“Science is what we have learned about how to keep from fooling ourselves.”– Richard Feynman
The quote above from the physicist Richard Feynman implies that things are not always as they seem. What we perceive as indubitable reality is often far from it. Indeed, many of science’s most profound discoveries have been counterintuitive, requiring scientists to transcend both commonsense explanations of phenomena and that which may be directly observed through the senses. The study of language, to the extent that it examines how languages are represented in the minds of speakers, is no different.
In the field of linguistics, the traditional notion of bilingualism is that multiple grammars are represented in the bilingual brain (Poplack, 1981; Myers-Scotton, 1993; MacSwan, 2014). More recently, however, certain scholars have challenged this stance. Instead, they argue that what appears to be multiple grammars is really one unified set of linguistic features—a unitary grammar (Otheguy et al., 2019). Which conception of bilingualism has more theoretical and empirical validity? The answer to that question has important implications, not just for scholars studying language and the mind but also for educators engaged in teaching emergent bilinguals across a variety of instructional contexts.
The Proposed Paradigm Shift: Translanguaging Theory
At first glance, it may seem as though bilinguals communicate by switching from one set of separate and distinct grammar rules to another, depending on the language they chose to employ. For example, if I were to utter the sentence “I am angry” in Spanish, I usually wouldn’t bother to include the subject pronoun I, which in Spanish is yo. I would simply say “estoy enojado,” because Spanish does not require the subject pronoun to be explicitly voiced. English, in contrast, does require the subject pronoun’s articulation, barring special cases such as imperative sentences (e.g., “Go to your room”).
In an apparent effort to stop fooling ourselves, translanguaging theorists posit that what we perceive as different languages (e.g., Spanish, English, Swahili, etc.) do not exist, at least psychologically speaking (Pennycook and Makoni, 2005; Pennycook, 2006). Rather, named languages, they contend, operate as mere social constructions, emerging through power-infused interactions among and within people groups over a sustained period. As a result, translanguaging theorists deny bilinguals have anything like two or more discrete grammars instantiated in their minds. According to Otheguy, García, and Reid (2015), what bilinguals do possess are “unitary collections of features, and the practices of bilinguals are acts of feature selection, not of grammar switch” (p. 281). Thus, when bilinguals appear to switch between grammars of different languages, like in the example given above, they are simply selecting features from within a unitary mental grammar.
Scholars from the code-switching camp maintain that bilinguals do in fact develop discrete yet overlapping mental grammars. One such scholar, Jeff MacSwan (2017), argues that a bilingual’s linguistic repertoire contains what he calls “language-specific internal differentiation” (p. 181). On the one hand, this stance recognizes that no two people come to acquire exactly the same language. It’s long been known that differences exist within individuals’ linguistic systems, even among those from the same speech community (Gumperz, 1964; Chomsky, 1986). Yet for most linguists, the construct of language is a useful idealization akin to how physicists use frictionless planes; moreover, linguists contend that theoretical and empirical justification exists for using the descriptor language-specific in describing internal linguistic phenomena. As such, code-switching researchers staunchly defend bilingualism as psychologically real.
Since MacSwan has proven the most outspoken critic of translanguaging theory, and because I view his proposed minimalist approach to code-switching (MacSwan, 2000) as particularly well-developed, my subsequent analysis will adopt many aspects from his framework. It is important to recognize, however, that disagreements abound among researchers in the code-switching camp. Nevertheless, they all agree on the existence of some type of internal distinction between the grammars of different languages. For the purposes of this article, then, I will refer to those who subscribe to the language-specific internal differentiation hypothesis as code-switching theorists and those who adhere to the unitary grammar hypothesis as translanguaging theorists.
One clear example code-switching theorists offer to support the argument that language-specific internal differentiation exists within the bilingual mind is adjective–noun word order in Spanish–English code-switching. In English, adjectives come before nouns (e.g., the white house), while in Spanish, adjectives typically come after nouns (e.g., la casa blanca). MacSwan (2017) cites research showing English–Spanish bilinguals strongly prefer code-switching constructions such as the white casa over constructions such as the casa white. He and other scholars observe that sentences such as “I am going with friends to the casa white” are extremely rare in English–Spanish code-switching corpora due to certain cognitive constraints imposed by the respective grammars; moreover, they are consistently judged by English–Spanish bilinguals as ungrammatical under rigorous experimental conditions (Belazi et al., 1994; Cantone and MacSwan, 2009; Gumperz, 1967; Lipski, 1978; Poplack, 1981; Sankoff and Poplack, 1981).
Translanguaging Theorists’ Rebuttal
Translanguaging theorists Otheguy, García, and Reid (2019) counter by claiming MacSwan’s adjective–noun word order examples fail to meet “basic requirements of observational adequacy” (p. 629). To support this criticism, they cite various utterances one or more of the authors overhear throughout the course of their daily lives. Such examples include a woman asking her child the following: “¿Querés ir al basement de la otra tienda y compramos esos cositos [sic] cool? (“Do you want to go to the basement of that other store and buy those little cool thingies?”) (p. 642). On the surface, the phrase esos cositos cool seemingly contracts code-switching theorists’ claims of adjective–noun word order generation in English–Spanish code-switching data. Another example provided in Otheguy et al. is of a conversation one of the authors overhears between a couple while they are making their bed where one interlocutor said ‘Estas sábanas no son deep, ‘only to be answered ‘¿Por qué no compras sábanas deep?,’ where sábanas deep is again presumably not allowed by the rules that MacSwan tells us establish the systematicity of code switching” (p. 643).
How might we arbitrate between these two theories of bilingualism? To keep from fooling ourselves, as Feynman suggests, we must examine that which we do not see, or in this case hear. Indeed, a commonality among all explanatory theories underlying the core of language and bilingualism is their internalist nature. Any theory that purports to explain a portion of the human cognitive architecture, as code-switching and translanguaging theories do, must clarify how the human mind constructs internal representations in such a way that aligns with empirical data. To better explain why this must be so, a quick detour into epistemology, or how knowledge is acquired, is necessary.
Theory-Laden Observation, Fallibility, and Illusion
How do magicians, or illusionists, make objects as small as a coin or as big as the Statue of Liberty disappear into thin air? They exploit the fact that the “reality” we all think we so clearly observe is ultimately interpreted through, and even constructed by, our cognitive faculties. That is, what we experience is necessarily contingent upon our brains filling in gaps and making predictions based on denigrated and incomplete sensory data. Our brains fill in these gaps by utilizing both bottom-up theories, such as how our visual systems automatically detect contrast without our conscious awareness—a magician can make objects appear to float by exploiting this knowledge—and top-down theories, such as the intention we ascribe to someone’s physical movements—a magician who scratches his or her head in the middle of the show is probably not doing so to relieve an itch but rather to deceive the audience in some way or another (Macknik and Martinez-Conde, 2010).
In fact, observation devoid of theory is impossible, for the very reason we are astounded by a magic trick in the first place is because our prior theories about how something is supposed to function are disrupted. For example, our theories of biology tell us that a woman in the process of being cut in half is supposed to be screaming in agony, not smiling gleefully while lying in a box. Theoretical physicist and philosopher David Deutsch (2011) puts it this way:
All observations are… theory-laden, and hence fallible, as all our theories are. Consider the nerve signals reaching our brains from our sense organs. Far from providing direct or untainted access to reality, even they themselves are never experienced for what they really are—namely crackles of electrical activity. Nor, for the most part, do we experience them as being where they really are—inside our brains… Our sense organs themselves, and all the interpretations that we consciously and unconsciously attach to their outputs, are notoriously fallible—as witness the celestial-sphere theory, as well as every optical illusion and conjuring trick. So we perceive nothing as what it really is. It is all theoretical interpretation: conjecture. (p. 10)
In sum, two important principles are present in the excerpt above as well as in my preceding commentary: 1) theoretical interpretation undergirds all observation, and 2) our theories or interpretations of sensory data are, as Deutsch rightly states, notoriously fallible.
Applying said principles to the current debate surrounding the bilingual brain elucidates the divide between the two theories. The code-switching orientation of the kind subscribed to by MacSwan is grounded in an explanatory theory of the internal workings of the bilingual mind, which is integral in constructing our perceived linguistic reality. In contrast, translanguaging advocates base their theory more directly on the lived experiences of bilinguals, which, by definition, place the emphasis on observation instead of explanation.
To further unpack this important divide, consider the following difference between code-switching as conceptualized by MacSwan and translanguaging theory. MacSwan’s version of code-switching theory presumes the existence of two important theoretical constructs, internal language, or I-language for short, and external language, or E-language. Noam Chomsky (2006), widely considered the father of modern linguistics, defines I-language as linguistic knowledge internal to the individual, the product of a language-specific computational system. E-language, in turn, is defined as externalized linguistic output, or language people experience and produce as they go about their daily lives within their communities.
While the counterexamples provided by Otheguy et al. (2019) (“esos cositos cool” and “sábanas deep”) would be considered E-language, theoretical linguists, in line with principle two, are committed to studying I-language. For just as we don’t experience nerve signals for what they really are, the linguistic output that is spoken, heard, or seen in the world is rarely an accurate indication of our linguistic competence, or the linguistic knowledge we all come to possess in our minds. Translanguaging theorists fail to fully recognize this by rejecting the construct of I-language, along with the much deeper explanations that accompany it.
One particularly profitable approach to studying linguistic competence is to posit a condition that states linguistic features encoded in a lexicon (i.e., case, number, person, and gender) must be compatible across a phrase or sentence. This is due to a phenomenon called feature checking. If the abovementioned features are incompatible, then the utterance is ungrammatical, regardless of whether it is a code-switching or monolingual utterance.
For purposes of illustration, consider the simple phrase the casa, consisting of the English determiner the, which marks for person and number, and the Spanish noun casa. Since casa is a gendered noun, it requires a determiner that also marks for gender (e.g., la or una), which the does not (Moro, 2014). The consequence of this feature mismatch is that the utterance crashes, which explains why it is consistently judged as ungrammatical in experimental settings (Aguiree, 1976; Lipski, 1978). Constructions such as the casa white crash for reasons involving not only feature checking but also another technical phenomenon in linguistics called movement. The interested reader will find a technical discussion of adjectives and word order as it pertains to code-switching in Cantone and MacSwan (2009).
Regarding Otheguy et al.’s counterexamples (i.e., sábanas deep and esos cositos cool), no sufficiently detailed theoretical account as to why these constructions are grammatical is ever posited, nor is sufficient empirical evidence presented in support of such a claim. Again, the authors seem to rely on observational data devoid of the type of explanatory framework cited above. A comprehensive explanation of the tools linguists employ to analyze language is well beyond the scope of this article. Nevertheless, in briefly discussing these few examples in passing, I hope to highlight the fact that when researchers adopt an explanatory approach to the study of language, they often uncover layers of abstract and counterintuitive rules, structures, principles, and constraints, which bear little resemblance to the actual speech we produce and process in our everyday lives.
It’s telling that instead of utilizing a construct such as I-language, Otheguy et al. (2015) opt for a more “theory-neutral” term called idiolects, which they define as “structured lists of lexical and grammatical features… a mental grammar that is acquired primarily through, and deployed mostly in, social and personal interaction” (p. 289). However, in line with principle one, a theory-neutral construct, by definition, lacks explanatory adequacy. As MacSwan (2017) states, “an explicit linguistic theory is one that does not rely on tacit assumptions about how language works but rather overtly states all relevant details” (p. 186). Indeed, to glean insights into the internal representations that exist in a bilingual’s mind requires linguists to make bold conjectures; that is, they must posit explanatory theories regarding such representations and how they may emerge. Then, linguists must test said theories against the data by conducting rigorous experiments, not by uncritically relying on data gleaned from overheard conversations.
To be fair to Otheguy et al. (2019), placing unmerited faith in observational data is a common error committed by many scientists and linguists alike. Oddly, it has become the dominant philosophical orientation in cognitive science today. Unfortunately, adherence to this practice has significantly deterred efforts to, as Feynman puts it, keep from fooling ourselves, as it replaces science’s fundamental purpose—the search for good explanations (Deutsch, 2011)—with superficial data analysis.
A Brief Message to Bilingual Educators
Putting theoretical differences aside, any approach to bilingual instruction should be judged on the extent to which it fosters academic achievement and a high degree of bilingualism and biliteracy in students. Educators are also faced with a moral imperative to value the language practices and identities bilinguals bring with them to the classroom. Resources such as García, Johnson, and Seltzer’s (2017) book The Translanguaging Classroom are commendable attempts to achieve the latter.
Still, if educators choose to create translanguaging spaces in their classrooms, they shouldn’t do so based on an illusion. Students may not be developing a unitary grammar, as translanguaging theorists would have educators believe. They may, however, be developing integrated yet discrete mental grammars that are deployed in strategic and rule-based ways. As MacSwan (2017) puts it, “codeswitching research has shown through detailed analysis that bilinguals are exquisitely sensitive to an incredibly rich and intricate underlying system of rules for both languages in their repertoires” (p. 190). It would be regrettable if educators were made to dispense with such an established view based solely on translanguaging theory’s conception of language and bilingualism.
Available at www.languagemagazine.com/references-david-burns-feb-2022
David Burns has worked as an ESL/bilingual teacher and coach in Puerto Rico and in various public schools across the US. He currently serves as director of language and literacy development for a national learning organization dedicated to improving academic outcomes for K–12 students.