Code-switching and multilingualism in literature

1 From oral to written code-switching

The past 50 years have seen an exponential growth both in the study of bilingualism, and in one of its most salient conversational manifestations: code-switching (CS). Linguists have long since debunked Weinreich’s conception of the ‘ideal bilingual’, and the claim that s/he ‘switches from one language to the other according to the appropriate changes in the speech situation … but not in unchanged speech situations and certainly not within a single sentence’ (Weinreich, 1953: 73). On the contrary, we now have a number of insightful approaches and models that help us to understand when and why bilinguals switch languages both within and between sentences, even in ‘unchanged’ speech situations.

In contrast to the long-standing attention paid to spoken CS, the study of written, and specifically literary, code-switching, has only recently enjoyed a surge of interest. The conference which gave rise to most of the articles in this issue, Code-switching in Literature, was organized at Birkbeck, University of London on 5 July 2013. Other international conferences on similar themes include The Poetics of Multilingualism, Eötvös Loránd University, Budapest, 4–6 April 2013; Ecrire entre les langues/écrire en langues at INALCO, Paris, on 8 November 2013; and Inverted Runes: New Perspectives on Literary Translingualism in Uppsala University, Sweden, 4–5 September 2015. Panels on literature and bilingualism were held at the International Symposium on Bilingualism in Oslo in 2011 and at the World Congress of Comparative Literature in Paris in 2013. The conference on Historical Sociolinguistics entitled Code-switching: the Next Step, held in Tampere, Finland, on 11–13 June 2014, showed that historical sociolinguists are now willing to consider both literary and non-literary CS within a common framework.

As noted by McClure (1998) Jonsson (2010) and Sebba (2012), linguists have paid comparatively far less attention to written CS than to conversational CS. One explanation for this lack of research is undoubtedly the sociolinguistic tradition which emphasizes the primacy of spontaneous spoken language, concentrating largely on phonological variation. This goes back to Labov’s principles, enunciated early on and widely followed within the discipline: ‘The vernacular, in which the minimum attention is paid to speech, provides the most systematic data for linguistic analysis.’ (1972: 208). Writing some 10 years later, Lipski (1982) commented that ‘because of the very fact that written documents, particularly those classed as literary, involve not only conscious reflection but also the inherent correction, editing, and rewriting process that accompanies any act of writing, it cannot be claimed that such texts may be used as specimens of naïve, spontaneous linguistic production’ (Lipski, 1982: 192; see also Coupland and Jaworski 2009: 5; Lillis, 2013). But although spontaneous linguistic production remains the bread and butter of sociolinguistic theory and analysis, the relationship between such production and other forms of language is both extremely important and relatively unexplored. It is precisely the gap between spontaneous oral production and written forms of the language that can tell us how each culture or community envisions what a language is. In the case of historical evidence, which is confined to written genres, deductions can, similarly, be made from the differences between different types of document (see Schendl’s comments in this issue on CS in business texts as described by Wright 2011), even if some questions will not be answerable on the basis of written evidence alone.

There is no doubt, however, that a methodological quandary exists: as many of the models we have for CS are based on oral data, it is unclear to what extent they can be applied to the written word. Sebba (2012) notes that recent studies of CS in writing have focused on computer-mediated communication (Androutsopoulos, 2007; Montes-Alcalá, 2007; Siebenhaar, 2006), which is no coincidence given that this dialogic form of discourse often resembles a hybrid of speech and writing. He has misgivings about the applicability of oral models of CS to monologic written language, especially when it comprises multimodal features. He even questions whether the term “code-switching” is applicable to this medium at all, or if it describes ‘substantially different phenomena’ (Sebba, 2012: 98).

If sociolinguists have tended to neglect literary evidence, literary scholars have been equally remiss in exploring multilingualism, despite the fact that this has been a resurgent feature of the language of literature in both the 20th and 21st centuries. Traugott and Pratt (1980: 392), writing over 30 years ago, note that ‘[t]here is very little that has been written as yet on experiments with multilingual literature’. In her recent study Multilingualism in Modernist Fiction, Taylor-Batty (2013: 7) reaffirms this is still the case today. She attributes this lack of attention to ‘the continuing cultural and political conception, particularly amongst Western native speakers of world languages such as English, of monolingualism as the “norm”, and the resultant perception of cases of literary multilingualism as exceptional, unusual, extraordinary’. Sociolinguists have long been aware that, in terms of speakers if not writers, it is in fact monolingualism which is the global exception rather than the rule. The study of literary CS has thus fallen between two stools, proving to be equally unattractive to both literary and sociolinguistic scholars for distinct reasons. This is unfortunate given that, as Lipski (1982: 192) rightly noted, ‘there is … a vast amount of linguistic, psychological, and aesthetic information to be obtained from a careful consideration of code-switching in its written form’.

2 Multilingualism in literature

Despite this background, significant studies of multilingualism in earlier periods, of necessity based on written evidence, have begun to emerge in the last 10–15 years (for the Ancient World, see Adams et al., 2002; Mullen and James, 2012; for medieval Britain see Trotter, 2000; for the medieval francophone world see Kleinhentz and Busby, 2010). Earlier research on written multilingualism often concentrated on representations of speech within specific written texts. Examples include Stolt’s (1964) study of Luther’s code-switching based on the notes taken by his followers; Timm’s (1978) study of Russian–French switching by the Russian aristocracy in War and Peace; or Iglesias-Rabade’s (1996) study of multilingual preaching. Much of the more recent work focuses on multilingualism in the written language more broadly. Schendl’s work covers multilingualism in a range of text types in the medieval period (Schendl, 1997, 2000); Pahta (2004) studied multilingualism in medical writing; and Wright (2002) has studied the languages of business texts.

Studies dealing specifically with literary contexts are less common, but this situation is now changing (Callahan, 2004; Davidson, 2010). Kleinhentz and Busby’s (2010) volume on the medieval francophone world contains articles on literary as well as non-literary themes, as does Sebba, Mahootian and Jonsson (2012). For the Ancient world, Adams, Swain and Janse (2002) cover multilingualism in a range of ancient languages and text types, literature being represented by Swain’s study of bilingualism in Cicero. Mullen (2011) and Mullen and James (2012) provide an overview of material in which Latin was in contact with other languages. They discuss possible typologies of types of multilingualism represented in these works, and mention various sources for the study of literary multilingualism, for example in the comedies of Terence and Plautus (Jocelyn, 1999; Shipp, 1953).

Multilingualism in literature takes many forms, and, as Lipski pointed out, a vast amount of linguistic, historical, psychological and aesthetic information can be derived from studying it. Translingualism, defined as writing in more than one language or a language other than one’s primary one, is discussed in two books edited by Kellman (2000, 2003). ¹ Writing in another language gives writers a means of escaping the constraints of their mother tongue, as when Rilke wrote poetry in French, or Milton wrote to his Italian lover in Italian. By the 19th century, however, due to the development of nationalism, this had become more unusual. In the 20th century, ‘concrete poetry’ such as that of the Austrian Ernst Jandl provided a more extreme example of how to use language as a pure material, devoid of sense and associations. Samuel Beckett is often given as the primary example of a translingual writer, with his preference for writing plays in French rather than in his native English. Famous for writing polished prose in his third language, English, Joseph Conrad’s style also shows influences from his first two languages, Polish and French, ² which give his English its particular character. The instances in which authors demonstrate a stylistic influence from their L1 are, however, distinct from passages where two languages are juxtaposed for a specific literary function (see Hampson, 1990, for a related discussion of Conrad).

Translingualism is connected to, though distinct from, the process of (self-) translation, another important field within multilingual studies. Nabokov, the prime example of a perfectly balanced bilingual, or ‘ambilingual’ writer, in the terminology of Hamers and Blanc (2000), translated many of his own works (as did Beckett). His multilinguality is a crucial aspect of his work, which in some cases incorporates actual code-switching and in other cases multilingual word-play (Appel, 2000), the classic example of the latter being that of Joyce in Finnegan’s Wake. As Gumperz (1982) notes, CS often involves a form of self-translation (which he calls ‘reiteration’), whereby bilingual speakers immediately repeat what they have just said in one language, in another language. Auer draws out this comparison more explicitly by calling reiterations ‘quasi-translations into the other language’ (Auer, 1995: 120). As we see in this special issue, this is an area where the differences between CS and translation may become blurred in written texts. However, to the extent that translation involves recasting a text in toto from one language into another, primarily to facilitate comprehension, this differs from the motivations underlying spontaneous CS. In fact, translingual writers often consciously avoid mixing their languages as doing so might detract from their achievement. In her memoir Lost in Translation, Eva Hoffman uses only the occasional Polish word (e.g. tęsknota, ‘yearning’), the central theme of her book being the incommensurability of languages (Hoffman, 1998). Others, such as Beckett or Alexakis, have, by contrast, expressed their multilingualism by systematically translating their own work. In certain periods, whole genres of poetry were introduced into other languages thanks to poets translating their works into the vernacular, e.g. Joachim du Bellay from Latin into French and Jan van der Not from French into Dutch (Forster, (2009 [1970])). It was also common to write certain works in one language and others in another, often according to genre, e.g. poetry v. prose (Langslow, 2002). In some periods reformulating the works of others was a common form of literary production and the ability to do this well was valued more than originality.

Multilingual writers bring about linguistic and cultural change by introducing terms and concepts from one language into texts in another, a process akin to linguistic borrowing (Mackey, 1993: 57). Translations may also be manipulated so as to evoke the original text behind the translation (Taylor, 2005). In the Early Modern period, producing multilingual poetry was designed to reach a wider audience familiar with Latin as well as various vernaculars. A poet was expected to be able to compose in several languages just as an accomplished artist could create in several media, each language offering different possibilities (Forster, (2009 [1970])). Such literary exploits relied on there being di- or triglossic multilingualism among the audience, an educated elite. In this they differ from cases, such as that of contemporary Alsace, where a vernacular literature co-exists – or is combined with – one in the official language(s) of the state, to which most ordinary educated readers have access.

3 Multilingualism within the text: Code-switching in literature

The use of several languages or varieties within the same text, or code-switching, has a multitude of possible functions within multilingual literature. It includes different languages being used for different characters or voices; to mark out different parts of the text; to represent a mixed speech mode which characterizes the community; or to bring in different registers or sets of allusions. In the case of intense switching within the grammatical unit (sentence or word), it may be used for deliberate comic effect. Troubadour poetry such as that of Raimbautz de Vaqueiras was written in a mixture of Romance languages (Dionisotti and Grayson, 1949). So-called ‘macaronic’ writing, a parodic form with dense code-switching – often involving Latin combined with various vernaculars – is thought to originate in a late 15th-century comical poem by Tifi Odasi (written in mixed Latin and Italian). Other examples of macaronic writing can be found in Middle English (Schendl, 2000) and Renaissance Crete (Vincent, 1991). Some of these uses represent what happens in spontaneous speech while others do not, a point to which we return in our own article in this volume. In fiction, code-switching most often occurs in dialogue and in stream of consciousness writing (Callahan, 2004). It gives the overall impression of an informal register, of a rejection of literary standards, and thereby of realism or piquancy; in short, it heightens the orality of a text.

Other examples of CS as used in literature, often in a humorous or satirical way, include the representation of the speech of speakers of other languages, as in Shakespeare, Henry V; Charlotte Brontë, Jane Eyre; Tolstoy, War & Peace; or Thomas Mann, The Magic Mountain. Such instances lend more authenticity than the representation of ‘foreign’ speech through monolingual means, as attempted by Hemingway in For Whom the Bell Tolls or Roth in Call it Sleep (Traugott and Pratt, 1980). Code-switching itself may also represent hesitant speech in characters who are not using their native tongue (as in Twain’s The Innocents Abroad). Idiosyncratic multilingual speech can be used to draw attention to particularities of certain characters. In Pantagruel VI, Rabelais mocks ‘l’escholier limousin’ who tries to show his learning by speaking Franco-Latin; the troubling Salvatore in Eco’s The Name of the Rose speaks a multilingual jargon, and in Tony Kushner’s play Homebody/Kabul, the former librarian is considered mad because of her multilingual babble (Molloy, 2002).

Part of the aim of this volume is to consider to what extent theoretical frameworks developed for the study of multilingual speech can be applied to the analysis of literary texts (Sebba, 2012). Lipski (1982) rightly points out that one cannot assume that there are two quite separate grammars, one for literature and one for normal speech contexts, with readers somehow bridging the gap. Indeed some literary scholars have directly appropriated models (e.g. grammatical ones) put forward by linguists and applied them to literature. For example, Diller (1997–1998) applies Myers-Scotton’s ‘Matrix Language Frame’ ³ model (Myers-Scotton, 1993) to a body of Medieval English Drama; but as Schendl (this issue) points out, this gives rise to some problems. This is partly because neither this nor any other grammatical model for spoken code-switching is generally accepted; ⁴ and partly because literature takes in different genres and text types which follow different conventions. Callahan (2004), for example, found the Matrix Language Frame worked well for analysing narrative sections of Hispanic American literature, but broke down when applied to the dialogue. Schendl (this issue) also describes a literary debate about the discourse functions of CS, in relation to Piers Plowman.

The findings are suggestive of at least a partial overlap between the functions of CS in speech and in literature. Of course, the written language has to be more explicit, and code-switching in this modality is therefore more ‘conscious’. It is this fact which allowed individual authors such as George, Rilke, or Eliot to use several languages within their texts for specific effects (Forster, (2009 [1970])) in a way which clearly was not designed to mimic spoken CS. Surrealist and Dada writers used multilingual verbal collage both as an overt literary and artistic technique (Butler, 1994). Oral code-switching is very often a sub-salient phenomenon arising from speech patterns in the community (Bullock and Toribio, 2009; Gardner-Chloros, 2009; Muysken, 2000). By contrast, the use of CS for more specifically literary purposes is exemplified by authors such as James Joyce and Ezra Pound, whose writings do not stem from a multilingual upbringing, and who both learned other languages later in life. Contemporary writers who code-switch in their daily lives sometimes avoid CS in their written work, either for normative reasons or so as not to limit their readership to those who are equally multilingual.

In writing, code-switching can serve a number of functions: authors in countries emerging from colonialism may use it to get round the dilemma of either exclusively using the language of the former oppressor or using local languages, where the latter would risk limiting their readership and not being commercially viable (e.g. Nigerian novelist Chinua Achebe, discussed in Traugott and Pratt, 1980: 386–389). In communities of immigrant origin, bilingual speech modes are often cultivated and celebrated, and their representation in literature may serve purposes that go beyond simply representing the speech-modes of their members. Examples of this include Creole literatures in the UK (Sebba, 2000; Smith, 2000) and Hispanic literatures in the USA (Rudin, 1996). Callahan (2004) considers that code-switching in such literatures – in contrast to the elite multilingual productions of Eliot or Pound – is a symptom of macro-level change in the community which can help to redefine the social reality; others, such as Ortiz (1986; quoted in Callahan, 2004), distance themselves from linguistic treatments of the subject and view code-switching as a symptom of conflict with the dominant group.

In more modern literature, and especially in the contemporary period, code-switching is found in a variety of genres. It is probably closest to oral code-switching in the letters of multilingual correspondents (Nurmi and Pahta, 2012; Pahta et al., 2010; Swain, 2002); in private diaries, such as those of Cardinal Mazarin (Madelin, 1931); and in comic or journalistic prose where it parodies actual modes of speech (Wrenn, 1993). Its continuing prevalence in poetry (Lipski, 1982) is related both to the short unit of composition and its usefulness for stylized or ludic uses of language. For example, special effects are sometimes achieved in poetry by rhyming lines written in different languages (Diller, 1997–1998: 525; Forster, (2009 [1970]). In recent times, poetry’s lack of commercial appeal has freed it from the constraints suffered by novelists, who need to consider the marketing of their work, though some multilingual poets such as Benjamin Zephaniah in Britain or Gérard Leblanc in Canada have achieved considerable public success.

In terms of scholarly attention, literary code-switching has so far been most thoroughly studied in two widely different sociocultural and historic contexts: on the one hand in medieval literature (Diller, 1997–1998; Schendl, 2000, 2002; also articles in Kleinhentz and Busby, 2010) and on the other hand, in contemporary Hispanic literature in the USA, where the genre is flourishing (Callahan, 2004; Jonsson, 2005; Kanellos, 1995; Lipski, 1982; Montes-Alcalá, 2012, and this issue) and where a deliberate mixed genre is developing among some writers (Manguel, 2004). CS has been studied more sporadically in relation to other bodies of literature (James, 2010; Lalla, 1998; Mackey, 1993; Nassar, 1993; Omole, 1987; Wrenn, 1993). Comparative studies of the uses of multilingualism in a range of literatures, relating the types of multilingualism found in them to the sociolinguistic backgrounds which give rise to them, are still lacking, and this is a stimulus for this project. The three-way relationship between spoken, written and literary multilingualism also remains to be clarified.

4 Contributions to this special issue

The authors contributing to this issue include linguists, literary, historical, cultural scholars – and one classicist; several of them straddle two or more of these categories. They have taken up the challenge of exploring literary CS from very different starting points.

Starting with the chronologically earliest example, Alex Mullen’s article is an analysis of an aspect of elite Roman Latin–Greek bilingualism. She describes the CS in Fronto’s correspondence using data from a new database of Greek CS in Roman epistolography. She finds Fronto’s use of Greek to be more restricted and less multi-functional than in Cicero’s better-studied correspondence, which is considered more ‘conversational’ despite the careful way in which it was constructed. This echoes contemporary studies which show marked variations in CS according to individuals or sub-groups within the same language pairs and the same communities.

Herbert Schendl’s analysis of CS in early English literature shows how the prevailing diglossic relationship between Latin and English, with French as a further language of literacy, was reflected in CS which carried specific functions rather than in the development of a mixed code as such. The variety of genres in which it appears allows us to observe a wide range of relevant functions, many of which are familiar from modern sociolinguistic studies of CS. At a grammatical level, Schendl shows that we are dealing with quantitative tendencies towards certain patterns being more common in spoken and in written corpora, rather than with an absolute divide between the two.

Cecilia Montes-Alcalá examines a corpus of bilingual poetry, drama and fiction from authors belonging to three US Latino groups: Nuyorican (mainland Puerto Rican), Chicano and Cuban American. She illustrates the specific socio-pragmatic functions code-switching serves in her corpus, and uses a quantitative methodology to compare the occurrence of these functions across group and literary genre. These data lead her to conclude that the motivations for code-switching in her literary corpus bear a strong resemblance to those that motivate natural speech production and that biculturalism accounts for the majority of switches. Thus, such instances of literary CS serve to reproduce the bilingual and bicultural milieu of a given writer and his/her characters, which is crucial for establishing a text’s authenticity.

Katharina Müller contrasts the emergence of Italo-Brazilian literature within the Brazilian regions of Rio Grande do Sul and São Paulo. In Müller’s analysis, code-switching in literature is multi-functional: as with speech, it can lead to a variety of socio-pragmatic and stylistic effects while also demonstrating societal-level changes in language use. Unlike speech, however, it also serves to codify and thus standardize hybrid dialects, a process that has occurred throughout history, and is ongoing in the present day.

In our own article, Weston and Gardner-Chloros, we take up Lipski’s implied invitation (see above) to extract information from written CS, from an explicitly sociolinguistic standpoint. Our study is motivated by a desire to explore how CS in literature can inform our understanding of CS more generally, using this more considered and deliberate medium to reinforce or challenge findings from the spoken sphere. Such an approach is, by definition, less concerned with literary issues such as the impact of CS on characterization or narrative. Indeed, as we try to show, this difference of approach can sometimes lead to radically different analyses of CS from the two disciplines.

5 Conclusion

Just as spoken CS provides a window on exactly how bilingual speakers deploy their linguistic resources, so too does the use of CS in literature clarify underlying structures and intentions that may be less apparent in a monolingual text. Our hope in presenting these articles is partly that they will encourage greater collaboration between linguists and literature scholars, and partly that both linguists and literary scholars will benefit by referring to a broader range of types of CS. The conventions and constraints of speaking and writing may be different, but the broad semiotic consequences of setting up contrasts by alternating languages are common to both. Considering the vast knowledge being accumulated now through the study of historical sociolinguistics, we strongly believe that modern sociolinguistic studies of bilingualism would benefit from integrating this broader frame of reference. The argument is not simply that the exploration of multilingualism in contemporary literature will facilitate diachronic comparisons with the multilingual literature of earlier periods, although that is certainly the case. It is also that the insights and methods of historical sociolinguists deserve a wider currency within sociolinguistics as a whole, and can provide another dimension to sociolinguists’ partly habit-driven reliance on spontaneous spoken data. Last but not least, we would argue that multilingualism and CS are promising fields for interdisciplinary, linguistically based studies of literature.