Canonical Values vs. the Law of Large Numbers: The Canadian Literary Canon in the Age of Big Data

 Abstract

In this article, I propose an alternative technique to the traditional method of constitution of the literary canon. Instead of basing the determination of the canon on different values, I scrutinize the Modern Language Association International Bibliography database in order to determine the most cited authors and literary works. Specifically, I study Canadian literature. Thus, through the process of data mining, I obtain a sample of over 25,000 references that allows us to observe the chronological evolution and the linguistic distribution of the critical bibliography about Canadian literature. This quantitative technique yields a corpus of 151 titles and 295 writers that are cited more than 10 times in the database. Consequently, this bibliography is not the result of subjective selection criteria, but is based on the law of large numbers. Furthermore, this study shows that the quantitative analysis of bibliographic databases is an effective way to bring new light to the field of literary studies.

 [Keywords: literary canon, Canadian literature, bibliographic databases, bibliometrics, data mining, Big Data]

 Canon formation

In spite of its apparent simplicity, the concept of canon is a difficult one to define. According to certain texts (Guillory 1995; Lentricchia and McLaughlin 1995), etymologically, the word finds its source in the Greek word kanon that signifies ruler or measuring stick. Initially, the term was used to identify those texts from the Old and New Testaments that are approved by the ecclesiastic authorities as the Word of God, thus, the texts that constitute the Sacred Scriptures. During the last decades, it has been considered that the canonization of literary texts operates in a way similar to the process of the constitution of the biblical canon.

In 1995, at the beginning of his book The Western Canon. The Books and School of the Ages, the American critique Harold Bloom determines that: “Originally the Canon meant the choice of books in our teaching institutions, and despite the recent politics of multiculturalism, the Canon’s true question remains: What shall the individual who still desires to read attempt to read, this late in history?” (Bloom, 15).

As such, the fundamental question about the canon seems easy to deal with. However, it is a very complex concept that has caused considerable discussions. For instance, in 1983, Critical Inquiry published a volume completely dedicated to the concept of the canon. In the introduction, Robert von Halberg establishes that “Interest in canons is surely part of a larger inquiry into the institutions of literary studies and artistic production. ‘Politics,’ ‘economics,’ ‘social,’ ‘authority,’ ‘power’ –these are some of the terms that recur throughout these essays; we are most curious now about those points where art seems less private than social” (von Halberg, iii).

In 1999, Nel van Djik analyzes the canon formation from different viewpoints: nationalism, literature, and institutions. According to van Djik:

 The list of works that count as our western society’s literary inheritance is no longer prescribed by the church and the state, but by authorized institutions such as literary criticism and literary education. In addition, scientific developments in the past decades have resulted in the widespread conviction that literary value is not an intrinsic but an attributed quality. This quality results from the consensus that exists between the members of a literary institution at a certain moment (121).

 For him, the researches working on this subject are divided into two groups: those that consider it from an ideological-hermeneutic approach and those who prefer a sociological perspective.

A year later, Anderson and Zanetti (2000) determine that the discussion about the canon has been conducted according to two opposite poles. On the one hand, those that belong to the right wing of the political spectrum defend a traditional canon. On the other hand, those that have leftist ideas declare that the canon is an obsolete artefact. Between these two poles, we find a series of perspectives that open up the notion of the canon in order to include minorities or that consider the existence of multiple canons.

In 2003, Jeffrey Insko picks up the debate about this concept, bringing back the tension between the imaginary canon (Guillory) and the pedagogical canon (Gallagher). The following year, Frank Kermode, in his book Pleasure and Change. The Aesthetics of Canon, pushes aside the ideological aspects of the canon constitution in order to introduce three new characteristics into the process: pleasure, change, and chance.

My purpose in mentioning these studies about the concept of the canon is not to carry on this debate, but to show that the constitution of a canonical corpus is a highly complex process. My aim is to introduce an alternative method that makes it possible to identify those literary works most frequently analyzed within a national literature[i]. In other words, instead of presenting a literary canon, I will trace a literary cartography. Specifically, I will focus on Canadian literature.

 Theoretic and methodological approaches

With the purpose of developing this experimental method, I base my research in the articulation of several essential notions. From the theoretical viewpoint, I build my work, on one hand, upon the concept of literary field established by Pierre Bourdieu (1992), and, on the other hand, on scientometrics (De Solla Price, 1963; Garfield, 1980; Leydesdorff, 1998).

Methodologically, there are two essentials aspects: firstly, the analysis of field of knowledge (Albrechtsen, 1997; Hjorland, 2001; Hjorland and Albrechtsen, 1995), and, secondly, theories of citations (Kaplan, 1965; Moravcsik and Murugesan, 1975; Gilbert, 1977; Small, 1978, 1998; Leydesdorff, 1998; Enger, 2009).

Although several scientometrists consider that quantitative methods cannot be used in the humanities due to the differences in terms of citation across the disciplines (Archambault et al, 2006, Cole, 1983, Cozzens, 1985, Larivière et al, 2006, Nederhof et al, 1989), recently, we have witnessed several bibliometric analysis in the humanities in general (Linmans, 2010; Moed et al, 2002; Osca-Lluch and Haba, 2005), and in literary studies in particular (Ardanuy et al, 2009; Hammarfelt, 2011; Herubel and Goedeken, 2000). The latter indicate that, despite the abovementioned citing differences, scientometrics is a relevant approach to increase our understanding of the behaviour of the literary field.

In this research, I use the Modern Language International Bibliography database; from now on I will refer to it as MLAIB. This electronic bibliography, renown as the most important one in literary studies, contains over 2,107,000 references and includes approximately 4,400 journals. Besides the articles, the MLAIB database includes references to books, book chapters and thesis[ii]. In terms of chronology, it covers the literary critique from 1886 to the present.

Through the techniques of data mining (Han et al, 2012; Witten et al, 2011), and keywords (Callon et al, 1993), I initially obtain a sample of the critical bibliography about Canadian literature. Then, I extract the Canadian literary corpus as well as a list of the main Canadian writers. The data is directly transferred in parcels of 500 references each time to a RefWorks account. The latter data management tool allows me to automatically convert the series into a spreadsheet format. Thus, from that point on, I can efficiently work with the series in order to obtain indicators, graphics, and tables and to qualitatively query their contents. The cutting date for the analysis is 2010.

 Literary cartography of Canadian literature

Through the application of the data mining technique, I obtained a sample of 25,102 references covering a period of 124 years, from 1886 until 2010. Figure 1 represents the results of this data mining process. The series begin to grow by the end of the 1950s. Since 2001, the average number of references has reached a volume of 792 publications per year. Journal articles are the most important type of documents, 67% of the sample, followed by book chapters, 26%. In terms of the language of publication, Figure 2, the sample is divided into 70% in English and 27% in French. Individually, other languages correspond to less than 1%, and altogether they only reach 3% of the references.

  Fig. 1. Chronological evolution of publications about Canadian literature, MLAIB 1886-2010.

  Fig. 2. Documents about Canadian literature by language of publication, MLAIB 1886-2010.

 Writers and Texts

In order to identify the most studied Canadian literary works and writers, I interrogated again the MLAIB database. I selected those writers and works that have been the object of at least 10 publications. The results of these interrogations were a corpus of 151 titles and a list of 295 authors. Table 1 corresponds to the top 20 Canadian titles according to MLAIB.

Table 1. Corpus of the top 20 Canadian literary works, MLAIB 1886-2010.

As we can observe, Margaret Atwood is the most outstanding author with 7 titles on the corpus. Next on the list are Michael Ondaatje, with 4 titles, and Anne Hébert, with 2 titles. Among this list, there are only 4 titles in French: Les fous de Bassan and Kamouraska by Anne Hébert, Bonheur d’occasion by Gabrielle Roy, and Maria Chapdelaine by Louis Hémon. This short list includes works from 1916 until 2000. However, the complete corpus spans from 1832 to 2008.

In Table 2, I present the top 20 Canadian authors. Again, Margaret Atwood is by far the writer with the highest number of references. In this case, we observe a perfect equilibrium between English and French Canadian literature, since there are 10 writers that represent each language.

Table 2. List of the top 20 Canadian writers, MLAIB 1886-2010.

Concluding remarks

At the beginning of this research, I presented several aspects that, according to various scholars, make the process of identifying canonical literary works very difficult. Then, I introduced an alternative method based on data mining. Thus, I was able to obtain a corpus of literary works that reflect the interest that the academic critique has expressed for Canadian literature through their publications for over 120 years. It seems to me that this method overcomes the different tensions involved in the process of canon constitution and, at the same time, it responds to the issues signified by Nel van Djik as well as to Harold Bloom’s essential question about the canon.

According to van Djik, the canon is prescribed “by authorized institutions such as literary criticism and literary education” (van Djik, 121). In this sense, the method I propose here is based on a very significant part of the critical activity of scholars. When using and exploiting the MLAIB database with scientometric methods, the bibliography obtained is not the result of subjective selection criteria, but is based on the law of large numbers. Actually, in the case of Canada, the results compiled in this research are the reflection of more than 25,000 references obtained with quantitative techniques that can be reproduced and verified.

At the same time, the results presented here respond to Harold Bloom’s question. Adapted to the literature analyzed in this case: “What shall the individual who still desires to read [Canadian literature] attempt to read, this late in history?” (Bloom,15). It seems to me that the corpus obtained provides an appropriate answer to what Bloom considers to be the interrogation that lays at the center of canon constitution.

Moreover, I consider that the corpus here obtained is clearly superior to any canonical list. Firstly, these results may be classified according to different parameters: geographic, chronologic, and linguistic. Secondly, these lists are dynamic as they can be actualized by periodically interrogating the MLAIB database. Finally, the corpus presented here, that includes 151 titles, is much larger than Bloom’s list for Canada, which contains only 9 titles (Bloom, 530).

It seems to me that this analysis is a clear demonstration of the relevance of using scientometric methods in the study of literature as they allow us to increase and deepen our knowledge about the literary field.

Notes