in (6), where the subject is an expletive. This argument frame is called an arg0-frame. (6) It rains. 2.7 The valence mechanism Instead of the valence features that are used in HPSG (see Figure 1), valence features for each of the argument roles are assumed, as illustrated in Figure 2.6 So instead of having valence features that reflect the surface function of the arguments, valence features that reflect the “deep” function of the arguments are assumed. Figure 1: Valence in HPSG Figure 2: New valence features 6 The arg5-role is not displayed here, and will not be discussed further in this paper. 126 Haugereid Verbs are specified with regard to what argument frame they have via the feature VAL-TYPE (see Figure 2). The transitive verb admire is given the VAL-TYPE value arg12, as illustrated in Figure 3. Note that the arg2-role of admire is constrained to be a nominal via the ARG2 feature. Some verbs like eat are compatible with more than one argument frame. eat can be both transitive as in (7a) (arg12-frame) and unergative intransitive as in (7b) (arg1-frame). These verbs are given VAL-TYPE values that reflect what argument frames they can appear in. In the case of eat the VAL-TYPE value is arg1-12. I will come back to these cases in Section 3.4. Figure 3: Lexical entry of admire (7) a. John eats fish. b. John eats. 3. Syntax Four classes of rules make up the basis of the syntactic analysis: - Valence rules: o Binary valence rules: Realize arguments in their canonical position. o Extraction valence rules: Extract arguments that are realized by the head filler rule. o Unary valence rules: Realize the unexpressed subject. Apply to functional elements such as infinitival markers and imperative morphology. - Force rules. Apply on the top on the tree. Categorize the clause as a proposition, question or command, and check that all the arguments are realized. - Head filler rule. Applies on the bottom of the tree. Realizes the element on the SLASH list as its first daughter. - Merge rule. Combines the projection that realizes the subject (the left branch) with a non-finite verbal category. A constructional approach to syntax and the treatment of passive in Norwegian 127 3.1 Valence rules There are four realizations of each of the valence rules above (except from the unary rules), one for each of the argument roles. The arg1 binary rule is illustrated in Figure 4.7 The rule in Figure 4 shows how the ARG1|LINK value is switched from a positive value in the head daughter arg1+ to a negative value in the mother arg1-. All the other valence features are copied up. The rule also introduces a parsons-style arg1-relation in C- CONT|RELS that links the non-head daughter to the LTOP of the head projection. The arg2, arg3 and arg4 binary rules are similar to the arg1 binary rule, except that corresponding LINK values are switched and corresponding parsons-style relations are introduced. Figure 4: Arg1 binary rule The valence extraction rules are similar to the binary rules except that the argument is not realized as the non-head daughter, but inserted into the SLASH list of the head daughter. An abbreviated version of the arg1 extraction rule is given in Figure 5. 7 I refer to instantiations of phrase types as rules and to phrases types as phrases. 128 Haugereid Figure 5: Arg1 extraction rule By doing linking with parsons-style semantics as illustrated here, it is possible to underspecify the arity of the semantic predicate. It will be the syntactic rules that in the end determine the number of semantic arguments, since the semantic arguments directly correspond to the syntactic rules. If the verb eat occurs in a transitive clause as in (7a), the number of semantic arguments will be two, as illustrated in (8a). If eat occurs in an intransitive frame (see (7b)), the number of arguments will be one, as illustrated in (8b). (8) a. b. As is shown in (8), the argument relations are linked to the predicate via the LBL of the relation. The semantic representations are called RMRSs (see [Cop03]). An approach that does linking in the syntax has an advantage to approaches that do linking in the lexicon, since A constructional approach to syntax and the treatment of passive in Norwegian 129 the latter approaches must specify precisely in the lexical entries how many semantic arguments they have, and so there is no room for underspecification. 3.2 Force rules Force rules are unary rules that apply on the top of the sentence and mark the sentence as a proposition, question or command. All the valence features of the daughter are constrained to have a negative LINK value, and the SLASH list is constrained to be empty. This is illustrated for the yes-no force rule in Figure 6. Figure 6: Yes-no force rule 3.3 Head filler rule The head filler rule applies at the bottom of the tree. This differs from the general assumption that the head filler rule applies at the top of the tree. It is the fact that the extraction rule always dominates the head filler rule that makes it possible to have the head filler rule at the bottom of the tree. The extraction rules enter the extracted element into the SLASH list of the daughter, as I showed in Figure 4. It is then copied from mother to head daughter until it reaches the head filler rule that fills it in. The head filler rule is illustrated in Figure 7. Figure 7: Head filler rule Since Norwegian is a V2 language, I assume that the constituent before the finite verb in a main clause is extracted and filled in by the head filler rule. This also holds for the subject of 130 Haugereid the main clause.8 So the analysis of an intransitive sentence becomes like the one in Figure 8.9 Figure 8: Analysis of Jon sover Figure 9: Analysis of En blomst gir Jon Kari (`A flower Jon gives Kari’) In Figure 8 the top unary rule (S) is the main clause force rule. The second rule (VP1) is the arg1 extraction rule, which inserts the extracted argument into the SLASH list of its daughter. The third rule (VP/NP) is the head filler rule which fills in the extracted subject. Figure 9 shows the analysis of a ditransitive sentence where the direct object is topicalized. The topicalized element is inserted into the SLASH list by the arg2 extraction rule (VP2), and transported down the tree until the head filler rule (VP/NP) fills it in. 3.4 The valence mechanism In order to control that a verb only occurs in the argument frames that it expects, a type hierarchy of linking types is introduced. Figure 10 illustrates how argument frame types are subtypes of certain constellations of LINK values. The top types in the hierarchy (arg1+, arg1-, arg2+, arg2-, arg3+, arg3-, arg4+ and arg4-) are possible LINK values. They tell whether an argument is realized or not. If an arg1 rule and an arg2 rule has worked in a clause, the LINK values of the head word of the clause projection will have the LINK values arg1+, arg2+, arg3-, and arg4-. These values are unified in the head word and unified with the VAL-TYPE value. As the hierarchy in Figure 10 shows, the greatest lower bound of arg1+, arg2+, arg3-, and arg4- is arg12, and if this type is not compatible with the VAL- TYPE value of the main verb, the parse fails. As I showed in Figure 3, the transitive verb admire has the VAL-TYPE value arg12. 8 A similar procedure is discussed in [PS94, 381] and is applied for Norwegian in [Ell03]. 9 The use of integers in the labels of the tree are there to show that an argument role has been realized by a rule. So the node VP1 indicates that the phrase has head value verb and that it realizes the arg1-role. A constructional approach to syntax and the treatment of passive in Norwegian 131 Figure 10: The link hierarchy The bottom types in Figure 10 are possible argument frames. The hierarchy also has a number of intermediate types. These are designed for verbs that can enter more than one argument frame. The type arg1-12 for example has two subtypes, arg1 and arg12. This means that a verb like eat, that is specified with arg1-12 as value of VAL-TYPE, will be compatible with both the arg1 frame and the arg12 frame. The tree in Figure 11 illustrates in detail how the LINK values end up in the head word. Here the head word is the verb beundrer (‘admires’). The tree also shows how the linking relations are introduced, and how they connect the predicate to the arguments. 3.5 Merge rule In the approach suggested in this paper, the main verb of the clause is not necessarily the head word of the clause. In a subordinate clause the complementizer will be the head word and in a main clause with an auxiliary, the finite auxiliary will be the head word. The function of the main verb in these cases will be to give the head projection its valence requirements. This is done in a rule called merge rule which copies the valence requirements of the non-head daughter onto the head projection. The merge rule is illustrated in Figure 12. In the merge rule the value of VAL in the second daughter is unified with the VAL values of the first daughter and the mother. This ensures that the head projection gets the valence requirements of the main verb. The merge rule makes use of a feature MERGE. This feature lets auxiliaries constrain the tense of the verbal constituent that they combine with. The first daughter serves as the syntactic head of the rule (the rule is head-initial), while the second daughter serves as the semantic head. The HOOK (that is, the LTOP and the INDEX) of the second daughter is unified with the HOOK of the mother. The merge rule takes as its first daughter a constituent that has realized the subject, and as its second daughter a verbal category that conforms with the tense requirements of the first daughter. This is exemplified in Figure 13. 132 Haugereid Figure 11: Transitive main clause A constructional approach to syntax and the treatment of passive in Norwegian 133 Figure 12: Merge rule Figure 13: Analysis of Jon har lest avisen (‘Jon has read the newspaper’) In Figure 13 the auxiliary har (‘has’) is the head word. The subject is realized by the arg1 extraction rule (AUXP1) in combination with the head filler rule (AUXP/NP) before the main verb is combined with the head projection by the merge rule (AUXP). The main verb transfers its valence requirements to the head projection and ensures that a proper argument frame is generated.10 4. Passive In this section I will show how the arg1-role can be expressed as passive voice. I follow the assumption of [Jae86], [Bak88] and [Åfa92] that there is a syntactic element PASS. I will show how this element has two realizations in Norwegian, namely as a passive auxiliary bli (bli-passive) and as an s-morpheme that is attached to the finite main verb (s-passive). First I will present some data. 10 The analysis suggested here is slightly different from the one implemented in Norsyg pr. September 2006. In Norsyg, the merge rule is head-final. In addition a more complicated procedure that I will not go into here, ensures that the valence requirements of the main verb end up in the head projection. 134 Haugereid 4.1 Data In Norwegian there are two kinds of passive, periphrastic passive (bli-passive) and morphological passive (s-passive). The periphrastic passive uses the auxiliary bli, (see (9b)), and the morphological passive attaches the suffix -s to the finite main verb (see (9c)). There is a slight semantic distinction between the two forms, which I will not go into (see [Hov77, 35-39]). The data I present here are well known in the literature (see e.g. [Hov77] and [Åfa92]). (9) a. En spiller smasher ballen. a player smashes ball-DEF ‘A player smashes the ball.’ b. Ballen blir smashet. ball-DEF becomes smashed ‘The ball is smashed.’ c. Ballen smashes. ball-DEF smash-PASS ‘The ball is smashed.’ In the examples (9b) and (9c), the subject (Ballen) would have been the direct object if the sentences were active. In (10), the three passive variants of the active clause Jon gir Marit en is (Jon gives Marit an ice cream) are given. (10) a. Marit blir gitt en is. Marit becomes given an ice-cream ‘Marit is given an ice cream.’ b. En is blir gitt Marit. an ice-cream becomes given Marit ‘Marit is given an ice cream.’ c. Det blir gitt Marit en is. it becomes given Marit an ice-cream ‘Marit is given an ice cream.’ In (10a) what would have been the indirect object in an active clause is the subject. In (10b) what would have been the direct object in active is the subject, and in (10c) the expletive det is the subject. 4.2 The passive types The passive auxiliary is assumed to be different from other words that may be the head word of a clause, in that it does not unify the LINK values in VAL, but rather in a feature called FIRST-VAL. The reason for this is rather technical. The auxiliary is assumed to introduce an arg1-relation in the same fashion as for example the arg1 binary rule (see Figure 4). The auxiliary does however not have a head daughter to relate its valence features to. (In the arg1 binary rule the ARG1|LINK value is switched from arg1- in the mother to arg1+ in the head daughter.) Instead the passive auxiliary relates its valence features to FIRST-VAL as shown A constructional approach to syntax and the treatment of passive in Norwegian 135 in Figure 14.11 If the passive auxiliary is the head word of the projection, the LINK values are unified in FIRST-VAL. Figure 14: The passive-aux-lxm type The s-pass-word, is an inflectional rule that adds an s-morpheme to main verbs. Figure 15: The s-passive-word type Since the passive auxiliary and the passive inflection absorb the arg1-role of the main verb, the subject must be realized by an element that does not have the arg1-role.12 11 The reason for having the FIRST-VAL feature in HEAD is that it will then be accessible to the merge rule in case the auxiliary is not the head word of the clause. In the implemented grammar it is the FIRST-VAL feature and not the VAL feature that is copied from the non-head daughter to the head daughter in the merge rule. But of expository reasons I have not shown that in the analyses given here. 12 It has been brought to my attention that in Yucatec Maya the verb corresponding to learn may have the following chain of suffixes: V - PASS - CAUS - PASS, and that the meaning corresponds to being taught, as illustrated in (i) on page 136 (see [Mül06]). In order to account for this kind of data, it seems that I would have to assume that there are two argument frames, one for the learning predicate and one for the causative morpheme, and that thetwo passive morphemes each realize an arg1-role. 136 Haugereid 4.3 Analysis The tree in Figure 16 shows in detail how linking is done in a passive transitive clause with the auxiliary bli. There are two different signs that do linking in the tree. The passive auxiliary (AUX1) adds an arg1-relation and shifts the arg1- link type in VAL to arg1+ in FIRST-VAL. This ensures that the arg1-role is realized. The arg2-extr-phrase (AUX2) adds an arg2-relation that it links to the extracted local and shifts the arg2- link type in the mother to arg2+ in the daughter. This realizes the arg2-role. The tree shows how all the link types arg1+, arg2+, arg3- and arg4- end up in the FIRST-VAL of the auxiliary. The tree does not show that the types are unified, since that would make the illustration more difficult to follow. When the types are unified, we get the type arg12, exactly like in the active counterpart in Figure 11. 5. Conclusion I have presented a constructional approach to syntax where different constellations of five syntactic argument roles are assumed to constitute argument structure frames. This, together with a hierarchy of link types (see Figure 10) has made it possible to give packed representations of possible argument frames of verbs without the use of lexical rules or multiple lexical entries. The assumption that passive is a syntactic element also has made it possible to account for passive without the use of multiple lexical specifications. (i) k=u ká `an -s -á`al le teòria-o` INCOMPL=3.ERG learn.PASS -CAUS -PASS.IMPF Det theory-D1 ‘The theory is being taught.’ (Somebody causes that the theory is being learned) A constructional approach to syntax and the treatment of passive in Norwegian 137 Figure 16: Transitive main clause with periphrastic passive 138 Haugereid References Baker, Mark C. (1988) Incorporation. A Theory of Grammatical Function Changing. The University of Chicago Press. Borer, Hagit (2005) Structuring Sense. An Exo-Skeletal Triology. Volume I. In Name Only. Oxford University Press. Borer, Hagit (2005) Structuring Sense. An Exo-Skeletal Triology. Volume II. The Normal Course of Events. Oxford University Press. Copestake Ann (2003) Report on the design of rmrs (preliminary version). Technical report, Cambridge, April 11. Croft , William (1991) Syntactic Categories and Grammatical Relations. Chicago. Ellingsen , Liv (2003) Norwegian word order in head-driven phrase structure grammar. phenomena, analysis, and implementation. Master's thesis, Institutt for Lingvistiske Fag, Universitetet i Oslo. Hovdhaugen , Even (1977) Om og omkring passiv i norsk. In Thorstein Fretheim, editor, Sentrale problemer i norsk syntaks, pages 15-46. Universitetsforlaget. Jaeggli, Osvaldo (1986) Passive. Linguistic Inquiry, 17:587-622. Müller, Stefan (2006) Phrasal or lexical constructions? Language, 82(4), To appear 2006. Pollard , Carl J. and Sag , Ivan A. (1994). Head-Driven Phrase Structure Grammar. University of Chicago Press, Chicago. Åfarli , Tor (1992) The Syntax of Norwegian Passive Constructions. Amsterdam: Benjamins. Åfarli , Tor (2003) Har verbet argumentstruktur? Motskrift, (2):87-99. Author’s e-mail address: petterha@hf.ntnu.no WORKING PAPERS ISK 3/2006 139-149 ISK, NTNU Diagnostic use of nonword repetition for detection of language impairment among Turkish speaking minority children in Norway Olaf Husby Background The communication skills of a child are considered delayed when the child is observed being noticeably behind his or her peers in the acquisition of speech and/or language skills. The impairment is related to the child’s capacity in speech/language comprehension and/or production and will significantly have an impact on the child’s educational progress compared with their age cohort. The delays and disorders range from simple sound substitutions to inability to understand or use language, or to use the oral-motor mechanism for functional speech and feeding13. A more detailed list of impairments will include delays and disorders related to – speech processing (recognition and processing of sounds in words and speech) – language comprehension (understanding words and sentences in a message) – language expression (ability to express messages using words and sentences). – speech production (ability to pronounce sounds correctly in speech. – language use (context relevant production and interpretation of language) By appropriate forms of specialist teaching, support and programs the language impairment of the child can be treated or reduced. A specialist works closely with family, teachers and counsellors and suggests effective strategies. Early identification of the nature of the impairment is essential for the success of interventions. It is important that a diagnosis is made and verified as early as possible. “There is considerable evidence that groups of children with language impairment repeat nonsense words less accurately than do their peers developing language normally” (Dollaghan and Campbell 1998). In order to locate language impaired children as early as possible, The Norwegian Ministry of Education and Research launched its strategic plan Equal education in practice. One of the aims is to develop appropriate tools for locating such children. This article is the result of a preparatory work to develop Turkish nonwords for nonword repetition tasks for immigrant children with Turkish as a first language. Nonword repetition Memory span is often interpreted as a measure of short-tem memory capacity (Hulme, Maugham and Brown, 1991). The span is related to the time taken to recite lists of span length (Schweickert and Boruff, 1986). When repeating gradually lengthened lists, greater 13 http://www.nichcy.org/pubs/factshe/fs11txt.htm 140 Husby parts of the information is lost. This is assumed to be caused by a passive decay process. This can be overcome by rehearsing the traces of the decaying items (Hulme, Maugham and Brown, 1991). It is shown that with repetition of lists that were just above span length, the probability of recalling the repeated list gradually increased. It is assumed that this is partially dependent on long-term memory representations. Watkins (1971) found that recall of items is sensitive to word frequency as introduction of high-frequency word contributes to an increase in the memory span. The high frequency words were also articulated more rapidly. One way of counteracting the contribution caused by familiarity with words, is to use nonwords, i.e. nonexistent words that conform to phonological rules of the language in question. It is shown that English-speaking children with specific language impairment consistently “score significantly lower than their age-matched typically developing peers […] and language-matched typically peers […] on tests of working memory. Research has focused on both phonological working memory – for example nonword repetition […] and sentence repetition” (Stokes et al. 2006). This findings have been corroborated by several studies, for instance Dollaghan and Campbell 1998; Gathercole et al. 2001; Alloway et al. 2005. “There is considerable evidence that groups of children with language impairment repeat nonsense words less accurately than do their peers developing language normally” (Dollaghan and Campbell 1998). It is even suggested that nonword repetition provides a phenotypic marker for some forms of developmental language impairment. Locating minority children with specific language impairment In 2004 The Norwegian Ministry of Education and Research launched its strategic plan Equal education in practice14. This is one of the measures implemented to raise the quality of education for language minorities. The target groups are minority language children of pre-school age, pupils and apprentices, as well as adults who still do not “reap the learning and social benefits of education stated in the primary objectives for Norwegian educational policy”. The primary goals of the strategic plan are (p. 33): 1. To ensure that minority language children of pre-school age have a better understanding of the Norwegian language 2. To improve the educational achievements of minority language pupils 3. To increase the percentage of minority language pupils and apprentices who begin and complete their upper secondary education 4. To increase the percentage of minority language students in higher education 5. To improve the Norwegian language skills of minority language adults The project described below is related to children of pre-school age. In 2002 there were around 33,000 minority language children from one to five years old. The majority of these had parents from non-western countries. The children with minority language background are under-represented in day-care institutions. While 66 per cent of all children of that age group attended day-care centres, only 33 per cent of all minority language children did so. 14 http://odin.dep.no/kd/norsk/dok/dok/handlingsplaner/045071-120010/dok-bn.html Diagnostic use of nonword repetition.Turkish speaking minority children in Norway 141 In addition, they in average spend less time there before starting school than majority language children. Research shows that the provision of good facilities for minority language children in day-care centres has a positive influence on the child’s school start. The strategic plan launches a set of measures to meet the challenges described above. The measures are set out in the order of the plan’s five primary goals. Goal 2 “To improve the educational achievements of minority language pupils” points at 19 measures. The one this paper is related to is measure 10: Language minorities with a need for specially adapted education: “The Norwegian Board of Education has been assigned the task of setting up a network to enhance competence related to language minorities who need specially adapted education. Relevant players in this network – in addition to the National Centre for Multicultural Education – include Torshov and Bredtvet resource centres [in Oslo]. Through the network, observation and mapping materials are to be developed along with methods connected to specially adapted education for pupils from language minorities who need such tuition. Human resource development in this field will also be considered, for example for school administrators and Educational-Psychological Services in primary and lower secondary education.” Responsibility is allocated to The Norwegian Board of Education and the National Centre for Multicultural Education in addition to school owners. The timeframe for the project is set to 2004-2007. Within measure 10 Torshov and Bredtvet resource centres are responsible for development, testing and evaluation of tools related to locating children from language minorities with specific language impairment. This project will focus on - tools for systematizing reports about the child’s bilingual development given by parents and pedagogic personnel - translation, development, testing and evaluation of various tests (Sentence memory: Test for Reception of Grammar – TROG; British Picture Vocabulary Test - BPVS). - tools for describing phonological aspects of children’s bilingual development As nonword repetition has been described as a clinical marker for special language impairment, a tool will be developed within measure 10. Tools for nonword repetition will ble developed for five minority languages: Turkish, Tamil, Urdu, Vietnamese and Albanian. Nonwords Nonwords are pronounceable strings of sounds which have no meaning. These words are also called pseudowords, invented words, nonsense words, or made-up words. For example the words HEASE, and MIVE seem to be English. They conform to English orthography, 142 Husby are all pronounceable, but they don't mean anything15. As meaning is absent, the repetition of such words relies upon phonological memory. Nonwords have to obey the constraints of the language that they putatively belong to. Spoken words must reflect phonotactic constraints, written words graphemic constraints. There are different kinds of nonwords. A monosyllabic word will function as a meaningless stem. Polysyllabic nonwords may be constructed by combining ±meaningless mono- or polysyllabic stems. Regular morphemes (roots, affixes, inflectional forms) may be combined with nonwords. As this project will use monosyllabic words only, I will not discuss possible structures any further. (Part of an outline is presented by Gürel, 1999.) As the child will repeat the nonwords orally after hearing them, it is important that the nonwords are constructed according to the phonological rules of the relevant language. However, in order for the test administrator to produce a valid reading of a nonword of a given language, the word should be written according to the conventional orthography rules of the language16. If different readings are possible (cf. the Norwegian nonword “fost” which may be read /fust/ (rhyming with “kost” - “broom”) ( or /fost/ rhyming with “post” – “mail”), instructions for pronunciation must be given. Dollaghan and Campbell (1998) present certain requirements for nonword repetition tasks. The nonwords should be designed to ensure that they are equally unfamiliar to children with impaired language] (LI) and children developing language normally] (LN). In this way one can avoid that the poor repetition of children with LI “could be attributed to their reduced language knowledge rather than to a fundamental psycholinguistic deficit. […] This requires that nonword repetition tasks be designed such that neither the nonwords nor their constituent syllables correspond to lexical items; further the predictability of individual phonemes within the nonwords should be minimized. In addition, nonwords ideally would include phonemes that are acquired early in the development (so that poor repetition performance cannot be attributed to articulation deficits) and are acoustically salient […]. Finally the presentation of the nonwords should be standardized to ensure that stimuli are presented with consistent rate, accuracy and intonation.” Monosyllabic nonwords must be constructed according to rules governing the syllable’s onset, nucleus and coda. In addition the word must follow the rules governing the inventory of the different positions (possible phonemes, their number, and their order). Voicing phenomena, like final devoicing, must be taken into account as well as suprasegmental features like duration and word tone. A larger set of nonwords should also reflect the relative frequency of phonemes in all positions of the syllable. For the actual project using Turkish nonwords the number of words is too small for the last requirement to be respected. However, nonfrequent features will not be implemented. For polysyllabic words there are more features to be aware of, among them stress, stress placement, weak forms, assimilation across syllable border, vowel and consonant harmony, juncture. For the project described here, only mono- and bisyllabic nonwords will be constructed. 15 Example words from http://www.readingsuccesslab.com/Glossary/NonwordTest.html 16 The discussion related to languages using ideographic writing will not be pursued here. Diagnostic use of nonword repetition.Turkish speaking minority children in Norway 143 Turkish - A short introduction to phonology and morphology Turkic languages are spoken from Balkan to north eastern Siberia. These languages share several salient features like agglutination, vowel harmony, verb-final word order and nominalised subordinate clauses (Kornfilt 1987). The biggest of the Turkic languages is Turkish. It is the main language of Turkey where it is spoken by about 60 million people, i.e. about 90% of the population17. Turkish is also spoken in Cyprus, Bulgaria, Macedonia, Greece, Romania and Uzbekistan. Below is a brief description of Turkish. For more details, see f.ex. Kornfilt (1997). Alphabet The Turkish alphabet is based on the Latin letters. There are 8 vowel graphemes and 21 consonant graphemes. The letter represents a central approximant /j/, and should not be confused with <ü> (which represents the phoneme /y/). <ğ> (“soft g”) always follows a vowel. It represents a rather weak front velar approximant between to front vowels. In word-final position or preceding a consonant, it has no independent pronunciation but lengthens the preceding vowel: - /da:/, - /da:da/ . abcçdefgğhıijklmnoöprsştuüvyz abcçdefgğhıijklmnoöprsştuüvyz ”The orthographic conventions correspond roughly to those of broad phonemic transcription” (Kornfilt ibid.). Some predictable alternations (voiced and unvoiced forms due to final plosive devoicing, voicing assimilation, vowel harmony) are expressed graphically. Alternations between palatal and velar realisations of /k, ɡ, l/ are not expressed graphically. The following letters do not correspond with phonemes that an be identified from the orthographic expression : <ğ > (silent), - /dʒ/, <ç> - /tʃ/, <ş> - /ʃ/, - /z/, - /j/. Vowels The vowel system of Turkish is symmetric. The eight vowel sounds can be grouped according to three features: height, backness and rounding as shown with letters in fig. 1. All combinations of the distinctive features ±back, ±round, ±high are observed. [- back] [+back] [-round] [+round] [-round] [+round] [+high] i y ɯ u [-high] e ö a o 17 Combined information from Kornfilt (1987) and CIA World Fact Book https://www.cia.gov/cia/publications/factbook/geos/tu.html. 144 Husby All vowel are phonemically short. However, on the surface level long vowels are found in18: 1. borrowings with unpredictable long vowels [ma:zi] – “past” 2. compensatory lengthened words of Turkic origin where a voiced velar fricative (present in orthography as <ğ>) used to follow a vowel, i.e. in syllable-final or word-final position. “çağ” – “epoch” is pronounced [tʃaː], in locative “çağda” - [tʃa:da]. 3. emphatic lengthening (Brendemoen and Hovdhaugen 1992:26) Diphthongs There are no diphthongs in Turkish. When two subsequent vowels occur, they belong to different syllables. A word like ait – ”belonging to” is therefore bisyllabic. Consonants Bi- Labio- Dental Alveolar Post- Palatal Velar Glottal labial dental alveolar Stop p b t d cj k ɡ Fricative f v s z ʃ ʒ h Affricate tʃ dʒ Nasal m n Tap ɾ Lateral l approximant Approximant j ɣ Final devoicing Syllable-final plosives /b, d, ɡ/ are devoiced (thus kitab-a /citaba/ ”book-dative”, but kitap /citap/ ”book”, kitap-lar /citaplar/ - ”book-plural/). Suffix-initial plosives assimilate progressively in voicing (thus kitap-da /citapta/ ”book-locative but araba-da /arabada/ ”car-locative”19. Another feature is the final devoicing of /r/. In initial and medial position /r/ is pronounced as an voiced alveolar tap. In final position it is pronounced as an unvoiced postalveolar fricative [ʃ]. 18 Examples from Kornfilt, ibid. 19 Examples from Oflazer and Inkelas (undated) Diagnostic use of nonword repetition.Turkish speaking minority children in Norway 145 Syllable structure According to Durgunoğlu (2006) 98% of syllables belong to the four forms V, VC, CV, CVC. Of these, the most frequent form is CV. There are a few cases of initial clusters CC, most of them in borrowed words. These words are mainly structured as C/r/ as in tren – ”train”, kral – ”king”. In colloquial pronunciation the consonants clusters are normally split by an epenthetic vowel. The result is tiren and kıral. Turkish is more tolerant for clusters in coda position. In codas of the type C1C2#, C1 is a non-plosive, C2 a plosive as in kurt – ”wolf”, borç – ”debt”. Some borrowed words contain non-acceptable consonant clusters. In general a high epenthetic vowel is inserted between the consonants. The vowels in general are governed by vowel harmony rules. (Arabic ism – ”name” is turned into Turkish isim, and Arabic nabz – ”pulse” into nabız (Brendemoen and Hovdhaugen 1992:34). Stress In general Turkish has stress on the word-final syllable: . arabá “car”, araba-lár “car- plural”, araba-lar-dán “car-plural-ablative”, bırák “leave!”, bırakacák “leave-future”, bırak-acak-lár “leave-future-3.plural”. The final stress rule is overridden in words containing lexically stressed roots, lexically stressed or pre-stressing suffixes and certain types of compounds. Most compound types have main stress on the first member (baʂ+bak- an → báʂbakan “head look after-relative participle” = prime minister’). Phrases typically have main stress on the final word or on the immediately preverbal constituent. Many roots and suffixes have exceptional stress placement, e.g. tarhána “dried curd” (root stress), pénaltı “penalty” (root stress), bırak-árak “leave-adverb” = by leaving’ (suffix stress)20. Assimilation and harmony Voicing assimilation There is a harmonizing principle at work for consonants. Initial plosives and affricates in suffixes assimilate with respect to voicing to the last sound of the word they are connected to. The underlying form of ablative case suffix is –DAn where D surfaces as /t/ or /d/ in accordance with the final sound of the preceding element. sokak-tan “from the street” kadın-dan “from the women” street-ablative woman-ablative kitap-tan “from the book” kitaplar-dan “from the books” 20 Examples form Inkelas and Ogun (2003) 146 Husby Vowel harmony A prominent part of Turkish phonology is the vowel harmony. Any of the eight vowels phonemes may occur in the first syllable of a Turkish word. “The distribution of vowels within a word is governed by vowel harmony, i.e. vowels share the specification for the feature [back] and, if they are high, they also share the specification for [round) (Kornfilt, ibid). Not all harmony principles (height, backness and rounding) are applied to all suffixes. Most affixes come in [-back] and [+back] forms like the plural form which surfaces as -ler and -lar. Locative case is signalled by the suffix –de/-da as in Türkiyede “in Turkey” but kapıda “at the door”. When all three features are at work suffixes may surface as in Türkiyedir - “it is Turkey”, kapıdır - “it is the door”, gündür - it is day", paltodur "it is the coat". The interplay between the three features is shown below for words kol – arm”, gül – “rose”, tür – “kind” and kum – “sand”. Durgunoğlu (2006) shows how these factors sum up as iterative loops. The noun tabak -”plate” - gets its plural form tabaklar - ”plates” - by adding the plural suffix -lar. The form of this suffix is governed by the [+back] vowel /a/ in the root tabak. Through suffixation of -daki we obtain the noun tabaklardaki: ”That present on those plates”. This can be turned into plural form by suffixing the plural morpheme presented earlier, but as the vowel now preceding the plural morpheme is the [-back] vowel /i/, the morpheme surfaces as -ler: tabaklardakiler – ”those present on those plates”. tabak plate - tabaklar plates tabak-lar tabaklardaki that present on those plates tabak-lar-daki tabaklardakiler those present on those plates tabak-lar-daki-ler There are several exceptions, most of them found in borrowed words. If a second vowel does not harmonize with the previous, the third vowel will start its own harmony domain. Kitap – “book”, which is borrowed from Arabic, will have –lar as plural suffix, a suffix that is harmonised to the second syllable. Vowel harmony is an important feature in Turkish word formation as there are agglutinating suffixes expressing person, number, case, tense, aspect, mood, voice, negation. Assimilation and harmony interplay A layer of complexity is added as vowel harmony and voicing assimilation interplay. In this way some suffixes have several forms. The underlying form of ablative case suffix is –DAn where D surfaces as /t/ or /d/ in accordance with the final sound of the preceding element and A will surface as /a/ or /e/ depending on the [±back] feature of the last vowel. orman-dan “from/via the forest” köprü-den “from/via the bridge” ağaç-tan “from the tree” eşek-ten “from the donkey” Diagnostic use of nonword repetition.Turkish speaking minority children in Norway 147 Morphology Turkish morphology is agglutinating and suffixing. Derivational suffixes precede inflectional ones. The suffixes added are stacked in a fixed order so a seemingly complex word may easily be analysed to its morphemic constituents as the examples gelmiyorum and kitapcıdaydık show: gel--m---iyor--------um “I am not coming” come-neg-progressive-1.singular kitapcı----da--(y)dı--k “We were at the bookseller's” bookseller-LOC-COPULA-1p.plural Turkish nonwords Based on the description given above a list of approximately 240 words were constructed. Brendemoen and Tanrıkut (1980) was used as reference. The list was later checked by two Turkish speaking informants, one of them a native Turkish speaker. 35 words were identified as either being real words, word roots or very close to actual Turkish words. These words were removed from the list. The final list contains 196 words, 116 monosyllabic and 80 bisyllabic. Of the monosyllabic words there are 4 CV-word, 29 VC and 82 CVC. There are no words with initial letter here. Brendemoen and Tanrıkut (1980) contain only 10 entries which all are loanwords. Dictionary Monosyllabic words Bisyllabic words. entry a ab atmat, armı, ayun, alun, ağmin b bam, bik, buk, bum, baf bunuk, bömün c cik, cüf, cal, cüm cimet, celi, cetur ç çemi, çıl, çom, çül çampa, çemür, cıfrı, çamık d di, dül, dör, dis, div devar, dümes, dimek, derek, duvek e em, ef ecik, eksal, elam, ektar f fam, for, feş, fım, fük fikar, fori, fazı, fedra, fade g gım, göv, gap, gil, gan gatlak, gımat, gömde, gürte h hon, het, hit, hün, hul hilka, herli, hütük ı ıb, ım, ır, ıv, ık ıtar, ına, ıta i ib, im, ir, iv, ik ida, ispon, ivam, izir, infor, itmut k ket, küv, köç, kam katem, keper, kümbe, kötlü l let, lo, löş, lef, lök laba m mon, mü, mür, mar, mıp mezlu, mime, müfet n niz, nug, naş, nok, nök noka o ob, om, oş omun, otmok 148 Husby ö öm, ök, öş özme, ötmöt p pit, pap, pöt pama, pitmi r rek, rüm, reş, rüt, rül rolap, rekme s se, sım, sep, saş sılma, söre, süplü ş şe, şim, şep, şat, şup şuma, şita t tal, tem, teş, toş, tüz tebek, tehmit, temlu, tora, tüvek u ub, um, uş, uv, ul ufri, utak ü üb, üm, üv, ül, üş ülü, ümtek v ven, viz, vuk, vaş, vel vekaş, vufas, vomkal y yam, yev, yöl yımı z zak, zel, zem, zür, zif zıpa As I have not been able to locate information about the relative frequency of phonemes in all positions of the syllable, no measures are taken to ensure that the word list is representative in that way. Based on a brief calculation of the number of words listed for each letter in Brendemoen and Tanrıkut (1980) a list was created to express relative frequencies. Words starting with initial letters with relative weight 1 are the less frequent. Words starting with letters of relative weight 4 have approximately twice as many entries as letters with relative weight 2 and so on. Letters with relative weight 1 are all small in number, but vary from about 60 entries (ı) to 230 entries (n). These relative frequencies are not compared to those of bigger dictionaries, but it is expected that the ratio will be approximately the same. The relative frequency should be referred to when a greater number of nonwords are put together to form a representative list. 1 2 3 4 5 6 7 cçefıj gpş abdhy i s mt k lnoö ruüvz Acknowledgement I would like to express my thanks to Mamo Tasan and Emel Türker. I am grateful to Thorstein Fretheim for several important remarks. Diagnostic use of nonword repetition.Turkish speaking minority children in Norway 149 References Alloway, T.P.; Gathercole, S.E; Adams, A.-M.; Willis, C.; Eaglen, R. and Lamont, E. (2005) Working memory and phonological awareness as predictors of progress towards early learning goals at school entry. British Journal of Developmental Psychology 23, 417-426. Brendemoen, Bernt and Tanrıkut, Yaşar (1980) Tyrkisk-norsk ordbok. Türkçe-Norveççe sözlük. Oslo: Universitetsforlaget Brendemoen, B. og Hovdhaugen, E. (1992) Tyrkisk grammatikk. Oslo: Universitetsforlaget. Dollaghan, C. and Campbell, T.F. (1998) Nonword Repetition and Child Language Impairment. Journal of SPeech, Language, and Hearing Research 41, 1136-1146. Durgunoğlu, Aydın Yücesan (2006) How Language Characteristics Influence Turkish Literacy Development. In: Malatesha Joshi, R and Aaron P. G. (2006) Handbook of Orthography and Literacy. New Jersey: Lawrence Erlbaum Gathercole, S.; Pickering, S.J.; Hall, M. and Peaker, S.M. (2001) Dissociable lexical and phonological influences on serial recognition and serial recall. The Quarterly Journal of Experimental Psychology, 54A (1), 1-30. Gürel, Ayşe (1999) Decomposition: To What Extent? The Case of Turkish. Brain and Language 68, 218– 224 Güngör, Tunga and Kuru, Selahattin (undated) Representation of Turkish morphology in ATN www.cmpe.boun.edu.tr/~gungort/papers/Representation%20of%20Turkish%20Mor phology%20in%20ATN.doc Hulme, C.; Maugham,S, and Brown G.D.A. (1991) Memory for Familiar and unfamiliar Words: Evidence for a Long-Term Memory Contribution to Short-Term Memory Span. Journal of Memory and Language 30, 685-701. Inkelas, Sharon and Orgun, Cemil Orhan (2003): Turkish stress: a review. Phonology 20 (2003) 139–161. Kornfilt, J. (1987) Turkish and the Turkic Languages. In: Comrie, B. (1987): The World’s Major Languages. London: Croom Helm Kornfilt, J. (1997) Turkish. London: Routledge Oflazer, K. and Inkelas, S. (undated) A Finite State Pronunciation Lexicon for Turkish http://people.sabanciuniv.edu/oflazer/archives/papers/pronlex.pdf. Schweickert, R. and Boruff, B. (1986) Short-tem memory capacity; Magic number or magic spell? Journal of Experimental Psychology: Learning, Memory and Cognition 12, 419-425. The Norwegian Ministry of Education and Research (2003): Equal education in practice http://odin.dep.no/filarkiv/212768/Strategiplan_ENG.pdf Stokes, S. F.; Wong, A. M.-Y.; Fletcher, P. and Leonard, L.B. (2006) Nonword Repetition and Sentence Repetition as Clinical Markers of Specific Language Impairment: The Case of Cantonese. Journal of Speech, Language, and Hearing Research 49, 2, 219-235. Watkins, M.J. (1971) The intricacy of memory span. Memory and cognition, 7, 411-419. Author’s e-mail address: olaf.husby@hf.ntnu.no 150 . WORKING PAPERS ISK 3/2006 151-160 ISK, NTNU Annotating and sharing language paradigms online Dorothee Beermann and Atle Prange Figure 1. Home of TypeCraft - www.typecraft.org 1 Introduction Dynamic language documentation is among the essential tasks of Modern Linguistics and one of its central concerns. It explores and redefines the borderlines between ‘field linguistics’, computational linguistics, and theoretical linguistic research, and if we were to name just one important aspect of the enterprise we would point to the potential that lies in the combination of traditional field methods with new technologies. With TypeCraft we present a project that focuses on the documentary and exploratory mode of research. Its aim is to generate reusable language data-sets with the foremost goal of advancing the standardization of glossing and to create resources of annotated natural language paradigms that capture central linguistic construction types. In this perspective we will describe and discuss TypeCraft’s central objectives, which are:  To provide an easy-to-use online tool for the annotation of natural language paradigms.  To make documentation of glossing conventions and descriptions of construction types immediately accessible to the annotators. Annotating and sharing language paradigms online 152  To foster glossing as a community effort, allowing the interactive use of annotation standards and their further development.  To make annotation and classification of language tokens subject to peer review in a wikipedia-like design.  To provide access to high-quality annotated natural language paradigms.  To develop a repository of data from less-studied languages, as well as less-known paradigms from well-studied languages.  To dynamically (re)structure the database to represent a repository of multi-lingually relevant construction types. Our paper is organized as follows: In section 2 we address some of the challenges within language documentation. In section 3, we discuss the main design properties of TypeCraft. In section 4, we address its future. Section 5 concludes the paper. 2 Language Documentation: Challenges and Goals The build-up of language data from less-studied languages faces two problems. The first problem concerns data preservation, while the second problem resides in data presentation. Forms of data preservation are dependent of the means chosen by the individual field linguist, and may vary greatly from the use of a tape recorder + on-paper documentation to digital techniques such as database systems for audio, video and written data. Within the latter category, The Field Linguist’s Toolbox (http://www.sil.org/ computing/catalog/), a light weight data management system distributed by SIL (the Summer Institute of Linguistics - http://www.sil.org/), is a system that has gained recognition among field linguists. However, the material gathered, including those stored as digital resources, are in many cases not accessible to a wider research community, as there is often little incentive and no means (with few exceptions, like the ‘Ailla’ databank (http://www.ailla.utexas.org/site/welcome.html) for the Indigenous Languages of Latin America) for making these resources available other than for individual publication. Forms of data presentation are likewise highly dependent on the individual linguist. Among attempts at standardization of glossing conventions, a well known effort is for example the Leipzig Glossing Convention (http://www.eva.mpg.de/lingua/ files/morpheme.html) or the Gold Initiative under EMELD ((http://emeld.org/ index.cfm). Still, as can be observed in theoretical writing as well as in raw data collections, open class items are in practice often glossed by corresponding words in English (or French) without explicit indication of what the word category in the relevant language really is, while closed class items are described using sets of abbreviations that vary from language to language, and from linguist to linguist. Several problems arise: To start with, we do not have an 153 Beermann and Prange international glossing convention which satisfies the level of standardization needed for digital language documentation, and efforts to develop gold standards for glossing and annotation sets often do not penetrate to the linguistic community as a whole. Although linguistic publications generally observe certain glossing standards, they often do not have the necessary level of explicitness needed. Secondly, what has emerged as a standard for the glossing of the well-studied languages within the Indo-European language family, and which has inspired automatic annotation tools, is not sufficient for most other languages. 3 What is TypeCraft(TC)? We start with a short first description of TypeCraft (TC) and give some background information, we then discuss its main features in detail. 3.1 A short description of Type Craft TC is a database with a user interface that allows linguistic annotation of sentences, featuring a two line glossing facility, as exemplified below: Sami : Locative deixis Bievden li girje Bievden li girje bievdde -n li girje table -INE.SG be.3PL.PRS book.NOM.PL N COP N There are books on the table Contribution by: Kristin Lindbach TypeCraft token reference: 1158842886-Sami-Lule Sami-Kristin Comments: In predicative constructions, the order of the NPs expressing 'location' and 'locatee'(the thing located) effects the interpretation of the 'locatee': If the 'locatee' follows the copula, its interpretation is indefinite (existential). (Cfr. possessive constructions: Máhtun li bednaga) < Figure 2. A TC-token - Lule-Sami 'There are books on the table' The string in bold in figure 2 corresponds to the script field in the annotation interface, while the first line in the table represents its Latin transliteration, which in Sami is identical to the script. The next line indicates morphological boundaries, followed by two lines of glossing. The first of these is reserved for translational glosses for the open class items and functional gloss symbols for the closed class items. The second gloss line contains parts of speech information. We have chosen a two line glossing, since, according to our Annotating and sharing language paradigms online 154 experience, one line glossing fosters misleading annotations. For instance, in a one line glossing, open class items are given as translational glosses only, which means that possible categorical mismatches will not be reported. The contributor line below the example is derived from one of the metadata fields in the annotation interface. Figure 2 shows the view of a language token when displayed as a search result. The search parameters, that is ‘language’ and ‘construction’, are displayed as headings of the token as a whole in the left corner above the token, while the dialect name, another part of the meta information, is displayed as part of the reference number. 3.2 First Phase of TC TC21 was originally developed from a masters level data documentation course. Its goal was two-fold: to make annotated linguistic data (in this phase mainly from the Volta-Basin languages of West-Africa and from the Semitic languages of East-Africa) accessible to a wider research community, and to provide basic education in language documentation to students. To begin with, data collection was done with the help of the The Field Linguists Toolbox. Born from the need to have an exchange platform that was more generally accessible, TC was started; we loaded our toolbox project files to a database and continued annotating online. 3.3 What is a good annotation tool? How must an annotation tool be designed to foster standardization of glossing, linguistic typing of language paradigms and the communication between linguists that work with this tool? Secondly, how can the design of an annotation tool contribute to the publication of authentic language data with annotations that are linguistically convincing? To our mind annotation must be effortless.22 An annotation tool must be as simple as a good word processing program, the screen must be like a piece of paper, automatic processes should be visualized and choices should be possible on the basis of a one screen design. Let us look at the editing interface of TC in figure 3 in detail: Sentence strings are entered directly into the annotator's browser, once in original script, once in a Latin transliteration. The annotator assigns the morpheme boundaries (there is no automatic word parsing as in Toolbox) and then starts on the two-line annotation in the Editing field. Underneath the Editing field is a Parse field where the annotator can view his input interactively. TypeCraft has an inventory of annotation 21 In 2004 for the first time, education in data documentation took place in the Linguistics Department of the Norwegian University of Science and Technology in Trondheim as part of a seminar in language typology. One of the aspects we were particularly interested in was the typing of language data according to construction types. We discovered soon that a linguistically motivated categorization of language data is not an easy task, neither is the annotation of the individual tokens, taken seriously – so the name 'TypeCraft' was coined. 22 Failing to observe this fact may in fact deepen the so-called digital divide (cf. http://www.developmentgateway.org/?goo=147 and http://en.wikipedia.org/wiki/Digital_divide) rather than bridging it: sub-optimality of working infra-structure or over user-friendliness might simply render unmanageable tools. The need to download additional and too much navigation inside instructions is in our experience an additionnal ‘blocker’. 155 Beermann and Prange symbols separated into two types:23 parts of speech symbols and functional glosses; in addition the annotator supplies translational glosses. In the parse field, the latter will appear in a brown color, while the former (POS symbols and functional glosses) will appear in green, if used according to the TC standard. If a user enters a functional gloss that is not recognized by TC, the symbol will be parsed as a translation gloss, which means it will appear in brown rather than in green in the parse field. Thus the annotator knows that his gloss is not known to TC. In order to see which glosses are known he Figure 3. TC-interface for data annotation now can open a drop-down window as part of the same screen, which allows him to inspect available gloss symbols. If he is uncertain about the nature of the gloss symbol he can go to the HELP button on the same page, from where he can open a new window which will expose a browseable file that displays more information about the use of glossing symbols. (Different from similar resources in Gold, symbols are thematically grouped together and symbols that are used interchangeably are exposed; for example, symbols standardly used to gloss nominal inflection are discussed together, the same is true for glosses used to annotate morphological aspect.) This procedure also holds for POS symbols (for these, red is the color signalling that a symbol is not known to TC). Notice, however, that nothing will prevent the user from integrating an unknown symbol into TC. Instead of being prevented 23 TCs inventory of annotational glosses is derived from the Leipzig Convention, enriched by glosses found mainly in the theoretical literature on African languages, and mark-ups found in corpora, mainly the British National Corpus and in EMELD GOLD. Annotating and sharing language paradigms online 156 from using unknown symbols, the user in encouraged to communicate his annotational choices. Two tools are offered at present: (i) The annotator is encouraged to make use of the comment field at the bottom of his screen, where additional linguistic information can be communicated in text. If a user feels that, e.g., the first verb in a serial verb construction is neither a full verb nor an auxiliary and that we therefore should recognize a part of speech for example called preverb (PV), the comment field is the place for entering this information. In this way no information gets lost simply because existing standards have failed to make a distinction needed otherwise. (ii) The annotator can take contact with TC by pressing a button called FEEDBACK. At present this is a direct link to the TC administrator, but in the future this will be connected to other TC users (see section 4). Here he can ask for support, not only concerning the use of glossing symbols but any issue related to the annotation process. In summary, the tools described above are a beginning for making standardization of glossing a process supported by the community of annotators themselves and for making standards available without enforcing them. Let us now turn to quality insurance. Any token from the private domain of a TC user has to be approved by three parties: the peer, the administrator and the contributor himself. TC thus implements a workflow that hardwires three distinct user-roles in the TC domain. We assume that it is in the common interest of these three user profiles to keep TC data correct and valid. The contributor collects and submits the data to TC, he or she is also the owner of the data. Peers are other TC users as well as a group of people selected by TC to contribute with their experience as senior linguists to the quality of the annotations. For the Beta version of TC, we were able to engage two expert peers for the two of the language groups in TC, the Scandinavian languages and the languages of the Volta basin. As an expert peer the user status is granted with access to the private domain of users that work on data connected to the expert’s language group. The peer expert can take direct contact to the annotator via the TC internal messaging system, but he can also overwrite data entered by the annotator. However, tokens are never lost from TC, we keep all token in store and users can via the annotation screen look at the history of changes. Peers and administrator cannot publish data without the contributor’s consent. After changes made by any one of the three guardians of the data, the evaluation process has to run again through one full circle. We are as concerned with the owner rights of the contributors, as we are concerned with the quality of the published data. Although quality assurance is a costly feature because of its administrative cost, we at present see no other way to equally well assure the rights of the owner and the quality of the data. 3.4 Types of objects in TypeCraft In this section we would like to talk about the type of objects we store in our database. The goal of any form of language documentation is that the data be in some relevant sense representative for the language to be stored. 157 Beermann and Prange Present ways to achieve this goal are corpora, sound and video documents as well as lexical databases. From a linguistic point of view, the character of a language is also present in the system of construction types it allows, that is, which kinds of diathesis alternations are represented, which kinds of NP dislocations are allowed, whether we find control-structures, and so forth. As opposed to Toolbox which is optimized for the morpho- phonemic mark up of text and dictionary building, TC has been designed for the morpho- syntactic annotation of sentences and phrases of medium length. The system interprets these entries as tokens of a certain construction type. In the following we would like to describe what we understand under a 'construction type' and develop the notions constructional paradigms and constructional networks. 3.4.1 What is a constructional paradigm? The term construction is here used as a descriptive term to denote a set of sentential or phrasal properties. A construction in this sense is a set of grammatical parameters that conspire to form a sentence signature with some properties being less central to the signature than others. Languages on the whole differ in which sentential signatures they allow and how these signatures are connected. A constructional paradigm arises from different views of a network of construction types. cleft spatio-temoral modificatio impersonal passive periphr.passive morph.passive passive resultatives causative tr. verb frame impersonal middle intr. verb frame extraposition topicalization Figure 4. Partial description of a constructional network Let us take an example: Figure 4 represents a partial constructional paradigm for the fictional verb V1 of the language L1. According to figure 4, V1 may enter into a intransitive frame. Taking a procedural view, we may say that V1, from a intransitive Annotating and sharing language paradigms online 158 frame, enters into a transitive frame. On this view, lines in figure 4 correspond to lexical rules and while strings correspond to nodes in a constructional derivation. The verb V1, once it has entered a transitive frame, furthermore allows an impersonal passive as well as a causative subsequent to a passive; a causative can likewise apply directly to a transitive frame. Orthogonal to these options is the circumstance that passives in L1 can be alternatively built morphologically or as periphrastic constructions. Finally all of these constructions may at any time undergo spatio–temporal modification, and they may alternatively or at the same time be realized as clefts, extrapositions or topicalized structures. A constructional paradigm, on this view, corresponds to a view of the dependencies shown in figure 4. 3.4.2 How can we represent constructional dependencies in TC? Consider the following quote from Senft 2002 Let me start with the following practical example: Suppose I am interested in the following two linguistic topics: serial verb constructions and nominal classification systems: I want to get as much information as possible from those languages that are documented in an on-line archive. If such an archive would be ideal, I could do the following search and get the following kind of information and data. I visit the website… and find a SEARCH function … I type “serial verb construction” and “nominal classification”. The search machine presents me the results of the search listing the languages and the files.” (Senft 2002, page3) Tokens in TC are typed according to which language and which construction they represent; a SEARCH in TC will allow you to specify the language as well as the constructions you are interested in, thus fulfilling one of the desiderata given in the quote above: Figure 5. SEARCH for Language and Construction 159 Beermann and Prange A different, more difficult, task is how we can indicate, for each annotated sentence at the time, its constructional dependencies or correlates. Different annotators of TC have found different solutions to this problem. For example, some annotators have provided tokens belonging to the same constructional paradigm with the same comment in the comment field. The effect is that for each individual token, the main constructional properties are repeated as an additional comment. Here is an example of this pattern: Figure 6. Complex Passives in Norwegian – a token as part of a constructional paradigm A further attempt to preserve construction type information is the naming of the construction (done in the construction field); subtypes of constructions have been indicated via hyphenated names, as exemplified in the following.: Figure 7. Subtypes of constructions indicated by construction name Finally a third way to indicate constructional dependency is to introduce a cross reference marker in the in the comment field of one example thus indicating the dependency to another token in the base. Annotating and sharing language paradigms online 160 4 The Future of TypeCraft This paper describes the beta version of TC and, relative to the desiderata described above, TC still faces many challenges. The communication between TC users is at present only implemented between the administrator and the users: points of general interest can not be made accessible to all TC users, thus that users working on similar constructions or encountering similar problem cannot benefit form prior discussion. We hope that a forum connected to TC will remedy the problem. An evaluation of the expert peer review system is at this point too early. The beta version has 35 users, senior linguists as well as students. Data sets are slowly emerging, but they still have to mature in order to be of interest to expert reviewers. The implementation of constructional paradigms is another challenge and we showed above that users of TC find individual solution to the problem. As part of the TC design we work at present at the visualization of networks, and the visualization of the position that a TC token has within a constructional network. Finally, TC has emerged from a North-South cooperation project between NTNU and the University of Ghana, sponsored by the Norwegian institution NUFU.24 Such cooperative projects are likely to constitute an essential developmental environment for TC. For its sustained activity, support from its home department provides its basic funding. 5 Conclusion As a response to the problems of data preservation, glossing standardization and general accessibility of data, TC offers a public access point to linguistically annotated natural language data. TC as an annotation tool seeks to establish a community site with routines that can serve in ensuring the quality of the data presented and that will allow the user to trace information back to its source. In an additional effort, TC at present focuses on improving and extending standards in glossing. The basic data type in TC are annotated sentential and phrasal strings, representing a certain construction typ. The implementation of the construction type is one of the immediate challenges that TC faces. A second line of development concerns an application that we hope to offer to linguists who would like to use the editing tools offered by TC, but who partially work without net access. References Dorothee Beermann, Lars Hellan and Jonathan Brindle. TypeCraft: a natural language database; paper presented at the Legon-Trondheim Linguistics Project Meeting in Accra, January 11, 2006. Bird Steven & Gary Simons 2003, Seven Dimensions of Portability for Langugage Documentation and Description. Language 79. 24 Cf. www.siu.no. The project is entitled "Computational Lexicography, Typology and Adult Literacy", and also known as the 'Legon-Trondheim Linguistics Project'. 161 Beermann and Prange Senft 2002c "What should the ideal online-archive documenting linguistic data of various (endangered) languages and cultures offer to interested parties? Some ideas of a technically naive linguistic field researcher and potential user", in: Peter Austin, Helen Dry, and Peter Wittenburg, eds. Proceedings of the International LREC Workshop on Resources and Tools in Field Linguistics, Las Palmas, 26-27 May 2002. 15-1-15-11. Las Palmas: European Language Resources Association. http://www.mpi.nl/lrec. Links Summer Institute of Linguistics: http://www.sil.org/ Leipzig Glossing Convention: http://www.eva.mpg.de/lingua/files/morpheme.html EMELD project: (http://emeld.org/index.cfm) Authors’ e-mail addresses: dorothee.beermann@hf.ntnu.no atle.prange@businesscape.no Working Papers 3/2006 isk Contents: Metaphor. How an analysis of metaphors can shed light on the relationship 5 between implicature and explicature Randi Waade On the functional independence of explicatures and implicatures 15 Thorstein Fretheim Anaphoric and non-anaphoric uses of the Norwegian adverb først ('first'): A 27 pragmatic analysis based on a univocal lexical meaning Thorstein Fretheim The metarepresentational use of main clause phenomena in embedded clauses 41 Thorstein Fretheim A relevance-theoretic analysis of UNLESS 59 Thorstein Fretheim Explaining connections in Akan discourse: the role of discourse markers 85 Nana Aba Appiah Amfo ‘Abroad’ and semantically related terms in some European languages and in 107 Akan (Ghana) Thorstein Fretheim and Nana Aba Appiah Amfo A constructional approach to syntax and the treatment of passive in 125 Norwegian Petter Haugereid Diagnostic use of nonword repetition for detection of language impairment 139 among Turkish speaking minority children in Norway Olaf Husby Annotating and sharing language paradigms online 151 Dorothee Beermann and Atle Prange