Navigation – Plan du site
1. Standardisation and Variation in English Language(s)

The non-standard in writing: A look at West African and Southeast Asian literature



Cet article analyse la représentation de variétés non-standard de l'anglais dans les textes littéraires. En utilisant une méthode de corpus et en mettant l'accent sur des textes provenant d'Asie du Sud-Est et d'Afrique de l'Ouest, et en incluant également des textes écossais à des fins de comparaison, l'article emploie une analyse quantitative qui démontre l'existence de profiles linguistiques distincts pour chacune des régions étudiées, qui pourtant suivent un principe commun. Une analyse diachronique montre que ces profiles ne sont ni statiques ni stables, mais changent d'une décennie à l'autre.

Haut de page

Texte intégral

1. Introduction

1Standard language and written language are intertwined concepts. This rings true for English in particular, given the role played by the increasing use of English in writing (Schaefer 206) and the introduction of the printing press in the acceptance, diffusion, and maintenance of the emergence of the current standard of English in the late Middle English period (Nevalainen and Tieken-Boon van Ostade 275–278; Moessner 172). The bond between standard and written language pertains to published materials, which are expected to use a more formal register than other written genres, such as e-mails or personal letters. However, the written medium of literary works, while generally employing the standard, abounds with instances of non-standard usage. The reasons for this can be manifold. One of the explanations that has been proposed relates to authors’ endeavour to render their characters’ oral speech in the written medium, that is using the non-standard as a mimetic device to add a layer of authenticity and realism to the direct speech featured in their works (Hakala; Hodson 196–236). Alternatively, the non-standard may be used to assign characters specific attributes commonly (if not stereotypically) associated with a given non-standard dialect (Agha, “Voice, Footing, Enregisterment”; Agha, Language and Social Relations; Lippi-Green). A further motivation for using the non-standard may have little to do with characters and the authenticity of their speech, but rather concerns an ideological stance taken by certain authors, which is to challenge the standard (Hodson 121) and support local varieties of a language by participating in the creation or preservation of a literary tradition which relies on local linguistic forms (Vicinus 189–195; Shorrocks 396–399).

2The question arises why literary texts should be approached from a linguistic perspective rather than left entirely to the academic field of literary criticism. After all, Edgar W. Schneider's categorisation of text types according to their proximity to speech (Schneider 73), which distinguishes five distinct categories, namely (1) Recorded (2) Recalled (3) Imagined (4) Observed and (5) Invented, describes the text type of literary dialect as characteristic of invented speech, being based on hypothetic rather than real speech. Its value as a data source for the linguistic description of non-standard varieties is therefore very limited. Although the speech in literary texts may be invented, authors can rely on real-life models and “have at their disposal all the codes and resources of language” (Simpson 164). As such, the linguistic study of non-standard usage in literary texts has opened up an array of research agendas which go beyond variety description and documentation: (1) the identification and description of strategies used for conveying features of oral speech in a written medium, (2) the question whether these strategies are common to all literary traditions, or alternatively, whether each literary tradition, or even individual authors, employ their own methods for representing the oral non-standard, (3) the scope of represented features, which may range from a very mimetic approach, meaning an attempt to include all features found in the actual variety that can be represented by the writing system without resorting to specialised methods such a phonetic transcription, to a symbolic representation, which uses only the necessary amount of features to render the variety recognisable to the reader, (4) the selection of features used in a symbolic approach, and the position of such features on the indicator-marker-stereotype continuum (Labov), as well as their status as "enregistered" features (Agha, “The Social Life of Cultural Value”; Agha, “Voice, Footing, Enregisterment”; Agha, Language and Social Relations; Johnstone), particularly in light of differences between authors who are part of the speech community in question, and those who are external to it, and (5) the usage of literary dialects from a diachronic perspective, including an assessment of the influence of historical and societal developments on the prevalence and assertiveness on non-standard forms in literary works.

3The present paper addresses questions (2) and (5) in particular by creating and comparing non-standard feature profiles for different literary traditions, specifically those of non-standard varieties of English, and observing diachronic developments within these traditions.

Previous research

4The fact that this convergent area of literary studies and linguistics has already been treated in previous publications should come as no surprise. Earlier coverage of the topic includes monographs by Norman Blake and Norman Page. Blake takes a diachronic approach to the representation of non-standard language and its functions in literary works, concluding that in literature, the use of non-standard forms of language are associated with stereotypes of human types and behaviours. Page demonstrates that the language of fiction is of particular interest in that it enables researchers to confront the nature of speech itself as well as literary issues. Speech, specifically direct discourse, plays a central role in the creation of illusion. Chapter 3, entitled “Speech and Character: Dialect” (Page 55–96), makes the point that regional and social dialects are not entirely independent. Christian Mair defines a "typology of literary nonstandard" distinguishing three functional dimensions: firstly, the motivation for using the non-standard (mimesis versus symbolism); secondly, language attitudes (conventional versus critical); and thirdly, the degree of integration of the non-standard (loose, characterised by non-standard dialogue and standard narration, versus tight, utilising the hybridisation of narrative idiom or the fusion of narrative perspectives).

5The case of emerging literary traditions have also been discussed, for example by Josef Schmied (119–137), who dedicates the fifth chapter of his monograph on English in Africa to the use of English in African literature, concluding that English has become an African means of literary expression, representing, but not necessarily reflecting, the wide range sociolinguistic variation of English in Africa. More recent discussions of the use of English in African literature include a study by Eyamba Bokamba, which offers a qualitative analysis of prominent West African novels, focusing on language issues in creative writing by relating them with societal linguistic choices and language policies.

6Much of the work hitherto mentioned has provided important insights into the motivations and functions of dialect representation in literature, but remains mostly qualitative and impressionistic in nature. Increasingly, recent and ongoing case studies rely on a quantitative approach, for example a monograph dedicated to the representation of African-American Vernacular English in four novels (Minnick), a study of the phonology of the contemporary humorous localised dialect literature of Liverpool (Honeybone and Watson), or a project investigating the literary dialect representation in British fiction from 1800 to 1836 (Hodson and Broadhead).

7The publications mentioned evidently do not cover the wealth of research on the topic. Still, most endeavours are narrowed in their scope, be it the representativeness of the analysed material (four novels studied by Minnick), the linguistic category under observation (sole focus on the representation of phonology by Honeybone and Watson), the period covered (focus on the early nineteenth century by Hodson and Broadhead), or a regional focus (focus on Africa by Bokamba). As such, there is, at least to my knowledge, no large-scale corpus-based comparative study of various dialects in literary works.

8To remedy this, a project named “The representations of oral varieties of language in the literature of the English-speaking world”, and started at the University of Strasbourg in 2013, compares the representation of non-standard language in Anglophone literature, covering multiple English-speaking regions, linguistic features, and spanning several decades. The findings from project-related publications pertain to the expression of local identities via linguistic means in literary prose (Paulin and Percillier), the creation of character and narrator profiles via varying densities and types of non-standard features (Percillier and Paulin, “A Corpus-Based Investigation of World Englishes in Literature”), and the accuracy of represented linguistic features in comparison to actual varieties (Percillier and Paulin, “Postcolonial Literature and World Englishes”).


9Ideally, a comparative study of the use of non-standard language in literary texts should draw from all English-speaking regions of the world and their respective English-language literary traditions. This includes areas in which English serves the function of a first language, so-called English as a Native Language (henceforth ENL) speech communities, or the Inner Circle to use the terminology defined by Braj Kachru, as well as English as a Second Language (henceforth) ESL or Outer Circle speech communities. English as a Foreign Language (EFL) communities, or the Expanding Circle, are not included as the existence of a local literary tradition in English is deemed too unlikely. The world regions worth investigating are shown in Figure 1.

Figure 1: Map of Inner Circle (blue) and Outer Circle countries (orange) with likely English-language literary traditions, grouped by world regions

Figure 1: Map of Inner Circle (blue) and Outer Circle countries (orange) with likely English-language literary traditions, grouped by world regions

10The overview given in Figure 1 shows a grouping of English-speaking countries by world regions, albeit with a number of isolates, such as Belize and Guyana, which, although located in Central and South America respectively, may or may not be considered part of the Anglophone Caribbean. Namibia, although included in the map, became an Anglophone country only recently when it appointed English as its sole official language despite having no English-speaking communities (Buschfeld and Kautzsch). It remains to be seen if, and how quickly, a local literary tradition emerges in that country.

  • 1 The West African texts were selected and annotated by Catherine Paulin (University of Strasbourg).

11Given that analysing the non-standard in literary texts requires in-depth knowledge of local varieties as well as at least rudimentary knowledge of local languages other than English, any attempt to cover the English-speaking world as a whole will result in a superficial analysis at best. It is therefore preferable, in a first phase, to focus on regions well-known to the author and subsequently encourage experts on other varieties to join the project and analyse texts from their respective regions using a common methodology. As such, the present chapter will focus on Southeast Asia and West Africa,1 and will also include Scottish texts as an Inner Circle point of comparison.

Outline of the present paper

12The present paper first provides an overview of the project methodology, followed by a quantitative analysis of non-standard features observed in the literary texts from regions under investigation, the goal being to determine to what extent the literary traditions of these regions use the non-standard. Subsequently, diachronic developments are investigated in order to observe whether any observed feature patterns are stable practices of a given literary tradition or whether they are rather prone to fluctuations.

2. Methodology

Text selection

13From the three main literary text types, poetry and drama may pose challenges as for poetry, rhyme and metre may influence syntax in addition to non-standard grammar, while in drama, usually written to be performed orally, non-standard features such as accent may be indicated in stage directions rather than found in the dialogue itself. This does not imply that these texts are not worth including, but simply that the first text type to be addressed is prose.

14While short stories can be included in their entirety, an important question concerns the choice of using whole novels or samples. An advantage of using entire novels lies in the ability to examine the use of the non-standard throughout a given novel, including the role of the non-standard in structuring the work. However, using entire novels also means restricting the number of texts and authors, while only using excerpts from novels allows for a greater variety of texts and authors and consequently a better representativeness of the literary landscape at hand.

From printed page to annotated corpus

15As a vast majority of selected texts are not in electronic form, it is necessary to digitise them by scanning and applying Optical Character Recognition (henceforth OCR) software. In spite of the accuracy of OCR, errors do occur, typically the exchange of similar characters such as “O” (capital o) and “0” (zero), “I” (capital I), “l” (lower case l) and “1” (one), as well as character combinations mistaken for a single character or vice versa, such as “rn”/“m” and “cl”/“d”. For this reason, all scanned texts were verified manually.

16Further, the characters used for quotations were made consistent for all texts. Single quotations marks (‘ ’) were automatically changed to double quotation marks (“ ”) using a Python (Python Software Foundation) script. This was done in order to allow a subsequent Python script which adds basic XML (eXtensible Markup Language) annotation to mark passages of direct speech. As the character for single quotation marks can be confused with the apostrophe character, using double quotation marks proved to be the most consistent method for achieving accurate marking. In addition to annotating character speech, the script also assigns each paragraph a running number for identification purposes.

17The texts are then annotated manually for any non-standard features encountered. In order to avoid inconsistent marking, any manual typing of XML tags should be avoided. For this reason, a simple XML editor named XmlCat (Percillier, XmlCat), short for "Xml Corpus Annotation Tool", was developed to fit the project's needs.

18The annotation process in XmlCat involves selecting a stretch of text containing a non-standard linguistic feature and specifying descriptive information about said feature. The software then inserts valid XML tags around the selected stretch of text: an opening tag in the format <category feature=" " attribute=" "> and a closing tag in the format </category>. The possible attributes are "observed" (non-standard variant observed in the text), "standard" (alternative expected in a standard variety), "language" (for elements in a language other than English) and "meaning" (for lexemes having other or additional semantic content or foreign language content). The used tag scheme distinguishes five main categories, with are shown in Table 1 using short text snippets as examples.

Table 1: Overview of the main categories of the tag scheme and tag structure


Untagged example

Tagged example


in de street

in <phonology feature="consonant_place" observed="d" standard="D">d</phonology>e street


I marry that woman when she be like this.

I <grammar feature="VP_tense" observed= "PRS_marry" standard="PST_married"> marry</grammar> that woman when she <grammar feature="VP_tense" observed= observed="INF_be" standard="PST_was"> be</grammar> like this.


You got handphone?

You got <lexis feature="lexeme" standard= "mobile phone">handphone</lexis>?


The old Pak Ciks in checked sarongs who stop for teh tarik

The old <code feature="mixing" language= "Malay" meaning="HON_uncle">Pak Cik</code>s in checked <code feature= "mixing" language="Malay" meaning="long loin cloth">sarong</code>s who stop for <code feature="mixing" language="Malay" meaning="pulled tea">teh tarik</code>


So anywey the priest comes.

So <spelling observed="anywey" standard= "anyway">anywey</code> the priest comes

19The distinction between the categories “phonology” and “spelling” deserves further explanation. Features tagged under the category “phonology” correspond to a pronunciation that is different from the standard pronunciation, whereas features classified as “spelling” refer to spelling substitutions that have no bearing on pronunciation. The latter category corresponds to the concept of “eye dialect”, as coined by George Philip Krapp (228).

Methods of data analysis

20The annotated texts can be analysed in three ways. Being able to read the annotated texts is important for a qualitative analysis. As the presence of tags makes reading cumbersome, a method for convenient reading while maintaining access to the annotation is crucial. Given that XML and HTML (HyperText Markup Language) are both derived from SGML (Standard Generalized Markup Language), XML files can be opened and displayed in web browsers (normally used to display HTML documents on the web). Doing so will show the text without any tags. Using CSS (Cascading Style Sheets), normally used for styling websites, the tags can be defined to highlight tagged passages, and further to display the tag contents when the mouse cursor is hovered over the highlighted passages. With this method, tags do not hinder reading the texts while remaining accessible.

21For analyses requiring searches for specific tags and seeing their direct contexts, concordance software such as AntConc (Anthony) can be used. For a more thorough quantitative analysis, a Python script was written which collects all tags and lists them in a CSV (Comma Separated Value) file for easy importation into R (R Core Team), a programming language for statistic analysis and plotting.

22A more detailed account of the project methodology, including the source code for the Python and R scripts, as well as the CSS, is given in a dedicated publication (Percillier, “Creating and Analyzing Literary Corpora”).

23The goal of the text collection process is to achieve a word count of approximately 100,000 word per corpus component. For the Outer Circle regions of Southeast Asia and West Africa, that target has been reached. The Inner Circle component currently contains Scottish texts only, and will be augmented with texts from other Inner Circle countries to reach the 100,000 target. Importantly, the text selection process is not limited to canonical or particularly feature-rich texts, as such restrictions would result in a biased corpus. Instead, the sources from which texts were selected consist primarily of anthologies and short story collections. The only criterion that a text needed to fulfil was the presence of at least one non-standard feature. The current status of the corpus is shown in Table 2 and serves as a basis for analysis in the present paper.

Table 2: Overview of current corpus structure







Inner Circle



Hanley, Clifford

The Orange and the Green



Kelman, James

How Late It Was, How Late






McCabe, Brian

The Face



McLellan, Robert

The Cat



Rose, Dilys

Friendly Voices






Spence, Alan

Its Colours They Are Fine



Southeast Asia




Faoite, Marc de

Mamak Murder Mystery



Gomez, Brian




Maniam, K.S.

The Return



Samarasan, Preeta

Rukun Tetangga



Zaini, Adib

The Runner



Zaman, Dina

After Dark, My Love





Dalisay, Jose

Killing Time in a Warm Place



Soledad's Sister



Gamalinda, Eric

Fear Of Heights



Elvis Of Manila



Yes, Jesus Loves Me



Groyon, Vicente Garcia

The Sky over Dimas





Alfian Sa'at

The Borrowed Boy



Bosco, Don

Everyday will be like Sunday



Cheong, Colin

The Stolen Child



Poisson Ivy



Fernando, Lloyd

Scorpion Orchid



Goh Poh Seng

If We Dream Too Long



Jeyaretnam, Philip

Making Coffee



Lim Thean Soo

Ricky Star



Lim, Suchen Christine




Yeo, Robert

The Adventures of Holden Heng



West Africa




Aidoo, Ama Ata

Male-ing Names in the Sun



Darko, Amma

Beyond the Horizon





Lewis, Ophelia

The Dowry of Virgins



Heart Men





Achebe, Chinua

Things Fall Apart



No Longer at Ease



A Man of the People



Girls at War



Adichie, Chimamanda Ngozi

Half of a Yellow Sun



The Arrangers of Marriage






Ekwensi, Cyprian

People of the City






Emecheta, Buchi

The Joys of Motherhood



Double Yoke



Saro-Wiwa, Ken




Sierra Leone

Cole, Robert Wellesley

Kossoh Town Boy



3. Regional features profiles

  • 2 Figure 2 and all subsequent figures were generated using R (R Core Team) and the packages ggplot2 ( (...)

24In order to verify whether various literary traditions use different methods for representing the non-standard, a comparison of non-standard features grouped by category can provide insights into patterns of use. Such a comparison is given in Figure 2.2

Figure 2: Frequency of non-standard features per region, normalised to 100,000 words

Figure 2: Frequency of non-standard features per region, normalised to 100,000 words

25The most striking observation in Figure 2 is the clearly higher frequency of features in Scottish texts. In addition to varying frequencies, the three regions display different proportions of categories. These stark differences (χ2 = 3230.7, df = 8, p < 0.001***) concern not only frequencies, but also proportions of represented linguistic categories. A more comprehensible way of visualising the different variety profiles can be achieved by comparing the proportions of different categories in percent, as shown in Figure 3.

Figure 3: Proportion of non-standard features per region in percent

Figure 3: Proportion of non-standard features per region in percent

26The patterns shown in Figure 3 vary greatly: while Scottish texts mostly rely on phonology, Southeast Asian texts use code-switching features most frequently, while West African texts mostly contain grammatical features. In spite of these obvious differences, a common underlying pattern can be observed: rather than having an equal distribution across various linguistic categories, an obvious focus of non-standard representation is placed on a single linguistic category, accounting for more than half of all instances. The question remains what factors determine the different choices for each region. A working hypothesis links the choice of the dominant category to the perception of maximum distance to the relevant standard variety. This would mean that Scottish authors wishing to portray the non-standard consider the phonology of their variety to bear the most striking differences from the standard variety, while Southeast Asian authors stress code-switching practices and West African authors view local grammatical patterns as noticeably different. This pattern was already noticed at earlier stages of the data collection process (see Paulin and Percillier 64–65; Percillier and Paulin, “A Corpus-Based Investigation of World Englishes in Literature” 134–135), and the addition of further texts has only confirmed the pattern. It remains to be seen whether this pattern remains when the Inner Circle component is completed, or when further regions from the Outer Circle are represented in the corpus.

27A more fine-grained view of features represented in one region in comparison to others can be obtained by plotting Pearson residuals (observed frequencies minus expected frequencies, divided by the square root of expected frequencies), as shown in Figure 4. Residual values of 0 represent the expected value if all regions displayed the same patterns, so that positive values stand for higher frequencies than expected, and negative values for lower values than expected.

Figure 4: Pearson residuals of non-standard features per region

Figure 4: Pearson residuals of non-standard features per region

28In addition to the dominant categories noticeable from Figure 3, a secondary focus category can be discerned for Scotland and Southeast Asia, namely spelling and lexis respectively. The categories that are mostly shunned by authors of a given region can also be identified. The fact that Scottish authors do not make use of code-switching to the same extent as Southeast Asian authors does not come as a surprise, as their communities do not display the same multilingual usage patterns as postcolonial countries, however code-switching to Gaelic is a strategy that Scottish authors have at their disposal, but choose to disregard entirely. In a similar vein, the avoidance of phonological features by Southeast Asian writers cannot be said to originate in a lack of accent features in the region’s varieties (see Mesthrie for a synopsis of the phonology of Southeast Asian Englishes), but rather in the authors’ choice to focus on other categories which they deem more salient.

4. Variation within regions

29One should of course be careful with sweeping statements about possible patterns across regions, as texts may vary greatly within a region as to how they use non-standard language. In other words, are there any typical Scottish, Southeast Asian, or West African patterns of non-standard representation, or are the observations made above based on averages of widely different practices within a given region? To answer this, the feature density of individual texts is depicted in a box and whiskers plot in Figure 5. Such a plot visualises the distribution of values by drawing a box representing the interquartile range, i.e. 50% of the data around the median value, and "whiskers" that have a maximum length of 1.5 times the interquartile range. Data points that are not within reach of the whiskers can be considered outliers.

Figure 5: Box and whiskers plot of feature density per text, grouped by region

Figure 5: Box and whiskers plot of feature density per text, grouped by region

30The feature densities of individual texts shown in Figure 5, where each point represents an individual text, reveals that each region contains exactly one text that is more feature-dense than is characteristic for the region. The outliers for Scottish and West African texts are extremely remote from the trend set by the other texts within the region, whereas the outlier for Southeast Asian texts is comparatively close. The outliers are "The Cat" (McLellan) for Scottish texts, "Rukun Tetangga" (Samarasan) for Southeast Asian texts, and an excerpt from Sozaboy (Saro-Wiwa, Chapters 1-2) for West African texts. The contrast between an outlier text and a text representing the average is illustrated for West African texts by excerpts (1) and (2).

    Although, everybody in Dukana was happy at first.
    All the nine villages were dancing and we were eating plenty maize with pear and knacking tory under the moon.
    Because the work on the farm have finished and the yams were growing well well. And because the old, bad government have dead, and the new government of soza and police have come. (Saro-Wiwa 1)

  2. Customs formalities here took thrice as long as at Liverpool and five times as many officials. A young man, almost a boy in fact, was dealing with Obi's cabin. He told him that the duty on his radiogram would be five pounds.
    "Right," said Obi, feeling his hip-pockets. "Write a receipt for me." The boy did not write. He looked at Obi for a few seconds, and then said: "I can be able to reduce it to two pounds for you."
    "How?" asked Obi.
    "I fit do it, but you no go get Government receipt." (Achebe 24)

31One of the obvious differences between excerpts (1) and (2) lies in the narrator type, as excerpt (1) features an intradiegetic and homodiegetic first person narrator, whereas excerpt (2) has an extradiegetic and heterodiegetic third person narrator. As the narrator in excerpt (1) represents a character voice, it is not surprising to observe a higher feature density. However, the integration of the non-standard goes further, as even the meta-text makes use of the non-standard, as seen in the chapter heading "LOMBER ONE" ("Number One"). Such a degree of integration is untypical of other texts from the region, which explains the status of Sozaboy as an outlier.

32A further observation based on Figure 5 is that Southeast Asian and West African texts do not appear to be as different in terms of density as previously suggested by Figure 2. As the box plots for Southeast Asian and West African texts look very similar, the possibility that the density differences observed previously are due to outliers. A non-parametric statistical test, such as the Wilcoxon Rank Sum Test, can compare two samples without being thrown off by outliers. Such a test fails to find any statistical differences between the densities of Southeast Asian and West African texts (W = 181, p = 0.8777), whereas there are significant differences between the feature densities of Scottish and Southeast Asian texts (W = 16, p = 0.0002717***), as well as between Scottish and West African texts (W = 120, p = 0.00147**).

33Considering this, the feature profiles drawn in the previous section should be revisited with the exclusion of outliers. A contrast of proportional feature profiles with and without outliers, given in Figure 6, shows minimal deviations from the patterns first observed in Figure 3.

Figure 6: Proportion of non-standard features per region in percent, all texts (top, corresponds to Figure 3) and without outliers (bottom)

Figure 6: Proportion of non-standard features per region in percent, all texts (top, corresponds to Figure 3) and without outliers (bottom)

34On the other hand, the normalised frequencies without the outlier texts, shown in the bottom panel of Figure 7, depart from the original pattern observed in Figure 2, as the normalised feature frequencies for Southeast Asian and West African texts now appear to be very similar. This suggests an Inner Circle/Outer Circle divide regarding feature densities, although it remains to be seen whether the addition of texts from further Inner Circle regions maintains this separation.

Figure 7: Frequency of non-standard features per region, normalised to 100,000 words, all texts (top, corresponds to Figure 2) and without outliers (bottom)

Figure 7: Frequency of non-standard features per region, normalised to 100,000 words, all texts (top, corresponds to Figure 2) and without outliers (bottom)

5. Diachronic developments

35When considering the patterns shown in Figures 6 and 7, the question remains whether they are stable patterns characteristic of the literary landscape in general, or whether they fluctuate across time. For the completed components of the corpus, i.e. those comprising Southeast Asian and West African texts, an analysis of feature profiles by decade can be performed.

Figure 8: Frequency of non-standard features by decade in Southeast Asian texts, normalised to 10,000 words

Figure 8: Frequency of non-standard features by decade in Southeast Asian texts, normalised to 10,000 words

36For Southeast Asian texts, the coverage extends from the 1970s to the 2010s. The analysis of frequencies of non-standard features across decades for Southeast Asian texts, shown in Figure 8, reveals that the use of the non-standard cannot be said to be stable, nor can the changes be described as regular. From the 1970s to the 1990s, the density of non-standard features gradually decreases, then reaches a sudden peak in the following decade, only to drop again in the 2010s. The sharp rise from the 1990s to the 2000s, corresponding to an almost threefold increase, is in need of an explanation. The Speak Good English Movement in Singapore, launched in 2000 (Wee 99), may play a role in this development, as speakers of Singapore English, including local authors, may have reacted to a movement that aimed at promoting standard English at the expense of the local variety by emphasising their vernacular in order to safeguard their means of expressing their local identity in English. Interestingly, the sudden increase in question does not correspond to a higher rate of code-switching and code-mixing practices, which remain relatively stable. Rather, grammatical and lexical features undergo rapid growth, only to return to their previous status in the following decade.

Figure 9: Proportion of non-standard features by decade in Southeast Asian texts, in percent

Figure 9: Proportion of non-standard features by decade in Southeast Asian texts, in percent

37Complemented by a proportional view shown in Figure 9, the analysis also reveals that code-switching and code-mixing features are not the dominant category in a consistent manner as suggested earlier in Figure 3. At the onset of the covered period, i.e. the 1970s, grammatical features are in fact the dominant category. In the following decades, code-switching and code-mixing features dominate the representation of local linguistic features, although the sharp rise of lexical and grammatical features, already mentioned in the discussion of Figure 8, created a more balanced profile of three almost equal categories for the spell of a decade.

Figure 10: Frequency of non-standard features by decade in West African texts, normalised to 10,000 words

Figure 10: Frequency of non-standard features by decade in West African texts, normalised to 10,000 words

38For West African texts, the coverage extends from the 1950s to the 2010s. The frequencies of non-standard frequencies per decade, shown in Figure 10, reveal a singular peak in the 1980s, with a feature density comparable to that observed for Scottish texts in Figure 2. As addressed in the previous section, this surge is due to an exceptionally feature-dense text taken from the novel Sozaboy by Nigerian author Ken Saro-Wiwa, first published in 1985. The non-standard title of the novel, meaning “soldier boy”, already hints to the intense use of the non-standard. With a normalised density of 1,626 features per 10,000 words, the text is denser than the average observed for Scottish texts. Due to the presence of this clear outlier, the feature densities in other decades appear to be low, although they are in fact as high or higher than in Southeast Asian texts (see Figure 8).

Figure 11: Proportion of non-standard features by decade in West African texts, in percent

Figure 11: Proportion of non-standard features by decade in West African texts, in percent

39The proportional feature profiles per decade in West African texts, shown in Figure 11, suggest that there is no consistent dominant category used to represent the local non-standard across decades, and in a more striking manner than in Southeast Asian texts (see Figure 9). West African writers may feel the need to forge forms of expression that create a tension between realism and creativity in their writing. Language issues are central and writers draw from a diversified range of linguistic means which vary according to the themes they develop, such as the rural or urban set up in which the story unfolds, the difficulties or even horrors their society undergoes, or the political role they believe literature plays in society. Generally speaking, grammatical features are the dominant category in West African texts, as determined previously in Figure 3, however they are a minor category at the onset of the period covered (1950s), and are reduced to a sliver in the 2000s. Other categories exhibit similar fluctuations: phonological features are clearly in the minority from the 1970s to the 1990s, then dominate in the 2000s, only to be entirely absent in the following decade; lexical features alternate from being marginal in certain decades, e.g. in the 1960s and 2000s, to being central in other decades, e.g. in the 1990s and 2010s; code-switching and code-mixing features appear to alternate almost regularly from a substantial role (ca. 25%) to a marginal role (ca. 10%) from decade to decade.

6. Conclusion

40The quantitative analysis undertaken in the present study showed varying densities for regions under investigation, as well as distinct feature profiles, which, although different, follow a common underlying pattern which selects a linguistic category on which to focus the representation of the non-standard. The choice of this category may lie in the perceived maximum distance from the respective standard variety. Further, the identification and exclusion of outliers has revealed that in addition to these different profiles, there appears to be a divide between Inner Circle and Outer Circle texts regarding feature densities. These differences may be due to a longer English-language literary tradition in Inner Circle regions. This deserves further investigation once further Inner Circle regions are added to the corpus.

41From a diachronic perspective, observed changes from decade to decade, which on occasion prove to be important, suggest that authors may react to social and political factors, such as the Speak Good English Movement in Singapore. Sudden changes can also be due to the idiosyncrasies of a single text or author, as in the case of Sozaboy by Ken Saro-Wiwa. This raises a question pertaining to the existence of author-specific practices of representation in addition, or even in opposition, to variety-specific practices. Further, the fluctuations observed from decade to decade indicate that authors may also react to previous representations of their variety by avoiding simple imitation of their predecessors and repeatedly devising innovative ways of representing their local variety in their works.

42The explanations offered for the choice of the dominant linguist category in a given region, as well as for the observed diachronic changes across decades, remain speculative, no matter how plausible they may be. A method for obtaining more tenable explanations lies in providing questionnaires to authors and publishers, thereby asking them to provide information on attitudes towards non-standard language, the influence of social, cultural, or political events, as well as other writers and their practices, and the issue of combining the integration of local cultural and linguistic identity with the commercial goal of reaching an international audience, alongside the different views that authors and publishers may have on the topic.

43A further next logical step in the continuation of the research project includes corpus growth. This does not equate to a sheer increase of the corpus size, but implies coverage of further Outer Circle regions, as well as the completion of an Inner Circle component, as Scottish texts cannot serve as an Inner Circle yardstick by themselves. For a diachronic analysis in new corpus components, care should be taken to collect a balanced spread of texts across decades for the entire period covered.

44The distinctions between narratorial passages and character speech, narrator profiles, and character profiles have already been addressed in an earlier publication (Percillier and Paulin, “A Corpus-Based Investigation of World Englishes in Literature” 135–141). An aspect that remains unexplored as of yet is the use of the non-standard in different types of character discourse, be it direct discourse, indirect discourse, free direct discourse, or free indirect discourse. The question is not limited to how readily the non-standard occurs in each discourse type, but also relates to how character discourse that is not overtly marked by quotatives and/or quotation marks, i.e. free direct and free indirect discourse, can be made apparent as not belonging to narrator discourse via the use of non-standard linguistic features.

45Lastly, a cognitive approach, possibly of an experimental type, could provide a better comprehension of the workings of symbolic representation, that is a representation that does not reflect the real language variety to its full extent. For instance, would speakers of a given variety, when asked to read aloud a text containing mostly grammatical and lexical features, also produce the phonological features of the variety even when these are not represented in the text? In other words, could a partial/symbolic representation serve as a cognitive trigger leading readers to “fill the gaps” for themselves, thus effectively generating a full/mimetic representation in the course of the reading process? Such a symbolic representation would have the benefit of functioning on two separate levels of interpretation for different readers: readers familiar with the variety, who may recreate it for themselves, and a wider international audience less familiar with the variety, for whom the representation would not yield a mimetic reading, but instead remain symbolic in order to mark the non-standard language as noticeably different.

Haut de page


Achebe, Chinua. No Longer at Ease. Penguin, 2010.

Agha, Asif. Language and Social Relations. Cambridge UP, 2007.

Agha, Asif. “The Social Life of Cultural Value.” Language & Communication, vol. 23, no. 3-4, July 2003, pp. 231–273, doi:10.1016/S0271-5309(03)00012‑0.

Agha, Asif. “Voice, Footing, Enregisterment.” Journal of Linguistic Anthropology, vol. 15, no. 1, June 2005, pp. 38–59, doi:10.1525/jlin.2005.15.1.38.

Anthony, Laurence. AntConc. 2016,

Blake, Norman F. Non-Standard Language in English Literature. Deutsch, 1981.

Bokamba, Eyamba G. “African Englishes and Creative Writing.” World Englishes, vol. 34, no. 3, July 2015, pp. 315–335, doi:10.1111/weng.12145.

Buschfeld, Sarah, and Alexander Kautzsch. “English in Namibia: A First Approach.” English World-Wide, vol. 35, no. 2, 2014, pp. 121–160, doi:10.1075/eww.35.2.01bus.

Hakala, Taryn. “A Great Man in Clogs: Performing Authenticity in Victorian Lancashire.” Victorian Studies, vol. 52, no. 3, 2010, pp. 387-412, doi:10.2979/vic.2010.52.3.387.

Hodson, Jane. Dialect in Film and Literature. Palgrave Macmillan, 2014.

Hodson, Jane, and Alex Broadhead. “Developments in Literary Dialect Representation in British Fiction 1800–1836.” Language and Literature, vol. 22, no. 4, Nov. 2013, pp. 315–332, doi:10.1177/0963947013497876.

Honeybone, Patrick, and Kevin Watson. “Salience and the Sociolinguistics of Scouse Spelling: Exploring the Phonology of the Contemporary Humorous Localised Dialect Literature of Liverpool.” English World-Wide, vol. 34, no. 3, 2013, pp. 305–340, doi:10.1075/eww.34.3.03hon.

Johnstone, Barbara. “Ideology and Discourse in the Enregisterment of Regional Variation.” Space in Language and Linguistics, edited by Peter Auer et al., De Gruyter, 2013, pp. 107–127, doi:10.1515/9783110312027.107.

Kachru, Braj B. “Standards, Codification and Sociolinguistic Realism: The English Language in the Outer Circle.” English in the World: Teaching and Learning the Language and Literatures, edited by Randolph Quirk and Henry G. Widdowson, Cambridge UP; The British Council, 1985, pp. 11–30.

Krapp, George Philip. The English Language in America. The Century Company, 1925.

Labov, William. Sociolinguistic Patterns. Blackwell, 1972.

Lippi-Green, Rosina. English with an Accent: Language, Ideology, and Discrimination in the United States. Routledge, 1997.

Mair, Christian. “Literary Sociolinguistics: A Methodological Framework for Research on the Use of Nonstandard Language in Fiction.” Arbeiten Aus Anglistik Und Amerikanistik, vol. 17, no. 1, 1992, pp. 103–123.

McLellan, Robert. “The Cat.” Scottish Literature in the Twentieth Century. an Anthology, edited by David McCordick, Scottish Cultural Press, 2002, pp. 480–482.

Mesthrie, Rajend. “Synopsis: The Phonology of English in Africa and South and Southeast Asia.” Varieties of English, Volume 4: Africa, South and Southeast Asia, edited by Rajend Mesthrie, Mouton de Gruyter, 2008.

Minnick, Lisa Cohen. Dialect and Dichotomy: Literary Representations of African American Speech. University of Alabama Press, 2004.

Moessner, Lilo. “Chapter 10: Standardization.” Early Modern English, edited by Alexander Bergs and Laurel Brinton, De Gruyter, 2017, pp. 167–187, doi:10.1515/9783110525069-010.

Nevalainen, Terttu, and Ingrid Tieken-Boon van Ostade. “Standardisation.” A History of the English Language, edited by Richard Hogg and David Denison, Cambridge UP, 2008, pp. 271–311.

Page, Norman. Speech in the English Novel. Humanities Press International, 1988.

Paulin, Catherine, and Michael Percillier. “Oral Varieties of English in a Literary Corpus of West African and South East Asian Prose (1954–2013): Commitment to Local Identities and Catering for Foreign Readers.” Études de Stylistique Anglaise, no. 9, 2015, pp. 59–79,

Percillier, Michael. “Creating and Analyzing Literary Corpora.” Data Analytics in Digital Humanities, edited by Shalin Hai-Jew, Springer International Publishing, 2017, pp. 91–118, doi:10.1007/978-3-319-54499-1_4.

Percillier, Michael. XmlCat. 2015,

Percillier, Michael, and Catherine Paulin. “A Corpus-Based Investigation of World Englishes in Literature.” World Englishes, vol. 36, no. 1, Mar. 2017, pp. 127–147, doi:10.1111/weng.12208.

Percillier, Michael, and Catherine Paulin. “Postcolonial Literature and World Englishes: A Corpus-Based Approach of Modes of Representation of the Non-Standard in Writing.” International Journal of Literary Linguistics, vol. 6, no. 1, 2017, pp. 1–24, doi:10.15462/ijll.v6i1.102.

Python Software Foundation. Python. 2016,

R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, 2016,

Samarasan, Preeta. “Rukun Tetangga.” KL Noir: Red, edited by Amir Muhammad, Fixi Novo, 2013, pp. 39–56.

Saro-Wiwa, Ken. Sozaboy: a Novel in Rotten English. Longman, 2006.

Schaefer, Ursula. “Standardization.” Middle English, edited by Laurel Brinton and Alexander Bergs, De Gruyter, 2017, pp. 205–223, doi:10.1515/9783110525328-011.

Schmied, Josef J. English in Africa: An Introduction. Longman, 1991.

Schneider, Edgar W. “Investigating Variation and Change in Written Documents.” The Handbook of Language and Variation and Change, edited by J.K. Chambers et al., Blackwell, 2002, pp. 67–96.

Shorrocks, Graham. “Non-Standard Dialect Literature and Popular Culture.” Speech Past and Present: Studies in English Dialectology in Memory of Ossi Ihalainen, edited by Juhani Klemola et al., P. Lang, 1996, pp. 385–411.

Simpson, Paul. Language Through Literature: An Introduction. Routledge, 1997.

Vicinus, Martha. The Industrial Muse: A Study of Nineteenth Century British Working-Class Literature. Barnes & Noble Books, 1974.

Wee, Lionel. “‘Burdens’ and ‘Handicaps’ in Singapore’s Language Policy: On the Limits of Language Management.” Language Policy, vol. 9, no. 2, May 2010, pp. 97–114, doi:10.1007/s10993-009-9159-2.

Wickham, Hadley. Ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York, 2009,

---. Scales: Scale Functions for Visualization. 2016.

Haut de page


1 The West African texts were selected and annotated by Catherine Paulin (University of Strasbourg).

2 Figure 2 and all subsequent figures were generated using R (R Core Team) and the packages ggplot2 (Wickham, Ggplot2) and scales (Wickham, Scales).

Haut de page

Table des illustrations

Titre Figure 1: Map of Inner Circle (blue) and Outer Circle countries (orange) with likely English-language literary traditions, grouped by world regions
Fichier image/jpeg, 380k
Titre Figure 2: Frequency of non-standard features per region, normalised to 100,000 words
Fichier image/jpeg, 172k
Titre Figure 3: Proportion of non-standard features per region in percent
Fichier image/jpeg, 160k
Titre Figure 4: Pearson residuals of non-standard features per region
Fichier image/jpeg, 176k
Titre Figure 5: Box and whiskers plot of feature density per text, grouped by region
Fichier image/jpeg, 164k
Titre Figure 6: Proportion of non-standard features per region in percent, all texts (top, corresponds to Figure 3) and without outliers (bottom)
Fichier image/jpeg, 236k
Titre Figure 7: Frequency of non-standard features per region, normalised to 100,000 words, all texts (top, corresponds to Figure 2) and without outliers (bottom)
Fichier image/jpeg, 260k
Titre Figure 8: Frequency of non-standard features by decade in Southeast Asian texts, normalised to 10,000 words
Fichier image/jpeg, 180k
Titre Figure 9: Proportion of non-standard features by decade in Southeast Asian texts, in percent
Fichier image/jpeg, 176k
Titre Figure 10: Frequency of non-standard features by decade in West African texts, normalised to 10,000 words
Fichier image/jpeg, 184k
Titre Figure 11: Proportion of non-standard features by decade in West African texts, in percent
Fichier image/jpeg, 196k
Haut de page

Pour citer cet article

Référence électronique

Michael PERCILLIER, « The non-standard in writing: A look at West African and Southeast Asian literature », E-rea [En ligne], 15.2 | 2018, mis en ligne le 15 juin 2018, consulté le 20 juin 2018. URL : ; DOI : 10.4000/erea.6312

Haut de page



University of Mannheim
Michael Percillier is a postdoctoral researcher at the University of Mannheim. His research interests cover corpus linguistics, World Englishes, literary linguistics, and historical linguistics. The current paper derives from the research project “The representations of oral varieties of language in the literature of the English-speaking world”, undertaken at the University of Strasbourg.

Haut de page
  • Logo Laboratoire d’Études et de Recherche sur le Monde Anglophone
  • Logo DOAJ - Directory of Open Access Journals
  • OpenEdition Journals