THE HANDBOOK OF SOCIOLINGUISTIC

UNIT 11. CODE-SWITCHING

1. Introduction

In bilingual communities, code-switching (CS) occurs when speakers mix elements from two or more languages within a
conversation. Research since the late 1970s has provided a detailed understanding of CS and its distinction from other
bilingual language contact phenomena.

This overview addresses misconceptions, such as the belief that CS results from limited proficiency or random language
choice. Most researchers agree that language choice in CS is not entirely free, though they differ on how constraints apply.
CS also shares grammatical principles with phenomena like first language attrition. Additionally, its similarities and
differences with borrowing are key to understanding these processes.

2. Goals

The primary goal of this overview is to provide a deeper understanding of code-switching (CS) by examining its
grammatical structure and its role as a sociolinguistic and psycholinguistic phenomenon. This includes situating CS within
broader linguistic behaviors, such as the interaction between languages or the influence of one language on another.
Another aim is to survey contemporary research on CS, focusing specifically on switching between languages rather than
between dialects or styles.

Early research explored the social functions of CS, often viewing it as a strategy to manage interpersonal relationships. For
instance, Gumperz (1982) and Auer (1984) described CS as a “contextualization cue,” signaling speaker intentions through
discourse. Researchers like Myers-Scotton emphasized CS as a negotiation tool for interpersonal relationships, where
speakers weigh social costs and rewards. In contrast, Heller and others focused on CS as a reflection of ethnic group
competition within broader political contexts.

While most studies examine CS at the interpersonal level, few have analyzed it from a macro perspective, such as its link to
group identity. This gap is partly due to challenges in quantifying CS or relying on self-reports. Poplack’s work on
French-Canadian communities, however, highlighted that community norms, rather than individual proficiency or
demographic factors, are the strongest predictors of CS usage. This suggests that social influences often outweigh
individual abilities in determining CS behavior.

In the 1980s, research on code-switching (CS) shifted toward understanding the morphosyntactic constraints of
intrasentential switching, focusing on where in a sentence language changes occur. This structural focus became a central
theme, with studies like those by Poplack and Joshi providing foundational insights. By the 1990s, structural constraints
were established as a dominant area of study, though research on the phonology of CS, such as Grosjean and Miller’s
findings on phonetic changes during switches, remained limited.They concluded: “When bilingual speakers insert a word or
phrase from the guest language into the base language, the switch usually involves a total change, not only at the lexical but
also at the phonetic level”

Morphosyntactic CS research branched into two main approaches. One used CS as a lens to study lexical entries and
language production processes, while the other tested syntactic theories using CS data. Despite this focus, social and
discourse motivations for CS continued to gain attention. Researchers began linking structural CS characteristics to social
and psycholinguistic traits of community groups, exploring how switching reflects intergroup dynamics, bilingual
proficiency, or attitudes toward group affiliations. Examples include studies on interethnic tensions in Brussels and
generational differences among Cantonese-speaking immigrants in Britain.

Some scholars also examined CS as a discourse-organizing tool, emphasizing its role in structuring conversational turns and
signaling social meanings. Auer’s concept of “juxtaposition” highlights the importance of sequential context in interpreting
code-switching. New directions in CS research include viewing it within broader bilingual speech production and applying
insights from CS constraints to second-language acquisition models. This approach integrates CS into larger discussions of
bilingualism and language learning.
3. The structural description of Code-switching

There is no universal agreement among researchers on what constitutes code-switching (CS). Some classify certain
linguistic phenomena, like Hindi-Marathi convergence (Pandharipande, 1992) or heavily influenced speech from a first
language undergoing attrition (de Bot and Schreuder, 1993), as CS, while others might label them differently. This lack of
consensus underscores the need for clear definitions and boundaries.

To analyze CS systematically, the Matrix Language Frame (MLF) model (Myers-Scotton, 1993a, 1995a) provides a widely
used framework. In this model, the Matrix Language (ML) sets the grammatical structure of mixed sentences, including
morpheme order and system morphemes (e.g., inflections, function words). The Embedded Language (EL) contributes
additional elements, typically content morphemes such as nouns, adjectives, or verbs.

Within the MLF framework, a clause or complementizer phrase (CP) can include three types of constituents:

1. Mixed constituents: These combine ML system morphemes with EL content morphemes while following the
ML grammatical frame. For example, in the Swahili-English sentence ni-ko SURE (“I am sure”), Swahili provides
the system morphemes ni- and -ko (ML), while English supplies the content morpheme sure (EL).
2. ML islands: These are fully composed of ML morphemes and adhere strictly to the ML grammar.
3. EL islands: These consist solely of EL morphemes structured by the EL grammar and are often adjuncts rather
than core arguments.

An example from Swahili-English CS illustrates these constituents. The phrase after two days serves as an EL island
(English adjunct), while u-ta-i-pata kwa wingi (“you will find it in abundance”) is an ML island entirely framed by Swahili.
Mixed constituents, like ni-me-DECIDE (“I have decided”), blend elements from both languages but follow the ML
grammatical rules. Cultural borrowings, like Uchumi supermarket, represent integrated terms rather than structural
switches.

A key idea in the Matrix Language Frame (MLF) model and similar frameworks is the unequal participation of languages
in code-switching (CS). While both languages are active during CS, the Matrix Language (ML) governs the grammatical
structure, creating a parsimonious explanation for the organization of mixed constituents. This distinction, where the ML
serves as the grammatical frame and the Embedded Language (EL) contributes additional elements, has been
acknowledged by researchers like Joshi (1985) and earlier scholars who referred to the ML as the "host" and the EL as the
"guest" language.

Earlier attempts to explain CS constraints focused on surface-based rules, such as linear order (e.g., the Equivalence
Constraint by Poplack) or specific grammatical categories like switches between pronouns and verbs (Timm, 1975).
However, these approaches faced numerous counterexamples, prompting shifts toward more abstract models. For instance,
Government and Binding (GB) theory and its variations (e.g., Woolford, 1983) offered alternatives but also struggled with
overgeneralization, allowing structures not observed in real CS data.

The MLF model avoids these pitfalls by prohibiting certain combinations, such as syntactically active system morphemes
from the EL, which other models might allow. This restriction ensures predictions about grammatical structure are more
robust and generalizable across different contexts.

In the study of code-switching (CS), understanding the distinction between intrasentential and intersentential CS is
essential. Intrasentential CS occurs within a single complementizer phrase (CP) that contains morphemes from two
languages, forming mixed constituents. These constituents combine elements from the Matrix Language (ML), which
provides the grammatical framework, and the Embedded Language (EL), which contributes specific lexical items. In
contrast, intersentential CS involves switching between CPs that are entirely in one language, where the ML-EL distinction
is irrelevant as each CP adheres fully to the grammar of one language.

The ML is defined as the language setting the grammatical frame of the CP, determined by factors such as the dominance of
its morphemes in the discourse and speakers’ perception of it as the primary language of interaction. This construct
explains morphosyntactic structure within intrasentential CS, distinguishing it from discourse-level constructs like the
unmarked choice, which reflects social expectations in language use.

Importantly, the ML is dynamic and can shift within discourse due to changes in topic, participant alignment, or
sociolinguistic context. Such shifts often occur among bilingual speakers navigating multiple identities, as seen in studies of
Turkish-Dutch bilingual youth. These shifts highlight the ML’s basis in sociopsycholinguistic factors, including language
attitudes and proficiency.

Intersentential CS, on the other hand, involves distinct CPs in different languages, such as in Pandharipande’s example of
Marathi-English switching: “to ghari ālā ani ø enjoyed the ice cream” (“he came home and enjoyed the ice cream”). Here,
the Marathi CP and the English CP remain structurally separate, with no mixing of morphemes. In sum, the ML-EL
framework provides valuable insights into intrasentential CS, while intersentential CS represents a simpler alternation
between monolingual structures. Together, these phenomena demonstrate how structural and sociolinguistic factors shape
bilingual discourse.

4. Code-switching compared with other bilingual speech

Code-switching (CS) is a hallmark of proficient bilingual speech, characterized by speakers' ability to produce
grammatically well-formed constituents in both languages. This proficiency does not necessarily adhere to standard dialect
norms but reflects the speakers' functional dialects. While speakers often show greater fluency in one language, their ability
to consistently project grammatical frames in either language underpins the structural coherence of CS.

CS occupies one pole of a bilingual speech continuum, defined by predictable grammatical structures and deliberate
speaker motivations. Its structural integrity stems from the consistent application of the Matrix Language (ML) framework,
wherein the ML provides the grammatical frame while content morphemes may come from either the ML or the Embedded
Language (EL). For instance, mixed constituents reliably adhere to the ML's morpheme order and system morphemes,
while EL elements contribute specific content.

Speaker motivations for CS include enhancing the socio-pragmatic force of their discourse persona, marking discourse
boundaries (e.g., topic changes or emphasis), and addressing lexical or semantic gaps in the ML through EL borrowings.
These motivations demonstrate how CS serves communicative intent, often operating below conscious awareness. A
notable example is a Senegalese politician who seamlessly integrates Wolof and French in speeches while denying he mixes
languages—a testament to CS as a deeply ingrained linguistic strategy.

CS contrasts with language attrition, another phenomenon in the bilingual continuum, which reflects declining proficiency
in one language. Initially, bilingual speech affected by attrition may resemble CS, as the waning language sets the
grammatical frame while introducing content morphemes from the dominant, waxing language. Such speech may include
predictable structures like those seen in CS. However, over time, attrition results in a shift of the ML to the waxing
language, potentially producing hybrid grammatical frames—a divergence from CS's deliberate and structured nature.
Moreover, attrition-driven bilingual speech often arises out of necessity, unlike the intentional deployment of CS.

In summary, CS is distinguished by its structural predictability and intentionality, serving communicative and
socio-pragmatic goals. While it shares initial structural similarities with language attrition, the latter diverges over time as
linguistic proficiency in one language declines, leading to changes in grammatical framing and motivations.

First-language attrition, common among adult immigrants, often arises in contexts where a new language is
sociolinguistically dominant. Immigrants who become bilingual in the dominant language may shift entirely to it by the
second generation. Alternatively, they may rely on code-switching (CS) out of necessity, particularly when their first
language (L1) declines to the point that it no longer supports fully grammatical speech. In such cases, CS often includes
numerous embedded language (EL) islands from the second language (L2), reflecting the dominance of the L2 in their
environment. For instance, German immigrants in Australia frequently incorporated locative prepositional phrases in
English, as observed by Clyne (1987).
Over time, CS patterns in speakers experiencing attrition or acquisition often
shift the Matrix Language (ML) from L1 to L2. Studies of children provide clear
evidence of this phenomenon. A Hebrew-speaking child, for example,
transitioned to using English as her ML after moving to the United States, even
though Hebrew verbs were initially part of her speech. Similarly, Turkish girls
who returned to Turkey from Germany initially used German as the ML in their
CS but shifted to Turkish as the ML over the course of the study.

Second-language acquisition (SLA) also fits within the continuum of bilingual speech and overlaps with CS. While SLA
speech (or interlanguage) often lacks fully bilingual utterances, learners frequently blend grammatical patterns from their
L1 and L2. If CS is broadly defined to include grammatical structures as well as morphemes from multiple languages, then
SLA and even pidgin or creole formation can be seen as akin to CS. In these cases, however, the ML is not derived from a
single language; both the L1 and L2 contribute content morphemes and EL islands.

In early SLA stages, bilingual speech tends to be unpredictable and primarily consists of content morphemes. For instance,
an early German learner from Spain produced an utterance that included a Spanish EL island ("mucho trabajo") alongside
emerging German grammatical structures. This reflects the composite nature of early SLA speech, where both languages
actively shape utterances but without the stable grammatical coherence seen in typical CS.

Thus, CS, attrition, and SLA demonstrate a continuum of bilingual speech phenomena, with varying degrees of structural
predictability and dominance between the participating languages. Each highlights the dynamic interplay of linguistic
systems as speakers navigate and adapt to bilingual contexts.

In second language acquisition (SLA), system morphemes from the first language (L1) rarely appear in learners’ speech,
except within embedded language (EL) islands. For instance, advanced Dutch learners of English incorporated Dutch
content morphemes but avoided Dutch inflections, though some Dutch determiners, which are system morphemes,
appeared. This aligns with the Matrix Language Frame (MLF) model for code-switching (CS), where the L1 resembles the
EL and rarely contributes system morphemes to mixed constituents.

Pidgin and creole formation also intersect with CS and SLA. In early stages, pidgin utterances resemble those of SLA,
relying predominantly on content morphemes as speakers struggle to establish grammatical frames in a language they are
acquiring. Over time, SLA learners surpass this stage as they gain control over the target language's grammatical structure.
In contrast, pidgin/creole development involves greater unpredictability, as speakers cannot establish a consistent Matrix
Language (ML), differentiating it from CS.

CS is distinct from lexical borrowing, although the two share motivations. Borrowing involves incorporating elements from
a donor language (EL) into a recipient language (ML) to meet expressive needs. Unlike CS, borrowing does not require
bilingualism. Borrowed lexemes are permanently integrated into the recipient language's mental lexicon and can appear in
monolingual speech. Conversely, CS lexemes retain their EL identity and rely on bilingualism for use. Borrowing also
reflects sociopolitical dominance, typically incorporating elements from a more prestigious language into a less dominant
one, whereas CS does not always follow this pattern. Borrowing occurs in two forms:

1. Core Borrowings: These represent concepts already covered in the recipient language and often
transition from CS forms. As their frequency increases in bilingual contexts, they are integrated into the
recipient language's lexicon over time.
2. Cultural Borrowings: These fill lexical gaps, entering abruptly to address immediate needs, such as
terms for novel concepts or technologies (e.g., "telephone").
Despite their differences, borrowed and CS lexemes undergo similar morphosyntactic integration when framed by the ML.
They adopt the ML’s system morphemes, inflections, and syntactic patterns. For example, in Wolof/French bilingual
speech, both borrowed and CS lexemes were embedded seamlessly into Wolof grammatical structures.

Ultimately, while distinctions between CS and borrowing are relevant for understanding processes like sociopolitical
influence and bilingualism, the structural treatment of borrowed and CS lexemes is largely identical. Thus, when analyzing
CS constraints, it is unnecessary to exclude singly occurring EL lexemes based on their potential overlap with borrowed
lexemes. Both contribute to the mixed constituents that models like the MLF framework aim to explain.

5. CS and other contact phenomena

Convergence occurs when the grammatical structure of one language is influenced by another, leading to a "rearrangement"
of grammatical frames while retaining all morphemes from a single language. Unlike attrition, where structural
simplification dominates as a result of a diminishing ability to use grammatical frames, convergence involves active
modification without necessarily indicating language decline. Attrition often includes code-switching (CS) to a dominant
language, whereas CS is not a structural necessity for convergence.

Convergence is frequently observed in communities with strong sentiments for language preservation or where speakers
of the language are numerous despite dominance by another language. For example:

● German-speaking communities in the U.S.: Influenced by English, they show changes such as the use of
"du" as a pro-form and altered past participle placement.
● Puerto Rican Spanish in New York: Displays English influence, such as time phrase placement unmarked
for Spanish and the use of calques like cuatro años atrás (“four years ago”) instead of the Spanish hace
cuatro años.For instance: “Mucho cubanoh, cuatro años atrás se mudaron pa'ca “Muchos cubanos, hace
cuatro años, se mudaron para aquí” (Torres, 1989: 428-9)

Further examples include German immigrants in Australia, where English influences grammatical rules, such as
incorrect verb placement and overgeneralization of German conjugation. The word schert in their speech resembles
English sheared, suggesting lexical interference.

Grammatical Modifications and Mixed Languages:

Convergence involves partial changes to grammatical frames rather than complete turnover. Early changes often
include morpheme order. However, in rare cases where one grammatical system replaces another while retaining
many content morphemes from the original language, a "mixed language" emerges. Mixed languages typically arise
during incomplete language shifts and demonstrate the blending of lexical and structural elements from distinct
sources. Examples include:

● Ma’a (Bantu grammar, Cushitic lexicon): A mixed language discussed by Thomason and Kaufman.
● Michif: French nouns are combined with Cree verbal morphology.

A possible pathway for mixed languages involves communities where CS dominates. Due to sociopolitical shifts,
the Matrix Language (ML) of CS is replaced with the dominant sociolinguistic language, reanalyzed as the
community language. The lexicon retains features of the original Embedded Language (EL), but the grammatical
structure aligns with the new ML.

6. The sociolinguistics of CS

6.1. CS patterning

Code-switching (CS) is a structurally unified phenomenon, but its specific patterns vary across communities. These
variations, referred to as "preferences," are influenced by sociolinguistic and psycholinguistic factors unique to each
community. In most cases of CS, one language tends to dominate in either intersentential or intrasentential CS. This
dominance is evident in three primary ways:

1. The Unmarked Choice in Interaction:

According to the Markedness Model (Myers-Scotton, 1983, 1993), there is typically an "unmarked choice" in
linguistic interactions—an option more commonly used and aligned with social norms. In CS, the unmarked choice
often reflects in-group solidarity rather than sociopolitical prestige.
○ Examples: Alsatian, rather than French, dominates in Alsatian-speaking families during CS. In contrast,
in Arabic-speaking communities, CS occurs in formal interactions, such as interviews, where speakers
combine “oral educated Arabic” with colloquial Arabic to express both education and authenticity.

2. Direction of Switching: In many communities, CS often occurs more frequently in one direction than the other,
reflecting the unmarked choice. For instance, switches in Gaelic-speaking Ireland are predominantly from Irish to
English, not vice versa. Similarly, in Bukavu (Swahili/French) and among Moroccan immigrants in the Netherlands
(Arabic/Dutch), switches generally occur toward the more dominant sociolinguistic language.

3. Setting the Discourse Frame:

The unmarked choice also establishes the broader discourse frame, influencing narrative structures or argument
organization. Although this phenomenon has not been widely studied, the dominant language in CS often shapes
such framing.

In some communities, particularly multilingual urban areas in Africa, India, or immigrant families in Europe and
North America, CS itself becomes the unmarked mode of informal in-group communication. Despite this, one
language still typically exerts dominance by setting the structural and organizational aspects of the CS discourse.

CS patterns reflect both community-specific social dynamics and broader linguistic principles. While CS may vary in
directionality and function, the unmarked choice consistently plays a central role in determining which language
dominates the interaction.

6.2. CS and community norms

Patterns of code-switching (CS) often reveal how speakers perceive themselves within their sociopolitical and cultural
contexts. When CS becomes the primary in-group communication medium, it signifies that speakers value both languages
as integral to their identities. This dynamic is common in postcolonial contexts where indigenous languages are used
alongside colonial languages, as seen in Baba Malay/English in Singapore or SeSotho/English in South Africa.

The prevalence of CS also depends on the nature of relationships between linguistic groups. In communities where
languages have positive associations with their respective groups, CS reflects shared identity and reduces polarization.
Conversely, in contexts of intergroup tension, such as Brussels today, CS tends to occur less frequently.

In communities with strong prescriptive norms emphasizing linguistic purity, CS is minimized. For instance,
French-speaking immigrants in Basel, Switzerland, and their children tend to avoid CS, reflecting cultural resistance to
language mixing. In contrast, second-generation Italians in Basel or Zurich exhibit high rates of CS. Among Francophone
Swiss, any CS that occurs typically happens between conversational turns rather than within single phrases or clauses,
showing a preference for linguistic separation.

CS patterns are deeply influenced by cultural attitudes, intergroup dynamics, and societal norms. Whether it is
embraced as a marker of shared identity or minimized due to cultural or social pressures, CS reflects the broader values and
relationships within a community.

6.3. CS and language proficiency

The study of Moroccan Arabic/French bilinguals by Bentahila and Davies (1992) highlights how age and language
proficiency influence code-switching (CS) patterns. Older bilinguals, who were educated during or shortly after the French
protectorate period, have high proficiency in French and frequently use full French clauses in their CS. When switching
within sentences, these speakers often embed French noun phrases (NPs) within an Arabic grammatical frame, as shown in
examples like hda LE DIX-SEPTIEME ETAGE f dak LE FEU ROUGE (“near the seventeenth story at the red light”).

In contrast, younger speakers, educated under Arabization policies and thus Arabic-dominant, incorporate fewer full
French phrases. Instead, their CS involves integrating French content morphemes, such as verb stems, into Arabic
grammatical structures, including using Arabic suffixes.

A similar proficiency-based distinction is observed in Turkish/Dutch bilinguals in the Netherlands (Backus, 1994). In
the 1980s, Turkish-dominant speakers used more intrasentential CS, mixing Dutch noun phrases into Turkish structures and
employing hybrid constructions like the “do” form (afstuder-en yap-tl—“do graduate”). By the 1990s, newer bilinguals,
who were proficient in both languages due to their education in the Netherlands, favored intersentential CS, producing
complete clauses in either Turkish or Dutch. This shift reflects the influence of balanced bilingualism on CS patterns.

Language proficiency shapes CS practices significantly. Bilinguals dominant in one language rely more on integrating
morphemes or fragments from the second language, while balanced bilinguals produce more full clauses in both languages.
This trend underscores the role of education and linguistic environment in determining CS styles.

Why should highly proficient bilinguals bother to engage in CS at all, which in both the Morocco and Tilburg case
goes well beyond filling lexical gaps? The answer is that speakers wish to signal their memberships in the communities of
speakers of both languages, not just one. CS is emblematic of dual membership, which is also the conclusion of Lüdi
(1992), who discusses in-country Francophone immigrants to German-speaking areas in Switzerland.

7. Variation in structural patterns across communities

Variation in how the Embedded Language (EL) is incorporated into code-switching (CS) reflects differences across
communities, language pairs, and linguistic structures. For example, Moroccan Arabic/French bilinguals show age-based
differences: younger speakers inflect French verbs with Arabic suffixes, while older speakers do not. Such patterns
underscore that EL roles in CS are not universal.

Studies reveal different strategies for integrating EL verbs into the Matrix Language (ML). When the ML is an
agglutinative language (e.g., Finnish, Swahili, Shona), EL verbs are often inflected with ML suffixes. However, this is not
always the case, as shown in Turkish/Dutch CS, where such forms are absent. Interestingly, non-agglutinative languages
like Irish also exhibit EL verb integration, as seen with English stems inflected for Irish grammar. Another frequent strategy
is the "do construction," where an ML verb (e.g., do) carries all verbal inflections while the EL provides a nonfinite verb
form (e.g., CONFUSE-paNNiTTAan in Tamil/English).

The study of EL material can also focus on lexical categories, such as singly occurring EL lexemes. Nortier's research on
Moroccan Arabic/Dutch CS in Utrecht found that single EL words were the most frequent type in her data, emphasizing
their role in CS.

Another area of study is the occurrence and function of EL "islands," which are complete EL structures embedded within
ML discourse. Their frequency and structural significance provide additional insights into EL-ML interactions.

Variation in how EL contributes to CS reflects both linguistic and sociolinguistic factors. The integration of EL verbs, the
dominance of certain lexical categories, and the role of EL islands offer a rich field for comparative research, revealing
patterns shaped by the typological and social characteristics of the languages involved.