Types of intertextuality in Chairman ’ s statements

Intertextuality is a pervasive feature of all disco urse, but norms and conventions vary widely across domains. Academic conventions ca n use difficulties for those who have been exposed to, or move on to, doma ins with other practices. Academic conventions are well documented; here we e xamine those of business writing. We created a corpus of chairman’s statemen ts from annual corporate reports and searched them for signalled and unsigna lled intertextual relationships. We hypothesise that statements from the same compan y will be linked by both repeated phraseology and acknowledged intertextuali ty.


Introduction
Intertextuality, the idea that texts are made of other texts, has been a commonplace since Kristeva (e.g., 1980) and Bakhtin (1986).We only know what to say and how to say it because we have heard or read what others have said or written.But intertextuality comes in many different forms and different discourse communities use it for different functions (Scollon 2004).Teachers of language for specific purposes (LSP) must facilitate their students' acquisition of the communicative practices of their target communities.However, this can be problematic if they differ substantially from the practices of the academic community.Some of the problems that may arise are due to the range of textual practices that students are familiar with from the lifeworld or from other domains, and that they bring with them into whichever new domain the LSP teacher is trying to introduce them to.In many ways the practices to be learned conflict with those which students have been exposed to in school or have observed in other, visible and public domains such as journalism.The delivery of adequate LSP instruction depends in part on the teacher's awareness of such potential conflicts.This is, naturally, true of a range of language features; intertextuality is a case in point.
Intertextuality has notoriously been defined in a number of ways.One characteristic often associated with it is polyphony (Ducrot 1984, Nølke 1994, Fløttum 2004), the recognition that texts contain different 'voices' encoded in various ways.However there are intertextual links among texts that have not been captured by investigations of polyphony, and there are polyphonic features that have nothing to do with relations between two texts.Fairclough (1992) provides a linguistic typology of what is called discourse representation, a concept close to Ducrot's polyphony.Discourse representations are divided at the first level between boundary-unmarked and boundary-marked.Boundary unmarked discourse representation covers presupposition, negation, metadiscourse and irony, cases where the author adopts or implies another voice than their own in the text.Boundary-marked ('explicit') discourse representation is subdivided into direct quotation, indirect speech, and the use of scare quotes.
We would, however, accept Hohl Trillini and Quassdorf's (2010: 272) still broader definition: 'intertextual processes involve, minimally, an earlier and a later text and an element from the former that is discernible in the latter'.Further we would argue that in non-minimal cases there may be multiple earlier texts contributing a particular element to a later text.Given this definition, it is possible that not all 'discourse representation' is intertextual, and that neither is all intertextuality polyphonic.Fairclough's (1992) unmarked categories, particularly negation, presupposition, and irony, do not in general relate to actual identifiable other texts; in fact they construct other voices independent of real texts.Similarly, a text in a given genre or register will have phraseological likenesses to another text in the same genre or register (e.g.Wray 2002Wray , 2006) ) which are not intended to evoke any other voice and in fact are intended, if there is any intention, to confirm that this text expresses the same collective voice as the others.It is debatable whether this would be regarded as polyphony.However, Hoey (2005) argues that our awareness of formulaic language as well as genre and register conventions, comes from the fact that we have been repeatedly exposed to conventional forms of expression and thus are 'primed' to produce them.In that sense, the influences of the earlier texts which are the sources of exposure make these language features intertextual.
Investigations of intertextuality examining both specific earlier and later texts have included Pecorari (2008) and Hohl Trillini and Quassdorf (2010).Pecorari focused on academic writing and the intertextual ties between dissertations and theses, and their sources.Hohl Trillini and Quassdorf used a large database of literary, literary-critical, and other 'later' texts all referring to the same 'earlier' text, Hamlet, and derived categories from the relations they found.In both studies marking and modification emerge as central criteria.Marking for intertextuality, the latter found, can be done by any or a combination of the following: the name of an author or work, a 'verbum dicendi or other metalinguistic marking', a typographical device, a textually implicit marker such as a syntactic anomaly, register mix, or anachronism, the receiving text genre (such as an anthology), or 'context' or of course there can be no marking (Hohl Trillini and Quassdorf 2010:280).Modification covers the degree of verbal or other identity between the two linked texts, on a continuum from quotation, in which the relationship is both lexical and semantic, to paraphrase, in which the only relation is semantic.
In Pecorari and Shaw (2012) we sketched a typology of intertextuality for applied language studies, based on the idea of identifiable similarity across texts rather than polyphony within a text.It is based on three primary criteria which emerged in interviews with experienced academic writers about the intertextual relationships they identify in student writing.These are the retrievability of a particular target text, conformity to community norms in respect of modification and marking, and writer intention.
Our first category, which we called indirect intertextuality, covered a range of features found in numerous earlier texts, rather than a single specific source.In some cases there is no intention to evoke a separate voice, as with repetitions of language in discussing the same topic or realizing the same genre.In others the earlier texts are not specific ones; an example is the productive expression 'x is the new y', as in 'pink is the new black' or 'sleep is the new sex'.Even if a specific origin for the expression can be identified (and O'Connor, 2007, does so), it is ordinarily used with an awareness of the fact that it is a commonplace expression, and without an awareness of a specific origin.
Our second category was conventional intertextuality, covering acknowledged references and deliberate allusions which conform in terms of marking and modification to the demands of the community of practice within which they are produced.Our last two categories were (3) unconventional and (4) deceptive, where content or language is borrowed without prescriptively required acknowledgement and, respectively, without or with an intention to deceive.Understanding these categories requires an understanding of the demands and expectations of the community in question, so that these conventions become a case in point for the sort of awareness of textual features required of an LSP professional.
This is all the more necessary because conventions vary across domains.Using Fairclough's categories, Scollon (2004) compares discourse representation, more or less what we call conventional intertextuality, in academic writing, newspaper reports, and advertisements.He finds that the same categories can be applied across what he calls communities of practice, (adopting Lave and Wenger's 1991 terminology), but there are quantitative and qualitative differences.Academics use a very wide range of discourse representation types, with rather little direct quotation.Journalists use direct quotation very much more than academics, but basically restrict their evocation of other voices to marked forms-direct quotation and indirect speech.Advertisers use a very wide range of representations of fictional and fictionalized discourse but do not quote the verbatim utterances of nonfictional individuals.These differences, Scollon argues, are due to discourse representation serving widely different functions in the different communities.He gives the three communities their own voices: The academic's position will be the most familiar to many LSP teachers, but as we said initially, it is not the position most commonly encountered outside the academic community.This is significant for two reasons.First, new entrants to that community may be expecting something quite different from what they find and need help with adapting.Second, students completing their studies need to be aware that academics' uses of intertextuality are not likely to be what they encounter outside the academy and they need to be alert for signs of expectations in working life.And to the extent that LSP instruction is intended to equip students with the ability to engage in communicative practices in specific domains, the intertextual practices of those domains may well be an element of course content.We therefore need descriptions of intertextual conventions in a wider variety of public fields, so that teachers can understand and respond to students' needs.
In this article we examine two aspects of intertextuality in a highprofile business genre with very different conventions from academic writing and from the other text-types examined by Scollon (2004), using a corpus approach and the definitions and framework just presented.We have chosen the genre of the chairman's statement (CS) within the company annual report.This is a well described genre whose intertextuality has not been investigated, and exemplifies a discourse which students may well need to use.
Annual reports are documents which are generally agreed to include instances of several different genres (de Groot 2006) with different registers closely linked intertextually).Chairman's statements (or management's statements) are identified by de Groot (2008) as components of annual reports alongside 'corporate profile ' and 'operational review' as other component genres. DeGroot (2008:73) identifies the topics of the Chairman's statement as "personal opinion about (financial) result, management situation, future outlook", its aims as "offering an informative and parental top-line overview of results, contextualizing information in succeeding sections, providing the company with a personal face, establishing reader-writer relationship" and its expected readers as a broad audience with a focus on shareholders.
A key function is impression management (Goffman 1959;Clatworthy and Jones 2006), so chairman's statements are interested texts like house agents' details as opposed to surveyor's reports (Shaw 2006), and are read with the knowledge that they are interested.Like other interested texts, the component genres of annual reports make use of a variety of different discourses (Bhatia 2004), but chairman's statements consistently use what Bhatia calls a 'public relations discourse'.They are public documents and thus may be read or written with other exemplars of the genre from other companies in mind, and can be expected to share their discourse.They are carefully produced, via multiple drafts, with multiple actual authorship, even if there is an individual nominal author (Davison 2011).Chairman's statements are members of (annual) series and thus could be expected to have conventional diachronic intertextual links with previous members of series.
Our aim here is to investigate intertextuality in this genre, using a corpus approach.We use the ample documentation of intertextual links in academic writing (Thomson and Ye 1991;Charles 2006;Swales 1990Swales , 2004;;Thomas and Hawes 1994;Hyland 1999) as a heuristic for the investigation of links in this promotional business genre, where we expect a different pattern.Thus we ask questions about the form of reference to the source, the extent of self-citation, the balance of quotation and paraphrase, etc, as well as about the number of shared ngrams and the implications for characterizing the discourse.The variety of intertextual link types revealed by Hohl Trillini and Quassdorf (2010) means that it is impossible to attempt to describe all types in a text collection of any size.We have selected two types for study, one marked and one unmarked.
Given the care with which large companies produce their annual reports, and particularly this key genre within them, the pattern revealed will be the one which is conventional in the genre.Our interest is in the way in which this pattern in business writing relates to the contrasting conventions of the academic domain.For example, templates, allowing the same message to be given in the same words on different occasions, are natural in many domains (e.g., the tax accountants examined by Flowerdew andWan 2006, 2010) but their acceptability in academic writing is highly contested.At the same time CS need to suggest something unique and essential about the given company.The focus of our investigation is the extent to which these high-stakes, highly crafted, but also very uniform documents make use of or avoid similar language.

Methods
Both quantitative/corpus and qualitative/discourse analytical methods were used.A corpus was compiled consisting of chairman's statements from the annual reports of 36 companies, most of which were listed on the London FTSE 100 as of 15 July 2012 (a full list is available in Appendix A).The statements were gathered from the year 2000 onwards, though because the availability of past reports varied, not all years are represented in the corpus for all companies.The corpus consists of 251 statements and just over a quarter of a million words.
An integrative approach (Charles, Pecorari and Hunston, 2010) of corpus and discourse-analytical methods was used.Marked direct reference was investigated by close reading followed by corpus search.First, some twenty statements from different companies were read through carefully and notes were taken of all intertextual links marked at Hohl Trallini and Quassdorf's highest levels: name, metalinguistic lexical item, and typography.The items that identified the intertextual element were then searched for in the whole corpus.These items included reporting verbs, and the names of specific texts, such as 'last year's report'.This search threw up a large number of examples and reading through these suggested further lexical markers that could be searched for.A new list of markers was drawn up and a second search produced a set of KWIC lines representing a high proportion of the instances of marked direct intertextuality in the corpus, which was then analysed in terms of form of marking and frequency.For the purposes of the corpus investigation we did not attempt to analyse noun uses (although, as noted above, we used nouns as search terms) and the focus of our quantitative study is on instances of intertextuality with reporting verbs.
Intertextual relationships among the statements were searched for by means of identifying n-grams, that is, strings of words of length n which appear in more than one text, extracted with the AntConc concordancer (Anthony 2007).Because the statements contain a great deal of financial information, the process of cleaning the corpus included standardising the use of symbols and words for 'dollars' 'pounds', etc., and substituting numbers with a placeholder, so that phrases such as the one in extract 1a and 1b could be indentified.
Extract 1a: The Board is recommending a final dividend of 3.35 pence per ordinary share (Aggreko 2002, p. 11)1 Extract 1b: Board is recommending a final dividend of 3.45 pence per ordinary share (Aggreko 2002, p. 4) N-grams are not necessarily indicative of a template writing strategy.They have normally been investigated (e.g., Biber, Johansson, Leech, Conrad & Finegan, 1999;Ädel & Erman, 2012) to make rather the opposite point, namely that the very frequent ones, while represented orthographically as multi-word units, have some of the characteristics of single lexical items.However, lexical bundles are relatively short units; most frequently studied are 3-and 4-word bundles, which according to Biber et al.'s (1999) criteria must occur ten times per million words and across five separate texts to qualify for bundle status.Five-word bundles are so much rarer that the frequency criteria are relaxed.If significantly longer n-grams are found, then that is likely to be indicative of copying, rather than phraseological status for those units2 .
The corpus was searched for n-grams occurring at least twice, from 100 words (the maximum permitted by AntConc) to 30 words.Since longer n-grams contain shorter ones, as longer strings were identified, the statements in which they appeared were removed from the corpus, so that results for shorter strings were entirely fresh.The result was a list of groups of statements with at least one shared chunk of language among them.These were then analysed manually for similarities and differences in content, structure, organisation and phraseology.

Results
In this section we report the results of our investigation of marked and unmarked intertextuality in the corpus.

Direct intertextuality
Close reading of a sample of statements revealed that explicit word-forword quotation is quite unusual (as it is in academic writing in many disciplines).It occurs in occasional citation of slogans to identify advertising campaigns by retailers (sixteen instances retrieved from the corpus by our procedure), as in Extract 2.
Extract 2: a relationship summed up so well by the 'your M and S' campaign (M & S 2006) Otherwise, intertextual reference involved paraphrase or simple naming of documents.The close reading phase showed that explicit references to other texts were signalled by using nouns referring to a (folk-)genre or verbs referring to a speech act of some kind.Some of these are comparable in level of precision to references to sources in academic texts.Extract 3 contains both elements in a very complete form, such that there is unequivocal reference to a definite source with a definite author.
Extract 3: In my half year statement I reported that NAME had retired from the Board on 18 April.(ABF 2007) In Extract 3 My half-year statement gives the date (six months before the present) and author ("my") of the text referred to, and in this respect is equivalent to an academic reference of the Name-Date type.One striking difference is that this example gives the genre of the cited text along with the citation (rather than showing it in conventionalized form in a reference list).The source author is a syntactic constituent of the sentence and so in the terms used for academic writing it is an integral reference (Swales, 1990).However, because the genre cited is given in the text, the actual referring structure is of the form "In (X's) 2006 statement, X reported …" which is not typical of academic citation.
Sometimes prototypical citations of this kind seem to refer to a third text, as in Extract 4.Here we are given the genre (report), author (my) and date (last year) and a paraphrase of content (our plan for) introduced by a speech-act verb (detailed).But this content itself refers to something (plan) which might be a text.A related form is exemplified by Extract 5.
Extract 5: In November we announced our plan to sell our European vending business Selecta As. (Compass 2006) Here something that might be a text (plan, decision) is the object of the speech act verb.The announcement is a cited text and if the plan or decision is to be regarded as a text, the plan or decision itself is another.
For the announcement, author, and date information are given, as well as a paraphrase of content.The genre information, if that is what it is, refers to the text announced not to the announcement (which may have been in the press release genre, for example).Alongside these quite full citation forms, there are others which only realize parts of this 'full' form, and cannot be directly compared to academic references.It is common for references of this kind to refer to a dated group of texts/utterances (Extract 6): Extract 6: However, as we said throughout the second half of 2007/08, consumer budgets are clearly under pressure (Sainsbury 2008).
Others have a reporting verb without either explicit genre or date (Extract 7): Words like decision, plan, agreement, settlement often occur without a reporting verb.Here there is a kind of cline of intertextuality.In Extract 5 above, for instance, plan might refer to a document or it might simply mean 'intention'.Sometimes it is likely that what is being referred to is a text.In Extract 8 the word decision probably refers to an instance of the genre 'court decision' Extract 8: the recent appeals court decision to reject the US federal government's US$280 billion claim against the US tobacco industry is obviously encouraging (BAT 2004) However other cases of the use of the words decision cannot be said to refer to specific texts or utterances.In Extract 9 it is not a text which is regretted although obviously the decision was expressed in words and in that sense makes the text polyphonic.
Extract 9: we regret her decision to leave and thank her for the significant contribution.(Whitbread 2007) Intertextuality permeates all texts and it would be impossible to catalogue every possible manifestation in our corpus.The focus of our quantitative study is therefore on instances of intertextuality with reporting verbs.
Neither the close reading nor the automatised corpus search produced any instances of quotation from literature or other canonical sources or of phrases like according to.That is, where a named source for a statement was given, it was always associated with a reporting verb.The two remaining categories for investigation were therefore direct quotations with some kind of typographical indication and 'reported speech' with a reporting verb.
The only direct quotations that could be examined via the corpus were those presented in quotation marks, and there were few of these.A handful of the statements included a genre-breaking section cast as an interview with the Chairman and CEO, and ten utterances in one of these were presented as direct quotations from an oral interview.Otherwise nearly all the quotations found were in statements from retailing or banking companies, and many were used as the names of campaigns of various kinds, as in Extract 2. Extract 10 gives a further example from the sixteen different quotations all of which refer to phrases which must have been instantiated in very many earlier advertising, marketing, and internal texts (the phrase has 779 hits on Google).
Extract 10: … running marketing campaigns such as 'feed your family for a fiver' (Sainsbury 2009) The list of reporting verbs investigated is neither exhaustive nor uniform: some verbs are clearly signals of intertextuality, others more marginal.Table 1 lists the verbs found in instances of intertextual and intratextual reference.(These verbs occurred equally frequently as metalinguistic rather than intertextual devices allowing evaluative comments I am delighted to announce that.Such instances have been excluded.)Table 1 shows that the corpus search for verb forms found 452 cases where there was clear reference to another text (in the vast majority of cases) or to another part of the Annual Report in question.While there are probably a comparable number of intertextual references where the indicator of intertextuality is a noun and some where a verb not searched for is used, one of the main findings of our survey is that explicit intertextuality is relatively infrequent in these texts.Instances marked by one of the verbs chosen occur at an average rate of less than 2 per statement (0.84 per thousand words), and overall 3 instances per statement (about 1.5 per thousand words) would be a reasonable estimate of frequency, confirming the impression from the close reading.By contrast Hyland (1999) found 10.7 citations (of a different form) per 1000 words of running text, as an average for all disciplines, in his RA corpus.
The sample is dominated by the fairly general verbs announce, say, and report.Instances of these verbs make up more than half the total found.Since each verb has idiosyncratic uses and for some (like agree) it is not clear when the reference is intertextual and when it is not, it is these three which are examined more closely to get a quantitative picture of explicit intertextual reference in this corpus.Table 2 classifies the instances according to the presence or absence of a precise date which would enable retrieval of the text referred to, and by the source citedthe logical subject of the reporting verb.Three instances of report and ten of says are omitted because they were intratextual, referring in the cases of report to other sections of the Annual Report, and in those of says to 'speakers' in the genre-breaking text presented as an interview.It is mostly we who announce and report things, while I is quite often the subject of say.The group, your Board, and the names of units within the company are common sources too.The names of individual company employees occur mainly when they announce their retirement, that is most achievements are presented as collective.
In academic writing self-citation is also a common feature.In a study of the citations of eight prominent researchers, White (2001) found that the author whom each re-cited most frequently was in fact him-or herself.However, there were still many more citations to the cumulative works of all the other scholars they cited.In addition, in academic writing self-citation often has a self-effacing quality; first-person references do occur in academic texts (Hyland, 2003), but it is also common for a writer to cite him-or herself by surname and in the third person, precisely as other authors are cited.In the present corpus there are no instances of this type.This difference reflects not only the genre but also the topic-one purpose of the CS is to review what 'we' have been up to in the last year.
The quantitative analysis (Table 2) confirms that a majority of intertextual references are rather specific in terms of source document date (and therefore retrievability), as in Extract 3 above.Even where the date is vague or unspecified the instance often refers to specific documents, very often using phrases like 'as previously reported'.Other instances refer to purported multiple texts (Extract 13): Extract 13: our aspiration … remains achievable, although, as we have repeatedly said, there may be peaks and troughs along the way.(Aggreko 2010) As noted above, an outsider such as a reader from the academic discourse community is struck by intertextual episodes in which the the genre of the source text is specified (In my 2009 report, as in Extracts 3 and 4), and/or the subject of the source text is a third text (We announced a plan as in Extracts 5, 6, and 7).Although quite normal, such cases were not particularly frequent in the corpus as a whole.

Lexical similarity
The previous section demonstrated that the chairman's statements contain relatively little intertextuality in the form of direct and explicit references to other texts.However, the analysis of n-grams revealed a great deal of intertextuality in the form of chunks of language which co-occur across the statements.Units of at least 30 words, shared by two or more statements, were found in the annual reports of 24 of the companies (see Figure 1 for an example).In all cases these were in different annual reports from the same company; that is to say, there were no cases of a chunk of this length appearing in the reports of two different companies.Here too, intertextuality is an in-house affair.
Using a minimum frequency of two occurrences, the corpus contains 1,196 30-gram types and 2,797 tokens.In other words, nearly 84,000 words, or one third of the running words in the corpus, are part of a 30gram which occurs at least twice.When a more restrictive minimum frequency of five tokens is applied, the corpus is nonetheless found to contain 27 types and 198 tokens, or 780 per million words.This exceptionally high frequency speaks to the extremely formulaic nature of this genre.
It should be noted that these frequency figures are not truly comparable with those for lexical bundles found in larger, general corpora (e.g., Biber et al. 1999).In such corpora lexical bundles demonstrate that some multi-word units co-occur with such regularity that they can be considered to have some of the properties of orthographic words.When examining texts from a particular domain, however, the frequent occurrence of very long strings of words does not attest the word-like nature of the strings; rather it indicates that the writer of a given statement was influenced by the earlier statements.
Despite the earlier caveat about the difficulty of establishing a numerical threshold which can be considered indicative of repetition rather than autonomous composition, we maintain that the presence of a string such as the one in Figure 1 below in two texts strongly suggests that it was copied (in-house) from one into the other.The Group's underlying profit, which we define as profit before taxation, exceptional items and amortisation of intangible assets, was [X] million pounds compared to [X] million pounds in [year].This represents underlying earnings per share, on a diluted basis, of [X] pence … Figure 1.A 42-gram, found in ABD 2006A 42-gram, found in ABD , 2007A 42-gram, found in ABD , 2008 To the extent that this part of the investigation was designed to understand whether earlier chairman's statements are used as templates for later ones, that inference was fundamental.The fact that there are many, long shared strings of text, and that they comprise a large proportion of the corpus as a whole is, we argue, evidence for templates and repetition in the production format for these texts.
This conclusion is supported by a qualitative analysis of the statements.Across companies, the statements show considerable regularity of content and rhetorical structure, suggesting that it is a well established and relatively clearly defined genre.In addition, within the same company, similarity of topic, structure and phraseology can be seen.A comparison of successive years' statements shows signs of clear development from one year to another.For reason of space constraints, we will illustrate this common pattern with reference to the chairman's statements of a single company.
Aggreko is a FTSE-100 company which 'provides power and temperature control solutions to customers who need them either very quickly, or for a short or indeterminate length of time' (Annual Report 2011, p. 7).The Aggreko chairman's statements which are included in the corpus date from 2000-2011, inclusive, and are on average 1421 words in length.During this period, two chairs' names appeared at the end of the statements, one in 2000 and 2001, and one beginning in 2002 through 2011.
The statements for that period are intertextually linked by covering a range of topics typical for this genre as illustrated in Figure 2 (see also de Groot 2008: 73).
1. Overview of some of the year's salient activities and projects 2. Statement of the company's strategy and objectives 3. Summary of financial performance, e.g. trading performance, revenue, return on capital 4. Overview of funding, e.g., debt, capital expenditure, etc. 5. Information about dividends and other information for shareholders 6.Statement of appreciation of the company's employees 7. Review of changes to the board and senior management 8. Statement about ethical concerns 9. Statement about the outlook for the company in the coming year 10.Chairman's signature and date Figure 2. Thematic sections in CSs3 The presence of these topics, the order in which they appear and the level of detail given to them are regular but not fixed.For example, every statement begins with a paragraph mentioning some of the highlight events of the year, but in some statements (e.g., 2002) this extends to more than one paragraph and in some (e.g., 2000), the theme recurs after other subjects have been dealt with.In no case does appreciation of the company's employees receive more than one paragraph, and it is absent from several statements.Information about financial performance can come at the beginning, middle or closer to the end.A statement about ethical issues appears only in 2010 and 2011.Despite this variation, the statements form a coherent body, as witnessed by the fact their content falls comfortably under the same set of headings.This generic coherence demonstrates that the writers of the statements have a clear awareness of the relatively sharp constraints which dictate appropriate and desirable content, and suggests that their understanding of appropriate content is guided by earlier statements.
The second conclusion is supported additionally by the phraseological similarities across the statements which is evidence for a 'template' composing strategy.Two points about the phraseological likenesses need to be made.The first is, simply, that they are numerous and occur in all thematic sections: there are many long, identical or very similar strings of language among the statements throughout the texts.In Table 3 below, the numbers in the left-hand column refer to the section of the statement (from the list above) and the year from which the quotation is taken appears in brackets after the quotation.Italics are used to highlight differences between the two years, where they occur.For the sake of space, only two instancess of each chunk are given; this does not mean that they appear only twice in the corpus.
The second noteworthy point about the language of the statements is the evolution across them.As Table 3 shows, some co-occurring chunks of language feature variation.In some cases this is due to the cyclical nature of the reporting process: a change which is announced one year may be the subject of evaluation or follow-up in a later year.Thus, in row 8, a change is announced in 2010 ('the Board has now taken the further step…') and reviewed the following year ('Last year the Board took the further step…').In other cases, though, the changes appear to be motivated more by a desire to vary and paraphrase than any need to adapt earlier phrasings to a new context; in row 9, for instance, there is no clear necessity to transform 'limited visibility with respect to the outcome' to 'limited visibility of the likely out-turn'.[2004] As ever at this early stage, there is limited visibility of the likely out-turn for the current year… [2005] Thus, even as the similarities between statements are clear, so is their evolution, with the result that while points of phraseological identity or similarity can be found between any two years, 2011's statement differs substantially from 2000's.This evolution is shown in Table 4, which shows the formulations used to hedge the predictions for the coming year.
What this analysis demonstrates is such a high degree of interrelatedness among the CS as to suggest a production strategy of using one year's statement as a template for the next.In the academy a strong emphasis in teaching student writing is placed on autonomous expression (Pecorari 2008).On the basis of these findings, that is not the common working practice of writers in the workplace, or at least not those called upon to produce this particular genre.2007 Looking ahead, at this early stage it is always hard to come to a definitive view of the outcome for the year as a whole, and particularly so when faced by the current level of uncertainty about the future direction of the various economies… 2008 It is always difficult at this early stage to come to a definitive view of the likely outcome of the year, and never more so than in the current economic environment.2009 It is always difficult at this early stage to come to a definitive view of the likely outcome of the year, and never more so than in the current economic environment.2010 The current instability in some countries in the Middle East and Africa makes the task of predicting the outcome for the year more than normally difficult; It must be acknowledged that in this section we have used textual evidence of similarity among the statements to deduce an intertextually influenced writing process.Sceptical readers may believe that some or all of the similarity documented here is coincidental, due to these statements' common purpose, necessarily similar content, discourse community-specific language, and in some cases indentical authorship.
While acknowledging that, by virtue of our method of analysis, the evidence for a template writing strategy presented here is circumstantial, we also believe it is strong.However, a more relevant point may be what intertextual similarity tells us about the nature of the written product.Writing demonstrably produced with a template strategy would presumably exhibit a degree of intertextuality approximately equivalent to that found here; therefore a template strategy would appear to be a useful one for writers aspiring to (learn to) produce exemplars of this genre, whether or not such a strategy has actually been used here.

Discussion
The present investigation examined two sorts of intertextuality-explicit reference and recycled phrasing-in an important business genre.While revealing that intertextuality is a pervasive feature of this genre, the findings have also demonstrated that in terms of its frequency, form and the inferred process, this genre differs substantially from academic genres.
Academic texts make prolific use of direct, identifiable references to earlier texts, and writers are expected to demonstrate that they have read widely on their topic, something which-despite the prevalence of selfcitation-means incorporating other voices than the author's own.However, in the CS, other texts referred to are primarily the earlier utterances of the chairman himself4 , a text produced by the company, or a first person plural source which is or must be inferred to be the corporate body.The references in these statements are therefore inwardlooking in a way which would not be conventional in academic writing.
The format in which citations are made is rigidly dictated by convention in academic writing, with broad similarities across academic texts and absolute uniformity required within a single publication.Bibliographic information is detailed, in order to permit readers to verify the claims writers make on their source authors' behalf.Here too the CS has a different profile.While many of the references in this corpus provided information similar to that found in academic textsauthorship, date, genre-the presentation of this information was simultaneously less detailed and less conventionalised in presentation.
A further important difference lies in the processes by which some of the intertextual relationships must be presumed to have come about.Evidence was presented above suggesting that a prior year's statement serves in many cases as a starting point for producing the current year's.This is directly at odds with expectations for academic writing.Authors who recycle portions of their earlier publications are frowned upon.For example, in a guide prepared for the Office of Research Integrity, a part of the US federal government, Roig (nd) places the practice of recycling a description of research methods from one paper to another under the heading '"Borderline"/unacceptable cases of text recycling' (p.23) and cites an editor of an academic journal who characterises such a template approach as self-plagiarism.
Researchers who publish the words of other authors without explicitly marking them as quotation and identifying the source are labelled plagiarists.Student writers who engage in either of those practices in assessment writing may find themselves charged with collusion or plagiarism.Positive prescriptions for carrying out an assessment writing task-such as those given by teachers of English for Academic Purposes-involve starting with an independent task conception and creating an original text to match the purpose.A template approach to academic writing would thus miss the mark in several ways.
Descriptions of the features of academic genres have observed that the superficial differences have their basis in the values and substantive practices of the communities in which they are produced.For example, the fact that integral citation is more common in the humanities than in the natural sciences is often attributed to the fact that the knowledge claims made in the latter area are (supposed to be) subject to verification and reproduction, and thus their source is relatively less important, while knowledge claims in the humanities tend to be inherently contestable, and thus can be evaluated more easily when their human source is identified and taken into account (Hyland 1999).
The intertextual practices found in the chairman's statements can similarly be interpreted in terms of the texts' rhetorical purposes and role within the discourse community.For example, an explicit reference in APA format to 'feed your family for a fiver' would be unnecessary on at least two grounds-nobody is likely to want to consult the source, and it has been publicised widely enough already to be familiar to readers of the statement-and, for the latter reason, it would be very difficult, if not actually impossible, to identify such a thing as an original source.
The objective of LSP instruction is to provide the set of knowledge and skills needed for for the specific communicative events which are characteristic of a particular domain.The LSP practitioner must therefore possess an awareness of the genres within a domain and the features which characterise them.If this awareness is not based on the teacher's personal knowledge of the domain, then s/he needs access to empirically based descriptions.This investigation of a widespread discoursal feature-intertextuality-in a particular domain-business texts-is a contribution to that descriptive project.
The different practices reflect different text functions and different conditions for text production and reception.In the context of higher education, assessment of writing skills is often done through the traditional academic genres such as the essay.The assessment criteria for such texts are well established and quite specific.With regard to intertextuality, good practice includes reading widely and citing a range of sources to support an independently developed argument.Students are instructed in, and expected to develop proficiency in formal features such as reporting verbs (distinguishing between 'Smith states' and 'Smith suggests'), mechanics (i.e., APA versus MLA, etc.) and other highly conventionalised aspects of source reporting.Original work is prized; unattributed source-dependent work is prohibited.However, these are neither the conventions of the lifeworld the students come from nor those of the professional worlds most of them go to.
The findings of this study suggest that LSP instruction in higher education contexts may present a problematic meeting ground for the practices of different domains in at least two ways.Students arrive at university with a great deal more exposure to the generic and discoursal practices of visible, public domains such as advertising and journalism, than they have to those of academia.Some students may have experience of the workplace, and exposure to the communicative practices of additional domains.Their beliefs about appropriate intertextuality practices do not equip them effectively for what they will encounter at university.The problematic areas may actually be intensified by the fact that students are not tabulae rasae; the knowledge base they bring from other areas may make it more difficult for them to see the the new knowledge they are required to assimilate.
A second area of difficulty arises when students of, for example, engineering, information technology, or natural sciences have completed a period of study and go into the workplace armed with what they have learned about how to write in university contexts-knowledge which, on the basis of the evidence presented here, will not easily transfer to workplace writing tasks.This is problematic in the context of the current emphasis that is placed in western countries on employability as an outcome of tertiary education.The Joint Declaration of the European Ministers of Education of 1999 (the 'Bologna Declaration') proposed radical changes to the organisation and administration of higher education in Europe 'in order to promote European citizens' employability and the international competitiveness of the European higher education system' (1999: 3), a goal which was ratified by the Communiqué of the Conference of European Ministers Responsible for Higher Education as recently as 2009: With labour markets increasingly relying on higher skill levels and transversal competences, higher education should equip students with the advanced knowledge, skills and competences they need throughout their professional lives.(2009: 3) Students whose professional work may require them to write texts in an environment like that of the chairman's statements would be better prepared if features such as the use of templates, promotional referencing style, etc. were placed as foils to academic-writing patterns rather than ignored.As noted above, it is likely that students bring with them to the university intertextuality patterns which clash with academic norms but are actually standard elsewhere, and this prior knowledge should be acknowledged and made use of.
We do not wish to suggest that resolving this clash of writing cultures is the primary purpose of LSP instruction, nor that doing so entails abandoning academic genres in favour of those used in the workplace.However, to the extent that such sharp differences have been shown to exist, effective LSP instruction should be based on a conscious and principled decision about which genres to teach and assess, rather than an assumption by default in favour of academic genres.

Extract 4 :
In my report last year I detailed our plan for restoring the fortunes of Marks and Spencer (M & S 2002).
Extract 7: … providing more delivery choices, something customers have said is important to them(M & S 2011) Extract 11: this policy was announced by the Chancellor of the Exchequer in the June budget.(RBSG2010)    Extract 12: Turning to broader societal issues, a great deal has been said in recent months about the role of bonuses in the banking system.(Lloyds2008)

Table 1 .
Reporting verbs functioning intertextually found in the corpus

Table 2 .
Instances of intertextuality with reporting verbs, by source of utterance and

Table 2
shows that intertextual reference in Chairman's Statements is overwhelmingly to texts produced in-house.The few outside sources are regulatory organizations (the European Commission, the OFT), political actors (Extract 11) or unspecified debaters (Extract 12).

Table 4 .
Evolution in hedging statements 2003 at this early stage of the year and subject to exchange rate variations… 2004 As is always the case at this time of year, we have limited visibility with respect to the outcome for 2005.2005 As ever at this early stage, there is limited visibility of the likely out-turn for the current year… 2006 …so it is always difficult at this early stage to predict the year's performance.