Looking for Textual Evidence Digital Humanities, Middling-Class Morality, and the Eighteenth-Century English Novel Ralf Schneider, Marcus Hartner, Anne Lappert 1. Introduction In our contribution to this edited volume we present a discussion of an attempt to identify and locate literary manifestations of the idea of the “virtuous social middle” in a large corpus of eighteen-century English nov­ els with the help of methods and tools from Digital Humanities (DH).1 This attempt was situated within the larger context of a research project on com­ parative practices in the eighteenth-century novel as part of the Collaborative Research Center (CRC) 1288 “Practices of Comparing” funded by the German Research Foundation (DFG). Our project started from three assumptions. The first was the traditional assumption held in literary history about the close connection between socio-historical developments and the “rise of the novel” 2 from “the status of a parvenu in the literary genres to a place of dominance” during the eighteenth century.3 Second, we assumed that the cultural construction of the “the middle order of mankind” 4 and its concom­ itant claims about a supposedly heightened sense of ‘middle-class’ morality 1 Wahrman, Dror, Imagining the Middle Class: The Political Representation of Class in Brit- ain, c. 1780 –1840, Cambridge: Cambridge Universit y Press, 1995, 64. 2 Watt, Ian, The Rise of the Novel: Studies in Defoe, Richardson, and Fielding, Berkeley: Uni- versit y of California Press, 1957. 3 Rogers, Pat, Social Structure, Class, and Gender, 1660 –1770, in: J. A. Downie (ed.), The Oxford Handbook of the Eighteenth-Centur y Novel, Oxford: Oxford Universit y Press, 2016, 39. 4 Goldsmith, Oliver, The Vicar of Wakefield, Oxford: Oxford Universit y Press, 2006 [1766], 87. Ral f Schneider, Mar cus Har tner, Anne L apper t24 0 was accompanied by a range of social processes of comparing.5 Af ter all, con­ structions of social identity tend to rely heavily on processes of othering, and comparing plays a vital role in the construction of self and other. Third, we assumed that in the emerging medium of the novel in the period under inves­ tigation, concepts of middle-class social identity were negotiated through particular literar y strategies of comparing, whose textual manifestations can be found specifically in textual representation of characters and charac­ ter constellations. Ultimately, the underlying value system concerning class identity in a novel ought to manifest itself also in the way that the behavior or dispositions of characters are described and evaluated in comparison as either desirable and adequate, or as despicable and inappropriate. As part of our strategy of substantiating those three assumptions, our project aimed at providing a more extensive review of the textual representations of social virtues and vices in the eighteenth-century English novel than available in traditional scholarly accounts of the topic so far. In order to achieve this aim, we decided to turn to the methods of DH. We planned to identify, with the help of dif ferent types of word searches (see below), recurrent expressions that refer to social behavior in either pos­ itive or negative terms. We expected a diachronic development to be visible across the corpus, e. g., similar to the way concepts of gentility changed their semantics during the period under consideration.6 None of our expectations were met, however, as we will demonstrate below. This prompted a reconsid­ eration of our search strategies and ultimately led to the insight that prac­ tices of comparing and social-identity construction may be more implicit in 5 In the following, we will employ the term ‘middle-class’ as a synonymous st ylistic varia- tion to expressions such as ‘middle order’, ‘middle rank’, the ‘middling sorts’, etc. We are aware that the application of the terminolog y of class to discussions of eighteenth-cen- tur y societ y is contested and comes with certain conceptual problems. For introductions to the term and concept of class in early modern Britain, see Corfield, Penelope J., Class by Name and Number in Eighteenth-Centur y England, in: Histor y 72 (1987) and Cannadine, David, Class in Britain, London: Penguin, 2000, 27, 31. 6 The concept of the gentleman, for example, changed from the narrow denotation of a man of noble birth to the more widely applicable notion of a man displaying a set of ‘genteel’ (moral) qualities and behaviours. During this “social peregrination” of the term, it lost “its oldest connotations of ‘gentle’ birth and ‘idle’ living, so that, in the later eigh- teenth centur y, individual vintners, tanners, scavengers, potters, theatre managers, and professors of Divinit y could all claim the status, publicly and without irony” (P. J. Corfield, Class by Name and Number, 41). L ooking for Tex t ual Evidence 241 literature than in other discourses and function in dif ferent ways. In what follows, we will first sketch the socio-cultural context of our corpus, in which the novels contribute to the negotiation of middle-class morality. We will then brief ly engage with the question of the applicability of DH methods in the analysis and interpretation of literature, before we document some of our text searches and discuss the results. 2. Inventing the superiority of the middling classes On the opening pages of Daniel Defoe’s The Life and Adventures of Robinson Crusoe (1719) the title character’s elderly father lectures the youthful pro­ tagonist on his place in the social fabric of eighteenth-century Britain. In his attempt to dissuade the restless and adventure-seeking Robinson from “[going] abroad upon Adventures”, he emphasizes his son’s birth into the “the middle State” of society.7 This he declares to be “the best State in the World, the most suited to human Happiness” as it is neither “exposed to the Miseries and Hardships, the Labour and Suf ferings of the mechanick Part of Man­ kind”, nor is it “embarrass’d with the Pride, Luxury, Ambition and Envy of the Upper Part of Mankind”.8 While those remonstrations unsurprisingly fail to convince the young Robinson Crusoe, they articulate a sentiment of ‘middle-class’ complacency found with increasing frequency in literary and philosophical writings over the course of the eighteenth century. Defoe’s fictional character constitutes only one voice in an increasingly audible choir within the cultural discourse of the period that promotes the idea of the ‘middle order’ as possessing a distinct and superior quality. Though this idea was neither new nor universally acknowledged,9 it became increasingly 7 Defoe, Daniel, Robinson Crusoe, ed. Michael Shinagel, New York: Norton, 1994 [1719], 5. 8 Ibid. 9 On competing models of the social structure of the period, such as the notion of a bipolar “crowd-gentry reciprocity” (Thompson, E. P., Customs in Common: Studies in Traditional Pop- ular Culture, New York: New Press, 1993, 71) and the persistent traditional belief in a prov- identially ordained, universal and hierarchical order of social layers (e. g. Tillyard, E. M. W., The Elizabethan World Picture: A Study of the Idea of Order in the Age of Shakespeare, London: Chatto  & Windus, 1967 [1942]), see the discussion in D. Cannadine, Class in Britain, 24–56. With regard to the notion of the superiorit y of the ‘middle-class’, see also French, who argues that the aristocracy and gentr y retain their dominant economic and politi- Ral f Schneider, Mar cus Har tner, Anne L apper t242 attractive to those who saw themselves as belonging to this particular seg­ ment of society.10 Building on the notion of a “virtuous social middle”11 ini­ tially developed in Aristotle’s Politics,12 they actively engaged in the discur­ sive construction of the middle order as a distinct social group not only by discussing its political and economic importance for the nation,13 but also by emphatically emphasizing its moral value.14 David Hume, for example, thought that the upper classes were too immersed in the pursuit of pleasure to heed the voices of reason and morality, while “the Poor” found themselves entirely caught up in the daily struggle for survival.15 As a result, in his view, only the “middle Station” af fords “[…] the fullest Security for Virtue; and I may also add, that it gives Opportu- nity for the most ample Exercise of it […]. Those who are plac’d among the lower Rank of Men, have little Opportunity of exerting any other Virtue, besides those of Patience, Resignation, Industry and Integrity. Those who are advanc’d into the higher Stations, have full Employment for their Gen- erosity, Humanity, Af fability and Charity. When a Man lyes betwixt these two Extremes, he can exert the former Virtues towards his Superiors, and the latter towards his Inferiors. Every moral Quality, which the human Soul is susceptible of, may have its Turn and, and be called up to Action: And a Man may, af ter this Manner, be much more certain of his Progress in Virtue, than where his good Qualities lye dormant, and without Employment.”16 cal power in Britain throughout the eighteenth centur y and beyond (French, Henry, Gen- tlemen: Remaking the English Ruling Class, in: Keith Wrightson (ed.), A Social History of England: 1500 –1750, Cambridge: Cambridge University Press, 2017, 269, 280). See also Mul- drew, Craig, The ‘Middling Sort’: An Emergent Cultural Identit y, in: Keith Wrightson (ed.), A Social History of England: 1500 –1750, Cambridge: Cambridge University Press, 2017 on the emergence of the “Middling Sort” as a cultural identit y during the early modern period. 10 D. Cannadine, Class in Britain, 32–33. 11 D. Wahrman, Imagining, 64. 12 Aristotle, The Politics, trans. Carnes Lord, Chicago: Universit y of Chicago Press, 1984, IV.11. 13 D. Cannadine, Class in Britain, 42. 14 The protagonist Charles Primrose in Oliver Goldsmith’s The Vicar of Wakefield, for exam- ple, sees “the middle order of mankind” as the social sphere that is home to “all the arts, wisdom, and virtues of societ y” (87–88). 15 Hume, David, Of the Middle Station of Life, in: Thomas H. Green/Thomas H. Grose (eds.), David Hume, The Philosophical Works, Aalen: Scientia Verlag, 1964 [1742], 4:376. 16 Ibid., 4:376–4:377. L ooking for Tex t ual Evidence 24 3 The passage indicates that Hume sees the middling class’s superior virtue as the result of a sociological process. By being exposed to a wider and more complex range of social life, individuals from the middle ranks are forced to develop greater moral sensitivity and power of judgement. While he thus attempts a philosophical explanation,17 other contemporary authors cham­ pion middle-class virtue in a more simplistic fashion by rhetorically fore­ grounding the idea of a stark contrast between the “generous Disposition and publick Spirit” of members of the middling ranks and the “Depravity and Selfishness of those in a higher Class”.18 It is important to note once more that such arguments about the (moral, economic, political, etc.) superiority of a distinct middle order or class, were less “an objective description of the social order” in Britain than “a way of constructing and proclaiming favourable ideological and sociological ste­ reotypes” of those who found themselves hierarchically situated between the poor and the powerful.19 In this context, the development of the eigh­ teenth-century novel as a distinct literary genre on the fast-growing market for printed material can be seen instrumental in the emergence of the (self-) image of the middle class as an economically relevant and culturally powerful social group.20 Written by (predominantly) middle-class authors for a (pre­ dominantly) middle-class audience,21 the novel played an important role in the invention and promotion of this group’s social identity, especially by con­ 17 For a discussion of Hume’s position in relation to that of Aristotle, see Yenor, Scott, Da- vid Hume’s Humanit y: The Philosophy of Common Life and Its Limits, Basingstoke: Pal- grave, 2016, 114–119. 18 Thornton, William, The Counterpoise: Being Thoughts on a Militia and a Standing Army, London: Printed for M. Cooper, 1752. Quoted from the unpaginated preface. 19 D. Cannadine, Class in Britain, 32. 20 The connection between the “rise of the novel” and the emerging middle class was first discussed in I. Watt, Rise of the Novel, and Habermas, Jürgen, The Structural Transforma- tion of the Public Sphere: An Inquir y into a Categor y of Bourgeois Societ y, trans. Thomas Burger, Cambridge: Polit y, 2015 [1962]. For a survey of perspectives af ter those authors, see Cowan, Brian, Making Publics and Making Novels: Post-Habermasian Perspectives, in: J. A. Downie (ed.), The Oxford Handbook of the Eighteenth-Centur y Novel, Oxford: Oxford Universit y Press, 2016. 21 Hunter points out that the readership of the novel was never restricted to one specific group only. In contrast to the argument presented here, he holds that “the characteris- tic feature of novel readership was its social range […] and the way it spanned the social classes and traditional divisions of readers” (Hunter, Paul J., The Novel and Social/Cultural Ral f Schneider, Mar cus Har tner, Anne L apper t24 4 tributing to the illustration and dissemination of the concept of middle-class morality.22 As a result, a preoccupation with the figure of the individual forced to navigate morally complex situations, together with the frequent vilification of characters from aristocracy and gentry, as well as a complacent middle-class contentedness with being placed in the ‘best’ social stratum, set the tone for much eighteenth-century prose writing.23 However, while the general connection between “the emergence of a bourgeois public sphere” and “the rise of novel writing and -reading” has long been treated as “a stan­ dard feature” of the period’s literary and cultural history,24 the aesthetic and narratological dimensions of the “invention” of middle-class superiority25 still remain a productive field of study. For this reason, our research project within the CRC 1288 “Practices of Comparing” set out to investigate the novel’s contribution to eighteenth-cen­ tury negotiation of social identity and morality by focusing on the play of nar­ rative and stylistic strategies that constitute an important aspect of this con­ tribution. As we are traditionally trained literary scholars, the methodological thrust of our project lay in the informed manual analysis of an ambitious, yet manageable corpus of some twenty carefully selected novels from the period. We specifically decided to focus on classical narratological analyses of aspects such as narrative situation, focalization, and perspective structure26 as well Histor y, in: John Richetti (ed.), The Cambridge Companion to the Eighteenth-Centur y Novel, Cambridge: Cambridge Universit y Press, 1996, 19). 22 Nünning, Vera, From ‘honour’ to ‘honest’. The Invention of the (Superiorit y of ) the Mid- dling Ranks in Eighteenth Centur y England, in: Journal for the Study of British Cul- tures 2 (1994). 23 For more detailed surveys of the eighteenth-centur y novel and its contexts, see Nünning, Ansgar, Der englische Roman des 18.  Jahrhunderts aus kulturwissenschaf tlicher Sicht. Themenselektion, Erzählformen, Romangenres und Mentalitäten, in: Ansgar Nünning (ed.), Eine andere Geschichte der englischen Literatur. Epochen, Gattungen und Teilge- biete im Überblick, Trier: WVT, 1996, and the contributions in Richetti, John (ed.), The Cambridge Companion to the Eighteenth-Centur y Novel, Cambridge: Cambridge Uni- versit y Press, 1996 and Downie, J. A. (ed.), The Oxford Handbook of the Eighteenth-Cen- tur y Novel, Oxford: Oxford Universit y Press, 2016. 24 P. Rogers, Social Structure, 47. 25 V. Nünning, From ‘honour’ to ‘honest’. 26 Fludernik, Monika, An Introduction to Narratolog y, London: Routledge, 2009, Wenzel, Peter (ed.), Einführung in die Erzähltextanalyse: Kategorien, Modelle, Probleme, Trier: WVT, 2004, Nünning, Ansgar, Grundzüge eines kommunikationstheoretischen Modells L ooking for Tex t ual Evidence 24 5 as on the representation of fictional characters.27 Our individual (close) readings indeed produced results that hermeneutically seem to confirm our assumptions of a middling-class preoccupation with social identity. Nev­ ertheless, we remained painfully aware of the limited scope of our project design regarding the number of texts that we were able to incorporate into our investigation. And we wondered if we could complement the traditional literary analyses of our research by turning to DH in the attempt to engage with at least some aspects of our research on a digital and somewhat broader textual basis. 3. Between close and distant reading: using DH methods for literary analysis and interpretation While the tentative origins of DH reach back into the first half of the twen­ tieth century,28 most of its methods and research questions fully emerged only during the past few decades. One branch of the wider field of DH has concerned itself with literary texts; and its exploration of the relationship between literature and the computer has taken many shapes. One major issue is the production and increasing availability of electronic (and schol­ arly) editions of primary and secondary works. This development has sig­ nificantly widened access to literary texts and now plays a vital role in the preservation of books and other textual materials;29 it has forced libraries and academic institutions to develop new data policies and technological solutions for storing and providing access to primary and secondary lit­ erature. Also, not only computer-related genres such as literary hypertexts der erzählerischen Vermittlung: Die Funktion der Erzählinstanz in den Romanen George Eliots, Trier: WVT, 1989. 27 Margolin, Uri, Character, in: David Herman (ed.), The Cambridge Companion to Narra- tive, Cambridge: Cambridge Universit y Press, 2007, Eder, Jens/Jannidis, Fotis/Schneider, Ralf (eds.), Characters in Fictional Worlds: Understanding Imaginar y Beings in Literature, Film, and Other Media, New York: de Gruyter, 2010. 28 Thaller, Manfred, Geschichte der Digital Humanities, in: Fotis Jannidis/Hubertus Kohle/ Malte Rehbein (eds.), Digital Humanities. Eine Einführung, Stuttgart: J. B. Metzler, 2017, 3–4. 29 Shillingsburg, Peter L., From Gutenberg to Google: Electronic Representations of Literar y Texts, Cambridge: Cambridge Universit y Press, 2006. Ral f Schneider, Mar cus Har tner, Anne L apper t24 6 have emerged,30 but digital technologies have dramatically changed both the publishing and book industry, so that many literary and scholarly texts now­ adays are read not from the printed page but from the displays of e-book devices. In the context of those developments, the impact of the digital revolu­ tion on the academic infrastructure of the humanities is without question. But while computers have long found their place even in the of fices of the most technophobic academics, and while even the rear-guard of traditional literary scholars use digital information retrieval systems such as electronic library catalogues and databanks, there is still widespread resistance to some other applications of digital methods in literary research. And indeed, in the realm of literary analysis and interpretation things look a bit complicated. On the one hand, textual analysis can very well apply digitized methods, in ways comparable to the strategies of computational and corpus linguistics. In the wide field of stylometrics, for instance, large corpora of texts can be scanned for the co-occurrence of particular textual features, which can then help trace historical developments in literary language, attribute authorship, or define genres.31 Also, the themes that dominate a text can be extracted by topic modeling.32 On the other hand, when it comes to the interpretation of literary works, there is some skepticism as to the ability of computer pro­ grams to support human readers in tasks of that complexity. Although tex­ tual analysis is always the basis for interpretation, interpretation is usually performed, af ter all, by highly educated, well-informed academic readers with a hermeneutic interest in exploring the meaning – or meanings – of a text. The main interest in interpretation lies in investigating a text’s combi­ 30 Ryan, Marie-Laure, Avatars of Stor y, Minneapolis: Universit y of Minneapolis Press, 2006, Ensslin, Astrid, Hypertextualit y, in: Marie-Laure Ryan/Lori Emerson/Benjamin J. Robert- son (eds.), The Johns Hopkins Guide to Digital Media, Baltimore: Johns Hopkins Univer- sit y Press, 2014. 31 Burrows, John, Delta: A Measure for St ylistic Dif ference and a Guide to Likely Authorship, in: Literar y and Linguistic Computing  17 (2002), Jannidis, Fotis/Lauer, Gerhard, Burrows’s Delta and Its Use in German Literar y Histor y, in: Matt Erlin/Lynne Tatlock (eds.), Distant Readings: Topologies of German Culture in the Long Nineteenth Centur y, Rochester: Camden House, 2014. See also the extensive introduction and survey by Juola, Patrick, Authorship Attribution, in: Foundations and Trends in Information Retrieval 3 (2006), 233–334, http://dx.doi.org/10.1561/1500000005. 32 Jannidis, Fotis, Quantitative Analyse literarischer Texte am Beispiel des Topic Modelings, in: Der Deutschunterricht 5 (2016). http://dx.doi.org/10.1561/1500000005 L ooking for Tex t ual Evidence 247 nation of thematic, aesthetic and rhetorical features which are understood to be culturally embedded in complex ways. Both ample contextual research and the close scrutiny of textual features are therefore generally considered prerequisites of literary interpretation. The ‘distant’ reading, i. e., the computerized analysis of textual patterns in texts, that DH have introduced to literary scholarships, thus looks fairly incompatible at first sight with the close reading and interpretation strate­ gies practiced by the scholar trained in literary hermeneutics. Franco Moretti famously spoke of distant reading as “a little pact with the devil: we know how to read texts, now let’s learn how not to read them”.33 But the advantage of distant reading is that it allows scholars to detect features across a number of texts that could only with dif ficulty and considerable use of resources be tackled by individual close readings. While the computer may lack the ability to detect ‘qualitative’ dif ferences, it is its promise of a seemingly boundless quantitative analytical scope that turns it into a potentially powerful ana­ lytic tool. Moreover, DH not only of fers the opportunity to extend existing research strategies in a quantitative fashion, but the playful exploration of digital tools may also lead to unexpected results and even contribute to the emergence of new research strategies. Emphasizing the productive power of playfulness and creativity, Stephen Ramsay advocates an informal “Herme­ neutics of Screwing Around” as a valid computer-based research strategy for the Digital Age in an inf luential paper.34 Concerned with the limited scope of the hermeneutical (close) readings in our project, we were intrigued both by this lure of quantitative analysis and the emergence of the “somewhat infor­ mal branch of text interpretation delightfully termed screwmeneutics” af ter Ramsay.35 Therefore, we decided to embark on a complimentary investiga­ tion of the textual manifestations of some concepts of middle-class virtue in the eighteenth-century novel with the help of DH. 33 Moretti, Franco, Distant Reading, London: Verso, 2013, 48. 34 Ramsay, Stephen, The Hermeneutics of Screwing Around; or What You Do with a Million Books, in: Kevin Kee (ed.), Pastplay: Teaching and Learning Histor y with Technolog y, Ann Arbor: Universit y of Michigan Press, 2014 [2010]. 35 McCurdy, Nina et al., Poemage: Visualizing the Sonic Topolog y of a Poem, in: IEEE Transac- tions on Visualization and Computer Graphics 22 (2016), 447. Ral f Schneider, Mar cus Har tner, Anne L apper t24 8 4. From search to research: some examples Our approach to using DH was unusual in so far as we did not take the more common route from distant to close reading but proceeded vice versa. Since we had already invested considerable ef fort in the (close) reading and analy­ sis of our original corpus of ca. twenty eighteenth-century novels, we began our journey into the field of DH equipped with a solid set of expectations about the literary negotiation of social identity during the period under investigation. Starting from the hermeneutical findings of our investigation, we then attempted to corroborate our results, by taking our research into the realm of computing, more precisely, by expanding the corpus of novels under investigation and developing ideas on how DH tools could help us to support our arguments. Our first step in this process was to expand our text base by creating a digital corpus of 55 novels (see the list in the appendix to this article), thus more than doubling the number of texts. We decided to look at some of the most well-known novels from the eighteenth century as well as to include some lesser known works that were however well received during the period in question. Further, we intentionally included works from dif fer­ ent genres such as sentimental novels, gothic novels, coming-of-age stories and adventure novels, in order to do some justice to the considerable variety and diversity in eighteenth-century literary production.36 Already during the process of compiling and preparing the corpus, how­ ever, we encountered the first methodological challenges. While DH of fers a great variety of tools and approaches, digitized texts are only ever suitable for a research purpose as they are prepared accordingly. In other words, if we were to look for complex sentence structures, or even narrative patterns conveying middle-class ideology, these structures would have to be tagged beforehand in each text. This means that passages that we consider as good examples for such patterns would have to be identified and electronically annotated accordingly in the hidden plane of text information, the markup. Not only did we need digital copies of all novels, but a lot of tagging by hand would have been necessary. The reason is that no program can automatically 36 See J. Richetti, The Cambridge Companion, Nünning, Ansgar and Vera, Englische Literatur des 18. Jahrhunderts, Stuttgart: Klett, 1998, and Backscheider, Paula R./Ingrassia, Catherine (eds.), A Companion to the Eighteenth-Centur y English Novel and Culture, Chichester: Wiley-Blackwell, 2006. L ooking for Tex t ual Evidence 24 9 mark up more complex structural features such as comparisons between characters that are not made explicit on the textual level, but are evoked through characters acting dif ferently in comparable situations, a strategy frequently used in prose fiction. To tag the texts for such features would be a very time-consuming process that presupposes an answer to our original question, namely what role practices of comparing play on a structural level in the textual constructions of social identity. This question would need to be answered before the markup could begin, since these structures would have to be analyzed before they could then be tagged in all texts of our cor­ pus. We would further risk to exacerbate the danger of confirmation bias that is structurally inherent to our approach anyway, as we would run the danger of finding exactly what we placed there during the tagging process. The sheer number of working hours that would have to be put into creating new digital versions with tags made this type of digital research impractical for a first, tentative and playful digital exploration of our expanded corpus of eighteenth-century novels. As a consequence of these first challenges we moved away from the idea to investigate complex syntactic and narrative structures, and turned to word and phrase searches as a feasible alternative, for which an array of DH tools are available, and for which simple text files suf fice.37 In this context, our assumption was that key terms denoting middle-class virtues and vices would be detectable in abundance across the novels of our corpus. While pro­ grams such as AntConc are especially promising when zooming in on indi­ vidual texts, Voyant proved to be more ef ficient when searching larger col­ lections of texts. Generally speaking, it is interesting to look at the frequency of words within one text and within a corpus, since words that occur very frequently (except for functions words such as conjunctions or articles, which we excluded from all searches) are likely to hint at the thematic focus of a text. Sometimes, however, the opposite of an expected word frequency may be revealing, too, as was the case in the searches we document below. Since we were also interested in diachronic developments, we began by using Voy­ ant, which of fered a direct comparison of word frequencies and the context 37 Project Gutenberg is the most easily accessible online text collection for such purposes. Although random checks of Project Gutenberg texts against the printed scholarly edi- tions we had read suggested that the former are not always entirely reliable, we decided that for the first stage of word searches, the results were unlikely to be heavily distorted. Ral f Schneider, Mar cus Har tner, Anne L apper t25 0 of their appearance across the corpus as a whole. We added the year of pub­ lication to the title in order to have the novels appear in chronological order of their publication, so that any diachronic changes would be immediately visible. Since the larger framework of our project was the study of the forms and functions of practices of comparing, our very first tentative approach was to run searches for words and particles that explicitly produce comparisons (such as more/less than, and words containing comparatives or superlatives ending on -er and -est). The result was that comparative words and particles occurred indeed frequently in our corpus (“more” = 14196 times, “less” = 2531, “than” = 12161, “like” = 4555). However, looking closer at our results it became apparent that words such as more were not always used to create an explicit comparison, but in many cases appeared in other contexts, such as to empha­ size the expressed meaning ( ‘still the more’), or to indicate temporality (‘once more’) in phrases like ‘little more than’, ‘still the more’, ‘many more’ and ‘once more’ (see fig. 1). Hence, the results of the context search put the result of the word frequency in question and provided a first indication that comparing in prose fiction might work in less explicit ways than in some other discourses. Fig. 1: Word search for “more” and immediate contexts We then turned to other word searches. Collecting results from our (close) reading of the selected text from our original corpus and in the playful spirit of “screwmeneutics” 38 we developed a list of terms that describe behavior and 38 N. McCurdy et al., Poemage, 447. L ooking for Tex t ual Evidence 25 1 dispositions in negative and positive ways that we considered to be important for the negotiation of social identity in eighteenth-century English novels. In particular, we decided to look for positively and negatively connotated adjec­ tives, but also noun phrases used in characterization by narrators and other characters, or in self-characterization. With this we aimed to make apparent the contrast between what were considered desirable or undesirable charac­ ter traits and actions and how these conceptions changed throughout eigh­ teenth-century literature. For this purpose, we created two lists of adjectives we came across in our close reading process and in our reading of second­ ary literature on the construction of social identity in the eighteenth centu­ ry.39 In the group of positive terms, we had collected such words as “gentle”, “gallantry” and “virtuous”; the negative ones included “foppish”, “conceited”, “impertinence”, etc. We then added other terms from these and related semantic fields and complemented the adjectives and adverbs with the per­ tinent noun phrases in an attempt not to overlook relevant textual manifes­ tations. This gave us a list that included the words “gentle” and “gentleman”, “gallant” and “gallantry”, “grace”, “graceful”, “gracious” and “graciousness”, “polite” and “politeness” “virtuous” and “virtue” for the positively conno­ tated behaviors and attitudes; the negatively connotated ones included “fool” and “foolish”, “fop”, “foppish” and “foppery”, “disagreeable”, “conceited” and “conceitedness”, “vulgar” and “vulgarity”, “impertinent” and “impertinence”, “impetuous” and “impetuosity”, as well as “negligent” and “negligence”. For ef ficient text searching, the truncated forms of these words were used.40 Our list then had for instance gentle*, tender*, grac*, gallant*, polit*, sweet* virtu*, modest*, moderat*, on the positive side, and fop*, fool*, disagreeab*, conceited*, vulgar*, impertinen*, impetuo*, negligen* on the negative. Figure 1 shows the frequency of both the negative and the positive search terms across our corpus. Since the corpus was organized chronologically, the graphic ought to show whether certain terms were used more or less frequently in later publications than in earlier ones. As we see in the dia­ gram, usage did vary considerably, but this variation shows no indication 39 V. Nünning, From ‘honour’ to ‘honest’, D. Wahrman, Imagining. 40 Using truncated forms, i. e., a word stem closed by an asterisk, allows the system to find instances of the stem in all variations and word classes; for example, gentl* would not only include the results for “gentle”, but also for “gently”, “gentleman”, “gentlemanly”, etc. Ral f Schneider, Mar cus Har tner, Anne L apper t252 of being related to diachronic changes during the time period. Frequencies rather vary from text to text. In fact, while individual texts may deviate from the median in a significant fashion, the overall frequency of the terms under investigation seems to remain more or less consistent over the entire eigh­ teenth century as far as our corpus is concerned. The underlying assumption guiding our approach was that the social changes in the understanding of the virtues and vices listed above would somehow be ref lected by changing word frequencies. Especially for gen- tle* did we expect to find a significant diachronic development, as notions of gentility changed from a rather narrow denotation of gentle birth to an understanding of polite behavior by the end of the century that made it pos­ sible for men from a significantly wider range of society to claim the status of a “Gentleman” (see FN 6). Contrary to our expectations, however, we were unable to discern significant developments in our search result. While gentle* indicated at least a slight discernible decrease of usage (see fig. 2), none of the other terms of fered a visible indication of a diachronic development. Put dif ferently, word f requencies did not hint at the emerging construction of a middle-class identity during the period, as described by eighteenth-century social history. One possible explanation for this may be that frequency cannot capture what a term means: While narrators and char­ acters in late eighteenth-century novels may use all variations of the words “gentle”, “gentleman”, etc. as frequently as those in the early phase, they may simply mean dif ferent things by those terms. With this possible explanation in mind, we decided to turn away from questions of diachronic development within the eighteenth century. Our next step was to look at the total word frequency of our search terms in the entire corpus. In order to corroborate our assumption that these terms play a significant role in the topics of the novels, we checked their position in the list of the most frequently appearing words within the body of novels under consideration. However, we were once more disappointed. The word count showed that out of the words we were looking for, most were situated in the lower ranks of the count, whereas words such as ‘said’, ‘Mr’, ‘time’ and ‘lit­ tle’ came up top of the list (see fig.  3). From our search list, only gentleman managed to enter the top 100 at position 81, followed by virtue at 239. The results for our negative terms proved to be even less impressive with, for instance, fool, reaching only the top 2000 of the most frequently used words in the corpus. Our positive terms generally ranked higher than our negative L ooking for Tex t ual Evidence 25 3 Fi g. 2 : F re qu en cy o f n eg at iv e a nd p os it iv e t er m s a cr os s t he co rp us Ral f Schneider, Mar cus Har tner, Anne L apper t25 4 terms with virtue at position 239 (1370 occurrences), agreeable at 440 (892 occurrences), sweet at 450 (881 occurrences), and tender at 299 (1163 occur­ rences). None of the negative terms made it above fool at position 1420 and with 329 occurrences. With all our negative terms ranking rather low and quite a number of our positive terms ranking comparatively higher and with a look at the most frequent words (especially “dear”, “great”, and “good”), one may speculate whether character traits might have been negotiated more in terms of stating an ideal during the period. This would mean that texts rather state what should be aimed for, while at the same time only implicitly hinting at negative traits and behaviors and hence, at what to avoid. On the other hand, our experience with words and particles that explicitly produce comparison showed that word frequency tells us little about the contexts of use, and hence little about the diverse meanings individual words can take on in dif ferent contexts. Such a bold claim would therefore need more data via context searches or a more elaborate analysis via close reading. Fig. 3: Top 40 most f requent words in the corpus L ooking for Tex t ual Evidence 25 5 Fig. 4: Position of the word “conduct ” in the word count Ral f Schneider, Mar cus Har tner, Anne L apper t25 6 Af ter none of our search terms had turned out to feature prominently among the most frequent words in the corpus, our next step was to turn to what we could find on the list of most frequent words (figs. 3 and 4). For this we went through this list looking for terms we felt to exhibit some kind of relationship to contemporary discussions of social identity. In this way, we found that com­ parably frequently used in our corpus were “honour” (no. 54 in the word-fre­ quency list), “poor” (no.  47), “character” (no.  141) and “conduct” (no.  187; see fig.  4), with the word “conduct”, referring to the overall comportment of a person. From those results, we considered “conduct” to be particularly inter­ esting. The term, appearing most frequently in Wollstonecraf t’s Maria, Or, The Wrongs of Woman (1798) and least frequently in Fielding’s Shamela (1741), is not only eponymous to the eighteenth-century genres of the conduct book and the conduct novel, but generally constitutes a key concept of the literary and cultural movement of sensibility.41, 42 For this reason, we decided to play around some more and searched for the word “conduct” in the sentimental novels of our corpus separately.43 Once more, we received a fairly inconclu­ sive diagram (fig. 5): Between the middle and the end of the eighteenth cen­ tury, sentimental novels feature the term “conduct” in varying ways. While interesting for the formulation of new research questions,44 this did not help us in terms of our thesis on the literary negotiations of social identity. In fact, the visualization suggested that a diachronic change in the 41 V. Nünning, From ‘honour’ to ‘honest’. 42 The low result for Shamela could be interpreted in dif ferent ways. On the one hand, it could mean that this text, being a parody of one of the most inf luential of the early sen- timental novels, wanted to avoid the term by way of taking a critical stance on the genre of the sentimental novel, which was heavily inf luence by the conduct book. On the other hand, Fielding may simply have counted on the reader to realise that both the original and the parody deal with conduct, without having to make that explicit. 43 The sentimental novels or parodies thereof in our corpus are, in chronological order: Samuel Richardson’s Pamela (1740), Henr y Fielding’s Shamela (1741) and Amelia (1751), Lau- rence Sterne’s Tristram Shandy (1759), Oliver Golding’s The Vicar of Wakefield (1766), Henr y Mackenzie’s The Man of Feeling (1771), Tobias Smollet’s Humphrey Clinker (1771), Frances Burney’s Evelina (1778), Maria Edgeworth’s Castle Rackrent (1800) and Jane Austen’s Sense and Sensibility (1811). 44 Such as: Is a separation of a genre and its parodies necessar y, and if so, how can such a distinction be upheld? In how far does the illustration present a visualization of genre negotiations by means of comparison? Is this wave movement even coincidental due to the novels in the corpus? L ooking for Tex t ual Evidence 25 7 Fi g. 5 : O cc ur re nc e o f “ co nd uc t” in se le ct ed se nt im en ta l n ov el s Ral f Schneider, Mar cus Har tner, Anne L apper t25 8 usage of particular words is rather dif ficult to argue for, based on the type of distant reading we engaged in our work with Voyant. Another visualization tool by Voyant of fers users the opportunity to look for the context and the co-occurrence of individual terms in a corpus. Here it became apparent that “conduct”, while mainly appearing as a noun in connection with adjectives that qualify it, also appears as a verb, and does so most frequently in our Gothic novels (figs. 6 and 7). With their tendency to set the action in regions both temporally and spatially remote from eighteenth-century England, the Gothic novels can comment on contemporary English society at best by implication, so that the latter finding pointed once more at the need for fur­ ther close reading and interpretation. Fig. 6: Examples for sentences containing “conduct ” (1) Fig. 7: Examples for sentences containing “conduct ” (2) L ooking for Tex t ual Evidence 25 9 5. Discussion The results of our investigations with Voyant were unexpected to say the least. They are not only at odds with important voices in secondary litera­ ture,45 but they also contradict our own close reading experiences that con­ firm the conceptual relevance of the listed virtues and vices in the portrayal of characters in the eighteenth-century novel in general. Our expectation was to find diachronic developments of the words used to describe presum­ ably middle-class virtues and f laws displaying an increase of frequency towards the end of the eighteenth century. We based our expectations on the assumption that the social identity of the ‘middling’ classes began to be con­ structed in negotiations in and beyond literature during this time period.46 By use of visualization tools we expected to be able to localize the moment these negotiations entered literature on a word level, but instead the results indicate that as far as our searched terms and our corpus are concerned no such change is traceable. Confronted with these findings, we naturally began to question our search strategies, including the list of terms we had thought to be so prominent in eighteenth-century discussions of virtues and vices. But we also wondered whether the infrequent appearance of those terms and the lack of clearly discernible diachronic developments in their application could also be explained dif ferently, for example, by considering the traditional distinction made in literary studies between telling and show- ing.47 Thus, we speculated that our findings may indicate a tendency to show virtues by means of the description of behaviors rather than by naming them explicitly. However, such a claim can only be upheld by a closer analysis in terms of close reading as a complementary method to the usage of DH tools. Further, it seemed that when working with computation techniques, there is the danger that significant dif ferences between texts belonging to the various subgenres of the novel that constitute the overall corpus may dis­ appear from view. While a scholar has certain background information on literary and cultural history available in close reading, a computer is rather 45 E. g., A. Nünning, Der englische Roman. 46 D. Wahrman, Imagining, Schwarz, L. D., Social Class and Social Geography: The Middle Classes in London at the End of the Eighteenth Centur y, in: Social Histor y 7 (1982). 47 Herman, David, Stor y Logic: Problems and Possibilities of Narratolog y, Lincoln: Universit y of Nebraska Press, 2002, 171–172. Ral f Schneider, Mar cus Har tner, Anne L apper t26 0 ignorant towards contextual details in its application of distant reading on a text. This bears problems as well as promises. But we wondered whether these subgenre specific groups such as Gothic novels, or Sentimental novels, had not better be analyzed by searching them separately. The justification for dealing with these separate groups of novels belonging to dif ferent sub­ genres separately lies in literary-historical conventions and definitions of, e. g., the Sentimental Novel, or Gothic Novel. The very fact that DH overlooks such conventions and definitions in the production of data, makes us aware of their potential relevance for analysis and interpretation. In the words of McCarty, DH forces us to “ask in the context of computing what can (and must) be known of our artifacts, how we know what we know about them and how new knowledge is made”.48 Just as in the case of words and particles which explicitly produce comparisons, and with the dif ferent rankings of positive and negative terms, our usage of DH tools challenged us to acknowl­ edge that computing can only ever give us information on texts in form of data. How we read and interpret these numbers and results foregrounds the responsibility of informed research. It is easy to quickly jump to false conclusions if the numbers seem to support the desired argument. But espe­ cially when we combine traditional research with DH methodology taking into consideration all the dif ferent aspects that inf luence the results (e. g., the corpus, the genre, the scope of each text, the relation to other literary works of the same time period, etc.) becomes a dif ficult yet, important task for every scholar in the humanities. For our usage of Voyant this meant we had to realize that even when we received results that seemed to corroborate our assumptions, this did not really mean direct support for our argument in terms of numbers. It only meant that we needed to question these results again in order to avoid run­ ning the risk of prematurely interpreting unanticipated quantitative data in the light of our underlying argument. In the case of the term conduct, for example, we seemed to have found a frequently used word that could sup­ port our argument of the negotiation of social identity in terms of morals in the eighteen-century English novel. Instead, further testing via other visualization tools of fered by Voyant made clear that this seemingly simple link between word frequency and research question of fered a false security 48 McCarty, Willard, Encyclopedia of Librar y and Information Science, New York: Dekker, 2003, 1231. L ooking for Tex t ual Evidence 26 1 (figs. 2, 4, and 5). Looking at the context of the usage of the word “conduct”, we could not support our argument but had to face that the various dif ferent contexts of occurrences of “conduct” varied significantly in meaning. This means that only in a few cases of the many occurrences did “conduct” actu­ ally appear in contexts that we had in mind and that supported our argument (figs.  7 and 8). Another example for the need to treat numeric results with caution was the result of the search term fool. The 1741 novel Shamela by Henry Fielding, which was written as a parody of Samuel Richardson’s highly inf lu­ ential sentimental novel Pamela, showed a peak in the frequency of the word “fool” (551 occurrences) in comparison to the other novels. Strikingly, the sec­ ond highest frequency of the word “fool” was actually found in Richardson’s Pamela, with 175 occurrences. The temptation to construct some intertextual correlation between both texts with regard to their top positions in the word count for “fool” was great: Fielding might have picked up an inherently sig­ nificant feature of Richardson’s novel and exaggerated that for the purposes of satire. However, when we took into account the overall length of the two texts, this argument collapsed: While 551 occurrences of fool seem notewor­ thy in the relatively short novel Shamela (14.456 words), there is nothing sig­ nificant about the term’s appearance in Pamela given the total length of this work. With 227.407 words Richardson’s novel is over fif teen times longer than that of Fielding. Thus, given the massive text of Pamela, the count of 175 occurrences of fool dwindles into comparative insignificance. We were lef t with the paradox that while being considered more precise and accurate in terms of quantitative and statistical occurrences than tra­ ditional methods of close, DH actually seemed to blur any assumption of a precise answer to questions of literary analysis. In our case, DH appeared to be more suitable for finding new questions than to of fer or support conclu­ sive answers to interpretative assumptions. Voyant was able to give us the exact number of word frequencies, to tell us which word appeared how of ten in which novel, and even of fered us to compare these frequencies across the corpus directly, while allowing us at the same time to look for the specific contexts of the words. All of this was very helpful, but mainly to question our own approach and its underlying categories. We set out to look for literary negotiations of social identity and how these were inf luenced by practices of comparison, just to be faced with the problem that comparison was already included in every aspect of our own approach. Instead of making clear dis­ tinctions more apparent, DH made us question these distinctions from the Ral f Schneider, Mar cus Har tner, Anne L apper t26 2 start. If this were the end of it, we would come out of this experiment quite disillusioned. Instead we are inspired by what seems to of fer a new method­ ology for approaching literary texts. While the usage of computing in literary studies is of ten feared to turn literary analysis into a mere equation whose solution would render all further examination of a text vain and shallow, the opposite seems to be true. DH of fers a chance to engage in a more playful, more open-minded yet at the same time equally critical approach to liter­ ature and its study that eventually draws research back to the text and the question of how texts are embedded in various discourses. 6. Conclusion At first sight, our engagement with text search and visualization tools for the analysis of a corpus of eighteenth-century English novels could be sum­ marized in terms of discouragement and frustration  – an experience that appears to be shared by scholars in other DH projects but that is apparently rarely admitted in DH. According to Jasmine Kirby “[w]e don’t talk enough about failure in the digital humanities”.49 Our failure to corroborate some of our assumptions with numerical data, and the necessity to proceed from the observation of word counts to the wider contexts of our findings in fact triggered two insights. First, if the actions and dispositions of humans in social interaction that the eighteenth-century novel negotiates as desirable or undesirable are much less explicitly mentioned than expected, the novel must have other ways of presenting them. Second, the practices of comparing, too, appear to be situated on other levels than that of the text surface, at least in the corpus under scrutiny in our project. The lack of simple numerical proof garnered from distant reading was, in our case, a productive ‘failure’, because it helped us formulate the hypothesis that literary practices of comparing involve the structural juxtaposition of characters in comparable settings and plot segments. As Nina McCurdy and her colleagues have demonstrated, there is some irony in the fact that the more precision a DH tool of fers, the more it makes sense to ‘screw around’ with it to render new interesting and exciting research questions. Narrow research questions in DH of ten of fer a 49 Kirby, Jasmine S., How NOT to Create a Digital Media Scholarship Platform: The Histor y of the Sophie 2.0 Project, in: IASSIST Quarterly 42 (2019), https://doi.org/10.29173/iq926. https://doi.org/10.29173/iq926 L ooking for Tex t ual Evidence 26 3 variety of open, inconclusive results, while ‘screwing around’ seems to lead to unexpected, innovative questions.50 None of this narrows literary research down to a question of sof tware engineering and mathematical bean counting, but rather computation techniques in form of tools of fer a playful exchange between the traditionally trained scholar and DH to find ever new ways of reading texts together in the midst of the “beautif ul mess” that is literature.51 What our search for textual evidence also appeared to show was that available strategies of tagging the words and passages of a text – the produc­ tion of markup  – could much profit from taking into account the research questions of literary scholarship. Existing markup algorithms performed autonomously by computer programs, may be helpful and time-saving, and they certainly have improved much in recent years; still, they rarely capture any of the more content-related questions pertaining to literary analysis, let alone interpretation. What, in the case of our project, really would have helped would have been the automatic isolation and tagging of passages that contain comparisons; this however, is nowhere in sight. We also encoun­ tered problems in the visualization of results, even though our corpus was, in DH terms, very small. How could meaningful illustrations be produced if hundreds, or even thousands, of books were subjected to data-mining? Visu­ alization tools will also have to be further developed to match the research designs of the humanities better. Af ter our venture into DH, we still believe that no computer can ‘find out’ anything about the meaning of a text on its own. Therefore, while the scholar’s limitations are quantitative, those of computer programs appear to lie in the quality of their findings. Nor will a text be ‘readable’ to a computer at all, if it has not been previously read, processed and digitized by humans, increasingly automatized programs of parsing and tagging notwithstanding. The solution to the apparent incompatibility of close and distant reading lies, unsurprisingly, in the fact that the two strategies can, and ought to be, regarded as complemen­ tary rather than competitive, as Stephen Ramsay, among others, has argued.52 As we have shown, to make use of DH methods can help literary scholars to focus and re-formulate their questions and research strategies, and to recon­ sider their assumptions about what literary texts do and how they do it. 50 N. McCurdy et al., Poemage, 447. 51 Ibid., 445. 52 S. Ramsay, The Hermeneutics. Ral f Schneider, Mar cus Har tner, Anne L apper t26 4 Appendix 1: Extended corpus of English eighteenth-century novels 1688 Oroonoko Aphra Behn 1713 The Amours of Bosvil and Galesia Jane Barker 1714 Exilius Jane Barker 1719 Robinson Crusoe Daniel Defoe 1720 Memoirs of a Cavalier Daniel Defoe 1722 Moll Flanders Daniel Defoe 1723 The Lining of the Patch Work Screen Jane Barker 1724 John Sheppard Daniel Defoe 1726 Gulliver’s Travels Jonathan Swif t 1740 Pamela Samuel Richardson 1741 Shamela Henry Fielding 1743 Jonathan Wild Henry Fielding 1748 Roderick Random Tobias Smollett 1749 Fanny Hill John Cleland 1749 Tom Jones Henry Fielding 1750 Harriot Stuart Charlotte Lennox 1751 Amelia Henry Fielding 1751 Betsy Thoughtless Eliza Fowler Haywood 1751 Peter Wilkins Robert Paltock 1752 The Female Quixote Charlotte Lennox 1759 Rasselas Samuel Johnson 1759 Tristram Shandy Laurence Sterne 1760 The Adventures of Sir Launcelot Greaves Tobias Smollett 1762 Millenium Hall Sarah Scott 1764 Castle of Otranto Horace Walpole 1766 The Vicar of Wakefield Oliver Goldsmith 1768 Maria; Or, The Wrongs of Woman Mary Wollstonecraf t 1769 Emily Montague Frances Brooke 1771 Humphrey Clinker Tobias Smollett 1771 The Man of Feeling Henry Mackenzie 1778 Evelina Frances Burney 1778 The Old English Baron Clara Reeve 1782 Cecilia Fanny Burney 1784 Imogen William Godwin 1786 The Heroine Eaton Stannard Barrett L ooking for Tex t ual Evidence 26 5 1786 Vathek – An Arabic Tale William Beckford 1788 Mar y – A Fiction Mary Wollstonecraf t 1789 Castles of Athlin and Dunbayne Ann Radclif fe 1790 A Sicilian Romance Ann Radclif fe 1791 A Simple Stor y Elizabeth Inchbald 1791 Charlotte Temple Susanna Rowson 1791 Romance of the Forest Ann Radclif fe 1793 The Castle of Wolfenbach Eliza Parsons 1794 Caleb Williams William Godwin 1794 The Mysteries of Udolpho Ann Radclif fe 1796 Memoirs of Emma Courtney Mary Hays 1796 The Monk Matthew Lewis 1798 Wieland Charles Brockden Brown 1799 St Leon William Godwin 1800 Castle Rackrent Maria Edgeworth 1806 Leonora Maria Edgeworth 1806 Wild Irish Girl Sydney Owenson 1806 Zof loya Charlotte Dacre 1811 Sense and Sensibilit y Jane Austen 1812 The Absentee Maria Edgeworth Bibliography Aristotle, The Politics, trans. Carnes Lord, Chicago: University of Chicago Press, 1984. Backscheider, Paula R./Ingrassia, Catherine (eds.), A Companion to the Eigh­ teenth-Century English Novel and Culture, Chichester: Wiley-Blackwell, 2006. Burrows, John, Delta: A Measure for Stylistic Dif ference and a Guide to Likely Authorship, in: Literary and Linguistic Computing 17 (2002), 267–287. Cannadine, David, Class in Britain, London: Penguin, 2000. Corfield, Penelope J., Class by Name and Number in Eighteenth-Century England, in: History 72 (1987), 38–61. Cowan, Brian, Making Publics and Making Novels: Post-Habermasian Perspectives, in: J. A. Downie (ed.), The Oxford Handbook of the Eigh­ teenth-Century Novel, Oxford: Oxford University Press, 2016, 55–70. Ral f Schneider, Mar cus Har tner, Anne L apper t26 6 Defoe, Daniel, Robinson Crusoe, ed. Michael Shinagel, New York: Norton, 1994 [1719]. Downie, J. A. (ed.), The Oxford Handbook of the Eighteenth-Century Novel, Oxford: Oxford University Press, 2016. Eder, Jens/Jannidis, Fotis/Schneider, Ralf (eds.), Characters in Fictional Worlds: Understanding Imaginary Beings in Literature, Film, and Other Media, New York: de Gruyter, 2010. Ensslin, Astrid, Hypertextuality, in: Marie-Laure Ryan/Lori Emerson/Benja­ min J. Robertson (eds.), The Johns Hopkins Guide to Digital Media, Balti­ more: Johns Hopkins University Press, 2014, 258–265. Fludernik, Monika, An Introduction to Narratology, London: Routledge, 2009. French, Henr y, Gentlemen: Remaking the English Ruling Class, in: Keith Wrightson (ed.), A Social History of England: 1500–1750, Cambridge: Cambridge University Press, 2017, 269–89. Goldsmith, Oliver, The Vicar of Wakefield, Oxford: Oxford University Press, 2006 [1766]. Habermas, Jürgen, The Structural Transformation of the Public Sphere: An Inquiry into a Category of Bourgeois Society, trans. Thomas Burger, Cambridge: Polity, 2015 [1962]. Herman, David, Story Logic: Problems and Possibilities of Narratology, Lin­ coln: University of Nebraska Press, 2002. Hume, David, Of the Middle Station of Life, in: Thomas H. Green/Thomas H. Grose (eds.), David Hume, The Philosophical Works, Aalen: Scientia Ver­ lag, 1964 [1742], 375–380. Hunter, Paul J., The Novel and Social/Cultural History, in: John Richetti (ed.), The Cambridge Companion to the Eighteenth-Century Novel, Cam­ bridge: Cambridge University Press, 1996, 9–40. Jannidis, Fotis, Quantitative Analyse literarischer Texte am Beispiel des Topic Modelings, in: Der Deutschunterricht 5 (2016), 24–35. Jannidis, Fot is/Lauer, Gerhard, Burrows’s Delta and Its Use in German Liter­ ary History, in: Matt Erlin/Lynne Tatlock (eds.), Distant Readings: Topol­ ogies of German Culture in the Long Nineteenth Century, Rochester: Camden House, 2014, 29–54. Juola, Patrick, Authorship Attribution, in: Foundations and Trends in Infor­ mation Retrieval 3 (2006), 233–334, http://dx.doi.org/10.1561/1500000005. http://dx.doi.org/10.1561/1500000005 L ooking for Tex t ual Evidence 26 7 Kirby, Jasmine S., How NOT to Create a Digital Media Scholarship Platform: The History of the Sophie 2.0 Project, in: IASSIST Quarterly 42 (2019), https://doi.org/10.29173/iq926. Margolin, Uri, Character, in: David Herman (ed.), The Cambridge Companion to Narrative, Cambridge: Cambridge University Press, 2007, 66–79. McCart y, Willard, Encyclopedia of Library and Information Science, New York: Dekker, 2003. McCurdy, Nina et al., Poemage: Visualizing the Sonic Topology of a Poem, in: IEEE Transactions on Visualization and Computer Graphics 22 (2016), 439–448. Moretti, Franco, Distant Reading, London: Verso, 2013. Muldrew, Craig, The ‘Middling Sort’: An Emergent Cultural Identity, in: Keith Wrightson (ed.), A Social History of England: 1500–1750, Cambridge: Cambridge University Press, 2017, 290–309. Nünning, Ansgar and Vera, Englische Literatur des 18.  Jahrhunderts, Stutt­ gart: Klett, 1998. Nünning, Ansgar, Der englische Roman des 18.  Jahrhunderts aus kulturwis­ senschaf tlicher Sicht. Themenselektion, Erzählformen, Romangenres und Mentalitäten, in: Ansgar Nünning (ed.), Eine andere Geschichte der englischen Literatur. Epochen, Gattungen und Teilgebiete im Überblick, Trier: WVT, 1996, 77–106. Nünning, Ansgar, Grundzüge eines kommunikationstheoretischen Modells der erzählerischen Vermittlung: Die Funktion der Erzählinstanz in den Romanen George Eliots, Trier: WVT, 1989. Nünning, Vera, From ‘honour’ to ‘honest’. The Invention of the (Superiority of ) the Middling Ranks in Eighteenth Century England, in: Journal for the Study of British Cultures 2 (1994), 19–41. Ramsay, Stephen, The Hermeneutics of Screwing Around; or What You Do with a Million Books, in: Kevin Kee (ed.), Pastplay: Teaching and Learn­ ing History with Technology, Ann Arbor: University of Michigan Press, 2014 [2010], 111–120. Richetti, John (ed.), The Cambridge Companion to the Eighteenth-Century Novel, Cambridge: Cambridge University Press, 1996. Rogers, Pat, Social Structure, Class, and Gender, 1660–1770, in: J. A. Downie (ed.), The Oxford Handbook of the Eighteenth-Century Novel, Oxford: Oxford University Press, 2016, 39–54. https://doi.org/10.29173/iq926 Ral f Schneider, Mar cus Har tner, Anne L apper t26 8 Ryan, Marie-Laure, Avatars of Story, Minneapolis: University of Minneapolis Press, 2006. Schwarz, L. D., Social Class and Social Geography: The Middle Classes in London at the End of the Eighteenth Century, in: Social History 7 (1982), 167–185. Shillingsburg, Peter L., From Gutenberg to Google: Electronic Representations of Literary Texts, Cambridge: Cambridge University Press, 2006. Thaller, Manf red, Geschichte der Digital Humanities, in: Fotis Jannidis/ Hubertus Kohle/Malte Rehbein (eds.), Digital Humanities. Eine Ein­ führung, Stuttgart: J. B. Metzler, 2017, 3–12. Thompson, E. P., Customs in Common: Studies in Traditional Popular Cul­ ture, New York: New Press, 1993. Thornton, William, The Counterpoise: Being Thoughts on a Militia and a Standing Army, London: Printed for M. Cooper, 1752. Tillyard, E. M. W., The Elizabethan World Picture: A Study of the Idea of Order in the Age of Shakespeare, London: Chatto & Windus, 1967 [1942]. Wahrman, Dror, Imagining the Middle Class: The Political Representation of Class in Britain, c.  1780–1840, Cambridge: Cambridge University Press, 1995. Watt, Ian, The Rise of the Novel: Studies in Defoe, Richardson, and Fielding, Berkeley: University of California Press, 1957. Wenzel, Peter (ed.), Einführung in die Erzähltextanalyse: Kategorien, Modelle, Probleme, Trier: WVT, 2004. Yenor, Scott, David Hume’s Humanity: The Philosophy of Common Life and Its Limits, Basingstoke: Palgrave, 2016.