‘Could be the translation, of course’. Analysing the perception of literary fiction and literary translations

Abstract

The advent of information technology has greatly expanded the methodological toolkit of literary scholars. In the digital humanities project The Riddle of Literary Quality, methods from literary studies, linguistics and computer science are combined with sociological approaches. I will describe how this project went about mixing and blending methods from different research disciplines in search of more knowledge about the modern-day Dutch and European literary systems.

1 Introduction

The advent of information technology has greatly expanded the methodological toolkit of literary scholars. In the digital humanities project The Riddle of Literary Quality, methods from literary studies, linguistics and computer science are combined with sociological approaches. I will describe how this project went about mixing and blending methods from different research disciplines in search of more knowledge about the modern-day Dutch and European literary systems. My special focus is on translations of modern fiction and how they are valued by readers in comparison with fiction written in their native language. I will conclude by discussing how the greatly changed content of the toolbox may influence the planning and execution of a literary research project and what this could imply for the future of literary studies in general.

The Riddle of Literary Quality (2012–2019) combines a computational analysis of low-level (i.e. easy to measure) linguistic features and high-level (i.e. much more difficult to measure) linguistic features¹ with the results of a reader survey, in order to unlock part of the mystery that surrounds ‘literary quality’ as experienced by both general and specialist readers. The aim is to uncover what the current conventions of literariness are in the Netherlands, predominantly using methods that enable verification and replication of the results. In addition to myself as the principal investigator/project leader and temporary collaborators from various areas, in the earlier phase of the project (2011–2016) the team also comprised two PhD students and a scientific programmer funded by the Royal Netherlands Academy of Arts and Sciences (KNAW) in its Computational Humanities Programme (2011–2016). The project is now in its final phase.²

The corpus that we selected for our research contains the 401 best-selling or most often borrowed (from public libraries) novels in the period 2010–2012 in the Netherlands, in order to ensure that we could gather enough opinions per book for statistical observations. As we wanted to exclude as much as possible the influence of long-standing ideas about which titles belong to the literary canon and which do not, we selected novels that were generally not published in Dutch before 2007. Our selection criteria led to a corpus that consisted not only of high-brow literary novels, but also of many popular novels that until recently were not often considered as objects of research by students of literature, such as fiction for young adults and suspense, romance and fantasy novels. Another unusual choice was to include novels translated into Dutch, as long as they matched the selection criteria. Translations play a very important role in the Netherlands. In our corpus, 152 titles were written in Dutch and 249 were translated into Dutch and published in Dutch for the first time in 2007–2012. Of these 249 novels, 180 were translated from English. The remaining 69 were translated from nine other languages. One of the project’s research questions was whether there are any differences in how Dutch originals are valued on the scale of literariness compared with translations into Dutch.

Readers opinions were gathered in a large online survey, the National Reader Survey (Het Nationale Lezersonderzoek), which ran from March to September 2013. Almost 14,000 respondents completed the survey. They provided some personal information and answered questions relating to their reading habits, and indicated which of the 401 novels they had read. We then presented a randomly selected list of seven of the books they had read and asked them how good they found each of these books on a scale of 1 (very bad) to 7 (very good), and how literary they found the book on a scale of 1 (not literary at all) to 7 (highly literary). Respondents were then offered the option to rate seven more titles. For one of the books, we asked them to account for their score on the scale of literariness in the form of a written motivation. For seven books a respondent had not read, we asked for a comparable rating. We did not ask them to motivate their score for a book they did not read.

2 Research Questions

Dutch literature in its international context is the topic of a recently published volume titled Doing Double Dutch,³ which focuses on how Dutch literature has been and is being transferred to other language areas through translations and adaptations. Translators are key players in this kind of cultural transfer and several chapters deal with specific literary works and how they have been translated into different languages and cultures using slightly different approaches.

In the Netherlands, 65–70 per cent of all fiction titles are translations,⁴ mostly from English. But English-language areas are not the most important receivers of translations from Dutch. Most scholars contributing to Doing Double Dutch agree that it is the German language area that is most important and often also the means by which interest in Dutch literary works is generated in yet other language areas. It is well known that interest in other literatures is neither mutual nor mirrored, and that imbalance is one of its key characteristics. There is a clear hierarchy in the ‘world republic of letters’,⁵ wherein English currently occupies the centre and Dutch literature is one of the many literatures on the periphery. The contributors to Doing Double Dutch repeatedly state, however, that cultural transfer can only take place when the periphery is somehow present in the centre, and vice versa. The methods they use are mainly qualitative by nature, and they do not make use of interviews, surveys, statistical approaches or computational stylistics (stylometry)—all methods that play a role in The Riddle of Literary Quality.

Against this backdrop I will first focus on Dutch readers. Using the results of the National Reader Survey, I will address the following three research questions:

• 1. How aware are Dutch readers that they are reading a translation?

• 2. Do Dutch readers give translated novels different scores compared with novels written in Dutch?

• 3. Do Dutch readers use different criteria when accounting for their appreciation of translated fiction compared with fiction written in Dutch?

The answers to these three questions lead to a new question: do we see the same patterns in other countries, or are there differences? To answer that I will shift my focus to a larger geographical area and describe an international experiment involving the Netherlands, Germany, France and the United Kingdom and summarize its preliminary findings. These findings will lead back to The Riddle of Literary Quality and a fourth research question, which will for now be addressed only very summarily:

• 4. What does a stylometric comparison of translated novels and novels written in Dutch reveal?

3 The National Reader Survey

The almost 14,000 respondents who completed the National Reader Survey are representative of the Dutch reading audience. Just over 70 per cent of the respondents are female and about the same percentage have a high level of education (university or higher vocational education); however, the men on average are slightly higher educated than the women. These are the people who read, buy novels, join reading groups and make use of libraries. The respondents also include professional readers, such as reviewers. We know this because they sometimes mention this in the open field that asks them to account for one of their ratings.

3.1 Translation Awareness

How aware are Dutch readers that they are reading a translation? Participants in the National Reader Survey were asked to account for their rating on the scale of literariness of a randomly selected book to which they had given a relatively high or a relatively low score in comparison with the other books they had rated. Some of the explanations they gave suggest that readers are very aware of whether they are reading a translation. Some respondents indicated they had read a book in the original language:

Heb het in het Frans gelezen. Prachtig taalgebruik en actuele problematiek zeer origineel en tussen de regels weergegeven.

Read it in French. Beautiful language use, current issues, very original and ‘between the lines’.

Personen worden nogal vlak neergezet. Stijl is matig. (In het Engels gelezen) Veel herhalingen.

Characters are rather flat. Style not very good (read it in English). A lot of repetition.

Respondents probably often gave their opinion of a book based on the original and without having read the Dutch translation. We do not think this is problematic for our survey—we were happy with an opinion even when it was based on the original. Some respondents addressed this issue explicitly, commenting on the quality of our survey:

Ik mis de vraag of ik het in het Engels heb gelezen: ja Julian Barnes schrijft fraaie, korte zinnen. De opbouw van al zijn boeken is systematisch de verhaallijn(en) is/zijn duidelijk. Zijn Engels is zeer goed te begrijpen in tegenstelling tot bijv. Coetzee die veel 'gemaakter' schrijft. (…)

You did not ask if I have read it in English: yes. Julian Barnes writes beautiful, short sentences. All his books are well structured and the plot line or lines is or are clear. His English is easy to understand, compared with for instance Coetzee, who has a much more ‘artificial’ style. (…)

Literair zegt voor mij iets over het taalgebruik (daarom heb ik dit boek dan ook in het Engels gelezen). Soms is een boek in de oorspronkelijke taal voor mij literair, terwijl de Nederlandse vertaling dat absoluut niet is!! Slechte vertaling of matige vertaling kan van een literair boek een flodder maken. Dat betekent dus ook dat ik uw insteek van deze enquête niet goed vind. Ik wil namelijk niet een vertaler op zijn literaire waarde beoordelen, maar wel de schrijver. En ik weet van veel anderen dat zij zich ook vaak ergeren aan slecht vertaalde boeken.

To me, ’literary’ has to do with language use (that’s why I read this book in English). Sometimes, I find a book literary in its original language, while the Dutch translation is definitely not literary! A bad translation or one of moderate quality can turn a literary book into a mess. This also means that I don’t approve of your approach in this survey. I want to evaluate the literary quality of the author, not of the translator. And I know that many other people are also annoyed by badly translated books.

Respondents praised a translation only occasionally. More often, they tried to soften their bad opinion of a book by adding a phrase such as ‘could be the translation, of course’.

Te oppervlakkig. Gaat niet diep genoeg op de materie in. Kan ook aan de vertaling liggen.

Too shallow. Does not delve deep enough into the topic. Could also be the translation.

Verhaal een beetje over de top. Niet altijd mooi van taal. (Kan ook aan de vertaling liggen natuurlijk!)

Story a bit over the top. Language not always agreeable. (Could also be the translation, of course!)

3.2 Differences in Ratings

Do Dutch readers give translated novels different scores compared with novels written in Dutch? I will only go into the scores on the scale of literariness. For all 401 novels, the mean score was 4.67 on a scale of 1 (not literary at all) to 7 (highly literary). Splitting the corpus into two revealed an interesting result: the mean score on the scale of literariness for all 152 Dutch novels is 4.95, whereas for all 249 novels translated into Dutch it is 4.47.

We need to look at different genres, however. There are 147 literary novels in the corpus—novels published with the label ‘literary novel’ given to it by the publisher and by libraries. The most-represented genre fiction in our corpus is the suspense novel (186), followed by the romance novel (41). In a fourth category, ‘Other’ (27), we put genres with only a couple of representatives, such as fantasy and regional novels, and titles that were marketed as novels but may also be labelled as literary non-fiction. We also put several volumes of short stories in this category. These 27 titles are thus a mix of different kinds of texts and are excluded from the comparison because there are too few of them to yield trustworthy statistics. Table 1 shows the number of titles in each of the categories, and Table 2 presents the mean scores on the scale of literariness.

Table 1 Number of titles in the corpus per genre

Table 2 Mean score on the scale of literariness per genre

What we see here is that the mean score on the scale of literariness for Dutch titles in the category of the literary novel is higher than for translated titles, whereas the opposite is found for the suspense novel and the romance novel, where the mean scores for translated titles were higher than for Dutch titles.

3.3 Differences in Criteria

Do Dutch readers use different criteria when accounting for their appreciation of translated fiction compared with fiction written in Dutch? Here, the explanations given by the respondents functioned as a proxy, since they do not provide the explicit criteria that the respondents used. But their explanations can perhaps help us to come a bit closer to understanding their criteria, whether applied consciously or not. I made use of AntConc, the freely available concordancing software developed by Laurence Anthony.⁶

I created two files, one containing all explanations relating to the 152 Dutch titles and the other with all explanations related to the 249 translated books. I then compared the vocabularies of both files, first listing the words that in a relative sense occur far more often than expected in the explanations related to Dutch titles than in those related to translated books, then listing the words that are relatively far more frequent in the explanations related to translated titles compared with those related to the Dutch books. For this, I used the keyness function in AntConc, which is based on log likelihood. In describing the results, I focus on the appellative nouns with the highest keyness, calling these ‘keywords’. Table 3 shows that the list of top key nouns in explanations related to the scores on the scale of literariness of Dutch novels contain several words referring to writing style and language use, and words referring to close family members. The top key nouns in the explanations related to translated novels mainly refer to genres, plot and serialization, and also include the word ‘translation’.

Table 3 Keywords (nouns) in explanations related to Dutch titles compared with translated titles (3a) and keywords in explanations related to translated titles compared with Dutch titles (3b)

The Riddle corpus contains an almost equal number of literary novels written in Dutch (73) and literary novels translated into Dutch (74). This means that the number of translated genre fiction titles is much higher than the number of original Dutch genre fiction titles, and this will certainly have influenced the results in Table 3. I therefore selected only the explanations related to the category of literary fiction and compared the vocabularies in the explanations related to Dutch literary novels with those related to translated literary novels. This led to Table 4, which again shows the nouns with the highest keyness in each of these files.

Table 4 Keywords (nouns) in explanations related to Dutch literary titles compared with translated literary titles (4a) and keywords in explanations related to translated literary titles compared with Dutch literary titles (4b)

Compared to Table 3 and 4 shows a partly different picture, especially regarding the explanations related to the translated literary novels, where genre names do not reappear. However, the keywords related to the Dutch literary novels still show slightly more nouns associated with style and language use. The keywords in the explanations related to translated literary fiction may be characterized as focusing on narrative structure and themes in the story.

4 A New Experiment

The answers to the first three research questions suggest that when asked for their opinion about a novel’s literariness, Dutch readers see Dutch novels in the category ‘literary novel’ as slightly more literary than translated novels in the same category. Both Dutch suspense novels and romance novels, however, get a slightly lower mean score on the scale of literariness than translated novels in these categories. In accounting for their scores, Dutch readers more often referred to language use and style (and family relations) when dealing with Dutch titles than when they accounted for their opinions of translated books. These Dutch readers thus seem to be inclined to see their ‘own’ literature as somewhat more literary than translated literature. Do we see the same pattern in other countries?

The third chapter in Doing Double Dutch (which was not yet available when we developed the experiment I will report on below) deals with a small part of this topic. Nico Wilterdink gives a very short overview of the international reception of contemporary Dutch literature in the United States, the United Kingdom and France, and he pays special attention to the historically changing literary relation with Germany. Wilterdink views the literary world-system as consisting of nationally or linguistically bounded literary fields that are interconnected through asymmetrical relations of power and influence. Worldwide, more than 50 per cent of translations are from English, but only 4 per cent of all books published in English are translations. Wilterdink observes: “Apart from some academic circles, there is hardly any public concern about a lack of openness to foreign (i.e. originally non-English) literature, if this is observed at all.”⁷ For completeness’ sake: in Germany and France, the share of translations is 14–18 per cent.⁸

Wilterdink notices a growing interest in Dutch literature since the 1980s for the countries he deals with, an observation he based on a growing number of translations and reviews of these translations.⁹ He draws attention to the fact that some Dutch authors are more famous in Germany than in the Netherlands. And in France, he states, the Netherlands (from a literary perspective) is seen as a country that is far, far away in the north—part of the Scandinavian countries. In the United Kingdom, translated literature from the Netherlands is even less visible than in France. The British do not seem to like books by authors with funny names, and Wilterdink assumes widely divergent stylistic conventions in the UK and on the European mainland may also explain this invisibility.

If we summarize the observed differences between the four countries in a few catchwords, we may say that the American attitude to Dutch literature is predominantly one of indifference and ignorance outside a small circle of professional specialists; the English attitude a mixture of arrogance and self-criticism; the French attitude well-meaning but superficial; and the German orientation open, interested, often enthusiastic.¹⁰

Wilterdink tries to explain this situation by looking at the global literary field or world-system as part of the more encompassing cultural world-system:

In general terms, we can say that the geographical, social and cultural distance between two nations correlates inversely with the intensity of transnational cultural exchange: the larger the distance, the less cultural exchange. Among the four nations compared, the distance with the Netherlands is the largest for the United States, the smallest for Germany, with France and England in between; and the differences in attention to Dutch literature correspond with this variation.¹¹

For our research, we wanted to look beyond only translated Dutch literature. We especially wanted to find out whether readers in other countries or language areas also look differently at novels written in their own language as compared with novels translated from other languages. We decided to perform an experiment in which we focused on novels from the literary category, excluding suspense and romance novels to not overburden the survey with too many variables. Our research questions for this experiment were:

• a) Do respondents’ scores on the scale of literariness differ for novels written in their native language compared with novels translated into their native language?

• b) If there is a difference, is there a certain hierarchy in the countries or languages of origin of the translated book and, if so, is this ranking the same in different countries?

We hired an international panel of readers with backgrounds comparable to those of the participants in the National Reader Survey. The gross representative sample consisted of participants aged 18 to 65 years. The net number of participants per country was 150.

We aimed for an equal number of novels in each of the main languages of four to six European countries, with an equal number of novels written by female and male authors. Only novels that were available in all chosen languages were selected, which proved to be a very difficult criterion. We hoped we could use titles that were also in the corpus of The Riddle of Literary Quality and the National Reader Survey, but this was only partially possible. To my dismay, the most challenging part was finding enough titles written by women.

From Wilterdink’s perspective it is probably not due to chance that we ended up with the same selection of countries as he did, only excluding the United States. We gathered a list of 24 novels, six from each of the following countries/language areas: the Netherlands (Dutch), the United Kingdom (English—we deliberately limited our selection of novels to authors from the British Isles), France (French) and the German language area (German—with titles published in Germany, but some of the authors originating from an adjacent country). Table 5 presents the complete list.

Table 5 The corpus of 24 novels available in Dutch, English, French and German

We asked the respondents to indicate which of these books they had read, and also to score all of them on the scale of literariness (instead of asking this only for the books they had read). Compared with the results of the National Reader Survey, this implied that most scores from the new experiment would very probably be primarily based on the reputation of a text or author (or language or national literary field) and not on the reader’s actual experience of the text itself. We are, of course, aware that even when people read a book, reputation also plays a role. But at least in theory, when they did read a book, their score may also be informed by the actual reading experience and thus also by stylistic, narrative, etc. elements.

We used the same 7-point scale as in the National Reader Survey and presented the 24 novels in random order. In the Netherlands, we presented the Dutch titles of the translated novels, followed by the author’s name and the country in which the book was originally published. In the United Kingdom, the first column had all titles in English, in France all titles in French and in Germany all titles in German. I will not describe all questions and all results of this experiment, but focus on answering the two research questions presented above. Statistical details will also be published elsewhere.

4.1 Differences in Ratings

Do respondents’ scores on the scale of literariness differ for novels written in their native language compared with novels translated into their native language? Yes, there is a difference. The National Reader Survey results showed that Dutch readers on average gave a slightly higher score on the scale of literariness to novels written in Dutch than to novels translated into Dutch. In the international panel survey, we see the same results for the Netherlands, the United Kingdom and France. Germany is an exception, in two ways. One of these relates to the novel Feuchtgebiete by Charlotte Roche. In Germany, this book received a (statistically significant) lower score than in the three other countries where it received a (statistically significant) higher score. The German panel was also exceptional in another respect. Compared with the panels in the other three countries, the representative panels in France, the United Kingdom and the Netherlands gave significantly higher scores to only two of their ‘own’ novels, and to Feuchtgebiete. The German panel had only one title with a significantly higher score for books from each of the four countries. For the German language area this was Der Turm by Uwe Tellkamp, for English The Sense of an Ending by Julian Barnes, for French Réparer les Vivants by Maylis de Kerangal, and for Dutch De Verdronkene by Margriet de Moor. The German panel seems to be more fair and more open to translated novels than the other panels, which seems to accord with the observations by Wilterdink in his chapter in Doing Double Dutch. Of course this is a very preliminary conclusion and it needs to be followed up with more in-depth research. It could, for instance, also be the case that national and international literary prizes had a larger impact on the German panel members than on the members of the other panels.

4.2 A Literary Hierarchy of Countries

Apart from the literary appreciation scores, we also have the answers to a couple of other questions we asked the representative panels in the four countries. After they had scored the 24 novels, we asked them how they valued the literatures of the different languages in comparison with each other. They were not allowed to go back in the survey to see the scores they had given to the 24 novels. The German panel, for example, was asked (in German):

If you compare English literature to German literature, which of the following statements do you find most adequate?

• 1. There is no difference.

• 2. English literature usually has a higher literary quality.

• 3. German literature usually has a higher literary quality.

• 4. I do not know.

The same was question was then asked for French compared with German and Dutch compared with German. The panels in the other countries were asked to make the same comparisons of the other languages with their own language. The order of the three questions about the three languages was always random. Next, the panel members in all countries were asked to rank the four languages, putting the language with the highest literary quality first and that with the lowest literary quality last.

These questions also yielded very interesting results. For the first set of three, many respondents selected the first option: there is no difference. But there were also enough respondents opting for another choice. Taking all responses into account, we can summarize that the English panel had a higher regard for English literature than for Dutch and German literature, and sees it as comparable to French literature. The French nicely mirrored this: the French panel had a higher regard for French literature than for Dutch and German literature, and saw it as equal to English literature. The Dutch panel opined that Dutch literature did not reach the high level of literariness of English literature but was certainly comparable to French and German literature. And finally, the German panel indicated that German literature did not reach the level of English literature but was comparable to French literature, and was definitely more literary than Dutch works. The respondents’ ranking of all four national literary fields is presented in Table 6.

Table 6 Respondents’ ranking of the literariness of the literatures of all four countries

It will be clear that according to the members of the four national panels, there is a hierarchy. It is also clear that this hierarchy differs in each country. In Germany, German literature is valued higher than literature from the other three countries, although only just beating English. In France, French literature is valued most and in the United Kingdom English literature is valued most. Only the Dutch panel sees its ‘own’ literature as second to another one, namely English literature. But this is not true for the scores the Dutch panel gave—there, the panel members (unconsciously?) put Dutch literature first. And this was also the outcome of the National Reader Survey.

The German panel’s scores for the 24 novels also do not quite agree with the results of the explicit rankings. This definitely needs more research, but for now, I expect that the mean of the ratings of all the books may give a better idea of the situation than the explicit rankings. In comparing or ranking the four literatures, respondents may quite easily answer in a more politically correct way, whereas their individual scores may bring to light any unconscious, implicit biases they have.

If we combine the rankings of all panels, we see a result that seems to agree with current ideas about the relative place of the four languages in the hierarchy of the world republic of letters (Table 7).

Table 7 Combined ranking of the literariness of the literatures of all four countries

After the respondents had presented their ranking of the four literatures, we asked them if they could explain why they considered novels from their chosen first rank to be the most literary. We have 469 responses to this question, some of which only state that they cannot explain their choice, that it is just a feeling or just their own opinion. Others refer to the style, or that they know the language and the culture and thus cannot make a fair comparison. Deeper layers as well as plot lines are also mentioned.

It is rather difficult to do a vocabulary analysis of these answers, since they are in four different languages. A word frequency list also does not help that much. It does help, however, to provide a quick overview of all the words used, to quickly find for instance all personal names that are mentioned: William Shakespeare (9 times), Jane Austen, Charles Dickens (4 times), Harry Potter (3 times), J.K. Rowling, Brontë sisters (twice each), Arnon Grunberg, Honoré de Balzac, Julian Barnes, Brecht, Charlotte Brontë, Camus, Eliot, Jane Eyre, Goethe, Hesse, Mann, Terry Pratchett, George Orwell, Siegfried Sassoon, Thackeray and J.R.R. Tolkien (once each). Below I present some of the explanations given by respondents favouring English literature, to illustrate the kind of answers we received:

Aus dem gefühl heraus

Just a feeling

Based on my own nationality

Beaucoup d'auteurs littéraires

Many literary authors

Beter woordgebruik en andere thema's

Better language use and different themes

Haben meist tolle handlungen und schauplätze

Usually have great plots and settings

Harry Potter is the best thing ever

Hebben bij mijn weten de meeste grote auteurs voortgebracht

To my knowledge have the largest number of great authors

Het is gewoon een feit, net als britse films of series

It’s just a fact, same as British films or series

We are the home of it :)

We have Dickens and Eliot etc.

We have the best authors

We have the most famous writers in history and we have learnt from them.

And some examples from those favouring German literature:

Allein durch die Namensangabe

Just by looking at the names

Anything is better than English

Ben de taal magtig [sic: spelling error for ‘machtig’]

I master the language

Better arcing of plot lines.

Das land der dichter und denker

The country of poets and thinkers

Eigentlich kommt es nicht darauf an aus welchem Land ein Buch kommt. Es gibt in jedem Land bessere und schlechtere Autoren. So wie bei allen Sachen...

In fact it doesn’t matter which country a book comes from. Each country has good and bad authors, as with anything…

Finde ich am besten geschrieben

I think they are the best written

Plus detaillé

More detailed

I will not go deeper into this here. A more thorough analysis will be part of another publication.

5 Stylometry

As shown above, direct questions to readers may not provide us with the best answers to our research questions. But a broader perspective, combining scores and explanations, seems more useful because then we can actually see that respondents are not always aware of any biases they have. The scores of the novels in the new international experiment showed that representative panels of readers from three of the four countries tended to rank literature written in their own language as the most literary. This agrees with the results of the National Reader Survey.

The next question is then: why? Can it be that some or all of the actual translations are of a lower linguistic and literary quality than the originals, as some of the respondents in the National Reader Survey suggested in their explanations? Even if that were not the case, we may also consider a more sociological explanation. Perhaps readers are not inclined to consider a translation as a typical or prototypical literary novel. This may be because in the literature classes at the schools they attended as young adults they were trained to pay attention only to literature in the native language, which implicitly demoted the worth of translations. This could be researched by, for instance, looking at how the literary canon was and is taught in different countries. Could it be that in the Netherlands, the United Kingdom and France, schools promote/promoted a clear divide between national literature and such a thing as ‘world literature’ or the classics in the high school curriculum, and that this is/was different in Germany, to name but one possibility? I will not go into this here, but I did take a first step towards answering the question concerning translation quality by performing a small stylometric experiment.

What does a stylometric comparison of translated novels and novels written in Dutch reveal? I again used the keyness option in AntConc to address this question. This time, I compared the vocabularies of a subset of the literary novels written in Dutch with novels translated into Dutch to see whether this would give us any new points of connection for further research. In selecting the subset, the gender of the author also played a role. As PhD student Corina Koolen shows in her thesis about The Riddle of Literary Quality,¹² in the National Reader Survey novels written by male authors tended to be seen as more literary than those by female authors. For my small stylometric experiment, I therefore used only an all-male or an all-female authored subcorpus. My aim was to compare a roughly equal number of Dutch original literary novels with a set of novels translated from one language, all in the same category or genre. Table 8 shows that this almost automatically leads to the choice of literary novels written by female authors in Dutch and translations into Dutch from English novels written by women. The Riddle corpus has 23 original Dutch literary novels written by women and 25 translations from English novels authored by women.

Table 8 Literary novels

Table 8 shows that of the 401 Dutch best-sold and most-borrowed titles, 50 novels written by male Dutch authors made it to the list, and only 23 by women did. Slightly more translated titles were written by male authors (40) than by female authors (34). The number of translations from English was higher for the translated titles of female authors (25) than male ones (19).

When we compare the vocabularies of all 23 novels written in Dutch by women with those of the 25 novels by women translated from English, the first thing that shows up is that most of the top 100 keywords are personal names. This is not unexpected. Proper names usually end up high on keyword lists, because they are mostly closely linked to the specific story that is being narrated and do not occur in many others. An analysis of the kinds of names in both sets is certainly interesting, but will have to wait for another occasion.

What is interesting in both top 100 lists is the occurrence of personal pronouns. In the keywords characteristic of original Dutch novels we find the following pronouns: “ik”, “me”, “mij”, “mijn” (‘I’, ‘me’, ‘my’, ‘mine’). In the keywords for the translated novels, however, we find “ze”, “haar”, “hen” (‘she/they’, ‘her’, ‘them’). This seems to indicate a possible difference in perspectives: are the most popular Dutch novels by women perhaps more often first-person narrations, while the translated ones prefer a third-person approach? Or are the Dutch novels more ‘egocentric’ in some other way, for instance through a more prominent role of dialogue?

Another observation concerns nouns referring to close family members. The top 100 Dutch novels’ keywords include “mama” (‘mummy’), “oma” (‘granny’) and “vader” (‘father’), whereas the translated novels has “mam” (‘mum’) in the keywords. What especially interests me in this list is the contrast between “mama” and “mam”: Dutch authors give preference to “mama” and translations into Dutch more often use “mam”. This could be an indication that the word form “mam” is influenced by English “mum”, and thus that the relatively higher use in translations from English somehow shows the English-language origin and might say something about the nature (and perhaps even the fluency?) of the translation. One should remark, however, that some Dutch women prefer to be called “mam” and not “mamma”—which ofcourse in its turn may be interpreted as an indication of a higher status of words with an English connotation.

From a translation perspective, the next two observations are also very interesting. Among the top 100 keywords of the translated novels, we find a number of verbs that probably mostly function to indicate turns in dialogue: “vroeg” (‘asked’), “wist” (‘knew’), “zei” (‘said’). This could mean that turns in dialogue are usually more explicitly described in the English originals than is usual in Dutch novels, or it could mean that the translations did not quite represent it correctly in the Dutch target language—and thus could be a sign of a fluency issue in the translations. The same can be said for the following words with high keyness in the translated novels: “en” (‘and’), “toen” (‘then’), “dat” (‘that’), “voordat” (‘before’), “terwijl” (‘while’), “hoewel” (‘although’). These conjunctions occur many more times in the translations than in the originals, and this again could indicate where the translations did not quite match the style as we find it in the selected novels written in Dutch.

Without further analysis we cannot exclude the possibility that the above observations are indications of translations not quite being up to the standards that, in this experimental setup, the originals can be seen as providing. This definitely could be part of the answer to the question why Dutch readers value the literariness of Dutch literary novels slightly higher than translated ones. However, more in-depth research from a translation perspective is necessary, using the tools and methods of literary stylometry and corpus linguistics.¹³ We could, for instance, trace and analyse the usage of the words with high keyness in the individual novels in the two subcorpora we compared, and see whether the individual translations differ in their use of these words and if any differences correlate with the mean ratings for the individual titles. Are some translations considered to be more literary than others? And are these the ones with a less deviating frequency of some of the words with keyness in the translation corpus? We should also bring in the English originals in digital form, for instance using an alignment tool, to make it easier to check the occurrences of the keywords in the translations against the original English phrases they were meant to represent. This may lead to plenty of interesting follow-up research, criss-crossing even more disciplinary boundaries than has already been done in this research project.

6 Conclusion

In seeking to uncover the current Dutch conventions of literariness, the Riddle of Literary Quality project has crossed several disciplinary borders. We have embraced many of the opportunities created by technological advancements. First, the Netherlands has one of the highest internet penetration rates in the world—in both 2011 and 2016, 94 per cent of all households had internet access, and in the latter year 92 per cent of all households actively made use of it.¹⁴ This made it possible to take the National Reader Survey online. Setting up and conducting the survey online is financially much more feasible than a manual approach. Using email and various kinds of blogs enabled us to find plenty of respondents in a time-efficient manner—we even had enough budget left to hire a public relations specialist to help us. It was also quite easy to gather the results and analyse them. We made use of a professional market researcher, Erika Nagelhout, and her national and international network, and this led to a very interesting and inspiring collaboration, resulting in several small follow-up experiments, including the international experiment described above. We think that this approach of finding respondents online and hiring representative panels makes our research and its results much more representative than many experiments done by psychologists whose test groups consist of first-year psychology students taking a survey for credits or money.

Second, the advent of the e-book came just in time. When I drafted the first version of the research plan in 2010, publishers in the Netherlands where still very hesitant to produce e-books. But by the time the proposal was reviewed and granted and we got ready to start in 2012, we were pleased to find that most of the 401 novels on our list were available in digital form. We only had to scan and OCR around 30 titles, so our digitization work took much less time and energy than we had anticipated.

Third, our driving force was always the research questions we have, and we did not intend to limit ourselves to the methods that had been used in previous literary studies. Our curiosity thus led us to cross boundaries. Interdisciplinary and transdisciplinary collaboration was therefore key in our project. First, we made sure that our team consisted of people with a mix of disciplinary backgrounds and expertise. The core team had a background in literary studies, computational linguistics, natural language processing, stylometry, lexicology and computer science. Second, we initiated collaborations with people from other disciplines, such as psychology and statistics, and from libraries and publishers, and with translators; sometimes, people approached us to initiate a joint project. And each of these collaborations led to new ideas. But collaboration only works well when the ultimate question is interesting for all those involved, so that each collaborator identifies with the shared research and research questions and has a stake in the outcome or some of the outcomes. From the first preparations for the project onwards, we found that we were not the only ones interested in the question of literary quality. This made collaboration a lot easier. For many of our collaborators there were also interesting results to expect, and we may need many more years to follow up on all the ideas and plans that were shaped.

Finally, this kind of collaboration needs to be on an equal footing, which implies that we need to share the results and ideas with each other. Perhaps this is the most difficult issue for many of us, as we come from a branch of humanities where the lonesome individual genius is still highly respected.

To sum up: I am convinced that the rise of information and communication technology has greatly enhanced the possibilities for humanities research—as long as researchers are willing to cross disciplinary boundaries themselves or are willing to collaborate on an equal footing (which also implies crossing at least some boundaries) with scholars from different disciplines. In the near future, these kinds of developments may greatly change the location of these disciplinary boundaries and even change the way that humanities research fields are organized.