Metadiscourse in English and Swedish Non-fiction Texts and their Translations

This study compares the metadiscourse (i.e. the mea nings which relate to the writers and readers of a text) in two samples of English an d Swedish non-fiction texts and their translations in the English Swedish Parallel Corpus. Using an integrative approach to metadiscourse (Ädel & Mauranen 2012:2), it finds that there is a considerably higher frequency of metadiscourse in t he Swedish original texts and a somewhat larger proportion of interpersonal metadis course, which represents the writer’s attitude towards the propositional content and the readers themselves. In particular, there is a more frequent usage of boost ers. In both of the translation samples, there is an increase in transition markers , which raises the level of explicitness in the text. In the translations into English, a tendency was also found for translators to reduce emphasis by omitting boosters and, in some cases, inserting hedges. This, coupled with the higher frequency of boosters in the Swedish original texts suggests that there may be differences in wri ting conventions in English and Swedish non-fiction texts, for instance, when it co mes to increasing the emphatic force of propositions.


Introduction
Texts may be seen as consisting of different levels of meaning, a propositional content level, which refers to actions, events, states of affairs or objects in the world portrayed by the text, and a writer-reader level, where the writers interact with their readers, explicitly guiding them through its structure and organisation, commenting on the writing process itself or expressing their opinions and beliefs concerning its content.The meanings expressed on the writer-reader level of the text have been referred to by the umbrella term metadiscourse, i.e. "the self-reflective linguistic expressions referring to the evolving text, to the writer, and to the imagined readers of that text" (Hyland 2004:133).Typical linguistic expressions of metadiscourse include, for instance, conjunctions and conjuncts, first and second pronouns referring to the writer and reader, interrogatives and imperatives addressing the reader, and references to the text itself, etc.As metadiscourse is "a manifestation of the writer's linguistic and rhetorical presence in a text."(Hyland 1998a:3), expressing the writer's "personality, audience-sensitivity and relationship to the message" (Hyland 1998c:438), it is one of the means by which writers attend to the rational, credible and affective appeals of persuasive rhetoric (logos, ethos and pathos) (Hyland 2005:63-85).
Metadiscourse varies depending not only on the writers' purpose and their relationship to their readers, but also the social and cultural context in whích writing takes place (Hyland 2005:113-137).Anglo-American writers of research papers have been found, for instance, to use a greater amount of metadiscourse in order to explicitly guide their readers through their texts than Finnish writers, who use a generally more implicit rhetorical strategy with less reflexivity and emphasis (Mauranen 1993: 252-259).According to Mauranen, this reflects a tendency for Anglo-American writing to be more rhetorically explicit than Finnish writing.Similarly, in a comparison of metadiscourse in English, Norwegian and French economics and linguistics texts, Dahl (2004Dahl ( :1821) ) found that the English and Norwegian writers used more metadiscourse than the French writers.Other contrastive studies have also found differences in the usage of metadiscourse in English and other languages, e.g.English and Slovene research papers (Pisanski Peterlin 2005), English and Spanish editorials (Milne 2003), and English and Spanish economic texts (Valero-Garces 1996).According to Hinds (1987:143), English represents a "writer responsible" culture, i.e. writers are expected to take responsibility for the clarity of their texts by providing signposts for the reader to ease processing, as opposed to a "reader responsible culture" which tends to be more implicit, laying more responsibility on the reader for the success of the communication.This is related to what Chesterton (1997:114), refers to as the "significance threshold" in communication, i.e. the point above which something is felt to be worth saying, and below which it is not felt necessary to say anything at all.This may vary from culture to culture and appears to be somewhat lower in English than in Finnish, for instance.When translating from Finnish into English, translators may therefore feel a need to strengthen the text by adding features of metadiscourse, whereas in translations into Finnish they may feel a need for the text "to be 'toned down' somewhat in order for it to meet the target culture's different tolerance of rhetorical display" (Chesterton 1997:115).Translating a text involves therefore taking into account the fact that the usage of metadiscourse in the target language may be influenced by different cultural preferences and norms of politeness.Williams (2010) found, in a study of students' translations from French to English, for instance, that when some of the students failed to translate some of the features of metadiscourse appropriately, this resulted in the loss of some of the nuances, which, according to Hyland (2005:39), contribute towards making the content of a text "coherent, intelligible and persuasive to a particular audience".Similarly, Pisanski Peterlin (2008) found that translators of Slovene research articles into English made a considerable number of changes in the metadiscourse, both omissions and insertions.
Swedish advanced learners of English have been found to use more metadiscourse in their argumentative writing than native speakers (Petch-Tyson 1998, Ädel 2008).In particular, there are more overt references to the discourse participants and more taking into account the imagined reader.There is also a greater density of metadiscourse elements (Ädel 2008:54).According to Ädel (2008:59), one of the chief influencing factors, as well as general learner strategies and a lack of genre awareness, may be different Anglo-Saxon and Swedish writing conventions, in particular a strong tendency towards informality in Swedish writing.It is possible, then, that in certain circumstances Swedish and English may have a different significance threshold as far as the usage of metadiscourse is concerned.This study aims, therefore, to investigate whether this may be the case.For this purpose, it will first compare the usage of metadiscourse in a sample of English and Swedish original texts (five texts in each language consisting of altogether approximately 60,000 and 64000 words, respectively) and then examine how the metadiscourse has been dealt with in their translations into English and Swedish (approximately 73,000 and 57,000 words, respectively).The original texts and their translations have been selected from the nonfiction category of the English Swedish Parallel Corpus (Aijmer et al. 1996).Each language sample consists of extracts from five texts.As the non-fiction category contains a wide variety of text types, ranging from parliamentary speeches and company reports to biographies and historical accounts, similar text types have been selected from each language as far as possible.Each language sample therefore comprises two extracts from biographies, two extracts from travel books and one extract from a historical account.A full list of the texts and the codes used in the examples quoted here are given below.
Section 3 compares the metadiscourse in the original texts.Section 4 compares the metadicourse in the translations and examines what changes have been made.Section 5, finally, discusses what conclusions can be drawn from these findings.First, in the next section, the model used in the classification of metadiscourse will be presented.

Classification of metadiscourse
There are two main approaches to analysing metadiscourse, an integrative approach which sees textual interaction between the writer and reader as its main defining feature and a non-integrative approach which follows a narrower definition of metadiscourse as reflexivity only, i.e. language commenting on language itself (Ädel & Mauranen 2010:2).It is the former, broader approach which will be adopted here, following Hyland's classification (1998aHyland's classification ( , b & c, 2000Hyland's classification ( , 2004Hyland's classification ( , 2005)), which is a development of the taxonomy originally proposed by Vande Kopple (1985) and later revised by Crismore et al. (1993).This model makes a distinction between interactive metadiscourse, which is used to organize the propositional content of the text, and interactional metadiscourse, which alerts readers to the author's perspective towards the propositional information and the readers themselves (Hyland 2005:50-54).In this study, I will, however, refer to these as textual and interpersonal metadiscourse, respectively.Each of these types of metadiscourse are illustrated here by examples taken from the samples of English and Swedish translations.
Textual metadiscourse consists of the sub-categories: transition markers, frame markers, endophoric markers, code glosses and evidentials.Transition markers express semantic relations between stretches of discourse, i.e. they explicitly establish "preferred interpretations of propositional meanings by relating individual propositions to each other and to readers" (Hyland 1998b:228).They signal, for instance, additive, contrastive and resultative relations, and they are realized by a wide variety of linguistic markers ranging from conjunctions and conjuncts to prepositional phrases, etc.
( Textual metadiscourse consists of the subcategories: hedges, boosters, attitude markers, engagement markers and self mentions.Hedges withhold commitment to a proposition, e.g.troligen ("probably") and probably in (9).In doing this, they indicate the writer's decision to acknowledge the possible existence of other voices and viewpoints and thereby be open to heteroglossic negotiation with the reader (Bakhtin 1986, Martin & White 2005:105).
The cathedral dedicated to St Lawrence was probably begun in 1085, when King Canute (later to be called Canute the Holy) created economic conditions for the construction.(LI) Boosters, e.g.without doubt and utan tvivel ("without doubt") in (10), increase the writer's commitment to a proposition and demonstrate a confident, decisive image (cf Hyland 2000:236).Like hedges, they open up the content to heteroglossic negotiation but at the same time they contribute to closing down the argument (Bakhtin 1986, Martin & White 2005:133).Some boosters emphasise the remarkability of the proposition, e.g.rentav and even in (11).( 10 The Lund spirit is supposed to contain a generous dose of scepticism.In some people this scepticism reaches such heights that they even deny the existence of a Lund spirit.(LI) Attitude markers show the writer's opinion of the content, expressing, for instance, affective attitudes of surprise, e.g.paradoxalt ("paradoxically") and Strange to say in ( 12), or regret, e.g.Sadly and Sorgligt nog ("sadly enough") in ( 13).
Strange to say, Lund became even more of a university town during these years of postwar expansion.
(LI) (13) Sadly, it no longer harbours the deer which once provided sport for kings, the disturbance caused by increased public pressure having driven them away.
Sorgligt nog har hjortarna, en gång kungligt villebråd, försvunnit härifrån på grund av de störningar samhällsutvecklingen orsakat.Engagement markers explicitly address readers and draw them into the discourse.They are typically second person pronouns referring to the reader, e.g.you and du in ( 14), and first person plural reader-inclusive pronouns we and vi, as in ( 15), and interrogatives and imperatives addressing the reader, as in ( 16) and ( 17). ( 14 At what do our riksdagsmän stare during their debates in their renovated riksdag?A painting of people at work, factories and smoke stacks, cars and houses?No, an enormous tapestry representing the land-and-seascape of the skerries, without a single house or human being in sight.(IU) (17) Nå låt oss lämna vår fiskande vän och återvända till Ett Svenskt Hem.
Well now let's leave our friend the director with his net and return to The Swedish Home.(IU) Self mentions are typically first-person pronouns I and jag, which make the writer's presence known in the text, as in ( 18).
And do I need to say that his childhood home was painted red and had white-painted corners?(IU) All the instances of metadiscourse in the English and Swedish samples were collected manually and then classified according to the model of metadiscourse above.Only explicit linguistic realisations have been included, although metadiscoursal meanings may also be inferred in the text.When several types of metadiscourse combine with each other, as in ( 17) above, where engagement markers (the imperatives, Låt oss lämna and Let's leave) function at the same time as frame markers indicating a shift of topic, and in ( 18), where the self mentions I and jag combine with engagement markers (the interrogative clauses, behöver jag säga det … and And do I need to say …), each function has been counted as a separate feature.The next section compares the metadiscourse in the English and Swedish original texts.

Metadiscourse in the Original Texts
Table 1.compares the metadiscourse in the English and Swedish original texts.In the sample of English original texts, there are 1453 metadiscourse items altogether, and their frequency is 24.2 times per 1000 words.In the sample of Swedish original texts, in contrast, the total number of metadiscourse items ( 2090) is much higher (statistical significance p<.001 2 ), and their frequency is 32.6 times per 1000 words. 3A similar higher frequency of metadiscourse in Swedish nonfiction texts was found in a study carried out by Ädel (1999).
All the different types of metadiscourse occur more frequently in the Swedish sample than in the English sample, with the exception of engagement markers, which are, conversely, slightly more frequent in the English sample (1.9 vs. 1.5 times per 1000 words).This is probably due to the fact that the second person pronoun you in English can both be an engagement marker addressing the reader and at the same time have generic reference, whereas Swedish makes a distinction between the second person pronoun engagement markers, du ("you", singular) and ni ("you", plural), and the impersonal pronoun man ("one"), which is used for generic reference.(This is exemplified by example (36) below).The greatest difference between the samples is found in the interpersonal metadiscourse, which is 2 Statistical significance has been calculated using the Sigil Corpus Frequency Test Wizard (sigil.collocations.de/wizard.html) 3 There is a great deal of variation between the individual texts.In the Swedish original texts, the frequency of metadiscourse ranges from 21.7 to 55.36 times per 1000 words.In the English original texts, it ranges from 15.7 to 39.0 times per 1000 words.altogether 1.5 times more frequent in the Swedish sample than in the English sample (9.2 vs. 6.1 times per 1000 words) compared to the textual metadiscourse, which is 1.3 times more frequent (23.4 vs. 18.1 times per 1000 words).There is thus a somewhat larger proportion of interpersonal metadiscourse in the Swedish sample than in the English sample (28.2% vs. 25.4%).The features which differ most in frequency are boosters and self mentions.Boosters occur more than twice as frequently in the Swedish sample as in the English sample (3.7 vs. 1.7 times per 1000 words) and self mentions occur 26 times in the Swedish sample and only once in the English sample.
In sum, there is more metadiscourse in the sample of Swedish original texts, in particular, interpersonal metadiscourse.The total amount of metadiscourse found in both of the samples is much lower than that which has been found, for instance, in studies of English academic writing, such as research articles and university course books, where metadiscourse features occur three times more frequently (66.2 and 68.5 times per 1000 words, respectively) (Hyland 12005: 102).The most striking difference is the less frequent usage of hedges, which occur only 1.8 and 2.5 times per 1000 words in the English and Swedish original non-fiction texts, respectively, in contrast to 16.7 and 6.4 times per 1000 words in research articles and university course books, respectively (Hyland 2005:102).It appears, thus, that in the type of non-fiction writing examined here, writers tend to intrude less into their unfolding text to influence their reader's reception of it.This is most probably due to the fact that they do not cast their claims as individual and contingent to the same degree as writers of research articles, and therefore there is not the same need to "ground propositions in an explicitly acknowledged degree of subjectivity" (Hyland 2005:93).
I will now go on to examine what happens to the metadiscourse when it is translated.

Metadiscourse in the translations
Table 2 compares the total amount of metadiscourse in the English-Swedish and Swedish-English original texts and their translations.It includes the numbers of matches, i.e. metadiscourse features which correspond to similar features in the original texts, and the numbers of changes, i.e. metadiscourse features which have been omitted or inserted in the translations.In the English-Swedish translations, the total number of metadiscourse items ( 1452) is similar to that in their original English texts (1453).
Altogether 1319 of these are matches (approximately 90% of the metadiscourse in the translations).134 items in the English original texts (approximately 9% of the total number) have been omitted in the Swedish translations and 133 items (approximately 9% of the total number) have been inserted.In the Swedish-English translations, the total number of metadiscourse items has increased slightly from 2090 in the Swedish original texts to 2121 in the English translations (statistical significance p <.001).Altogether 1851 of these are matches (approximately 87% of the metadiscourse in the translations).The correspondence between the metadiscourse in the Swedish-English translations is thus slightly lower than in the English-Swedish translations.239 items in the Swedish original texts (approximately 11% of the total number) have been omitted in the English translations, and 270 items (approximately 13% of the total number) have been inserted.There is thus a slight increase in the total amount of metadiscourse in the translations from Swedish into English, which may be a translation bias due to the influence of a high frequency in the source texts (cf.Gellerstam, 1994:61).In both of the translation samples, the translators have made a number of changes, both insertions and omissions of metadiscourse features.I will now examine these in more detail.
According to Chesterton (1997:88-115), changes made in translation are syntactic, semantic, and pragmatic strategies used by the translator in order to achieve "what they regard as the optimal translation".Syntactic strategies manipulate the clause and sentence structure of the text.Semantic strategies change its meaning, by, for instance, changing emphasis (Chesterton 1997:104).Pragmatic strategies, which typically incorporate syntactic and semantic strategies, manipulate the message itself, depending on the translator's knowledge of the prospective readership of the translations.These include explicitness changes, which affect the level of explicitness of the text (Chesterman 1997:108), information changes, which add or omit information that cannot be inferred from the surrounding text (Chesterman 1997:109), interpersonal changes, which alter the relationship between the author and the reader (Chesterman 1997:110), illocutionary changes, which are changes in speech acts (Chesterman 1997:110), and visibility changes which are changes in the authorial presence in the text (Chesterman 1997:112).
4 Underlying these strategies is the tendency for translators to make compensatory changes, i.e. to compensate for items that have been omitted, added or changed in the translation at some other point of the text.
In the following, I will examine how the translators have used these strategies in the translation of metadiscourse, looking first at the textual metadiscourse.

Textual metadiscourse
Table 3 compares the textual metadiscourse in the original texts and their translations, including the numbers of matches, i.e. the textual metadiscourse features in the translations which correspond to similar features in the original texts, and the numbers of changes, i.e. the textual metadiscourse features which have been omitted or inserted in the translations.Textual features have increased in number in both of the samples (from 1085 to 1115 in the English-Swedish translations and from 1501 to 1531 in Swedish-English translations).Altogether 1010 and 1348 of these are matches (approximately 90% and 88% of the textual metadiscourse in the translations).75 items in the English original texts have been omitted in the Swedish translations and 105 items have been inserted.153 items in the Swedish original texts have been omitted in the English translations, and 183 items have been inserted.In the following discussion of the changes in features of textual metadiscourse, I have treated the insertion and omission of transition markers, endophoric markers, frame markers and evidentials as explicitness changes and the insertion and omission of code glosses as information changes.I will exemplify each of these changes as follows.

Explicitness changes
The insertion of transition markers, endophoric markers, frame markers and evidentials raises the level of explicitness by making explicit relations which are implicit in the source text, as in ( 19), where the translator has made the implicit causal relationship in the original text explicit by inserting the transition marker accordingly, and ( 20), where the translator has inserted the endophoric marker, i den här guiden ("in this guide book"), thereby making explicit reference to the text itself.Similarly, in (21), the translator has inserted the evidential men skrev till honom ("but wrote to him"), making the source of the following quoted extract explicit.
( Conversely, omission lowers the level of explicitness, as in ( 22), where the transition marker so has not been translated, leaving the causal relationship implicit, and ( 23), where the translator has omitted the endophoric marker som nämnts ("as mentioned"), which refers to an earlier passage in the text.Similarly in (24), the translator has omitted the frame markers, för det första ("for the first") and för det andra ("for the second"), which indicate the organization of the discourse in separate stages of argumentation.
( ("For the first must already constructed foundations be removed, for the second was there ...") In Luleå this knowledge was priceless because the project was risky from the start.Existing foundations had to be removed, and cash was constantly short, which led to constant improvisations.(RE) Some explicitness changes are due to syntactic differences between the two languages.Non-finite -ing clauses in English, for instance, have no directly corresponding non-finite form in Swedish and therefore often correspond to finite clauses linked by a transition marker.In (25), for instance, the -ing clause, Going to sea young, has been translated into a finite clause, Jens Engebreth gick till sjöss tidigt ("Jens Engebreth went to sea early") which is coordinated by the transition marker och ("and").
(25) Going to sea young, Jens Engebreth had had only elementary schooling.This had not prevented his learning how to navigate or rising to the top.
Jens Engebreth gick till sjöss tidigt och fick bara elementär skolutbildning.Detta hindrade honom inte från att lära sig navigera eller nå en hög position.(RH) ("… went to sea early and had only …") As shown in Table 3, the textual feature that has been changed most is the transition marker.In both translation samples more transition markers have been inserted (96 and 165 in the English-Swedish translations and Swedish-English translations, respectively) than omitted (64 and 135 in the English-Swedish and Swedish-English translations, respectively), which results in an increase in the total numbers of transition markers (from 939 to 971 in the English-Swedish translations, and from 1155 to 1185 in the Swedish-English translations).(These differences are statistically significant at p<.01 in both translations.)This increase in transition markers reflects the tendency for translators to raise the level of explicitness in the text (Blum Kulka 1986:292).The other textual features which alter explicitness, i.e. endophoric markers, frame markers, and evidentials have been changed a small number of times, but these changes do not result in significant differences in their numbers in the translations.

Information changes
The insertion of code glosses provides new information which the translator believes the target language readers may need in order to interpret the text.This is based on the translators' assumptions concerning the target language readers' knowledge of the cultural environment of the source language.In ( 26), for instance, the translator has inserted the code gloss typically a small wooden house in the English translation to explain stuga ("cottage"), a Swedish expression which the target readers are not expected to be familiar with and therefore may need explaining.
They had lived through the winter in a modern flat, but now they could move to a small stuga, (typically a small wooden house), in Djupdalen, 30 km south of Stockholm, whence he could supervise the building of von Platen's house.(RE) In ( 27), the translator has inserted the code gloss so to say.This is a metalinguistic comment highlighting the double-layered meaning of the verb spread in combination with colour in the colour spread.Insertions of code glosses such as these reflect the translator's increased awareness of the language itself during the translation process.
("…spread itself the colour to finer timber houses …") During the sixteenth century the roofs of palaces, big houses and churches were painted red.During Sweden's Great-Power period (1560-1718) the colour spread, so to say, to larger timber-built houses; their owners wanted to imitate brick.(IU) Conversely, the omission of a code gloss removes information that translators believe are irrelevant for the target language readers' interpretation of the text.In (28), for instance, the translator has omitted the code gloss eller "läkekvinnor", hur man nu vill kalla dem ("or 'women healers', how one now wants to call them"), which is a paraphrase of the expression kloka gummor ("wise old women").
("… or 'women healers', how one now wants to call them") Similarly, in ( 29) the translator has omitted the code gloss so-called which in the English original text indicates that students cap is an expression that the writer believes English readers may not be familiar with.The Swedish target language readers, on the other hand, have a similar tradition of students caps and may therefore be expected to be familiar with this expression.
(29) This was the so-called "students cap", a grey peaked quasi-military affair with a tassel dangling from the top.
Det var studentmössan, grå, en smula militärisk i stilen och med en tofs som hängde ned från kullen. (RH) As shown in Table 3, code glosses have been inserted slightly more often in both translations (6 and 16 times in the English-Swedish and Swedish-English translations, respectively) than omitted (5 and 11 times in the English-Swedish translations and Swedish-English translations, respectively), which may reflect a tendency for the translators to add information which assists interpretation.
In sum, the main change which the translators make in textual metadiscourse is to raise the level of explicitness by increasing the number of transition markers.This occurs in both translation directions and is inherent in the translation process.

Interpersonal metadiscourse
Table 4 compares the interpersonal metadiscourse in the original texts and their translations, including the numbers of matches, i.e. the interpersonal metadiscourse features in the translations which correspond to similar features in the original texts, and the numbers of changes, i.e. the interpersonal metadiscourse features which have been omitted or inserted in the translations.Interpersonal features have decreased in number in the English-Swedish translations (from 368 to 337), but remain almost the same in the Swedish-English translations (589 in originals and 590 in translations).Altogether 309 and 503 of these are matches (approximately 91% and 85% of the interpersonal metadiscourse in the translations).59 items in the English original texts have been omitted in the Swedish translations and 28 items have been inserted.86 items in the Swedish original texts have been omitted in the English translations, and 87 items have been inserted.In the following discussion I have treated the insertion and omission of boosters and hedges as emphasis changes, the insertion and omission of engagement markers, which include the reader in the discourse, as interpersonal changes, and changes from or into interrogative and imperative clauses as illocutionary changes.The insertion and omission of self mentions and attitude markers have, finally, been treated as visibility changes.I will exemplify each of these changes as follows.

Emphasis changes
The insertion of a booster increases the force of a proposition, as in (30), for instance, where the translator has inserted the booster, e.g.faktiskt ("in fact").The force of a proposition is also increased by the omission of a hedge, as in ( 31), where the translator has omitted the hedge what may be interpreted as.
( Tre dagar tidigare hade kapten Thaddeus Bellingshausen, en rysk sjöofficer som sänts ut av tsar Alexander i ett anfall av expansionsiver, ungefär 2 800 km längre österut antecknat att han siktat den antarktiska iskalotten där den möter havet.(RH) ("… recorded that he sighted the Antarctic ice cap …") Conversely, the omission of a booster "tones down" the force of the proposition, as in (32), where the translator has omitted the booster, säkerligen ("certainly").The force of a proposition is also "toned down" by the insertion of a hedge, which signals the writer's lack of commitment to its content, as in (33), for instance, where the translator has inserted the hedge, kanske ("perhaps").( 32 As shown in Table 4, the changes in boosters and hedges work in different directions.In the English-Swedish translations, the numbers of insertions and omissions of boosters (16 and 13) and hedges (10 and 11) do not result in any great change in their numbers in the translations.In the Swedish-English translations, on the other hand, there are twice as many omissions of boosters as insertions (51 vs. 25), which results in a decrease in the total number of boosters (from 236 to 210, statistical significance p<.01).There are also slightly more insertions of hedges than omissions (18 vs. 10).It appears, thus, that the translators into English but not Swedish have felt the need to "tone down" propositions by omitting a number of boosters and inserting a few hedges.

Interpersonal and Illocutionary changes
The insertion of engagement markers increases the reader's involvement in the text, as in (34), where the translator has changed the third person expressions in the Swedish original, varje människa ("all people"), de ("they") , dess handhavare ("their users"), by using inclusive we in the English translation, thereby presenting the content from a shared writer and reader perspective, and ( 35), where self mention by the author in the original text has been extended into inclusive vi ("we") in the translation, thereby including the reader.
( Vi skall här helt kort granska två av dessa nya företeelser för att därefter i följande kapitel undersöka den reformerade Jahvereligionen.(KAR) ("We shall here quite briefly examine …") Similarly, the reader's involvement in the text has been increased in (36) by using you when the Swedish original has the impersonal generic pronoun man ("one"), and in (37) by changing a declarative clause into an interrogative which functions as a rhetorical question directed towards the reader.("In case a twelve-year-old boy taken himself in to a building site and bites broken the crane so happens it in hundred per cent against the parents' will and knowledge") one of the cranes with his teeth, doesn't he do it, to one hundred percent, without the knowledge and consent of his parents?(IU) Conversely, the omission of engagement markers reduces the reader's involvement in the text, as in ( 38), where inclusive vi här ("we here") in the Swedish original text has been translated into in Sweden they, thereby adapting the text to the readers of the English translation, and (39) where the engagement marker you, has been omitted.4, the changes in engagement markers (32 omissions and only one insertion in the English-Swedish translations, and 29 insertions and 20 omissions in the Swedish-English translations), occur mainly in one translation (SUG) in the English-Swedish translation sample and in two translations, (IU) and (LI), in the Swedish-English translation sample.In (SUG), most of the omissions are where the translator has used the impersonal generic pronoun man when the English original has you, as exemplified by (37) above.These changes are, therefore, chiefly due to language differences (i.e. the fact that the second person pronoun you in English also has generic reference, corresponding to the impersonal prounoun man ("one") in Swedish).In the two translations in the Swedish-English sample, one translator (IU) has frequently inserted the engagement marker we when the Swedish original has a third person perspective, as exemplified by (34), above, and the other (LI) has, conversely, changed the shared author-reader perspective denoted by inclusive vi ("we") in the original text to a third person perspective in the translation, as exemplified by ( 38) above.The changes in engagement markers in the two samples appear, therefore, to be partly due to language differences and partly due to individual choices by translators, rather than to overall differences in the usage of engagement features in Swedish and English non-fiction texts.

Visibility changes
The insertion of self mentions and attitude markers increases the visibility of the author, as in (40), where the translator has inserted to me and also uses an active verb and first person reference (I have here quoted) to correspond to the objective agent-free passive citeras ("is quoted") in the Swedish original and (41), where the translator has inserted an attitude marker (although, looking at its motorway, one is hard put to believe this) expressing an opinion concerning the content.As shown in Table 4, self mentions and attitude markers have only been changed a small number of times in both samples, and these changes do not result in a significant difference in their numbers in the translations.In the Swedish-English translations, there are, however, slightly more insertions of both of these features (9 attitude markers and 6 self mentions) than omissions (4 attitude markers and 1 self mention), which is perhaps due to the influence of a higher frequency of these features in the Swedish source texts.
In sum, the main changes which the translators make in interpersonal metadiscourse are to reduce emphasis and, in some texts, to alter the interpersonal relationship between the author and the target language reader.The former change takes place in the translations into English only, which suggests that translators may be adapting their texts to a lower level of emphasis which they perceive to be required in the target language.The latter change appears to depend on choices made by individual translators to, for instance, adapt the author-reader relationship of the original text to the target language readers.

Conclusion
This comparison of metadiscourse in a small sample of English and Swedish original non-fiction texts and their translations has found a considerably higher frequency of metadiscourse features in the sample of Swedish original texts than in the English sample.It has also found some qualitative differences, i.e. that there is somewhat more interpersonal metadiscourse in the Swedish texts, in particular a more frequent usage of boosters.In the translations, a number of changes were made in the metadiscourse, and all of the different kinds of metadiscourse features were both inserted and omitted in varying numbers.For some features, these changes led to an increase or decrease in their proportions in the translations.The main change was in transition markers, which were inserted more often than omitted, thereby increasing their total number and raising the level of explicitness in the translated texts.This occurred in both translation directions and appears therefore to be an inherent part of the translation process.The other main changes were in engagement markers and boosters.The changes in engagement markers were chiefly restricted to two translations in one sample and one translation in the other, and appear therefore to be mainly due to choices by individual translators rather than to the translation process itself.Boosters, on the other hand, were omitted more often than inserted in the translations from Swedish into English only.There was, thus, a tendency for the translators to reduce emphasis in English by omitting boosters and, in some cases, inserting hedges.This, coupled with the higher frequency of boosters in the Swedish original texts suggests that there may be differences in preferences in English and in Swedish when it comes to increasing the emphatic force of propositions.
The samples investigated here are small and therefore conclusions drawn from them must be tentative.In general, however, the findings suggest that, in its usage of metadiscourse, Swedish is similar to English in being a writer-responsible writing culture.In fact, the larger amount of metadiscourse found in the Swedish original text suggests that this may be true to an even larger extent of Swedish.The results of the comparison also suggest that this is particularly true of certain kinds of interpersonal metadiscourse, such as emphasis and also, to some extent, self mention.As both of these features are characteristics of informal writing, the findings here provide further support for Ädel's observation (2008: 54) that there may be a strong tendency towards informality in Swedish writing, and this influences the usage of the metadiscourse in Swedish advanced learners' writing in English.To conclude, then, it appears that, in non-fiction texts such as those investigated here, there may be differences in Anglo-Saxon and Swedish writing conventions so that the "significance threshold" in Swedish may be somewhat lower than in English when it comes to expressing certain kinds of interpersonal metadiscourse.Further investigation of larger samples and other text types is, of course, needed.
that the Renaissance was flourishing in Italy at that time, in Sweden they continued to build in the tried and tested brick Gothic style.(LI) (39) This book does set out to show you a side of the British capital usually reserved for residents!Avsikten är att visa den sida av den brittiska huvudstaden som vanligtvis är förbehållen dess invånare.(SUG) ("the intention is to show this side of the British capital …") As shown in Table

Table 1 .
Metadiscourse in the English and Swedish original texts

Table 2 .
Metadiscourse in the translations

Table 4 .
Interpersonal metadiscourse in translations Conversely, the omission of self mentions and attitudinal markers reduces author visibility, as in (42) where jag citerar ("I quote") in the original is translated into a nonfinite verb to quote, and (43) where the attitude marker med all rätt ("with every right") has been omitted.