David Bellos ’ Indirect Translation of Ismail Kadare ’ s The File on H : A Contextual Analysis

This article is a linguistic study of David Bellos’ indirect translation of Ismail Kadare’s The File on H (1997), a novel first published in 1980-1981 in the Albanian literary review Nëntori, and translated into English on the basis of Jusuf Vrioni’s French version, Le Dossier H (1989). Also called “double”, “mediated” or “second-hand”, indirect translation is an understudied phenomenon, often criticised by scholars because of its greater distance to the original. Cay Dollerup (2000: 23), for example, argues that the grammatical structure of the mediating language (ML) obscures the distinctions made in the source language (SL), and that possible “mistakes” in the ML may be repeated in the target language (TL). Do fidelity and loyalty to the author become weakened in Bellos’ indirect translation? To what extent is such weakening discernible linguistically? And does this particular case of indirect translation reveal notable patterns or recurring types of linguistic shifts between ST and TT? Showing that some of the features specific to Kadare’s Albanian writing are tempered in the doubly-translated English text, yet highlighting that similar shifts occur in the three language directions involved, this article demonstrates that changes between ST and TT may occur in indirect translation regardless of the strategies adopted by MT – thus challenging the hypothesis that linguistic shifts in indirect translation follow a single or consistent pattern.


Introduction
"There is no such thing as a perfect, ideal, or 'correct' translation" (Newmark 1988: 6).In every representation of a text in another language modifications of the original occur.The term commonly used to refer to these changes is "shifts" (Catford 1965;Popovič 1970;Toury 1980;Leuven-Zwart 1989).Originally defined by John C. Catford (1965: 73) as "departures from formal correspondence in the process of going from the SL to the TL", the concept was later broadened by Anton Popovič to include not only linguistic phenomena, but also replacements arising from textual, literary or cultural considerations.For Popovič (1970: 79), "[a]ll that appears as new with respect to the original, or fails to appear where it might have been expected, may be interpreted as a shift".Popovič further stresses the inevitability of certain linguistic differences and the influence that the context in which translation occurs may have on the emergence of translation shifts.In his approach, shifts are not something to be avoided or repudiated, but mere manifestations of the relationship between source and target texts.
Critics have frequently dismissed translation shifts as "errors" or "mistakes" (Shuttleworth and Cowie 1997: 153), and negative accounts making use of this evaluative terminology prevail particularly in the literature on indirect translation (St. André 2009: 230).As a relatively understudied practice, also referred to as "double", "mediated" or "second-hand", indirect translation has often been frowned upon by scholars because of its greater distance to the original (Ringmar 2012: 141).Cay Dollerup (2000: 23), for example, argues that the grammatical structure of the mediating language (ML) obscures the distinctions made in the source language (SL), and that possible linguistic mistakes in the ML may be repeated in the target language (TL).Moreover, Dollerup claims, translators tend to take more liberties with MTs than with STs because on arrival in the target culture a text already translated has a lower status than an original texttherefore, the translator is less inclined to preserve its features if they deviate from target culture norms.In this sense, "fidelity and loyalty to the author become weakened", according to Dollerup (2000: 23), "not out of ill will, but for practical reasons".
The English version of Dosja H, a short novel by the Man Booker International Prize winner Ismail Kadare, presents a fascinating case of indirect translation.First published in 1980First published in -1981 in the Albanian literary review Nëntori, and translated into English via Jusuf Vrioni's French translation, Le Dossier H (1989), Kadare's novel links three relatively distant Indo-European languages: Albanian (the modern representative of a distinct branch of the Indo-European language family), French (a Romance language deriving from Latin) and English (a West Germanic language closely related to German and Dutch).Do, as Dollerup suggests, fidelity and loyalty to the author become weakened in Bellos' indirect translation?To what extent is such weakening discernible linguistically?And does this particular case of indirect translation show notable patterns or recurring types of linguistic shifts between SL and TL?In other words, how does Bellos' English translation (E) relate to Kadare's Albanian original (A), and to what extent is it influenced by Vrioni's French version (F)?
The present study examines the nature of the changes arising between ST and TT, and the role played by MT in these shifts, through comparative analysis of the linguistic composition of Ismail Kadare's Dosja H (1990), its French version Le dossier H (1989) by Jusuf Vrioni, and the French-based English translation The File on H (1997) by David Bellos.Its main objective is to interrogate the hierarchy underlying the binary opposition direct vs. indirect at the centre of Dollerup's claim that translating via a mediating text may result in a less faithful translation.In the first section, I present the theoretical, historical and linguistic background of the current investigation.I begin by exposing the central issues surrounding the phenomenon of indirect translation, before moving onto a discussion of the historical, political and literary context of the writing of the novel; I finally turn to a description of the contrastive linguistic model adopted and its relevance for this inquiry.The second section is devoted to the linguistic study itself.Divided into five sections, it examines the semantic, syntactic and contextual shifts arising in the three language directions involved (A → F, F → E and A → E), at a word, phrase and sentence level.The final section summarises the findings of the analysis and draws some tentative conclusions about the main assumptions surrounding the phenomenon of indirect translation.

Indirect translation
In their Dictionary of Translation Studies, Shuttleworth and Cowie (1997: 76) define indirect translation as "the procedure whereby a text is not translated directly from an original ST, but via an intermediate translation in another language".According to this definition, an indirect translation is a translation of a translation, where the source language (SL), the mediating language (ML) and the target language (TL) differ.Indirect translation is a common and longestablished practice, and examples can be found without searching too far.The history of Bible translation, for example, provides many striking instances of second-hand translations.In fact, "it is almost impossible to examine literary exchange, especially historically, without coming across this phenomenon" (Ringmar 2007: 4).Despite its omnipresence, indirect translation remains a marginal area of research in translation studies (Pięta 2012: 311).The main reason for this negligence, according to St. André (2009: 230), is that it replicates the stigma attached to translation itself.As Toury (1995: 143) explains, "the concept of translation itself ha[s] been undergoing [changes] towards an ever-growing emphasis on adequacy, which inevitably [involve] a concomitant reduction of tolerance for indirect translation as a whole".However, indirect translation forms a "legitimate object for research" (Toury 1995: 130;italics in Toury) and interest in the phenomenon is growing (Pięta 2012: 311).The burgeoning literature on the subject has been accompanied by a fluctuating terminology which reveals a lack of consensus concerning the metalanguage used, as well as the difficulty to theorise the process across the variety of practices encountered.Kittel (1991: 26), for example, differentiates between "intermediate" (first-hand) and "mediated" (second-hand) translation, whereas Toury (1995: 129) uses the terms "intermediate", "indirect" and "second-hand" interchangeably.Dollerup (2009) contrasts indirect translation (where the mediating translation serves merely as a stepping stone toward the final translation) and relayed translation (where the mediating text is intended for a real audience, as it is the case in the present research).Further, Koskinen and Paloposki (2010: 294) point out that indirect translation is sometimes considered a subset of retranslation -although, in Gambier's view (1994: 413), the latter term should be reserved for multiple translations of a single source text into one target language.A quick overview of the main terms used in the most recent English-language literature on the subject provides a good illustration of this diverse terminology (Table 1).

Indirect
Because of its increased distance from the source text, indirect translation has been perceived as an inferior form of translation.Toury (1995: 129) describes the prevailing attitude towards the phenomenon as if it were "some kind of disease to be shunned".Evidence of this widespread suspicion can be found in many areas, including institutional translation, translation theory and literary translation (see Ringmar 2007: 2).The hierarchy underlying the opposition direct vs. indirect constitutes the starting point of the present article and the principal aspect that it aims to explore through analysis of the linguistic shifts at play in this particular case of mediated translation.Thus, the main focus of the analysis is the overarching trajectory (ST1 → MT/TT1/ST2 → TT2), with special attention given to the role of MT in the modifications emerging during the transfer from ST1 to TT2.In addition, two direct forms of translation are considered (ST1 → TT1 and TT1/ST2 → TT2), for in any given doublytranslated text the following operations take place: The analysis further includes a third form of direct translation (ST1 → TT2), since, as we shall see in the following section, Bellos had partial access to Kadare's original text on several occasions., a forty-year totalitarian rule characterised by an atmosphere of suspicion and excessive nationalism.To eliminate dissent, the government imprisoned people in forced-labour camps; travel abroad was forbidden after 1968 to all but diplomats; and Western European culture was looked upon with deep suspicion, often resulting in bans on unauthorised foreign material and arrests of those who owned it.Usually ironic and darkly comic, Kadare's novels criticise Hoxha's regime, and The File on H is no exception.

Ismail
First published in two parts in 1980 and 1981, the novel tells the story of two Irish-American scholars from Harvard who travel to Albania in the 1930s to investigate the tradition of oral epic poetry.Their aim is to find out whether Homer was the author of the Iliad and the Odyssey, or simply adapted them from existing epics.With this objective in mind, they set out to record samples of Albanian rhapsodies in order to analyse their changes over time.When they reach their destination, they are confronted by suspicious provincial townspeople and a paranoid local governor who sends informers after them, convinced that they are spies.They manage to record the songs of several rhapsodists and to identify interesting variations among the songs, but in the end their work is destroyed by a Serbian pastor who thinks that the scholars' research will establish Albanian epic poetry as the original and the Serbian as an imitation.Used as an indirect way of speaking about the atmosphere under Hoxha's totalitarian rule (the "H" in The File on H standing for both "Homer" and "Hoxha"), the novel can also be seen as a metaphor for translation.
The File on H has been selected specifically for analysis from the numerous cases of indirect translation in Kadare's oeuvre, because translation represents a central theme in the novel.The dual attitude of suspicion and admiration surrounding the two foreigners echoes the notion of belles infidèles according to which beauty and fidelity cannot coexist in translation (see Salama-Carr 2011: 406-408).In fact, multilingualism plays a crucial role in the narrative, triggering plot development and provoking some of the most comic situations in the novel.Furthermore, the rhapsodiststhe researchers' main object of studyare themselves symbols of translation.As travellers, they epitomise cultural exchange and invoke the spatial etymology of the term "translation" (from the Latin translatio, "carried across").Their practice, which consists of adapting and diffusing existing works, resembles that of translators, and the question of variation is central to both oral poetry and to translation.
Variations between source and target texts are of particular importance when analysing indirect translations because the presence of a mediating text is expected to intensify the gap between the original and the final translation.Differences are supposedly even greater when the intermediate translation has a genuine audience in the first target language, since the translator is more likely to be affected by the norms of the intermediary culture than when he is not swayed by considerations for a real readership (Toury 1995: 129-146).David Bellos' version of the The File on H is based on a translation designed for a real audience.Jusuf Vrioni's French translation was published by Les Editions Fayard in 1989 for an authentic francophone readership, and with no intention of serving as a basis for further translation.In fact, David Bellos' English version did not appear until 1997 and Bellos' involvement in the project was impromptu: When I was asked ten years ago by Christopher MacLehose, the literary director of The Harvill Press, to take on the translation into English from the French translation of Ismail Kadare's Dosja H, I was initially dubious in the extreme.I knew no Albanian (and even now know only the tiniest scraps of that strange and difficult tongue).I also had principles!Enough damage can be done in one language shift to make a double shift seem like a recipe, if not for disaster, then at least for pretty thin gruel.On the other hand, the book fascinated me; as I began to read more of Kadare in French I also realized that what was being put on my plate was a writer of the first importance, with an oeuvre that was wide-ranging, coherent, intricately connected… and certainly worth bringing into English as well as could be done.But why not get it translated direct?I asked MacLehose.He raised his long arms to the heavens.If you only knew… was his mysterious and unanswerable reply.(Bellos 2005: 1) Bellos' description of indirect translation as a "recipe for disaster" in this passage echoes the general distrust surrounding the phenomenon, as discussed in 2.1, and posits the publisher's choice of indirect translation in this instance as inexplicable.
There may exist various reasons why a mediating text is chosen for translation, the most obvious being the lack of competent translators in the SL (Ringmar 2012: 141).This is certainly a plausible reason for The File on H, since contact with Western countries was prohibited during Enver Hoxha's dictatorship.This meant that a very limited number of Albanians were allowed to travel abroad and few native English speakers could visit the country to learn the language.Consequently, only a handful of people were able to undertake literary translations of this scale.Moreover, given the country's special relationship with France, more translators were available to translate into French, including Jusuf Vrioni and Tedi Papavrami, two of Kadare's best-known French translators.According to Bellos, the political context played an important role in The Harvill Press' decision to base the English translation of Dosja H on Jusuf Vrioni's French version, instead of Kadare's Albanian original.During the communist era, he explains, intellectual property was not recognised as such in Albania, and there were no Albanian rights of Kadare's originals available for purchase during that time.The French texts by Vrioni, by contrast, were Fayard editions.They followed the Universal Copyright Convention and were therefore available for purchase by publishers wishing to translate them into English (Bellos 2008: 17).
The circumstances under which the writing and publishing of Dosja H and its subsequent French and English translations took place have a direct bearing on this investigation.Translations form part of a wider context of norms and conventions, which may influence the translator's choices or have an indirect effect on his interaction with the text being translated.Decision-making is never purely linguistic in translation; or rather, linguistic choices will themselves be affected by political, social or historical factors.Contextualising this particular case of indirect translation is thus essential for understanding the role that such elements may have played in the linguistic shifts occurring between A, F and E.
In "The Englishing of Ismail Kadare: Notes of a Retranslator" (2005), for example, Bellos draws attention to a crucial aspect of his translating task: the fact that he was told by an Albanian doctor, Ylli Hassani, where the French adaptation departed a lot, or a little, from the source text.Bellos does not specify these passages or what proportion of the work they represent.However, Hassani's involvement suggests that Bellos had partial access to the source text and that his version might on several occasions be a direct rendering of Kadare's Albanian text rather than a translation of Vrioni's French translation.

Methodology
In this article I compare the linguistic composition of the Albanian, French and English versions of Kadare's Dosja H, following Joseph L. Malone's classification of translation shifts in The Science of Linguistics in the Art of Translation: Some Tools from Linguistics for the Analysis and Practice of Translation (1988).According to Malone, translation shifts can be divided into five main categories, four of which are split into further subcategories ( Malone's model suits the purpose of this article particularly well as it provides a concise, yet elaborate, framework for analysing the shifts at play in this particular case of indirect translation. Other typologies have been proposed by scholars for characterising translation shifts.In Comparative Stylistics of French and English ([1958]1995), Jean-Paul Vinay and Jean Darbelnet identify a total of seven translation procedures, organised into two distinct groups: on the one hand, loan, calque and literal translation are presented as "direct" strategies; on the other hand, transposition, modulation, equivalence and adaptation are described as "oblique".Vinay and Darbelnet's succinct taxonomy is effective, but not suitable for this article since their four "oblique" categories, the ones that involve notable linguistic modifications, are insufficiently detailed to account for the variety of literary alterations identified.Conversely, Kitty van Leuven-Zwart's comprehensive comparative model (1989), which comprises 37 subcategories grouped into three major branches (modulation, modification, and mutation), is unnecessarily complex.Similarly, Andrew Chesterman's typology of 30 strategies (1997), which would be too long to list here, is very extensive and therefore not adapted for this small-scale research.I have chosen to use Malone's taxonomy because, with its five categories and nine subcategories, it is sufficiently fine-grained, without being overly exhaustive or convoluted.It thus allows for both precision and clarity.
The analysis of the translation shifts between A, F and E is thus framed within the subsequent model, adapted from Malone (Table 4): I use Malone's subcategories of Equation, Substitution, Reordering, Amplification, Reduction, Diffusion, Condensation, Divergence and Convergence to describe the variety of translation shifts arising in all three language directions examined, focusing mainly on alterations that tend to attenuate or emphasise Kadare's linguistic mannerisms.

Linguistic comparison of A, F and E
In this section, I examine the nature of the shifts at play in the double translation of Dosja H through a comparative analysis of the linguistic composition of its Albanian, French and English versions.I inspect the differences arising between ST1 and TT2, and the role played by MT (TT1/ST2) in the emergence of these shifts, based on 15 examples of modifications which appear to dissolve the nuances or subtleties of Kadare's original text.Each of the instances discussed draws attention to deviations that may alter some of the traits characteristic of Kadare's writing (semantic ambiguity, critical irony, idiomatic expressions, etc.).These examples are taken from the first 15 pages of the novel (7-22 in the Albanian version)presumably the reader's first point of entry into Kadare's text.Five main types of deviations are identified and three examples are provided to illustrate each of them.These examples are discussed in so far as they confirm or contradict the main assumptions about indirect translation discussed in 2.1.Their aim is not to quantify the number of deviances arising between the texts examined, but to characterise the variety of patterns, or lack thereof, informing the shifts at play in this particular case of indirect translation.
In other words, their role is not to establish how much the three language versions of Kadare's novel differ from one another, but rather to describe how (i.e. the different ways in which) they do so.

Deviations from A to F, where E follows F
Below are instances of F deviating from A (A ≠ F), and E repeating the changes initiated by F (F = E).They are examples of what Dollerup calls the repetition of "mistakes" from MT to TT, which appear to confirm his hypothesis that fidelity to the author may become weakened in indirect translation because possible deviances in the ML will be repeated in the TL.This dynamic can be represented as follows: A ≠ F = E.
(1) SUBSTITUTION (A) Pra, pa i lejuar vetes t'i japë këshilla ndokujt (Kadare 1990: 19-20) Thus, without allowing himself to give advice to anyone (F) Ainsi, sans prétendre vouloir prodiguer des conseils à qui que ce fût (Kadare 1989: 19) Thus, without claiming to want to give advice to anyone (E) So without claiming to offer advice to anyone (Kadare 1997: 13) The Albanian uses the reflexive verb "lejuar vetes" ("to allow oneself"), which in the French translation is replaced by "sans prétendre vouloir" ("without claiming").The most evident counterpart to the Albanian verb in French is not "prétendre vouloir" but "se permettre de" ("to allow oneself") or "s'autoriser à" ("to authorise oneself").This substitution is carried over onto the English version, which, as one might expect, translates "prétendre vouloir" as "claiming" by contracting the MT item without omitting a layer of meaning.The modification initiated by F and condensed in E attenuates in both texts the idea of selfcensorship conveyed by A's "pa i lejuar vetes" ("without allowing himself").
(E) (1) That is how the dead are buried, (2) she found herself thinking, but she cut off that thought, as she always did whenever a (3) macabre or even painful idea came into her mind.(Kadare 1997: 7-8) Here, the English translationfollowing the mediating French textcarries more information than the Albanian original on three occasions.(1) Firstly, F's "C'est de cette manière qu'on ensevelit les morts" ("that is how the dead are buried") amplifies A's "kështu e mbulojnë njeriun" ("that is how people are covered") by converting A's suggestive image of a covered body into the blatant representation of a buried corpse.E's "that is how the dead are buried" replicates F's explicitation and thus reduces mimetically the ambiguity of A's evocative phrasing.
(2) Secondly, the clause "nisi të mendonte" ("she began to think") in A is amplified in F ("se prit-elle à songer"), which E faithfully renders as "she found herself thinking".Whilst A simply signals an initiation of thought, F and E further express an element of surprise, which magnifies the character's state of alienation.(3) Thirdly, F translates "të shëmtuar" as "macabre", thus infusing the adjective with a more specific meaning than its Albanian counterpart.The epithet "të shëmtuar", used to qualify ugliness when applied to physical appearance, refers more broadly to a negative moral quality when associated with ideas or thoughts.Kadare could have chosen "i tmerrshëm", a more common equivalent for "macabre" in Albanian, but he uses the general qualifier "të shëmtuar", which allows for more ambiguity as to the exact nature of the character's ideas.By conferring this epithet a more precise meaning, interpreted from its use in context, F and E thus accentuate the amplification operated in the first clause ("That is how the dead are buried").In addition, both F and E omit "papritur" ("suddenly"), further confirming the influence of F on E's interpretation of A.
(3) AMPLIFICATION (A) "Pra ejani pakëz këtu, pëllumbthat e mi, lërini një çikë ato eposet e magnetofundet dhe uluni të bëni një punë për lalën."(Kadare 1990: 9) "Come here, my little doves, leave for a moment those epic songs and taperecorders, and sit down to do something for lalën [an authority figure]." (F) "Venez donc par ici, mes petits agneaux, laissez tomber un petit moment vos épopées et vos maniétofonds, et asseyez-vous qu'on discute un brin.» (Kadare 1989: 9) "Come here, my little lambs, leave for a moment those epic songs and taperecorders, and sit down so we can discuss a little bit." (E) "Come this way, my lambs.Let's forget about heroic songs, tapregorders and such alike for a while, all right?Sit down, let's talk things over."(Kadare 1997: 5) In this example, F and E introduce italics to indicate the neologisms "maniétofonds" and "tapregorders" even though the corresponding element in A ("magnetofundet") uses none.This typographic amplification suggests that Bellos' reason for matching the French text in this instance does not result from his lack of knowledge of Albanian but from a deliberate choice.Given the incongruity of the word "maniétofonds" in French, it would be very surprising if Bellos had not compared Jusuf Vrioni's translation with Kadare's original.The possibility that Bellos did not consult the original cannot be entirely excluded of course, but it seems unlikelyespecially in light of his close collaboration with the Albanian doctor Ylli Hassani.The governor's wrong use of the modern French term "magnétophone" in A is designed to portray the protagonist's provincial character.By highlighting this neologism typographically, F and E amplify the satirical undertone of Kadare's text.The dark, critical humour of Kadare's novel, subtly conveyed between the lines in A, is exacerbated in both F and E.
In this section, I have discussed cases where MT departs from ST and where these departures are preserved in TT.These examples confirm the influence of F on E, both as a result of the process of indirect translation (the fact that E is primarily based on F) and as potentially deriving from the translator's deliberate choice (the fact that E follows F despite partial access to A).

Deviations from F to E, where F doesn't follow A
Below are examples of linguistic changes where F departs from A (A ≠ F), and E deviates from F (F ≠ E).The changes in the first part of the trajectory (A → F) are emphasised in the second part of the trajectory (E → F).They are examples of what Dollerup calls the aggravation of "mistakes" in indirect translation (Dollerup 2000: 23), and seem to confirm Bellos' description of this activity as "a recipe for disaster" (Bellos 2005: 2).This dynamic can be represented as follows: A ≠ F ≠ E.
(4) SUBSTITUTION (A) do ti kapte me presh në duar (Kadare 1990: 9) he would catch them with leeks in their hands (F) il les prendrait la main dans le sac (Kadare 1989: 9) he would catch them with their hand in the bag (E) he would catch them in flagrante (Kadare 1997: 4) This is an example of substitution from A to F, further reinforced from F to E.
Referring to the act of being caught, A uses the idiomatic expression "do ti kapte me presh në duar" ("he would catch them with leeks in their hands"), to which F substitutes "la main dans le sac" ("the hand in the bag") and E "in flagrante".F replaces the vegetable metaphor with that of the bag, but maintains the image of the hand; whereas E loses the imagery of the Albanian phrase entirely.The English expression "in flagrante", from the Latin "In flagrante delicto" ("in blazing offence"), is a legal term used to indicate that a criminal has been caught in the act of committing an offence.The meaning conveyed in E is thus in line with both F and A, but its register is slightly more elevated than its colloquial French and Albanian counterparts.In A, the idiomatic expression "me presh në duar", used as part of the governor's stream of consciousness, serves to signify the character's provincial mentality.F preserves the colloquial aspect of A, but attenuates its rural element.With the higher register of the Latin expression in flagrante, E further deflates the satirical representation of the governor's rustic character, conferring him a more educated stature.
(5) DIFFUSION (A) ajo e ndjehu veten në krahët e njërit prej tyre, pastaj te tjetrit duke vallëzuar "Tangon e xhelozisë" (Kadare 1990: 12) she felt herself in the arms of the one, then in the arms of the other dancing "the Tango of jealousy" (F) elle s'était sentie dans les bras de l'un, puis dans ceux de l'autre, dansant le tango Jalousie ( Kadare 1989: 12) she felt herself in the arms of the one, then in the arms of the other dancing the tango Jealousy (E) she imagined herself in the arms of the one, then in the arms of the other, dancing the tango to the tune called "Jealousy" (Kadare 1997: 7) This scene depicts the governor's wife, Daisy, fantasising about the two foreigners, whom she hasn't met yet.She imagines herself dancing with them to a song entitled "Jealousy", as she is passed from one man's arms to the other.In A, the formulation "duke vallëzuar 'Tangon e xhelozisë'" ("as she danced the 'Tango of jealousy'") suggests that she is both dancing to a tune called "the Tango of jealousy" and that she is performing a dance provoking jealousy.This double meaning disappears in F, where "jalousie" is no longer a genitive modifying the noun "tango" as it is in A, but an italicised substantive apposed to "tango", thus unequivocally referring to the title of a song.The ambiguity of A is furthermore depleted in E with the addition of "to the tune called", which diffuses the MT noun phrase "le tango Jalousie".This shift is interesting because TT expands MT without adding an extra layer of meaning.However, the wider significance of ST is altered, since in E the character simply dances to a tune suggestively called "jealousy" whereas in A she actually performs this tuneshe enacts the dance both literally and metaphorically.
(E) The Minister smiled to himself … "You nincompoop!" he said contemptuously.(Kadare 1997: 4) In the last sentence of this excerpt, the reader is given to hear the Minister's disdainful opinion about the governor.A's formulation -"Idiot, tha me vete" ("Idiot, he said to himself") -suggests that the Minister's criticism is interiorised.
Even though the wording does not exclude the possibility that the utterance may have been said out loud, the reflexive construction "tha me vete" ("said to himself") suggests that it is more likely to be an inner, silent thought.F's use of the verb "songer" ("to think") reduces the range of interpretations of "tha me vete" as "thought" or "said to himself".In F, it is clear that the Minister's reaction takes place in silence, since it is assuredly neither muttered nor said.In E, on the other hand, thought ("songea-t-il") is not only acted out ("he said contemptuously"), but it is further amplified by the addition of "contemptuously".Despite F's univocal interpretation of Kadare's text, E operates a significant shift from both F and A. Whilst in A and F the criticism centres on the governor, the exaggeration of the Minister's scornful reaction in E tends to vilify the latter rather than the formeran effect which is reinforced by the outdated register and frivolous nature of the word "nincompoop" compared to "idiot" and "crétin".
In this section, I have analysed instances of MT departing from ST which are further accentuated in TT.These shifts illustrate the intensification of alterations which indirect translation (as a double process of transformation) is expected to activate.They tend to confirm the claim that translating via a mediating text will further remove the final TT from the original ST (Ringmar 2012: 142).

Deviations from F to E, even though F follows A
Below are examples of variations where E deviates from F (F ≠ E), even though F equates A (A = F).In these instances, E differs from A despite the clear clues given by F. These cases challenge the assumption that linguistic shifts in indirect translation are the result of the obscuring role of ML (Dollerup 2000: 23).This dynamic can be represented as follows: A = F ≠ E.

She shifted again (…). Why do I have to let myself have such thoughts?
(E) She shifted once more (…).Why did she have to let herself have such thoughts?(Kadare 1997: 10) In the second sentence, the governor's wife, Daisy, tries to push her negative thoughts away as she lies in the bath.The personal pronoun "je" in F clarifies the first person inflection of A's conjugated verb "çoj" ("I take").As a null-subject language, Albanian can omit subject pronouns: the grammatical person is thus implied and reflected by the inflection of the verb.On the other hand, English and French require an explicit subjecthence the addition of the pronoun "je" in F. Despite F's clear indication, E substitutes a third-person pronoun ("she").The use of the first person in the Albanian original is important because it signals a skilful narrative shift.In the previous sentence, the scene is recounted from a third-person perspective: "Ajo lëvizi prapë" ("She moved again").In contrast, the switch to the first person in the following sentence marks a change in point of view, giving access to Daisy's inner thoughts through free indirect discourse (i.e. from her own point of view).This effect is tempered in E, which uses the third person in both sentences, first to describe the character's actions and then to report her inner thoughts in the form of free indirect discourse.The entry into the protagonist's stream of consciousness thus becomes less direct in E, despite F's close adherence to A.
(8) REDUCTION (A) dhe posht emrit fjalët "Chirurg-gynecolog" (Kadare 1990: 16) and beneath the name the words "Surgeon-gynaecologist" (F) et, au-dessous, les mots "Chirurgien-gynécologue" (Kadare 1989: 15) and beneath, the words "Surgeon-gynaecologist" (E) and beneath it the word gynaecologist (Kadare 1997: 9) This is a case of reduction from A to E. Despite F following A to the letter, E omits the first part of the French compound noun "Chirurgien-gynécologue" ("surgeon-gynaecologist"), a word-for-word rendering of the Albanian "Chirurggynecolog".The title "surgeon-gynaecologist" does exist in English.However, it is arguably less used than its Albanian and French counterparts in their respective contexts.Apparently insignificant, this omission is noteworthy because the redundant presence of the term "surgeon" in A and F enforces the social status of the doctor, and serves as a hint to the couple's fertility issues described in the following paragraph.
(9) REDUCTION & AMPLIFICATION (A) si të çliruara nga një ankth (Kadare 1990: 16-17) as if freed from an anguish (F) comme désangoissées (Kadare 1989: 16) as if tranquilised (E) all anguish gone (Kadare 1997: 10) Here, E loses the comparative adverb "si" used in A's "si të çliruara nga një ankth" ("as if freed from an anguish"), despite the presence of the French conjunction "comme si" ("as if").Moreover, the phrase "të çliruara nga një ankth" ("freed from an anguish"), condensed in F ("désangoissées"), is amplified by E's addition of "all": A's "an anguish" thus becomes "all anguish" in E. This double shift of omission and amplification has an important altering effect on the text: "anguish" is presented as a reality in E (the fact that it disappears implies that it was genuinely felt in the first place), whereas in both A and F anguish belongs to a realm of possibilities, which the character is not entirely aware of, for it is as though an anguish had dissipated.
The variations discussed in this section show that deviations between TT (E) and ST (A) occur in this case of indirect translation not only as a result of the misleading composition of MT (F), but despite the latter's close adherence to ST (A).

No major shifts between A and E, despite F deviating from A
Below are examples of minor changes between A and E (A = E), in spite of possible alterations in F (A ≠ F).In such instances, E deviates from F (F ≠ E), but accurately renders the original A (A = E).These are surprising cases, which go against the expectation that MT has necessarily a negative, if any altering, effect on TT's relationship to ST in indirect translation.This dynamic can be represented as follows: A = E despite F ≠ A.
(10) DIFFUSION (A) tymtarët e pullazeve matanë rrugës dukeshin si të përthyer (Kadare 1990: 11) the chimney stacks across the street looked crooked (F) de l'autre côté de la rue, les cheminées des toits paraissaient dessiner des lignes brisées (Kadare 1989: 11) on the other side of the street, the chimney stacks looked as though drawing broken lines (E) the chimney stacks on the other side of the street looked crooked (Kadare 1997: 6) The treatment of "dukeshin si të përthyer" ("looked as though crooked") in F and E reveals a surprising case of E equating A ("looked crooked"), despite F diffusing A: "paraissaient dessiner des lignes brisées" ("looked as though drawing broken lines").Moreover, E maintains the position of A's "matanë rrugës" ("across the street"), even though F moves the corresponding clause ("de l'autre côté de la rue") to the front.Here, the composition of E is closer to A than to F, showing TT matching ST rather than the MT on which it is supposedly based.
(11) DIVERGENCE (A) fluturimin e fantazisë (Kadare 1990: 15) the flight of fantasy (F) le vol de son imagination (Kadare 1989: 14) the flight of her imagination (E) the flight of her fancies (Kadare 1997: 8) The Oxford Albanian-English Dictionary (Newmark 1999) provides the two following definitions for the Albanian word "fantazi": "1.creative imagination, creative ability; 2. Fantasy, pure fantasy."The occurrence of the word in the accusative case "fantazisë" in A can thus refer either to the faculty of the imagination or to a fantasy.Here, the metaphor "fluturimi e fantazisë" ("the flight of her imagination/fantasy") describes the imaginative impulse of the governor's wife when thinking about the two foreignersthoughts which are undeniably tainted with sexual undertones.The double understanding of "fantazi" as both imagination and fantasy is thus crucial in this context.Interestingly, E maintains this ambiguity, despite F's diverging choice of "imagination", which reduces the range of interpretations.In E, the use of the term "fancy" in its plural form "fancies" emulates A's subtle innuendo, since "fancies" both designates "the faculty of the imagination", and echoes the word "fantasy", from which it derives as a contraction of its late Middle English form.Thus, the imagery of A's metaphor finds a perfect match in E, which makes astute use of the existing English saying "being prone to flights of fancy", despite the slightly restrictive, if not misleading, choice operated in F. E's closer proximity to A than to F in this instance is all the more remarkable considering that generally, even though Albanian shows no close affinity to any other Indo-European language, its grammar and some of its formal features are reminiscent of Romance languages (like Romanian and French).
(12) REDUCTION & SUBSTITUTION (A) çishte nevoja ta sëmurte atë (Kadare 1990: 17) what was the need of making it ill (F) pourquoi en venait-elle à l'altérer (Kadare 1989: 16) why did she end up altering it (E) she didn't need to make herself ill (Kadare 1997: 10) This clause describes the conflicting thoughts of the governor's wife, Daisy, whose excitement upon hearing the news of the two foreigners' visit to town is shadowed by dark thoughts about infertility.Here, Kadare presents a distinct use of the verb "sëmurte" ("make ill"), usually appearing in the reflexive form "sëmur veten" ("make oneself ill"), but employed in this particular instance with the pronoun "atë" ("it"), which refers to Daisy's prior excitement.In this context, A's "çishte nevoja ta sëmurte atë" ("what was the need to make it ill") conveys the idea "why did she have to spoil her excitement".However, the selection of the verb "sëmurte" carries more information, since it reflects the character's medical preoccupations.F's choice of "altérer" ("to alter") reduces the meaning of A's "sëmurte", and further dilutes the idea of irrepressible urge contained in A's "nevoja" ("need") by substituting "pourquoi en venait-elle" ("why did she end up").On the other hand, although E loses the interrogative construction present in both A and F, it surprisingly maintains the imagery of "sëmurte" ("make ill"), as well as the sense of urgency conveyed by the word "nevoja" ("need").Considering F's obscuring role, E's flagrant resemblance with A is compelling, and suggests that this sentence might be an example of a revision made by Bellos after consulting Ylli Hassani, the Albanian doctor who helped him to cross-check Vrioni's version against the original Abanian.Interestingly, however, despite its close adherence to A's idea of illness, E replaces Kadare's transitive use of the verb "sëmur" with a reflexive form, which conveys a much stronger meaning by giving the impression that the character is torturing herself, when in fact she is just spoiling her own joy.Bellos' possible knowledge of F's distance from A suggests that in this occurrence E's compliance to and departure from Kadare's text are deliberate rather than coincidental or driven by practical reasons.
In this section, I have discussed instances in which TT is closer to ST than to MT, even when MT deviates significantly from ST.After highlighting the incongruity and unexpectedness of these cases, I argued that they could signal exceptional linguistic similarities between SL and TL, as well as occasional access to the source text and deliberate choices on the part of the translator.

All directions: Mixed strategies
Below are examples of changes between A, F and E, where alterations in the three directions involved follow different patterns or reveal mixed translation strategies.These cases undermine the hypothesis that linguistic shifts in indirect translation follow a singular or consistent pattern.Due to the complexity and specificity of each case, no diagrammatic representation is proposed for such configuration.
(13) SUBSTITUTION, REDUCTION & AMPLIFICATION (A) Ndonëse i specializuar për përgjim me vesh, vazhdonte më poshtë spiuni, ai qe përpjekur ta kryente sa më saktë këtë detyrë (Kadare 1990: 19) Although he was a specialist of aural spying, the spy continued below, he had tried to accomplish his duty as meticulously as possible (F) Bien que sa spécialité fût le guet auditif, poursuivait-il, il s'était efforcé de s'acquitter avec le maximum de scrupules de sa mission (Kadare 1989: 19) Although his specialty was aural spying, he continued, he had tried to accomplish his mission as scrupulously as possible (E) Although his particular branch was aural, he continued in his longwinded way, he had tried to accomplish his mission as scrupulously as possible (Kadare 1997: 13) In this example, E adopts contradictory strategies in relation to A and F. It follows F in replacing A's subject "spiuni" ("the spy") by the personal pronoun "il" ("he"), but it deviates from F's omission of "më poshtë" ("below").In fact, not only does E the Albanian phrase "më poshtë", but it also amplifies it with the expression "in his long-winded way".In this scene, wherein the governor is reading the informer's report of the two foreigners' activities, E's addition of "in his long-winded way" expresses the governor's contempt for the spy's writing style.However, such judgment is not present in A, which merely uses the neutral expression "më poshtë" ("below").Based solely on this example, it would be difficult for any reader of all three languages to know which text served as a source for the English translation, since E follows neither entirely F nor A.
(F) Ne m'en parlez pas, pour l'amour du Ciel, j'en ai presque les reins brisés.(Kadare 1989: 22) Don't talk to me about them, for the love of Heaven, my kidneys are almost broken.
(E) Don't talk to me about them suitcases, for God's sake, they nearly broke my back!(Kadare 1997: 15) The treatment of this sentence in F and E, which reports the direct speech of a baggage porter, reveals a complex mix of translation shifts.Firstly, both F and E diffuse the colloquial Albanian expression "lëre" ("leave it").F chooses to render the Albanian phrase by the idiomatic French expression "Ne m'en parlez pas" ("Don't talk to me about them"), wherein the polite address "vous" (instead of the expected "tu") signals a more elevated register than A's dialectal "lëre".E further diffuses F and A with the addition of the word "suitcases".Combined with the retention of the pronoun "them" to which it is juxtaposed, the occurrence of the word "suitcases" in E recreates the colloquialism of the Albanian text.Through the addition of a single word, E thus rectifies F's change of register, whilst equating almost to the letter its idiosyncratic formulation "Ne m'en parlez pas" ("Don't talk to me about them").Secondly, F and E both retain the imagery of A's "pash zotin" ("have God") with their respective translations "pour l'amour du Ciel" ("for the love of Heaven") and "for God's sake".Lastly, E shows closer adherence to A than to F in the third clause.F departs from A's "desh më thyen kërbishtat" ("they [the suitcases] almost broke the bones of my lower back") in three ways.Firstly, in F the grammatical subject is no longer the suitcases, implied by the third-person plural inflection of the verb "më thyen" (" [they] broke") in A, but the personal pronoun "je" ("I").Secondly, as a result of the change of subject, the active form of the verb "më thyen" ("[they] broke") in A is replaced by a passive construction in F: "j'en ai presque les reins brisés" ("my kidneys are almost broken").Thirdly, F substitutes the standard term "reins" ("kidneys") to A's colloquial "kërbishtat" ("the bones of the lower back").Despite the fact that E is supposedly based on F, its composition in this third clause equates none of the changes arising in F. E does deviate from A, but in a different way from F. Although more accurate than its French counterpart "reins", E's "my back" remains less colloquial than A's "kërbishtat", used here as part of a typical Albanian expression.Moreover, like F's "presque" ("almost"), E's choice of "nearly" dilutes the idiosyncrasy of Kadare's text by ignoring the archaic aspect of the word "desh" ("almost"), whose modern Albanian equivalents would be "gati" or "përpak".In simultaneously following and departing from both F and A, E presents a strikingly convoluted association of translation strategies, which seems to lack any form of pattern or unity.
(E) Dull Baxhaja, otherwise known as "The Eaves", was the best informer on the books at N- (Kadare 1997: 12) In this example, E adopts three different strategies.First, E chooses to follow A in the retention of the proper name Dullë Baxhaja.In doing so, E's "Dull Baxhaja" departs deliberately from F, which on the other hand chooses to translate the informer's name as Dul Lasoupente.The character's last name Baxhaja in Albanian is the definite form of the substantive "baxha", meaning "the skylight".Possible counterparts of "baxha" in French are "lucarne" ("skylight"), "mansarde" ("attic") or "soupente" ("loft").F's "Dul Lasoupente" is thus a literal rendering of A's "Dullë Baxhaja".Jusuf Vrioni's strategy is not arbitrary.Kadare's choice of "Baxhaja" as a last name for this informer relates to a particular episode in the novel during which the character eavesdrops on the two foreigners from the loft above their room.F's translation maintains this intratextual reference, whereas E retains the Albanian name (suggesting that Bellos had direct access to and deliberately followed Kadare's original), whilst adding another layer of meaning with the Anglicisation of the first name "Dull", thus signifying a quality of boredom.However, E does not ignore the name's reference to the eavesdropping episode altogether; rather, it chooses to diffuse it by adding the nominal clause "otherwise known as 'The Eaves'", a sobriquet which aptly evokes both the action of eavesdropping and its location.The double-entendre of the character's name, condensed in both A and F, is thus diffused in E. In the second phrase on the other hand, E follows F very closely.F amplifies A's "më i miri ne N." ("the best in N.") to "le plus coté auprès des autorités de N…" ("the most valued by the authorities of N").In adherence with F, E adds "on the books", thereby also infusing a sense of public reputation which is not present in A's factual predicament "më i miri" ("the best").Thirdly, contrary to both A and F, E inserts a verb ("was") and displaces the substantive "informer" after that verb.In these two clauses, we thus see E alternately following A, adhering to F and departing from both A and F.
The examples discussed in this section reveal the complexity of E's relation to its two prior versions, F and A. These occurrences of conflicting translation strategies within a single clause or sentence highlight the lack of consistency underlying the shifts at play in this case study and cast doubt on the suggestion that the modifications arising from a double process of translation follow a unique pattern.

Conclusion
This contrastive analysis of the linguistic composition of the Albanian, French and English versions of Ismail Kadare's Dosja H reveals a complex mixture of translation patterns, and lack thereof.Firstly, cases wherein F's structure weakens E's adherence to A confirm Dollerup's thesis that so-called "mistakes" in the MT will be repeated in the TT.Secondly, instances of E accentuating the changes operated by F verify the claim that in a double process of transformation like indirect translation alterations activated in the MT may become exacerbated in the TT.Thirdly, instances of modifications from A to E despite F's close coherence to A indicate that linguistic shifts may occur in indirect translation not only as a result of the confusing composition of MT, but despite the latter's clarifying role.Fourthly, occurrences of E matching A despite F's obscure alterations illustrate exceptional similarities between ST and TT, most likely triggered by the translator's occasional access to the original text.
Lastly, examples of inconsistent changes in the three language directions observed highlight E's compound relation to its two prior versions, thereby undermining the assumption that linguistic shifts in indirect translation follow a singular or consistent pattern.
In response to the initial research questions, we can say that the English version of Kadare's Dosja H alternately matches, alters and adjusts Jusuf Vrioni's French version, both deliberately and for practical reasons.In this particular case of indirect translation, the linguistic shifts emerging between ST and TT are not only prompted by the doubling of the translating process or by the relative distance between the linguistic systems involved, they also stem from the translator's own choices and complex relationship to the mediating source texts.Bellos' English translation reproduces F's changes despite his occasional access to A, and it departs from A even when provided with clear clues by F.Moreover, on the one hand, E seems to claim its indirectness by highlighting the influence of the ML (F) within the text and in author interviews.On the other hand, it appears to translate directly from ST (rather than from MT) on several occasions, as though seeking to escape its own status as a second-hand translation.These conflicting attitudes and absence of coherent patterns, which become all the more apparent when condensed within a single sentence or clause, highlight the difficulty of theorising the linguistic shifts at play in indirect translation.
Overall, several assumptions surrounding the phenomenon of indirect translation are called into questionstarting with the hypothesis that the presence of a mediating language has a particular, if any, altering effect on TT's relationship to ST.In addition, by demonstrating that the modifications emerging between ST and TT may not only be due to the relative distance between the languages involved but also to the translator's own choices, the examples discussed here invite us to think of the linguistic shifts arising in indirect translation not so much in terms of "mistakes" or "errors", but rather as possible effects or attitudes towards its prior versions.Bellos' treatment of Kadare's Dosja H draws attention to a paradoxical, yet fascinating, possibility of indirect translation: the lesser level of "indirectness" when both ST and MT are invoked as source texts.For, as Jaroslav Spirk explains in Censorship, Indirect Translations and Non-Translation, the predicament that an indirect translation "becomes less indirect if the translator subsequently checks with the original poses a theoretical dilemma", which accentuates the fact that "(in)directness is not a binary opposition, but rather a continuum, raising questions of the degree of (in)directness" (Spirk 2014: 137-8;italics in Spirk).

Table 1 :
Terms used to refer to indirect translation (adapted from Ringmar 2007: 2)

Table 2 :
Kittel & Frank's model of indirect translation

Table 3 :
Taxonomy of translation shifts adapted from Malone (author's examples)