语义韵最早的文献Semantic_prosody_Bill_Louw_1993

Irony in the Text or Insincerity in the Writer? The Diagnostic Potential of Semantic Prosodies Bill Louw University of Zimbabwe Abstract One consequence of the advent of large corpora has been their increased potential for reveal­ ing consistencies in the influence of collocation on the behaviour of particular linguistic forms. Irony relies for its effect on a collocative clash which is perceived, albeit subliminally, by the reader. In order for a potential collocative clash to attract the ironist's interest, there must be a sufficiently consistent background of expected collocation against which the instantiation of irony becomes possible. A consistent aura of meaning with which a form is imbued by its collocates is referred to in this paper as a semantic prosody. Semantic prosodies have been largely inaccessible to human intuition about language and they cannot be retrieved reliably through introspection. Evidence is emerging that departures in speech or writing from the expected profiles of semantic prosodies, if they are not intended as ironic, may mark the speaker's real attitude even where s/he is at pains to conceal it. This paper explores the ques­ tion of the diagnostic capability of semantic prosodies and suggests what role they may play in the composition and grading of suasive language and in 'radical stylistics' in the future. Happy (trying to quiet Willy): Hey, Pop, come on now ... Willy (continuing over Happy's line): They laugh at me, heh? Go to Filene's, go to the Hub, go to Slattery's, Boston. Call out the name Willy Loman and see what happens! Big shot! (Arthur Miller, Death of a Salesman, Act I) 0. Introduction As irony is to form the subject of this paper, one can perhaps be forgiven for allowing certain agreeable ironies in the history of stylistics to contextualise this study. In 1970, Donald C. Freeman edited a collection of articles under the general title Linguistics and Literary Style. In his brief introduction, Freeman 158 BILL LOUW (ibid: 6) alludes to Bernard Bloch's definition of style and criticises it: Another difficulty in the work of the 'style as deviation' school of linguistic stylistics is its definition of the norm from which an author's style is sup­ posed to differ in certain ways. For example, Bernard Bloch defines style as 'the message carried by the frequency distributions and transitional prob­ abilities of [a discourse's] linguistic features, especially as they differ from those of the same features in the language as a whole. ' This definition is a chimera. The frequency distributions and transitional probabilities' of natural language are not known and never will be, and even if they could be ascertained they would constitute no particularly revealing insight into either natural language or style. (emphasis added) Ironically, all the supposed impossibilities on this fairly comprehensive agen­ da have been made possible in the past twenty-two years by the author of a paper written in 1966 and reprinted in Freeman's collection, Professor John Sinclair. Sinclair's research in lexis and collocation over the 26 years can now furnish evidence that the description of vocabulary patterns is both necessary and possible. He writes, for example (1987; 1991) of the patterning associated with the verb happen and the phrasal verb set in. Happen is habitually as­ sociated with unpleasant events. The irony in the quotation from Arthur Miller with which this paper begins hinges on the fact that the reader knows that an evaluation like Big shot! is unlikely to follow See what happens! Of set in, Sinclair says the following (1987:155-156): The most striking feature of this phrasal verb is the nature of the subjects. In general they refer to unpleasant states of affairs. Only three refer to the weather; a few are neutral, such as reaction and trend. The main vocabulary is rot (3), decay, malaise, despair, ill-will, decadence, impoverishment, infection, prejudice, vicious (circle), rigor mortis, numbness, bitterness, mannerism, anticlimax, anarchy, disillusion, disillusionment, slump. Not one of these is desirable or attractive. The subjects of set in are also ... largely abstractions: several are nominalisations of another part of speech. This is effectively the first computationally derived 'profile' to appear in print of the phenomenon which Sinclair has begun to refer to as a 'semantic proso­ dy' (personal communication 1988), applying the term 'prosody' in the same sense that Firth (in Palmer 1966:40) used the word to refer to phonological colouring which was capable of transcending segmental boundaries. The nasal prosody in the word Amen would be an example: we find that the vowels are imbued with a nasal quality because of their proximity to the nasals m and n. THE DIAGNOSTIC POTENTIAL OF SEMANTIC PROSODIES 159 In the same way, the habitual collocates of the form set in are capable of col­ ouring it, so it can no longer be seen in isolation from its semantic prosody, which is established through the semantic consistency of its subjects. It is difficult to establish whether the existence of this phenomenon was perceived by traditional semanticists before the advent of linguistic comput­ ing. Bréal (1897) seems to have come closest to understanding it. He referred to transference of meaning which is the product of habitual collocation as 'contagion'. Ullmann (1963:185) refers to Bréal's explanation in these words: Among its widespread semantic ramifications, Latin 'persona' has become a negative particle in French. This change is due to purely linguistic rea­ sons: contiguity with the negative particle ne has 'infected' this word, in the same way as rem, passum, punctum, by a process which Bréal has termed 'contagion'. Bréal, as the term 'contagion' suggests, seemed to sense that there is in the language a preponderance of pejorative examples of this phenomenon. A fur­ ther detail in which Bréal and Ullmann were correct is that 'contagion' is a general linguistic phenomenon which pervades every type of language. In other words, no amount of genre-based or register-based study in this century could ever have revealed its presence. However, Bréal offered too few ex­ amples for us to infer that semantic prosodies, and in particular their potency and proportions, might be accessible to our intuition. They are essentially a phenomenon that has been only revealed computationally, and whose extent and development can only be properly traced by computational methods. 1. Semantic prosodies in Larkin's poem's 1.1 Semantic prosodies in Larkin's 'First Sight' Let us turn now to consider a semantic prosody Philip Larkin's poem First Sight, which Sinclair analysed in 1966, in the paper referred to above: First Sight Lambs that learn to walk in snow When their bleating clouds the air Meet a vast unwelcome, know Nothing but a sunless glare. Newly stumbling to and fro All they find outside the fold, 160 BILL LOUW Is a wretched width of cold. As they wait beside the ewe, Her fleeces wetly caked, there lies Hidden round them, waiting too, Earth's immeasurable surprise. They could not grasp it if they knew, What so soon will wake and grow Utterly1 unlike the snow. The last line is particularly interesting, as is illustrated by the concordance for utterly drawn from the original 18 million word corpus at Cobuild. There are 99 citations altogether, and every third one is reproduced here: 1 nothing. The farmers were utterly against the union and utterl 2 rly against the Union and utterly against the Wages Board. Now 3 f it everything seemed so utterly altered that I felt illogica 4 But Io has no wind and is utterly arid. Most likely this erosi 5 rst thing we'd ever seen, utterly blackened now, the skin on t 6 in his diary: "Whitehall utterly burned to the ground, nothin 7 In my experience it gets utterly confused. And there are seve 8 s ambitious wife, are not utterly convincing. Miguel Fernandes 9 nfident, well-trained and utterly dedicated to the idea of win 10 feet. Its hopes appeared utterly demolished in 1956, when Mr 11 outwards from the centre utterly destroying everything in its 12 the island the view was utterly different. The filmy enchant 13 . I think it would be an utterly different kind of programme 14 probability's sake, not utterly disconfirming the tale of a 15 ng which could be called utterly effete, self-centred and poi 16 them. The idea was to be utterly exhaustive, to examine the n 17 how kind he was, oh how utterly good and what trouble he too 18 ivor which worked for me utterly. I mean I'm not really inter 19 to reveal a man who was utterly insensible to the gravity of 20 monies latent in her was utterly Jumbled by the clumsiness an 21 s listeners but which is utterly meaningless to himself and t 22 iver. When you think how utterly obsessed by self-denial Engl 23 our not uniquely evil or utterly out of line with custom and 24 oops would have found it utterly ridiculous. Dean had worked 2 5 e, but only if they feel utterly secure, and this is a contra 26 er words, of course, are utterly stupid, "a hard day's night" 27 he nearest town. "I'd be utterly terrified to go up, wouldn't 2 8 lways making? We fail so utterly to reward responsible marrie 29 ore 1914. Little of this utterly transformed scene can be tra 3 0 and illusion, then it is utterly unreasonable to suppose that 31 rrassment. Posy had been utterly unsympathetic when she heard 32 d I feel I fail to do so utterly. Why? The stipulation of the 33 d of ours and destroy it utterly. "You can work miracles", sa If we study the collocates to the right of utterly we find a phenomenon similar to that identified for set in. The concordance shows that utterly has an over­ whelmingly 'bad' prosody: there are few 'good' right-collocates. Indeed a 'bank' of the bad ones would be easy to establish and would be bound to have its own unique profile. THE DIAGNOSTIC POTENTIAL OF SEMANTIC PROSODIES 161 If one applies this knowledge to Larkin's poem, one discerns at once that the poet's intuition was keenly aware of the presence of a bad semantic proso­ dy on utterly. The catch-all word unlike after utterly in the closing line of the poem allows the cumulative force of the bad prosody, in all its disagreeable variety, to be mustered from our prior knowledge as readers and then ranged against the deceptive homogeneity of the new born lambs' first problem ever: snow's vast unwelcome. The poet is content to hint at the myriad cruelties of the world the lambs have just entered and the burden of that hint is borne by the word utterly. Sinclair (1966:141) comments on the power of the line Utter- ly unlike the snow in critical language which the evidence of semantic proso­ dies can now corroborate: "It is the traditional 'brush off' structure: 'Some­ thing I'd prefer not to talk about' 'Nothing you won't know all about in time'". Sinclair makes the point that any attempt to spell out these sinister im­ plications would of necessity be couched in language which paints in too much context. In other words, in order to bring a prosody to the fore, it would have to be laboured upon, and this would be unacceptable for the same reason as the "rarity, in love poetry, of lines like: Her smile was not in the least like the grin of a decomposing vampire" (ibid: 141) Let us suggest another prosody in the poem: Larkin's depiction of wak­ ing. As corpora at Cobuild have grown since the first 18 million word corpus, we find that we are able to extract examples of the sinister sense of wake followed by and, as used by Larkin. The following citation is drawn from a more recent 37 million word written corpus, now incorporated into the Bank of English:2 ... ancestor lying asleep with his head on his arm who was one day to wake and gobble up everything that lives. It is becoming clearer that the great 'sleeping giant' in the application of Sinc­ lair's work to stylistics lies in matching texts against corpora: this supplies the 'bottom up' textual evidence for what has until now been regarded as 'top down' prior knowledge in the act of reading (Louw 1989). 1.2 A semantic prosody in Larkin's 'Days' Sinclair's stated position has long been that the pursuit of independent word meaning has been as illusory as it has been sustained. Testimony to this can be readily demonstrated by concordancing for two or more consecutive words chosen from a literary text, rather than just one word, as is often the practice. 162 BILL LOUW The correlation between repeatable aspects of text process and repeatable events outside the text is far more striking than our intuition could ever have predicted. Here is another example, this time from Larkin's poem Days: Days What are days for? Days are where we live. They come, they wake us Time and time over. They are to be happy in: Where can we live but days? Ah, solving that question Brings the priest and the doctor In their long coats Running over the fields. The line Days are where we live purports to offer happy associations, but it is a line which leaves the reader with inexplicable feelings of melancholia. It appears to foreshadow the theme of death with which the second half of the poem becomes preoccupied and from which it never escapes. Whatever it is in the reader's prior knowledge that accounts for this, it must be verifiable as a piece of text process in our experience of the whole language. In more than two thirds of the following concordance from the 18 million word corpus, days are is followed by words like gone, over, and past. Days are not so much where we live as where we have lived and where we are likely, possibly soon­ er rather than later, to die: 1 t it yourself the prices these days are absolutely astronomica 2 ite 'The world is wide, no two days are alike, nor even two ho 3 ays are gone whenel. But those days are almost twenty years go 4 gless extinction when the grey days are done but who are reaso 5 o men for unequal pay. But the days are gone whenel. But those 6 or do I. The big beer drinking days are gone. They drank becau 7 nd cry for peace. My political days are good and over. I'm not 8 ople making these things these days are making money out of th 9 erage trawler when its fishing days are over - as the Morning 10 it was before. Those good old days are over because trout fis 11 Lourenco Marques. Alas those days are over. What did he die 12 o walk means that his babying days are over. The stroking cea 13 ade me regret that my dancing days are over. Rudolph couldn't 14 fate of Czechoslovakia. These days are over, and that is what 15 ng after me Grandad's working days are past walk along with m 16 a black black sky. But those days are rare and usually to be 17 f I had a striking clock. The days are stretching out again a 18 ness and constancy of country days are the very qualities tha 19 e the only movies I see these days are these nights, on the 1 20 finances of old people these days are very much better than 21 rate that situation. The hard days are with us and they are c THE DIAGNOSTIC POTENTIAL OF SEMANTIC PROSODIES 163 The proportions of this semantic prosody can be further verified from the larger 37 million word corpus. In it the term days are appears 104 times and corroborates the profile which emerges from the concordances given above. 2. Irony and semantic prosodies The remainder of this paper has four objectives: (i) It will seek to demonstrate the role of semantic prosodies in making certain forms of irony possible. (ii) It will argue that if irony can be shown to be a phenomenon foregrounded by virtue of the fact that it runs counter to a semantic prosody, then this must provide one criterion for the existence of the semantic prosody in question. (iii) It will argue that where forms are retrieved which run counter to a seman­ tic prosody but which cannot be shown to have a deliberate ironic intention, an alternative explanation might be inferred from the utterance's context of situa­ tion. It may be possible to uncover details about the attitude of speakers or writers which indicate that their utterances are insincere. The paper will pro­ pose that semantic prosodies, used in this way, help to resolve some of the problems which Grice (1978:124) poses about irony, and thus they could eventually make a contribution to the debate on felicity conditions or prag­ matics in general. Where a speaker or writer's insincerity cannot be demon­ strated from the context, other hypotheses may be entertained, for example that the speaker is young or is not using his or her first language . (iv) It will suggest that semantic prosodies are a powerful component in sua- sive writing and that they might be pressed into service for the purpose of producing data-driven suasive writing. For example, this section will propose that semantic prosodies can be used to grade advertising copy in order to assess its 'sincerity'. It is now worth returning to the concordance for utterly (see section 1.1 above) with some of these objectives in mind. If one moves through the con­ cordance lines in search of exceptions, i.e. any apparently 'good' collocates instead of 'bad' ones to the right of utterly, one finds only four in the entire KWIC concordance drawn from the original 18 million word corpus:3 1 nfident, well-trained and utterly dedicated to the idea of win 2 h how kind he was, oh how utterly good and what trouble he too 3 it up - "I think it's oh utterly grand of you to give us all 4 t. Edward III, placid and utterly venerable, his face flowing These citations of utterly are an excitingly clear example of Mukarovsky's (1932) notion of foregrounding, demonstrated computationally. All of the 164 BILL LOUW forms listed here carry a fairly obvious ironic intention. Citation (1) {confi- dent, well-trained and utterly dedicated) may appear at first to be a dubious candidate, but the wider context, when accessed, makes it plain that the sub­ jects are Nazi storm-troopers and that their 'profession' is extermination! Cita­ tion (2) is taken from a telephone conversation in the spoken corpus in which the participants are criticising someone else. The irony in (3) is self-evident and the irony in (4) becomes more evident once one realises that the reference is to the king who is, at the time of utterance, dead. Examples of this kind begin to assist us in determining criteria for recog­ nising semantic prosodies. It is only because the prosody on utterly is as con­ sistent as it is that it admits the possibility for irony. Prosodies are undoubt­ edly the product of a long period of refinement through historical change and even though new prosodies may be in the process of being formed, they can­ not be used for the purpose of instantiating irony until their prosody predomi­ nates sufficiently strongly. It will be obvious so far that semantic prosodies must furnish one of the most compelling arguments for building and using larger and larger corpora. Their precise characteristics will only be revealed and be seen to stabilise once the content of a prosody can be collected extensively. The nature of the ex- ponency of irony through exceptions to a prosodic trend will, equally, only begin to gel once one has a reasonably large collection of what is a fore­ grounded phenomenon and therefore an extremely rare one. The original 18 million word corpus has disclosed only the tip of the iceberg in indicating the prevalence and extent of prosodies in the language. The investigator cannot but yearn for more and more evidence as the Bank of English increases in size and generic range. Consider for example the following short passage f