(Semi)peripheries in contact – Indirect translation of novels into Swedish 2000–2015 Anja Allwood University of Gothenburg, Sweden ABSTRACT This article widens the perspective of indirect translation (ITr) research by focusing on the range of mediating languages and with a corpus of all indirect translations into a specific language (Swedish) during 2000–2015. The following issues will be described and explained: which languages are used as mediating languages (MLs), what their respective proportions look like, and which possible rea- sons for using a language other than English as the ML can be identified. The corpus reveals that out of all novels translated into Swedish during the period under study, 1.3% (70 novels) were indirect translations, and out of these 70, more than two thirds have English as the ML. A search into the cases where English has not been used produced a list of suggested reasons regarding the choice of ML. The most often occurring explanation seems to be that no English translation existed, or that such a translation was already indirect. Keywords: indirect translation, literary translation, novels, triangulation, Swedish language (Pol)periferije v stiku – Posredni prevod romanov v švedščino med letoma 2000 in 2015 IZVLEČEK Prispevek se s korpusom vseh posrednih prevodov v švedščino med letoma 2000 in 2015 osredoto- ča na razpon posrednih jezikov in s tem razširja pogled na raziskave posrednih prevodov. Opisane in razložene bodo naslednje teme: kateri jeziki se uporabljajo kot posredni jeziki, kakšen je delež posameznega posredniškega jezika in katere možne razloge za uporabo drugih posrednih jezikov poleg angleščine je mogoče identificirati. Korpus pokaže, da je bilo med vsemi romani, ki so bili v obravnavanem obdobju prevedeni v švedščino, 1,3 % (70) posrednih prevodov, med temi sedem- desetimi prevodi pa sta bili dve tretjini prevedeni prek angleščine kot posredniškega jezika. Pregled primerov, pri katerih se angleščina ni uporabila kot posredni jezik, pokaže seznam potencialnih razlogov za izbor posredniškega jezika. Zdi se, da je najpogostejša razlaga ta, da angleški prevod ne obstaja ali pa da je že angleški prevod sam posredni prevod. Ključne besede: posredni prevod, literarni prevod, romani, triangualcija, švedski jezik 57 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 1. Introduction Indirect translation is not just a thing of the past but still very much a current practice, not only in non-literary translation and subtitling, but also in literary translation (see e.g. Ringmar 2007; Pięta 2019, 25). Despite the growing body of research on indirect translation since the turn of the millennium, there are still issues to be resolved, espe- cially in different language areas. Alvstad (2017, 152) calls for – among other things – research on “the relative percentages of direct and indirect translations, the degree of overtness about the indirectness, and whether indirect translation is becoming more or less common” . Alvstad addresses the situation in Sweden, but these questions are obviously applicable more widely as well. Much empirical research thus far has focused on single language pairs: Chinese–Spanish (Marín-Lacarta 2012); Polish–Portuguese (Pięta 2012); T urkish–Slovene (Pokorn 2013), and Modern Greek–Finnish (Ivaska 2016). Here, the focus is on all indirect translations of novels into a specific language, i.e. Swedish, during a specific time, i.e. 2000–2015, in order to capture the nature of the practice on a wider scale. This approach allows us to answer the questions of how, and from which languages, indirect translation is used at present in Sweden. The main focus in this article is on the mediating languages and on possible reasons for using a language other than English as a mediating language. The corpus used in this study was compiled within the framework of a PhD project on indirect translation. It consists of all first editions of novels published in book form, translated either directly or indirectly, into Swedish during 2000–2015. The Swedish context is analysed and potential explanations for factors governing the choice of the mediating languages are discussed. 1.1 Terms, definitions and abbreviations In this article, the term indirect translation (ITr) will be used to denote a “translation based on a source (or sources) which is itself a translation into a language other than the language of the original, or the target language” (Kittel and Frank 1991, 3). There are two main advantages of the definition proposed by Kittel and Frank: 1. It does not limit the involved languages to only three – source language, mediat- ing language and target language – but opens up for an unlimited chain of trans- lations of translations, even though ITr typically involves only three languages. 2. The definition opens up the possibility to include compilation translation in the term indirect translation, i.e. translations with more than one source (language) (see Ivaska 2021). 58 Anja Allwood: (Semi)peripheries in contact Other terms and definitions have been suggested for this phenomenon but the termi- nology seems to have started to stabilize on indirect translation as the preferred term (cf. Ringmar 2007, 3; Assis Rosa, Pięta and Bueno Maia 2017, 114–17). 1 In this article, the term source language (SL) is used to denote the language of a work be- ing translated, whereas ultimate source language (ultimate SL) specifies the original au- thor’s first language choice. In a similar way the term ultimate target language (ultimate TL) is used to denote the final translation of interest for this study – however, nothing prevents the Swedish texts from being used as mediating texts for further translations in the future. The term mediating language (ML) will be used for the language into which a translation is done which is later used as a source text for a new translation. 1.2 The influence of (hyper)central languages and literatures English is “the most central language in the international translation system” (Heilbron 1999, 434). Heilbron’s findings are supported by more recent studies by e.g. Lindqvist (2015, 81) and Ringmar (2015, 159), who come up with percentages of 60% or more for English as the SL in the world at large and in Europe. However, all these calculations are based on UNESCO’s Index Translationum, which is not always relia- ble: the different national statistics that Index is based on are not entirely comparable (Heilbron 1999, 433), and the language classification systems vary – e.g. the two writ- ing standards of Norwegian have been classified into three languages (Ringmar 2015, 157). Despite these shortcomings, the Index is the most comprehensive database for world literature in translation and allows for large-scale comparisons. The (hyper)centrality of English and the other central languages, i.e. French and Ger- man, and possibly Russian, can in many ways be seen in semi-peripheral language areas, such as the Swedish, Spanish and Japanese (Lindqvist 2015). 2 Not only are they the source languages for a majority of the translations, according to Heilbron (1999, 436) a literary work from a (semi)peripheral language needs to have been translat- ed into a central language before it will be translated directly into another (semi) peripheral language, cf. Casanova (2005), as well as Heilbron and Sapiro (2007) on 1 For example both Dollerup (2000, 2014) and St André (2009) use the term relay translation. See e.g. Washborne (2013) and Pięta (2017) for further discussions regarding terms denoting this phenomena. 2 Lindqvist (2015, 81) compares Heilbron’s statistics over the global field of translation 1980–1989 with Sapiro’s statistics for 1990–1999 and her own research for the year 2012. Even though the statistics for the different years are not entirely comparable they show that the position of Russian as ultimate SL has decreased from 10% of the world’s translations in the 1980s to 1.5% in the 1990s and then up to 5% in 2012. 59 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 consecration and Lindqvist (2011) on double consecration. Marín-Lacarta (2012) draws a similar conclusion in her study on Chinese novels in Spanish translation 2001–2009: The dependence on the English (and to a certain degree the French) literary system re- sults in that the “same works are almost simultaneously published in different Western literary systems; a homogenized reception is taking place” (Marín-Lacarta 2012, 5). The importance of the central languages and, in the contemporary world, English above all, is thus valid for indirect translation as well (cf. Marín-Lacarta 2012). As stated in Ringmar (2015, 165) the mediating text “will be in a central language, where- as the original SL (and often the end TL) is peripheral. ” An example of this is that for ITr from Polish to Portuguese, French was the most used ML until the 1990s when it was overtaken by English (Pięta 2012). 2. Research questions The aim of this article is to investigate the use of ITr in novels translated into Swedish between the years 2000 and 2015, with a view to understanding the conditions in which different languages function as MLs in this particular context. To achieve this aim, three research questions are formulated: 1. Which languages are used as MLs for ITr during this period, and what are their respective proportions? 2. What could be the possible reasons for using a language other than English as the ML? 3. Which principles governing, or influencing the choice of ML can be identified in the material? 3. Research design Translations “need to be teased out of general bibliographies” (Paloposki 2018, 18), and ITrs, for their part, need to be teased out of translation bibliographies. Many ITrs are covert: indirectness is often obscured in translation metadata (see Ivaska 2020), possibly because of the cultural preference for directness in translation. Gideon Toury (1988, 139) observed more than 30 years ago the “prevalent cultural norm, ascribing upper- most value to translating from the original” (original emphasis), and this norm still seems to hold true: according to Alvstad (2017, 151), “[p]resent-day cultural policies in Scandinavia clearly favor direct over indirect translation.” ITr generally is “regard- ed as less desirable” (Assis Rosa, Pięta and Bueno Maia 2017, 113, cf. Ringmar 2015, 168–169). 60 Anja Allwood: (Semi)peripheries in contact The identification of both the overt and covert ITrs of novels into Swedish and the build- ing of the corpus, therefore, necessitated both quantitative and qualitative methods. A triangulation method was used where information was first gathered from Libris, the Swedish National Union Catalogue. The collected corpus was then subjected to a contex- tual analysis, where information about the specific authors and translators provided the background against which it was possible to verify the stated SLs – and, where present – the stated MLs of the translations (see further section 3.1–3.2; cf. Pięta 2012, 2019, 29–30). 3.1 Corpus The corpus of the study forms the main material used to answer research question 1: which languages are used as MLs, and what are their respective proportions? The cor- pus was designed to include first editions of all printed novels translated into Swedish during the period under study. There were some important restrictions: children’s literature as well as audio- and e-books and books published by Harlequin were ex- cluded from the material. E-books are not included in the current scope of research because, first, e-books were not systematically registered in Libris before 2012 when there was a change in legislation concerning submitting them for inclusion (Riksdags- förvaltningen 2012), and second, e-books do not compete on the market on terms equal to those of printed books. For example, the V AT on e-books in Sweden in 2015 was 25% and on printed books only 6% (Nilsson et al. 2015). Furthermore, e-books were borrowed at the public library rather than bought during the period under study (Nilsson et al. 2015; Facht 2012). Harlequin was a big player in the field (approxi- mately a third of all translated novels in a pilot study of the years 2000–2005), but the company’s business model in Sweden differed from that of other publishers (it only translated books with English as ultimate SL; Silberstein 2019, personal com- munication). Therefore, their inclusion would not have generated any indirect trans- lations. Furthermore, Harlequin books in Swedish have been thoroughly researched elsewhere (cf. Hemmungs Wirtén 1998; Lindqvist 2002). A search formula with Boolean operators was designed to extract the novels from Libris. The search had to be repeated individually for each of the 16 years of research data. Each title in the corpus was accompanied by information on author(s), translator(s), year of first publication, and source language(s), in addition to whether it was overtly acknowl- edged that the work had been translated indirectly. Basically, everything that Libris had classified as a novel was taken to be a novel. 3 Unwanted hits were excluded manually. 3 Libris on its part makes the decision on what material to classify as a novel based on how the material is presented in the book or by the publishing house (Persson 2019, personal communication). 61 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 These were mainly due to cataloguing errors and included reprints, collections of short stories, children’s books, poetry collections, audiobooks and e-books. To spot covert ITrs, contextual information on authors’ and translators’ language backgrounds was needed. Therefore, two registers were created. The first register con - tained information on the authors’ origin and language background. The second reg- ister contained information on the translators’ confirmed language skills as stated in digital databases from Swedish translators’ associations, e.g. SFÖ and Översättarcen- trum, as well as other available on-line resources, such as translators’ personal web pages, and news articles about the translators. 4 In some cases, where the informa- tion from Libris was incomplete or seemed particularly odd, the peritexts (cf. Genette 1997) of the target texts were consulted in order to establish if the oddities stemmed from the published novel itself or from Libris. In a few cases, where no other option was available, direct contact was made with translators or publishers via e-mail or phone calls, to confirm whether a translation was direct, or indeed indirect. It can obviously not be assumed that the translators only used the language edition stated in the Libris catalogue or elsewhere as the ML, or that they did not look for solutions or interpretations in translations of the same work into other languages. However, the most important fact for this study is that the publishing houses in ques- tion chose to publish an indirect translation – whether openly stated or not – instead of a direct one from the language the novel was originally written in. Direct transla- tion is, after all, what is normally to be expected, as stated earlier. The triangulation method used in this study can be compared to the one used by Pięta (2010, 2012), who studied more than a hundred direct and indirect book-length translations published between 1855–2010 from Polish to Portuguese. Pięta started with compiling a list of potential titles from different sources, “such as catalogues, databases, bibliographies, memoirs, periodicals, encyclopaedias and the like” (Pięta 2010, 4). Her material consisted of narratives, dramas and poetry, and in order to es- tablish which books were translated directly and which were translated indirectly she analysed first peritexts and then available epitexts. 5 Finally, where needed, Pięta did a comparative textual analysis of the Polish source texts, possible mediating texts and the Portuguese target texts. 4 SFÖ stands for “The Swedish Association of Professional Translators” and Översättarcen- trum is a Swedish non-profit association mediating contacts between translators and translation buyers. 5 Peritexts could be e.g. blurbs, prefaces, and annotations, whereas epitexts are located outside of (the covers of ) the novel such as reviews, archival documents and bibliographies (see Genette 1997; Pięta 2019). 62 Anja Allwood: (Semi)peripheries in contact Pięta thus studied a longer time span and a more restricted language set of transla- tions while my research focuses on a shorter period but takes in all translations of novels from all and any languages. The contemporaneous nature of the material under study in the present article means that: • All initial information could, at least in theory, be gathered from one and the same source (Libris), because the Swedish obligation for publishers to send their works to the National Library for registration is well established. • Much information on authors, translators and translations could be found online. Translators could also be contacted to confirm whether a specific translation was indeed direct or indirect. • The majority of the books were available for peritext consultation from the uni- versity library. Therefore, for the present article no general micro-level comparisons between differ - ent language versions were necessary. 3.2 Databases, libraries and book sites In order to answer research question 2, concerning possible reasons for using a lan- guage other than English as the ML, different library catalogues were consulted. These included the online versions of the British Library, Literature Online, Contemporary Authors Online, WorldCat and Princeton University Library. The aim was to establish whether the novels in question existed in English and, if so, what year they were first published. The publishing year was needed to confirm whether the English transla- tion could have served as a basis for the translation into Swedish or not, from a time perspective. Searches were conducted starting either with the title in the ultimate SL or with the name of the author and then narrowing the scope into books published in English. For research question 3, concerning which principles governing or influencing the choice of ML that can be identified in the material, further online libraries and oth- er websites were consulted, e.g. the German National Library, Amazon and Ency- clopædia Britannica. For the books where no potential principle was discernible from the online research, information was complemented by interviews with the Swedish publishers and one translator. 63 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 4. Results and discussion For the years 2000–2015, a total number of 5,259 first editions of both directly and in- directly translated printed novels were identified and collected from Libris. Of these, 70 novels (1.3%) were ITrs. English was the ML in 48 of the 70 novels, and 22 ITrs had a ML other than English. Of these 22 novels five were translated by the authors themselves into a more central language than that of the original text (see section 4.3 for an example), and these self-translations were then used as mediating texts for the translations into Swedish. 6 When the ultimate SL text and the self-translation into the ML were published in the same year, the Swedish language version of these ITrs was only included in the corpus if it could be confirmed through epitexts which language version was the ultimate SL and which was the ML. That the ultimate SL was written first did not necessarily always mean that it was published first. Novels for which it could not be established that the two language versions were not written simultane- ously were not included in the material (cf. Kellman 2003 for such examples). The self-translated novels are included in the corpus since their existence means that the Swedish publisher had the choice between two language versions and chose not to use the first, ultimate SL as the basis for the Swedish translation. 7 The reliability of the complied corpus is deemed high since the national database it is based on is governed by a legal obligation for the publishers to send in their works for registration. Sweden is not alone in this, and a similar system is in place in Bel- gium (Linn 2006, cited in Paloposki 2018, 18) and Finland (Therman 2007, cited in Ivaska 2020, 76). However, some errors and missing titles cannot be ruled out, due to – among other things – unclear peritexts, varying cataloguing practices among librarians, and general human factors. Most of the discovered misclassifications were, however, less severe: second editions classified as first editions, children’s books clas- sified as novels, and missing language codes when the SL was English. 6 Here, only the ones are counted where no other translator is mentioned in connection with the mediating text, and not translations made by a separate translator ”in collaboration with the author” , such as Bernardo Atxaga’s Soinujolearen semea (2003 ), in English The Accordionist’s Son (2007), which was written in Basque and translated into Castilian Spanish by Asun Garikano and Bernardo Atxaga (Libris; Santana 2009, 218). 7 It can of course be discussed if a self-translation should be regarded as a translation at all, or just as another language version by the same author. For T oury (2012, 100) a mediating translation in ITr has to be done by someone else than the original author, but in, for example, Assis Rosa, Pięta, and Bueno Maia (2017), where many different classifications and types of indirectness are discussed, there is no mention of how many agents should be involved for a text to be counted as an ITr. See also Ringmar (2015, 168) for a reflection on the usefulness of considering which language self-translators choose to first write in. 64 Anja Allwood: (Semi)peripheries in contact 4.1 Mediating languages As regards the first research question – Which languages are used as MLs for indirect translation during this period, and what are their respective proportions? – it can be concluded that 48 (68.5%) out of the identified 70 indirectly translated novels used an English translation as the mediating text. These novels will only be briefly dealt with in section 4.3, and what follows will focus on the remaining 22 novels, their MLs and their way into the Swedish literary system. Table 1 gives an overview of the numbers for the ultimate SLs of these 22 novels, the MLs and ultimate TL, i.e. Swedish. As shown in Table 1, French, with six novels, is the most used ML after English. French has mediated four different ultimate SLs: Albanian (2), Dari (1), English (1) and Turkish (2). After French follows German, mediating four novels and languages: Hebrew (1), Kerewe–Swahili (1), Kurdish (1) and Turkish (1). Russian also medi- ates four novels, but from three different ultimate SLs: Kazakh (2), Turkmen (1) and Ukrainian (1). Spanish mediates four novels from two ultimate SLs: Basque (1) and Galician (3). Dutch has been used as a mediating language for three novels, all of them with Bahasa Indonesia as the ultimate SL. Turkish, finally, was not only a language being mediated by other languages (English, French and German), but also served as a mediating language itself for one novel originally written in Kurdish. Table 1: The 22 novels translated into Swedish and mediated via a language other than English 2000–2015, sorted by ML. Ultimate SL Novels Mediating language Ultimate TL Albanian 2 French 6 Swedish 22 Dari 1 English 1 Turkish 2 Hebrew 1 German 4 Kerewe/Swahili 1 Kurdish 1 Turkish 1 Kazakh 2 Russian 4 Turkmen 1 Ukrainian 1 Basque 1 Spanish 4 Galician 3 Bahasa Indonesia 3 Dutch 3 Kurdish 1 Turkish 1 65 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 As is evident from Table 1, the MLs for novels translated indirectly into Swedish dur- ing 2000–2015 are both central languages such as French, German and Russian, as well as semi-peripheral languages, such as Spanish, and peripheral languages, such as Dutch and Turkish. From this we can draw the conclusion that centrality is not the only decisive factor for the choice of ML. 4.2 Reasons for using mediating languages other than English The second research question – What could be the possible reasons for using a language other than English as ML? – will be dealt with next. In the corpus, 22 of the 70 indi- rectly translated novels were mediated via a language other than English, i.e. 31.5%. Five plausible categories of explanations are identified as to why English was not used as ML for these 22 novels. Nine of the novels did not exist in English at the time of the Swedish translation. Here, obviously, the publishers had to take recourse to a ML other than English. However, by looking at the year of publication these circumstances were obvious only for eight out of the nine novels. For the ninth novel – a tertiary translation of a Kerewe novel self-translated into Swahili and then translated into German and finally into Swed- ish – the English version was published six years before the Swedish version. Here the publisher and the translator were consulted in order to find out why this English version was not used as the ML. It turned out that the novel had been translated from German into Swedish a few years before the English version was published, but that it, for copyright reasons, had been stranded at the publishing house for many years before it was finally published in Sweden. For seven of the 22 novels the English language version was already an indirect trans- lation, which would have made a Swedish translation of these works tertiary transla- tions. With the general negative attitudes towards indirect translation (cf. section 3), it is easy to see why the publishers would want to avoid that, if possible. It is interesting to notice however, that five out of these seven novels where published the same year in both English and Swedish, or with only one year’s difference. Therefore, it seems un - controversial to assume that the decisions to publish these works in English may have influenced the Swedish publishing houses’ decisions to also have the works translated. One of the 22 novels was published in English in 2013 and in Swedish in 2014. The English version was translated directly from the Galician ultimate SL (1998) where- as the Swedish one was translated from the Spanish translation (1998). From a time perspective, it is not likely that this English version was available as a mediating text for the Swedish translation. However, the English version may have been acting as a 66 Anja Allwood: (Semi)peripheries in contact marketing factor. The English and Swedish translations are published so close to each other in time – and more than fifteen years after the ultimate source text – that it is possible that the forthcoming English translation may have influenced the Swedish publisher to initiate their own translation, even though the English language version probably could not function as the ML in this case. 8 One novel was originally written in English and translated by the author herself into French. This French translation was published one year before the English original. Here it is possible that the staff at the Swedish publishing house in question did not know which language version was the original, especially since this author often writes directly in French and the French version was published first. Alternatively, it did not matter to them since it was the author who had done the translation. Finally, there were four novels (in addition to the Kerewe–Swahili–German–Swed- ish novel mentioned above) where the English language versions were direct transla- tions from the ultimate SL, and were published more than a year before the Swedish versions, but despite this English was not used as the ML. At first sight, there were no obvious reasons for not choosing English as the ML here. These four translations came from the same publishing house, three with the language combination Bahasa Indonesia–Dutch–Swedish and one with Dari–French–Swedish. The Dari–French–Swedish case seems the more straightforward one and the choice of French as ML not that surprising. Biographical information about the author revealed that he had moved from Afghanistan to France and started to write in French, and one of his later books, with French as ultimate SL, had already been translated directly from French into Swedish. It may therefore have seemed natural to continue translat- ing from French (albeit with a new translator), instead of using the existing English language version as the mediating text or maybe trying to find a translator from Dari. For the other three novels, more research was needed to establish why they were not translated via English as could have been expected, and thus the publisher was inter- viewed. The novels were translated from Bahasa Indonesia to Dutch and then into Swedish. According to the publisher, he tried to have the novels translated directly from Bahasa Indonesia and, when this failed, from English, but the result was not the expected. In the end the publisher chose to have the translations done from the Dutch language versions, which he assumed to be reliable, not least considering the histori- cal ties between Indonesia and the Netherlands. 8 The fact that a British or American prestige publisher is publishing a translation from a more peripheral language, is one of the best sales arguments for the original publisher when it comes to selling translation rights to other countries (Heilbron 2000, 17; Lindqvist 2012, 205). 67 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 The possible reasons for why English was not used as the ML for 22 of the indirectly translated novels 2000–2015 are summed up in Table 2. Table 2: Summary of reasons to choose a ML other than English for translation of novels into Swedish 2000–2015. No English language version existed at the time of the Swedish translation 9 The English language version was already an indirect translation 7 The English language version came out only one year before the Swedish and was probably not available as a source text 1 The specific author was previously translated via another language 2 Direct translation and translation via English had not generated the expected results 3 Total 22 4.3 Principles influencing the choice of mediating language The third research question – Which principles governing, or influencing the choice of ML can be identified in the material? – is dealt with below. As stated previously in section 4, English is the ML in almost 70% of the cases (which also corroborates earlier studies’ findings about the hyper-centrality of the English language, even in the case of ITr). Hence, the principle of using English as ML in indirect translation whenever possible is the most predominant and general principle during the period under study. However, the reasons for the use of other languages in ITr have not been explored in depth in the ITr literature until now. In order to give possible explanations for the language choices when English is not used as the ML seven additional principles were formulated: • The principle of self-translation • The principle of author–ML proximity • The principle of historical–cultural proximity • The principle of previous translations • The principle of the status of the mediating text • The principle of the personal choice of the publisher or translator • The principle of the only ML available. These categories are not hard and fast, as the principles may overlap, or co-exist, for some of the novels, and it is not feasible to judge which principle weighs heavier than another. However, as a rule of thumb, author-related principles are considered more prominent than linguistic principles, which in turn have been given more prominence than principles that concern the publisher, translator or other, more extrinsic, factors. 68 Anja Allwood: (Semi)peripheries in contact The principle of self-translation: Five novels, by four different authors, are self-transla- tions from the ultimate SL to the ML. An example of this is the Galician author Do- mingo Villar, who translates himself from Galician into Castilian Spanish, from which the translation into Swedish was made. Since Villar is living in Spain, where both his ultimate SL and the language he translates himself into are spoken, he could also be classified as belonging to the principle of author–ML proximity. The principle of author–ML proximity: If we let this principle include the physical proximity between the author and the geographical place where the ML is spoken, as well as the geographical proximity between the places where the two languages gen- erally are spoken, this principle is applicable to 11 of the 22 novels. An example is the Albanian author Ismail Kadare, who has lived in France since 1990 and is translated from Albanian via French into Swedish. This result is in agreement with the results in Pięta (2012, 324), where a correlation is noted between “the country of residence of the Polish émigré author at the time of publication of the Polish ST and the mediating language through which their work was translated into Portuguese” . The principle of historical–cultural proximity: In the current material, this principle concerns the historical ties between the Netherlands and its former colony Indone- sia. The principle is applicable to three novels, all written by the Indonesian author Pramoedya Ananta Toer, whose novels were translated from Bahasa Indonesia into Dutch and then further into Swedish. As noted previously, the publisher thought that the Dutch translation ought to be a good representation of the ultimate source text because of the historical ties with the ultimate SL culture. The principle of previous translations: This principle concerns two authors with one nov- el each. Both authors have changed the language they write in, and this is seen as the strongest factor here; it seems that in these two cases the Swedish publishers decided to use the same SL for the same author – even though they did not use the same translators as before. This seems true for Atiq Rahimi, who changed from writing in Dari to writ- ing in French, and probably, but to a lesser degree, for Nancy Huston, who has written most of her novels in French (and a few in English). However, both authors also live in a country where the ML is spoken, which means that the principle of author–ML proxim- ity also could have played a part in the Swedish publishers’ decisions regarding the ML. The principle of the status of the mediating text: This is a rather tentative principle as it only contains a single example in the corpus, i.e. the novel by the T urkish author Sema Kaygusuz. For the Swedish translation, the choice of possible MLs seems to have been between German and French, and the publishing house in question has German as one of its three focus languages (Ersatz 2019). Still, French was chosen as the ML. The only discernible difference between the German and French translations is that the 69 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 French version had been awarded two prestigious prizes before the publication of the Swedish translation, while the German language version was not similarly acclaimed (Prix Balkanika in 2008 and Prix France-Turquie in 2010, see France culture 2019). It is not possible to claim with certainty that this was the decisive factor for the publish- ing house regarding ML, but it seems plausible. The principle of the personal choice of the publisher or translator: This principle might be more common than is shown in this article, since it is likely that publishers, and translators suggesting novels for translation, use their own language skills and inter- ests as a starting point when looking for possible translations. However, in this corpus there is only one example that could be explained by this principle: the Kurdish author Mehmed Uzun who lived in Sweden for almost 30 years. His first novel translated into Swedish was written in Kurdish and the translator used German as the ML. At the time of the Swedish translation, French and Norwegian language versions were also available (the Norwegian was translated indirectly via German). The Swedish au- thor and translator, Ingmar Björkstén, was a personal friend of Uzun. In his memoirs (2002), Björkstén recounts how he learned that Uzun’s Swedish publisher could not find anyone to translate the novel directly from Kurdish. Björkstén offered to do the translation from the German language version, to which the publisher agreed. The translation was done in cooperation with the author himself. Björkstén stated that he preferred to translate from German, even though he also looked certain things up in the French and Norwegian versions (Björkstén 2002, 221). 9 Finally, there are three novels where the choice of the ML was the only option, as there were no other language versions published at the time. The principle for these three novels can be called the principle of the only ML available. For example, the Hebrew author Lizi Doron’ s (most often spelled Lizzie Doron in Swedish) novel was only trans- lated into German at the time of its translation into Swedish. For the sake of clarity, it should be pointed out that the proposed principles were formulated on the basis of analysis of the Swedish target culture and that they cannot be fully transferred into other target areas. I believe, however, that some principles, especially self-translation and author–ML proximity, could be taken as starting points for research in other areas and might be relevant in broader contexts. These principles seem to govern or influence the choices of the publishers when they make decisions on how to have a certain novel translated. Nonetheless, there are also other factors to be taken into account, such as the factors that “pull” literary translations into the 9 The possibility that translators of other novels in the database can also have used further translations into other languages cannot be excluded. However, the focus of this article is on novels that are not translated directly, not on tracing the genealogy of those texts. 70 Anja Allwood: (Semi)peripheries in contact target culture (e.g., publishers and translators) as well as the factors that “push” liter- ary translations into the target culture (e.g., agents and source culture institutions). 10 5. Conclusions The general assumption is that there is a hierarchy between direct and indirect trans- lations within the system where a literary work is translated from one language into another. If possible, a direct translation is preferred. If it is not possible to have a di- rect translation, e.g. because no translator of that particular language combination is available, indirect translation (possibly a compilation translation) is resorted to. The analyses in this study lend support to that assumption for translations into Swedish between the years 2000 and 2015. Out of the total of 5,259 translated novels in this period, 70 (1.3%) were ITrs. The first research question of this study concerned which languages were used as MLs for ITrs in Sweden in 2000–2015, and their respective proportions. As we have seen, English is used as the ML for translations of novels into contemporary Swedish in 48 of the 70 identified cases of ITr (making up 68.5% of all ITrs). The findings are thus in line with the general tendency of using a (hyper)central language as the ML in ITr, as discussed in section 1.2. The other languages used as mediating languages are, in descending order: French, German, Russian, Spanish, Dutch and Turkish. The study thus shows that even if the most common ML is English, the publishers sometimes have to choose a ML other than English. This leads us to the second re- search question, which was concerned with the possible reasons for using a language other than English as the ML. In the Swedish context, the most frequent reason was that the novel did not exist in English. This was the case with nine of the 22 novels that did not have English as the ML. In seven cases, the English version was itself an ITr, which may have made the publishers deem them not suitable as mediating texts. In total there were seven ITrs where the Swedish language version was published the same year or the year after the English version, but the English version was not used as the ML. In these cases it is likely that the Swedish publishers had knowledge about the upcoming English translation, and that this knowledge encouraged the Swedish publishers to organize their own translations, since these novels were likely to get in- ternational recognition which could increase the sales. From this we can draw the conclusion that the hyper-central English-language liter- ary system influences the Swedish literary system in more ways than first meets the 10 For further references to research on factors pulling and pushing literary translations see Paloposki (2018, 23–4), for research on supply-driven translation see Vimr (2020). 71 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 eye. Not only is English the SL for 61% of the 5,259 direct translations of novels into Swedish over the period 2000–2015, but it is also the ML for 68.5% of the 70 indirectly translated novels. In addition to this at least one third of the ITrs that are not mediated via English still seem to have been affected, in one way or another, by the fact that an English translation was being produced, or had just been published. Thus, to answer the third research question concerning which principles that can be identified in the material regarding the choice of ML, we can conclude that the overar- ching principle to use English as the ML whenever possible is very strong in Sweden. However, it does not seem stronger than the rule not to use English as the ML if this would create a tertiary translation into Swedish. In the choice between using English as the ML and at the same time creating a tertiary translation, or using another ML, another language is chosen. Here we can also notice that even though the central languages of French and German were the most common MLs after English other, more peripheral, languages could also function as the ML. In the Swedish context, the principle to use a translation made by the original author (i.e. a self-translation) as the mediating text also seems quite strong, and it sometimes coincides with the use of a (semi)peripheral language as the ML, e.g. Galician–Spanish–Swedish. Six further principles for the choice of ML were also identified in this study, and these were author–ML proximity, historical–cultural proximity, previous translations, the status of the mediating text, the personal choice of the publisher or translator and the translation in question being the only ML version available. Hopefully these principles could be further developed and in the future tested on and compared to the situation regarding ITr in other modern language situations. More research is also needed on the actual relationship between direct and indirect translations. For example, it would be informative to know if the Swedish figure of 1.3% indirectly translated novels out of all the 5,259 novels translated 2000–2015 is comparable to the numbers in other (semi)peripheral literary systems. In a Swedish context, the publishers’ role and reasoning in connection to ITr should be investigated further. After all, these agents are the ones who generally make the decision about whether a text should be translated directly or indirectly. However, for now we can conclude that (semi)peripheral languages sometimes func- tion as the ML for ITr into Swedish, and that there are novels that have been translated indirectly from a (semi)peripheral language into Swedish, even though they have not been translated into English yet. Heilbron (2000) describes such cases as the excep- tion that proves the rule, but these could also be seen as special instances where the agents involved, in particular, should be investigated further. The central languages are certainly important for ITrs, and centrality of the ML is also important, but these are not the only decisive factors for the choice of ML for an ITr. 72 Anja Allwood: (Semi)peripheries in contact References Primary sources Amazon. Accessed October, 2019. https://www.amazon.com/. British Library. Accessed October, 2019. http://explore.bl.uk/primo_library/libweb/ac- tion/search.do?vid=BLVU1. Contemporary Authors Online. Accessed October, 2019. https://www.gale.com/intl/c/ contemporary-authors-online. Encyclopædia Britannica. Accessed October, 2019. https://www.britannica.com. Ersatz. Accessed October, 2019. http://www.ersatz.se/about.htm. France culture. Accessed October, 2019. https://www.franceculture.fr/personne-sema-kay- gusuz. German National Library. Accessed October, 2019. https://www.dnb.de/DE/Home/ home_node.html. Libris. Accessed October, 2019. https://libris.kb.se/. LION (Literature Online) Reference Tool. Accessed October, 2019 via Gothenburg Uni- versity Library. Persson, Ingvar, librarian at Kungliga biblioteket, Stockholm. Personal e-mail 2019-12-09. Princeton University Library. Accessed October, 2019. https://library.princeton.edu/. SFÖ. “The Swedish Association of Professional Translators”. Accessed May, 2020. https:// sfoe.se/en/. Silberstein, Anna, editor at HarperCollins Nordic. Personal e-mail 2018-08-24. WorldCat. Accessed October, 2019. https://www.worldcat.org/. Översättarcentrum. Accessed October, 2019. http://www.oversattarcentrum.se/sv.html/ english. Secondary sources Alvstad, Cecilia. 2017. “ Arguing for Indirect Translations in Twenty-First-Century Scan- dinavia. ” Translation Studies 10 (2): 150–65. https://doi.org/10.1080/14781700.2017.1 286254. Assis Rosa, Alexandra, Hanna Pięta, and Rita Bueno Maia. 2017. “Theoretical, Method- ological and Terminological Issues Regarding Indirect Translation: An Overview.” Translation Studies 10 (2): 113–32. https://doi.org/10.1080/14781700.2017.1285247. Björkstén, Ingmar. 2002. En dag, en tid: möten och minnen. Rimbo: Fischer & Co. Casanova, Pascale. 2005. “Literature as World. ” New Left Review 31: 71–90. Dollerup, Cay. 2000. “Relay and ‘Support’ Translations. ” In Translation in Context: Selected Contributions from the EST Congress, Granada, 1998, edited by Andrew Chesterman, 73 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 Natividad Gallardo San Salvador and Yves Gambier, 17–26. Benjamins Translation Library. Amsterdam and Philadelphia: John Benjamins Publishing. Dollerup, Cay. 2014. “Relay in Translation.” In Cross-Linguistic Interaction: Translation, Contrastive and Cognitive Studies: Liber Amicorum in Honour of Prof. Bistra Alexieva, Published on the Occasion of Her Eightieth Birthday, edited by Diana Dimitrova Jank- ova, 1–12. Sofia: St. Kliment Ohridski University Press. https://cms13659.hstatic.dk/ upload_dir/docs/Publications/232-Relay-in-translation-(1).pdf. Facht, Ulrika. 2012. “Aktuell statistik om e-böcker sammanställd av Ulrika Facht.” In Läsarnas marknad, marknadens läsare: en forskningsantologi utarbetad för Littera- turutredningen, edited by Ulla Carlsson and Jenny Johannisson, 231–40. Göteborg: NORDICOM. Genette, Gerard. 1997. Paratexts: Thresholds of Interpretation. Cambridge: Cambridge University Press. Heilbron, Johan. 1999. “Towards a Sociology of Translation – Book Translations as a Cul- tural World-System. ” European Journal of Social Theory 2 (4): 429–44. Heilbron, Johan. 2000 “Translation as a Cultural World System. ” Perspectives 8 (1): 9–26. https://doi.org/10.1080/0907676X.2000.9961369. Heilbron, Johan, and Gisèle Sapiro. 2007. “Outline for a Sociology of Translation – Cur- rent Issues and Future Prospects.” In Constructing a Sociology of Translation, edited by Michaela Wolf and Alexandra Fukari, 93–107. Benjamins Translation Library 74. Amsterdam and Philadelphia: John Benjamins Publishing. Hemmungs Wirtén, Eva. 1998. “Global Infatuation: Explorations in Transnational Pub- lishing and Texts: The Case of Harlequin Enterprises and Sweden. ” Skrifter utgivna av Avdelningen för litteratursociologi vid Litteraturvetenskapliga institutionen i Uppsa- la 38. PhD diss., Uppsala: Uppsala universitet. http://www.littvet.uu.se/digitalAs- sets/24/24781_global-infatuation.pdf. Ivaska, Laura. 2016. “Uncovering the Many Source Texts of Translations: Indirect Trans- lations of Modern Greek Prose Literature into Finnish 1952–2004. ” Poster presented at the EST Congress 2016, Aarhus, Denmark. https://www.academia.edu/28664155/Un- covering_the_Many_Source_Texts_of_Translations_Indirect_Translations_of_Mod- ern_Greek_Prose_Literature_into_Finnish_1952_2004_poster. Ivaska, Laura. 2020. “Identifying (Indirect) Translations and Their Source Languages in the Finnish National Bibliography Fennica: Problems and Solutions. ” MikaEL 13: 75– 88. Ivaska, Laura. 2021. “The Genesis of a Compilative Translation and its de Facto Source Text.” In Genetic Translation Studies: Conflict and Collaboration in Liminal Spaces, edited by Ariadne Nunes, Joana Moura and Marta Pacheco Pinto, 71–88. London: Bloomsbury. http://dx.doi.org/10.5040/9781350146846.ch-005. 74 Anja Allwood: (Semi)peripheries in contact Kellman, Steven G. 2003. Switching Languages: Translingual Writers Reflect on Their Craft. Lincoln: University of Nebraska Press. Kittel, Harald, and Armin Paul Frank. 1991. “Introduction.” In Interculturality and the Historical Study of Literary Translations, edited by Armin Paul Frank and Harald Kit- tel, 3–4. Göttinger Beiträge Zur Internationalen Übersetzungsforschung 4. Berlin: Er- ich Schmidt. Lindqvist, Yvonne. 2002. Översättning som social praktik: Toni Morrison och Harlequinse- rien Passion på svenska. Acta Universitatis Stockholmiensis, New Series 26. Stockholm: Almqvist & Wiksell International. Lindqvist, Yvonne. 2011. “Dubbel konsekration – en förutsättning för svensk översättning av utomeuropeisk litteratur? Maryse Condé som exempel. ” Språk och stil 21: 140–70. Lindqvist, Yvonne. 2012. “Det globala översättningsfältet och den svenska översättnings- marknaden: Förutsättningar för litterära periferiers möte”. In Läsarnas marknad, marknadens läsare: en forskningsantologi utarbetad för Litteraturutredningen, edited by Ulla Carlsson and Jenny Johannisson, 197–207. Göteborg: NORDICOM. Lindqvist, Yvonne. 2015. “Det skandinaviska översättningsfältet – finns det?” Språk och stil 25: 69–87. Marín-Lacarta, Maialen. 2012. “ A Brief History of Translations of Modern and Contem- porary Chinese Literature in Spain (1949–2009).” 1611: Revista de historia de la tra- ducción 6: 1–7. Nilsson, Skans Kersti, Elena Maceviciute, Tom Wilson, Annika Bergström, and Lars Hö- glund. 2015. “The Tensions of E-Book Creation and Distribution in a Small-Language Culture. ” Northern Lights 13 (1): 29–47. https://doi.org/10.1386/nl.13.1.29_1. Paloposki, Outi. 2018. “The Missing Needle: Bibliographies, Translation Flows and Re- translation.” In Translating Scandinavia: Scandinavian Literature in Italian and Ger- man Translation, 1918–1945, edited by Bruno Berni and Anna Wegener, 15–28. Ana- lecta Romana Instituti Danici. Supplementum 50. Rome: Edizioni Quasar. Pięta, Hanna. 2010. “Portuguese Translations of Polish Literature Published in Book Form: Some Methodological Issues. ” In Translation Effects: Selected Papers of the CET- RA Research Seminar in Translation Studies 2009, edited by Omid Azadibougar, 1–25. Leuven: CETRA. https://www.arts.kuleuven.be/cetra/papers/files/hanna-pieta-portu- guese-translations-of-polish.pdf. Pięta, Hanna. 2012. “Patterns in (in)Directness: An Exploratory Case Study in the Exter- nal History of Portuguese Translations of Polish Literature (1855–2010). ” Target 24 (2): 310–37. https://doi.org/10.1075/target.24.2.05pie. Pięta, Hanna. 2017. “Theoretical, Methodological and Terminological Issues in Research- ing Indirect Translation: A Critical Annotated Bibliography.” Translation Studies 10 (2): 198–216. https://doi.org/10.1080/14781700.2017.1285248. 75 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77 Pięta, Hanna. 2019. “Indirect Translation: Main Trends in Practice and Research. ” Slovo. Ru: Baltijskij Accent 10 (1): 21–36. http://dx.doi.org/10.5922/2225-5346-2019-1-2. Pokorn, Nike. 2013. “Experience through Translation – the Translated Experience: The Turkish Presence in Slovene Literature and Translation.” Across Languages and Cul- tures 14 (2): 167–81. https://doi.org/10.1556/Acr.14.2013.2.2. Riksdagsförvaltningen. 2012. Lag (2012:492) om pliktexemplar av elektroniskt material, Svensk författningssamling 2012:2012:492 t.o.m. SFS 2018:1922 - Riksdagen. https:// www.riksdagen.se/sv/dokument-lagar/dokument/svensk-forfattningssamling/ lag-2012492-om-pliktexemplar-av-elektroniskt_sfs-2012-492. Ringmar, Martin. 2007. “‘Roundabout Routes’: Some Remarks on Indirect Translations. ” In Selected Papers of the CETRA Research Seminar in Translation Studies 2006, 1–17. Leuven: CETRA. https://www.arts.kuleuven.be/cetra/papers/files/ringmar.pdf. Ringmar, Martin. 2015. “Figuring Out the Local within the Global: (Sub)Systems and Indirect Translation. ” In IberoSlavica Special Issue of IberoSlavica: Translation in Iberi- an-Slavonic Cultural Exchange and Beyond, edited by Teresa Seruya and Hanna Pięta. 153–78. Santana, Mario. 2009. “On Visible and Invisible Languages: Bernardo Atxaga’s Soinujo- learen Semea in Translation. ” In Writers in between Languages: Minority Literatures in the Global Scene, edited by Maria Jose Olaziregi, 213–29. Center for Basque Studies Conference Papers Series 5. Reno: Center for Basque Studies/University of Nevada. St André, James. 2009. “Relay.” In Routledge Encyclopedia of Translation Studies, edited by Mona Baker and Gabriela Saldanha, 2 nd ed., 230–32. London and New York: Rou- tledge. Toury, Gideon. 1988. “Translating English Literature via German – and Vice Versa: A Symptomatic Reversal in the History of Modern Hebrew Literature.” In Die literar- ische Übersetzung. Stand und Perspektiven Ihrer Erforschung, edited by Harald Kittel, 139–57. Berlin: Erich Schmidt. Toury, Gideon. 2012. Descriptive Translation Studies and Beyond: Revised Edition. Am- sterdam and Philadelphia: John Benjamins Publishing. Vimr, Ondřej. 2020. “Supply-Driven Translation: Compensating for Lack of Demand. ” In Translating the Literatures of Small European Nations, edited by Rajendra A. Chitnis, Jakob Stougaard-Nielsen, Rhian Atkin and Zoran Milutinović, 48–68. Liverpool: Liv- erpool University Press. Washbourne, Kelly. 2013. “Nonlinear Narratives: Paths of Indirect and Relay Translation. ” Meta: Journal Des Traducteurs 58 (3): 607–625. https://doi.org/10.7202/1025054ar. 76 Anja Allwood: (Semi)peripheries in contact About the author Anja Allwood is a PhD student at the Department of Swedish at the University of Gothenburg, Sweden. Her main research interests include indirect translation, liter- ary translation in a broad sense and publishing. She is a trained translator from Dutch and English to Swedish with an MA in Translation Studies. Her professional expe- rience spans from subtitling to translating for European Union institutions. She has taught university courses in practical translation, Dutch and Swedish. 77 Stridon. Journal of Studies in Translation and Interpreting, Volume 1 Issue 1, pp. 57–77