1.Introduction

The emergence of linguistic styles coincides with the development of social indexicalities, that is, associations between specific sets of linguistic features and social meanings (Hernandez Campoy 2016Hernandez Campoy, Juan M. 2016 Sociolinguistic Styles. London: Wiley. ). Referred to as enregisterment, a corollary of linguistic focusing, this process is inherently interactional, and implies the replication of commonly interpreted linguistic features across speakers (Agha 2007Agha, Asif 2007 Language and Social Relations. Cambridge: Cambridge University Press.; Sharma and Rampton 2011Sharma, Devyani, and Ben Rampton 2011 Lectal Focusing in Interaction: A new Methodology for the Study of Superdiverse Speech. Queen Mary’s OPAL 22.). This study proposes a discussion of style in the context of African urban environments where code-switching (CS) practices are widespread. More specifically, it focuses on CS strategies deployed within the multilingual and multidialectal Ovambo community, Namibia’s main ethnic group, among which a strongly stereotyped distinction is made between individuals who grew up in the traditional Ovambo heartland and those who grew up in towns. The specific questions that the study raises are: What are the CS patterns used by Oshiwambo speakers across the stereotypical distinction between more urban and less urban individuals, and to what extent do they form part of a common stylistic repertoire? To answer that question, the study closely examines a naturalistic multilectal interaction between four young Ovambo Windhoek residents. Combining the methodological example set by Li Wei (1998)Li Wei 1998 The ‘why‘ and ‘how‘ Question in the Analysis of Conversational Code-switching. In Peter Auer (ed.), Code-switching in Conversation. Language, Interaction and Identity, 156–179. London: Routledge. with a matched-guise experiment – it takes a sequential and holistic sociolinguistic perspective to the participants’ CS patterns with a view to establish how focused and representative their CS practices are in both a sequential and indexical sense.

The article is organized as follows. Section 2 jointly introduces the notions of style and code-switching. Section 3 provides some background information on the Oshiwambo speech community and its socio-historical context. Section 4 presents the data and the methodology used to analyse them. Section 5 provides an aggregate quantitative account of intra- and inter-speaker variation patterns observable in the data. Section 6 looks at sequential regularity in CS patterns and at its possible motivations. Section 7 looks at regularity in CS patterns from a social indexicality perspective. Finally, Section 8 places the findings in the broad perspective of CS practices observed elsewhere in urban Africa.

2.Code-switching and style

Sociolinguistic style can be broadly defined as a ‘coordinated configuration of linguistic features’ (Coupland 2011Coupland, Nikolas 2011 The Sociolinguistics of Style. The Cambridge Handbook of Sociolinguistics, 138–156. Cambridge: Cambridge University Press. ) or as ‘systems of distinction’ (Irvine 2001Irvine, Judith T. 2001 ‘Style’ as Distinctiveness: the Culture and Ideology of Linguistic Differentiation. In Style and Sociolinguistic Variation, ed. by Penelope Eckert and John R. Rickford, 21–43. Cambridge: Cambridge University Press.). The sociolinguistic study of style was initially embedded in the variationist paradigm, whereby stylistic variants located along a [+/−standard] continuum are regarded as manifestations of contextual attention paid to speech, and conceptually distinguished from social variants, correlationally linked to social categories (Eckert and Rickford 2001Eckert, Penelope, and John R. Rickford 2001 Introduction. In Style and Sociolinguistic Variation, ed. by Penelope Eckert and John R. Rickford, 1–20. Cambridge: Cambridge University Press.; Labov 1972Labov, William 1972 Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press.). Increasingly, the study of sociolinguistic style and of language variation in general has become entangled with interactional perspectives on social meaning: Speakers deploy linguistic variants not only as a function of context or of social group membership, but also for enacting mutually opposable voices whose alternation signals strategic conversational shifts (Irvine 2001Irvine, Judith T. 2001 ‘Style’ as Distinctiveness: the Culture and Ideology of Linguistic Differentiation. In Style and Sociolinguistic Variation, ed. by Penelope Eckert and John R. Rickford, 21–43. Cambridge: Cambridge University Press.; Coupland 2001Coupland, Nikolas 2011 The Sociolinguistics of Style. The Cambridge Handbook of Sociolinguistics, 138–156. Cambridge: Cambridge University Press. ). In view of the tendency to map social meanings directly onto linguistic variants, Ochs (1992)Ochs, Elinor 1992 Indexing gender. In Rethinking Context: Language as an Interactive Phenomenon, ed. by Alessandro Duranti and Charles Goodwin, 335–358. Cambridge: Cambridge University Press. reminds that the semiotic linkage between the two evolves through distinct stages, which coincide with distinct types of indexicality. At first, linguistic variants come to index specific stances, that is, subjective orientations to ongoing talk via processes of replication across interactions. Over time, such ‘direct indexicalities’ may combine with ‘indirect indexicalities’, that is, semiotic associations stereotypically linking them to social categories through which they become enregistered as components of styles (see further Jaffe 2009Jaffe, Alexandra M. 2009 Introduction: The Sociolinguistics of Stance. In Stance. Sociolinguistic Perspectives, ed. by Alexandra M. Jaffe, 3–28. Oxford: Oxford University Press. ; Du Bois 2007Du Bois, John W. 2007 The Stance Triangle. In Stancetaking in Discourse. Subjectivity, Evaluation, Interaction, ed. by R. Engelbretson, 139–182. Amsterdam: Benjamins, ; Agha 2007Agha, Asif 2007 Language and Social Relations. Cambridge: Cambridge University Press.). The notion that there are different types of indexicality is reflected in indexicality scales such as the one proposed by Labov (1972)Labov, William 1972 Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press., who distinguishes between ‘indicators’, or low-indexicality linguistic variants that are unconsciously used, markers, which are salient enough for stylistic work, and stereotypes, which form the topic of social comment with or without reflecting actual language use (see further Hernandez-Campoy 2016). Initially focused on monolingual repertoires, reflections on style have increasingly become part and parcel of the study of CS in multilingual environments.

CS refers to the conversational alternation between outwardly distinguishable codes (Auer 1998Auer, Peter 1998 Introduction: Bilingual Conversation Revisited. In Code-switching in Conversation. Language, Interaction and Identity, ed. by Peter Auer, 1–24. London: Routledge.). Although ‘codes’ can be defined inclusively as ‘ways of speaking’, as it is used in this present study, the term CS is generally encountered in bilingualism studies where it is applied to the alternation between languages rather than between dialects or styles (cf. Bullock and Toribio 2009Bullock, Barbara E., and Almeida J. Toribio 2009 Themes in the Study of Code-switching. In The Cambridge Handbook of Linguistic Code-switching, ed. by Barbara E. Bullock and Almeida J. Toribio, 1–18. Cambridge: Cambridge University Press. ; see further Alvarez-Cacamo 1998Alvarez-Cacamo, Celsio 1998 From ‘Switching Code’ to ‘Code-Switching’: Towards a Reconceptualisation of Communicative Codes. In Code-Switching in Conversation, ed. by Peter Auer, 29–50. London: Routledge.). When described in structural terms, CS falls into four main categories. Insertional CS involves one dominant grammatical frame to which elements from the other language are adjusted. Congruent lexicalization is a typical by-product of frequent CS in contexts of minimal typological distance which makes the identification of one specific matrix language difficult. By contrast, alternational CS typically occurs with typological distance, and is typically manifested by switching in between clausal units. Akin to alternational CS is backflagging, which involves CS between clausal elements and syntactically peripheral ones, such as, in particular, discourse markers (Muysken 2013Muysken, Pieter 2013 Language Contact Outcomes as the Result of Bilingual Optimization Strategies. Bilingualism: Language and Cognition 16(4): 1–22. ). CS can arguably give rise to linguistically regular styles where insertions and discourse markers from the donor language acquire the status of borrowing while remaining salient enough to be semiotically opposable to their historical equivalents in the host language. Also, congruent lexicalization can ultimately give rise to ‘fused lects’ (Auer 1999Auer, Peter 1999 From Codeswitching via Language Mixing to Fused Lects: Toward a Dynamic Typology of Bilingual Speech. The International Journal of Bilingualism 3(4): 309–332. ), which may stylistically contrast with the original donor languages. Even though they do not give rise to fused lects, alternational CS patterns can still develop patterns of recurrence. The higher the frequency of alternational CS (and of CS in general), the lower the salience of individual switches, and the higher the probability that it may be constitutive of a multilectal style strategically opposable to other multilectal styles articulated around insertions and discourse markers, as well as to monolectal ways of speaking (cf. Auer 1999Auer, Peter 1999 From Codeswitching via Language Mixing to Fused Lects: Toward a Dynamic Typology of Bilingual Speech. The International Journal of Bilingualism 3(4): 309–332. ; Alvarez-Cacamo 1998Alvarez-Cacamo, Celsio 1998 From ‘Switching Code’ to ‘Code-Switching’: Towards a Reconceptualisation of Communicative Codes. In Code-Switching in Conversation, ed. by Peter Auer, 29–50. London: Routledge.; Meeuwis and Blommaert 1998Meeuwis, Michael, and Jan Blommaert 1998 A Monolectal View of Code-switching: Layered Code-switching among Zairians in Belgium. In Code-switching in conversation, ed. by Peter Auer, 76–100. London: Routledge.). Ultimately, the extent to which a CS pattern functions as a style can only be established by observing it in its interactional use.

One way of establishing from an analyst perspective whether CS patterns form part of a system of stylistic distinction is to assess how sequentially focused their distribution is. Turn-by-turn linguistic convergence in code choices may be accommodation-driven, and as such subconscious and devoid of stylistic function (Trudgill 2008Trudgill, Peter 2008 Colonial Dialect Contact in the History of European Languages: On the Irrelevance of Identity in New-dialect Formation. Language in Society 37: 241–280. ). However, turn-by-turn linguistic convergence can also form part of a strategy of ensuring conversational cohesion in complementarity with turn-by-turn linguistic divergence, which intervenes as a strategy for marking conversational shifts such as change in frame, footing and speaker alignment. An illustration comes in the form of convergent and divergent code choices that are typically effected for marking ‘preferred’ and ‘dispreferred’ second pair parts, respectively (Li Wei 1998Li Wei 1998 The ‘why‘ and ‘how‘ Question in the Analysis of Conversational Code-switching. In Peter Auer (ed.), Code-switching in Conversation. Language, Interaction and Identity, 156–179. London: Routledge.; Auer 1998Auer, Peter 1998 Introduction: Bilingual Conversation Revisited. In Code-switching in Conversation. Language, Interaction and Identity, ed. by Peter Auer, 1–24. London: Routledge.; see further Sacks 1995Sacks, Harvey 1995 Lectures on Conversation. London: Blackwell. ; Schegloff 2007Schegloff, Emmanuel A. 2007 Sequence Organization in Interaction. Volume 1: A Primer in Conversation Analysis. Cambridge: Cambridge University Press. ). The linguistic contrast between the alternating languages may on its own provide the salience that is necessitated for marking conversational shifts. However, a contrasting code can be chosen for marking a conversational shift for the reason that it additionally possesses social indexicality in the view of its user (Li Wei 1998Li Wei 1998 The ‘why‘ and ‘how‘ Question in the Analysis of Conversational Code-switching. In Peter Auer (ed.), Code-switching in Conversation. Language, Interaction and Identity, 156–179. London: Routledge.). Transition points between distinctive patterns of CS (e.g. insertional vs. alternational) can also possess salience as contrasting CS patterns have been shown to operate as distinctive styles in specific societal settings (Meeuwis and Blommaert 1998Meeuwis, Michael, and Jan Blommaert 1998 A Monolectal View of Code-switching: Layered Code-switching among Zairians in Belgium. In Code-switching in conversation, ed. by Peter Auer, 76–100. London: Routledge.). Ultimately, CS can be deployed independently of sequential factors to project specific stances and/or specific social values linked to broader social imageries linked to group membership. Uses of CS in such functions can be established via ethnographic means in the form of members’ perspectives and of general sociolinguistic data relevant to defining the context of the interaction (Auer 2007Auer, Peter 2007 Introduction. In Style and Social Identities: Alternative Approaches to Linguistic Heterogeneity, ed. by Peter Auer. 1–24. Berlin: De Gruyter.; Jaffe 2007Jaffe, Alexandra M. 2007 Codeswitching and Stance: Issues in Interpretation. Journal of Language, Identity & Education 6(1): 53–77.; see further Gumperz 1982Gumperz, John J. 1982 Discourse Strategies. Cambridge: Cambridge University Press. ). In the interest of both systematic description and in-depth sociolinguistic contextualization, this study proposes to combine sequential and ethnographic approaches for analysing multilectal language behaviours observed among a sample of mutually interacting urban Ovambos with contrastive sociolinguistic biographies.

3.Oshiwambo speakers in the context of Namibia’s sociolinguistic ecology

Namibia is a multi-ethnic and multilingual society with a strong legacy of segregation inherited from South African colonial rule. Under the South African regime (1915–1990), Namibia (then known as ‘Southwest Africa’, henceforth ‘SWA’) was subjected to South Africa’s apartheid legislation whose purpose was to support White minority rule. To that end, the South African regime enforced ethnic separation through territorialisation in ethnic ‘homelands’, as well as through compulsory mother-tongue instruction in primary school (UNIN 1981UNIN 1981 Toward a Language Policy for Namibia. Lusaka: United Nations Institute for Namibia.). Although it also had English as an official language, SWA gave prominence to Afrikaans in public life and in secondary education (Maho 1998Maho, Jounie F. 1998 Few People, many Tongues. The Languages of Namibia. Cape Town: Gamsberg Macmillan.; UNIN 1986UNIN 1986 Namibia. Perspectives for National Reconstruction and Development. Lusaka: United Nations Institute for Namibia.). As a result, Afrikaans had by 1990 consolidated the status of inter-ethnic lingua franca that it had begun to acquire from the late 18th century in the territory’s southern and central regions (cf. Prinsloo et al. 1982Prinsloo, K. P., D. J. Stoker, A. M. Lubbe, A. E. Strydom, H. A. Engelbrecht and D. P. van Vuuren 1982 Aspekte van taal- en kommunikasie-aangeleenthede in SWA/Namibië. Volume 13. Pretoria: Human Resources Research Council.; Stals and Ponelis 2001Stals, Ernst, and Fritz Ponelis 2001 Só het Afrikaans na Namibië gekom: Afrikaans-Hollands in Namibië vóór die koloniale tyd. Windhoek: Gamsberg McMillan.). In the northern districts, however, its presence was weaker while English was more visible as a medium of instruction, such as in particular in ‘Ovamboland’, that is, the territorial unit designated as homeland for the demographically dominant Ovambo ethnolinguistic group (cf. Fourie 1991Fourie, D. J. 1991 Language Conflict in Owamboland. South African Journal of Linguistics 9(3): 69–73. ; Harlech Jones 1990Harlech-Jones, Brian 1990 ‘You taught me language’. The Implementation of English as a Medium of Instruction in Namibia. Oxford: Oxford University Press.). The abolition of apartheid by independence brought radical socio-political transformations. The Ovambo majority effectively gained political hegemony, and the dismantlement of influx controls triggered a wave of migration from the rural northern districts towards the more urbanized central regions (Elischer 2013Elischer, Sebastian 2013 Political Parties in Africa: Ethnicity and Party Formation. Cambridge: Cambridge University Press. ; Pendleton 1996Pendleton, Wade 1996 Katutura: A Place where We Stay. Life in a Post-apartheid Township in Namibia. Athens: Ohio University Press.). At a linguistic level, English became Namibia’s only official language, as well as the main medium of instruction at the expense of indigenous languages and Afrikaans (Brock-Utne and Holmarsdottir 2001Brock-Utne, Birgit, and Hall B. Holmarsdottir 2001 The Choice of English as a Medium of Instruction and its Effects on the African Languages in Namibia. International Review of Education 47(3–4): 293–322. ).

Oshiwambo is Namibia’s most spoken home language (48.9%), followed by Khoekhoegowab (11.3%), Afrikaans (10.4%), Rukavango (9%), and Otjiherero (8.6%). English is still only marginally declared as a home language with 3.4%. Post-independence migration patterns have brought some changes to the geographic distribution of home languages. Oshiwambo became the main home language in two regions outside of historical Ovamboland, namely, in the urbanized regions of Khomas, where Windhoek, the capital city, is located, and Erongo (Namibian Statistics Agency 2012Namibia Statistics Agency 2012 Namibia 2011. Population & Housing Census Main Report. Windhoek: National Planning Commission., 2003Namibia Statistics Agency 2003 Namibia 2001. Population & Housing Census. Windhoek: National Planning Commission.). Since its introduction as a medium of instruction from Grade 4, literacy rates in English have dramatically increased, to the point of overtaking literacy rates in Afrikaans (Namibia Statistics Agency 2003Namibia Statistics Agency 2003 Namibia 2001. Population & Housing Census. Windhoek: National Planning Commission.). The scope for indigenous languages to develop as inter-ethnic lingua francas has been limited. Post-independence surveys of school language policies show that English and/or Afrikaans rather than indigenous languages are used as mediums of instruction at ethnolinguistically diverse urban schools (cf. Wolfaardt 2001Wolfaardt, Dolores 2001 Facilitating Learning: An Investigation of the Language Policy of Namibian Schools. Doctoral Dissertation. Bellville: University of the Western Cape.; see further Stell 2016Stell, Gerald 2016 Current Trends in Linguistic Diversity in Post-apartheid Windhoek: A Qualitative Appraisal. Language matters 47(3): 326–348. ).

Oshiwambo, Namibia’s most spoken language, is a Southwest Bantu language. The term Oshiwambo collectively refers to eight dialects, out of which three have been used as written standards, namely Oshindonga, Oshikwanyama, and Otshikwambi. The former two were introduced as mediums of instruction during the South African period and are still used as such (Fivaz 1986Fivaz, Derek 1986 A Reference grammar of Oshindonga (Wambo). Windhoek: The Academy.; Legère et al. 2000Legère, Karsten, Richard Trewby, and Mariana van Graan 2000 The Implementation of the Namibian Language Policy in Education: Lower Primary Grades and Pre-service Teacher Education. Windhoek: Namibia Educational Research Association.). While just before independence Oshikwanyama had a larger number of native speakers than Oshindonga within Ovamboland, Oshindonga was more widespread among Ovambos outside of Ovamboland and it has been in more widespread use as a written standard (cf. Malan 1990Malan, J. S. 1990 Aspekte van identiteitsvorming en – verandering onder die Wambo. South African Journal of Ethnology 13(1): 1–10.; Fivaz 1986Fivaz, Derek 1986 A Reference grammar of Oshindonga (Wambo). Windhoek: The Academy.; Prinsloo et al. 1982Prinsloo, K. P., D. J. Stoker, A. M. Lubbe, A. E. Strydom, H. A. Engelbrecht and D. P. van Vuuren 1982 Aspekte van taal- en kommunikasie-aangeleenthede in SWA/Namibië. Volume 13. Pretoria: Human Resources Research Council.).11.Post-independence censuses make no distinction between Oshiwambo dialects. The latest estimates come from Prinsloo et al. (1982)Prinsloo, K. P., D. J. Stoker, A. M. Lubbe, A. E. Strydom, H. A. Engelbrecht and D. P. van Vuuren 1982 Aspekte van taal- en kommunikasie-aangeleenthede in SWA/Namibië. Volume 13. Pretoria: Human Resources Research Council.. While there could be levelling towards the standard varieties, incipient koineization might also be at work, as Haugh (2013)Haugh, Wendy 2013 Against Apartheid: Nationalists, Cosmopolitans, and Language Ideology in Northern Namibia. Anthropological Quarterly 86(1): 191–220. observes that the deployment of multidialectal Oshiwambo repertoires is regarded as a token of social mobility. English is also present in local repertoires with a similar symbolic meaning, and it is as a result significantly used in monolingual mode among younger generations in the north, or in combination with Oshiwambo (Ausiku 2010Ausiku, Justus K. 2010 An Evaluation of the Implementation of the Namibian Language-in-education Policy in the Upper Primary Phase in Oshana Region. MA dissertation. University of South Africa.). Ovambos born in cities outside the historical Ovambo heartland are perceived as sociolinguistically distinct. Referred to as ‘Ombwiti’ (‘rootless’, see Pendleton 2002Pendleton, Wade 2002 Katutura and Namibia: The People, the Place, and the Fieldwork. In Chronicling Cultures: Long Term Field Research in Anthropology, ed. by Robert Kemper and Anya P. Royce, 34–58. Walnut Creek, CA: Altamira Press.), they are typically exposed to other Namibian languages and to Afrikaans, while they may have lost fluency in Oshiwambo (Stell 2016Stell, Gerald 2016 Current Trends in Linguistic Diversity in Post-apartheid Windhoek: A Qualitative Appraisal. Language matters 47(3): 326–348. ; see further Prinsloo et al. 1982Prinsloo, K. P., D. J. Stoker, A. M. Lubbe, A. E. Strydom, H. A. Engelbrecht and D. P. van Vuuren 1982 Aspekte van taal- en kommunikasie-aangeleenthede in SWA/Namibië. Volume 13. Pretoria: Human Resources Research Council.). While there is a negative perception of Ombwiti linguistic behaviours among Northerners, long-urbanized Ovambos derogatorily refer to northern-born Ovambos as ‘Kashuku’ (‘foreigners who know nothing’, see Peyroux 2004Peyroux, Elisabeth 2004 Windhoek, capitale de la Namibie. Changement politique et recomposition des périphéries. Johannesburg: IFAS.). Ovambos have since the 1980s been massively migrating to southern towns, especially Windhoek where they now form the largest ethnolinguistic group (Namibia Statistics Agency 2003Namibia Statistics Agency 2003 Namibia 2001. Population & Housing Census. Windhoek: National Planning Commission.; Sohn 2003Sohn, C. 2003 Changement gestionnaire et recompositions urbaines post-apartheid. La question foncière à Windhoek. Doctoral dissertation. Strasbourg: Université Louis Pasteur.). It is open to question to what extent urban Ovambos, who can be expected to have varying biographies of linguistic exposure and varying personal ties to the north, are developing a linguistic consensus that spans oppositions between urban and rural indexicalities.

4.Sample and methods

The speech data used for this study consist of a single informal interaction between four self-declared female Ovambo students recruited through the ‘friend of a friend approach’ (Milroy 1980Milroy, Lesley 1980 Language and social networks. Oxford: Blackwell.) from the Windhoek campus of the University of Namibia (UNAM). Before the interaction between the participants took place, each participant was interviewed on their linguistic backgrounds, linguistic behaviours, and sociolinguistic perceptions. Based on individual interviews conducted with them, the participants, referred to as Sp-1, Sp-2, Sp-3, and Sp-4, are presented below.

The four participants were invited to sit together in an insulated classroom on campus and video-recorded for roughly one hour. To reduce the Observer’s Paradox, the researcher remained absent throughout the recording. Besides, no guidelines were given to the participants except for ‘talking the way they usually do, in whatever language they usually use’. The recorded materials were transcribed by Oshiwambo speaking students using broad orthographic conventions. The total number of transcribed words is ca. 9,200, out of which 40.2% are Oshiwambo (mostly Oshindonga), 58.9% English, and 0.9% Afrikaans. The transcripts were subsequently shown to 15 self-declared Oshindonga speaking students (to whom I refer as the ‘judges’), all of them born and bred in the northern districts, who were requested via a matched guise experiment to provide linguistic comments on the participants’ use of Oshiwambo dialects, as well as to sociolinguistically stereotype the participants.

The analysis of the data follows three successive stages. Stage 1 involves the encoding of orthographic units as English, Oshiwambo, or Afrikaans. Additionally, the linguistic elements of the Oshiwambo part are categorized in terms of their dialectal origin based on the opinion of the judges and the available lexicographic sources for Oshindonga and Oshikwanyama (Tirronen 1986Tirronen, Toivo E. 1986 Ndonga-English Dictionary. ELCIN: Windhoek.; Turvey 1977Turvey, Basil H. C. 1977 Kwanyama-English Dictionary. Johannesburg: Witwatersrand University Press.). Based on the coded transcriptions, Stage 1 (Section 5) produces a general picture of CS types found in the data, of which a description is provided in terms of language/dialect combinations and grammatical types. Stage 2 (Section 6) focuses on signs of inter-speaker convergence and divergence in the use of these CS patterns that can be sequentially accounted for as a function of accommodation or of conversational mechanisms. Finally, Stage 3 (Section 7) examines what social indexicalities these CS patterns possess, if they possess any. Support for the analysis conducted during that stage is sought in interview data, data on Namibia’s sociolinguistic ecology, as well as in the sociolinguistic perceptions that were elicited from the judges by means of the following set of questions:

• What type of person is Speaker (1 to 4) in your opinion?

• Where do you think she is from?

• What do you think her family or circle of friends are like?

• To what extent does her linguistic behaviour make you think that?

Section 8 reviews the forms and functions of CS observed in the data and places them in the perspective of current sociolinguistic dynamics in Namibian society and of urban Africa in general.

5.Variation in the data: A general perspective

The most general way of describing CS patterns is to quantify language choices. The highest percentages of English lexical elements are found with Sp-4 and Sp-1 (80% and 72.6% out of all words used by each). Sp-2 and Sp-3 follow with 64.6% and 41%, respectively. The highest frequencies of Afrikaans lexical elements are found with Sp-3 (1.4% of all words used), followed by Speaker 4 (1.1%), Sp-2 (0.5%), and Sp-1 (0.1%). Insertions mostly comprise English insertions in Oshiwambo matrix clauses (253/297). The participants differ in terms of how frequently the English words that they use form insertions: The highest proportions of inserted English words are found with Sp-3 (25.5%), followed by Sp-2 (12%), while the lowest proportions are found with Sp-4 (4.9%) and Sp-1 (4.3%). The other patterns of insertional CS involve Oshiwambo insertions in English matrix clauses (36/297), and Afrikaans insertions in Oshiwambo matrix clauses (8/297). Sp-2 and Sp-4 comparatively insert more Oshiwambo words into English matrix clauses (17.1% and 12.5% out of all their insertions) than do Sp-1 and Sp-3 (7.4% and 4.9%, respectively). Most Afrikaans insertions are found with Sp-3. The judges generally agreed that the dominant Oshiwambo variety used throughout the interaction is Oshindonga, except in the case of Sp-3 who is perceived as frequently switching into Oshikwanyama. A total of 170 non-Oshindonga Oshiwambo lexical elements could be identified, mostly as Oshikwanyama (96%), Otshikwambi (3%), and Oshingandjera (1%). While all participants use Oshikwanyama elements, only Sp-2 uses Otshikwambi and Oshingandjera words.

Altogether, most English insertions are formed by nouns (59%), followed by verbs (15%), adjectives (12%), prepositional phrases (6%), numerals (5.5%), and function words (2.5%). The rarity of the last category summarizes the limited potential for congruent lexicalization between Oshiwambo and English, predicted by their typological distance (Muysken 2013Muysken, Pieter 2013 Language Contact Outcomes as the Result of Bilingual Optimization Strategies. Bilingualism: Language and Cognition 16(4): 1–22. ). In turn, typological distance explains why Oshiwambo insertions in English clauses mostly consist of syntactically peripheral elements in the form of mostly adverbials (81%), followed by nominal phrases (12%), and (emphatic) pronouns (7%), which in Muysken’s typology are countable as alternations. English insertions are generally fit in an Oshiwambo grammatical frame. The near-totality of English insertions take on the semantically void Oshiwambo affix -a unless they end on a vowel. The near-totality of inserted English NPs assume the Class 9 prefixes o- and Class 10 prefixes oo-, such as in o-commission-a, oo-connection-a, o-aunty, etc. Although most inserted English adjectives remain unajusted to Oshiwambo by not taking any concordial prefix (a possible sign of congruent lexicalization), the only two attributively used adjectives (out of thirteen) take a concordial prefix (e.g. oshistofele shimwe oshiblacka ‘one black stocking’, Sp-2). Inserted English verbs take on the Oshiwambo TMA affixes, such as completive -ile and passive -wa (1). One isolated sign of congruent lexicalization is found in the tendency among Sp-1, Sp-2, and Sp-3 to affix the English participial suffix -ing with a general incompletive meaning to inserted English verbs (2). Oshiwambo nominal insertions in English matrix clauses appear in subject position and do not take any English affixes (Example (3) where the inserted Oshindonga noun ntu ‘people’ appears with a Class 2 prefix aa-). The few Afrikaans insertions consist mostly of verbs, nouns, and adjectives (e.g. bliksem ‘to beat up’, bra ‘guy’, mal ‘crazy’).

The main non-Oshindonga elements found in the Oshiwambo data are affixes and personal pronouns. The former prominently include the Oshikwanyama Class 2 noun prefix ova- and the Class 2 concord prefix va- (whose Oshindonga equivalents are aa- and ya-, respectively). Non-Oshindonga pronouns in the data include 1st person singular ame (Oshindonga: ngame, ngaye), 2nd person singular ove (Oshindonga: ngoye, ngweye), 1st person pl. fye (Oshindonga: tse), 2nd person plural nye (Oshindonga: ne), 3rd person plural vo (Oshindonga: yo). All judges identified Sp-3 as the participant using Oshikwanyama grammatical and lexical forms most frequently. Out of the 113 Class 2 affixes that she uses, 43.2% are Oshikwanyama and 56.8% Oshindonga. By contrast, the pronominal forms that she uses are mostly Oshikwanyama (in eighteen cases out of twenty-two). Generally, Oshikwanyama and Oshindonga frames are kept distinct (5). Three cases of congruent lexicalization involving mixed Oshikwanyama-Oshindonga morphology are discussed in the next section. Functional elements from other dialects are rare. They include five Otshikwambi subject concords used by Sp-2 and one Oshingandjera first person pronoun used by Sp-1. Non-functional lexical elements of non-Oshindonga origin were less frequently pointed out by the judges. They mostly include Oshikwanyama verbs and nouns (e.g. Oshikwanyama hombola vs. Oshindonga hokana ‘to marry’, or Oshikwanyama polifi vs. Oshindonga polisi ‘police’).

Most switches in the data (72.4%) signal alternations between Oshiwambo and English, which confirms the relevance of typological distance to predicting grammatical CS types (Muysken 2013Muysken, Pieter 2013 Language Contact Outcomes as the Result of Bilingual Optimization Strategies. Bilingualism: Language and Cognition 16(4): 1–22. ). The highest proportions of English matrix clauses are found with Sp-1 and Sp-4 (61.7% and 60% of all verbal clauses, respectively).22.For the purpose of this study, matrix languages are assigned to matrix clauses based on the language of subject-verb agreement marking (see Deuchar et al. 2007Deuchar, Margaret, Pieter C. Muysken, and Sung-Lan Wang 2007 Structured Variation in Codeswitching: Towards an Empirically Based Typology of Bilingual Speech Corpora. The International Journal of Bilingual Education and Bilingualism 10(3): 298–340. , 309–11 for a similar approach). Matrix dialects are assigned based on the language of pronouns and of concordial prefixes. Verb-less utterances are excluded from counts. Sp-2 and Sp-3 follow with 53% and 30.5%, respectively. There are no instances of Afrikaans matrix clauses. Example (4) shows an alternation between an English and an Oshiwambo matrix clause that takes place at the clausal boundary between the two, that is, the slot occupied by the conjunction ndee ‘but’. Example (5) shows one of the three alternations between an Oshikwanyama clause, signalled by an Oshikwanyama pronoun (ove) and its associated concordial prefix ova-, and an Oshindonga matrix clause, signalled by the Class 2 plural prefix aa- of ntu (Oshikwanyama: ova-) and the distinctively Oshindonga determiner ayehe (Oshikwanyama: aveshe).

The alternations found in the data differ in terms of how clear-cut their boundaries are. Unlike (4), (6) forms a case of a blurred boundary between two matrix languages as an English matrix clause is seemingly ‘triggered’ under the effect of a contiguous English insertion in the preceding Oshiwambo matrix clause (see Clyne 2003Clyne, Michael 2003 Dynamics of Language Contact. Cambridge: Cambridge University Press. for ‘triggering’). Triggering across clausal boundaries – a feature of congruent lexicalization (cf. Muysken 2013Muysken, Pieter 2013 Language Contact Outcomes as the Result of Bilingual Optimization Strategies. Bilingualism: Language and Cognition 16(4): 1–22. ) – occurs most often with Sp-2, with 4.9% of her alternations affected, and least often with Sp-3, with 2.1% of her alternations affected.

Alternations involving non-verbal elements are found with all participants. They occur between verbal clauses and response markers (7), conjunctions (8), adpositions (9), and false starts (10).

A salient aspect of variation in CS observed in the data concerns the way that languages are distributed across turns. Table 1 shows that the participants differ in terms of how many of their turns can be categorized as having only English, only Oshiwambo, or both, as matrix languages. Sp-3 shows the highest frequencies of Oshiwambo matrix turns, followed with some distance by Sp-2. Sp-4 shows the highest frequencies of English matrix turns, followed by Sp-1. Finally, the highest proportions of turns featuring Oshiwambo and English as matrix languages are found with Sp-2.

Single matrix language turns are often marked as bilingual by insertions, adverbials, adpositions, and discourse markers, thus fitting Muysken’s definition of ‘backflagging’ (Section 2). An additional backflagging strategy for single matrix language turns consists in affixing the Oshiwambo semantically void -a suffix to the last word of the last English utterance, as illustrated by (11).

Table 2 shows the frequencies per speaker of single matrix language turns marked as bilingual by these means. Each participant tends to mark their Oshiwambo matrix turns as bilingual more often than their English matrix turns. Sp-1 stands out by marking her English matrix turns as Oshiwambo more often than the others, while Sp-2 least often involves Oshiwambo elements in her English matrix turns. Sp-4 tends more often than the others to keep her (sparse) Oshiwambo matrix turns monolingual. The participants differ in terms of how often they involve Afrikaans elements in their English and Oshiwambo turns. The Afrikaans element recurring most often is the conjunction maala (‘but’, from Afrikaans maar), which each participant uses in alternation with Oshiwambo nde(l)e(ne). The next most recurrent Afrikaans forms used by all participants in combination with both Oshiwambo and English matrix clauses are the indicator of shared knowledge mos and the question tag nè. Sp-3 stands out by involving Afrikaans elements more often than the others in her English and Oshiwambo turns, followed by Sp-4, Sp-2, and Sp-1. Finally, Sp-3 alternates between Oshindonga and Oshikwanyama within and across turns: Out of all her Oshiwambo matrix turns, she takes thirteen distinctively Oshikwanyama turns, eight distinctively Oshindonga turns, and six distinctively mixed Oshikwanyama-Oshindonga turns.

To summarize findings at this stage, the participants share CS patterns that generally show an alternational and insertional character with marginal congruent lexicalization. All participants tend to append syntactically peripheral Oshiwambo elements to English clauses. Additionally, they all append syntactically peripheral Afrikaans elements to English and Oshiwambo matrix clauses. This suggests that the participants might share a common stylistic repertoire. Still, there is significant inter-speaker variation in frequencies of English, Afrikaans, and Oshindonga/Oshikwanyama lexical items, in how much CS patterns quantitatively tend towards an insertional pole rather than to an alternational one, as well as in frequencies of Oshiwambo-marked English clauses. The next section assesses how sequentially focused these CS patterns are by establishing whether they perform any function linked to accommodation or conversation management.

6.Sequential variation in CS patterns

From an aggregate perspective, language choices in the exchange point towards mutual accommodation tendencies. The figures in Table 3 show the relative proportions of Oshiwambo versus English words used by each participant while interacting with other specific participants. The highest frequencies of Oshiwambo words are found in interactions with Sp-3, which suggests that the participants acknowledge and accommodate her Oshiwambo-dominant character. The interactions between Sp-1/Sp-2 and Sp-4 suggest some mutual linguistic accommodation. Sp-1 addresses Sp-4 exclusively in English, which seems to be an acknowledgement by Sp-1 of her English-dominant character. Sp-4 conversely uses more Oshiwambo with Sp-2 and Sp-3, which suggests that she is aware of their Oshiwambo-dominant character. Overall, Sp-3 shows the least accommodating patterns in recipient-based language choice as her proportions of Oshiwambo and English words remain stable across recipients.

Table 3.Language choices between participants, % words (O=Oshiwambo; E=English)33.Afrikaans words are subtracted from the count in this table, as well as in table 4, and treated apart further below. This table only takes account of turns whose recipient is signalled by the turn-taker's initial gaze direction (cf. Ho et al. 2015Ho, Simon, Tom Foulsham, and Alan Kingstone 2015 Speaking and Listening with the Eyes: Gaze Signaling during Dyadic Interactions. PLoS One 10(8). ; Rossano 2013Rossano, Frederico 2013 Gaze in Conversation. In Handbook of Conversation Analysis, ed. by Jack Sidnell and Tanya Stivers. 308–329. London: Blackwell.). The speaker averages are based on the whole exchange, and thus include speaker turns with no specific recipient.

In interaction with	Sp-1	Sp-2	Sp-3	Sp-4
Sp-1	Average: O: 27.5%; E: 72.5%	O: 30.2%; E: 69.8%	O: 36.5%; E: 63.5%	O: 0%; E: 100%
Sp-2	O:27.5%; E: 72.5%	Average: O: 35.5%; E: 64.5%	O: 38%; E: 62%	O: 13.5%; E: 86.5%
Sp-3	O: 62.75%; E: 37.25%	O: 67.5%; E: 37.5%	Average: O: 58.85%; E: 41.15%	O: 64%; E: 36%
Sp-4	O: 0%; E: 100%	O: 14.5%; E: 85.5%	O: 16%; E: 84%	Average: O: 27.75%; E: 72.

The overall convergent tendencies described above suggest that linguistic convergence takes place on a turn-by-turn basis. This suggestion is partly confirmed by a sequential perspective. Around 25% of all English turns (110/454) directly follow a preceding English turn. The fact that most English turn pairs (62/110) involve an initial turn taken by Sp-4 suggest that they are largely conditioned by the other participants’ awareness of her limited competence in Oshiwambo, and thus accommodation-driven. Following the logic of triggering discussed in the preceding section, English turns can also follow, or be followed by, part-English turns, as illustrated by Example (12). Out of the 118 turns featuring two matrix languages, 86 display an initial matrix language (in 38 cases English, in 22 cases Oshiwambo) that mirrors the matrix language of the preceding turn, or at least of the last sentential element of the preceding turn. Most of these turns are taken by Sp-2, followed by Sp-3, Sp-1, and Sp-4.

Section (5) pointed out that the participants occasionally use English matrix turns containing syntactically peripheral Oshiwambo elements. Most of these turns (64/80) form responses to (part-)Oshiwambo turns.44.I am including both verbal and non-verbal turns in this count. The context in which these turns are taken is illustrated by Example (13): Sp-4 meets Sp-3’s Oshiwambo request for an explanation (i.e. shaashi?) with an English matrix turn introduced by the same Oshiwambo element used as a conjunction. The seemingly triggered first element of Sp-4’s turn could be seen as a strategy of downgrading the contrastive value of an English matrix turn at this juncture by symbolically stressing its status as a ‘preferred’ second pair part of the question-answer pair initiated by Sp-3. Although they are used by all (see Section 5), most Oshiwambo-marked English matrix turns are taken by Sp-1 (26/64) and Sp-4 (20/64) in response to preceding (part-)Oshiwambo turns taken by Sp-2 and Sp-3. This suggests that English turns featuring Oshiwambo discourse markers form a prominent component of the linguistic strategies used by Sp-1/4 to accommodate the Oshiwambo-dominant character of Sp-2/3.

Unlike English-Oshiwambo CS, CS between Oshindonga and other Oshiwambo dialects seems only marginally predictable at a sequential level, as the Oshikwanyama features used by Sp-3 are generally not immediately replicated by her recipients (one rare exception is shown in Example (15), discussed in the next section). In fact, Sp-1 and Sp-4 use Oshikwanyama elements while responding to Oshindonga or English turns taken by Sp-2. However, it seems as if Sp-3 strategically reduces her use of Oshikwanyama features to converge with the other participants’ Oshindonga-dominant character. The clearest indication in this regard is that her three utterances featuring mixed Oshikwanyama-Oshindonga morphology include what seems to be the strategic insertion of the Oshindonga noun ntu (‘people’; Oshikwanyama: nu) with its Oshindonga plural prefix aa- instead of the Oshikwanyama form ovanu, which she uses elsewhere. This results in grammatically mixed constructions such as aantu vahapu (‘many people’; Oshindonga: aantu oyendji; Oshikwanyama: ovanu vahapu). Her (part-)Oshikwanyama turns are generally isolated. Finally, the Afrikaans elements used by the participants, i.e. mostly maala, mos, and nè (see Section 5), are in no instance used in adjacent turns.

There is some evidence that sequential inter-speaker divergence marks shifts in conversational structure. Some untriggered monolingual English turns seem to serve as first pair parts in sequences signalling transitions between conversational frames. Out of the 49 identified frames, 13 are introduced by an English or Oshiwambo turn, or by a mixed Oshiwambo-English turn with an English first sentential element, that in each case linguistically contrasts with the last turn closing the preceding frame. Example (14) illustrates a transition between two frames. Subsequent to turn (3), where Sp-4 acknowledges the information provided by Sp-2 without any additional request for information, Sp-2 takes the floor again in a humorous tone – enacting a fictional character while asking a formulaic question without next-speaker selection. She then goes on in turn (6) to shift the frame away from the video recording devices in the room towards TV careers. She achieves that by switching from English into Oshiwambo, simultaneously flagging the sequence as an actually intended first pair part.

The other two interactional contexts that to some extent show an association with CS are Turn-Relevant Places (TRP) and turn requests. A relatively small number of intra-turn switches (15/118) perform clearly recognizable functions as TRP markers (Pomerantz 1984Pomerantz, Anita M. 1984 Pursuing a Response. In Structures of Social Action : Studies in Conversation Analysis, ed. by J. Maxwell Atkinson and John Heritage, 152–163. Cambridge: Cambridge University Press.). Example (12) above illustrates this function: Sp-3’s turn comprises an acknowledgement of the preceding turn taken by Sp-2, emphasized by the fact that it linguistically mirrors Sp-2’s English turn. This acknowledgement is followed by an Oshiwambo question that announces a TRP by selecting Sp-2 as next-speaker. In a general interactional context dominated by Sp-2 and Sp-3 (41.3% and 31.8% of all words uttered, respectively), Sp-1 displays a strong tendency to linguistically mark her turn requests, generally by using English (see Wiemann and Kapp 1975Wiemann, John M., and Mark L. Kapp 1975 Turn-taking in Conversations. Journal of Communication 25(2): 75–92. for turn request marking strategies; see further Section 7).

To sum up, much of the turn-by-turn convergence and divergence between participants is accommodation-driven and/or channelled through interactional mechanisms. Alignment and conversational cohesion are signaled by linguistic continuity across turn transitions, which is ensured by convergent uses of matrix frames and by Oshiwambo-marked English matrix utterances following Oshiwambo utterances. Switches in between turns tend to manifest conversational shifts. In sum, a sequential analysis shows that the alternation between monolingual matrix frames and backflagging possess salience, which lends them conversational functionality, and makes them potentially usable as stylistic devices. However, a sequential analysis does not explain why CS often takes place without discernible conversational shifts, nor why dialect-mixing and Afrikaans are used in the exchange. The following section attempts to explain that while assessing by holistic sociolinguistic means the stylistic functionality of all CS patterns discussed so far.

7.CS patterns as styles

The deployment of styles has been tentatively modelled as a function of stancetaking (Section 2). In his study of a Pittsburghese multiparty conversation, Kiesling (2009)Kiesling, Scott F. 2009 Style as Stance: Stance as the Explanation for Patterns of Sociolinguistic Variation. In Stance. Sociolinguistic Perspectives, ed. by Alexandra M. Jaffe, 171–194. Oxford: Oxford University Press. contextualized stancetaking within a range of speech activities that he grouped into three categories, namely ‘social’ (which comprises ‘gossip’), ‘informational’ (which comprises ‘expert information/teaching’), and ‘discourse management’ (which comprises ‘facilitating’), and found that the use of Pittsburghese dialect features correlated more with the first category than with the others, with which standard features showed stronger correlations. Table 4 shows the proportions of English and Oshiwambo words used by this study’s participant across each of the three speech activities named by Kiesling. The figures show that all speakers display a slight preference for Oshiwambo for social and discourse management activities, while English is conversely associated with informational speech activities. The resulting suggestion that English may index objective stances conforms to the notion that English – being the overarching prestige language in the Namibian context – is naturally associated with formal contexts (Section 3).

The high status of English is particularly perceptible in the strategic use that Sp-1 makes of it, of which a brief description was given in Section (6). To claim the floor in an interactional context dominated by Sp-2/3, Sp-1 puts forward turn requests in English, irrespective of the language in which the preceding turn was taken. In doing so, she can presumably invoke the authority associated with English in the Namibian context, and thus maximize her chances for her turn requests to be noticed and granted by the other participants (see further Myers-Scotton 1993Myers-Scotton, Carol 1993 Social Motivations for Codeswitching. Evidence from Africa. Oxford: Clarendon. and Section 6). Such an interpretation conforms with the idea developed by Li Wei (1998)Li Wei 1998 The ‘why‘ and ‘how‘ Question in the Analysis of Conversational Code-switching. In Peter Auer (ed.), Code-switching in Conversation. Language, Interaction and Identity, 156–179. London: Routledge. that some social meanings are ‘brought along’ into conversations and manifest themselves in conjunction with interactional mechanisms. The consistent use that Sp-4 makes of English cannot be interpreted along the same lines. Since she had indicated prior to the interaction that her competence in Oshiwambo is low, her use of English can be assumed to be cognitively motivated to a large extent, which in turn explains why Sp-2/3 tend to respond to her in English (Section 6). However, an accommodation-based interpretation of the interaction between Sp-4 and Sp-2/3 fails to explain why the English turns that Sp-4 addresses to Sp-2/3 do not systematically trigger English responses from them. The motivation for this apparent non-accommodating behaviour can perhaps be found in a comment that Sp-3 made during the preliminary interviews. Asked about sociolinguistic distinctions that she sees among Oshiwambo speakers, Sp-3 expresses her disapproval of Ombwiti (Section 3), as ‘they [the Ombwiti] try to be something that they are not’, and ‘sometimes pretend they don’t know Oshiwambo’. Through her comments, Sp-3 ascribes pretension and illegitimacy to Ombwiti, against whom she might be positioning herself as a an ethnically loyal Ovambo. Her frequent avoidance of English in her responses to Sp-4 could therefore be interpreted as resistance to the ‘sell-out’ indexicality that she associates with monolingual English usage in the context of interactions between Ovambos.

Apart from interactional shifts (see Section 6), the alternating use of Oshiwambo and English in the exchange may reflect an on-going negotiation of ‘authenticity’ among the participants (Lacoste et al. 2014Lacoste, Véronique, Jakob Leimgruber, and Thiemo Breyer 2014 Authenticity: A View from Inside and Outside Sociolinguistics. In Indexing Authenticity: Sociolinguistic Perspectives, ed. by Véronique Lacoste, Jakob Leimgruber and Thiemo Breyer, 1–13: Berlin: De Gruyter.). Being Ombwiti or ethnically loyal involves stereotypes that may be situationally claimed, avoided, or combined, for the sake of projecting social personae unlikely to attract negative ascriptions. There are strong indications that Oshiwambo-marked English utterances (Section 6) serve the purpose of claiming ‘ethnic’ authenticity against a monolingual English background. Ethnic authenticity is a value that all participants seem keen to project in combination with a more urban or worldly authenticity, which could explain why too much monolingual Oshiwambo usage is avoided by regularly combining Oshiwambo with English. As an indicator, Table 5 shows the average number of consecutive English and Oshiwambo turns per participant.

Combined with Oshiwambo and English, Afrikaans discourse markers may also project urban or worldly authenticity, since Afrikaans is a language historically associated with inter-ethnic communication in southern regions and urban Namibia in general (Section 3). This analysis is confirmed by the fact that all participants acknowledged in the interviews that Afrikaans is useful in the context of Windhoek. The fact that Sp-3/4 use Afrikaans discourse markers most often reflects their biographies of exposure to multi-ethnic environments outside of the Ovambo heartland (Section 4). An additional layer of indexicality could involve the use of Oshiwambo dialectal elements, whose combination implies a biography of exposure to various Ovambo groups in urban areas, and is thus likely to convey mutually overlapping ethnic and urban authenticities (see further Section 3). Example (15) shows how the participants tend to compete for this category of authenticity. In response to an utterance by Sp-3 that contains an Oshikwanyama pronoun (nye ‘you’, Oshindonga ne), Sp-2 produces an Oshikwanyama and an Otshikwambi (underlined) clause embedded in an English-dominated turn. Her Oshikwanyama clause could form an instance of symbolic or partial accommodation. Haugh’s observation that Oshiwambo dialect-mixing is associated with ‘worldliness‘ (Section 3) suggests that, at the same time, Sp-2 may be signalling to Sp-3 that she is ‘worldly’ enough to match her use of Oshikwanyama, while her subsequent use of an Otshikwambi clause may be strategically projecting an alternative ‘Ovambo worldliness’ that lies beyond the reach of Sp-3, who uses Otshikwambi at no point in the exchange.

The matched guise experiment yielded insights into social indexicalities potentially attached to CS patterns while confirming some of the analyses presented above. Most judges55.The judges are referred to by means of the abbreviated form J(1–15). agree that Sp-4’s competence in Oshiwambo is limited (she makes ‘mistakes’, J-2; ‘doesn’t know the words’, J-13). Most judges ascribe an Oshindonga background to all four participants. Most differing interpretations are found in conjunction with Sp-3: She might be a ‘Ndonga with Kwanyama friends’ (J-6), a ‘Ndonga Kwanyama’ (J-7), or ‘a Kwanyama’ (J-3, J-4). Additionally, Sp-2 is ascribed a Kwambi background, and a mixed Mbalantu and Ngandjera background by two judges. It is clear to all that having grown up in town is the only explanation for Sp-4 not mastering Oshiwambo (‘She talks like a typical town girl’, J-6; ‘you can only understand her when she speaks English’, J-1). Also, four of the judges ascribe an urban background to Sp-1/2 on account of their extensive use of English (even though Sp-1/2 are in fact northerners, see Section 4). Other judges dispute the urban authenticity of Sp-1/2. For example, J-7 believes that the way that Sp-2 uses English is ‘fake’, and forms an attempt ‘to hide her rural background’. More scathing are the comments that Sp-2 is ‘just a Ndonga who wants to be seen as a Mbwiti’ (J-8), a ‘pretender from the village’ (J-9). There is more agreement as to the urban credentials of Sp-3, which in the view of J-3 are clearly manifested by her use of Afrikaans alongside English. While most judges interpret the participants’ dialect-mixing patterns as a function of their social circles, the authenticity of Sp-3’s dialect-mixing behaviour is disputed by one judge (she speaks ‘fake Oshikwanyama’, J-8). By contrast, Sp-1 is noted for speaking ‘good’ or ‘deep’ Oshindonga (J-2, J-7), as a result of which ‘you can tell her tribe easily’ (J-4). As Sp-3 did, some judges (born and bred in the northern districts, see Section 4) describe Ombwiti stereotypes in vivid terms: ‘Money-driven’ (J-6), ‘opportunistic’ (J-13). Such negative ascriptions show the delicate balancing acts that the participants may have felt under pressure to perform between claims of urban exposure and ethnic loyalty, via quantitatively contrasting yet indexically similar patterns of CS.

8.Discussion

The multiparty interaction on which this study is based revealed a range of structurally recognizable CS patterns found among all participants. The most widespread pattern involves alternational Oshiwambo-English CS, and insertional Oshiwambo-English CS, in both cases with marginal traces of congruent lexicalization. Other CS patterns involve backflagging in the form of Oshiwambo-marked English utterances, CS between Oshiwambo or English utterances and Afrikaans discourse markers, and mixing between Oshiwambo dialects. The distribution of these patterns display significant inter-speaker variation, but the fact that they are replicated across speakers offered a first suggestion that they are focused and thus possibly constitutive of styles. The sequential analysis only partly confirmed this suggestion: Only the alternation between monolingual Oshiwambo and English frames, and to a lesser extent Oshiwambo-marked English utterances, (occasionally) seemed conversationally functional, while the other patterns displayed a conversationally unconstrained distribution. The analysis of social indexicality conducted in Section 7 could shed light on whether the sequentially unaccounted-for variation is indexically focused or random. Notably, it could go some way towards explaining sequentially unaccounted-for alternation between Oshiwambo and English as the manifestation of a balancing act between opposed ethnic and urban indexicalities. It could also account for the use of Afrikaans elements and dialect-mixing for strengthening the overall urban authenticity associated with Oshiwambo-English CS. To sum up, the CS patterns observed among the participants manifest sequentially and/or indexically focused practices. Most focused is CS between Oshiwambo and English frames, and to a lesser degree CS between English utterances and Oshiwambo discourse markers. The occurrence of Afrikaans discourse markers and dialect-mixing may seem sequentially unfocused, but their use turned out to be strategic and socially indexical. All observed CS patterns thus seem to form part of a stylistic repertoire shared by the participants.

The linguistic practices discussed in this study reflect aspirations for social mobility, consonant with urbanization and multilectal exposure. They also reflect the operation of peer pressure for maintaining some degree of ethnic authenticity, whose absence invites social sanctions, one of which – the matched guise experiment revealed – is being dismissed as ‘fake’ or as a ‘pretender’, that is, as a ‘wannabe Ombwiti’. The value generally attached to ethnic authenticity and the disapproval of ‘too much urbanity’ among Ovambos is perhaps best summarized in the saying recorded by Pendleton et al. (2016Pendleton, Wade, and Ndeyapo Nickanor 2016 The State of Food Insecurity in Windhoek, Namibia. AFSUN: Cape Town., 7) that owu na okukala wu na omutala kegumbo (‘you should have a room for sleeping at home in rural areas’). The strategic balancing act between urban and ethnic indexicalities in the participants’ linguistic behaviours generally reflect practices that Myers-Scotton (1993)Myers-Scotton, Carol 1993 Social Motivations for Codeswitching. Evidence from Africa. Oxford: Clarendon. observed in urban Zimbabwe and Kenya where the young generations ‘are not satisfied with either the identity associated with speaking Shona or Swahili alone’ and ‘see the rewards in indexing both identities for themselves’ (see further Deumert 2013Deumert, Ana 2013 Xhosa in Town (Revisited) – Space, Place and Language. International Journal of the Sociology of Language 222: 51–75.). The regularity of that balancing act, that this study illustrated in the form of what seems to be a consensual cap on consecutive single matrix turns in either Oshiwambo or English, provides additional evidence that alternational CS unto itself may possess stylistic properties (see further Auer 1998Auer, Peter 1998 Introduction: Bilingual Conversation Revisited. In Code-switching in Conversation. Language, Interaction and Identity, ed. by Peter Auer, 1–24. London: Routledge.; Alvarez-Cacamo 1998Alvarez-Cacamo, Celsio 1998 From ‘Switching Code’ to ‘Code-Switching’: Towards a Reconceptualisation of Communicative Codes. In Code-Switching in Conversation, ed. by Peter Auer, 29–50. London: Routledge.). The lexically regular styles observed in the form of Oshiwambo-marked English utterances and CS between Oshiwambo/English utterances and Afrikaans discourse markers reminds of the properties of the Tsotsitaal ‘register’, widespread in urban South Africa, which consists of regular sets of core lexical items that can combine with multiple languages (Mesthrie and Hurst 2013Mesthrie, Rajend, and Ellen Hurst 2013 Slang Registers, Code-switching, and Restructured Urban Varieties in South Africa: An Analytic Overview of tsotsitaals with Special Reference to the Cape Town Variety. Journal of Pidgin and Creole Linguistics 28(1): 103–30. ). It can be suspected that Oshiwambo and Afrikaans discourse markers will play an increasingly emblematic role in a general Namibian context of rural-to-urban migration that is causing monolingual English to become more widespread.

9.Conclusion

This study examined CS patterns in a multiparty interaction between young Ovambo residents of Windhoek with different biographies of urban exposure. It found that – despite significant inter-speaker variation – they share qualitatively similar CS patterns. These CS patterns were studied from a joint sequential and indexical perspective to establish whether their use is focused. The main finding is that English-Oshiwambo alternational CS and Oshiwambo-marked English utterances are relatively regular in their sequential occurrence, which is reflected in the combined urban and ethnic indexicalities that they carry in a context of pressure for balancing Oshiwambo and English. The distribution of specific features, namely, dialect-mixing and Afrikaans discourse markers, proved to be sequentially random, but turned out to also possess social indexicality. It can be suspected – based on the limited data set presented in this study – that English with Oshiwambo and/or Afrikaans discourse markers might be set to increasingly assume visibility in a context of pressure for jointly indexing ethnic and urban authenticity against a backdrop of incipient language shift. It is hoped that subsequent studies-based on a socio-demographically more representative samples – will establish how generalized this trend is beyond the tertiary educational settings within which the present study is set.