After many conversations on the matter with my colleagues at Georgetown and beyond, and making this a topic of discussion with graduate students in my courses over decades, I want to take this opportunity to present my thoughts on the evolution of applied linguistics.

Let me first contextualize what follows and describe where I come from, intellectually. Input is essential; learning is error-driven. Bilingualism is not exceptional. Becoming multilingual is possible for everyone, but it is more likely to develop for some than for others. Heterogeneity in outcomes is natural because context matters and access to interaction – input, output, feedback – varies across learners. Just as importantly, what the learner contributes to the equation also matters. Learners vary in their capacities and willingness to deploy these capacities to learn from what their context offers. Crucially, it is the interaction between the individual and their context that accounts for language development. Such interaction is complex and in constant flux. Instruction can help, but it can also be a hindrance – just ask me about my ten-year experience with French in grammar school.

I am an applied linguist in a language department with a large doctoral program in the US, where I teach graduate and undergraduate courses in language acquisition, teacher education, and bilingualism broadly defined. I only started speaking Catalan at 18, which may seem strange to some, considering that I am a native of Barcelona, but at the time, political, educational, and social constraints set limits on the ability to speak and write in the minoritized language for many like me. In graduate school, my first publications were on Catalan sociolinguistics, because deep down, all Catalans are sociolinguists. But I had arrived in the US with the goal of becoming a better language teacher, so for my dissertation I decided on an experimental study on Processing Instruction under Bill VanPatten’s mentorship. Back then, I thought that applied linguistics was synonymous with second language learning and teaching, and that my other interests – sociolinguistics and psychology – belonged somewhere else. Still in graduate school, however, the interdisciplinary Second Language Acquisition adn Teacher Education (SLATE) program at University of Illinois at Urbana-Champaign (UIUC) and my first conferences – Second Language Research Forum (SLRF) at Michigan State, then at the American Association for Applied Linguistics (AAAL) soon after – helped me realize that the field was much broader than I had imagined, even though at the time, Teachers of English to Speakers of Other Languages (TESOL) and AAAL happened back-to-back in the same location, which made the AAAL program look more English-centered and pedagogically oriented than it is today.

My background has doubtless shaped the direction of my research interests, which are many. In fact, in a country where academics are hyperspecialized, deciding what to leave aside has been an excruciating experience. I learned to manage it over time, and now I find myself teaching graduate students to do the same. The topics of my publications span from sociolinguistics to applied linguistics, from pedagogical interventions to individual differences, from L2 development to L3 language development in bilinguals, from classroom learning to immersion abroad, and, of late, some L2 writing research.

Along the way, I have heard a number of provocative statements. In graduate school, I was told that context was too messy, and that I would do better focusing on the individual’s mind, which, after all, is universal. At EuroSLA in Barcelona in the late 1990s, the late, great Mike Long questioned my decision to pursue the study of bilinguals learning a third language in front of a large room full of big names. “How wise is your pursuit, considering how little we know about second language acquisition at this time?” he asked. And the first comment after my AAAL keynote questioned my defense of language pledges in study abroad programs in the age of translanguaging. Despite this push to let go and simplify, my research takes an approach to L2 development defined by the interaction between mind and context; I have contributed to the field of L3 development, which has grown exponentially since the 1990s, and we are now looking at students’ perspectives on language pledges during study abroad (https://www.middlebury.edu/language-schools/how-it-works/language-pledge). It is one thing for researchers to criticize the use of language pledges in their defense of translanguaging practices, and another for students to share that view of pledges. As a study abroad program director myself, I can see how some students see pledges as a way of protecting their right to exposure to and use of the target language from less motivated students.

It is not just that my interests are broad; applied linguistics is still too young – five decades old – compared to most disciplines, which means we have to exercise care before we push aside theories, approaches, hypotheses, or research questions. A keynoter at a recent AAAL meeting said, with disdain, that “they (researchers, approaches) do not mention ‘cognitive,’ but you can smell it.” Smell it!? He presented “cognitive” as something to be avoided, just as others will avoid considering the social context as messy. Searching for the most elegant theory – both parsimonious and capacious in its explanatory power – can reduce and oversimplify the very same phenomenon the theory is trying to explain. Is elegance always a reliable guide to scientific truth, then? In a world where there are more multilingual speakers than monolingual speakers, can a theory of language leave bilingual development aside in the name of elegance?

It is not my intention to persuade the reader that “anything goes,” or that it is fine to look at any variable and study any phenomenon minimally related to language as worthy of our energy and time. Nor am I suggesting that we walk into the darkness of applied linguistics, hands desperately extended until we can catch hold of something interesting to ponder over. If we do, we are likely to end up each grabbing and holding on to a different part of the proverbial elephant, enamored with our own partial vision, taking our explanans for the explanandum. Rather, my point is that achieving parsimony requires the generation and consideration of alternative theories, models, or methods that can enhance scientific work by opening a researcher’s mind to other possibilities (Casadevall & Fang, Reference Casadevall and Fang2018). We need to be careful but also open-minded. And I am saying this because I have heard colleagues complain on social media, in the hallways, and over wine at conferences, about how applied linguistics has lost its North Star. For some, qualitative analysis is too subjective and lacks explanatory power. Others see in the excessive preoccupation with social justice a distraction from matters at the core of applied linguistics. Still, for others, neurocognitive approaches are reductionistic. Not everyone is willing to accept the new directions that the field has been following, and it is not just because change, in general, is challenging. The reality is that applied linguistics has undergone a significant transformation, shifting from a traditional focus on grammar and the classroom to an interdisciplinary field deeply integrated with technology and social justice. And the change is accelerating. What has changed? In what direction? When did it start? What sparked the change?

Applied linguistics is currently moving through several “turns.” The Social Turn (Block, Reference Block2003) and its heightened focus on real-world social problems saw applied linguists increasingly address issues like linguistic human rights, the revitalization of endangered languages, and how language is used by powerful entities to create social inequities. Personally, the Black Lives Matter movement, the pandemic, and the current US administration have made me “see” a reality that I cannot unsee in my research. As for teaching, while faculty decide the intellectual direction of their courses, students are often closer to the ground and expect courses and research to reflect the Social Turn, too. This tension is not necessarily problematic, but it is there.

More recently, we witness the “Trans-” Turn. Traditional concepts of “native” versus “non-native” speakers have been replaced by the concepts of translanguaging and multilingualism and by a view of bilingualism as a dynamic, flexible resource rather than two sets of rigid, separate systems. The Social Turn and the Trans- Turn have fostered collaboration among applied linguists of different traditions, especially sociolinguists and psycholinguists, in a virtuous cycle. (cf., Casadevall & Fang, Reference Casadevall and Fang2018). One of the consequences of the Trans- Turn is the recognition that a bilingual is not two monolinguals in one, and that the monolingual baseline is not the one against which a bilingual’s production should be compared. In fact, studies that look into differences among bilinguals – age of acquisition, presence or absence of bilingual literacy, for example – are now considered more informative than comparisons between bilinguals and monolinguals.

The artificial intelligence (AI) Turn: I confess to being already tired of thinking about AI, and that change has only started! But it is unavoidable: AI is redefining the field, creating a new subdiscipline known as Digital Applied Linguistics that already has its own international association (Global Digital Applied Linguistics Association) and academic journal (https://www.castledown.com/journals/dal). Their mission is to foster inclusive, evidence-based dialog and collaboration among researchers, educators, students, developers, and policymakers who are shaping the future of language education and multilingual communication in digitally mediated contexts. AI-driven pedagogy is clearly important, but applied linguists are needed to mediate the tools, ensuring large language models and chatbots communicate ethically, and auditing AI for cultural bias. Decolonization has become a frequent term in Digital Applied Linguistics, as it has in the other social sciences. Applied linguists are now central to developing a variety of language processing tools, including voice assistants, translation apps, chatbots, and grammar checkers.

These turns are taking place at a time when, in Western countries, especially the US, the number of applied linguists in academia is shrinking as universities are merging linguistics, applied linguistics, and language departments with broader departments, and tenure-line faculty are being replaced by non-tenure-line faculty. This is certainly a painful development. I do not have numbers on this, but our graduates are having a torrid time these days. As a consequence, they are increasingly finding employment in government, tech, and business, where salaries are also higher but not conducive to research. Non-tenure-line faculty are almost always hired into the teaching track, without research expectations. Add to this enormous cuts to research grants in general and funds for bilingual education in the US in particular, and it is obvious that the future of applied linguistics research is compromised. This is no longer dominated by Western institutions; scholars in Asia and the Gulf States are increasingly exercising agency to address regional issues and reduce the global knowledge divide. While this is a welcome intellectual development for the field, it also means that competition to get your manuscript accepted to a top journal like Studies in Second Language Acquisition (12% acceptance rate) is now much fiercer than it was. And AI has leveled the linguistic playing field: I remember reviewing manuscripts that were almost incomprehensible. Not anymore, as authors can ask AI to rewrite their manuscripts in beautiful formal English, even when the author’s voice gets lost in the process.

The way we do science in general is changing, too. We now have more powerful data analysis techniques and more sophisticated equipment for making observations and running experiments. What we know has also gained in breadth and depth. My dissertation included fewer than 100 participants, only two independent variables – Time and Condition – and three dependent variables – each from a different task built along a continuum from less to more communicative, hypothesized to correlate with lower to higher demands on working memory capacity. Today, we can run that Multivariate Analysis of Variance in less than 2 minutes on a cheap laptop, dragging and dropping data in SPSS. Back then, I ran the analysis on the university’s computer mainframe; one of the committee members, a professor in the Department of Educational Psychology, assisted with the coding. I picked up the 40-page printout of continuous paper with the results from the general printer on the engineering campus and proceeded to separate the pages to cover my bedroom walls. Who prints their R output anymore? I relied on purely behavioral measures (accuracy – right/wrong). Others after me, using a similar paradigm, added working memory as a covariate or used latency or eye-tracking to provide a window into the process, adding explanatory power. And as for the analytical tools, Analysis of Variance is not extinct yet but is being replaced with large multimodal models (LMMs; Nassaji, Reference Nassaji2025) that are more flexible and provide solutions when statistical assumptions are violated. While LMMs are the current standard, the field is currently experiencing a second wave of methodological shift toward Bayesian mixed-effects models. Back in graduate school, there was one research manual for applied linguists. There is now a new edition of that volume, other manuals, a fairly new but highly cited journal on research methods and analysis, and a very active Facebook page on the topic. And with all this sophistication in tools for data collection and analysis, the research is only as good as its research questions. I have witnessed many a budding researcher miss the forest for the trees (i.e., know all the details about every possible working memory test, but forget the motivation behind the inclusion of working memory in the study).

It can all feel overwhelming, and the speed of the changes adds to that feeling. The rise of the Internet – which was made available to graduate students just as I arrived in Illinois – has facilitated the dissemination of scientific results more rapidly than ever, with studies made available online even before they are printed so they can be scrutinized, incorporated, or replicated at dizzying speeds. Access has been democratized, too. Open access has taken “free” (not necessarily really free to the authors and their institutions, unless authors publish in Diamond Open Access journals) distribution of information to a new level by making original scientific articles available to anyone with Internet access. Open access, registered protocols, materials in IRIS, and OASIS summaries: Ideas, plans, and data can now be easily shared regardless of distance or political and institutional boundaries. Despite the availability of these resources and the support of journals like SSLA and Language Learning, with dedicated sections for replications, replication studies are still rare, which has implications for the discipline. A single study’s false positive can become accepted as a scientific fact upon which to build a theory, and conscious or subconscious methodological shortcuts can go undetected; without conceptual replications, the field cannot establish the boundary conditions of an effect – the L1, the learning environment – and may interpret limited results as universal. In this regard, availability does not seem to translate into progress.

There are no Leonardos or Darwins anymore. Higher complexity and specialization have necessitated more cross-disciplinary collaboration than in the past. Advances in bioengineering have resulted in the COVID-19 vaccine, but collaboration between sociologists, anthropologists, economists, and mathematical modelers assisted in its deployment and in understanding how to prepare the world for the next pandemic. Scientists today are more likely to work in large teams – regardless of disciplinary specialization – and that includes applied linguists. Psycholinguists, sociolinguists, and statisticians work in collaboration with practitioners who are close to the issues and the data. Just take any volume of the major journals in the field and count the number of authors per article. This transition to team science, often transnational, has also been facilitated by the Internet. In contrast, most of the applied linguistics research I read in graduate school was single-authored, often by men (a shout-out to the great female trailblazers in the field!). Yet there are disadvantages in team science: author order and credit, interpersonal conflict, institutional coordination, and consensus-driven conservatism. We will always need the individual who takes creative theoretical risks. But I think that the disadvantages can be worked out, that the pros outweigh the cons in team science, and that four eyes see more than two, as we say in Spanish.

The previous paragraphs delineated the direction of multiple transformations that applied linguistics is undergoing and stressed the speed with which change is taking place. Change is inexorable. It is also uncomfortable because we are wired to seek safety in predictability. But change is also a necessary prerequisite for growth and innovation. Since we cannot – should not – stop change from happening, we should focus on our response to change. Depending on our position and responsibilities, we can affect the direction of change in many ways while wearing our different hats. The following paragraphs summarize ways in which I work with those changes to lead them in the direction I consider appropriate for my purposes, depending on that day’s hat. Specifically, I discuss the impact of AI on research generation and evaluation and on graduate student education; team science; theoretical turns; and narrow, tribal views of where applied linguistics should be headed.

Journal editors play a large role in taking the reins of AI, but as reviewers and authors, we can draw editors’ attention to the inappropriate use of AI in the process of both writing and reviewing. A few weeks ago, one of the reviewers for a forthcoming manuscript of mine (double-blind) cited another of my publications in their review, misrepresenting the design and making claims about its conclusions that were plainly false. This prompted me to change the way I engaged with the references from all three reviewers, instead of taking their accuracy at face value. This was, after all, a top journal. I have heard AI may be assisting journal editors. Perhaps, but it is certainly making their work more complicated too.

Still on the topic of AI, and wearing my teacher’s hat, I have recently noticed that graduate students’ papers include references that do not exist, or they do, but, like the reviewer’s report, attribute wrong results and conclusions to them. And even when they are correct, the ideas sound generic and uninteresting, and the choice of references does not have a logic – they are not the most cited, or the most recent, or the classic ones that first put the idea forward. Even before AI, students proposed hypotheses that had already been tested and discarded – uninteresting or impossible to test. In a field as new as applied linguistics – Language Learning was founded in the late 1940s but applied linguistics publications started in earnest around the 1970s and Krashen’s times – students dare to ask if they can cite literature published before 2020? In my graduate course, Instructed SLA for new doctoral students, we devote one week to each theory, each associated with three readings: an introductory chapter and two empirical studies, one a classic and the other very recent. We discuss where the theory started, how it relates to other theories, and where the theory is now. Students need to be aware of the evolution of ideas in applied linguistics in order to know how to move forward – rather than in circles. It is a way of taking control of change.

We also act as gatekeepers when we evaluate faculty dossiers for tenure and promotion. Co-authorship, totally indisputable and manageable in the hard and social sciences, is still challenged in language departments where votes from humanists often dominate. It falls on the external evaluator to do some pedagogy and devote part of their report on the applicant’s dossier to educate other colleagues on the applied linguistics profession’s standards, what applied linguists do and how they do it, and the ways in which doing science is changing, including team science in the social sciences.

I heard the unhappiness triggered by what felt like a significant move away from cognitive-oriented second language acquisition (SLA) and the classroom toward critical applied linguistics, with ways of doing research that are closer to anthropology. Given my background and my intellectual interests, I do not think that moving toward critical applied linguistics is negative, but we should move toward critical applied linguistics without leaving cognitive-oriented SLA behind. That said, I must admit I felt a bit awkward at the AAAL meeting in Houston, akin to not belonging. Every keynoter devoted an inordinate amount of time to detailing their “positionality.” The topics and perspectives were narrower, and I worried that the reviewers’ answers to the question on the abstract’s contribution to social justice might have left out important research. Interestingly, I see that the reviewer form for the AAAL27 Conference has removed that question.

In contrast, at the 2026 AAAL conference in Chicago, in part because I was doing my own very informal meta-analysis of AAAL meetings, I sat through every keynote, enjoyed all of them, and learned a lot from all the speakers, even when the topics, perspectives, and sources of evidence were widely varied. Because “applied linguistics addresses a broad range of language-related issues in order to understand their roles in the lives of individuals and conditions in society” (https://www.aaal.org/2027-call-for-proposals), it must be flexible so that it can accommodate new concerns and opportunities. Looking at the areas included, one may wonder what Antiracism, Decolonization, and Intersectionality for Systemic Transformation has in common with Language Cognition and Brain Research. Is raciolinguistics solely the matter of sociolinguistics and anthropology? Except that for some years now, Janet van Hell at Penn State has been conducting research on the neural and behavioral mechanisms behind accented-speech processing, speaker identity, and implicit bias rooted in the cognitive neuroscience of bilingualism (Grey & van Hell, Reference Grey and van Hell2017). Their research shows that a speaker’s foreign-accented identity changes how the brain processes language, connecting to broader systemic issues such as linguistic discrimination, with implications for the legal system, the workplace, and housing. It is thus possible to investigate research questions that address social issues using models and tools common in neurocognitive approaches to bilingualism/SLA. van Hell’s research represents just one example among many. Approaches, theories, methods – they are not an either/or; compatibilities are possible; it is not a zero-sum situation. So please, let’s not throw the baby out with the bathwater when we find the research before us leans too much toward the cognitive or the socio – the mind and its context – or is too qualitative or too quantitative or focuses too little or too much on the classroom. Change is in the nature of things, and we need to stay open and be informed to manage it.