The Riddle of Literary Quality: A Computational Approach

van Dalen-Oskam, Karina Book The Riddle of Literary Quality: A Computational Approach Provided in Cooperation with: Amsterdam University Press (AUP) Suggested Citation: van Dalen-Oskam, Karina (2023) : The Riddle of Literary Quality: A Computational Approach, ISBN 9789048558148, Amsterdam University Press, Amsterdam, https://doi.org/10.5117/9789048558148 This Version is available at: https://hdl.handle.net/10419/281237 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ the riddle of literary quality karina van dalen-oskam Karina van Dalen-oskam Literary quality of the riddle A ComputAtionAl ApproACh 9 789048 558148 AUP.nl hat is literature? Can we measure ‘literariness’ in texts themselves? The innovative Computational Humanities project ‘The Riddle of Literary Quality’ asked thousands of Dutch readers for their opinion about contemporary Dutch and translated novels. The public shared which novels they had read, what they really thought of them, and how they judged their quality. Their judgments of the same novels were compared with the results of computational analysis of the books. Using evidence from almost 14,000 readers and building on more textual data than ever before, Van Dalen-Oskam and her team uncovered unconscious biases that shed new light on prejudices many people assumed no longer existed. This monograph explains in an accessible way how the project unfolded, which methods were used, and how the results may change the future of Literary Studies. Karina van Dalen-Oskam is head of the Computational Literary Studies research group at Huygens Institute for the History and Culture of the Netherlands, an institute of the Royal Netherlands Academy of Arts and Sciences (KNAW). She is also professor in Computational Literary Studies at the University of Amsterdam, and Honorary Professor in the Humanities at the University of Wolverhampton (UK). 9789048558148 Hardback.indd 1 31-05-23 15:36 The Riddle of Literary Quality The Riddle of Literary Quality A Computational Approach Karina van Dalen-Oskam Amsterdam University Press Originally published as: K. van Dalen-Oskam, Het raadsel literatuur: is literaire kwaliteit meetbaar? Amsterdam University Press, 2021 Translated and adapted from the Dutch by the author. Cover design: Margreet van de Burgt Lay-out: Crius Group, Hulshout isbn 978 90 4855 814 8 e-isbn 978 90 4855 815 5 doi 10.5117/9789048558148 nur 621 Creative Commons License CC BY NC ND (http://creativecommons.org/licenses/by-nc-nd/4.0) K. van Dalen-Oskam / Amsterdam University Press B.V., Amsterdam 2023 Some rights reserved. Without limiting the rights under copyright reserved above, any part of this book may be reproduced, stored in or introduced into a retrieval system, or transmitted, in any form or by any means (electronic, mechanical, photocopying, recording or otherwise). Contents 1. The Riddle 9 Introduction The Riddle of Literary Quality 9 From Sociology of Art to Neo-Darwinism 13 Computational Literary Studies 18 Empirical Study of Literature 27 Conclusion 28 References 29 2. The National Reader Survey 32 Books, Readers, Opinions The Survey 32 The Books 35 The Readers 37 Categories of Books 42 Winners and Losers 44 Motivations 51 Contrasting Opinions 55 Conclusion 60 References 61 3. Romance, Suspense, and Translations 63 Mixed Methods in More Detail From Readers’ Opinions to Text Analysis 63 Romance and the Fifty Shades Trilogy 66 Crossing the Borders of Suspense 77 The Riddle of Translation Quality 80 Translations in the Periphery 89 Conclusion 91 References 92 4. Literary Novels Written by Women 95 Donoghue and Dorrestein The Glass Ceiling 95 Translations Compared 100 6 The RiddLe of LiTeRaRy QuaLiTy Donoghue, Kamer (Room) 103 The Dutch Perspective 106 Dorrestein in More Detail 115 Dutch Originals Opposed 121 Conclusion 122 References 123 5. Literary Novels Written by Men 124 From Barnes to Grunberg Barnes and the Booker 124 Translations Opposed 127 Dutch Originals Compared 132 A Literary Duel: Kluun versus Koch 141 Grunberg and Romance 148 Conclusion 154 References 156 6. Style, Gender, and Genre 158 Combining Multiple Perspectives Beyond Words 158 Syntactical Analysis 159 Parts of Speech 160 Cliché Expressions 161 Themes or Topics 162 Predicting Scores for Literary Quality 163 Vector Space Explorations 166 Sentence Length 168 Comparing Dutch Literary Novels by Female and Male Authors 170 Author Gender and Perceptions of Literary Quality 177 The Feedback Loop of Genre 181 Conclusion 184 References 186 7. The Riddle of Literary Quality Solved? 188 Reflection and Conclusion References 198 ConTenTs 7 Appendix 1: The Survey 200 Appendix 2: The Books 204 Appendix 3: The Website 214 Bibliography 215 Tables 222 Figures 223 Acknowledgements 225 Index 227 14 The RiddLe of LiTeRaRy QuaLiTy researchers, and teachers. A work appreciated by this group receives symbolic capital – not money, at least usually not enough in smaller language areas such as Dutch (although there are exceptions), but literary standing and prestige. Authors and publishers who position themselves in this part of the field do not opt for short-term security, but hope for long-term success: an ‘ever seller’ and perhaps inclusion in the literary canon. That may mean sufficient income for publishers, but usually not for the authors themselves. Bourdieu thus pointed out a number of fundamental oppositions. On the one hand there is the subfield of large-scale production, on the other hand that of limited production. Economic capital versus symbolic capital. Works that temporarily attract a lot of attention, bestsellers, versus timeless classics, ‘real’ literature. Authors of bestsellers earn lots of money. Authors of works of high literary quality are, by definition, struggling to make a living. But those who write for eternity do accumulate prestige and literary status, for example by being awarded literary prizes. The loser in economic terms is the winner in the race for literary prestige (1993, 39). The economy of the subfield of limited production was characterized by Bourdieu as an anti-economy (1993, 54). He stated that economic success is even considered suspect. If a book sells sensationally well, people unconsciously deem it unlikely to be ‘high literature’, Bourdieu stressed. In the past, this assumption led to the exclusion of popular literature from scientific research, which only paid attention to the literary canon (1993, 39) – as Karel van het Reve noted in his famous talk in Leiden to which I referred earlier in this chapter. The subfield of large-scale production, that is, of bestsellers, is conservative, as Bourdieu called it. Authors and publishers who focus on this subfield are very much aware of what the public has already appreciated before. They try to deliver what fits the existing taste of the public. This does not mean that nothing ever changes, but innovations ultimately arise from developments at the opposite pole, the subfield of limited production. There, publishers and authors who strive for literary prestige actually try to break through existing traditions and consciously come up with new things. A younger generation tries to distinguish itself from the older generation in this way. The avant-garde does the risky experimenting and only those elements that have been proven to appeal to the wider public are taken over in the pole of large-scale production. This, in turn, causes the avant-garde to want to distance itself from this in a kind of push and pull movement (1993, 58). The consequence is that players at both poles – but for opposite reasons – constantly have to slightly adjust their strategy in order to obtain the success they are looking for, be it money or status, economic or symbolic capital. It is clear why Bourdieu never commented on what this means in concrete terms for the texts themselves. His theory describes a process and not its tangible, in this case linguistic, output. In Bourdieu’s mind, the form of the texts entirely depends The RiddLe 15 on the context, and since the context is constantly changing, it is not surprising that the texts themselves are also constantly in flux. Moreover, as Bourdieu stated in the foreword to Distinction (1984/2010), the English translation of his book La distinction (1979) in reply to criticism, the precise workings of the sociological process can be slightly different in different countries and time periods. What mattered to him, however, was the underlying principle. Moreover, Bourdieu stated that it is of main importance for authors to publish their work in the appropriate place to be considered of literary importance. They need to be visible in the right magazine and need to be published by the right publisher (1993, 95). Awards are important, but here too, the work must be awarded the correct prize. An award that would delight an author from the other pole may have a very negative impact on status and prestige on the other (1993, 100). The manuscripts that a publisher receives can be seen as the result of a preselection made by the submitting authors themselves. Authors have written their work keeping in mind which subfield they want to play a role in, and they approach those publishers who they assume to be active in that subfield. The same applies to critics. They receive books for review that have in turn been pre-selected by the publisher who published them. Review copies are sent to those who the publisher expects to pay (positive) attention to the titles, based on their previous reviews (1993, 133–135). Everything is connected to everything else. Opinions may seem very personal, but are in fact always collective judgements that are directly or indirectly related to positions in the literary field and to social class. Social conventions, in other words, are the basis for literary conventions. They are what perceptions of literary quality are embedded in. Bourdieu maintained that the taste of readers depends primarily on the reader’s level of education and after that, on their social class (the focus of Bourdieu 1984/2010). Bourdieu observed that especially readers who do not write themselves, but who do have a higher education and are members of the intellectual elite, allow themselves to be influenced by the unwritten rules of the literary field (1993, 41). He did not go into how readers develop their literary taste, however. This is where other sociologists picked up and developed what they called ‘the new sociology of art’. They brought back into the picture the form and shape of the artworks themselves. Since this corresponds to how in the project The Riddle of Literary Quality we wanted to investigate what constitutes literary quality by looking at the texts themselves, their insights may be useful in this quest. In 2007, sociologist Antoine Hennion theorized how ‘amateurs’, non-specialists, develop their taste and how they arrive at a value judgement. He illustrated his theory with the description of the ways in which different mountaineers try to reach the same summit. They have similar resources at their disposal, but not the same physical condition and climbing experience. Climbers therefore choose 16 The RiddLe of LiTeRaRy QuaLiTy different routes to reach the top. The mountain is not a social construction, Hennion emphasized, but it does have the potential to make differences between people visible. One can climb a mountain via all sorts of routes and still end up at the same summit, sharing the same view. The same applies to taste. People can have the same taste and the same value judgement. But that doesn’t mean that the judgement has come about in the same way. They may have developed their taste in different ways. Taste is not the result of unconscious and passive sociological processes, but is actively acquired by each individual. Taste is therefore a choice, Hennion stated. The preferences one has developed in the past determine who one is. What those preferences are depends on one’s family, education, and the social environment in which one has functioned. In forming our opinions, almost all of us look at what others think and how they arrived at that judgement before we form our own. The sociological processes never completely disappear from view. But where Bourdieu looked at macro-processes in which taste was mainly seen as a characteristic of social class, Hennion concentrated on the analysis of processes at a micro-level, such as the ways in which art is advertised and how judgements about art are made. Seeking to connect the macro (Bourdieu) and micro (Hennion) approaches, sociologist Ori Schwarz suggested in 2013 that new research should focus on ‘tasting techniques’ – ways in which taste is developed. Schwarz believed that mastering certain techniques in order to arrive at a value judgement can indeed be related to a person’s social background and education. People from different social groups can thus have at their disposal different techniques for determining the value of a work of art – and ultimately arrive at the same judgement. Value judgements, in that sense, are not reducible to the influence of other people and their subjective experiences and preferences, but rather to the mastery of certain things, techniques for arriving at a judgement. At the same time, Schwarz argued, the ‘right’ way to consume art can be seen as a status symbol and an indication of one’s identity – which agrees with Bourdieu’s perspective. Schwarz raised several issues that were ignored by Hennion. Firstly, some genres are valued significantly more than others, and this high level of appreciation is usually linked to an emphatic aversion to other genres. Often, we see a clear relationship between a genre and a particular age group or social class – there will be very little overlap between groups that appreciate classical music and heavy metal, Schwarz assumed. Furthermore, it is precisely the differences in the ways individuals develop their taste that can be related to social structures. Hennion’s ‘amateurs’ clearly had characteristics of certain social classes. And thirdly, many people will have experienced how they themselves or others were excluded on the basis of their taste or the way in which they arrive at a value judgement about a work of art. Consciously or unconsciously, we judge someone’s professional capacity The RiddLe 17 or social suitability on the basis of their way of life, taste, or mode of judgement. The nature of attention to cultural products is thus partly determined by social stratification, but conversely there is also an influence on the social structure. This influence is visible not only when it comes to identifying suitable candidates for a particular job, but also, for example, when it comes to finding friends or a life partner. That last remark was prompted by psychological research with a (Neo-)Darwinian angle. Schwarz referred to a book published in 2010 by psychologist and linguist Paul Bloom, How Pleasure Works: The New Science of Why We Like What We Like. Bloom observed that people often attribute the reason for appreciation, pleasure, or whatever we want to call it, to properties of that which is enjoyed (sex, art, play) or consumed (food). Translating that to our research on literary quality, readers often seek the reason for the literary value they attribute to a text in the text itself. Bloom was convinced that this is only part of the story. He believed that enjoyment is influenced by other factors. It all starts with what we like or don’t like to eat. This can be predicted if it is known what part of the world someone originates from. A predilection for something has to be learned. And sometimes people say they like something because it indicates, for example, that they have enough money to afford that food. In other words: taste as a status symbol. Can we indeed compare reading literature to eating unpopular vegetables such as broccoli or Brussels sprouts? As an acquired taste? It takes a while to get used to the taste. And some people will never like them, but for the greater good – their own child’s education, for example – they are willing to pretend every now and then that they absolutely love them. Bloom did not go into this, however. One possibility he did mention is that when people look at art (or, for that matter, read literature), they consider how difficult it is to make. The more effort involved in the creation, the greater the appreciation. But even that is not enough. More important, and according to Bloom even the essence, is that we think that the effort put into a work of art is only successful if it is done by a creative genius. Because that is what we have learned about artists. So, paintings by Vermeer are highly valued art, but those by Vermeer imitator Van Meegeren are not, even though he might have spent just as much time on them. On this point, Bloom disagreed with Denis Dutton, who described the role art may have played in the evolution of mankind in the 2009 book The Art Instinct: Beauty, Pleasure, & Human evolution. According to Dutton, art is part of the process of sexual selection. He assumed that artistic men (yes, men…) are attractive partners for women and that, for this reason, creative genes are spread throughout the human population. It is difficult to create good art, Dutton argued, so an artist shows that he is clever, inventive, and can plan well. He is therefore a good partner. Bloom thought this short-sighted: it explains the attraction of the artist (the person) and not of the art (the object). He did not want to limit the sense of art to sexual selection from a 18 The RiddLe of LiTeRaRy QuaLiTy Darwinian perspective. After all, we are not constantly looking for a (genetically) suitable life partner; there are many other relationships that are important. We are also looking for friends, allies, and leaders. And so looking at what someone thinks about art can help in finding the right person for the right role. To translate this to our subject: a shared literary taste can be a way of connecting with others. Take a look at my bookcase, and you will know how likely it is that we will get along or that there is work to be done. Literary taste as a social lubricant. But that still leaves the question of why one novel is ‘allowed’ on one’s bookshelves and another is not. To what extent is symbolic capital – prestige – responsible for one’s literary taste and to what extent is the attributed value inspired by how the text is shaped in language? In the project The Riddle of Literary Quality we harnessed new approaches developed in computational literary studies to answer this question. Computational Literary Studies The quantitative approach to text analysis firmly sits in the discipline of what in Europe has recently become known as computational literary studies. The field as a practice and as a community has been around much longer, with labels such as ‘digital literary studies’, ‘stylometry’, or ‘distant reading’. The first use of ‘computational literary studies’ can be traced back to my professorial chair in computational literary studies at the University of Amsterdam, which was created in 2012. At the start the name was not used yet, nor in the first phase of the important COST Action Distant Reading for European Literary History (CA16204) which ran from November2017 until May2022 and was funded by the European Commission. It was introduced, however, in 2020 when the German priority programme SPP 2207 Computational Literary Studies was funded by the German Research Foundation (DFG). This priority programme aims to develop and establish innovative computational methods in the discipline of literary studies. The first phase of this programme funded eleven research projects. At the time of writing, the second phase of the programme was close to being launched. Another new project, developed as a follow-up to the COST Action Distant Reading, was named Computational Literary Studies Infrastructure. This project received funding from the European Commission in the Horizon 2020 framework infraia-02-2020 (Integrating Activities for Starting Communities). It runs from March2021 until March2025. As stated on the website of the European Commission, the aim of this project is to ‘help build the shared and sustainable infrastructure needed to undertake literary studies in the digital age (…).The resulting improvements will benefit researchers by bridging gaps between greater and lesser-resourced communities in computational literary studies and beyond, ultimately offering opportunities to create new research and insight The RiddLe 19 into our shared and varied European cultural heritage’ (‘Computational Literary Studies Infrastructure’). And in 2022 the first issue of the international, open access and peer-reviewed Journal of Computational Literary Studies (JCLS) was released, advocating itself on its website as responding to ‘the increasing differentiation of subfields within the Digital Humanities, an ongoing process in which Computational Literary Studies has already gained considerable maturity and visibility.’ The project The Riddle of Literary Quality is rooted in new developments in stylometry starting with a measure called Burrows’ Delta Procedure for authorship attribution and in the subsequent turn towards questions beyond authorship. Around the turn of the century, authorship attribution using computational means was a prominent topic in what was then called Humanities Computing. The assumption underlying this research is that comparison of a disputed text with a large corpus of texts written by authors who may be candidates for the authorship may help to identify the person who most likely was the real author of the disputed text. The corpus needs to consist of texts written in the same period of time as the disputed text and it must be certain who wrote them. Various methods can be used to examine the corpus. One of the most successful is the measure we now call Delta or Burrows’ Delta, after the Australian literary scholar John Burrows (1928–2019), who developed this method and first published about it in 2002. As I make extensive use of the Delta Procedure in this monograph, I will explain it in more detail than other methods that will be mentioned later in this chapter. In Burrows’ original implementation, the Delta procedure builds a frequency list of the 150 most frequent words for each text in the corpus and for the selected corpus as a whole, listing the relative frequencies for each of the selected words. It then compares the frequency of each word in each text to the relative frequency of that word across the entire text collection. For all words, the deviations of the frequency from the mean are added up, resulting in a number for each text, the ‘Delta’. That number expresses how much the total of all 150 most frequent words in a text differs in terms of frequencies – the number of times a word occurs in the text – from the occurrence of those words in the entire text collection. The author of the text with the Delta score closest to the score of the disputed work is most likely the author who should also be held responsible for that text. Many experiments using texts of which the authorship was undisputed have shown that the Delta procedure works very well indeed. Using a much longer list of the most frequent words works even better, as David Hoover has shown in several contributions (e.g. Hoover 2004). There are some limitations, however. The size of the used texts does matter; the preferred minimum size of texts ranges between one thousand to five thousand words (tokens) (cf. Eder 2015). As in other authorship attribution methods, the Delta score only indicates the degree of difference in word use and word frequencies for the texts that are part of the actual corpus, which 20 The RiddLe of LiTeRaRy QuaLiTy means that the researcher’s observations are only valid for that specific corpus. The procedure always yields a text that most closely resembles the disputed text, also if the real author is not represented in the corpus. Another issue to remember is that the method does not work if the disputed text is an author’s very first publication, because in this case, there is simply no work available with which the disputed text can be compared. And finally, the procedure usually is only trustworthy if the corpus consists of texts belonging to the same genre. Experiments have shown that authorship attribution methods yield the best results if all texts in the corpus belong to the same genre, for example only literary novels, or only plays, and not a combination of both (cf. Burrows 279). This suggests that there are genre-dependent conventions in language usage that are responsible for differences in word frequency patterns. With the launch of his Delta procedure, Burrows set in motion a whole new authorship attribution ‘craze’ in the digital humanities. Coming from a background in literary studies, he especially influenced literary scholars to use his method and further develop the procedure. The main reason why in computational literary studies Delta became such a huge success (next to its success rate in correctly identifying authors) is that it is transparent: it deals with words, interpretable linguistic units, and these words and their frequencies are relatively easy to inspect and as such help the researcher to get a closer look at the data, to explore the texts in more detail and to uncover the directions in which to continue the research on the way to a satisfying and fruitful interpretation. Delta helps us to creatively think about our textual data and how they function. Delta was implemented in the Stylo Package for R developed by Maciej Eder, Jan Rybicki, and Mike Kestemont and in this form reached a whole new audience of scholars who benefitted from the graphical user interface that made using Stylo and applying Delta a lot easier. Eder and the other members of the Computational Stylistics Group have been teaching how to use this package all over the world for the last decade or more. Their website lists the ongoing work and the many publications related to Stylo. I have benefitted greatly from their work, and a lot of my own research as presented in this book is done with the Stylo package; I will go into more detail about the different measures I applied in the next chapters of this book. What Delta has taught us is that by using digital text corpora and statistical measures we actually can recognize patterns in language use that are usually invisible to a human reader. Surprisingly, authors may be recognized by their word frequencies. And the biggest surprise is that it is mainly the very common words that betray the identity of an author. Words that appear in every text and in the works of every author. Articles, conjunctions, prepositions: no text goes without them. If we instruct software to count all the words in a text and compare the results with the number of occurrences of the same words in other texts, it turns The RiddLe 21 out that there are subtle differences in how often different authors use the same very frequent words. For example, one author chooses an article just a little more often than others. Another author uses more adjectives compared to other writers. Those differences may, of course, indirectly depend upon the topics authors deal with; they could also result from deliberately chosen and divergent approaches. We cannot yet determine exactly how it works, because there are so many different variables involved. For the time being, however, most researchers assume that the differences can partly be traced back to the unique development of the language skills of each individual author, as David Hoover, one of the scholars who has done a lot of work enhancing the Delta procedure, writes in his most recent book Modes of Composition and the Durability of Style in Literature (3–8). A logical next step after identification of individual authors was to apply authorship attribution models such as Delta to find out more about collaboration between authors. I will mention some examples from different languages, genres, and time periods. Joris van Zundert and I analysed a Medieval Dutch Arthurian novel that was supposedly left unfinished by the first author and was completed by another. We used a walking window version of Delta to find out where the second author took over from the first. Mike Kestemont, Sara Moens, and Jeroen Deploige explored how Hildegard von Bingen and her different amanuenses, all writing in Latin, collaborated. I looked into the collaboration of two eighteenth-century Dutch female authors who collaborated on four epistolary novels to see if critics denying one of the two any literary talent were correct or if we should give precedence to what the two women themselves stated again and again, that they collaborated on an equal footing. I also considered if the two authors created a character voice for each of their letter writing characters or not. Lisa Pearl, Kristine Lu, and Anousheh Haghighi followed up on this research with their computational analysis of character voice in Samuel Richardson’s epistolary novel Clarissa. Another interesting study on assumed collaboration was published by Rachel McCarthy and James O’Sullivan, testing an assumption of some readers that Emily Brontë’s Wuthering Heights must have been co-authored by her brother Branwell. For the modern time period, Jan Rybicki, David Hoover, and Mike Kestemont, for example, focused on works co-authored by Joseph Conrad and Ford Madox Ford trying to find out if Ford was indeed the author of a sizeable fragment ofNostromo, as he claimed. Jan Rybicki and Magda Heydel analysed the nature of a translation into Polish of a novel by Virginia Woolf that was started by one translator and finished by another. The new methods for authorship attribution as sketched above also inspired scholars to ask questions that go beyond authorship. An example is the analysis of the chronological development of one author’s style using the same measures as for authorship attribution (Hoover 2021, 122–142). One of the first researchers of 22 The RiddLe of LiTeRaRy QuaLiTy literature to look beyond verification of authorship was Matthew L. Jockers. In 2013 he published Macroanalysis, a monograph in which he reported on his computational research into a corpus of around 3,500 English-language novels, all first published in the nineteenth century. Jockers was interested in measuring such things as genre, origin, and influence. Can certain differences in style be traced back to, for example, the time period in which a work was written, or to the genre to which it can be said to belong? Is the geographical origin of the author – in addition to American works Jockers also included novels by English and Irish authors in his research – recognizable in the writing style? Do other external factors play a role in the design of a novel? And can we perhaps even find out which authors most influenced others, looking only at the digitized texts themselves? Jockers found that the overall style of the novels did not change very much throughout the nineteenth century. What did change over time was the presence of certain kinds of novels, genres that are recognizable by, among other things, shared topics and an associated shared style. Jockers concluded that genres develop and disappear over a period of about thirty years, which agrees with what other researchers have found (84). He also established that some authors have a much more varied style in their different works than others. He noted that the style of Jane Austen, for example, was very easy to recognize with computational means. All her novels are similar in terms of the use of the most frequent words; the differences only show in the topics of the narratives, which are reflected by less frequent content words. In contrast, Mark Twain and James Joyce offered such a wide variety of styles that the software had the greatest difficulty (so to speak) in recognizing these authors (93). Jockers also used the software he developed to predict the author’s gender based on the texts themselves. This proved to be correct in eighty per cent of the cases. He was particularly interested in the 20 per cent where things go wrong (93–95). All in all, Jockers’ methods worked best in recognizing differences in authorship. Then, in descending order of correct classification, followed genre, decade of publication, and finally gender of the author (97). Jockers also tried to delve deeper in terms of content. Can software help to discover themes in the novels? Which topics are characteristic of certain genres or authors or countries? For this he used a technique called topic modelling, which determines for the selected text collection which words often occur in close proximity (more about topic modelling in Blei 2012). The software produces a number of word lists, to which the researcher then has to apply a label that describes the subject connecting the words. An example from Jockers’ topic modelling is the list ‘heart, cabin, bog, county, mountains, Derry, house, air, evening, turf, friends, heaven’. Jockers labelled this list of co-occurring words as the topic ‘Ireland’ (124). With these lists of words, the proportion of the topics was measured in all novels. The results showed how they are distributed across The RiddLe 23 different authors, genres, time periods, origin, or gender of the author. In his nineteenth-century English texts Jockers found, for example, that the theme ‘Villains and Traitors’ occurred more often in novels by male than by female authors, just like ‘Quarrels and Duelling’ and ‘Enemies’. Female authors paid more attention to ‘Affection and Happiness’ and ‘Female Fashion’. The topic of ‘American Slavery’ appeared most often in novels from the United States and less so in those from Great Britain and Ireland, although the slavery discussion there was equally intense. This became apparent when Jockers compared the results with those of the theme ‘Native Americans’, where the imbalance was much greater between novels from the United States on the one hand and those from Ireland and Great Britain on the other (136–141). In 2016, together with Jodie Archer, Jockers published the book The Bestseller Code. Archer and Jockers now focused on modern novels, written in English or translated into English. More than fifty thousand novels are published in the United States every year. Around two hundred titles from that annual crop end up on the prestigious New York Times bestseller list, less than half a per cent of all published fiction. Archer and Jockers wanted to find out what the books that made it to the bestseller list have in common, hoping this would help them learn more about what makes a novel so good. For their analysis of the novels themselves, they applied the same collection of computer programmes that Jockers developed for Macroanalysis. The results helped to uncover which characteristics a novel should have to be bought and devoured by many readers, a ‘bestseller code’. At least 30 per cent of a bestseller consisted of one or two main topics. The more different topics, the less well the novel sold. The narrative did not need to have a lot of sex, drugs, and rock and roll. Yes, sex sold, but only in a niche market, with a relatively small group of adepts. Part of the bestseller code was that a novel must be about relationships between people – friendship, enmity, love, hate, and everything attached to that. And those relationships must necessarily even be one of the main topics of the novel. The authors of Archer’s and Jockers’ bestselling corpus who best addressed this in their novels were Danielle Steel and John Grisham. These two also applied another rule perfectly: one of the topics addressed in a novel must be dominant and sort of serve as the glue for the story that is being told. Steel is known for her novels in which women find their way to fulfilled lives with meaningful relationships. Grisham’s books revolve around characters with legal problems and how those problems turn their lives upside down. Realism was important in bestsellers, so topics like work, school, media, doctors, death, marriage, and taxes did well (52–68). Archer and Jockers calculated the varying proportion of words expressing positive or negative emotions for each novel and graphed their use throughout a novel. They suggested this may present an impression of the plot, the storyline of a book (97). They looked for novels in their corpus that had a similar line to E.L. James’s 30 The RiddLe of LiTeRaRy QuaLiTy Eve, Martin Paul. Close Reading with Computers: Textual Scholarship, Computational Formalism, and David Mitchell’s Cloud Atlas. Stanford University Press, 2019. Franssen, Thomas. How Books Travel: Translation Flows and Practices of Dutch Acquiring Editors and New York Literary Scouts, 1980-2009. 2014. University of Amsterdam, PhD dissertation. University of Amsterdam Digital Academic Repository, https://hdl.handle. net/11245/1.446961. Hennion, Antoine. ‘Those Things that Hold Us Together: Taste and Sociology.’ Cultural Sociology, vol.1, no.1, 2007, pp.97–114, doi:10.1177/1749975507073923. Herrmann, Berenike J., et al. ‘Revisiting Style, a Key Concept in Literary Studies’. Journal of Literary Theory, no.9, 2015, pp.25–52. Hoover, David L. ‘Testing Burrows’s Delta’. Literary and Linguistic Computing, vol.19, no.4, 2004, pp.453–475, doi:10.1093/llc/19.4.453. Hoover, David L. Modes of Composition and the Durability of Style in Literature. Routledge, 2021. Jockers, Matthew L. Macroanalysis: Digital Methods and Literary History. University of Illinois Press, 2013. Journal of Computational Literary Studies (JCLS), https://jcls.io/. Kestemont, Mike, et al. ‘Collaborative Authorship in the Twelfth Century: A Stylometric Study of Hildegard of Bingen and Guibert of Gembloux’. Digital Scholarship in the Humanities, vol.30, no.2, 2015, pp.199–224. doi:10.1093/llc/fqt063. Ketzan, Erik and Christof Schöch. ‘What Changed When Andy Weir’s The Martian Got Edited?’ Digital Humanities Conference 2017, Montréal, 8–11 Aug 2017. Conference Abstract, pp.285–288. Ketzan, Erik and Christof Schöch. ‘Classifying and Contextualizing Edits in Variants with Coleto: Three Versions of Andy Weir’s The Martian’. Digital Humanities Quarterly (DHQ) vol.15, no.4, 2021, http://digitalhumanities.org:8081/dhq/vol/15/4/000579/000579.html. Koolen, Corina. Reading Beyond the Female: The Relationship Between Perception of Author Gender and Literary Quality. 2018. University of Amsterdam, PhD dissertation. University of Amsterdam Digital Academic Repository, https://hdl.handle.net/11245.1/ cb936704-8215-4f47-9013-0d43d37f1ce7. McCarthy, Rachel and James O’Sullivan. ‘Who wrote Wuthering Heights?’ Digital Scholarship in the Humanities, vol.36, no.2, 2021, pp.383–391, doi:10.1093/llc/fqaa031. ‘Novel Perceptions: Towards an Inclusive Canon Project Website.’ University of Wolverhampton, http://novel-perceptions.thememorynetwork.com/. Pearl, Lisa et al. ‘The Character in the Letter: Epistolary Attribution in Samuel Richardson’s Clarissa’. Digital Scholarship in the Humanities, vol.32, no.2, 2017, pp.355–376, doi:10.1093/ llc/fqw007. Piper, Andrew. Enumerations: Data and Literary Study. University of Chicago Press, 2018. ‘The Riddle of Literary Quality Project Website.’ Huygens Institute, Royal Netherlands Academy of Arts and Sciences, https://literaryquality.huygens.knaw.nl/?page_id=179. The RiddLe 31 Rybicki, Jan, and Magda Heydel. ‘The Stylistics and Stylometry of Collaborative Translation: Woolf’s Night and Day in Polish’. Literary and Linguistic Computing, vol.28, no.4, 2013, pp.708–717, doi:10.1093/llc/fqt027. Rybicki, Jan, et al. ‘Collaborative Authorship: Conrad, Ford and Rolling Delta’. Literary and Linguistic Computing, vol.29, no.3, 2014, pp.422–431, doi:10.1093/llc/fqu016. Salgaro, Massimo. ‘How Literary Can Literariness Be? Methodological Problems in the Study of Foregrounding.’ Scientific Study of Literature, vol.5, no.2, 2015, pp.229–249, doi:10.1075/ssol.5.2.06sal. Schwarz, Ori. ‘Bending Forward, One Step Backward: On the Sociology of Tasting Techniques’. Cultural Sociology, vol.7, no.4, 2013, pp.415–430, doi:10.1177/1749975513480957. Stukker, Ninke, and Arie Verhagen. Stijl, taal en tekst: Stilistiek op taalkundige basis [Style, Language and Text: Linguistics-Based Stylistics, not available in English]. Leiden University Press, 2019. Underwood, Ted. Distant Horizons: Digital Evidence and Literary Change. University of Chicago Press, 2019. Van Cranenburgh, Andreas. Rich Statistical Parsing and Literary Language. 2016. University of Amsterdam, PhD dissertation. University of Amsterdam Digital Academic Repository, https://hdl.handle.net/11245.1/cb936704-8215-4f47-9013-0d43d37f1ce7. Van Dalen-Oskam, Karina, and Joris van Zundert. ‘Delta for Middle Dutch: Author and Copyist Distinction in Walewein’. Literary and Linguistic Computing, vol.22, no.3, 2007, pp.345–362, doi:10.1093/llc/fqm012. Van Dalen-Oskam, Karina. ‘Epistolary Voices: The Case of Elisabeth Wolff and Agatha Deken’. LLC: The Journal of Digital Scholarship in the Humanities, no.29, vol.3, 2014, pp.443–451, doi:10.1093/llc/fqu023. Van Dalen-Oskam, Karina. Het raadsel literatuur: Is literaire kwaliteit meetbaar? [The Riddle of Literature: Can We Measure Literary Quality?] Amsterdam University Press, 2021, https://karinavdo.github.io/RaadselLiteratuur/. Van Peer, Willie, editor. The Quality of Literature: Linguistic Studies in Literary Evaluation. John Benjamins Publishing Company, 2008. Van het Reve, Karel. ‘Literatuurwetenschap: het raadsel der onleesbaarheid’ [Literary Theory: The Riddle of Illegibility, not available in English]. Johan Huizinga Lecture, 8December1978, Leiden, https://www.dbnl.org/tekst/reve003lite02_01/index.php. 2. The National Reader Survey Books, Readers, Opinions Abstract: This chapter describes how the 2013 Dutch National Reader Survey was organized and what the main outcomes were, in such a way that the survey can be adapted for use in other countries or language areas. It presents the criteria for the selection of the four hundred plus one novels that participants could rate and zooms in on the four main categories of books on the list. This is followed by a description of the kinds of readers that took part and which books received the highest and lowest mean scores for general quality (how good the book was) and literary quality. It elaborately quotes from the motivations readers gave for their ratings and analyses the main trends by applying keyword analysis. Keywords: Literary quality, Survey Design, Reader Roles, Keyword Analysis, Computational Literary Studies, Distant Reading The Survey The novelty of The Riddle of Literary Quality is the combination of computational text analysis methods with the analysis of readers’ opinions about the literary quality of the same texts. The aim was to find out whether readers of certain novels agree or disagree about the literary quality of these novels, how large the variety of opinions may be, and whether particular trends or patterns may be observed. Another goal was to explore whether different kinds of readers or reader roles may be distinguished. Last but not least, the Riddle team was interested in how readers would motivate their opinions about literary quality when asked to describe the reasons for their ratings. The necessary data were gathered in 2013 in a large online survey – a logical choice, because in that year 95 per cent of the Dutch population already had access to the Internet.1 1 As stated by the Dutch Central Bureau for Statistics (CBS), https://www.cbs.nl/nl-nl/achtergrond/2013/48/ eenpersoonshuishoudens-maken-inhaalslag-met-internet. Dalen-Oskam, K. van, The Riddle of Literary Quality: A Computational Approach. Amsterdam: Amsterdam University Press, 2023 doi 10.5117/9789048558148_ch02 The naTionaL ReadeR suRvey 33 This chapter starts with a description of the composition of the survey and continues with a discussion of the selection of the corpus. It then presents a general overview of the results in more detail than was done in the 2020 article ‘Literary Quality in the Eye of the Dutch Reader: The National Reader Survey’ written by Corina Koolen, Karina van Dalen-Oskam, Andreas van Cranenburgh, and Erica Nagelhout. The outcomes that will be discussed inform the organization of the rest of this monograph, as will be made clear throughout the chapter and summarized at the end. For the exact numbers, percentages, and more details concerning the outcomes to be discussed, scholars may refer to the ‘litRiddle’ package for R that includes all data that were gathered in the survey (see Appendix 3). Specialized in economic psychology, market researcher Erica Nagelhout helped the Riddle team to design and carry out The National Reader Survey (Het Nationale Lezersonderzoek), as we named it. The agency that took care of the technical implementation to present the survey to a wide audience also provided the option to hire participant panels representative of the Dutch population in terms of gender, age distribution, educational level, and place of residence. We used such a representative panel twice; the first time to test the survey before we finalized it, and the second time just before we closed the survey to the public, to be able to compare the information collected so far with the results from the representative panel of around five hundred inhabitants of the Netherlands. This would help us to see how the participants in The National Reader Survey compared to the Dutch population as a whole. The final version of the survey can be found in Appendix 1. We asked respondents for their age, how many books they estimated to read per year, and whether they read only fiction, only non-fiction, or both. We also asked for their gender, highest level of education, and the first part of their postal code.2 To learn more about possibly different types of readers having different opinions about the same book, the survey included sixteen statements about reading and participants were asked to what extent they agreed with each. In addition, they were asked to indicate to what extent they considered themselves to be literary readers. Participants were presented with a list of the author’s name and title of 401 books and were asked to tick the check boxes of all books they could remember reading. Next, the survey software randomly selected up to seven of the books marked as read and then participants were asked to rate each of those books on two scales: in terms of general quality (how good or bad they thought the novel was) and in terms of literary quality. Score 1 was the lowest score and stood for 2 Dutch postal codes consist of four digits representing an area, followed by two letters indicating a street or part of a street in that area. For privacy reasons, we were only allowed to ask for the four digits and not for the additional two letters. 34 The RiddLe of LiTeRaRy QuaLiTy ‘very bad’ or ‘definitely not literary’ and 7 was the highest, ‘very good’ or ‘highly literary’. Each also had an eighth option ‘Do not know’. Participants could choose to rate seven more books and repeat this as many times as they wished for the books they had read. The survey software then selected one book that the participant had rated relatively high or low for literary quality compared to the other books that had been rated by the same person. For that book the respondent was asked: ‘Can you explain why you rate this book this way?’ A free field gave them ample room to motivate their score. Next, participants were shown the list of books without the ones they had previously indicated to have read. They were asked whether there were any books on the list that they had an opinion about, even without having read them. From the ticked boxes on this list, the survey software randomly selected up to seven books to be rated in the same way as the books they had read. Some of the first participants were rather surprised by the question about books they had not read. This led the Riddle team to post a list of frequently asked questions on the National Reader Survey website, including ‘Why do you also ask for opinions on unread books?’ Our answer was: ‘Sometimes people deliberately choose not to read a book. This suggests they do have an opinion about it. We would like to know that opinion.’ This was one of the two questions we received most often. The other question was ‘What do you mean by literary?’ Our answer was: ‘We would like to find out what readers consider literary. Many readers do have an idea about that. That’s why we don’t give a description of what we ourselves consider “literary”’. The National Reader Survey was launched on 4March2013 and the press release announcing it was widely picked up. Mentions appeared on the websites of various public libraries, publishing houses, and other book- and reading-related sites with a link to the survey. Some newspapers published an interview. When we noticed that proportionally fewer participants came from the Dutch provinces of Friesland and Limburg than from the rest of the Netherlands, we decided to launch two short promotional videos in Frisian and Limburgian, in order to attract even more respondents from those areas. With a message on the web magazine Mixed Grill we tried to get more men to participate. And to get more opinions about romantic-humorous books we advertised on the Dutch site chicklit.nl, at the time of writing in July2022 still a very lively Dutch website, although the label ‘chick lit’ has generally fallen out of fashion because of the inherent sexism of the name.3 3 Kyra Schiffers wrote a communication plan and went in search of opportunities to generate attention for the National Reader Survey. Bas Doppen designed a front page for the website that led to the survey, featuring the logo of an open book with the words ‘What do you think of me’ above it. Renske Siemens ensured, among other things, that a bookmark with the logo was made. 250 copies of these were delivered by NBD Biblion, an organization supporting Dutch libraries, to every public library in the Netherlands. The naTionaL ReadeR suRvey 35 After two weeks, we had received 635 completed surveys. After a month, the score was 1,078. Then the cpnb, the Foundation ‘Collective Promotion for the Dutch Book’, which aims to encourage people to read more books, sent out a mailing in which The National Reader Survey was prominently mentioned and a week later (16April2013) the counter suddenly stood at 7,480. From that moment on, the number of completed surveys went up steadily, with additional increases following a mailing from two bookstore chains and an advertisement at the check-out counter of the website Boekwinkeltjes.nl, where private book owners offer books for sale. On 20June we had received 9,227 completed surveys and on 8July11,283. On 27September2013, when 18,484 people had started the questionnaire, we closed the survey. The number of participants who stopped prematurely was higher than expected, 4,700. In the end, 13,784 participants had completed the survey. The Books We were interested in what publishers call ‘longer fiction’ and decided that the survey corpus should include only novels or separately published stories of a certain length (with a minimum of twenty pages). To make relevant observations, we wanted to collect at least fifty ratings for each title. We calculated that we needed four hundred books. 4 To ensure that our personal bias about which novels have or do not have any literary value would not inform our selection of the books we drew up an additional list of strict selection criteria. One of these related to the first time of publication. Because we were interested in the personal opinions of participants, unaffected (if possible) by what they had learned during their education, we selected only recently published fiction. We chose the year 2007 as the cut-off date, excluding books published before 2007. As explained in Chapter1, Johan Kwantes made two short films with authors Renate Dorrestein and Christiaan Weijts in which they asked readers to participate in the survey. The films to specifically recruit Frisian and Limburg participants were recorded by Ilja Nieuwland (Frisian) and Steven Surdèl (Limburgian). 4 We wanted to collect opinions on two scales: literary quality and general quality. Our initial assumption was that this should get opinions about four groups of texts: (1) books perceived as both literary and good; (2) books perceived as literary but bad; (3) books not perceived as literary but perceived as good; (4) books not perceived as literary and perceived as bad. In addition, we wanted opinions not only about books that had been read by the respondents, but also about books that had not been read. This resulted in a total of eight categories: read books that were considered literary and good, books not read that were perceived as both literary and good, and so on to books not read that were not perceived as literary and perceived as bad. We estimated that we would have enough information to make meaningful observations if each category contained fifty books. Based on this assumption, we went in search of four hundred books. 36 The RiddLe of LiTeRaRy QuaLiTy we included both novels originally written in Dutch and novels translated into Dutch. For comparable reasons we also did not want to limit ourselves to the ‘literary novel’. If we wanted to find out what ‘literature’ and literary quality are, we would have to be able to identify the differences between fiction that does and does not get high scores for literary quality. That is why we also paid a lot of attention to novels that publishers labelled differently than ‘Literary novel’, such as thrillers and romance, to name only the most well-known. To make our selection replicable we based it on the genre label the Dutch publishers attributed to the books.5 We did not include the label in the list of the books as shown to the respondents nor did we show book covers. Publishers often select a book cover with a particular category and audience in mind, to influence buyers and readers. As researchers such as Dixon et al. (2015) have shown, readers can often tell from the cover images what they may expect from a book. To minimize the influence on the participants in The National Reader Survey we only presented the author’s name and the book title. To ensure that we would select books read by as many people as possible – otherwise we risked not gathering enough opinions to make meaningful observations – we decided to compile the list based on the best-selling titles in bookstores and the most borrowed titles in public libraries in the three years preceding The National Reader Survey. 6 We completed the list with the three book week gifts from 2010, 2011, and 2012, the free book that buyers receive from their bookseller when they buy a book during the yearly ‘book week’ organized by the Foundation ‘Collective Promotion for the Dutch Book’ (cpnb). The book week gifts are not included in sales lists, but because their circulation figures are considerably higher than those of most other books – several hundreds of thousands as opposed to 5 In 2013, Dutch publishers made use of NUR, which stands for ‘Nederlandse Uniforme Rubrieksindeling’ (‘Dutch Uniform Classification’). The code consists of three digits and indicates the main subject (e.g. detective, adventure novel), the form (paperback, comics general), or the target group (literary fiction general, fiction 15+) of a book. Booksellers use the NUR code to decide where to place a book in their shop: on the table with new suspense novels or literary fiction, or on the shelves with e.g. poetry or books for young adults. In online bookshops the label associated with a specific NUR code is used to facilitate finding books with a similar content, form or target group. For our research into what constitutes literary quality we focused on NUR code 300, Literary Fiction General, and selected all longer fiction from this category. We added subcategory 285, Fiction 15+ (also known as Young Adult Fiction) from NUR code 200, Children’s books. 6 For the selection we used the Top 100 lists of cpnb (the Foundation ‘Collective Promotion for the Dutch Book’) for the years 2010, 2011, and 2012, and the lending figures of the Stichting Leenrecht (Foundation for Lending Rights) for July2009–June2010 and for the same period in 2010–2011 and 2011–2012. Significantly more fiction was bought from bookstores than borrowed from public libraries during 2009–2012 and many of the books that were borrowed often were also on the best-selling list. We addressed this by first selecting 375 titles from the top of the list of best-selling books, supplementing these with 22 titles that were not included in the list of most borrowed books. The naTionaL ReadeR suRvey 37 several thousands to tens of thousands of other books – their inclusion on our list was justified. Of course, we do not know whether all the copies given away by bookshops were actually read but the same is true for all the other books that were bought or borrowed. Because the survey opened only days before the start of the 2013 Dutch Book Week, we added that year’s book week gift as a bonus so the final list consisted of 401 books. In the course of our research we discovered that our list has a few titles that had slipped through our selection net. We did collect opinions about these books. I will clearly indicate whether these are included in the analyses I present in this book.7 The Readers We asked respondents to indicate not only their age and part of their postal code, but also their gender and their level of education. It may be useful to note that the Dutch survey did not include questions about ethnicity and about class, but that both of these were added to the Big Book Review running in the United Kingdom at the time of writing this book. When asking participants for their gender we invited them to select one of only three options: female, male, or no information – we are thoroughly aware of the fact that gender is not a binary concept, but a longer list of options would not have been particularly useful for our analysis. Of the Dutch participants who indicated that they read at least one book a year, 72.3 per cent identified as women, 27.1 per cent as men, and 0.6 per cent (ninety people) did not share information. The number of female participants was even higher in the age group of 16–20 years (82 per cent) but slightly lower in the age group of 65 years and older (over 62 per cent). This result agrees with earlier research that has shown that in the Netherlands most readers of fiction were women. That 61.3 per cent of readers aged fifty and over were women is also in line with previous research. 7 Occasionally a novel turned out to have been published before 2007, such as Saskia Noort, De eetclub (The Dinner Club). Several separately published long stories turned out to be from much earlier published collections of short stories, for example Tessa de Loo’s De grote moeder (*The Great Mother, the asterisk indicates the book is not available in English) from her 1983 debut collection De meisjes van de suikerwerkfabriek (*The Girls from the Confectionery Factory). And a number of books published as novels turned out to be short story collections. Some readers may also wonder why the list includes books such as Reizen zonder John (In America: Travels with John Steinbeck) by Geert Mak and Baltische zielen (*Baltic Souls) by Jan Brokken, which many readers would rather call (literary) non-fiction. The publishing house where these books appeared gave them the label literary novel, and that’s how they ended up on our list. Baltische zielen by Jan Brokken also contains various contributions and could therefore have been regarded as a collection, so that in retrospect the book should not have been admitted to our list. 38 The RiddLe of LiTeRaRy QuaLiTy More than 70 per cent of the respondents indicated that they attained higher vocational education (37.6 per cent) or university education (34.2 per cent). This is 69.4 per cent for women and 78.6 per cent for men. Of the women 27 per cent only completed secondary education while for the men this is 18.1 per cent. The percentages of women and men with primary education as their highest level of education were not that far apart: 2 per cent of women and 1.5 per cent of men selected this option. We did not receive any indication of education level for 1.6 per cent of the women and 1.8 per cent of the men. Participants represented all regions of the Netherlands. For about 14 per cent the place of residence is not known. Furthermore, a few hundred participants came from outside the Netherlands, mostly from Belgium. They either did not enter a postal code or they input a postal code identical to the first part of a Dutch postal code.8 Participants from Flanders might therefore have filled in the same postal code as some participants from the Netherlands. This had a very limited influence on the results, however.9 The participants of The National Reader Survey mainly represented readers with a high level of education and as such were not representative of the entire Dutch population. We know from the annual Intomart GfK survey on reading behaviour in the Netherlands in 2012, 2013, and 2014, for example, that around 83 per cent of the Dutch population read at least one book a year.10 For the 13,784 people who took part in the National Reader Survey, the figure was 98.3 per cent. More details can be found in Figure2.1, which shows the results of the question ‘How many books do you read on average per year?’ On the left we see that 6100 participants filled in a number from 1 to 20, almost 3800 participants read 21 to 40 books per year, and as the number of books per year increases, the number of participants rapidly decreases from there. On the right, the figure shows the results for bins of 10: 2350 participants read 1 to 10 books per year, 3700 read 11 to 20, and 2450 read 21 to 30 books, et cetera. In total, 224 participants filled in a number higher than 100, and there were ten people who indicated reading more than 200 books a year, one of them even boasted the amazing number of 700. An interactive version of this graph as well as next graphs (in colour) can be found on the companion website to this book (see Appendix 3). 8 See note2 to this chapter. 9 We also know the survey attracted participants with a first language other than Dutch, for example from this motivation – I quote the exact text as written by the respondent: ‘Sorry my Dutch is not good enough to write an answer to this question. (Sorry mijn Nederlands is niet goed genoeg voor deze vraag)’ (ID_5767). The novel this participant had rated was a translation. Most of the times it is not possible to ascertain whether respondents rating translations had read the rated novel in Dutch or in its original language, but we do not consider this to be a problem. 10 See https://www.leesmonitor.nu/nl/wie-leren-er, with references to more detailed reports there. The naTionaL ReadeR suRvey 39 figure2.1: number of books respondents indicated to have read per year A small number of participants indicated reading less than one book a year. Several hundred also read only non-fiction. We were especially interested in those reading fiction. The total number of fiction-reading participants was 13,332 and they read an average of 28.5 books per year, which amounts to a book about every fortnight and perhaps a few extra during the holidays. Another study from early 2013 found that residents of the Netherlands read an average of fifteen books a year all the way through (that study did not count books that readers did not finish).11 This confirms that the National Reader Survey attracted a lot of frequent readers. To find out whether the respondents to The National Reader Survey may be distinguished into different types of readers I collaborated with the American literary theorist and statistician Allen Riddell. Many theories address how readers read, but most of these are difficult to test empirically. This is the case, for example, for the four modes of reading proposed in 2008 by literary theorist Rita Felski. She summarized each of these in one word: recognition (to get to know oneself better), 11 See https://www.kvbboekwerk.nl/app/uploads/2017/09/2013-smb-gfk-meting-23-regulier.pdf, accessed through https://www.kvbboekwerk.nl/consumentenonderzoek/consumentenonderzoek. 46 The RiddLe of LiTeRaRy QuaLiTy novel written by a woman ranks 36: Margriet de Moor, De schilder en het meisje (*The Painter and the Girl, not available in English, mean score 5.9). Both these novels will receive due attention in Chapter4, Literary Novels Written by Women. A closer look at the bottom ten, however, is also extremely interesting (Table2.3). At the very bottom, E.L. James’ Vijftig tinten grijs (from English Fifty Shades of Grey) has a score of 2.1. The next two parts of this trilogy both scored 2.6 and are at the bottom of the list at places 12 and 14. The hugely popular Fifty Shades of Grey trilogy is a phenomenon that deserves special attention. In Chapter3, Romance, Suspense, and Translations, I will discuss these books in detail and will also address the difference in score between Volume1 and Volumes 2 and 3. Table2.3: Top 10 least literary of the 401 books Mean score Author, Title (Title of published English version; *not available in English in 2022) 2.6 Jill Mansell, Drie is te veel (Two’s Company) 2.6 Katie fforde, Trouwplannen (Wedding Season) 2.5 Lauren Weisberger, Champagne in Chateau Marmont (Last Night at Chateau Marmont) 2.5 Jill Mansell, Versier me dan (Take a Chance on Me) 2.5 sophie Kinsella, Ken je me nog? (Remember me?) 2.5 sophie Kinsella, Mini shopaholic (Mini Shopaholic) 2.5 Lauren Weisberger, Chanel chic (Chasing Harry Winston) 2.4 sophie Kinsella, Mag ik je nummer even? (I’ve Got Your Number) 2.4 sophie Kinsella, Shopaholic en baby (Shopaholic and Baby) 2.1 e.L. James, Vijftig tinten grijs (Fifty Shades of Grey) All ten novels at the bottom of the literary ranking were written by female authors and, apart from Vijftig tinten grijs (Fifty Shades of Grey), they all belong to the category of Romance. Moreover, all ten are translations from English. At place 16 from the bottom, we find the first book by a male author: Een dief in de nacht (*A Thief in the Night) by Appie Baantjer and Simon de Waal, which received an average score of 2.7. That literary quality and general quality indeed are two different things for the participants in The National Reader Survey becomes apparent when the list is sorted from best to worst. Vijftig tinten grijs (Fifty Shades of Grey) is still at the very bottom but the average score of 3.5 for general quality is, on a scale of 7, already less serious than the 2.1 for literary quality. Vijftig tinten donkerder (Fifty Shades Darker) is now also in the last ten; the third volume can be found at place 13 from the bottom. Vijftig tinten vrij (Fifty Shades Freed) scored an average of 4.6 and again we see a striking difference between the first volume on the one hand and the second and third volumes on the other. The other eight at the bottom are completely different from the bottom of the list for literary quality. The Romance novels that received such low ratings for literary quality interestingly are not rated as the worst The naTionaL ReadeR suRvey 47 novels one could read. Next to the two Fifty shades volumes, only one of the ten is a translation: Elizabeth Gilbert, Eten, bidden, beminnen (Eat, Pray, Love), which scored a 4.6 for general quality. What is also striking is that the other seven books are originally Dutch novels, written by both female and male authors; four of these are from the category Literary novel, two from Suspense, and one from Romance. Readers did indeed make a difference in their judgements between literary quality and general quality. This becomes even clearer when we compare the average scores for all 401 novels. Figure2.4 shows how many books have an average score ranging from 0–1, 1–2, et cetera, to 6–7.15 The lowest average score for general quality, i.e. how good readers thought the book was, is 4. The lowest average score for literary quality is 2. So we see a broader range for literary quality: from 2 to 6. If we compare the mean scores for literary quality and general quality for all books, we see that two thirds of the titles labelled by the publishers as Literary novel score lower on literary quality than on general quality, while one third score the same or slightly higher. Books from the other two major genres in our corpus, Romance and Suspense, without exception score higher on general quality than on literary quality. The biggest difference in score for Romance novels is 3.1, for Mag ik je nummer even? (I’ve Got Your Number) by Sophie Kinsella, with a 5.5 for general quality and 2.4 for literary quality. In the Suspense category, the biggest difference is 2.7, for 61 Uur (61 Hours) by Lee Child (overall quality: 5.9, versus literary quality: 3.2). figure2.4: Mean scores for all 401 books 15 In other words, Fig.2.4 presents the aggregated number of books having a mean score rounded to the nearest whole number. 48 The RiddLe of LiTeRaRy QuaLiTy Not reading a book can also reflect an interesting opinion. Participants in The National Reader Survey could indicate whether there were any books on the list that they had not read but did have an opinion about. Participants could then rate a maximum of seven of these books. The lowest number of ‘not read, but have an opinion’ for a book is 19, the highest 4310. As a reminder, a total of 13,784 people completed the survey. The top 10 of books participants most often have an opinion about is presented in Table2.4. Table2.4: Top 10 books participants had not read but had an opinion about Number of participants Author, Title (Title of published English version; *not available in English in 2022) 4310 e.L. James, Vijftig tinten grijs (Fifty Shades of Grey) 2139 e.L. James, Vijftig tinten donkerder (Fifty Shades Darker) 1849 a.f.Th. van der heijden, Tonio (*Tonio) 1728 Kluun, Haantjes (*Alpha Males) 1716 e.L. James, Vijftig tinten vrij (Fifty Shades Freed) 1019 stieg Larsson, Mannen die vrouwen haten (The Girl with the Dragon Tattoo) 983 Jonas Jonasson, De 100-jarige man die uit het raam klom en verdween (The 100-Year- Old Man Who Climbed Out the Window and Disappeared) 976 Tatiana de Rosnay, Haar naam was Sarah (Sarah’s Key) 976 herman Koch, Zomerhuis met zwembad (Summer House with Swimming Pool) 951 saskia noort, Koorts (*Fever) The list in Table2.4 probably primarily highlights which books were most visible in bookshops, in newspapers, on television and elsewhere during the time preceding our survey in 2013 and were most often discussed and talked about. The top 10 includes the Dutch translation of all three volumes of E.L. James’ Fifty Shades trilogy, and two Dutch novels that appeared almost simultaneously in early 2011: Zomerhuis met zwembad (Summer House with Swimming Pool) by Herman Koch and Haantjes (*Alpha Males) by Kluun (a one-word penname of the author publishing in English as Ray Kluun). This almost simultaneous publication received a lot of media attention for months on end. In Chapter5, Literary Novels Written by Men, I will analyse the head-to-head literary battle this resulted in. We noted an interesting trend when we compared the mean scores for all 401 books by readers and non-readers. Novels ending up at the bottom of the literary rankings were usually judged even more harshly for their literary quality by non-readers than by readers. The books at the top of the literary rankings, however, received roughly the same score from non-readers as from readers and sometimes the scores were even a little higher. This firmly illustrates that the participants in The naTionaL ReadeR suRvey 49 The National Reader Survey did have an idea of the literary value of books they had not read. The implication is that not only the books themselves played a role in the assessment but also their reputation or that of their author. In short: there were indeed certain value conventions and these were apparently recognizable to readers without even having to read the book. The four main categories of books also received clearly different mean scores (Table2.5). The average score on general quality is 5.81 (on a scale of 7) for all 147 novels in the category Literary novel, 5.66 for all 186 novels in the category Suspense. The 27 titles labelled by us as Other scored an average of 5.59 and the 41 in Romance received an average of 5.27 from readers, which is the lowest average score per category. So, the differences in overall quality (how good is the book?) between the four main categories were not much more than half a point. The averages were calculated on the basis of all reader opinions together. Taking the average score per book, the numbers were different but the trend remained the same. This also means that the corpus would be of little help if we wanted to find out what distinguishes a good book from a bad one: it simply did not contain enough bad books. Luckily, for literary quality the differences were considerably larger. Romance scored lowest, with an average of 2.95, and the category of the Literary novel received the highest mean score, with 5.44. Suspense ended up in between these two categories with 3.94. Other is just above Suspense with 4.07. Table2.5: Mean scores for literary quality and general quality per category Category Number of books Literary quality General quality Literary novel 147 5.44 5.81 suspense 186 3.94 5.66 Romance 41 2.95 5.27 other 27 4.07 5.59 It is also informative to look at the distribution of rounded mean ratings for literary quality across all categories (Figure2.5). We see, for example, that the majority of Literary novels received an average score that was rounded to 5 or 6 for literary quality, and most Suspense had a mean rate rounded to 4). Both categories most often received a rounded to 6 for general quality (Figure2.6). In other words, most Literary novels had a higher score on literary quality than most other books. Apparently, a suspenseful book was hardly ever highly literary. Romance had even less chance of being rated as highly literary. There was an overlap, but the ranking of the genres was clear. Literary novels were at the top, and Romance at the bottom of the genre hierarchy. 50 The RiddLe of LiTeRaRy QuaLiTy figure2.5: for all 401 books the mean score for literary quality per category figure2.6: for all 401 books the mean score for general quality per category An additional feature we considered is whether a book was written in Dutch or was translated into Dutch from another language. Taking all reader opinions on all 401 books into account, we found that their average score for literary quality was 4.67 (on a scale of 7 being the highest). The 152 original Dutch novels received an average of 4.95 from the participants in The National Reader Survey, which was slightly higher than the overall average, while the 249 translated novels received an average of 4.47, slightly lower than the average of 4.67. When it comes to literary quality, translations were thus rated slightly lower than originals, and this difference is statistically significant. The naTionaL ReadeR suRvey 51 The possible influence of the translation of a novel on the reader’s judgement is further explored in Chapter3, Romance, Suspense, and Translations. Zooming in on individual books on the list also helps to find out more about the diversity in readers’ ratings. The novel that was most read, Haar naam was Sarah (Sarah’s Key) by Tatiana de Rosnay, was rated for general and literary quality by 1117 participants. Not all of the 7135 readers who had indicated to have read this novel were presented with the title for assessment; our survey software ensured that as many different titles as possible were rated by the participants. As a reminder: score 1 was the lowest score and stood for ‘definitely not literary’ and 7 was the highest, ‘highly literary’. Participants could only enter whole numbers, but the averages we calculated obviously have decimal points. In the eye of the readers, Haar naam was Sarah (Sarah’s Key) scored an average of 4.7 for literary quality and 5.7 for general quality (how good the book was considered to be). Figure2.7 shows that for this novel, the most commonly awarded score for general quality is a 7 and for literary quality a 6. For the sake of completeness: the non-readers with an opinion about this book gave it slightly lower scores: an average of 4.0 for literary quality and 4.4 for general quality. figure2.7: Ratings of Tatiana de Rosnay, Haar naam was Sarah (Sarah’s Key) Motivations The National Reader Survey not only yielded ratings of the 401 books on the list, but for one book they rated participants were also asked to motivate their score for 52 The RiddLe of LiTeRaRy QuaLiTy literary quality in a free text field. The survey software selected a book that ranked relatively high or low on the literary quality scale compared to the other books rated by the same person. Around 11,951 of the 13,784 respondents made use of this opportunity. The answers ranged from one word (often ‘no’) up to several sentences. With just a couple of exceptions these motivations were written in Dutch. I will present them in my own English translation; the respondent ID added between brackets after my translation may lead the interested scholar to the original Dutch text in the litRiddle package for R (see Appendix 3). By way of introduction, I present a selection of motivations about Haar naam was Sarah (Sarah’s Key) by Tatiana de Rosnay. One of the participants justified the score 7, ‘highly literary’, as follows: ‘Haar naam was Sarah has several storylines that adequately come together at the end. Furthermore, it is written in such a way that the person reading it feels emotionally involved. Also, the author’s choice of words is “more difficult” compared to, for example, books by Collins and Rowling’ (ID_7261). Another reader wrote: ‘Several storylines run through each other, which gives the book extra power. The plot is clear, and you can read on without having to refer back to earlier parts of the book. There are beautiful sentences in it’ (ID_8559). A score 6, ‘literary’, was motivated as follows: ‘The story, the construction, the suspense, the human condition, beautifully rendered’ (ID_1693). The next participant found the novel ‘tending towards not literary’ (score 3): ‘I find the style extremely annoying. In fact, it is a glorified kitchen maid’s novel’ (ID_10070). Score 2, ‘not literary’, was justified by a respondent as follows: ‘It is mainly a compelling story, but it is not well written. Nothing is left to the imagination of the reader’ (ID_8092). And ‘definitely not literary’, the lowest score of all, was given to the novel by the following reader: ‘A cliché story, structured according to a fixed pattern, everyday language, nothing surprising. All emotions overworked’. (ID_9302). On average, the motivations consist of 15.6 words. All motivations together form a quite substantial research corpus that can provide us with a lot of useful information. Printed they take up 550 pages and it would take weeks to categorize or label them in a useful way for further analysis. However, with a few simple actions in a simple computer programme, it is possible to gain insight into the most important trends very quickly. I used the concordance software AntConc, developed by Laurence Anthony (2019), to get an overview of all the words used and their frequencies and in this way I selected which features to further zoom in on. The programme indicated that all motivations together contained 186,530 words (tokens), divided into a total of 10,728 different words (types). More than half of these types occurred only once, a total of 5792. A very simple first step is to explore the words used most frequently in all motivations. Amongst the hundred most frequent words were nouns such as (I The naTionaL ReadeR suRvey 53 have translated the Dutch words into English): story, style, language, depth, layers, sentences, characters, way, plot, structure, content, theme, and adjectives such as good, literary, beautiful, and suspenseful. Most prominent were motivations in which style played an important role, both positively and negatively. I did not care for the writing style (ID_9745). Because of the style, the construction of the story, the terms used, and especially the subject matter, described from an unusual point of view (ID_8476). Very often, participants also mentioned the absence or presence of deeper layers as the reason for their score for literary quality. Book has a lot of depth and multiple layers (ID_6542). It is suspenseful, the storyline is well crafted, but in literary books I expect an extra layer (ID_9532). Different arguments were regularly combined. For instance, reference to the characters in the story was accompanied by a mention of the style and deeper layers. Beautiful sentences and word choice. Characters studied in depth. Double layer (ID_2409). It’s just a fun book, purely for relaxation. The characters don’t develop, there are no beautiful sentences, you don’t have to search for the deeper meaning, etc. (ID_9230). The same applies when respondents mentioned the story and the structure, the plot – usually style and deeper layers are also addressed. Complicated structure, beautiful sentences, nice use of words, cultural theme, great attention to atmosphere (ID_12253). Both the story and the construction of the story are special, the book has deeper layers. It also has intertextuality. Furthermore, I would call the writing style literary (ID_11713). Some participants explicitly wondered what literature or literary quality actually is, or even went in search of a definition – which we deliberately did not provide because we did not want to influence the respondents. Literary, of course, is a tricky word. What is literature but what those who know about literature think is literature? Nevertheless, I think this is literature. It is well constructed, and the characters are well developed (ID_3837). 54 The RiddLe of LiTeRaRy QuaLiTy From Wikipedia I get: In general, literature is understood to be the entirety of texts considered to have universal value. In my opinion, this is not the case with Herfstlied (*Autumn Song). It is purely for entertainment (ID_5903). Some participants referred to what they learned in high school for their answers. Had to read it for the reading list. Not a very suspenseful story. There are questions but they didn’t stimulate me (ID_8088). Complexity of the story, character development, writing style. More or less the characteristics I used to learn in high school, although these can be applied to more books than my literature teacher listed (ID_4065). The association of literature with difficult and boring came up several times. A book without a good story that does not give pleasure in reading. That has got to be literature (ID_3894). When I think of literature, I think of a bit of pomposity, with a lot of intertextuality. This book was so easy to read that it could hardly be literature (ID_3159). I found it easy to read and sometimes have the feeling that literature has to be difficult to read to be called literary (ID_13077). Motivations regularly included a reference to the name of an author the respondent attributes more value to than the one in our list of books they were asked to substantiate their rating about. Dutch or Flemish authors were mentioned, such as the well-known (for Dutch readers) literary author Harry Mulisch (1927–2010, not represented on our list), but also foreign authors whose work may have been read in the original language or in translation in the Netherlands. I think the book is more of a side dish, a real women’s book, not suitable for people who like literature. Thinking about literature, I’d rather think of Harry Mulisch: difficult words, fat books that make you think and that’s not what I think of this book. Although I liked the book! (ID_6205). Literary has some depth in its themes and it also has to do with writing style. I consider Sebald and Roth to be literature – without short-changing Nico Dijkshoorn. His observations are sharp and accurate, but there is no great idea behind them – other than to inform the reader about his youth (ID_3832). Several readers wrote that they did not want to distinguish between what is literary and what not. The naTionaL ReadeR suRvey 55 No, I think it’s nonsense to ask these questions anyway, what difference does it make if something is literary (ID_13629). I’m not knowledgeable about literature and I’m not interested in the literary value of any book (ID_8402). On the one hand, because it lacks layers and intertextuality (those are the things that would make me consider a book literary), and on the other hand, since reading Karel van het Reve’s Huizinga lecture, I’ve actually doubted the usefulness of those criteria as arguments for literature – so maybe it’s just a feeling (ID_12486). That their rating for literary quality was ‘just a feeling’ is what made it difficult for several respondents to explain their opinion. No, I personally don’t have a good definition of literature. It usually is a feeling (ID_12081). No I can’t, it’s a feeling (ID_12350). Contrasting Opinions In the previous section, a look at the one hundred most frequent words in all the motivations together presented us with a fascinating first view on how the participants in The National Reader Survey motivated their opinions about literary quality. A good way to get more detailed information is to compare the vocabularies of different subsets of the motivations to see where they may differ. For these experiments I again made use of the AntConc software developed by Laurence Anthony (2019). I used AntConc’s keyword function to find out which words occur much more often than would be expected in one set of motivations as compared with another. For each of my keyword measurements I used the following values: Log-likelihood (4-term), statistic threshold p < 0.05 (+ Bonferroni), effect size measure Dice coefficient.16 Using the litRiddle package for R and AntConc, my measurements can be easily reproduced, so I will focus on the main results and observations, again presenting English equivalents of the Dutch keywords. To get an idea of the differences in motivations of a high score for literary quality with those of a low score, I compared motivations for ratings 1, 2, or 3 (low literary quality) with those for ratings 5, 6, and 7 (high literary quality), leaving out 4, the rating exactly in the middle, ‘bordering literary and non-literary’. The 16 More information about keyword analysis and a good example of how it may be used for a stylistic analysis of a literary text can be found in Culpeper 2015. 62 The RiddLe of LiTeRaRy QuaLiTy Dixon, Peter, et al. ‘Judging a Book by its Cover.’ Scientific Study of Literature, vol.5, no.1, 2015, pp.23–48, doi:10.1075/ssol.5.1.02dix. Felski, Rita. Uses of Literature. Oxford: Blackwell P, 2008. Koolen, Corina, et al. ‘Literary Quality in the Eye of the Dutch Reader: The National Reader Survey.’ Poetics, vol.79, no.101439, 2020, pp.1–13, doi:10.1016/j.poetic.2020.101439. Riddell, Allen, and Karina van Dalen-Oskam, ‘Readers and their Roles: Evidence from Readers of Contemporary Fiction in the Netherlands.’ PLoS ONE, vol.13, no.7, 2018, pp.1–12, doi:10.1371/journal.pone.0201157. Von Heydebrand, Renate, and Simone Winko. Einführung in die Wertung von Literatur: Systematik – Geschichte – Legitimation [Introduction to the Evaluation of Literature: Systematics – History – Legitimacy, not available in English]. Ferdinand Schöningh, 1996. 3. Romance, Suspense, and Translations Mixed Methods in More Detail Abstract: Translations were usually valued slightly less than Dutch originals. Participants attributed the lowest mean scores for literary quality to E.L. James’ Fifty Shades of Grey. The Dutch publisher marketed the translations of the Fifty Shades trilogy as Literary novels. To find out if this label indeed suits the novels best, the chapter applies cluster analysis using Burrows’ Delta and principal components analysis (PCA). The same methods are used to explore five novels by Swedish author Henning Mankell, to see whether the style of his Literary novels differs from his Suspense novels. Finally, keyword analysis is applied to get an indication of the fluency of translations, to see whether the results may explain the differences in ratings of originals and translations. Keywords: Literary quality, Fifty Shades trilogy, Henning Mankell, Translation quality, Computational Literary Studies, Distant Reading From Readers’ Opinions to Text Analysis This chapter presents the results of several experiments with the mixed-methods approach I selected with the aim to learn more about the conventions of literary quality in the Netherlands in 2013. It sketches the text analysis methods I applied, shows how they were combined with the results from The National Reader Survey described in the previous chapter, and outlines the ensuing observations. Most of the novels discussed in this chapter are translations and most belong to the categories of Romance and Suspense or, for one reason or another, skirt the lines between Literary novel, Suspense, and Romance. Special attention will be paid to the issue of translation quality. In Chapter2 I presented the top 10 most literary and least literary novels from the list of 401 books in the eyes of the Dutch readers who participated in the survey. As a reminder: The highest score a reader could attribute to a novel was 7, ‘highly literary’, the lowest 1, ‘definitely not literary’. The novel to which readers attributed Dalen-Oskam, K. van, The Riddle of Literary Quality: A Computational Approach. Amsterdam: Amsterdam University Press, 2023 doi 10.5117/9789048558148_ch03 64 The RiddLe of LiTeRaRy QuaLiTy the highest mean rating, Julian Barnes’ Alsof het voorbij is (The Sense of an Ending) received an average of 6.6 and the one with the lowest, E.L. James’ Vijftig tinten grijs (Fifty Shades of Grey) scored a 2.1. For all 401 books in the corpus the mean scores for both literary quality (the perceived literariness) and general quality (how good the book was considered to be) can be found in Appendix 2, ranked from highest score for literary quality to lowest. My first experiment was an exploratory analysis of these twenty novels applying a cluster analysis based on the hundred most frequent words. I selected the ‘Classic Delta’ measure that was based on Burrows’ Delta (described in Chapter1) and made use of the Stylo Package for R (Eder et al.2016). In a cluster analysis, the most similar texts in terms of words and word frequencies are statistically linked in pairs after which the most similar clusters are also linked in pairs, and so on. The further to the right in the graph two books are connected, the more similar they are in the frequencies of the selected words – the graph is ‘built’ from right to left. The result of this first measurement is visualized in Figure3.1. The short names for each novel consist of the author name and, after the underscore, the first three words of the title that are not an article. The vertical line connects the books or clusters of books that are most similar in the frequencies of the hundred most used words. The author signal is clearly visible: all of the novels by Weisberger, by Mansell, and by Kinsella, all represented by more than one novel in the ten least literary novels on the list, cluster together in separate author clusters. Furthermore, the ten books with the lowest scores for literary quality (starting with L_, in a lighter colour in Figure3.1) end up together in one of the two main clusters, and eight of the highest-scoring novels (starting with H_) are also clustered together. On the companion website to this book (see Appendix 3) this figure can be found in colour, added with visualizations from more measurements and with some more information. For now, it suffices to note that for the selected hundred most frequent words, the novels Norwegian Wood (Norwegian Wood) by Haruki Murakami and Alsof het voorbij is (The Sense of an Ending) by Julian Barnes resembled the ten novels awarded the lowest literary quality slightly more than they resembled the other eight highly literary novels. An inspection of the hundred most frequent words in these twenty novels with the AntConc tool by Laurence Anthony showed only two of these are content words: the nouns ‘eyes’ and ‘hand’. The other 98 are function words: articles, pronouns, prepositions, and verb forms of the verbs to be and of modal verbs. When I applied cluster analyses with more words (from 130 most frequent words onwards), Murakami and Barnes moved into the cluster with the highest literary quality, but they were joined by five of the least literary ones. RoManCe, susPense, and TRansLaTions 65 figure3.1: The ten novels with the highest (h) and lowest (L) scores for literary quality, Cluster analysis (100 most frequent words) Applying the same measurement to the fifty most literary and fifty least literary novels, which together took up a quarter of the complete corpus of 401 novels, yielded roughly the same results as for the ten most and least literary novels (graph not shown). More than forty books from both ends of the list clustered together. Barnes and Murakami here ended up in the cluster in which the highly literary books dominated. Here, seven of the novels with highest literary quality ended up in the cluster that included most of the novels that received the lowest scores for literary quality. They are listed below; an asterisk before the English titles indicates that the novel is not available in an English version. In those cases, I present an English translation of my own that is certainly not always the best possible translation, but will give at least some idea about the book to those who do not read Dutch. 66 The RiddLe of LiTeRaRy QuaLiTy These provisional translations are also included in the litRiddle package for R (see Appendix 3). Anna Enquist, De verdovers (*The Anaesthetists) Arnon Grunberg, De man zonder ziekte (*The Man Without Illness) Arnon Grunberg, Huid en haar (*Hair and Hide) P.F. Thomése, De weldoener (*The Benefactor) Arthur Japin, Dooi (*Thaw) Bernard Dewulf, Kleine dagen (*Little Days) J.J. Voskuil, De buurman (*The Neighbour) The cluster dominated by novels that received the highest scores for literary quality included six books that belonged to the fifty novels with the lowest scores: Jennifer Weiner, Sommige meisjes (Certain Girls) Candace Bushnell, 1 Fifth Avenue (One Fifth Avenue) Mart Smeets, Een koud kunstje (*A Breeze) Mart Smeets, De afrekening (*The Reckoning) Suzanne Vermeer, Cruise (*Cruise) Suzanne Vermeer, Après-ski (*Apres Ski) Although there is again a measurable difference between most of the novels that received the highest and the lowest scores for literary quality, this difference does not explain yet how the novels differ exactly, nor why the readers rated them the way they did. An important point to note, however, is that the fifty most literary novels were almost all marketed as Literary novels, whereas most of the fifty least literary books had a different label. They included a few Suspense novels, five regional novels, and a lot of novels we categorized as Romance (more about the categorization in Chapter2). It is therefore much more likely that the results of these first measurements did not mainly reflect the difference in literary quality but rather a difference in genre. They do not solve the riddle of literary quality but point the way to further explorations. Especially the outliers are intriguing and several of them will be discussed in more detail in the upcoming chapters. Romance and the Fifty Shades Trilogy The first books I will elaborate on are the three volumes of the Fifty Shades Trilogy. Vijftig tinten grijs, the translation of E.L. James’ novel Fifty Shades of Grey, received RoManCe, susPense, and TRansLaTions 67 the lowest mean score for literary quality out of all 401 books. Participants who did not read this novel and did contribute an opinion about it, rated it even lower than those who did read it. Readers rated the literary quality of this first volume of the trilogy with an average of 2.1, the second and third volume both received a mean score of 2.6. The averages were determined based on a total of 627 ratings for the three volumes together of which the first one received the most, 280. Volume2, Vijftig tinten donkerder (Fifty Shades Darker), received 169 ratings and Volume3, Vijftig tinten vrij (Fifty Shades Freed), 179. Figure3.2 shows that considerably more readers gave the very lowest score, 1, to the first volume of the trilogy than they gave to the second and third volume. Many of these readers probably did not continue reading volume2 and 3. Those who did continue reading can be expected to have given higher ratings, with as a result the somewhat higher average scores for the second and third volume.0 figure3.2: Ratings for the literary quality of e.L. James, Vijftig tinten-trilogie (Fifty Shades Trilogy) To find out why readers were so disenchanted by the literary quality of the trilogy, I looked at the justifications readers provided for their rating. For the three novels together we recorded 143 motivations, nine of which were written by male participants. The words niet (not), geen (none), and weinig (little) occurred frequently but so did veel (many). Negative adjectives dominated: slecht (bad) and simpel (simple). The word goed (good) also frequently occurred, but as with veel (many), this may be immediately after a negation. Readers are referred to the litRiddle package for R if they would like to verify this themselves. I will quote a couple of the motivations as a teaser, starting with four of the very few respondents who gave one of the three volumes a score of 6 (‘literary’) or 7 (‘highly literary’). I present my own English 68 The RiddLe of LiTeRaRy QuaLiTy translation of the motivations; the original Dutch text can be found in the litRiddle package for R by means of the added ID. It is simply wonderful (Female, age 18; ID_975). Because it is suspenseful, contains a story, and is written with love (Female, age 60; ID_12026). Because it is well written since you are just completely drawn into the story (Female, age 16; ID_7676). A story that really touches you (Female, 16; ID_5223). The age of these respondents is striking: three very young participants (two aged sixteen and one aged eighteen – potential participants under sixteen were not allowed to take part in the survey) who indicated to be swept away by the novel and one aged sixty who appreciated the suspense of the story and remarked that it is ‘written with love’. Interesting as this may be, we cannot draw any solid conclusions from this precisely because the number of motivations for these high scores is so small. We may learn more from looking at motivations for the abundantly available low scores. These readers agreed about a poor writing style and a bad story. Some wondered aloud why they read it or why others read it. The writing style is abominable, the story is devoid of substance, and it is obvious that the author has a very limited vocabulary. It is still a mystery to me why I read this book in the first place (Female, 36, score 2; ID_11795). Language use, characters, and content. Lots of repetition of what happened, it was innovative because of the bondage aspect. More of an entertaining book and a hype (Female, 56, score 1; ID_5714). Because not a single literary sentence can be found in the entire book! This book is only about the subject matter. Never read such a badly written book. Content was a revelation but I don’t need to read a book for that (Female, 55, score 1; ID_2470). As the second quote above shows a low score can go hand in hand with positive comments: the book was entertaining. Describing the novel as innovative was probably also meant positively. Something similar can be seen in the third quote, in which a reader admits that the content of the story was nothing less than a revelation to her – only to immediately undo the compliment. Some readers found the emphasis on sex in the story sufficient reason to give it a low score for literary quality. Other readers wrote that the sex is well described: RoManCe, susPense, and TRansLaTions 69 Too much focus on sex (Female, 33, score 1; ID_13638). Because the story seems closer to a book for older girls, rather than being original. I do find the descriptions of sexual acts and physical things very good (Female, 62, score 3; ID_11347). However, good descriptions of sex were not automatically perceived as literary descriptions: This is absolute pulp, barely worth the paper it is written on. I don’t understand why it is so highly praised. Fine what someone does in the bedroom but literature … Marquis de Sade was literature, not this (Male, 54, score 1; ID_10479). Another comparison was much more prominent in readers’ substantiations of their opinions of the Fifty Shades trilogy. A great many readers compared the novels to popular romantic books on sale in supermarkets: Ladies’ novel. Even worse than the average Harlequin novel (Male, 53, score 1; ID_13048). To me, this book feels like a more expensive edition of the Harlequin novels and other books about love and sex that can be found at the cash register in a supermarket (Female, 31, score 1; ID_11012). In Chapter2 I mentioned that the publisher of the Dutch translation of the three volumes of the Fifty Shades trilogy marketed the books with the label of Literary novel. Because the Riddle team was not convinced by this categorization, we decided to keep the novels apart from the other Literary novels in our corpus by putting them in our fourth category, Other. To find out whether the language of the trilogy is closer to the other Romance novels, as we suspected, or that they indeed are more alike to the other Literary novels in our corpus, I have taken several measurements, each time comparing the Fifty Shades trilogy with books from a different category. The results of this experiment are best illustrated by the visualizations of a principal components analysis (PCA) of the thousand most frequent words. PCA applied to the vocabulary of texts analyses correlations in the frequencies of the words used in the novels. It is a mathematically sophisticated procedure which maps the very complex abstract space of a large set of input variables (in this case: word frequencies) onto a conceptually simple, two-dimen- sional scatterplot. Such a visualization is much more intuitive for a researcher than multidimensional setups. It may be helpful to think of a designer trying to draw a building on a sheet of paper: the usual approach is to represent the most 70 The RiddLe of LiTeRaRy QuaLiTy important side (for example the front wall) as precisely as possible, while the side walls are deliberately transformed, to represent as much information as possible. After the transformation, the side wall coordinates are embedded into the X and Y dimensions. PCA performs a very similar transformation of the input variables, in order to identify the dominant information, namely PC1 and PC2. The other principal components can still be accessed and shown on a separate plot, but they usually contain very little information about the input data, and I refrain from showing them in the graphs in this book. So, the first component, depicted on the horizontal axis in visualizations, reflects the cluster of correlated variables that together explain most of the variation in the data. Previous research on literary texts has found that the first principal component is usually related to the authorship of the texts. The second principal component, shown on the vertical axis, reflects the largest amount of correlating variables out of the remaining variables (i.e. those that do not correlate with the first component). Earlier research has shown that this can often be linked to a difference in genre or subgenre of the texts. More information about PCA in computational literary research can be found in Jockers (65–67 and 100–104), Hoover (26 and onwards) and Craig and Greatley-Hirsch (30–39). As before, I used the Stylo Package for R. For all PCAs I used the correlation version. In the visualizations of the measurements, books that are very similar are close together, and books that differ from the other selected novels end up as outliers. When exploring these visualizations it is important to remember that the results only show how the books that were selected for the experiment compare to each other. As soon as books are added or removed, the picture will change. Still, these visualizations may be of great help to get a rough idea about how the selected novels compare to each other in terms of most frequent words. In figure3.3, I have visualized the comparison of the trilogy with the 41 Romance titles. On the basis of a statistical comparison of vocabulary and word frequencies, the software has situated the trilogy far in its own corner compared to the other novels. Thus, the Fifty Shades trilogy is visibly different from the titles in the Romance category. It is important to remember that we did not have any representatives of the Harlequin type of novels in our corpus. To check whether the trilogy also deviates from this kind of popular romantic books we would need another corpus of texts, with representatives from the time period just before Fifty Shades of Grey took the book market by storm. For now, it is sufficient to note that the trilogy seems to be far removed from the novels in our corpus that have been labelled as Romance. RoManCe, susPense, and TRansLaTions 71 figure3.3: e.L. James, Vijftig tinten-trilogie (Fifty Shades Trilogy) and Romance, PCa (1000 most frequent words). novels translated from english are indicated with an e_ in front of the author name and shortened title and novels originally written in dutch are indicated with n_. The o_ in front of the Fifty Shades titles stands for other. The various groups have each been given their own shade of grey to make them easier to recognize. Next, I compared the trilogy with Literary novels. For this, I did two sets of measurements, one comparing it with the Literary novels written by men and the other with those written by women – the visualizations would otherwise be too cluttered for an interpretable result. The graphs are not shown here, but they are available on the companion website to this book, see Appendix 3 for more information. Both times the trilogy stood well apart from the other novels. However, the distances are somewhat smaller than in the comparison with the Romance category. I would conclude that labelling the volumes of the trilogy as Literary novels is generous but perhaps not as far-fetched as we initially thought. 78 The RiddLe of LiTeRaRy QuaLiTy mother, and her daughter. The other literary novel by Henning Mankell in our corpus is Minnet av en smutsig ängel, which appeared in Swedish in 2011 and was translated into Dutch by Clementine Luijten the same year as De geschiedenis van een gevallen engel (A Treacherous Paradise). Here, too, a woman is the protagonist of the story. The novel tells how Hanna Renström in 1904 moves into a hotel in Lourenço Marquez, which turns out to be a brothel, and how she copes with this environment. The three Suspense novels by Mankell that met the criteria for inclusion in our corpus (most sold from bookstores and/or most borrowed from public libraries in 2009–2012 and first published in Dutch in 2007 or after) are Kennedy’s brein (Kennedy’s Brain), De Chinees (The Man from Beijing), and, from the Kurt Wallander series, De gekwelde man (The Troubled Man). Kennedy’s brein and De Chinees also have female protagonists. The five novels by Mankell have each been translated into Dutch by a different translator (see Table3.1). All five books were rated by between 440 to 504 readers. The average scores for literary quality were between 4.75 and 5.00. The two literary novels received the highest scores but they were not that much higher than the scores for the three Suspense novels. These three ended remarkably high up in the literary ranking, wedged in between many other books from the category of the Literary novel. This leads to the question whether Mankell’s Suspense novels are perceived to be more like Literary novels than other books categorized as Suspense. In a small experiment, I analysed all forty translated Literary novels by male authors in our corpus, which include Mankell’s. To this selection I added his three Suspense novels. A whole series of PCA analyses yielded almost the same outcome. As an example, I show the visualization of the result of the PCA of the thousand most frequent words (Figure3.5). Mankell’s three Suspense novels appear somewhat separately from the other novels, together with his Literary novel De geschiedenis van een gevallen engel. However, his other Literary novel De Daisy Sisters is very far removed from the other 4. When I added more thrillers to the comparison, this pattern did not change. The Suspense novels by, for example, Stieg Larsson and Jussi Adler-Olsen then took up the space between the Mankell cluster at the bottom of the graph and the rest of the field, being more similar to the other literary novels than the four books by Mankell (see the website for visualizations of these additional measurements). The conclusion is that, in terms of vocabulary and word frequency, De geschiedenis van een gevallen engel (A Treacherous Paradise) is close to the Suspense novels that Mankell wrote in the same time period. It seems, therefore, that although Mankell published books in two different genres, Suspense and Literary novel, his style as reflected in his use of words, perhaps is not very different. The fact that De Daisy Sisters was an outlier does not disprove this because the original is much older (1982) and it is known that an author’s writing style may develop over time just like languages and the conventions that apply to different genres, as David Hoover (2021) recently demonstrated. Further research into Mankell’s complete oeuvre would RoManCe, susPense, and TRansLaTions 79 be needed to confirm this. If this discrepancy between genre label and Mankell’s writing style is correct, what is interesting is that such a development can even be found in translations. This probably reflects current translation conventions in the Dutch language area – and also needs further research. One of the Literary novels that shifted position after I added Mankell’s Suspense to the mix is De 100-jarige man die uit het raam klom en verdween, the Dutch translation of the debut novel of Swedish author Jonas Jonasson, Hundraåringen som klev ut genom fönstret och försvann (The 100-Year-Old Man Who Climbed Out the Window and Disappeared). The Dutch version was released in 2011 and was a smash hit. It was translated by Corry van Bree, who was also responsible for the translation of Mankell’s figure3.5: Translated Literary novels written by male authors and three suspense novels by Mankell, PCa (1000 most frequent words). Mankell’s suspense novels are indicated with an s_ in front of the author name and shortened title. The Literary novels are divided in those with the highest scores for Literary quality (h_), the lowest scores (L_) and the intermediate group (M_). The various groups have each been given their own shade of grey to make them easier to recognize. 80 The RiddLe of LiTeRaRy QuaLiTy De Chinees (The Man from Beijing). The story is just as unusual as its title. The 99-year- old Allan Karlsson is bored but not bored to death, to his great annoyance. When his hundredth birthday arrives and the administrators of his elderly home have prepared a party for him, he quietly climbs out of his window and sets off to the nearest bus station in his slippers. A slapstick world trip full of car chases and explosions ensues. Perhaps these explosions explain why, in terms of vocabulary and word frequency, the novel most resembles the thrillers among which it ended up in my last measurements. To return to Mankell, it is not clear how his reputation influenced readers in their ratings of the literary quality of these five books. Did his Suspense novels get higher scores than those of other Suspense authors because readers knew Mankell also wrote Literary novels? They did end up between the Literary novels in the ranked list of all 401 books. Or would the mean scores for the two Literary novels have even been higher if Mankell would not have been well-known as a Suspense author? These questions cannot be answered for now. The Riddle of Translation Quality The Mankell case also leads us to another question: Are the ratings the participants in The National Reader Survey attributed to translated novels similar to novels originally written in Dutch? Many Dutch readers also read other languages than Dutch, as can be seen time and again in their answer to the question ‘Can you explain why you rate this book as you do?’ I read the book in Spanish, beautiful diction! (ID_3137, about Carlos Ruiz Zafón, Het spel van de engel [The Angel’s Game]). Read it in French. Wonderful use of language and topical issues very originally and subtly presented. (ID_5763, about Michel Houellebecq, De kaart en het gebied [The Map and the Territory]). Read it in Danish, master it (Danish) sufficiently to find the book suspenseful but cannot give a literary judgement (ID_4967, about Jussi Adler-Olsen, De bedrijfsterrorist [*The Corporate Terrorist]). The story is very important in this book. The language a little less so. Nevertheless, the books in the Potter series are full of wordplay and hidden references that are largely lost in Dutch. NB: That’s why I read this book in English! (ID_3985, about J.K. Rowling, Harry Potter en de Relieken van de Dood [Harry Potter and the Deathly Hallows]). As I already indicated in Chapter1 of this monograph, translations play an important role in the Netherlands. Researcher Thomas Franssen calculated that of all poetry RoManCe, susPense, and TRansLaTions 81 and fiction titles published in 2009, 62.8 per cent were translations. Most translations came from English and belonged to the category of literary novels, many more were translated from different languages than in the categories of suspense and romance (Franssen 30–37). The Riddle research corpus shows the same distribution. Of the 249 translated books in our corpus of 401, 180 were translated from English, out of which 104 originated in the United States and 67 in the United Kingdom. The remaining 69 came – in order of most to least number of titles – from Swedish, Italian, Spanish, Danish, Norwegian, French, German, Icelandic, Japanese, Portuguese, and Afrikaans. I suspect that many respondents based their assessment of a book on reading it in its original language, but I do not think this problematic. Their opinions were still welcome. The previous chapter showed that participants in The National Reader Survey regularly referred to the fact that a book was a translation when substantiating their rating for literary quality. Older participants (sixty years and over) did so slightly more often than younger ones (sixteen–forty-five years). Many respondents were well aware that they were reading a translation. They regularly praised the translated book, but sometimes they were less enthusiastic or suspended a negative judgement about the literary quality by suggesting that the problem may also lie with the translation: Read the translation, which was unreadable because of the grammatically incorrect sentences, and therefore did not finish the book. Maybe I would have had a different opinion if I had read the original English edition (ID_5434, about Kathryn Stockett, Een keukenmeidenroman [The Help]). Very beautifully written, good plot. Good translation (ID_7312, about Tatiana de Rosnay, Die laatste zomer [A Secret Kept]). Maybe it was the bad translation, but the writing style was not exactly literary (ID_5465, about Silvia Avallone, Staal [Swimming to Elba]). Apart from a suspenseful plot the book has deeper layers for the attentive reader. Moreover, the author made extensive use of intertextuality and other so-called literary procedures. And not unimportant: much care was taken with the Dutch translation; it strikes me that this is not always the case with books I consider less literary (e.g. thrillers, which I often detest for that reason) (ID_10902, about Carlos Ruiz Zafón, De schaduw van de wind [The Shadow of the Wind]). Story a bit over the top. Not always beautiful language. (Can also be due to the translation of course!) (ID_11472, about John Boyne, Het winterpaleis [The House Of Special Purpose]). On the seven-point scale from 1 to 7, the 152 novels written in Dutch received an average score of 4.95 for literary quality, just above the average of 4.67 for all 401 novels, while the 249 translated novels ended slightly below that average, with a 82 The RiddLe of LiTeRaRy QuaLiTy score of 4.47. The lower score for translations may largely be due to the fact that the categories of Suspense and Romance contain considerably more translations than original Dutch novels and both are genres that are awarded less literary ratings than titles from the category Literary novel anyway. Translations did not receive similar scores for literary quality as novels written in Dutch in the same category. The scores for Dutch and translated Suspense were not that far apart, while the average score for translated Literary novels is lower than that for original Dutch Literary novels (Table3.3). That category also was the only one with an almost equal number of titles, namely 73 Dutch and 74 translated Literary novels (Table3.2). In the genres of Suspense and Romance, on the other hand, the translated books collectively received a higher score than the Dutch ones. The differences are statistically significant; they can be easily verified with the litRiddle package for R (see Appendix 3). Table3.2: Categories and number of books originally in Dutch and translated into Dutch Category Dutch Translated into Dutch Total Literary novel 73 74 147 suspense 58 128 186 Romance 536 41 other 16 11 27 Table3.3: Categories in Dutch and translated with mean score for literary quality Category Literary quality Category Literary quality dutch 4.95 Translated 4.47 Literary novel dutch 5.52 Literary novel translated 5.35 suspense dutch 3.76 suspense translated 4.02 Romance dutch 2.75 Romance translated 2.97 other dutch 4.74 other translated 3.54 Table3.4: Categories of Literary novel and Suspense in Dutch and translated with mean score for literary quality Novels Number of Literary novels Literary quality Number of Suspense novels Literary quality dutch 73 5.52 58 3.76 Translated from english 44 5.16 89 3.91 Translated from other languages than english 30 5.56 39 4.19 RoManCe, susPense, and TRansLaTions 83 Table3.4 shows that in the categories of Literary novel and Suspense, novels translated from English on average scored slightly lower than books translated from other languages. Translated Suspense from languages other than English even scored significantly higher than those written in Dutch. But in the Dutch literary field, as in our corpus, we found far fewer translations from languages other than English. It may be that it is precisely this relative rarity that makes translations from languages other than English seem slightly more valuable in the eyes of readers. Another explanation should also be considered, however: it may in fact be the case that translations from English are simply worse than those from other languages. And it would be extremely useful if we could somehow easily measure whether readers have a point in their criticism of a translation. Earlier in this chapter I quoted a reader stating a translation had ‘grammatically incorrect sentences’. Another participant in The National Reader Survey wrote: ‘Bad book, too American, unrealistic, badly translated, ugly sentences’ (ID_8793). These readers did not explain what exactly is incorrect or ugly about these sentences but that does not make their opinions any less relevant. Translation scholars have given considerable thought to what makes a good translation. Below, I cite only a few of their publications in English. First and foremost, they agree that the answer to this question is not easy. After all, translation is not simply replacing a word in one language (the source language) with an equivalent in another language (the target language). Languages and language systems are never exactly the same and a translation can therefore never be completely identical to the original. What a translator necessarily does is interpreting the source text and reformulating it in the target language, also taking into account the cultural context (Bassnett 3). Moreover, the ideas of what constitutes a good translation change over time and this underlines there is no fixed standard for a successful translation (Bassnett 81). Little research has been done into measurable aspects of translation (cf. Lewandowska-Tomaszczyk 6). Negative comments about translations are mainly made by people who know the original language well enough to think they can see that the translation is inadequate, according to translation scholar David Bellos in his book Is That a Fish in your Ear? The Amazing Adventure of Translation (36). This statement seems to fit perfectly with the audience that took part in The National Reader Survey, given the motivations cited earlier in this chapter in which readers explicitly stated that they had read a book in the original language. The most widely read and spoken language in the Netherlands other than Dutch is English and on the list of 401 novels, the majority of translations came from English. French, German, and many other languages followed at a great distance. If we are to believe Bellos, then the widespread knowledge of English may contribute to the observation that Dutch 84 The RiddLe of LiTeRaRy QuaLiTy readers seemed to trust translations from English less than translations from other languages. As I showed in the previous section, the average scores for literary quality indeed give the impression that it is mainly translations from English that readers find fault with (see Table3.4). I therefore examined these in more detail in two experiments. I made use of the knowledge that many readers consider a translation to be good if the language use does not reveal that it is a translation, or if the translation reads as if it were an original. The term that translators and translation scholars use for this is ‘fluency’; a text can be experienced as a fluent translation from the perspective of the receiving language and culture. In my experiments I limit my focus to linguistic features, assuming that the degree of fluency of a translation could be revealed by comparing the vocabulary of translated novels with original Dutch novels. The subject of the first experiment is a set of Dutch words that linguists call ‘modal particles’. These are common words that carry little meaning in themselves but which can be used, for example, to diminish or cushion an announcement. In a Dutch sentence such as ‘Ga maar even zitten’ (‘Just sit down’) the words maar and even can be described as modal particles. They are difficult to translate into English. In addition to maar and even we may also include dan, eens, misschien, nou, soms, and toch. English does not know them and has other ways to achieve the same goal. And that has consequences for Dutch-English and English-Dutch translators: they must consciously choose other constructions to express the same intention. Translation scholars have noted that texts translated from English into Dutch contain fewer particles than texts written in Dutch. Together with linguist Ton van der Wouden and data scientist Saskia Lensink I investigated whether this is also the case in the books translated from English in the research corpus of 401 novels (Lensink et al.). We focused on the categories Literary novel, Suspense, and Romance, and examined the occurrence of the modal particles in 305 novels, 136 written in Dutch and 169 translated from English. The measurements showed that the modal particles dan, ook, even, toch, and eens occurred significantly more often in novels written in Dutch. The modal particles maar, nou, soms, and misschien appeared significantly more often in novels translated from English into Dutch. For the Dutch modal particles even, toch, and eens, we also found a difference in usage between the various genres. Using an algorithm based on the occurrence of these words, we wanted to predict whether a novel was a translation or written in Dutch. It turned out that it was not possible to recognize translated books in the category of Literary novel, but that it was possible to recognize translations in the categories of Suspense and Romance. This indicates that in the use of these three modal particles, the translations from English into Dutch in the category Literary novel can be called fluent – the frequency of use was comparable to that RoManCe, susPense, and TRansLaTions 85 in the original Dutch literary novels. This also meant that the translations from the genres of Suspense and Romance were significantly less fluent in the use of the modal particles even, toch, and eens than those from the category of the Literary novel. I consider it quite possible that readers subconsciously noticed this and got the feeling that a translation was substandard. In the second experiment I applied a method I have used before in this book: comparing the word frequencies of all words in two text collections in order to find out which words are significantly more frequent in one group compared to the other. I used the keyword function of the AntConc tool developed by Laurence Anthony. For this purpose, I used a smaller selection of novels translated from English, namely from the category Literary novels written by women – 23 Dutch and 25 novels translated from English. Comparing the frequencies of the thousand most frequent words in these two groups with the help of text analysis software did not show a clear dichotomy. However, a comparison of the vocabulary in both groups did yield words that occurred significantly more often than might be expected in one of the two groups. On average, the 23 Dutch novels contained considerably more often the personal pronouns ik (I), me/mij (me), and mijn (mine). The translations contained ze (she), haar (her), and hen (them) relatively more often. This seems to have little to do with translation quality. It may be that the selected Dutch novels more often presented a first-person narration while the novels translated from English were written more often in a third-person perspective. This was confirmed by research conducted by Lisanne van Rossum in her master’s thesis at Utrecht University in 2020. In the 25 novels translated from English, a number of verb forms occurred significantly more often in the translated than in the original Dutch novels: vroeg (asked), wist (knew), and zei (said). These are often used to mark speech in dialogue. The pattern may also be coincidental, for instance because the English originals of the selected books contained more dialogue than the Dutch novels. Another possibility is that it was not the proportion of dialogue that differed, but the way in which it was introduced, for instance when it would be more usual in original Dutch novels of the same genre to omit such an indication. In that case, we could conclude that here we have encountered a reflection of different literary conventions in the two languages. Yet another possible explanation is that translators may have added the dialogue marks themselves from time to time. The English originals, like many Dutch novels, then did not always explicitly signal the interactions in a dialogue with a verb. Research has shown that translators often tend to render the text they translate more understandable and accessible, among other things by adding these kinds of words. For an example of changing dialogue markers in Swedish translations from English, see Bellos (188). If translators indeed added such dialogue markers, 86 The RiddLe of LiTeRaRy QuaLiTy I would argue that the translation can be labelled as less fluent. The Dutch translation may be seamless, but the text has been made easier to read than was probably intended in the original. This simplification may not fit the genre to which the novel belongs. The 25 translations also used the conjunctions en (and), toen (then), dat (that), voordat (before), terwijl (while), and hoewel (although) significantly more often. I suspect that the more frequent use of terwijl (while) in the translations has to do with the English participle constructions (such as being, walking, etc.), which in Dutch is very unusual and in translation is usually replaced by a subordinate clause, beginning with, for instance, terwijl. And hoewel (although) is probably the translation of the fairly frequent though or although in English. The word hoewel is particularly striking in this list, because for Dutch readers the word is archaic.1 I would conclude this is a good example of word usage that is less fluent from a Dutch perspective. For the more frequent conjunctions en (and), toen (then), dat (that), and voordat (before) in the translations, this conclusion is more difficult to draw. As with the more frequently used vroeg (asked), wist (knew), and zei (said), translators may have unconsciously added them to make the story easier to follow. This can be seen as a characteristic of a less fluent translation. Books in the Suspense category also had significantly more conjunctions in the translations from English than in the original Dutch books, which confirms that genre differences may play a role. For now, we can conclude that if we start with the idea that a good translation features fluent Dutch and is not recognizable as such, it seems that in some respects (the modal particles), translated literary novels in our corpus are slightly more fluent than novels labelled as Suspense or Romance. In other respects, however, genres show no differences (the conjunctions, especially terwijl [while] and hoewel [although]). To verify these observations I turned to another statistical approach that I used earlier in this chapter: principal components analysis (PCA) using the Stylo package for R. I use the visualizations of the results of the calculations for the first two components as an aid to see how the selected novels compare to each other. The largest difference is depicted on the horizontal axis, the first component, and the second largest on the vertical axis. Figure3.6 shows a visualization of a PCA of the thousand most frequent words. Titles printed in black (starting with E_) are translations from English and titles printed in grey (starting with N_) are written in Dutch. In this figure the Dutch novels and novels translated from English into Dutch are scattered across the graph, largely overlapping. This means that they cannot be clearly distinguished from each other in their use of 1 Hoewel is one of the words on the list of archaic words of the Genootschap Onze Taal, the Dutch language society, see https://onzetaal.nl/taalloket/modern-taalgebruik. RoManCe, susPense, and TRansLaTions 87 the thousand most frequent words. The picture changes, however, if we make a PCA of only the six conjunctions that in the keyword approach sketched above I found to be especially frequent in translations from English. Figure3.7 illustrates that novels written in Dutch and novels translated from English hardly overlap and occupy two clearly distinct areas in the graph. This confirms that there is a significant difference in the use of these six words in the originals and the translations. figure3.6: Literary novels by female authors written in dutch and translated from english, PCa (1000 most frequent words). novels translated from english are indicated with an e_ in front of the author name and shortened title and novels originally written in dutch are indicated with n_. The various groups have each been given their own shade of grey to make them easier to recognize. 94 The RiddLe of LiTeRaRy QuaLiTy Radway, Janice A. Reading the Romance: Women, Patriarchy, and Popular Literature. 1984. Updated ed., University of North Carolina Press, 1991. Slob, Marjan. Foute fantasieën: Kleine filosofie van de ontvankelijkheid. [Bad Fantasies: Small Philosophy of Receptivity, not available in English]. 2007. Updated ed., Rotterdam: Lemniscaat, 2017. Van Cranenburgh, Andreas. Rich Statistical Parsing and Literary Language. 2016. University of Amsterdam, PhD dissertation. University of Amsterdam Digital Academic Repository, https://hdl.handle.net/11245.1/cb936704-8215-4f47-9013-0d43d37f1ce7. Van Dalen-Oskam, Karina. Het raadsel literatuur: Is literaire kwaliteit meetbaar? [The Riddle of Literature: Can We Measure Literary Quality?] Amsterdam University Press, 2021, https://karinavdo.github.io/RaadselLiteratuur/. Van Rossum, Lisanne M. Towards Literary Narrative Perspective: The Relationship Between Narrative Perspective and Perceived Literariness. 2020. Utrecht University, research master’s thesis. Utrecht University Repository, https://studenttheses.uu.nl/ handle/20.500.12932/38257. Venuti, Lawrence. The Translator’s Invisibility: A History of Translation. 1995. 2nd ed., Routledge, 2008. 4. Literary Novels Written by Women Donoghue and Dorrestein Abstract: Literary novels written by female authors received lower scores for literary quality than those written by male authors. In this chapter, the literary novels written by women are analysed, starting with translations and then followed by Dutch originals. Cluster analysis using Burrows’ Delta, principal components analysis (PCA), and keyword analysis are applied to explore differences in word usage and word frequencies between highly rated and lowly rated novels in order to find out whether these may explain the ratings. Among the translations, Emma Donoghue’s Room and Elizabeth Gilbert’s Eat, Pray, Love receive special attention. Works by Dutch authors Margriet de Moor, Marente de Moor, Naima El-Bezaz, and Renate Dorrestein are described in more detail. Keywords: Literary quality, Emma Donoghue, Elizabeth Gilbert, Renate Dorrestein, Computational Literary Studies, Distant Reading The Glass Ceiling The previous chapter showed that literary quality in the eyes of the Dutch readers who participated in The National Reader Survey in 2013 was strongly linked to the perceived genre of a novel. Books labelled Romance in the research corpus received the lowest scores, those marketed as Literary novels generally received the highest scores, and Suspense sat in between. Continuing my quest in search of the conventions of literariness in the Netherlands in 2013, I will now focus on the category of Literary novels. As became clear in Chapter3, Literary novels by female authors generally received lower scores than those by male authors, and the fact that a novel was a translation may also have influenced readers’ ratings, sometimes negatively and sometimes positively. For these reasons, I will describe novels written by female and male authors in separate chapters. The first of these two chapters is dedicated to the Literary novels by female authors. I have included Dalen-Oskam, K. van, The Riddle of Literary Quality: A Computational Approach. Amsterdam: Amsterdam University Press, 2023 doi 10.5117/9789048558148_ch04 96 The RiddLe of LiTeRaRy QuaLiTy a novel co-authored by two women but excluded a novel by a mixed duo. After a brief glance at the position of female authors in the literary ranking, I will first zoom in on the 34 translations and then on the 23 Literary novels written in Dutch by women. The research corpus of 401 novels contained a total of 196 books with a female author. Of these, we labelled 57 as Literary novels, 83 as Suspense, 40 as Romance and 16 as Other. With all 401 novels ranked according to their mean score for literary quality, the ‘glass ceiling’ that still exists in Dutch society can be found between floor 31 and 32. The first 31 ranks are taken by Literary novels written by male authors, and the 32nd place has the highest ranking Literary novel written by a woman: Het grote huis (2010) by Nicole Krauss, translated from the (American) English Great House (2010) by Tjadine Stheeman and Rob van der Veer. Literary novels by female authors are mainly found between rank 32 and rank 136 – in Appendix 2 the complete list of books can be found, ordered from highest literary quality to lowest. Suspense written by women predominantly occurs between ranks 140 and 304. Romance, mainly written by female authors except for one title written by a mixed duo, ended up between ranks 331 and 400. Literary novels also occasionally occur lower down on the ranked list, and some titles from the Suspense and Romance categories can be found higher up than in the ranges I mentioned above. It is striking that the Dutch Suspense novel Dorp (*Village; the asterisk indicates no English translation is available and that the English title is my own provisional translation) by Anouschka Voskuijl is on rank 65, right in the middle of the Literary novels, and that we find the romantic Diner voor 2 (Turning Tables) by female duo Heather and Rose MacDowell on place 133, close to the Literary novels and still above most of the Suspense. Diner voor 2 was rated by 41 participants, which is somewhat below the minimum of 50 ratings that we would have liked to have for each book. Dorp was rated by 48 respondents, also just below the minimum. This slightly lower number of ratings may or may not have played a part in the exceptional position in the ranking, and therefore I will not be tempted to draw any conclusions about their ranking. A general glance at the list of the Literary novels written by female authors shows that the upper half contains more Dutch novels, and the lower half more translations from English. Translations from other languages can be found in different parts of the ranked list. Three novels translated from Spanish ended up in the top half, together with one translation from German and one from Italian. A second translation from German is on the negative side of the ranking. Three translations from French feature in the middle range, together with a set of translations from English. It is also striking that the four novels considered least literary are Dutch. Those who would like to explore the list in more detail are referred to Appendix 2 and to the litRiddle package for R (see Appendix 3). LiTeRaRy noveLs WRiTTen By WoMen 97 Nicole Krauss’ Het grote huis (Great House) narrates several life stories, in all of which the same antique writing desk plays a role. In The National Reader Survey, 483 readers rated the novel. The average of their scores for literary quality was 5.99 on a scale of 7. Most of them gave it a 6 (literary) for literary quality, followed by sevens (highly literary) and fives (somewhat literary) – the readers were unanimous in their praise (see Figure4.1). figure4.1: Ratings of nicole Krauss, Het grote huis (Great House) A total of 76 respondents presented us with some more information. Many of them referred to the language and the deeper layers in the novel to substantiate their ratings. What is striking about these motivations is that several participants reported that it was a book that begged to be reread. To quote one of them – those who would like to read the original Dutch can find it in the litRiddle R package through the ID number presented here by way of reference: ‘Wonderful writing style, beautiful story lines. A book to reread several times, then you get more and more out of it’ (ID_6381). Another reader wrote: ‘Read it in my book club and we discussed it in depth’ (ID_5366), perhaps suggesting that a book that lends itself to an in-depth discussion by definition has high literary quality. A final striking motivation I would like to quote is the following: ‘She clearly adopted her husband’s writing style, but to me literature is where good writing and a good story come together’ (ID_4077). The first part of this statement written by a female respondent seems rather accusatory – Krauss supposedly ‘nicked’ the writing style of her (ex) husband, Jonathan Safran Foer – author of, among other books, Everything is Illuminated and Extremely Loud and Incredibly Close. But the result is very good. 98 The RiddLe of LiTeRaRy QuaLiTy The respondent is positive, but not wholeheartedly so. Foer is not on our list, by the way. Another Literary novel translated from English ended up remarkably far on the negative side of the ranking. Shortly after its publication, the book was widely discussed and debated, but this apparently did not help it gain literary prestige. Eten, bidden, beminnen by Elizabeth Gilbert received an average score of 3.51 for literary quality. The full title of Elizabeth Gilbert’s book is Eten, bidden, beminnen. De zoektocht van een vrouw in Italië, India en Indonesië. It appeared in 2007 in the Dutch translation by Martine Jellema after the English Eat, Pray, Love: One Woman’s Search for Everything across Italy, India and Indonesia that was published in 2006. The publishing house of the Dutch translation labelled it as a translated literary novel. The book is presented as the author’s personal account of a year in which she tries to get her life back into balance. Just over six hundred participants in The National Reader Survey who indicated that they had read the book rated it on both literary and general quality, with widely varying scores, as Figure4.2 shows. When it comes to general quality – how good do you think this book is? – the scores selected most often were 5 and 6. For literary quality, 5 was also the most frequently given score, but the most frequent after that was a rating of 3. figure4.2: Ratings of elizabeth Gilbert, Eten, bidden, beminnen (Eat, Pray, Love) A total of 89 readers motivated their rating, all but four of whom were women. Below I present a small selection; for a list of all substantiations I again refer to the litRiddle package for R, suggesting to additionally use a translation tool to LiTeRaRy noveLs WRiTTen By WoMen 99 get a general idea about their nature in your own preferred language. The ratings are indicated before the quote, from score (1), definitely not literary, to (7), highly literary. The highest score was represented only once with a motivation; most of the substantiations were for low ratings for literary quality. (1) It simply is a true story of a woman in a personal crisis (ID_213). (1) The style was cringe-inducing (ID_8840). (1) What a boring, contrived, predictable book. I’m quite willing to empathize with other people, other perspectives on life, and other experiences, but this was so packed with platitudes that after doing my best for forty pages I diagonally read the rest at random. So I didn’t finish reading the book, which is very much against my habit (no matter how bad a book is) and it went straight to the second-hand bookstore in almost new condition! (ID_9474). (2) This book is nothing more than the lament of a woman in her midlife crisis. What a wimp that Elizabeth is (ID_7596). (2) Thought it rather resembled a Harlequin novel (ID_6310). (2) I liked the book a lot, it made me very happy, but it is written in a rather ‘straightforward’ way. I read it because of the interesting story, but not for deep insights or beautiful language (ID_4875). (3) Nicely entertaining book, not delving very deep (ID_4794). (3) The book has a straightforward style, it does not have any deeper layer. The language is not very good (ID_12986). (4) The story is predictable. It is not well written. Typical American whining story (ID_3802). (4) Actually, it is not a literary novel. It is an autobiographical work packaged as a novel. A cheap novel, which could not captivate me (ID_1289). (5) The main character grows as a person throughout the book. She actually understands less at the beginning than at the end (ID_9596). (6) Different from other novels. Especially because different aspects of cultures and religions are described (ID_13582). (7) The book is written in neat sentences and is at times profound (ID_12312). The motivations for their opinions about the literary quality seem to revolve around a genre aspect. Readers sometimes described Gilbert’s novel as non-fiction, as autobiographical, and therefore as not being literary. Repeatedly it is characterized as a women’s book, a ladies’ novel, a kitchen maid’s novel, or a Harlequin novel. Furthermore, the novel was mostly experienced as not profound, not beautifully written, and as not having any deeper layers. It was easy to read, chatty, a simple story. It was too popular to be literature, too American, and too superficial. And the narrator, Elizabeth, was having a midlife crisis and called simply a ‘wimp’. Is the distance between the 100 The RiddLe of LiTeRaRy QuaLiTy opinions about Krauss’ highly rated novel and those about Gilbert’s loathed book reflected in the texts themselves, analysed by means of text analysis tools? Translations Compared To find out how the 34 translated Literary novels written by female authors compare to each other in their vocabulary use I divided them into three groups: those with the highest mean scores for literary quality, those with the lowest scores, and the group in between. Group H: highest scores for literary quality – mean score Nicole Krauss, Het grote huis (Great House) – 5.99 Julia Franck, De middagvrouw (The Blind Side of the Heart) – 5.73 Isabel Allende, Het eiland onder de zee (Island Beneath the Sea) – 5.64 Isabel Allende, Het negende schrift van Maya (Maya’s Notebook) – 5.50 Emma Donoghue, Kamer (Room) – 5.45 Silvia Avallone, Staal (Swimming to Elba) – 5.45 María Dueñas, Het geluid van de nacht (The Time in Between) – 5.37 Alison Pick, Donderdagskind (Far to Go) – 4.97 Kathryn Stockett, Een keukenmeidenroman (The Help) – 4.95 Group M: middle scores for literary quality – mean score Jenna Blum, Het familieportret (Those Who Save Us) – 4.90 Hannah Shah, De dochter van de imam (The Imam’s Daughter) – 4.89 Tatiana de Rosnay, Het huis waar jij van hield (The House I Loved) – 4.89 Anita Amirrezvani, Dochter van Isfahan (The Blood of Flowers) – 4.89 Tatiana de Rosnay, Het appartement (*The Neighbour) – 4.82 Tatiana de Rosnay, Kwetsbaar (Moka) – 4.81 Sarah McCoy, De bakkersdochter (The Baker’s Daughter) – 4.77 Amanda Hodgkinson, Britannia Road 22 (22 Britannia Road) – 4.77 Jenna Blum, In tweestrijd (The Stormchasers) – 4.71 Margaux Fragoso, Tijger, tijger (Tiger, Tiger) – 4.71 Tatiana de Rosnay, Haar naam was Sarah (Sarah’s Key) – 4.67 Tatiana de Rosnay, Die laatste zomer (A Secret Kept) – 4.59 Sarah Blake, De laatste brief (The Postmistress) – 4.59 Nicci Gerrard, Het weerzien (The Middle Place) – 4.53 Janelle Brown, Alles wat wij wilden was alles (All We Ever Wanted Was Everything) – 4.23 Jodi Picoult, Negentien minuten (Nineteen Minutes) – 4.17 LiTeRaRy noveLs WRiTTen By WoMen 101 Group L: lowest scores for literary quality – mean score Joy Fielding, Roerloos (Still Life) – 3.88 Lynn Austin, Eindelijk thuis (Until We Reach Home) – 3.85 Elizabeth Gilbert, Toewijding (Committed) – 3.80 Jean M. Auel, Het lied van de grotten (The Land of Painted Caves) – 3.66 Monika Peetz, De dinsdagvrouwen (*Tuesday Women) – 3.59 Elizabeth Gilbert, Eten, bidden, beminnen (Eat, Pray, Love) – 3.51 Karen Kingsbury, Laatste dans (A Time to Dance) – 3.50 Emily Barr, Klasgenoten (Out Of My Depth) – 3.49 Karen Kingsbury, Nooit te laat (Redemption) – 3.47 As in Chapter3, I used the Stylo package for R (Eder et al.2016). My first experiment was a principal components analysis (PCA) of the thousand most frequent words in all 34 novels together. In the visualization of the measurement in Figure4.3, each group has been given its own grey scale, to make it easier to spot the major differences. More about the method can be found in Chapter3. In reading the graphs, the essence is that the closer books are positioned to each other, the more similar they are in their frequency distribution of the thousand most frequent words. Figure4.3 shows that the three groups overlap by a wide margin. This suggests that word usage cannot very well explain the differences in literary appreciation. It is noteworthy that the three Spanish novels are collectively positioned in the upper left quadrant, quite far away from the other novels receiving the highest scores for literary quality, which suggests to me that their shared Spanish origin is reflected in the style of the translations. Even when I selected other amounts of most frequent words or applied other kinds of measurements, the same novels ended up as relative outliers. In the group with the highest literary quality Emma Donoghue’s Kamer (Room) stood somewhat apart; I will examine this novel in more detail in the next section. In the middle group Sarah Blake’s De laatste brief (The Postmistress) was an outlier, and in the group with the lowest literary quality both titles by Elizabeth Gilbert, Toewijding (Committed) together with the book it was a sequel to, Eten, bidden, beminnen (Eat, Pray, Love), were outside of the core group. Toewijding, by the way, received a slightly higher average score for literary quality than Eten, bidden, beminnen: 3.80 as opposed to 3.51. I suspect that this was because those readers who gave Eten, bidden, beminnen very low scores did not read the sequel Toewijding, so that the mean scores based on the opinions of the repeat readers was slightly higher – the same pattern as described in Chapter3 for the three volumes of the Fifty Shades trilogy, where the second and third volumes received slightly higher scores for literary quality than the first one. As in Chapter3, I used the keyword function in Laurence Anthony’s AntConc software to find out which words occurred significantly more often than could 102 The RiddLe of LiTeRaRy QuaLiTy be statistically expected. For this, I left out the middle group and only compared the nine novels that readers found most literary with the nine novels that received the lowest average score for literary quality. This resulted in the observation that the novels with the highest literary quality contained relatively more personal pronouns indicating a first-person narrative (ik, mijn, me – I, my, me), and more auxiliary verbs in the present tense (zeg, zegt, kijk, kijkt – say, says, look, looks) than the novels that were awarded the lowest literary quality. Conversely, the nine novels with the lowest literary quality rating contained more personal pronouns ze (she/they) and haar (her), which may indicate a predominance of third-person narratives and possibly also a greater representation of female characters, but also more je (you/your, singular) and jullie (you/your, plural), which may indicate more dialogue. Furthermore, they showed many more past tense forms here: zei, figure4.3: Translated Literary novels written by female authors, PCa (1000 most frequent words). scores for literary quality: h_ (highest), L_ (lowest), and M_ (middle group). nicole Krauss, Het grote huis (Great House) hides among other books on the zero line to the left of the intersection (which in a way represents the average of the 34 novels). LiTeRaRy noveLs WRiTTen By WoMen 103 vroeg, keek (said, asked, looked). In the novels with low scores for literary quality we also see a larger share of the conjunctions toen (then) and dat (that; the use of some conjunctions was discussed in Chapter3). The text of the lowly rated Eten, bidden, beminnen (Eat, Pray, Love) agreed in more ways than one with the highest scoring novels. Two words that were characteristic for the novels with the highest scores can be traced back to the book that was a bit on the fringe in Figure4.3: Kamer (Room) by Emma Donoghue. They are ouwe (old), which invariably precedes the male personal name Nick in the novel, and wc (wc, toilet). The word wc occurs a total of ninety times in Kamer, which is a remarkably high number in the entire corpus of 401 books. Except for Een keukenmeidenroman (The Help) by Kathryn Stockett, where wc is found forty times, the other novels hardly refer to it. In the previous chapter, Kamer also ended up as an outlier in the use of the conjunction hoewel (although). In the next section I will look into this novel in more detail. Donoghue, Kamer (Room) Emma Donoghue’s novel Room was published in 2010. The Dutch translation Kamer by Manon Smits appeared on the market in the same year. The style of the Dutch version very closely recreates that of the original English text. To illustrate, I quote the opening sentences of the English version: Today I’m five. I was four last night going to sleep in Wardrobe, but when I wake up in Bed in the dark I’m changed to five, abracadabra. Before that I was three, then two, then one, then zero. ‘Was I minus numbers?’ ‘Hmm?’ Ma does a big stretch. ‘Up in Heaven. Was I minus one, minus two, minus three – ?’ ‘Nah, the numbers didn’t start till you zoomed down.’ ‘Through Skylight. You were all sad till I happened in your tummy.’ We are looking through the eyes of Jack, a five-year-old boy who all his life has never left the room in which he was born. His mother, the reader soon realizes, is kept prisoner in what looks like a small wooden shed. The shed has no windows except a skylight and is almost completely soundproof. Jack’s mother lived alone in the room for a very long time before Jack arrived. In Jack’s words and thoughts: ‘Then you wished and wished on your egg till you got fat.’ She grins. ‘I could feel you kicking.’ 110 The RiddLe of LiTeRaRy QuaLiTy As for the translated novels I started with a PCA based on the thousand most frequent words from all nineteen Dutch novels (Figure4.6). To make it easier to spot the titles from the three groups (highest and lowest literary quality and the middle group), they have each been given a different shade of grey in the visualization. I have explained PCA in more detail in Chapter3. As with the translated novels, the three groups of Dutch novels with different score ranges largely overlap, which means that differences in literary quality cannot be explained by the word frequencies that were examined. figure4.6: dutch Literary novels written by female authors, PCa (1000 most frequent words). scores for literary quality: h_ (highest), L_ (lowest), and M_ (middle group). The two almost completely overlapping novels below dorrestein’s De leesclub are Margriet de Moor, De schilder en het meisje (*The Painter and the Girl) and Marente de Moor, De Nederlandse maagd (The Dutch Maiden), the two novels with the highest mean scores for literary quality in this selection. LiTeRaRy noveLs WRiTTen By WoMen 111 Figure4.6 begs several questions about the relative outliers in the visualization. At the bottom of the graph we find Vinexvrouwen by Naima El Bezaz (1974–2020). In 38 short chapters, the book describes life in a ‘Vinex neighbourhood’ – a new neighbourhood with uniform and affordable housing intended for young families – in Almere, a satellite city for former inhabitants of Amsterdam. The first-person narrator Naima, her husband, and their two daughters are the first family of Moroccan descent in the neighbourhood, and this causes some unrest with the neighbours, but also with Naima herself. The book is not available in English; a literal translation of the title would be ‘Vinex Women’ but this will probably not be helpful to those not familiar with the Dutch Vinex concept. My provisional translation *Middle-Class Women in Modern Suburbia should be seen as a very rough approximation. In March2022, the translation database of the Letterenfonds (Dutch Foundation for Literature) did not mention any translations of any work of Naima El Bezaz yet, so below is a tentative translation of a quote from chapter24 (cf. Van Dalen-Oskam [2021] 162–163). It was Easter. And for the first time, I went to the Protestant church to attend the Easter celebration organized by my daughter’s school. I was as excited as my eldest and I wanted everything to be perfect. Muslim women wear headscarves when they go to the mosque. The Queen and others wear hats in church, I always saw that on TV when there was another wedding or christening. But I didn’t have a hat. My medicine supplier had some, but when we were trying on hats in her bedroom, we discovered that the circumference of our heads didn’t quite match. I needed something smaller, but where could I find a decent hat? My daughter admired herself in the mirror because she wanted to look beautiful, as she thought she should. The Christian school had clear rules, good standards, and values, so my eldest taught me a lot. Which words to use and which not. But now the hat. I found a number of them on Otto.nl. Some looked like the ones I saw on TV in a church. I ordered one and it would be delivered within a few days. The book, published as a novel, has obvious autobiographical traits and may be characterized as auto-fiction. The main character has the same name as the author, and the events described seem to have a high degree of realism, but how fictional they are remains unclear. The book was controversial and sparked outrage among readers. This is reflected in the motivations we received from participants in The National Reader Survey. One male reader who rated the book with a 3 for literary quality and a 6 for general quality wrote: ‘Because it’s a daily account of life there, without frills’ (ID_10043). Another male reader gave the book a 1 on both scales, stating: ‘No depth, stylistically very mediocre, content uninteresting. At the level of chit-chat’ (ID_12219). And to also give a voice to two female readers: a participant who gave the book a 1 for 112 The RiddLe of LiTeRaRy QuaLiTy literary quality and a 4 for general quality wrote that ‘[i]t is a light-hearted book, you don’t have to make any effort for it at all. But it is a nice book’ (ID_7103). Several women would rather classify the novel as belonging to another genre, including the following reader who gave it a 3 on both accounts: ‘In my opinion, the book had no deeper meaning. Besides, the book read more like a chick lit than a literary work’ (ID_3461). It is difficult to find out why this novel is a relative outlier compared with the other eighteen books in this group of novels. My hunch is that it may have to do with the auto-fictional nature of the book, and therefore that genre (or subgenre) plays a role here. Because there are too few comparable novels in the corpus, this possibility cannot be further explored. At the top of the graph in Figure4.6, as relative outliers, sit two novels by Simone van der Vlugt that may be characterized as historical novels. Here again, one might think that genre or subgenre – in this case the historical novel – may be responsible for differences in style that are reflected in vocabulary differences. However, there are several other historical novels in the selection: Susan Smit, Vloed (*High Tide), Josha Zwaan, Parnassia (*Parnassia), and also the two highest scoring novels by Margriet de Moor and Marente de Moor, the two novels that mostly overlap in the graph just below Renate Dorrestein’s definitely not historical novel De leesclub (*The Reading Group). This exception calls the genre explanation into question. Dorrestein is represented with two novels in this selection, which end up in different parts of the graph – a bit comparable to what we saw for the two literary novels written by Mankell in Chapter3. I will go into this in more detail in the next section of this chapter. What is also intriguing in the graph is the closeness of the novels by Margriet de Moor and by Marente de Moor. I have already introduced Margriet de Moor’s work, with special focus on De schilder en het meisje (*The Painter and the Girl), and presented some motivations for its ratings. I will now do the same for Marente de Moor’s work. So far, she has published six books, of which De Nederlandse maagd was the third. As of March2022, Marente de Moor’s work had been translated into sixteen different languages; four of her titles were available in German and two in Bulgarian, French, Italian, and Spanish. De Nederlandse maagd was the only novel in the set of Dutch Literary novels written by female authors that was available in an English translation. De Nederlandse maagd appeared in Dutch in 2010 and was awarded the prestigious Dutch AKO Literature Prize in 2011 and the European Union Prize for Literature in 2014. The English translation by David Doherty, The Dutch Maiden, was published in 2016. Eighteen-year-old Janna leaves Maastricht by train in September1936 to stay with an old friend of her father’s for a few months. Egon Von Bötticher lives in a village near Aachen and is a renowned fencing instructor (maître). Janna’s father is a doctor who met Von Bötticher in 1914 when he worked for the Red Cross as a LiTeRaRy noveLs WRiTTen By WoMen 113 medical aid worker on the battlefields just across the Dutch border. For reasons that are unclear to Janna, the relationship between the two men has long been strained. Janna came into contact with fencing at the 1928 Olympic Games in Amsterdam and has been enchanted by the sport ever since. Her father eventually gives her the opportunity to continue her studies at Von Bötticher’s, together with the beautiful identical twins Friedrich and Siegbert. Janna is completely fascinated by their instructor and his mysterious past, but the political situation in Germany in 1936 brings her stay in the house to a fateful end. Below I quote a longer excerpt from chapter9 of Part1 of the English translation, which carefully reflects the style of the original Dutch (cf. Van Dalen-Oskam [2021] 160). He [Von Bötticher] picked up a foil and stood in front of the mirror, poised to attack. As I had expected, his stance was perfect. All of the power in his tensed body was channelled into his elegant left hand, held effortlessly in mid-air as if his right held no weapon at all. Between the tip of his foil and the mirror there was no more than a breath, just enough to suggest that his reflection was an opponent of flesh and blood. A fraction closer and the illusion would have been shattered. ‘Two identical sabreurs, but trained differently. Your opponent becomes rattled: which one is he facing? He tries to respond exactly as he responded to your likeness, but the two of you fence differently every time. This unsettles him. Is this the sabreur who repeatedly braves his attacks, parrying nimbly, or is it the other, the one who always seems to be on the retreat…’ – without altering his stance he backed off at lightning speed – ‘… only to lunge forward and make good his position? Racked by doubt, he concedes a hit.’ It struck me as an absurd plan. Didn’t all fencers look pretty much alike when they were suited and masked? Besides, the names were announced before every match and opponents on both sides were changed at the same time. Nevertheless the twins were delighted. ‘Like Zorro!’ Friedrich cried. ‘No one knows Zorro’s true identity either!’ Von Bötticher frowned. He had no idea what the boy was talking about. Of the 629 readers who rated the book, 36 wrote a motivation for their score. A female reader gave the book the highest score, a 7 for both literary quality and general quality: ‘The style of writing, the sentence structure and the vocabulary. Also won a number of literary prizes, I believe’ (ID_1398). Another female participant gave it a 6 on both scales: ‘Unexpected, unusual theme, excellent writing style’ (ID_10970). A few people were considerably less positive; the following reader rated both literary as general quality with a low 2 and justified her decision as follows: ‘The style was stiff and long-winded. I found the book unreadable’ (ID_1201). The next extensive motivation was from a female reader who gave the book a 7 for literary quality and 114 The RiddLe of LiTeRaRy QuaLiTy a 6 for general quality: ‘You experience the development of the main character: you get to know her thoughts and feelings. Her doubts, but also her curiosity about the fencing teacher hold your attention. You increasingly feel the threat of an unrequited love. The book is written in beautiful Dutch: varied use of words, original way of phrasing. A language that is too beautiful for everyday life’ (ID_13197). In the visualization of the PCA based on the thousand most frequent words in Figure4.6 Marente de Moor’s De Nederlandse maagd (The Dutch Maiden) seemed to completely overlap with Margriet de Moor’s De schilder en het meisje (*The Painter and the Girl). One could hypothesize that the close relationship between the authors may have played a role: Marente is Margriet’s daughter. The Polish computational literary scholar Jan Rybicki has shown that the works of the English Brontë sisters always cluster together in similar stylistic analyses of nineteenth-century English literary fiction (Rybicki 235–236). When asked to explain this clustering during workshops he teaches on the Stylo package in R, Rybicki usually suggests this may have to do with the seclusion of the family in which the sisters grew up and shaped their writing careers. In 2021, this idea was confirmed by Rachel McCarthy and James O’Sullivan. To test whether the novels by Margriet de Moor and Marente de Moor resemble each other in diction because the authors have spent a lot of time in each other’s presence and have been influenced by each other in their work, I applied a cluster analysis in addition to the PCA. I used the Stylo Package for R and used Classic Delta as a measure; in the visualizations I made, the Delta score is shown on the horizontal axis. The statistical approach of the cluster analysis establishes pairs of most similar texts or clusters based on words and word frequencies. The further to the right in the graph two books are linked, the more similar they are in the frequencies of the words analysed. Cluster analyses based on different amounts of most frequent words presented a result that aligned with what Figure4.7 showed for the thousand most frequent words. However, in this different statistical approach compared to PCA, the novels of Marente de Moor and Margriet de Moor no longer closely resembled one another, but each clustered with one of the other novels. Still, the four of them were the most similar. Marente de Moor’s De Nederlandse maagd (The Dutch Maiden) most closely resembled another historical novel, Susan Smit’s Vloed (*High Tide). Together they cluster with the pair consisting of Margriet de Moor’s historical novel De schilder en het meisje (*The Painter and the Girl) and Renate Dorrestein’s De leesclub (*The Reading Group), the latter of which is definitely not a historical novel. What we can again conclude from this experiment is that genre is not a sufficient explanation for stylistic similarity. It is quite possible that the age of the authors also plays a role. Marente de Moor (born in 1972) and Susan Smit (born in 1974) are almost contemporaries. The other two women differ more in age: Margriet de Moor was born in 1941 and Renate Dorrestein in 1954. Whether age and generational LiTeRaRy noveLs WRiTTen By WoMen 115 differences do indeed influence writing style can only be confirmed when we conduct much more extensive research based on a much larger corpus of novels, so an answer to this question will not be available for now – and the same goes for the question if the family relation between Margriet de Moor and Marente de Moor has influenced similarities in their writing style. figure4.7: dutch Literary novels written by female authors, Cluster analysis (1000 most frequent words). scores for literary quality: h_ (highest), L_ (lowest), and M_ (middle group). Dorrestein in More Detail Renate Dorrestein (1954–2018) is mainly known as the author of novels but she was also active as a speaker and as a teacher of creative writing. Some of her non-fiction 116 The RiddLe of LiTeRaRy QuaLiTy publications dealt with writing, sharing her writing methods, and suggesting best practices to aspiring authors. Several of her novels are translated. According to the translation database of the Letterenfonds (Dutch Foundation for Literature), the number of receiving languages as of March2022 was seventeen. German topped the list with a total of fourteen titles, followed by Italian with eight, French and English both with five and Finnish and Swedish with three. The two Literary novels in our research corpus were not among the five novels translated into English. De stiefmoeder (*The Stepmother) is available in German and in Finnish; in the list of 401 novels ranked for literary quality, in our research corpus, it occupied the 64th rank with a mean score of 5.53. De leesclub (*The Reading Group) had not been translated; it can be found at rank 99 with a mean score of 4.98 for literary quality. In the PCA and the cluster analyses reported on and visualized in the previous section, De leesclub, published in 2010, invariably ended up in a different cluster than De stiefmoeder, published in 2011 (Figures 4.6 and 4.7). To explore their differences I will provide a very short summary of the two novels, select some motivations of readers’ opinions, and provide a very provisional translation of my own hand of a somewhat longer quote from the novels. De leesclub (*The Reading Group) is presented as a court hearing of which we only get the oral report of a reading group that went on a boat trip in Scotland with the Dutch writer Gideon de Wit. The reading group consists of seven middle-aged women who read a truly heroic number of books and always prepare their meetings with authors down to the last detail. The seven members Tillie, Jo (short for Johanna), Willemien, Annabel, Martha, Leonie, and Barbara together act as narrator, who always speaks of ‘we’ and ‘us’. The written account by this first-person plural narrator is interspersed with 45 footnotes of tourist information for the reader travelling in Scotland, compiled by Leonie. De Wit is the author of ‘one of the most talked-about oeuvres of this century and the last’, and is highly regarded and showered with awards, especially outside of the Netherlands. The women chose him as their tour guide for a boat trip themed ‘In the spirit of Moby Dick’. But that trip does not end well – the very first sentence of the novel suggests that the reading group ended up murdering de Wit. An important link in the chain of events is the following scene in the sixth chapter, in my translation (cf. Van Dalen-Oskam [2021] 167–168): [B]efore he [Gideon de Wit] embarked on his duties as a tour guide, he wanted to clear up a misunderstanding that had haunted him for years. He cast a meaningful look at each and every one of us. A misunderstanding that, according to some quacks, was even the main reason for his conversion to Islam: he would have claimed on several occasions that women could not write. Pure nonsense and slander. Such a statement was semantically impossible. In reality, he had only LiTeRaRy noveLs WRiTTen By WoMen 117 occasionally meekly suggested that in order to write well, one had to have balls. It was easy to spot the work of female authors. Books about trifles. Books full of nitty-gritty details. Books without weight and mass. Books without balls. That touches a nerve. The blood went to our heads. We had to bite our lips. An old nag. But perhaps it was just a question of taste, or perhaps a literary judgement was being passed that we as amateurs could never match. Nevertheless, our heart went out to the female authors we had met over the years in our series of literary encounters. We had devoured many of their ball-less books, they had sometimes changed our lives or at least shed new light on them. We had often found them to be wise women with humour and great psychological insight. We decided we had to stand up for them now. We had to defend the weight and mass of their work in no uncertain terms. We could not deny them without feeling indefinitely guilty, and didn’t guilt lead to penance? But would Gideon de Wit still take us seriously if we started talking about our beloved female writers?’ The designation ‘anti-roman’ (anti-novel) did not appear on the title page of the first print of De leesclub, but this is the label Dorrestein herself retrospectively attributed to the novel in the overview of her previously published works in her later books. She did not use that for any of her other works, as can be seen in the front matter of her 2017 book about writing that I will go into shortly. Of the sixteen motivations we collected about De leesclub in The National Reader Survey, I will quote two. This female reader gave the novel the maximum score of 7 for both general and literary quality: ‘A thin book with a lot of content. I read it twice in a row: the first time for the story, the second time for the rest. It leaves a lot of room for personal interpretation and imagination, not everything is told’ (ID_303). A completely opposite opinion can be found in the motivation of another reader: ‘Thought it such a stupid book and literary means that it has interesting layers and hands the reader something to take away. Humourless, boring, and totally uninteresting, I found it. I’m not a Renate Dorrestein fan’ (ID_2461). She gave the book a 1 for general quality and a 2 for literary quality. In De stiefmoeder (*The Stepmother) Claire Paagman goes on a solo trip for the first time in years. At the last minute, after a heated argument, her husband Axel does not come along. She travels to the English coast for an exhibition of her colourful quilts. The exhibition is organized by the Royal Quilting Academy, which has awarded Claire a prize that she has to collect with a speech. Shortly after arriving, Claire explores the town’s main shopping street and buys a Dracula cape for her stepdaughter Josefien. Back in her hotel she is interviewed by a female 118 The RiddLe of LiTeRaRy QuaLiTy journalist. I present a provisional, personal translation of a quote from chapter6 of Part1 ‘Claire’ (cf. Van Dalen-Oskam [2021] 171). The journalist fiddles with coffee, sugar, milk. She has remarkably graceful hands. At the same time, in a conversational tone of voice, she asks questions intended to break the ice. Crazy weather, isn’t it, Mrs Pacman? A heat wave like this is very rare here. Is this her first visit to the town, has she had time to look around? The old town is certainly worth a look. As is the river, with on the bank the inn where Bram Stoker once stayed. And, of course, the bench on the cliff where Stoker, looking out to sea, got the idea to land his Count Dracula right here. She nods, suppressing her boredom. It’s all in her guidebook. It’s strange how proud people in this seaside resort are that a fictitious monster washed ashore right there. But at least it inspires some snide remarks in the guise of literary allusions for her acceptance speech. She takes a sip of her coffee, which is of moderate quality. You’ll see that at least one photographer comes up with the idea of having her pose on Stoker’s bench today. Well, she can wear Josefien’s cape for the photo. And at this thought, out of the blue, she is overwhelmed by nostalgia for the time when she was casually rehearsing her stepdaughter’s French homework at the kitchen table. Well, Claire, I’ll call you Moon from now on. Followed by a screeching fit of laughter, with a lot of swishing of yard-long blonde hair. The place is never mentioned by name but it is clear Claire is in Whitby. During her stay Dracula will return in another form. In between acts, we learn about the origin of Claire’s marital problems, which turn out to revolve around a sensitive issue her stepdaughter brought up. In the second part of the novel we are presented with Axel’s perspective and in the final part with that of Josefien, Axel’s daughter and Claire’s stepdaughter. I quote the motivations of a couple of readers, again all women. A maximum score of 7 for literary quality and a 6 for general quality was motivated as follows: ‘I find Renate Dorrestein’s imagination highly original. Her writing style is concise, fast, surprising. She plays with language. Language seems to be an independent factor in her writing’ (ID_5426). The next reader was less enthusiastic. She gave the book a 4 for literary quality and a 5 for general quality: ‘Nothing more than a nice, smoothly written story’ (ID_3869). The novel received a rating of 5 for both general and literary quality from the following reader: ‘While the main character is staying abroad (and I think has just been abandoned by her husband or vice versa) she immerses herself in a river of thoughts and feelings. More is going on than appears at first glance and her personality also proves to be different from what the reader initially thinks’ (ID_11525). LiTeRaRy noveLs WRiTTen By WoMen 119 Renate Dorrestein not only wrote novels but also published her thoughts and experiences about writing, authorship, and being a female author in the Dutch literary landscape. In 2017, a big book titled Onder de motorkap van het schrijverschap (*Under the Bonnet of Authorship) brought together two earlier texts: Het geheim van de schrijver (*The Writer’s Secret) from 2000 and De blokkade (*Writer’s Block), from 2013. The first of these aimed to be a handbook for aspiring writers, a guide for readers to understand their own reading taste, and a kind of literary autobiography (2017: 12). The second described how Dorrestein suffered and overcame a writer’s block, which started right after De stiefmoeder (*The Stepmother) was published to great acclaim. The high praise she received for that novel was a big surprise to her and she could not explain why this novel finally brought her success (2017: 234, 280–281). This may also indicate that the genesis of the novel did not differ from that of earlier ones. Dorrestein described herself in Het geheim van de schrijver as a so-called ‘process writer’, someone who starts with an initially small idea. During the creative process of writing and rewriting that follows she sees which turns the story can take and finds out whether the developing manuscript has the potential to grow into a novel. She believed that the story is in fact already there, and that it must find her as a writer – and not that she was the one who consciously designed and structured the story. She experienced the search for the story that already exists somewhere as an intensive and exciting process, with an uncertain outcome. For some of her novels, it took ages before she was sure it would work. Sometimes she had to bin a concept, and other times it took dozens of versions before it finally became a good novel. But what makes a good and literary novel is not clear. Dorrestein was very much aware of the fact that it can take a while before a writer is read with pleasure and appreciation and that it is often unclear what exactly causes such a breakthrough. Has the work become better, or have the prevailing conventions changed? Dorrestein herself always explicitly stated that above all she wanted to tell a good story. She also often referred to the fact that for a long time it was actually a disqualification for a (female) author to be called a ‘storyteller’, but that recently, narrative prose gained considerable prestige (2017: 172). That is to say: literary prestige. Dorrestein emphasized that it is clear that many Dutch readers aspiring to be writers find literary quality important, but that it is not clear how exactly literary value is attributed (2017: 30–31). She summarized her view on how literary critics deal with the problem of literary quality as follows: The choice of subject does play a role. Some consider motherhood to be a completely acceptable literary theme; others believe that we may only speak of real literature when, following Hemingway, people are out on the streets at night in a novel, throwing up in gutters and visiting prostitutes. And apart from that, a 126 The RiddLe of LiTeRaRy QuaLiTy would like to explore the ranked list in more detail are referred to Appendix 2 and to the litRiddle package for R, see Appendix 3. The highest score for literary quality was reserved for Alsof het voorbij is (The Sense of an Ending) by Julian Barnes. Rank 2 up until 31 were also taken by Literary novels authored by male writers. Alsof het voorbij is was published in Dutch in October2011 and was translated by Ronald Vlek. The English original The Sense of an Ending also appeared in 2011, and in October of the same year received the prestigious Man Booker Prize for the best novel written in English and published in the United Kingdom or Ireland. The Dutch translation of the novel became available in the same month the prize was awarded, and it is very likely that Dutch readers were aware of the award and – consciously or unconsciously – took it into account in their assessment of the novel’s literary quality. The 817 received ratings resulted in an average score of 6.62 for both literary quality and general quality, where 7 was the highest score a reader could give to a book. In this, the participants showed a lot of agreement, as can be seen in the visualization of the distribution of the ratings in Figure5.1. figure5.1: Ratings of Julian Barnes, Alsof het voorbij is (The Sense of an Ending) Alsof het voorbij is (The Sense of an Ending) is narrated by the elderly Tony Webster. He describes the childhood friendship with his classmates Colin and Alex and later the gifted Adrian. But after secondary school they begin to drift apart. Adrian gets involved in a relationship with Veronica Ford, who used to be Tony’s girlfriend. Then the friends are informed that Adrian has committed suicide. Years later, Tony LiTeRaRy noveLs WRiTTen By Men 127 receives word that he has inherited a small sum of money and some documents from Veronica’s recently deceased mother. This reactivates Tony’s memory in an unexpected way; he must thoroughly revise his judgement of himself and others. One of the participants wrote: ‘This is the best book I have read in ages. It is very beautifully written. The subject matter is very surprising and very well developed. The book stays in your mind for a very long time’ (ID_8819). A total of 114 participants motivated their rating of the literary quality of the novel. Per usual, the writing style and the ‘deeper layers’ of the novel are frequently attributed as literary properties. This reader rated the novel with a 6 for literary quality: ‘The writing style and the fact that you can talk about it in a reading group’ (ID_5788) – the previous chapter referred to a similar motivation for the highest-ranking Literary novel by a female author, Het grote huis (Great House) by Nicole Krauss. The next reader also gave Barnes’ novel a 6 for literary quality, and was one of the few who combined a high score with a negative comment: ‘Expertly written fiction that, in this case, I actually don’t like’ (ID_1640); as for the other substantiations by respondents I quote the complete text this reader put in, so we do not get to know why the respondent did not like the novel. The vast majority of the readers who motivated their score of the book gave it the top score 7: ‘I found it a very intense book of which I enjoyed every sentence and that left a deep impression’ (ID_6627). ‘It is beautifully described how the main character finds out that his memories are faltering’ (ID_13162). ‘It is beautifully written, has a surprising storyline and when I finished it, I started to read it again’ (ID_9309). In the next sections of this chapter, a variety of other novels from this subcategory will be discussed in more detail, starting with translations. Translations Opposed To find out how the forty translated Literary novels by male authors compare to each other in their vocabulary and word frequencies, I divided them into three groups of roughly the same size: those with the highest mean scores for literary quality, those with the lowest scores, and the group in between. The borders between the three groups are arbitrary, because in contrast to the 34 translated Literary novels by female authors, high scores for literary quality predominate. Group H: highest scores for literary quality – mean score Julian Barnes, Alsof het voorbij is (The Sense of an Ending) – 6.62 Michel Houellebecq, De kaart en het gebied (The Map and the Territory) – 6.41 Haruki Murakami, Norwegian Wood (Norwegian Wood) – 6.31 Umberto Eco, De begraafplaats van Praag (The Prague Cemetery) – 6.24 128 The RiddLe of LiTeRaRy QuaLiTy David Mitchell, De niet verhoorde gebeden van Jacob de Zoet (The Thousand Autumns of Jacob de Zoet) – 6.19 Haruki Murakami, 1q84 (1Q84) – 6.19 Jonathan Franzen, De correcties (The Corrections) – 6.16 Jonathan Franzen, Vrijheid (Freedom) – 6.15 Laurent Binet, HhhH (HHhH) – 6.02 John Irving, In een mens (In One Person) – 5.88 Chad Harbach, De kunst van het veldspel (The Art of Fielding) – 5.88 Paolo Giordano, De eenzaamheid van de priemgetallen (The Solitude of Prime Numbers) – 5.88 Charles Lewinsky, De verborgen geschiedenis van Courtillon (*Saint John’s Eve) – 5.86 Group M: middle scores for literary quality – mean score Carlos Ruiz Zafón, De gevangene van de hemel (The Prisoner of Heaven) – 5.80 John Irving, De laatste nacht in Twisted River (Last Night in Twisted River) – 5.77 Charles Lewinsky, Het lot van de familie Meijer (Melnitz) – 5.72 Carlos Ruiz Zafón, Het spel van de engel (The Angel’s Game) – 5.66 Dave Eggers, Wat is de wat (What is the What) – 5.64 Irvin D. Yalom, Het raadsel Spinoza (The Spinoza Problem) – 5.63 Niccolò Ammaniti, Ik en jij (Me and You) – 5.61 Niccolò Ammaniti, Zo God het wil (As God Commands) – 5.56 Niccolò Ammaniti, Het laatste oudejaar van de mensheid (*The Last New Year’s Eve of Humanity) – 5.49 Paulo Coelho, De beschermengel (The Valkyries) – 5.48 Carlos Ruiz Zafón, De schaduw van de wind (The Shadow of the Wind) – 5.46 Karl Marlantes, Matterhorn (Matterhorn) – 5.42 Khaled Hosseini, Duizend schitterende zonnen (A Thousand Splendid Suns) – 5.41 Chris Cleave, Kleine Bij (The Other Hand) – 5.36 Group L: lowest scores for literary quality – mean score Niccolò Ammaniti, Laat het feest beginnen (Let the Games Begin) – 5.33 Lawrence Hill, Het negerboek (The Book of Negroes) – 5.32 Paulo Coelho, Aleph (Aleph) – 5.30 John Boyne, Het winterpaleis (The House Of Special Purpose) – 5.27 Ildefonso Falcones, De hand van Fatima (The Hand of Fatima) – 5.17 Henning Mankell, De Daisy Sisters (*Daisy Sisters) – 5.00 Henning Mankell, De geschiedenis van een gevallen engel (A Treacherous Paradise) – 4.92 Jonas Jonasson, De 100-jarige man die uit het raam klom en verdween (The 100-Year- Old Man Who Climbed Out the Window and Disappeared) – 4.81 LiTeRaRy noveLs WRiTTen By Men 129 George Mastras, Tranen over Kashmir (Fidali’s Way) – 4.79 Ken Follett, Val der titanen (Fall of Giants) – 4.57 Justin Cronin, De oversteek (The Passage) – 4.18 Nicholas Evans, De vergeving (The Brave) – 4.16 David Baldacci, Die zomer (One Summer) – 4.09 As in Chapter4 I started the text analysis experiments with several principal components analyses (PCA). I made use of the Stylo package for R (Eder et al.). Figure5.2 shows a visualization of the results of a PCA based on the thousand most frequent words in the texts. Each group has its own grey scale, to help to easily spot any major differences. More about the method can be found in Chapter3. The essence is that the closer books are placed to each other in the visualization, the more similar they are in the frequency distribution of their one thousand most frequent words. As for women’s Literary novels in translation and in Dutch, the three groups with the highest and lowest literary quality scores and the group in between show an uneven spread over the entire field of the graph. This means that here too, the difference between the most and least literary translations seemingly cannot be related to the use of the thousand most frequent words. There must be more to it than word choice and word frequencies alone. An outlier at the bottom of the graph is De hand van Fatima (The Hand of Fatima) by Ildefonso Falcones from the group with the lowest scores for literary quality. As of yet, it is unclear why this novel ended up in this isolated place. Falcones’ work has an exceptional size of 944 densely printed pages, but whereas it is commonly accepted as understandable that very short texts end up as outliers, the same is not true for very long ones. Still, when I redid the analysis with random samples of different lengths from all forty novels, the book was no longer an outlier (graphs not shown). Outliers at the top are Norwegian Wood (Norwegian Wood) by Haruki Murakami from the group considered most literary, and De Daisy Sisters (*Daisy Sisters) by Henning Mankell from the least literary group. In Chapter3 I paid special attention to the five novels by Henning Mankell in the corpus, and showed how De Daisy Sisters (*Daisy Sisters), one of his two Literary novels, deviated from the other Literary novels and the three Suspense books in a similar way. I hypothesized that this may be caused by the fact that this Dutch translation from 2009 goes back to a significantly older Swedish original from 1982. The same is true for Haruki Murakami’s Norwegian Wood (Norwegian Wood), which was originally published in Japanese in 1987 and was only published in a Dutch translation in 2009. The distance between the two novels by Murakami is much smaller than that between the two by Mankell. Whether the original publication date of the two novels is responsible for their outlier positions cannot 130 The RiddLe of LiTeRaRy QuaLiTy be said with certainty, but to me it does seem probable. More tests are needed, however, using a much larger corpus with representatives from different time periods. Next, I used the keyword function in Laurence Anthony’s AntConc tool to find out which words occurred significantly more in the novels with the highest scores for literary quality than those with the lowest scores, and vice versa. I excluded the middle group from this comparison. In the novels with the highest literary quality we see a relatively greater use of the personal pronouns ik (I), je (you), and mij (me), and of the present tense in the verb forms is (is) and heb (have). In the novels that were rated least literary, other personal pronouns were used more prominently: ze, which can be either a feminine singular (she, her) or a plural (they, them), hen (them), hij (he), and hem (him). The lowest-rated group featured some verb forms in the past tense much more frequently than expected: keek (looked), zag (saw), figure5.2: Translated Literary novels written by male authors, PCa (1000 most frequent words). scores for literary quality: h_ (highest), L_ (lowest), and M_ (middle group). LiTeRaRy noveLs WRiTTen By Men 131 antwoordde (replied), riep (yelled), and waren (were). Novels with the highest scores for literary quality seem to tend to be first-person narration due to their relatively larger share of ik (I) and mij (me). Another explanation may be that they have more dialogue because of the more prominent occurrence of je (you). The books that received the lowest scores for literary quality more often seem to be written in the past tense and as a third-person narrative. In the novels with the highest literary quality, the following words are used much more frequently than in the novels with the lowest literary quality: doch (yet), als (if), ook (also), of (or), and dus (so). Most of these are conjunctions. More prominent in the least literary group are de (the) and en (and). As such, there seems to be a significant difference in the use of conjunctions in the highest and lowest literary quality categories. The conjunction en (and) is more prominent in the novels with least literary quality, as opposed to of (or) and doch (yet) in the novels scoring highest for literary quality. The Dutch conjunction doch (yet) is archaic and therefore invites further inspection. 1 A closer look showed that it only occurs in nine of the forty novels, with a total of 323 occurrences. In eight of these it is only used once or twice, and the remaining 304 all come from David Mitchell’s historical novel De niet verhoorde gebeden van Jacob de Zoet (The Thousand Autumns of Jacob de Zoet). The novel about a Dutch bookkeeper on Deshima is set in Japan in the period 1799 to 1811, and was translated into Dutch by Harm Damsma and Niek Miedema. Below, I present the first two occurrences of doch (on the first two pages of the novel) in both translation and English original. Door het mousselinen gordijn heen fluistert de geneesheer Maeno: ‘Ik wilde mij er zelf van vergewissen hoe het kind ligt, doch’ – de bejaarde geleerde kiest zijn woorden met zorg – ‘dit schijnt niet gepermitteerd.’ Dr Maeno whispers through the muslin curtain. ‘I wanted to examine the child’s presentation myself, but…’ the elderly scholar chooses his words with care ‘… but this is prohibited, it seems.’ ‘Dokter Maeno, bent u het met me eens dat hier sprake is’ – ze gebruikt het Hollandse woord – ‘van een “dwarsligging”?’ ‘Wellicht,’ zegt de geneesheer in hun geheimtaal, ‘doch zonder nader onderzoek … ’ ‘Dr Maeno, would you agree the infant is in’ – she uses the Dutch term – ‘the “transverse breech position”?’ ‘Maybe,’ the doctor answers in their code-tongue, ‘but without an examination…’ 1 Doch is one of the words on the list of archaic words of the Genootschap Onze Taal, the Dutch language society, see https://onzetaal.nl/taalloket/modern-taalgebruik. 132 The RiddLe of LiTeRaRy QuaLiTy So it seems that the translators chose the somewhat archaic Dutch doch to represent the English non-archaic ‘but’. I suspect that Damsma and Miedema did this to compensate for comparable English archaisms elsewhere that could not be properly translated into equally archaic Dutch. In this way, the overall style of the translation still reflects that of the original. Dutch Originals Compared The corpus consisted of fifty male-authored Literary novels written in Dutch. Female authors had 23 novels in the list. Men therefore did considerably better on the Dutch bestseller list and in the ranked list of loans in Dutch public libraries that informed the selection of novels to be included in the corpus. And what is more: where at the time of writing of this book only one of the 23 Literary novels by women was translated into English, twelve novels by male authors were available in an English translation. This is a significant difference. As before, I have marked the titles I translated into English myself with an asterisk, as can be seen in the overview of the three groups below. Group H: highest scores for literary quality – mean score Erwin Mortier, Godenslaap (While the Gods Were Sleeping) – 6.60 Erwin Mortier, Gestameld liedboek (Stammered Songbook) – 6.54 Tom Lanoye, Sprakeloos (Speechless) – 6.37 A.F.Th. van der Heijden, Tonio (*Tonio) – 6.29 Stephan Enter, Grip (*Grip) – 6.26 J. Bernlef, Geleende levens (*Borrowed Lives) – 6.25 Tommy Wieringa, Caesarion (Caesarion) – 6.20 Gerbrand Bakker, De omweg (The Detour) – 6.19 Arthur Japin, De overgave (*Surrender) – 6.17 Thomas Rosenboom, Zoete mond (*A Mouth Full of Sweets) – 6.17 Thomas Rosenboom, Mechanica† (*Mechanics) – 6.16 Arnon Grunberg, Huid en haar (*Hair and Hide) – 6.16 J. Bernlef, De een zijn dood (*One Man’s Meat) – 6.14 Arnon Grunberg, Selmonosky’s droom† (*Selmonosky’s Dream) – 6.13 Arthur Japin, Vaslav (*Vaslav) – 6.11 Arnon Grunberg, De man zonder ziekte (*The Man Without Illness) – 6.11 Stefan Brijs, Post voor mevrouw Bromley (*Mail for Miss Bromley) – 6.09 Group M: middle scores for literary quality – mean score P.F. Thomése, De weldoener (*The Benefactor) – 6.09 Adriaan van Dis, Tikkop (Betrayal) – 6.07 LiTeRaRy noveLs WRiTTen By Men 133 J.J. Voskuil, De buurman (*The Neighbour) – 6.05 Arthur Japin, Dooi† (*Thaw) – 6.03 Bernard Dewulf, Kleine dagen (*Little Days) – 5.98 F. Springer, Quadriga (*Quadriga) – 5.95 Jan Siebelink, Het lichaam van Clara (*Clara’s Body) – 5.94 Maarten ’t Hart, Verlovingstijd (*Betrothal Time) – 5.86 Peter Buwalda, Bonita Avenue (Bonita Avenue) – 5.84 Kader Abdolah, De koning (The King) – 5.84 Tom Lanoye, Heldere hemel (*Clear Sky) – 5.83 Dimitri Verhulst, De laatste liefde van mijn moeder (*The Last Love of My Mother) – 5.82 Jan Siebelink, Oscar (*Oscar) – 5.81 Kees van Beijnum, Een soort familie (*A Kind of Family) – 5.76 Leon de Winter, Recht op terugkeer (*The Right to Return) – 5.75 Kader Abdolah, De kraai (*The Crow) – 5.51 Joost Zwagerman, Duel (*Duel) – 5.50 Group L: lowest scores for literary quality – mean score Leon de Winter, VSV of daden van onbaatzuchtigheid (*Acts of Selflessness) – 5.47 Ronald Giphart, IJsland (*Iceland) – 5.34 Remco Campert, Dagboek van een poes (*Diary of a Cat) – 5.33 Herman Koch, Het diner (The Dinner) – 5.12 Alex van Galen, Süskind (*Süskind) – 5.10 Herman Koch, Zomerhuis met zwembad (Summerhouse with Swimming Pool) – 5.09 Jaap Scholten, Kameraad Baron (Comrade Baron) – 4.97 Nico Dijkshoorn, Nooit ziek geweest (*Never Been Ill) – 4.79 Ravelli, De Vliegenvanger (*Fly Trap) – 4.75 Ernest van der Kwast, Mama Tandoori (Mama Tandoori) – 4.73 Ronald Giphart, Zeven jaar goede seks† (*Seven Years of Good Sex) – 4.65 Robert Vuijsje, Alleen maar nette mensen (*Decent People Only) – 4.46 Rik Launspach, 1953 (*1953) – 4.29 James Worthy, James Worthy (*James Worthy) – 3.79 Kluun, Haantjes (*Alpha Males) – 3.50 Mart Smeets, De afrekening (*The Reckoning) – 3.06 As for the Literary novels written in Dutch by female authors, the list contains four titles that consist of separately published stories. They are marked with a cross. I have not included them in the analyses I report on in this chapter, because their relative shortness would make them outliers, without providing any other useful information. 134 The RiddLe of LiTeRaRy QuaLiTy The Dutch-language Literary novel receiving the highest score for literary quality is Godenslaap (While the Gods Were Sleeping) by Erwin Mortier, presented as an old woman named Heleen’s retrospective on her youth during the First World War. It ended in second place on the literary ranking, right after Alsof het voorbij is (The Sense of an Ending) by Julian Barnes. The third place was also for Erwin Mortier, with his Gestameld liedboek (Stammered Songbook). This autobiographical novel sketches how the 66-year-old mother of the author suffers from Alzheimer’s disease. In short notes, ranging from one sentence or a couple of sentences to a few pages, Mortier records observations of his mother and her illness. He shows how more and more words disappear from her memory, and how her language ability fades away. The many blank lines on the pages visually echo the description of this process. The quote presented below from the English translation by Paul Vincent from around a quarter into the book shows a scene from the first stages of her disease (cf. Van Dalen-Oskam [2021] 199): I’d rather not imagine, but I do anyway: her walking doglike after him, the only person she still recognizes and on whom she fixates like a dove on the dovecote. Him first putting her on the sofa, picking up the vacuum cleaner again, but then giving up after she comes trailing after him for the umpteenth time. The macabre dance of the two of them behind the vacuum cleaner round the table, she getting her feet caught in the cable and falling to the floor. Her crying. The blood. His feeling of guilt. The doctor. I call back and ask: how can you stand it, I’m all in after a quarter of an hour. He says: It’s not that bad, I can take it.’ Most of the 576 participants in The National Reader Survey who rated the novel gave it the maximum score of 7 for literary quality (Figure5.3). We also received 105 substantiations of scores. A reader aged 28 wrote: ‘In a very beautiful way, with special phrasing, Erwin Mortier tries to make clear how he felt and what he observed during his mother’s Alzheimer’s’ (ID_12416). And a 73-year-old man stated: ‘Self-mockery and irony wrapped in Belgian humour. Mortier writes beautiful sentences. Belgians are often better at this than the Dutch’ (ID_4543). A man aged 58 pointed out: ‘Very subtle use of language, jumps in time, now and then it may be too much, especially when compared to Lanoye’s Sprakeloos [Speechless]’ (ID_9659). And this sixty-year- old man mentions Lanoye as well: ‘Because of Erwin Mortier’s very poetic use of language. Mortier knows, almost like no other, how to unlock a world of meaning in sparse text. Wonderful literature. (Just like his compatriot Tom Lanoye)’ (ID_4974). LiTeRaRy noveLs WRiTTen By Men 135 figure5.3: Ratings of erwin Mortier, Gestameld liedboek (Stammered Songbook) These motivations automatically lead us to a comparison with Sprakeloos (2009) by Tom Lanoye, which landed on the fifth place in the literary ranking. It was translated into English as Speechless, also by Paul Vincent. The subject is closely related to that of Gestameld liedboek (Stammered Songbook). Lanoye’s flamboyant mother, a well-spoken and enterprising amateur actress nearing her eighties, suffers a stroke that seriously impairs her speech. Lanoye writes about the personality of mother Josée, the relationship with her husband Roger, and what their last years looked like. Lanoye’s book is much longer than Mortier’s. The chosen form is clearly different, although, as already mentioned, the subject is very similar. The following quote from Sprakeloos towards the end of the book may well illustrate the difference with Gestameld liedboek. The author visits his mother in the institution where she has been admitted when she can no longer be cared for at home (cf. Van Dalen-Oskam [2021] 201). I enter her half-darkened room, prepared for the worst. All the same I am alarmed by the picture and the sound. She is sitting at the head end of the bed, with her back almost against the headboard and with a pillow on her lap that she is clinging to. She is rocking her body maniacally but softly backward and forward. It is her face that is most frightening. It is like a Greek mask, that of despair. She opens her mouth wide as can be and makes a quiet whining sound without crying and looks straight ahead, bewildered. With those deep furrows in her face, reinforced by the shadow of the curtains, and with her bulging eyes, her hollow cheeks and dishevelled hair she is reminiscent of a photo from the past: she as a judge, in 142 The RiddLe of LiTeRaRy QuaLiTy Herman Koch’s big breakthrough came in 2009, with his best-selling novel Het diner, which was also in our corpus of 401 novels. By March2022, Het diner was translated into 38 languages. Sam Garrett was responsible for The Dinner, the English translation that was first published in 2012. This book, which has also been turned into several movies with the same title, is about two couples who try to come to terms with the knowledge that their teenage sons have committed a serious crime. While on an international level Koch is mainly known as the author of literary fiction, the Dutch audience first got to know him as an actor and comedian in Jiskefet, a humorous, absurdist-satirical series (1990–2005) of Dutch public broadcaster vpro television that in some ways may be compared with the humour of Monty Python’s Flying Circus. This is something to keep in mind when reading motivations of ratings of Koch’s novels. Other works of both Koch and Kluun have also been translated, but Koch’s work is spread more widely than that of Kluun. A total of eight of Koch’s works have appeared in Italian, six titles are available in German, Finnish, Hebrew, and Russian, five in French, and four in Bulgarian, Danish, English, and Spanish. Koch’s next novel after Het diner was Zomerhuis met zwembad (Summerhouse with Swimming Pool), as of March2022 translated into 25 languages, clearly riding the wave of the success of Het diner. The global distribution of Kluun’s work is topped by four titles translated into Bulgarian, three into Czech, two into Complex Chinese (for Taiwan) and one into Chinese (published in Wuhan, in mainland China). Kluun’s Haantjes (*Alpha males) has until now only been translated into one language; it was published in Complex Chinese in 2012. Haantjes (*Alpha males) is all about Stijn van Diepen’s and Frenk Versteeg’s creative and strategic marketing agency Merk in Uitvoering – the name plays with the Dutch phrase werk in uitvoering (work in progress), so it signals something like ‘brand in progress’. Their one-year-old company is not making much progress in terms of assignments and income. They present themselves as new and hot and very much in demand, but according to their secretary Maud they mainly produce reports. Then, Frenk’s gay neighbour, the international model Charles, has a great idea. Soon the Gay Games will be held in Amsterdam. That will attract visitors from all over the world. And gays are big spenders. What if we replaced the colour red with pink in all the flags? And so the project Gay Flags is born. Stijn can already see the money coming in and also invests in the project himself. He informs his wife Carmen, nominated for the title Business Woman of the Year, only after he has put his money on the table. Stijn would have done better to involve her in the project much earlier… On Sunday 23January2011 the Dutch left-leaning quality newspaper de Volkskrant published an unexpectedly favourable review of Haantjes by literary reviewer Arjan Peters. Peters explicitly stated that he was negatively prejudiced when he LiTeRaRy noveLs WRiTTen By Men 143 started reading, mainly because of the vulgarity of Kluun’s earlier novels, but that all his objections disappeared while reading. Another influential literary reviewer, Arjen Fortuin, wrote extensively about both authors and both novels on 28January2011 in the quality newspaper NRC-next, comparing the novels with each other in both print runs and literary quality. Fortuin was also positive about Haantjes. He agreed with Peters that Kluun’s novel was moving away from his earlier ‘kitsch’ and suggested that Koch took the opposite route; from being an author with a specific literary audience in mind for high-quality literary novels, he became a writer for a much wider audience when he hit the jackpot with Het diner (The Dinner). Fortuin emphasized that although both Kluun and Koch have very different backgrounds, the print runs of their new novels were more or less in line with eighty thousand copies for Kluun and one hundred thousand for Koch. According to Fortuin, another similarity is that both novels in fact are ‘men’s books about being a man’. The main character in Herman Koch’s Zomerhuis met zwembad (Summerhouse with Swimming Pool) is general practitioner Marc Schlosser, married to Caroline and father of two beautiful daughters aged eleven and thirteen. Marc works in Amsterdam and he does things differently from other GPs. He spends twenty minutes per consultation – double of what other GPs do. Personal attention is his trademark, and it works. His patients are just as exclusive and include many celebrities. When actor Ralph Meier drops by without an appointment, this leads to further personal contact. Ralph invites Marc and his family to a barbecue, where they get to know Ralph’s wife Judith and their two sons. This leads to an invitation to visit the summer house Ralph and his wife have rented on the French coast during the next holiday. Feelings of discomfort that have arisen in the meantime come to fruition there, with irreversible consequences for Marc and his family. Literary critic Arjen Fortuin, in his piece for NRC-next, described the novel as a smooth read, funny, and not superficial, but all in all more of the same when compared to Het diner. Fortuin was somewhat disappointed with Koch’s Zomerhuis met zwembad. And although he was positive about Haantjes, he thought the literary battle was unequal: Fortuin rated Kluun as ‘playing a couple of divisions lower’ and marked Haantjes as very lightweight. To him, Koch seemed to have the best chance of winning both the literary duel and the overall number of sales. When we rank our corpus of 401 books according to number of sales and borrowings, the two novels ended up not very far apart: Koch’s novel at rank twelve and Kluun’s at seventeen. The complete list can be found on the companion website to this book (see Appendix 3) or using the litRiddle R package, sorting the books by their ID number. In The National Reader Survey, Koch’s novel received considerably more opinions than Kluun’s book (714 versus 208). Figures 5.6 and 5.7 show that Koch mainly 144 The RiddLe of LiTeRaRy QuaLiTy received scores of 5 or 6 for literary quality and that the opinions about Kluun’s book are much more evenly spread across scores 1 to 5. In the readers’ eyes, Koch clearly won the literary duel and Kluun definitely was the loser. They were less positively surprised by the new Kluun than the two literary reviewers mentioned earlier. figure5.6: Ratings of Kluun, Haantjes (*Alpha Males) figure5.7: Ratings of herman Koch, Zomerhuis met zwembad (Summerhouse with Swimming Pool) LiTeRaRy noveLs WRiTTen By Men 145 To gain some insight into this literary duel, I present a representative selection from the justifications readers gave for their scores. The ratings are indicated before the quote, from score (1), definitely not literary, to (7), highly literary. (1) Kluun: Poorly written, within three pages it is clear what is going to happen, very flat/vulgar characters (ID_4638). (1) Kluun: The story is based on the book Kaas [Cheese, by Willem Elsschot], so not very creative. Predictable storyline, not much depth (ID_1497). (1) Kluun: Spun-out anecdote not strong stylistically (ID_5961). (1) Koch: In my opinion, Zomerhuis met Zwembad [Summerhouse with Swimming Pool] is a typical example of a book written according to a certain concept. A book that was written purely to achieve a high position in the bestseller list. For me, literature is a text that contains an original voice, the drive to tell a story, originality in ideas, style, and use of language (ID_10693). (2) Kluun: Haantjes by Kluun has a bad story and is overrated by the public (ID_1575). (2) Koch: Plot is too simple and over the top. Popular (ID_12306). (3) Kluun: I liked Komt een vrouw bij de dokter [Love Life]. Haantjes [*Alpha Males] bored me (ID_4958). (3) Koch: I don’t like the style, the content is too predictable, and the characters lack depth (ID_6032). (4) Kluun: I like the way the book is written but I think the content is not very high brow (ID_3485). (4) Koch: Sometimes somewhat cheap effects and insufficient literary distance (ID_658). (5) Kluun: Badly written, probably meant to be literature, but does not meet my (subjective) standard regarding the quality of a novel (ID_11529). (5) Koch: Thinking required (ID_1824). (5) Koch: I don’t find that easy to answer. I like to read a good story. But the intensity of it, the impression the book leaves on me, is certainly important. A feeling of being part of the book. Well, difficult question (ID_10245). (6) Koch: Good story, developed with deeper thoughts. Interesting people who have to make choices in difficult situations (ID_7500). (6) Koch: Difficult, but in my opinion it is a stayer (ID_2745). (7) Koch: The possibility to identify and the recognizability of the character are developed to such an extent that the humour becomes almost visual (ID_9251). (7) Koch: Linguistically good. Tension is built up well, very humorously written. Good story line. Continuously captivating and interesting (ID_7844). The motivations show that Kluun’s novel was considered too anecdotal and popular in style, whereas Koch’s was perceived as having a literary style and deeper layers. 146 The RiddLe of LiTeRaRy QuaLiTy Several motivations are strongly reminiscent of those found in the previously described reactions in the media. It is certainly not unthinkable that literary reviewers were actually responsible for some of the motivations above. We do know that some participants of The National Reader Survey actually were literary reviewers, because some of them explicitly mentioned this in their motivation of a literary rating. To get a clearer view on the differences between the two novels themselves, I compared them with four other books, two with high scores and two with relatively low scores for literary quality: the literary top scorers Alsof het voorbij is (The Sense of an Ending) by Julian Barnes and Godenslaap (While the Gods Were Sleeping) by Erwin Mortier, and the much less literary Shopaholic & Baby (Shopaholic and Baby) by Sophie Kinsella and Nooit ziek geweest (*Never Been Ill) by Nico Dijkshoorn. Before you take a look at Figure5.8, have a taste of the style of the novels and predict the outcome… The first quote is from Chapter17 of Part I of Kluun’s Haantjes (in my own translation). The second is from Chapter11 of Koch’s Zomerhuis met zwembad in Sam Garrett’s translation (cf. Van Dalen-Oskam [2021] 223–224). Maud came in with the boy. He introduced himself as Hylke Strijpstra and apologized for his father’s last-minute absence (‘a big order in Dubai’). Because of the bad weather he was a bit late. Frenk said that he wanted to save the part of the presentation of Merk in Uitvoering until next time, when father Strijpstra would be present. He wanted to use this conversation to gather the information we needed to make the right decision. Mr Strijpstra’s youngest son embarked on his story. The Dokkumer Vlaggencentrale [The Flag Factory in Dokkum] had existed since 1797. They were the largest supplier in Western Europe. Of all the flags that we saw in the Netherlands, both commercial and private, over three quarters came from the Dokkumer Vlaggencentrale. Even the large champion’s flags hanging from the ridge of the Arena came from their company. With a wave of his hand Frenk silenced Mr Strijpstra jr. Frenk believed that the Dokkumer Vlaggencentrale could make flags like no other. What else did they supply apart from flags?’ ‘Daddy…’ I felt a hand on my elbow and turned to look. There stood Julia with the languid boy who had shaken my hand earlier, but who I had already forgotten was called Alex. Standing slightly behind them were two other boys and two girls. ‘Can we go out to get some ice cream?’ she asked. ‘It’s really close.’ In terms of timing it was both a good and a bad moment. There was a chance that the slight sultry edge to our – superficially – innocent conversation about teenage bedrooms, baby seal posters, and horse books might be lost for good. LiTeRaRy noveLs WRiTTen By Men 147 On the other hand, here I stood with my thirteen-year-old daughter, living proof that this charming man – me – was capable of siring a child. And not just any child, but a dreamy-eyed blonde who threw fifteen-year-old boys’ hormones into overdrive the moment they saw her. I won’t try to deny it: I take pleasure in being with my daughters in places where everyone can see us together. figure5.8: Kluun, Haantjes (*Alpha Males), versus Koch, Zomerhuis met zwembad (Summerhouse with Swimming Pool), Cluster analysis (1000 most frequent words). Compared with Barnes, Alsof het voorbij is (The Sense of an Ending), Kinsella, Shopaholic en baby (Shopaholic and Baby), dijkshoorn, Nooit ziek geweest (*Never Been Ill), and Mortier, Godenslaap (While the Gods Were Sleeping). The cluster analysis of which the visualization is shown in Figure5.8 was done with the Stylo package for R and uses as input the thousand most frequent words in all six novels. The further to the right books are connected by a vertical line, the more similar they are in word use and word frequencies. The result is that Haantjes and Zomerhuis met zwembad are much less distant from each other than readers’ opinions would suggest. They are both close to the novels Alsof het voorbij is by Julian Barnes and Godenslaap by Erwin Mortier, which are highly regarded for 148 The RiddLe of LiTeRaRy QuaLiTy their literary quality, and in this cluster of four clearly distinguish themselves from Shopaholic & baby by Sophie Kinsella and Nooit ziek geweest by Nico Dijkshoorn, which are considered to be far less literary. The measurements in this small selection of novels may support the idea that Kluun’s novel was undervalued by readers. So the unexpectedly positive reviews that Haantjes received in prestigious newspapers such as NRC-next and de Volkskrant did not boost Kluun’s low literary prestige. Grunberg and Romance Arnon Grunberg has been one of the most prolific and visible literary authors in the Netherlands since his debut novel Blauwe maandagen (Blue Mondays) appeared in 1994. He churns out novel after novel, writes newspaper columns, and is visible on television and at all kinds of literary events. Grunberg has received many literary awards for individual novels and in 2011 was awarded the prestigious Constantijn Huygens oeuvre prize, followed in 2022 by the equally prestigious P.C. Hooft oeuvre prize for prose fiction. In March2022 the translation database of the Letterenfonds (Dutch Foundation for Literature) mentioned translations of Grunberg’s work into thirty languages. Topping the list was German, with a total of eighteen translated titles, followed by French (11), Hungarian (11), English (10), Italian (9), Czech (8), Portuguese (7) and Turkish (6). Grunberg’s internationally most well-known novel was Tirza (2006), translated into 21 languages. Sam Garrett was responsible for the English version, Tirza, published in 2013. The two novels in our corpus have not been translated into English yet but are available in four (Huid en haar – *Hair and Hide) and six (De man zonder ziekte – *The Man Without Illness) other languages than Dutch. In Chapter3 I showed how Grunberg’s two novels end up between the Romance novels when I compared the most literary and least literary novels in our corpus using text analysis methods. In this section I will try to explain this. To give you a taster, I will start with a quote (cf. Van Dalen-Oskam [2021] 226). ‘What are you waiting for?’ asks Lea. She is wearing a black wool coat with a fur collar, bought second-hand. She cannot afford such coats new. Lea travels lightly. One backpack is all she needs for five days. A hairdryer gets most of the wrinkles out of your clothes. On her knee is a hand. But a hand on a knee is not intimacy. ‘What exactly are you a connoisseur of?’ A professor had asked her during a standing reception that evening, seemingly casually touching her upper arm. She had found it unpleasant. The question and the touch. LiTeRaRy noveLs WRiTTen By Men 149 An hour before, she had hung her dress in the bathroom on the rail to which the shower curtain was attached and treated it with the hair dryer. The wrinkles did not come out as well as she had hoped. Tomorrow morning she will go home, then she can have her dress dry-cleaned. These are the first lines – in my tentative translation – of Huid en haar (*Hair and Hide) by Arnon Grunberg. Roland Oberstein is writing a book about the economic bubble, and has a pretty complicated life. He is divorced and his son lives with his ex-wife, he has a girlfriend in Amsterdam, and currently works in Fairfax, in the United States. At a conference in Germany on the Holocaust – a subject he says he also finds interesting from an economic perspective – he meets the American scholar Lea, who is writing a book about Rudolf Höss, commandant of Auschwitz. Lea’s marriage has been without sexual intimacy for some time. For the first time since ages she is looking for a new sex partner and manages to seduce Roland. Back in the Netherlands, Roland allows himself to be seduced by a female student. Meanwhile, Lea’s husband has fallen fatally in love with an illegal Guatemalan male migrant, whom he forces into sex under the promise of a green card for him, his wife, and his child. The entanglements between all these characters become more and more complicated and ultimately end badly for some of them. A participant wrote about Huid en haar, giving the novel the highest score, 7, for both general quality and literary quality: ‘Grunberg knows how to draw you in and make you feel disgusted (about humanity?) in a way that is extremely clever. It does not make you happy’ (ID_6868). The next reader gave the same rating and motivated it in these words: ‘A lot of depth, also in terms of characters. Fascinating, very creative writing style. A game with language/structure/characters’ (ID_4039). Another participant also emphasizes the importance of style, scoring 5 for general quality and 6 for literary quality: ‘Because of the writing style. Grunberg does not necessarily accommodate the reader, he does not write a simple entertaining story’ (ID_3363). There are very few exceptions to the praise. ‘I think it’s lazy writing’ (ID_9814), wrote a reader who gave the novel a 4 for general quality and a 5 for literary quality. Grunberg’s other novel in our corpus, De man zonder ziekte (*The Man Without Illness), is about young Samarendra – usually ‘Sam’ – Ambani, who was born and raised in Switzerland, has a Swiss mother and a father who comes from India. He has a girlfriend, Nina, and works as an architect. He took part in a competition to design an opera house in Baghdad and was among the three finalists. At the invitation of Hamid Shakir Mahmoud, the man who organized the competition, Sam travels to Baghdad. Mahmoud personally guarantees Sam’s safety and Sam is looking forward to the meeting – which never happens. In Iraq, Sam has some gruesome experiences and his life completely changes, but his calling to design 150 The RiddLe of LiTeRaRy QuaLiTy buildings does not disappear. And when he has to go to Dubai for a new project, he even looks forward to it. That visit ends even worse, however. This novel begins as follows, again in my translation (cf. Van Dalen-Oskam [2021] 227): For his trip, Samarendra Ambani and his girlfriend bought a new suitcase. A grey suitcase with wheels. Because Samarendra’s girlfriend was afraid that he would not recognize his suitcase on the luggage belt, she tied a light green hair ribbon around the handle. He would have preferred not to have that hair ribbon. Samarendra, who is called Sam by most people, likes to come across as a professional traveller, someone who has been almost everywhere and therefore feels at home almost everywhere. Such a hair ribbon was more suited to a middle-aged tourist with homesickness and barely suppressed fears. But he did not want to disappoint his girlfriend. She had said: ‘Then you will always think of me when you see your suitcase.’ From the substantiations of ratings of De man zonder ziekte I quote two that mention the writing style in different ways. This reader gave the novel a 7 for both general and literary quality: ‘In this book I found his writing style very compact, he tells the story without unnecessary digressions, it was an easy read’ (ID_12100). The following somewhat longer motivation also refers to Grunberg’s use of language: ‘The structure, plot, and outcome of the story, and the suspense, the use of language, the contrasts between the main character, his personality, and the situation he finds himself in are described in such a way that you constantly want him to reconsider things. It keeps you active, you want to write along, to warn. I think it is literary if you can use language to bring that about in your reader’ (ID_468). This reader rated the general quality with a top score of 7 and the literary quality with 6. Also for this novel very few readers were not enthusiastic at all. The following participant gave the book a 1 for general quality and a 3 for literary quality: ‘The story is utter rubbish. A seriously overrated novel’ (ID_5671). Readers found the style of Arnon Grunberg’s two novels easy or enjoyable to read, found the use of language creative and compact, and appreciated this positively – these were the arguments for a high score for literary quality. None of the Grunberg readers who took part in The National Reader Survey and shared their opinions about one of his novels referred to Romance. The software did, as I already pointed out. Why? For a taste of this genre I again selected Shopaholic en baby (Shopaholic and Baby) by Sophie Kinsella that I used in a previous experiment in this chapter. The Dutch translation was produced by Mariëtte van Gelder. I quote the first paragraphs from the first chapter of the English original to give an impression of the writing style. Rebecca Brandon is pregnant and has absolutely no idea what to expect: LiTeRaRy noveLs WRiTTen By Men 151 OK. Don’t panic. Everything’s going to be fine. Of course it is. Of course it is. ‘If you could lift up your top, Mrs Brandon?’ The sonographer has a pleasant, professional air as she looks down at me. ‘I need to apply some jelly to your abdomen before we start the scan.’ ‘Absolutely!’ I say without moving a muscle. ‘The thing is, I’m just a teeny bit… nervous.’ I’m lying on a bed at the Chelsea and Westminster hospital, tensed up with anticipation. Any minute now, Luke and I will see our baby on the screen for the first time since it was just a teeny blob. I still can’t quite believe it. In fact, I still haven’t quite got over the fact that I’m pregnant. In nineteen weeks’ time, I, Becky Brandon, née Bloomwood… am going to be a mother. A mother! Luke’s my husband, by the way. We’ve been married for just over a year and this is a one hundred per cent genuine, honeymoon baby! We travelled loads on our honeymoon, but I’ve pretty much worked out that the baby was conceived when we were staying in this gorgeous resort in Sri Lanka called Unawatuna, all orchids and bamboo trees and beautiful views. Becky works as a personal shopper at a new prestigious department store, The Look, which is on the brink of collapse due to a lack of business/patrons, so she has all the time in the world to shop privately, looking for the perfect baby clothes, bedding, cots, pushchairs, and whatnot. Everything has to be of the highest fashion standard – and that includes the person who will guide her through the pregnancy and deliver the baby. As soon as Becky hears of Venetia Carter, ‘the must-have celebrity obstetrician’, she knows what she wants. But not everything in Becky’s pregnant life goes according to plan. Yet with her creativity and personality, she manages to solve problems in an impressive way. Here too, I quote a couple of substantiations written by readers who rated the book. ‘It is no literary masterpiece, easily written, but with a lot of humour’ (ID_13068) wrote a reader who gave the novel a 4 for both general and literary quality. Another one gave the book a 6 for general quality and a 2 for literary quality and goes a little deeper into the use of language than the previous participant: ‘It is written as an easy read, so easy words, sentences not too long. That’s not a bad thing at all, but not literary’ (ID_8052). A final quote, from a reader who gives the novel a 3 for overall quality and a 1 for literary quality: ‘Too many colloquialisms, implausible events, too many “easy” storylines that are predictable, chewed-out explanations.’ The extensive motivation continues: ‘This book is one of a series and all the books are structured the same, but with a different execution. No multiple layers in the book. What you read happens, no more and no less. So, not very original and just plain reading’ (ID_12557).