Five surprising truths about language mixing

On this International Day of Multilingualism, I celebrate coexisting languages and their speakers everywhere. I’m in good company, since more than half the world’s population is said to speak more than one language, often many more. This means that multilingualism is not the exception, but the norm. And yet this most ordinary state of affairs continues to be associated with a variety of deficits, mainly linguistic. One of the most salient and stigmatized is language “mixing”, widely considered to display laziness and ignorance, when not blamed for the deterioration or even demise of one or all of the languages involved. 

As a sociolinguist, my job is to investigate such stereotypes scientifically. I run a vibrant lab where the structure of bilingual speech represents a major research focus. Over the last several decades, my team and I have carried out cutting-edge analyses involving 13 language pairs, nearly 500 bi- or multilingual speakers, and hundreds of hours of their recorded conversations. Here’s an example of what we heard, produced spontaneously by one of our study participants. 

We systematically located, extracted and analyzed more than 43,000 instances of such language mixing in these data. Our results showed that much of the received wisdom about bilingual speech is simply wrong! Here are 5 surprising facts that explain why.

While language mixing may appear frequent (22,000 cases appeared in our study of the national capital region alone, for example), it actually pales in comparison to the unmixed stretches produced by the same speakers in the same interactions. This is true of every bilingual community we’ve studied, regardless of the specific languages involved: rarely does mixed material account for more than 1% of the verbal output! Needless to say, this is a far cry from what the media or speakers themselves (2) would have us believe. 

Starting a sentence in one language and ending in another, as in the Spanish/English example in (3), is a widespread stereotype of bilingual speech.  

But systematic quantitative analysis shows that mixed discourse is overwhelmingly (95%) constituted of inserting single words from one language – the donor – into a recipient, as in example (1). This is known as lexical borrowing. Most such insertions turn out to be established loanwords, attested for centuries in recipient-language dictionaries, like English-origin chum in example (4), or sandwich, gang, or bar, considered to be “French” words since 1801, 1837 and 1860 respectively.  

You don’t even need to be bilingual to use those; just witness the unfettered use of French-origin words like beef, fashion, government, literature and verdict (among thousands of others) by monolingual anglophones worldwide! In comparison, sentences involving longer stretches of different languages (i.e. multi-word code-switching) are thin on the ground – if used at all – in most bilingual communities systematically studied to date.  

Far from displaying laziness, ignorance or worse, we found language mixing to be orderly and systematic. With no explicit instruction on how to combine languages (obviously, since the practice is generally chastised), bilinguals follow the same unspoken rules. First, they confine their mixing to two major strategies: borrowing and code-switching.

The key mechanism behind borrowing involves converting (or integrating, in technical parlance) donor-language words into recipient-language grammar. In the process, bilinguals divest these words of their original grammatical features and imprint them with those of the borrowing language. Depending on the requirements of that language, this may mean inflecting them with prefixes or suffixes, assigning them a grammatical gender, and and/or positioning them in a specific order. In (5), for example, another participant conjugated the English verb groove with the appropriate French 1st person singular imperfect affix “ais”, gave the noun show French masculine gender, and postposed the qualifier rap to it, following French word order, instead of preceding, as per English. 

Remarkably, borrowed items are treated like their recipient-language counterparts regardless of the languages involved! This explains why the English-origin words in the examples below may seem unrecognizable to unilingual anglophones. Car is inflected with the Tamil accusative marker -ei in (6), consistent with its role as direct object of a Tamil verb, change is conjugated with the Igbo past tense affix in (7), while represent and conference in (8) carry the Ukrainian masculine singular instrumental and feminine singular locative case-markers respectively. And this doesn’t begin to exhaust the possibilities offered by the world’s languages. 

Remarkably, such appropriation of recipient-language grammar does not take centuries to achieve! It happens right away, when the word is first borrowed. This adaptive property of word borrowing has important consequences for understanding how mixing affects the languages involved. 

The other major mixing strategy is code-switching, defined as the juxtaposition of a multiword sequence of one language with that of another. Code-switching is applauded when a bilingual changes languages to accommodate to a monolingual participant, as happens so often, even in officially bilingual countries like Canada. But what fascinates linguists (and horrifies the casual observer) is when it occurs within the confines of a single sentence, as in (3) and (9). Such examples are often invoked to support the widespread claim that code-switchers can’t speak either language. 

Our research shows that nothing could be farther from the truth. On the contrary, the vast majority of such intra-sentential code-switches respect the grammar of not one, but both, of the languages involved simultaneously. How could this be? First, unlike borrowing, the internal constituency of each sequence remains that of its language of origin, explaining, for example, why the adjective dead precedes the noun bodies, just as required in English, rather than following it, as does the borrowed rap in example (5) above. But the degree of bilingual prowess required to code-switch felicitously is most evident from the way the switched segments are positioned. In striking contrast to received wisdom, they’re not just thrown in at random. Instead their placement follows robust, albeit implicit, rules, and as far as we can tell, these operate across all language pairs, regardless of their makeup. Simply put, languages are switched only between sentence elements that are normally ordered in the same way in the unmixed grammars. This rule, which has been formalized as the “Equivalence Constraint” (Poplack 1980), is schematized in the figure below.  

The dotted lines indicate permissible code-switch sites; crossovers indicate elements that remain in the same language (i.e. resist switching), because their word orders conflict. Here, the object pronoun him follows the verb in English, while its counterpart le precedes the verb in Spanish. A switch is thus licensed before the verb phrase or after, but not within it. (This speaker, like many others, opted to eschew two potentially permissible sites in favour of switching at a major sentence boundary (between main and complement clause). Quantitative analysis shows that this implicit constraint is obeyed 99% of the time, revealing that code-switching is a skill, rather than a defect. And further research associating it with the most proficient bilinguals in the community only confirms that fact! This flies in the face of received wisdom. 

Every spontaneous, or nonce, borrowing has the potential to become a bona fide dictionary-attested loanword. But our historical research (Poplack & Dion 2012) shows that few complete that trajectory. Based on data produced by francophone Quebeckers born between 1846 and 1994, we determined that under 7% of their English-origin words persisted over the century and a half studied. This mean that most borrowings are ephemeral. They simply don’t stick around long enough to have a lasting impact on the lexicon of the recipient language. 

It should now be apparent why language mixing is not detrimental to the languages involved. In the first instance, despite the salience of this discourse strategy, both its major manifestations are not only exceedingly rare, but also short-lived. In addition, because borrowed words jettison their grammatical structure to adopt that of the recipient language as soon as they are borrowed into it, it stands to reason that they can’t alter that structure. And since code-switching involves respecting (i.e. maintaining) the grammatical structure of each of the languages involved, it follows that this process doesn’t affect them either.  

So, in addition to the enhanced cognitive benefits, job prospects, sociocultural advantages and more conferred on us by multilingualism, let us also celebrate language mixing, this endlessly expressive and fascinating resource uniquely available to speakers of more than one language. Happy International Day of Multilingualism! 

Shana Poplack, C.M., D.Litt, FRSC, CRC (I) is Distinguished University Professor and Canada Research Chair in Linguistics at the University of Ottawa, and founding director of the Sociolinguistics Laboratory there. She and her team have spent decades scientifically studying bilingualism and its effects among speakers of more than a dozen language pairs, including Canada’s official languages. She is the author of Borrowing: Loanwords in the Speech Community and in the grammar (2018; Oxford University Press). 

* Examples are reproduced verbatim from audio recordings. Codes in parentheses identify the language, (corpus), speaker number and location in transcript or audio recording.