Racism and dehumanisation in Heart of Darkness and its Italian translations: A reader response analysis

This article presents the results of a reader response study of Joseph Conrad’s Heart of Darkness and two of its Italian translations. Specifically, data from an online questionnaire are used to test whether English and Italian readers respond differently to the potential racist implications of the fictional representation of the African natives. Whereas one translator removes completely all occurrences of nigger(s) and negro, the other adds additional uses of the slurs which are not present in the original. We explore with empirical methods whether these translational alterations have an effect on the readers’ perception of dehumanisation, discrimination and racism in the text, comparing responses to each translation with responses to the original. Our findings not only show evidence of significant differences in the responses between one translation and the original but also suggest that other linguistic and extra-linguistic factors could be influencing readers’ response. With this article, we aim to contribute to the under-researched application of reader response approaches to translation studies.


Introduction
It is widely accepted among translation theorists that translating, far from being an act of neutral reproduction, inevitably alters the original and results in differences between the source text (ST) and the target text (TT). As Hermans (2014Hermans ( [1985: 11) famously puts it, 'all translation implies a degree of manipulation of the source text for a certain purpose'. According to Lefevere (1992), the purpose is that of making the translation function in a given socio-cultural context in a certain way. Through 'rewriting' (Lefevere 1992) the ST, the translator can shape the TT in order to foster or hinder a given interpretational and/or ideological reading. Farahzad (2003), for example, finds that Persian translators alter syntactic and lexical features of an English text about feminism in order to fit it into their own ideological frameworks. Even though Farahzad (2003: 280) recognises that manipulative shifts can be either deliberate or unintentional, he maintains that manipulation is nevertheless ideologically motivated, whereas Dukāte (2009) acknowledges that a distortion of the ST can also result from accidental, hence un-ideological, errors of the translator. Dukāte (2009: 88) explains that a series of mistakesdue to carelessness or lack of language and/or world knowledge, for instance -'can have a cumulative manipulative effect and as a result a text may seem to be factually, linguistically or ideologically manipulated'. Venuti (2013) adds a third category, the 'unconscious', to these causes of manipulation. This category is drawn from psychoanalytic theory, particularly the work of Freud and Lacan, and refers to the differences in translation that are 'symptomatic of an unconscious motivation, a repressed anxiety, an unsatisfied desire' (Venuti, 2013: 55).
Although they are connected to the unintentional category discussed by Dukāte (2009), unconscious causes differ in being a product and a reflection of the translator's psyche, rather than a result of mistakes or oversights.
Whether intentionally made or unconscious, ideological or unmotivated, alterations in translation have the potential to trigger a reading of the TT that differs in some respect from that of the ST. Many studies have focused on identifying and examining the stylistic changes that can produce these different reading effects, comparing original and translated texts. It is only through comparison, Munday (2014: 14) explains, that 'any alteration, muffling, exaggeration, blurring, or other distortion of the authorial voice' will come to the fore. Despite the shared use of comparison as analytical technique, the range of linguistic features discussed by these studies, as well as the approaches adopted, vary greatly. Yu (2017), for example, carries out a register study, examining the relationship between dialects and varieties of standard language in a Chinese version of The Adventures of Huckleberry Finn and its original.
She explains that different varieties of standard language can be used as sociolects, against with dialects are contrasted. In her study, Yu (2017) finds that the use of standard language as a sociolect has been omitted in the Chinese translation of Huckleberry Finn; this alteration affects the linguistic hierarchy between language varieties in the novel (Yu 2007: 62). Boase-Beier (2014) brings narratology to the study of literary translation, examining narrative structures in the English translation of Herta Müller's novels. Through the discussion of six extract comparisons, she shows that narrative perspectives are changed in translation, altering the sense of focalisation in the TTs and the degree to which the reader has to rely on the text to fill context gaps due to lacking culture-specific background schemata. Morini (2007) uses instead pragmatics and conversation analysis to compare dialogues in Emma and in three Italian translations. He examines whether and how the multitude of implicatures and implicit meanings that characterise Emma's dialogues have been reproduced in the Italian TTs. Corpus approaches too have been extensively employed to support the comparative study of translation (see for example Author1 & Colleague, 2017;Johnson, 2016;Čermáková, 2015;Čermáková & Fárová, 2010;Winters, 2009Winters, , 2010Bosseaux, 2004Bosseaux, , 2006, not only because corpus tools can enhance the comparison of texts, but also because they offer a wider range of analytical possibilities unachievable without the help of the computer. The studies here mentioned have confirmed the theorists' claim that textual alterations are intrinsic features of the translated text, showing with concrete evidence that the act of translating creates a new text, as opposed to copying transparently an existing one. However, despite the wealth of research that has showed the linguistic nature of translational alterations, the discussion of the effects that these alterations can have on the target reader's response to the text remains mostly hypothetical.
The empirical study of translation response and reception is an under-researched area.
Research on the reading of translations has often been either evaluative, with a focus on translation criticism (Chan, 2016: 121), or mainly theoretical, based on abstract conceptualisations of the reader (Kruger, 2012: 218). However, 'the question of how particular translation strategies affect readers' responses […] is an empirical one, requiring carefully justified and methodically executed quantitative and qualitative research designs if it is to be answered in a reliable and responsible way' (Kruger, 2012: 218). This paper adopts an empirical approach to study translation response, aiming to redress the existing gap in the field.
It carries out a reader response analysis of Joseph Conrad's Heart of Darkness and two of its Italian translations. More specifically, the study reported in this paper uses an online questionnaire to survey readers' reactions to a central interpretational element of Conrad's short novel; that is, the representation of the African natives. By establishing a twofold comparison -ST vs. TTs and TT vs. TTthis study adds importantly to the limited existing discussion on reader response in translation studies, testing with data-driven methods whether translation alterations produce different readings of the translated text, compared to the original.

Heart of Darkness, racism, and reader response analysis
For the last forty years, Conrad scholars have been discussing the potential racist and dehumanising implications of Heart of Darkness. The debate originates from a 1975 lecture (and its 1988 published version) by Chinua Achebe, in which he accuses Conrad of being a 'thoroughgoing racist' (Achebe, 1990: 11) and Heart of Darkness of being a text 'which celebrates [the] dehumanization [of the African natives], which depersonalizes a portion of the human race' (Achebe, 1990: 12). Achebe (1990) pays particular attention to the way the African natives are represented in the text, as the passages in which they appear are, according to the critic, the 'most revealing' (Achebe, 1990: 5) of Conrad's racist attitude. He argues that the representation of the Africans only focuses on their physical appearance, on their 'blackness', without any mention of their thoughts or feelings. Africans are also deprived of human expression (Achebe, 1990: 8), lacking the ability to communicate even in their own language: '[i]n place of speech they [make] "a violent babble of uncouth sounds."' (Achebe, 1990: 8). As such, they do not play an active role in the fictional world, but are reduced to props of the setting. Ultimately, the natives are dehumanised: they are eliminated as human factor and 'devoid of all recognisable humanity' (Achebe, 1990: 12). Many literary critics have responded to Achebe's (1990) claims (see for example Zins, 1982;Watts, 1983Watts, , 1990Hawkins, 2006;Miller, 2006), and the resulting discussion is still very lively today. Surveying the various arguments in favour of and against Achebe's (1990) claims is beyond the scope of the current paper, but it is important to highlight here that the debate has shaped and keeps shaping the way Conrad and Heart of Darkness are read and interpreted (Allington, 2006). This debate has also potential implications for the translation of Heart of Darkness.
Translation is a phenomenon strictly tied to the relations between and within cultural systems (Even-Zohar, 1990). Translation is influenced by, and can influence, the sociocultural environment in which it takes place, inescapably interacting with or simply reflecting the target context of production (Lefevere, 2014). As Allington (2006: 133) explains, studying Heart of Darkness after Achebe's reading means 'to take a stand on a matter of controversy that, in the Sixties, simply did not existnamely, whether or not it is a racist book'; translators, as informed readers whose interpretative choices can shape the TT, cannot avoid taking a stance.
Whether the translator shares, disagrees, or even simply ignores Achebe's (1990) reading, intentionally or not, their decisions can nevertheless have repercussions on the TT and its reception. With their linguistic choices, translators can overstress or diminish either of the positions in relation to Achebe's (1990) critique, emphasising one reading over the other. At the lexical level, for example, the choice of which word to use or to avoid in the translation of ST items related to the representation of the Africans can influence the perception of the racist implications of the text. This is shown by Kujawska-Lis (2008), who compares two Polish translations of Heart of Darkness, one written before Achebe's (1990) lecture and the other after it. She observes that the lexical choices of the two translators to render terms like nigger, negro, savage, brute, etc., differ in the extent to which they emphasise or tone down the racist implications of the original: the early translator replaces potentially racist terms with neutral ones, while the later translator uses even more derogatory items than the original. Depending on which TT the target reader reads, Marlowthe main narrator of Heart of Darknessmay appear either 'as less verbally aggressive and mentally superior to the Africans' or as more 'biased both linguistically and intellectually', compared to Marlow in the ST.  arrives at a similar conclusion, using a corpus stylistic approach to compare four Italian translations of Heart of Darkness. He finds that the three translations produced in the same period (1989)(1990), the fourth being published in the 1920s) but published by different publishing houses differ in terms of the choices made to translate nigger(s) and negro. Whereas one translation erases all occurrences of both terms, replacing them with more neutral words (e.g. nero, the Italian equivalent of 'black'), the other two not only maintain all the uses of nigger(s) and negro, but also introduce additional occurrences not present in the original. These differences seem to indicate that the translators have responded, in one way or another, to the debate about racism in Heart of Darkness, as far as the depiction of the natives in the text is concerned. Both Kujawska-Lis (2008) and  suggest that the textual alterations they identified have the potential to alter the interpretation of this aspect of Heart of Darkness.
The reader can respond differently to the representation of the Africans in the text depending on which translation they read, affecting their perception of the text (and the author) as racist or otherwise. However, the potential of these alterations to manipulate the reader's response to the text, compared to the original, remains untested.
The current study builds on the findings of . It uses responses elicited through a questionnaire to test whether differences arise in the way English and Italian readers react to potential dehumanising, discriminating, and racist implications in Heart of Darkness and in two Italian translations. As such, this study sits within the remit of reader response analysis, whose application in the field of stylistics has gained popularity in the last decade. As Whiteley and Canning (2017) explain, this approach is not new, as it draws on existing and well-established methodologies and paradigms (see Whiteley and Canning, 2017 or Harding, 2014 for an overview); however, '[t]he impulse to collect extra-textual data about literary reading in order to inform, develop and reflect upon stylistic analysis is becoming increasingly widespread' (Whiteley and Canning, 2017: 72). The collection of data through interviews, surveys, focus groups, measurements of reading and reaction times, etc. provides the stylistician with an observational basis with which to test 'whether assumptions and frameworks of stylistic analysis are supported by evidence from real readers' (Peplow and Carter, 2014: 440). A wide range of empirical approaches and methodologies have been used to enrich and expand stylistic investigations, supporting the exploration of difficulty in poetry (Castiglione, 2017), shifts in narrative point of view (Sotirova, 2006;Cui, 2017), body language multi-word clusters in Dickens's novels (Mahlberg et al., 2014), and foregrounding (Zyngier et al., 2007), to name just a few. Yet, as mentioned in the previous section, the application of these methods to the study of reader response in translation is not as widespread.
Although the role of the reader has always been taken into account in translation studies, this role has often been held by a theorised reader, who represents an aspect of the target culture to which the translation is addressed, or as the translator-as-reader (Chan, 2016: 123), whose reading can shape the TT. Despite the recognition of the importance of the actual reader in translation research (Assis Rosa, 2006: 103), studies that concentrate on real readers and their responses are very few. For example, Li (2012) does consider real readers in his discussion of the relationship between ideology and translation, but mainly in terms of interpretative reading communities (cf. Fish, 1980), tracing the reception history of some ideologically-loaded in Afrikaans translated children's books, does not take into account the STs, but only compares responses to the TTs. One of the few studies that compares readers' responses to ST and TT is Chesnokova et al. (2017). They examine readers' reactions to the original and to the Brazilian, Russian, and Ukrainian translations of Poe's poem "The Lake" using a five-point semantic differential scale, where participants were asked to indicate their thoughts on the poem. The data show significant differences between the groups of readers, confirming that responses to the original differ from that of the translations, and between translations. However, Chesnokova et al. (2017) have no native speakers of English among their participants, so the reaction to the English text is based on the reading of speakers of English as a second language.
Although this does not affect their analysis (as they are interested in cross-cultural responses The current study not only compares reactions to different TTs that differ in terms of translation strategies used by the translators, but it also directly contrasts responses to the original with responses to each translation. In this way, it tests two hypotheses: (i) that responses to the ST can differ from responses to the TT as a consequence of the alterations that the text inevitably goes through during translation; and (ii) that responses to different TTs of the same ST can vary as a consequence of the diverging translation strategies adopted by and/or alterations introduced by the translators. The following section will detail the specific hypotheses that this study aims to test and the methods and procedure used. Section 4 will present the results of the study, which will be discussed in Section 5. Finally, Section 6 will provide some concluding remarks.

Methods
This article compares responses to the representation of the African natives in Heart of Darkness and two of its Italian translations. The aim is testing whether differences arise between the texts in the extent to which English and Italian readers perceive the descriptions of the Africans as being discriminating, dehumanising, and racist. The two TTs used in this study were produced in the same year, 1990. The first is the Mondadori edition, translated by Rossella Bernascone; the second is the Garzanti edition, translated by Luisa Saraval. These texts differ diametrically in the way they render the most openly racist terms in the original, nigger(s) and negro. As Author1 (2017) Achebe (1990) says about the way the Africans are described in Heart of Darkness, representing the linguistic instantiations of the dehumanising tendencies referred to by the critic (for a detailed description and analysis of the patterns, in the original and in the translations, see . The semantic preferences/prosody are present in the translations too; it seems that the translators only altered the most obvious reflections of the racist discourse (the words used to refer to the Africans), but maintained unaltered the linguistic patterns that create dehumanisation in the text (Author1, 2017: 172).
Based on these divergences between the three texts, it was hypothesised that there would be a difference between the ST and the TTs in terms of perceived dehumanisation, discrimination, and racism. Bernascone's translation would be perceived as more racist, dehumanising, and discriminating than the original, because it uses racial slurs more frequently than the ST; Saraval's translation would be perceived as less racist, dehumanising, and discriminating than the original, because it uses no racial slurs. We also tested whether there is a difference in the perception of dehumanisation, discrimination, and racism between the two translations, as a result of the opposing strategies used by the translators. Overall, the following hypotheses were tested: -There is a difference in the perception of dehumanisation, discrimination, and racism between the ST and the TTs: 1. Bernascone's translation is perceived as (1.a) more dehumanising, (1.b) more discriminating, and (1.c) more racist than the ST; 2. Saraval's translation is perceived as (2.a) less dehumanising, (2.b) less discriminating, and (2.c) less racist than the ST; -There is a difference in the perception of dehumanisation, discrimination, and racism between the translations: 3. Saraval's translation is perceived as (3.a) less dehumanising, (3.b) less discriminating, and (3.c) less racist than Bernascone's translation.
To collect data on readers' responses we used an online questionnaire in three versions, one for each text. The English version was addressed to native speakers of English to gather reactions to the ST. The two Italian versions were addressed to native speakers of Italian to collect responses on the TTs. 1 The Italian questionnaires were answered by different pools of participants; that is, Italian participants were either presented with Bernascone's translation or Saraval's. The questionnaires asked participants to read 17 extracts from the ST or one of the TTs and answer three sets of questions about these passages. The extracts were the same in all versions, but obviously differed in terms of language and text they came from. To put together the 17 extracts about the natives, we first selected all the passages in which nigger, niggers, and negro occur in the original, and their Italian versions. Then, we added to these the ones in which Bernascone introduced additional occurrences of the Italian equivalent negro ('nigger'), and their version in Saraval's translation and in the English original. The passages (usually one or two paragraphs long, between 336 and 37 words in the ST) were presented in the order they appear in the text and occasionally short notes were added between extracts to help the reader understand how one passage linked to the next (for example, 'Marlow reaches another station where he waits for months for his steamer to be repaired'). 2 Three sets of questions measured readers' perception of dehumanisation, discrimination, and racism in the passages. The first set comprised of six questions to which participants were asked to answer indicating their agreement on a 7-point Likert scale (see : 2). The first three questions measure perception of agency (capacity of self-control, capacity to act morally, capacity to plan), while the last three questions measure perception of experience (capacity to experience emotion, capacity to experience refined or uniquely human feelings, capacity to experience consciousness). The second set equally comprised of six 6-point Likert scale questions (see Appendix) aimed at measuring perception of discrimination. The items were extracted and adapted from the Privilege and Oppression Inventory (Hays et al., 2007), a model 'designed to measure an individual's level of awareness of social issues (i.e., privilege and oppression as it relates to four primary cultural dimensions)' (Hays et al., 2007: 68). Of the four dimensions (racial, gender, sexual orientation, religious identities), we used the one related to racial issues, selecting and adapting six items from the original 13-item inventory. Finally, to measure readers' perception of racism, we used a 6-point Likert scale question asking participants to indicate to what extent they agreed or disagreed with the statement 'The representation of the African natives in these passages is racist'. Participants were also given the option to add any comment on the statement above. We could not find an existing racism measure or inventory that would work in this study, as the vast majority of established scales (cf. Kressin et al., 2008;Gamst et al., 2011;Atkins, 2014) measure racism and racial prejudice in the participants themselves, or their ability to be aware of it, as opposed to their perception of racism in external material (such as our extracts). The questionnaire ended with questions on the age, gender, and native language of the participants, as well as on whether they had read and/or studied Heart of Darkness before. Although we recognise that exploring the effect that these variables can have on the perception of racism, dehumnisation, and discrimination is an endeavour worth pursuing, we could not do it here, as we did not have enough data of different types of participant groups to investigate whether these variables might have influenced our findings.
Dividing the participants of the three questionnaires into smaller sub-groups (e.g. male readers of the Bernascone version) would have fragmented further our pool of data. We therefore focused on the three largest data pools (readers of the Bernascone version, readers of the Saraval version, and readers of the original), in line with our research aims.
We collected 65 responses, 23 for the English version, 21 for the Bernascone version, and 21 for the Saraval version. English participants were between 18-and 61-years old, mostly women (19); 15 of them had already read Heart of Darkness and 9 had studied it too. Italian participants working with the Bernascone version were between 26 and 35 (11 female and 10 male participants). Most of them had never read (15) or studied (16) Heart of Darkness before.
Italian participants completing the Saraval version were between 29 and 38 (13 male and 8 female participants). In this case too, most of them had never read (16) or studied (19) Heart of Darkness before. Table 1 shows the ratings for the questions relating to dehumanisation, discrimination and racism, which will be examined in turn below. When looking at the ratings across the different question types, it is important to note that the scale for the dehumanisation questions is different to that of discrimination and racism. This was done to ensure that the results are comparable to the literature that has developed, evaluated, and implemented the instruments to assess dehumanisation and discrimination. In order to analyse the results statistically, when comparing ratings across all three versions the Kruskal-Wallis H test was used, while when comparing two versions the Mann-Whitney U test was employed. The six questions related to dehumanisation asked participants to evaluate, on a scale from 0

Results
(not at all) to 6 (totally), to what extent they thought the natives in the extracts would be capable of doing a number of things based on the way they were described in the passages, not nor amongst the texts (p's > .05). Notably, a comparison of agency and experience ratings reveals a significant difference (Z = -6.13, p > .001) across all text types, indicating that readers' perception of the natives' agency is significantly less than their ability to experience emotion, feelings, and consciousness.
Turning to discrimination, six questions asked participants to judge the extracts on a scale from 1 (strongly disagree) to 6 (strongly agree): (i) the whites generally have more resources and opportunities than the African natives; (ii) the whites have the power to exclude the African natives; (iii) there are benefits to being white in the social order/context described by these passages; (iv) white cultural characteristics are more valued than those of the African

Discussion
Based on the results described above, we can now confirm or reject the hypotheses formulated in Section 3, as shown in Table 2 below.  (Kennedy, 2002: 32), the use of which has been seen as highly racially offensive since 1800 (Hughes, 2006: 327). In contrast, the use of negro in Italy has been considered equivalent to nero ('black') until the 1970s and only at the beginning of the 1990s, with the debates on political correctness, the perception of the word has started to change (Faloppa, 2004(Faloppa, , 2011. What is more, Faloppa (2011: 10) explains that the perception of negro as a racial slur in Italy is also the direct result of the interdiction of nigger in Anglo-American contexts. Overall, this suggests that negro and nigger are not equivalent in terms of their racist load, which is intricately connected to their respective context of origin. Context cannot be transferred during translation and therefore the translated word loses its original contextual meanings and gains new ones in the target context. Even though Bernascone's version has more occurrences of the slur than the ST, the use of negro in Bernascone's translation may not be perceived as racist as that of nigger in the English original, whereas the presence or absence of the slur in Italian may not result in a diverse response to the two TTs.
On the other hand, other aspects and features of the text, for example dehumanisation and an intrinsic derogatory representation of the natives, could be sufficient for Saraval's translation to be perceived as racist as Bernascone's, despite the absence of the openly racist word negro.
Before moving to the conclusion, it is also worth discussing the fact that, across all text types, perception of natives' agency is significantly less than their ability to experience. The consistency of this result across ST and TTs suggests that this is a feature of the text itself, as opposed to being specifically of the original or the translations. Readers are more likely to recognise in the Africans the 'capacity to be subjected to sensations (e.g., emotions, consciousness, or personality)' than the 'mental capacities that enable decision-making and organizing behaviours (e.g., idea, judgment, self-control, or communication)' (Morena et al., 2016: 2). It can be argued that the text dehumanises the natives insofar as it denies them the mental capacities to have active agency. Morena et al. (2016: 5) explain that low levels of agency could be associated with animalistic dehumanisation (considering humans as animals), while low levels of experience could be linked to mechanistic dehumanisation (considering humans as objects or robots); only high levels of both would indicate perceiving someone as a complete human being. In the case of Heart of Darkness and its translations, the greater perception of experience in the natives compared to their low level of agency seems to point to the animalistic metaphor of dehumanisation. The fact that all text types, despite their differences, consistently trigger the same type of dehumanisation suggests that dehumanisation is not linked to the use or avoidance of racial slurs, but is rather conveyed by other independent linguistic features (for example, the lexico-semantic patterns briefly mentioned in Section 3and discussed in Author1 (2017) which have been reproduced by both translators).

Conclusion
Assessing the effects of translation alterations on readers' response is a complex endeavour. If, on the one hand, we demonstrated that removing racial slurs from an Italian translation of Heart of Darkness is connected to lower perception of dehumanisation, discrimination, and racism compared to the original, on the other hand, we also found that the opposite strategy (i.e. adding racist terms) does not produce the opposite result. We believe that this is a reflection of the multifaceted nature of the translation phenomenon: the words the translator uses are only part of the picture and the effect they can have on the reader is not one-dimensional, but can vary depending on intricate interactions with the context of reception and other linguistic features.
Although our results indicate that translation can manipulate reactions to the text, further research is needed to explore the multidimensional relationships between linguistic alterations and reader response in translation.