Introduction

The term retranslation is typically used in translation studies to describe either the act of translating a text which has already been translated into the same target language at some earlier moment in time, or the product of such an act (Tahir Gürçağlar, 2019). Both the practice and its results have been a topic of significant interest to translation scholars for at least three decades. Overviews of this field of study often highlight a special issue of the French journal Palimpsestes (1990) as the first collection of essays to provide sustained engagement with the phenomenon. Most notably, it was in this volume that the well-known scholar of literary translation Antoine Berman set out what has since become known within the discipline as the “retranslation hypothesis” (Chesterman, 2004, p. 8). This is the suggestion that, while first translations generally attempt to be as accessible as possible for the target culture, and are consequently often rather inaccurate and clumsy (“maladroite”) in their rendering of an original, later retranslations tend to move closer to the source text, striving to better reproduce the cultural, textual and other particularities of the original (Berman, 1990, p. 3). For Berman (1990, p. 7), this process of gradual improvement over time often culminates in the production of a “grande traduction”, a definitive translation which finally succeeds in capturing the essence of the original. Further retranslations may be produced in subsequent years for new audiences, but these inevitably find themselves being compared to the “grande traduction”.

As Deane-Cox (2014, p. 5) and others have since convincingly shown, Berman’s retranslation hypothesis is not only flawed in its uncritical adoption of a paradigm of progress and perfection, but it also overlooks much of the complexity of retranslation as a cultural phenomenon (see also Susam-Sarajeva, 2003, p. 2; Brownlie, 2006, p. 148). In its myopic focus on the nature of the relation of equivalence between source and target texts, it largely ignores the social, political, ideological, and economic context in which retranslations are necessarily embedded and downplays the wide variety of factors motivating each successive translator to provide a new reading of the original. Indeed, research over the past thirty years has repeatedly shown the extent to which retranslations—like all translations— are always situated within a complex web of ideologies, norms, beliefs and expectations, and are inevitably conditioned by these extratextual forces. For example, in a case study focused on five successive translations of Emile Zola’s novel Nana into English, Brownlie (2006) explored how changes in dominant middle-class attitudes in the English-speaking world towards references to sex and sensual parts of the body in public discourse have shaped the word choices of the translators of this work over the past 140 years. Brownlie (2006, p. 158) found that many sections of the original French text considered offensive were simply omitted or replaced with vague euphemism when the work was first translated (anonymously) into English in 1884: this was largely because the translation aimed at attracting a broad readership and the publisher did not want to risk prosecution following the passing of the Obscene Publications Act in 1857 (see also Nead, 2005, p. 158). Later retranslations by Duff (1956), Holden (1972) and Parmée (1992), on the other hand, were produced in contexts in which more liberal ideologies prevailed and for this reason, Brownlie argues, they do not show such levels of linguistic prudishness and self-censorship.

The present paper aims to develop further the investigation of shifts in interpretation over time that can be observed across multiple retranslations of a single source text by exploring changes in the characterization of ‘the common people’ in two English-language versions of Thucydides’ History of the Peloponnesian War, the first produced by Samuel Bloomfield in 1829 and the second by Steven Lattimore in 1998. Specifically, I seek to demonstrate the extent to which the political and ideological climates out of which these two retranslations emerged might have influenced the narrative construction of the ordinary, non-elite members of the ancient Greek societies described by each translator. Second, this paper additionally attempts to contribute methodologically to the study of characterization in (re)translated texts by illustrating the potential of a corpus-based approach developed as part of the Genealogies of Knowledge project. Both Bloomfield’s and Lattimore’s retranslations have been included in the Genealogies of Knowledge ‘Modern English’ corpus and are consequently available for analysis via the suite of corpus analysis tools developed specifically for this project and available for download from our website (Luz, 2011; Luz and Sheehan, 2014).Footnote 1 As I aim to demonstrate through the case study below, this methodology enables the researcher to identify the accumulation of textual cues through which any given narrative actor is constructed in each of the translated texts under investigation. I argue that it thus permits the systematic creation and efficient comparison of detailed character profiles in the two translations, and the observation of linguistic and narratological patternings unique to each.

Mediating Thucydides as a political scientist in the nineteenth and twentieth centuries

Thucydides’ History of the Peloponnesian War is a detailed contemporary account of a conflict fought between the rival city-states of Athens and Sparta in 431–404 BCE. In addition to its historical interest as the only eyewitness description of this war to survive into the modern era, this work has long been read as a source of political lessons of significant potential interest to statesmen and public alike. Indeed, a persistent feature of the prefaces and introductions to many of the translations of Thucydides’ Greek text now available in English is their assertion of the clear relevance of the work to political concerns in the target culture. For example, Thomas Hobbes’ (1629/1843, p. 5) version, the first to be done “immediately from the Greek”, includes a dedication “To the Right Honourable Sir William Cavendish” in which the translator recommends his author’s writings “as having in them profitable instruction for noblemen, and such as may come to have the managing of great and weighty actions”. Nearly four hundred years later, Steven Lattimore (1998, p. 13) similarly prefaces his translation of the History by signaling its topic as one of “enduring importance” for modern culture and political thought.

This trend can be explained at least in part with reference to Thucydides’ own framing strategies in the opening chapters of his text. Specifically, as Potter (2012) has pointed out, the classical Athenian writer claims to offer not only a discussion of the political causes of the conflict between Athens and Sparta in 431–404 BCE, but also those of similar events that might be expected to happen in the future. In a much cited passage (1.22.4),Footnote 2 Thucydides suggests that his aim has been to produce a “possession for all time” (“κτῆμά ἐς αἰεὶ”) of use to anyone who might “wish to look at the plain truth about both past events and those that at some future time, in accordance with human nature, will recur in similar or comparable ways” (trans: Lattimore, 1998, p. 14). In other words, the historian wants his readers to view his narrative of the events of the war both as a historical record and as an exemplary paradigm for understanding political history in general: he wants us to ask why Athens lost, to what extent its decline was caused by democratic decision-making, and what the proper relationship should be between people and power.

Nevertheless, it must also be recognized that the relevance of Thucydides to the modern world is far from self-evident and that the existence in his text of lessons of transhistorical value cannot be regarded as a given (Harloe and Morley, 2012, p. 12). Rather, modern readers’ assumption of this ancient text’s pertinence to their own societies has always been mediated by countless other narratives circulating in the receiving context, especially those concerning the relative merits and risks of monarchy, aristocracy and democracy. It is in this light that we should read nineteenth-century translator Samuel Bloomfield’s (1829, p. 6; emphasis in original) paratextual framing of the History as an illustration of the “evils of unbalanced democracy” as a deliberately political act. Indeed, as his dedication to the then Prime Minister, Arthur Wellesley the Duke of Wellington,Footnote 3 and his copious footnotes additionally show,Footnote 4 this translator had a clear political agenda in retranslating Thucydides’ text for his contemporary readers in the early nineteenth century: Bloomfield appears to have sought to give voice to the conviction held by many members of the ruling classes during this period that the common man was not sufficiently intelligent or responsible to be entrusted with participation in the political decision-making process (Cartledge, 2016, p. 299). For Bloomfield, as for many of his contemporaries (Hanson, 1989, p. 70), democracy meant class rule: the unchecked exercise of power by the least educated, least economically independent and least politically principled citizens (see also Williams, 1976, p. 94). Consequently, the drive towards greater democracy, led in Georgian Britain by prominent Whig politicians such as Charles Grey, would—he implied—necessarily imperil the social order and threaten the prosperity of the commonweal. As I seek to explore in the analysis that follows, this aristocratic ideology appears to have strongly shaped his characterization of the ordinary non-elite members of society when translating Thucydides’ account of classical Athenian domestic politics; his prior assumptions about the traits and habits of the common people as a political actor in his own society can consequently be seen to have led him to a very particular rendering of this ancient text.

In order to demonstrate the strength of these patterns in Bloomfield’s work, the early nineteenth-century translator’s version is compared here with Steven Lattimore’s retranslation, published in the US in 1998. This alternative interpretation is in no way assumed to be an entirely neutral or literal rendering of the Greek; rather, it is chosen as a target text produced “in a world that agrees on the importance and desirability of democracy” (Hanson, 1989, p. 68). Unlike Bloomfield’s Britain, the context out of which Lattimore’s translation emerged is fundamentally pro-democratic in its discourse and rhetoric. As Hanson (1989, p. 68) has explained, it would no longer be common for the members of the now transnational English-speaking public for which this late-twentieth-century rendering has been created to equate democracy primarily with a dangerous and unstable form of politics: the “odious connotations” democracy once held have now been replaced by new and generally positive associations with popular sovereignty and political equality. Mainstream politicians in the US, Britain and other Anglophone countries frequently highlight the democratic values on which their respective societies are built and make repeated reference to their dedication to the people as the sovereign power and source of authority in the state. I therefore argue that, as with Bloomfield, this ideological context must have shaped Lattimore’s characterization of ‘the common people’ in his translation of Thucydides.

Narrative characterization in historiographical texts

My application of the concept of characterization to the study of these two retranslations is grounded in the view that it is not merely possible but also productive to make use of the tools of narratology in the analysis of (ancient) historiography (Hornblower, 1994, p. 131). As demonstrated most prominently in the work of Barthes (1982), White (1987), Hornblower (1994), and de Jong (2014), this theoretical framework allows us to highlight the narrative mode of representation adopted by historians and concentrates our attention on the rhetorical storytelling techniques by means of which an account of the past is constructed by its author (and/or reconstructed in translation). Adopting this approach and conceptualizing Thucydides’ History first and foremost as a story does not imply that I seek to equate his text with fiction; that is, I am not suggesting the events he describes did not happen at all (see also Hornblower, 1994, p. 133). Rather, the claim is that the application of narratological theory can better enable us to explore the full range of devices and techniques through which the complex and often contradictory chaos of historical reality is made intelligible by historians for their readers (de Jong, 2014, p. 172; Rigney, 2014, p. 12; Barthes 1982, p. 20). Narratology can additionally help us to identify with greater precision the broader implications of the shifts in meaning which inevitably take place when a history is renarrated by a translator in another language for a new audience.

In this theoretical context characterization is understood as the process through which the various actors depicted in a narrative are distinguished from one another, are provided with distinct traits and are hence transformed into characters (Jannidis, 2013; Bal, 2017, p. 7).Footnote 5 This is achieved, as Culpeper (2014, p. 28) has shown, through the interaction of the text with its reader’s background knowledge. Characterization, in other words, is shaped both by an accumulation of more or less explicit textual cues embedded within the language of the narrative being told and by the application of prior knowledge, brought to the text by the reader (Culpeper, 2014, p. 28).

Textual cues supplied by the narrative text itself consist of a referring expression (the label), a predicate (the attribute) and often a modalizer which suggests degrees and qualifications (Garvey, 1978, p. 73; Margolin, 2002, p. 108). The referring expression will in many cases be a proper name like John, but personal pronouns such as he and she and definite descriptions such as the President of the USA can also be used (Margolin, 2002, p. 108). These characterization cues are most explicit in cases of “direct definition” (Rimmon-Kenan, 2002, p. 62), i.e., when the narrator himself or herself overtly assigns a specific quality to an actor. This is the case, for example, at 1.138.3 of the History of the Peloponnesian War, where Thucydides as narrator writes:

For, indeed, Themistocles was a man in whom was most clearly displayed the strength of natural understanding (trans. Bloomfield, 1829, v. 1, pp. 254–255).

Here, the name Themistocles acts as the referring expression, helping the reader to distinguish this actor from other participants;Footnote 6 “was a man…natural understanding” is the predicate, informing us of his gender and effortless intelligence; and “most clearly” serves as a modalizer, increasing the strength of the assertion.

Interventions of this kind are invariably the most authoritative and decisive delineations of character, especially in works such as the History in which Thucydides mainly assumes an omniscient, objective tone as narrator (Gribble, 1998, p. 42). It is for this reason that the present paper focuses largely on this type of characterization cue. Culpeper (2014) and Rimmon-Kenan (2002) point out, however, that direct definitions are far from the only sources of attributive propositions provided by the text. In the course of a narrative, individual actors may themselves be made to highlight particular qualities that they wish to promote in relation to their own character (self-presentation) or in relation to that of other actors in the narrative (other-presentation—Culpeper, 2014, p. 167). We can observe both of these processes at play in, for example, Alcibiades’ speech at 6.17.1 of the History,Footnote 7 where this prominent aristocrat foregrounds both his own youth (“I am in the flower of my days”) and the general impression held among the Athenian population of his rival Nicias as a successful military leader:

And now do not be afraid of me because I am young, but while I am in the flower of my days and Nicias enjoys the reputation of success, use the services of us both (trans: Jowett, 1881, p. 420).

As Culpeper (2014, p. 171) notes, such self-presentations of character are no less important than the direct definitions provided by the narrator, but the influence of the internal motivations of the characterizer should nevertheless be taken into consideration. The validity of Spartan King Archidamus’ suggestion in 1.84.3 that “Our habits of discipline make us both brave and wise” (trans: Jowett 1881, p. 53) is clearly more subject to suspicion of bias than if Thucydides speaking as narrator were to make this assessment (although this should not be taken to imply that Thucydides is entirely neutral in his narratorial characterizations of the Athenians and Spartans either—Podoksik, 2005).

Finally, readers’ impressions of character can also be formed out of what are widely described as indirect presentations, that is, where an actor displays a particular trait through their actions, manner of speech and external appearance, as well as through the settings in which we encounter them in the storyworld (Thompson, 2006, p. 18). Rimmon-Kenan (2002, p. 63) shows, for example, how readers may infer features of an actor’s character from “acts of commission” (acts performed by the actor), “acts of omission” (acts which the actor ought to have undertaken, but did not) and “contemplated acts” (the actor’s unrealized plans or intentions). Indeed, as Lattimore (1998, p. 143n) notes in a footnote to his own translation of Thucydides, the fact that Alkidas does not follow Teutiaplos’ suggested plan to attack Mytilene in 3.30.1–3.33.1, and that instead he seeks simply to return home to the Peloponnesus, contributes significantly to the characterization of this actor as a “stereotypical timid and unenterprising Spartan leader”. The repetition of such behavior at other points in the narrative effectively consolidates this impression, confirming our initial suspicions through the accumulation of supporting data (see Roisman, 1987, p. 385 for a useful summary).

No matter how explicit the textual indicators of character might be, it is important for our purposes here to underline the fact that “our prior knowledge about particular kinds of people affects the way we interpret new information” (Culpeper, 2014, p. 57). Readers inevitably categorize actors they encounter in narratives according to their own individual and preconceived schemata of interpretation (Eder et al., 2010, p. 14). Eder et al. (2010, p. 14) suggest these schemata help us to “fill in” pieces of information necessary for characterization but not directly provided by the text. This is evident in the fact that, despite Thucydides not once describing Themistocles’ physical features, for example, we nevertheless can and do form a mental image of this actor as we read the text, developed largely on the basis of our stereotypical knowledge about ancient Athenians, charismatic politicians and effective military commanders in general.

Culpeper (2014, p. 75) suggests three broad groupings for the schematic categories we tend to use as part of this process of top-down characterization: personal categories, social role categories and group membership categories. This typology has clearly been developed with the characterization of individual protagonists in mind (e.g., Pericles or Themistocles in Thucydides’ narrative), but it can nevertheless be usefully adapted to inform our analysis of the characterization of groups such as ‘the common people’. What I shall term character-specific categories of prior knowledge include readers’ understandings of the preferences, interests, habits and qualities of a named group, specific to a particular narrative and/or a particular cultural and historical context (comparable to Culpeper’s ‘personal’ categories; 2014, p. 75). This would take account of, for example, our knowledge of the ancient Athenian common people in particular, as opposed to the ordinary members of human societies in general. Such background information may be derived from various sources, including not least the writings of other ancient historians we might have read (e.g., Xenophon’s Hellenica), as well as the accounts provided in more modern history books, literature, art and perhaps even film (e.g., Philipp Foltz’s 1852 painting Perikles hält die Leichenrede or Roberto Rossellini’s 1971 film Socrates). Background knowledge of this type will inevitably inform our appreciation of an actor’s actions in the course of the narrative, although further attributes will also be added to our knowledge of this actor as we progress through the text and interact with the assortment of direct and indirect cues it contains. Social role categories, by contrast, concern “knowledge about people’s social functions” (Culpeper, 2014, p. 75), such as our pre-formed conceptions regarding the proper relationship between politicians and the public, or our understanding of the role women might have played in political decision-making in the ancient world. Once again, this knowledge will to a large extent determine how we initially conceptualize particular actors, but it also remains fluid and malleable, and may therefore be confirmed or contested through our encounter with the text. Finally, what I will refer to as stereotypical categories (comparable to Culpeper’s ‘group membership’ categories; 2014, p. 75) include the associations that develop around broad social groupings such as gender, race, class, nationality and religion. As an example, we might cite the common expectation that rich aristocrats will want to protect their property rights while poorer members of a society will seek a fairer distribution of wealth. Similarly, even if a modern reader of Thucydides’ History knows little about the content of this text, they will tend to approach it with a belief that the Athenians are the heroes in the narrative, largely on the basis of the prestige attached to this civilization in Anglo-American culture since the mid-nineteenth century (Whedbee, 2003, p. 65). Interestingly, many eighteenth-century readers of ancient history, by contrast, tended to see Sparta in a more positive light, celebrating the Spartan state’s “free, virtuous and long-lived regime” (Saxonhouse, 1996, p. 14) whilst denigrating Athens as “a tyranny in the hands of the people” (Mitford, [1784–1810] 1838, v. 4, p. 10; see also Demetriou, 1996, p. 285). These categories of knowledge are arguably the most decisive in shaping a reader’s characterization of a narrative actor; given their widespread entrenchment in society at large, they are also perhaps the most resistant to change.

A corpus-based methodology for the analysis of characterization in (re)translated text

An awareness of the extent to which the reader’s own prior knowledge is involved in the process of characterization is especially important when we consider that translators act not only as mediators of the source text for the target culture, but also as readers of this original work themselves (Hatim and Mason, 1990, p. 224). This means that the cultural schemata translators bring to their reading of a source text will inevitably shape their textual choices in the target-language version. As I attempt to demonstrate in the analysis below, their individual interpretation of the characterization of each actor will necessarily be inscribed in their translation (Baker, 2014; Venuti 2014, p. 7). This is perhaps particularly significant in the case of authors such as Thucydides, whose prose is in many passages famously opaque and open to multiple possible readings.Footnote 8

Somewhat surprisingly given the recent wave of interest in narrative and narratology that has swept the field of translation studies (Jones, 2019), the extent to which characterization may be impacted by the mediation of the translator has not been the focus of significant critical attention. In one of the very few studies that have addressed this issue, Prince (2013, p. 25) shows how the impossibility in English of inflecting an adjective according to the gender of the protagonist described affects the construction of literary characters in the minds of target-language readers when a narrative has been translated from languages such as Italian or French. He quotes as an example Lanser (1999, pp. 175–176), who points out that

[t]here is a moment in the first chapter of George Sand’s La Mare au Diable [The Devil’s Pool] when the masculine form of an adjective provides the sign that the narrator is male; in the English translation no such signal would be manifest. Conversely, translators of George Eliot’s Adam Bede into French would surely have found it difficult to avoid sexing the narrator of Chapter 17, who undertakes many acts of self-characterization, in such constructions as ‘I am content,’ that English does not, but French would inflect.

Prince’s discussion is more anecdotal than systematic, and he does not offer a coherent theoretical framework or methodology with which to develop future research on this topic. To the best of my knowledge, the only study to provide a detailed examination of the effects of translation on the characterization process is McIntyre and Lugea’s (2015) analysis of the deaf and hard-of-hearing (DHOH) subtitles produced for the popular American police procedural drama The Wire. The authors of this study effectively demonstrate the utility for translation studies of a narratological model of characterization, and show that viewers reliant on the DHOH subtitles will inevitably characterize the series’ protagonists in quite different ways to those who are able to access the spoken dialogue. They argue, for example, that the omission in the subtitles of what they term non-fluency features in the character Nakeisha Lyles’ dialogue in episode 1, scene 3 (for example, repetitions and phrases such as “you know”) ultimately reduces the possibility that DHOH viewers will come to conceptualize this actor as nervous and unsure of herself, a character trait which is immediately evident for hearing viewers (2015, p. 77). Yet, McIntyre and Lugea’s (2015) focus is exclusively on the kinds of losses in character-specific information that result from the shift from an oral to a written mode of communication; they do not, in other words, explore how the translator’s own background knowledge and socially determined ideological beliefs might impact the characterization process for the readers/viewers of their target-language text. One of the key aims of the analysis I present in the next section (“(Re)constructing ‘the common people’”) is to draw greater attention to the fact that different translators working in different socio-political contexts may produce translations which attribute divergent sets of character traits and qualities to the actors presented in the narrative. The study thus hopes to move discussion of the effects of translation on narrative characterization away from a paradigm of loss, deficiency and failure and instead to promote a perspective which embraces the productive role translators often play in reconfiguring the countless narratives through which we come to know, imagine and make sense of the past, our present and potential futures (Baker, 2014, p. 159; Harding, 2012, p. 22).

As explained in the introduction, I additionally seek to contribute methodologically to the study of characterization in (re)translated texts by illustrating the potential of a corpus-based approach developed as part of the Genealogies of Knowledge project. Such an approach is valuable, I argue, because it can enhance our ability to engage in a more efficient and systematic manner with the accumulation of characterization cues spread throughout a narrative text. While McIntyre and Lugea’s (2015, p. 69) manual analysis forced them to limit the scope of their investigation to the opening three scenes of episode 1, season 1 of The Wire, the affordances of corpus analysis software mean that such close textual analysis can easily be combined with broader quantitative interrogations of far larger datasets. As a result, rather than restrict our analysis of Bloomfield’s (1829) and Lattimore’s (1998) translations of Thucydides to the comparison of limited stretches of text, we can process their works by means of corpus analysis software to generate detailed word frequency data and/or concordances that display every instance of a particular lexical item as it appears across the full length of each translator’s version.

The value of such computer-assisted methodology has been demonstrated particularly clearly in recent research in the field of narratology. Hubbard (2002), for instance, has conducted a corpus-based exploration of the ways in which the two main protagonists in Jane Austen’s novel Sense and Sensibility are distinguished from one another and attributed particular character traits through specific features of their dialogue. This study involved a mainly quantitative analysis of indirect cues in each of the two sisters’ spoken discourse: for example, Hubbard established that the elder sibling Elinor uses causal conjunctions (so, because, therefore, for this reason) with significantly greater frequency than Marianne, a finding which is consistent with direct (narratorial) definitions of Elinor as the more sensible sister, calm and collected in her judgment. Identifying such a pattern in the dialogue of these actors across the full length of Austen’s 120,000-word text would have been less feasible—or at least hugely time-consuming—through manual analysis alone; by contrast, the software developed for corpus-based studies can reveal such patterns at the mere click of a button.

Another powerful illustration of the potential of corpus software for the study of characterization can be found in Balossi’s (2014) analysis of Virginia Woolf’s The Waves. Here too, corpus tools are shown to enhance the analyst’s capacity to observe repeated patterns in the textual presentation of each of the novel’s core characters. For example, Balossi’s (2014, p. 115) analysis of the ways in which Susan expresses herself shows a frequent preoccupation with the natural world, conveyed through the higher than expected prevalence of lexical items related to animals and plants. As the researcher notes, this is significant for the way it highlights “one of Susan’s most salient traits: her tendency to represent her feelings and emotions in highly physical terms” (i.e., through reference to and analogy with her surroundings), as well as her strong sense that the countryside is where she belongs (2014, p. 117). Once again, therefore, a corpus-based method is shown to allow the analyst to engage in a systematic manner with the character profile of each of the narrative’s protagonists, and to provide a more comprehensive account of their traits, habits and concerns. No study has so far considered how this computer-assisted approach might be developed in order to compare and contrast characterization patterns present within multiple retranslations of a single source. This is consequently the research gap the present paper seeks to address.

(Re)constructing ‘the common people’

Were this analysis focused on the characterization of an individual actor such as Pericles or Themistocles, a logical first step of the inquiry would have been to search for that character’s name in the corpus browser to generate a concordance of all the segments of each translator’s version of Thucydides’ History in which the name appears. An interesting—if rather more challenging—feature of the character under investigation in the present study (‘the common people’) is that it is referred to by several expressions in the Greek original (Ober, 1989, p. 10), as well as in the two translations examined here. It is variously referred to in Bloomfield and Lattimore as the people, the common people, the many, the masses, the multitude and by numerous other signifiers, often within the same passages but also across each text as a whole. Consequently, I begin the analysis by establishing which designations are chosen more or less frequently by the two translators and highlighting the implications of these textual choices for the characterization of the ordinary population of Thucydides’ Greece. I then discuss a number of the qualities that are attached by Bloomfield to one such label, the multitude, and contrast these attributive propositions with those found in Lattimore’s translation.

Table 1 shows the variety of expressions used to refer to the non-elite citizens that populate Thucydides’ History. This list was extracted from the output of the word frequency list plugin developed for the Genealogies of Knowledge project: this tool provides the user with a complete record of all tokens in a text or selection of texts and ranks them in order of their frequency of occurrence. Discarding function words such as a, in, and with, which tend to account for most of the high frequency items in any corpus, allows us to scan the lexical items in the list relatively quickly to identify most—if perhaps not all—of the designations which could conceivably be used to describe ‘the common people’, albeit sometimes only when preceded by the definite article, as in the many. Concordances were then generated for each of these tokens in turn (for example, people, masses and many) and arranged alphabetically by the item to the left of the node word in order to identify not only the frequency of single tokens such as people but also recurrent combinations such as the people and the common people which are used to refer to the actor under study.

Table 1 Frequency of lexical items used by Bloomfield and Lattimore to refer to the ordinary (non-elite) members of society

Also presented in Table 1 are figures indicating the frequency with which each translator uses these designations. These purely quantitative results should not of course be taken at face value: the concordances generated for each of these terms do in many cases contain a number of false positives, i.e., instances where the items searched for are not in fact used as labels to distinguish the ordinary citizens from any other actor in the History, but instead form part of a larger segment of meaning which is not relevant to our current purposes. This seems to be particularly the case for public, which is most commonly used as an adjective, as in at the public expense, rather than as a noun, as in the public decided.

Nevertheless, these figures can provide initial insight into the divergent characterizations suggested by each translation. In particular, we might note how the predilection apparent in Bloomfield’s version for describing this actor as the multitude (60 instances of multitude and 44 of the multitude, as opposed to 4 and 0 instances in Lattimore, respectively) contrasts with Lattimore’s counter-preference for expressions such as common people, the majority and the masses. As Raymond Williams (1976, p. 192) notes, the multitude has long been a term of “open political contempt” for describing the majority of a population: indeed, Williams suggests this was the key word in early modern England for describing the lower social classes, and that it has always been overwhelmingly negative or derogatory in its connotations. This is largely because the label multitude stresses the large quantity of individuals grouped under this category, being derived from the Latin adjective multus meaning ‘many’ or ‘a great number’ (Lewis, 1895, p. 521). Through this word choice, emphasis would thus appear to be placed on the heterogeneity of this group, the competing interests of its members and the disordered variety of their desires and aspirations. These connotations are explicitly verbalized by the adjectives modifying multitude in some of the lines extracted from Bloomfield’s translation (Fig. 1).

Fig. 1
figure 1

Concordance lines extracted from Bloomfield’s translation

Lattimore’s preference for common people (34 instances, vs. 11 in Bloomfield), by contrast, suggests a quite different characterization of this actor. For a start, it can be argued that common has rather more positive connotations than multitude. As exemplified by the other nouns modified by this adjective in Lattimore’s text (Fig. 2), common suggests solidarity and an ordered unity. It is associated with shared goals (‘common cause’, lines 2–5; ‘common interest(s)’, lines 11–13) and agreement (‘common resolution(s)’, lines 17–18; ‘common sense’, line 22). The ‘common people’ are therefore positioned not as a ‘heterogeneous multitude’, but as a solid aggregate, acting together with one force.

Fig. 2
figure 2

Concordance lines of common [+NOUN] extracted from Lattimore’s translation and ordered alphabetically by the word in position 1 to the right

The same can be said of the masses (6 instances in Lattimore vs. none in Bloomfield), another referring expression for ‘the common people’ which likewise conjures images of homogeneity as opposed to multiplicity (Williams, 1976, p. 195). Similarly, through its associations with democratic processes of voting, the majority (36 instances in Lattimore vs. 3 in Bloomfield) implies a certain unity of will and of purpose, suggesting adherence to a shared set of beliefs and objectives, again in marked contrast to Bloomfield’s use of the multitude.

The extent to which these choices are the result of deliberate, conscious strategies on the part of Lattimore or Bloomfield is of course extremely difficult if not impossible to assess. As noted in my introduction, many forces are at play in the retranslation process and the differences observed must consequently be explained with reference not only to each translator’s personal politics, but also to broader trends of linguistic and ideological change within society at large. It should additionally be acknowledged that this analysis does not take into consideration the specific terms used to refer to ‘the common people’ in the original Greek and thus the extent to which the divergences identified between the two translations may be attributed to a desire for greater fidelity to the source on the part of Bloomfield or Lattimore. Thucydides himself was a member of the Athenian aristocracy and it is now widely accepted that he was rather ambivalent in his opinion of democracy, remaining undecided in his assessment of the suitability of the ordinary Athenians to hold political power (Lattimore, 1998, p. 16). Unfortunately, however, the Genealogies of Knowledge corpus analysis tool does not allow for the automatic alignment of source and target texts, and so deeper quantitative analysis along these lines is not currently feasible. That said, it is in my view likely that the patterns discussed above can be attributed at least in part to the radically diverging sets of stereotypical knowledge that have formed the schemata of interpretation for each translator’s representation of the non-elite members of the ancient Greek societies described.

Having illustrated the extent to which the choice of referring expression might influence the characterization prompted by each translator’s target text, we can now proceed to explore in more detail the particular traits and qualities attributed to the ordinary non-elite actors of the History by Bloomfield in contrast with Lattimore. As explained above, given limitations of space, this next section of the analysis will focus only on investigating those direct character definitions associated with the multitude. The multitude has been selected for closer analysis as a referring expression in whose use quantitative analysis shows there to be the greatest disparity between the two translators (44 lines in Bloomfield, vs. none in Lattimore—see Table 1). It is also a label with especially strong pejorative connotations and which therefore promises to reveal particularly striking results when comparing the two retranslations.

In this second part of the analysis then a concordance was generated for the multitude as it appears in Bloomfield’s translation (Fig. 3), and all lines containing only indirect character cues (such as lines 4, 14, and 41), examples of other-presentation (such as lines 21 and 31) and/or false positives (such as lines 10 and 22) were removed using the Delete Line function of the concordance browser.Footnote 9 This filtering process left just the five concordance lines presented in Fig. 4. Line 1 in Fig. 4 can be considered an almost prototypical example of direct character definition for the way Bloomfield’s Thucydides bluntly attributes a specific named quality (a lack of diligence in the search after truth) to the multitude. Lines 2–5, on the other hand, are categorized as such for the way a particular action (or reaction) is presented as usual (lines 3, 4, and 5) or accustomed (line 2) behavior among the multitude, suggesting this can be seen as indicative of a generalized character trait, a habit that defines their nature.

Fig. 3
figure 3

Full concordance of the multitude generated from Bloomfield’s translation

Fig. 4
figure 4

Direct character definitions of the multitude identified in Bloomfield’s translation

Expanding the co-text displayed for these five concordance lines individually (Fig. 5) through the Extract function of the browserFootnote 10 shows that they all seek to ascribe strongly negative qualities to this actor. Specifically, in lines 1, 2 and 3, Bloomfield’s Thucydides appears to construct the multitude as careless, rash and impulsive: they are largely incapable of exercising independent critical thought, the implication being that they tend instead to accept blindly the arguments of others, to let themselves get swept up in the rhetoric of the assembly and to act as a result of their passions rather than out of logic, planning or strategy. This impression is further cemented in Line 4 where Bloomfield’s rendering presents the multitude as a somewhat cowardly group in society, “readily disposed to be orderly” and to accept the rule of a newly appointed board of elders in the face of the Spartan threat, while in line 5 the nineteenth-century translator attributes the quality of fickle inconstancy (“levity”) to this narrative actor as one of its major character flaws.

Fig. 5
figure 5

Comparison of Bloomfield’s and Lattimore’s translations of specific passages of Thucydides’ History

In each of these instances, Bloomfield’s rendering seems to leave little doubt as to the character to whom these overwhelmingly negative attributes are being attached: the multitude is clearly presented as a specific class in society, and Bloomfield’s Thucydides appears to be repeatedly highlighting the irrationality and irresponsibility of this group as a political actor. Lattimore’s translations of these same passages (found at 1.20.3, 6.63.2, 4.28.3, 8.1.3, and 2.65.3 of Thucydides text, respectively) offer evidence of the potential for quite a different reading of the original Greek. For a start, we can observe that, according to Lattimore, these statements are not made in relation to the behavior of the multitude, but to the actions of most people, a mob, a crowd, a democracy and a multitude. In this alternative, late-twentieth century rendering, Thucydides is making claims about human beings in general, rather than about a specific class. It is no longer the multitude in particular who are careless with the truth (line 1), but the population as a whole with few exceptions. Similarly, lines 2, 3, and 5 are transformed into theoretical musings on crowd psychology, that is, on the fact that when brought together as members of a large group, individuals often act in ways they would not usually, typically leading to irrational and/or aggressive behavior. Line 4, finally, has more to do with political theory and the advantages of democracy as a system of government in dealing with crisis situations: indeed, in a footnote to this passage, Lattimore suggests that Thucydides seeks to indicate here “that the resilience and flexibility essential to the Athenian recovery was derived from democratic institutions” (1998, p. 412n).

Given the ideological significance of this divergence between the two versions for the characterization of the ordinary people, examining the source text is helpful at this stage to establish the extent to which each translator has actively reinterpreted the original Greek. The noun used in lines 2 and 3 is ὄχλος (ochlos) which, as the Liddell-Scott-Jones Greek-English Lexicon confirms, can be understood to mean both multitude (Bloomfield), and mob or crowd (Lattimore). Similarly, the Lexicon suggests that most people, multitude and democracy are all valid translations of the terms used by Thucydides in lines 1 (πολλοί/polloi), 4 (δῆμος/dēmos), and 5 (ὅμιλος/homilos). Yet, it is important to note that, with the sole exception of line 1Footnote 11, the Greek definite article is missing from the relevant segments of the source text. To render these items repeatedly as the multitude can therefore be seen as an important interpolation by Bloomfield at these points in the text.

Other notable differences between the two translations include the absence of Bloomfield’s value-laden adverb clamorously from Lattimore’s translation of the passage in Line 3, and the much less explicit means by which this late-twentieth century translator renders the phrase translated by Bloomfield as “with the accustomed levity of the multitude” in Line 5 (emphasis added). Once again, inspecting the source text in these two cases provides further evidence of the extent to which Bloomfield has made more decisive an intervention in the characterization of the ordinary (non-elite) members of classical societies than Lattimore. No equivalent for clamorously can be found in the original Greek for Line 3, and the character trait of “levity” would appear to be the result of a strategy of extreme explicitation on the part of the nineteenth-century translator: “ὅπερ φιλεῖ ὅμιλος ποιεῖν” can simply be rendered “as any throng of people likes (or is wont) to do.”

Concluding remarks

In this last case especially, the differences between the two translations might be explained with reference to questions of translator’s style and perhaps the desire of Bloomfield to make his target text as immediately accessible as possible for his intended readers. Indeed, explicitation has long been a popular strategy adopted by translators to clarify the meaning of otherwise ambiguous passages of a source work (Blum-Kulka, 1986). The addition of the adverb ‘clamorously’ in Line 2 could similarly be motivated by a desire to increase the affective potential of Thucydides’ History through a more explicit appeal to the readers’ senses. Finally, we must also take heed of the fact that, as acknowledged in his introduction to the History (1998, pp. 13–21), Lattimore has benefitted from over a century of intense scholarship and scrutiny focused on almost every line of Thucydides’ text, a critical resource that simply was not available to Bloomfield at the beginning of the nineteenth century. Nevertheless, it is reasonable to suggest that, based on the analysis presented here, Bloomfield may have been influenced at least in part by his prior assumptions concerning the lower classes and their behavior in society as he worked on this translation. Based on the evidence discussed above, I have argued that this translator’s stereotypical and social role knowledge regarding the habits, traits and tendencies of the ordinary members of his own society acted as a lens, refracting his understanding of the Greek and the specific events, actors and situations Thucydides narrated. This contextually determined schema of interpretation seems to have led Bloomfield to label the general population of Athenian society contemptuously as the multitude with striking frequency and to ascribe to them a series of negative qualities that are not present in Lattimore’s twentieth-century version nor in the corresponding passages of the source text.

Future research could usefully investigate the more indirect character cues through which the multitude are constructed in Bloomfield’s translation. As discussed in the section entitled “Narrative characterization in historiographical texts”, characters are constructed in narratives not only through direct narratorial definitions but also through indirect acts of commission and omission: the traits implied by these acts could readily be identified by using the concordance browser to examine the verbs for which non-elite members of society serve as subject and object. It is interesting to observe, for instance, that while the peace treaty between the Athenians and the Spartans at 4.118.10 is simply “approved by the people” in Bloomfield’s translation, it is “resolved by the people” in Lattimore’s, the latter rendering suggesting a much more active part in the political decision-making process than the former. Corpus tools enable researchers to explore whether or not such language use is part of a wider pattern across the two target texts as entire works, and thus to investigate the extent to which each translator’s assumptions about the proper role of ordinary citizens in politics might have informed their translation of the History.

Finally, further studies could additionally explore the extent to which the patterns observed in Bloomfield’s and Lattimore’s translations can be found in other translations produced in the nineteenth and twentieth centuries. As mentioned in the introduction, the Genealogies of Knowledge Modern English corpus includes several other translations of Thucydides, such as those by Henry Dale (1848), Richard Crawley (1874), Benjamin Jowett (1881), and Charles Smith (1919), but the investigation could also productively extend to translations of many other classical historians and political philosophers, such as Herodotus, Xenophon, Plato and Aristotle. Multiple retranslations of the works of these authors too are available for interrogation via the Genealogies of Knowledge concordance browser and other software tools developed by the project team. It is hoped that the methodology developed here and the resources and software tools created by the project might help provide further insight into the complex interactions that occur between translation, politics and society as texts and ideas travel through time and across languages.