TYPE Conceptual Analysis

PUBLISHED 17 October 2025

DOI 10.3389/fpsyg.2025.1636777

Beyond comprehensible input: a

OPEN ACCESS neuro-ecological critique of
Krashen’s hypothesis in language
EDITED BY
Nilushika Prasangani,
Sabaragamuwa University, Sri Lanka

REVIEWED BY
Ahmed Ghorbel,
education
University of Jendouba, Tunisia
Jenrev Macapagal,
Pampanga High School, Philippines Quang Nhat Nguyen* and Dung Thi Hue Doan
*CORRESPONDENCE
Saigon International University, Ho Chi Minh City, Vietnam
Quang Nhat Nguyen
nguyennhatquang@[Link];
[Link]@[Link]

RECEIVED 28 May 2025

This article critically reassesses the Comprehensible Input (CI) hypothesis
ACCEPTED 22 September 2025 in language education by drawing on recent advances in neurolinguistics
PUBLISHED 17 October 2025 and an ecological perspective on learning. While the CI hypothesis claims
CITATION that language is acquired by understanding input slightly beyond a learner’s
Nguyen QN and Doan DTH (2025) Beyond
current competence (i+1), converging evidence from brain research shows
comprehensible input: a neuro-ecological
critique of Krashen’s hypothesis in language that language development is an active and embodied process supported by
education. Front. Psychol. 16:1636777. interaction, feedback, and multimodal engagement. From the ecological point
doi: 10.3389/fpsyg.2025.1636777
of view, affordances are the perceivable opportunities for action that arise in
COPYRIGHT the ongoing coupling of learner and environment. Using this combined neuro-
© 2025 Nguyen and Doan. This is an
open-access article distributed under the
ecological lens, the paper critically reviews empirical studies from the last
terms of the Creative Commons Attribution three decades and demonstrates that meaningful language growth depends on
License (CC BY). The use, distribution or learners detecting and acting on such affordances rather than merely processing
reproduction in other forums is permitted,
provided the original author(s) and the
linear and simplified input. Adaptive and AI-supported learning systems further
copyright owner(s) are credited and that the illustrate how contemporary technologies can operate these mechanisms and
original publication in this journal is cited, in offer individualized, scalable alternatives to the static i+1 model. The analysis
accordance with accepted academic practice.
No use, distribution or reproduction is
argues that CI should no longer serve as a central doctrine in language education
permitted which does not comply with these and calls for pedagogies that reflect the interactive, affordance-rich processes
terms. revealed by current brain and language science.

KEYWORDS

comprehensible input, neuro-ecological approach, affordance theory, adaptive

learning, AI in education, language learning

Introduction
A new wealth of interdisciplinary research in neurolinguistics, brain-based learning,
and ecological language perspectives has cast fresh doubt on the comprehensible input
hypothesis. Stephen Krashen’s Comprehensible Input (CI) hypothesis has historically
held sway in Second Language Acquisition (SLA) theory and pedagogy. Krashen (1982)
famously argued that learners acquire language “in only one way”, by understanding input
that is just beyond their current level of competence (the i+1 principle; Krashen, 1982).
Introduced as part of his Monitor Model in the late 1970s and 1980s, this hypothesis shifted
language teaching toward meaning-focused exposure and away from rote grammar drills
(Krashen, 1985). Indeed, Krashen’s ideas spurred approaches like the Natural Approach
and Communicative Language Teaching, emphasizing immersive exposure over explicit
instruction (Krashen and Terrell, 1983). Historically, the input hypothesis had a profound
impact, becoming “the single most important concept in SLA” in Krashen’s view (Krashen,
1996) and attracting many practitioners worldwide (Lightbown, 2019). However, from its
inception, the theory was also highly controversial. Critics argued that Krashen’s claims

Frontiers in Psychology 01 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

were vague, untestable, and overstated, lacking clear definitions and will spur acquisition, provided the input is understood, made
robust empirical support (McLaughlin, 1978; Gregg, 1984; Gass and comprehensible through context, or aided. Krashen maintained
Selinker, 2008). that acquisition is distinct from conscious learning: given sufficient
Findings from cognitive neuroscience show that language comprehensible input and low anxiety (a low affective filter),
learning is not a passive process of absorbing input, but an learners will subconsciously absorb the new language (Krashen,
active, embodied one, engaging multiple brain systems through 1985). Grammar teaching, error correction, or forced output
interaction, feedback, and multimodal experience (Pulvermüller, were deemed unnecessary or secondary; understanding input was
2018; Tomasello, 2019). Likewise, ecological theories (e.g., considered necessary and sufficient for acquisition in most cases
Van Lier’s affordance model) reconceptualize language learning (Krashen, 1985, 1996). This stance was encapsulated in Krashen’s
as emerging from learners’ dynamic interactions with rich bold claim that comprehensible input is the causal factor in SLA,
environments, rather than from simplified, teacher-controlled and that “all other factors. . . work only when they contribute to
input (van Lier, 2004; Larsen-Freeman, 2018). In parallel, comprehensible input and/or a low affective filter” (Krashen, 1985,
advances in personalized and AI-supported learning demonstrate p. 4). In other words, motivation, interaction, or output matter only
that a one-size-fits-all i+1 approach fails to address individual insofar as they increase understandable input or reduce anxiety,
learner pathways, whereas adaptive systems can finely tune according to Krashen.
input and practice to each learner’s needs (Nguyen, 2022a; The impact of this hypothesis on language education was far-
Paladines and Ramirez, 2020; Lin et al., 2023). In short, reaching. Krashen’s ideas helped catalyze a paradigm shift in the
Krashen’s comprehensible input hypothesis appears conceptually 1980s from grammar-intensive methods (e.g., audiolingual drills,
flawed, empirically outdated, and practically insufficient in grammar-translation) to more meaning-focused, communicative
modern contexts. approaches (Lightbown, 2019). Along with Terrell, Krashen
This article systematically critiques the CI hypothesis through developed the Natural Approach, encouraging teachers to flood
an interdisciplinary lens. We begin by outlining Krashen’s students with slightly challenging input in low-anxiety settings,
original hypothesis and its legacy. We then synthesize evidence postponing speaking until learners felt ready (Krashen and
from neurolinguistics and brain research showing that language Terrell, 1983). While rooted in broader sociolinguistic theory, the
acquisition is an active, neuroplastic process that depends on Communicative Language Teaching (CLT) movement found a
interaction and embodied experience rather than the passive strong theoretical justification in the Input Hypothesis’ emphasis
consumption of input. Next, we draw on an ecological framework, on meaningful exposure over explicit grammar instruction (Long,
including van Lier (2004) and Nguyen’s (2022a,b, 2025a,b) work 1996; Richards and Rodgers, 2014). Teachers were encouraged
on affordance theory, to argue that meaningful learning emerges to use simplified target language, visuals, and gestures to make
when learners engage with rich, multimodal environments instead input comprehensible, assuming that if students understood the
of relying on simplified input alone. Within this integrated neuro- message, even with some new structures present, their language
ecological view, experience-dependent brain plasticity operates in proficiency would naturally grow. This approach appealed to
tandem with the ecological processing of affordances, allowing practitioners for its intuitiveness (mirroring how children seem
linguistic knowledge to continually emerge and reorganize as to pick up language) and its humane, low-stress ethos. Krashen
learners act within their environments. We critically examine often invoked the example of children acquiring L1 through
the i+1 concept’s limitations in scaling and personalization, immersion in language they only partly understand, not through
contrasting it with modern individualized, data-driven learning grammar drills. By the late twentieth century, the comprehensible
pathways and AI-enhanced environments. A detailed neuro- input principle had become attractive to many language teachers
ecological review of recent empirical studies in neurolinguistics, worldwide, profoundly influencing curricula, teacher training, and
ecological language learning, and adaptive learning supports this even popular self-study methods. Despite, or perhaps because
critique. We conclude by calling for the abandonment of CI as of, its popularity, Krashen’s theory quickly attracted extensive
a guiding pedagogical doctrine and for a reconceptualization of criticism. Scholars noted that the Monitor Model, including the
language education using neuro-ecological and affordance-based Input Hypothesis, was “one of the most controversial theoretical
frameworks that better reflect how languages are learned in the perspectives in SLA in the last quarter of the twentieth century”
twenty first century. (Gass and Selinker, 2008, p. 298).

Krashen’s comprehensible input Insights from neurolinguistics

hypothesis: theory and impact on neurolinguistic evidence: language
language education learning as an embodied, interactive,
and neuroplastic process
Stephen Krashen’s Input Hypothesis emerged in the late 1970s
as part of his broader Monitor Model of SLA. The Input Hypothesis In contrast to Krashen’s theory, modern neurolinguistic
posits that “language is acquired by receiving comprehensible input research shows that language development is not a passive process
slightly above one’s current level of competence (i+1)” (Krashen, of absorbing comprehensible input. Instead, it is a constantly active,
1982, p. 21). In practical terms, if a learner’s proficiency is i, brain-modifying endeavor that requires sensorimotor engagement,
then exposure to language just one step beyond that level (i+1) social interaction, and continual feedback. Krashen’s hypothesis,

Frontiers in Psychology 02 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

by contrast, depicts the learner’s mind as a language “acquisition refer to neuroplastic adaptations, such as growth in neural
device” that merely needs understandable input and low anxiety connectivity or more efficient brain activation patterns, which
to operate (Krashen, 1982). Contemporary cognitive neuroscience are observed when learners engage in communicative interaction
paints a very different picture: understanding language is in the L2. One illustrative study found that adult learners who
inseparable from using language, and neuroplastic change is driven practiced conversation with feedback showed greater increases in
as much by practice and embodied experience as by exposure itself brain activity related to syntactic processing (and made larger
(Pulvermüller, 2018; Li and Jeong, 2020; Tomasello, 2019). proficiency gains) than those who spent equivalent time passively
First, consider the role of neuroplasticity and active watching L2 videos (Li and Jeong, 2020). Interactive conditions
engagement. Neuroplasticity refers to the brain’s ability to likely stimulate neurotransmitter systems (e.g., dopamine release
rewire by forming new neural connections in response to learning associated with reward and motivation) that facilitate memory and
and practice. To induce such changes, mere exposure is often learning (Ullman, 2016). Indeed, neuroscience has demonstrated
insufficient; active processing and retrieval of language seem to that novelty and active involvement trigger dopamine in the
be critical (Pulvermüller, 2018; Ellis, 2008). Studies have shown, brain’s reward pathways, strengthening new information retention
for example, that when learners actively produce language or (Lisman et al., 2011). A classroom solely centered on teacher-
engage in conversation, distinct brain networks are recruited provided comprehensible input, especially if it involves little
compared to when they only comprehend input. In a meta- novelty or student agency, risks lower engagement on a
analysis of neuroimaging studies, language production (speaking) neurological level. By contrast, a class that prompts learners
was associated with additional activation in right-hemisphere to interact, solve problems, and get immediate feedback on
frontal and temporal regions related to motor planning, whereas their output will activate attentional and reward networks
comprehension primarily engaged left-hemisphere language that enhance learning and memory consolidation. Feedback
regions (Indefrey and Levelt, 2004; Price, 2012). In other words, closes the loop for the brain: it allows the learner to correct
producing language uses broader neural circuitry than listening, errors and adjust their internal representations. Brain imaging
engaging areas involved in motor control, sensory feedback, and research on second-language training has shown associations
cognitive monitoring. This suggests that active use provides a between corrective feedback and activation in prefrontal and
different and arguably deeper kind of learning stimulus to the language-related regions, consistent with the idea that the brain
brain than passive input; it lights up more of the brain, creating updates linguistic representations during feedback-based learning
more pathways that can later consolidate into new language skills (Meyer et al., 2013; Li, 2010). These interactive, feedback-driven
(Pulvermüller, 2018). processes are not accounted for in Krashen’s input-centric model,
Additionally, brain-based research on embodied cognition where understanding input alone is meant to suffice. Modern
finds that understanding language often involves simulating neurolinguistics instead indicates that language acquisition is an
experiences; for instance, hearing action verbs can activate active skill acquisition process, more analogous to learning a
motor cortex regions as if the listener were performing the musical instrument or sport than to filling an empty vessel with
action (Pulvermüller, 2018; Barsalou, 2008). Such findings align content. As with those domains, practice (including output) and
with the idea that language is grounded in sensorimotor and feedback are key drivers of improvement, physically shaping neural
emotional systems (Tomasello, 2019). If language meaning is tied pathways over time (Pulvermüller, 2018; Ullman, 2016).
to perception and action, learning words and structures might Another line of evidence comes from embodied language
require connecting them to embodied experiences, not just hearing learning and multisensory integration. A growing body of research
them in simplified sentences. A purely input-focused approach shows that pairing language with sensory, motor, or emotional
might under-stimulate these essential brain processes (Gullberg, experiences can significantly boost acquisition (Gullberg, 2010;
2010). Verbotonalism, as discussed in Yang et al., 2023 and Cai Macedonia and Von Kriegstein, 2012). For example, young learners
et al. (2021), highlights the integral role of prosody, particularly who gesture or act out meanings while learning new vocabulary
intonation, in the perception and production of individual sounds. tend to remember those words better, and neuroimaging confirms
According to Cai et al. (2021), intonation is ever-present in spoken that their sensorimotor brain regions become linked to the word
language and influences the quality of vowels and consonants, representations (Macedonia and Von Kriegstein, 2012; Ping et al.,
suggesting that focusing on prosodic features can enhance 2014). A review of embodied learning in language education
pronunciation more effectively than segmental approaches alone. identified that embodied approaches (from drama and movement
Crucially, social interaction and feedback have been shown activities to VR immersion) improve language skills and increase
to enhance neural and cognitive outcomes in language learning, emotional and motivational engagement, providing a more holistic
beyond what input alone can achieve. Recent works on what learning experience for students (Macedonia and Von Kriegstein,
has been called the “social brain of language” emphasize that 2012). Such findings underscore that language is learned and stored
humans are biologically tuned to learn more effectively, but not in the brain alongside a rich tapestry of sensory and contextual
exclusively, through interaction, though learning is still possible information, not as isolated linguistic code. Comprehensible
through other mechanisms. Li and Jeong (2020) review evidence input, especially when “finely tuned” to be just a bit above the
that language learning “unfolds in socially interactive contexts” current level, often involves simplifying and decontextualizing
and report that “work from several recent L2 studies also suggests language (Ellis, 2008). This contradicts what we know about
positive brain changes along with enhanced behavioral outcomes memory: information with multiple modalities and rich context is
as a result of social learning” (p. 1). These positive brain changes remembered far better than bare-bones input. The CI hypothesis

Frontiers in Psychology 03 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

also largely ignores the affective dimension (beyond “keep anxiety 2004). Van Lier succinctly said, “Ask not what’s inside your head,
low”), but brain research indicates that emotional resonance and ask what your head’s inside of ” (van Lier, 2000, p. 6).
learner investment in content can drive attention and memory Whether something becomes an affordance depends on the
(Tomasello, 2019). organism’s capabilities and interests. In the context of language, an
In short, brain-based evidence rejects the notion of the learner affordance is a property of the meaningful or useful environment
as a passive sponge for input. Instead, it portrays learners as active for communication, relative to the learner’s goals and abilities
meaning-makers whose brains thrive on interaction, multisensory (van Lier, 2004). Building on van Lier’s (2000) call to move “from
context, emotional engagement, and producing language. An input input to affordance,” a growing body of scholarship now treats
hypothesis that does not incorporate these active ingredients is affordance as a relational, emergent property of learning ecologies
neurologically implausible. Understanding language is necessary, rather than a static feature of the environment. Work by Rietveld
of course, but understanding alone, without use or contextual and Kiverstein (2014) and Chemero (2009) has emphasized the
grounding, is not sufficient to rewire the brain’s linguistic capacities dynamic coupling of perception, action, and sociocultural norms,
(Nguyen, 2022b). As Gregg (1984) noted, Krashen offered no while Larsen-Freeman (2018) and Tomasello (2019) have shown
mechanism for converting comprehended input into acquired that language development unfolds through iterative interactions
competence. Modern neuroscience suggests that the mechanisms that blur the line between environment and cognition.
involve practice, feedback, and embodied experience, triggering Within this international conversation, Nguyen (2022a,b,
neuroplastic changes, factors largely outside the scope of the 2025a,b) and Nguyen and Doan (2025a,b) extend the concept
original Input Hypothesis. by theorizing affordance as a multi-layered neuro-ecological
network in which perceptual salience, learning valence,
normative constraints, and intentional agency interact with
technological mediation. Their Five-Dimensional Affordance
Ecological theory and affordances: Framework (perceptibility, valence, compositionality, normativity,
intentionality) resonates with Rietveld and Kiverstein’s “skilled
context, interaction, and intentionality” hypothesis while adding an explicit focus on
multimodality in language education AI-supported learning environments and learner-driven reflection.
Idiographic case analyses of AI-mediated interaction (as suggested
While neurolinguistics shines a light on internal brain processes in Nguyen, 2025b) further illustrate how individual learners notice,
and the importance of action for learning, ecological approaches revalue, and orchestrate affordances across time, a process that
to language learning shift the focus to the external environment, aligns with usage-based views of emergent grammar (Ellis, 2008)
the whole context in which learning occurs. An ecological and with research on self-regulated engagement in technology-rich
perspective, as articulated by scholars like van Lier (2000, 2004), settings (Godwin-Jones, 2019). Together, these strands suggest
posits that language acquisition emerges from the relationship that affordances in language education are understood as co-
between learners and their environment, not from input per se. constructed opportunities for action and meaning-making that
The environment offers affordances for learning, opportunities recruit neural, social, and technological resources, rather than as
for meaningful action and perception, and learners actively pick teacher-controlled inputs to tune to a notional i+1 finely.
up on these affordances to develop language (van Lier, 2000, This ecological view holds that language learning is
2004). This starkly contrasts with the Comprehensible Input fundamentally relational (van Lier, 2004; Larsen-Freeman,
(CI) model in which the environment is essentially a delivery 2018). It is about the learner in context, continuously adapting
mechanism for graded input. Here, we explore how affordance to and co-creating their environment. In such a view, simplified
theory and ecological thinking expose fundamental shortcomings input (the typical implementation of i+1) appears impoverished.
in the comprehensible input concept, arguing that rich, multimodal Comprehensibility is achieved not only by simplifying linguistic
interaction in context is the real driver of language development. forms but also by providing meaningful context and multimodal
Leo Van Lier famously suggested the field move “from input clues in the environment (Ellis, 2008; van Lier, 2004; Tomasello,
to affordance” in understanding language learning (van Lier, 2000, 2019). For example, caretaker speech to children (often cited
2004). Borrowing from Gibson’s (1979) ecological psychology, an by Krashen as evidence for the necessity of simplified input) is
affordance is defined as a “reciprocal relationship between an linguistically simpler and highly contextualized, about the here-
organism and a particular feature of its environment” (van Lier, and-now with shared situational cues (Snow, 1977; Tomasello,
2004, p. 5; Gibson, 1979). The environment has various properties 2019). Research has shown that even in first language acquisition,
that can afford (i.e., enable or invite) specific actions to an active caretaker or “motherese” speech is not uniformly simple in
organism. For example, a tree affords climbing to a child, shade to a structure; parents may use complex sentences, but communication
picnicker, or food to a caterpillar (Gibson, 1979; van Lier, 2004). For remains comprehensible due to gestures, intonation, and
instance, a billboard in a target language is an affordance: it offers immediate context (Snow, 1977; Tomasello, 2019). Critics of the
written language input, visual context, and perhaps cultural insight. Input Hypothesis pointed out that “comprehensible input” does
A conversation with a native speaker is a rich affordance: it offers not necessarily mean “linguistically simplified” input (White,
comprehensible input, but also the chance to negotiate meaning, 1987; Gregg, 1984; Ellis, 2008). Instead, input can be made
to gesture, to observe social cues, etc. Crucially, affordances do not understandable through contextual support. For instance, a teacher
cause learning automatically, just as a tree does not cause a child to could use a normal (even complex) sentence in the target language,
climb, but they invite active engagement. The learner must perceive accompanied by a picture, demonstration, or facial expressions,
and act upon the affordance for learning to occur (van Lier, 2000, so learners can grasp the meaning. The CI hypothesis did not

Frontiers in Psychology 04 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

adequately acknowledge this distinction; it tended to conflate 2019; Lai et al., 2015). Simplified input in a sanitized classroom,
comprehensible input with simplified input. White (1987) and by contrast, might be less effective because it strips away the very
Gregg (1984) noted that an overemphasis on simplification can be affordances (context, interaction, multisensory cues) that learners
misguided. If input is too sterilized and simplified, learners may be need to make form-meaning connections.
deprived of exposure to authentic language features and challenges Moreover, the ecological approach considers the broader
that prompt development. In an ecological approach, by contrast, environment of the learner, including out-of-class exposure,
the goal is not to simplify the language artificially but to amplify the community, technology, and social networks (Lai et al., 2015).
environment and provide a rich semiotic budget that the learner In today’s world, language learners often have access to a vast
can draw on (van Lier, 2004). This could mean diverse sources environment beyond the teacher’s input, from online media and
of input (visual, auditory, textual), interactive opportunities, and social platforms to multicultural peer groups. Ecological research
meaningful tasks that jointly make complex language accessible. on out-of-school learning and technology integration suggests that
Another key concept in ecological theory is agency, as learners the most successful learning happens when learners can navigate
are agents who actively explore and exploit their environment dynamic, unpredictable contexts and treat challenges as affordances
(van Lier, 2004; Larsen-Freeman, 2018). They are not passive for growth (Lai et al., 2015; Godwin-Jones, 2019). For example,
recipients of input; they seek out information, respond to feedback, a student might play a video game in the target language and
and initiate communication. The environment “affords” different initially understand little (far beyond i+1), but the game’s visuals
actions to those looking for them. For example, in a classroom and feedback loop gradually make more language comprehensible.
rich with print (posters, labels) and social interaction, an engaged They acquire some phrases not because someone finely tuned
learner might notice a new written word (an affordance for them to the student’s level, but because the ecological system of
reading), ask a classmate what it means (turning it into an player-goal-feedback in the game encouraged repeated exposure,
affordance for dialogue), and then use it in a project (an affordance hypothesis-testing, and social interaction with other players. This
for production). None of these steps is driven solely by externally kind of learning scenario is invisible in the CI framework, yet
provided input labeled i+1; they are driven by the learner’s increasingly common and powerful in reality.
interaction with a multifaceted environment (van Lier, 2004). Van An ecological perspective critiques the Comprehensible Input
Lier emphasized the “centrality of interaction” in affordances, hypothesis because learning is contextually situated, not a
noting that “affordances consist in the opportunities for interaction mere input-output process (Larsen-Freeman, 2018; van Lier,
that things in the environment possess relative to the sensorimotor 2004). Language emerges from meaningful activity in authentic
capacities of the animal” (van Lier, 2004, p. 92). In language terms, environments, where learners actively use environmental resources
any given utterance or object might afford different things to (including other people) to progress. Krashen’s i+1 concept
different learners. One notices a grammatical form they have almost isolates language from its environment, implying that one can
acquired (i+1 for grammar), another gets the primary meaning feed language at the right level into a learner, and acquisition
(i+0 for comprehension), and another perhaps finds it too difficult will happen. Ecological theory counter-argues that what matters is
(i+ TooHigh) but picks up a non-verbal cue instead. The rigid engaging learners in rich interactions (affordances) that naturally
notion of i+1 as a precise incremental step is incompatible with contain the ingredients for learning, whether or not they are at
this fluid, learner-centered view. Instead of an i+1 input, ecological a perfectly calibrated difficulty level. Indeed, Brown (2000) noted
theory talks about calibrating tasks and contexts to be challenging that input should not be so far beyond learners as to overwhelm
yet manageable, and about learners adapting their attention to them, but also “not so close to their current stage that they are
whatever input or cues are salient and valuable at the moment not challenged at all” (p. 277), advocating a balanced difficulty,
(Larsen-Freeman, 2018; Ellis, 2008). not a precise one-level step. The challenge and support come
Ecological perspectives also highlight the importance of from the environment as a whole, not just linguistic tweaking.
multimodality and authenticity. Language in the real world is Therefore, from an ecological standpoint, the Comprehensible
typically embedded in a multimodal context; gestures, tone of Input hypothesis is both conceptually and practically limited:
voice, facial expressions, physical surroundings, and cultural cues conceptually, it misses the relational, interactive essence of learning;
all provide meaning (Gullberg, 2010; van Lier, 2004; Tomasello, practically, it can lead to dull, contrived pedagogy that fails to
2019). Research on multimodal learning shows that adult learners exploit the whole ecology of learning opportunities available in and
benefit from face-to-face interaction with all these cues, allowing out of the classroom.
them to leverage non-linguistic information to comprehend and
acquire language (Gullberg, 2010; Muntaha et al., 2023). Thus,
comprehension can often be achieved without simplifying language
Personalized, adaptive learning in the
by offering multiple modes of understanding. For instance, learners
might not know a word, but seeing the speaker’s pointing gesture or twenty first century vs. the i+1
the object in question allows them to infer it. This process differs hypothesis
significantly from the CI notion of a teacher intentionally using
a word just slightly above the students’ level. It is more organic The i+1 formula for individuals: from
and learner-driven; the learner encounters richer input, making enlightenment to doubt
it comprehensible through contextual inferencing. Ecological
language learning theory argues that such rich, contextualized Krashen’s i+1 formula, the idea of always aiming input just
experiences enable deeper learning, mirroring how language is one notch above the learner’s current competence, was an elegant
used and learned outside the classroom (van Lier, 2004; Tomasello, pedagogical heuristic in the 1980s. However, as a model for guiding

Frontiers in Psychology 05 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

instruction, i+1 has severe limitations, especially considering Another drawback of i+1 is that it implicitly assumes the
modern personalized learning approaches. One major issue is teacher (or material designer) knows best what the learner is ready
scalability and individualization: in a classroom of diverse learners for next. It is a top-down approach, rooted in syllabus design
(or in self-study scenarios), how can one determine and deliver (like structural grading of textbooks) (Nguyen, 2022b). However,
the optimal i+1 input for everyone? Krashen acknowledged that language learning is not strictly linear, and oftentimes learners
finely tuning input is ideal but offered no concrete mechanism benefit from spaced repetition, recycling, and revisiting simpler
for assessing each learner’s input on the fly (McLaughlin, 1987; material even as they tackle harder input (Brown, 2000; Ellis, 2008).
Ellis, 2008). In practice, teachers use broadly simplified language The i+1 formula offers no room for such pedagogical strategies; it is
or textbook scripts labeled by level, a one-size-fits-all approach only forward-moving. In practice, good teachers often depart from
that may be i+5 for one student and i−1 for another. Measuring strict i+1 progression to review or to address gaps (which Krashen
“current level” is itself problematic, as McLaughlin (1987) pointed might say is an unnecessary “learning” activity, but research
out; language proficiency is not a single number, and learners have says otherwise). Meanwhile, adaptive learning systems naturally
spiky profiles (firm in some areas, weaker in others). Krashen’s incorporate review because they detect when a learner falters on
theory assumes a homogeneous, linear progression (stage 4 to stage something and can represent easier examples or provide hints. They
5, etc.), yet research shows that learners often acquire different also can adjust pacing: some learners might leap from i to i+3
parts of language along different trajectories. For example, a learner if they show rapid mastery (skipping intermediate steps), whereas
might have a high level of listening comprehension but a lower others might need many micro-steps. I+1 as a fixed step size is too
level of grammar accuracy; what is i+1 in this case? The rigidity crude for this nuance.
of the concept fails to accommodate these realities. Brown (2000) Scaling up to large classes or diverse populations further
noted “the indeterminacy of the sequences” in language learning exposes i+1s limitations. In a class of 30 students, each has
and argued that attempts to describe i+1 precisely “will practically a different i. A teacher speaking in one register will inevitably
lead us nowhere” (p. 277). be incomprehensible to some (i+TooHigh) and boringly easy to
others (i+0 or lower). Krashen’s advice here was that if input
is incomprehensible, add more visuals or simplify until it is
understood (Krashen, 1985). Nevertheless, simplifying to the lowest
The challenge of individualization and common denominator can hold back advanced learners, while
scalability using richer language to challenge them might lose the beginners.
This is a perennial dilemma in mixed-ability teaching. Modern
In contrast, twenty first century personalized learning models solutions emphasize differentiation, providing multiple streams of
aim to dynamically adapt to each learner’s needs in a way input or tasks at varying levels, often with the help of technology or
far beyond the static i+1 formula. Contemporary personalized collaborative grouping (Ellis, 2008; Paladines and Ramirez, 2020;
learning theories such as rhizomatic learning (Lian, 2004, 2011; Lin et al., 2023). For example, some teachers use online reading
Lian and Sangarun, 2023) and affordance-based language learning platforms that adjust the text complexity for each student while
(Nguyen, 2022a,b) reject linearity, advocating instead for non- covering the same content, or AI tutoring chatbots that students
hierarchical, multidirectional pathways of language growth, where can interact with at their own level, getting personalized feedback.
learning emerges from complex, dynamic connections among These approaches align with what an adaptive, individualized
experiences, resources, and participants. With the advent of learning path would prescribe, and they have shown improved
educational technology, big data, and artificial intelligence, it has outcomes in both language and other subjects (Godwin-Jones,
become feasible to do what human teachers cannot easily do: 2019; Paladines and Ramirez, 2020; Lin et al., 2023). By contrast,
continuously assess a learner’s performance and adjust content a teacher strictly adhering to comprehensible input might limit
difficulty in real time (Godwin-Jones, 2019; Paladines and Ramirez, all discourse to a narrow band of vocabulary and grammar (the
2020; Lin et al., 2023). These AI-powered adaptive learning systems presumed i+1 for the class), potentially under-stimulating many
highlight how simplistic the i+1 concept now appears. For instance, and still confusing some. Over time, this could result in fossilization
modern language learning platforms (such as Duolingo or Babbel) or boredom. The one-level-above approach does not scale well to
employ algorithms to estimate a user’s proficiency on various diverse needs, whereas intelligent systems or flexible pedagogies can
micro-skills and then select or generate practice items accordingly cope with variability.
(Paladines and Ramirez, 2020; Lin et al., 2023). Recent data- At this point, it is of paramount importance to present
driven personalized learning models integrate big data analytics combined insights from neurolinguistics and ecological psychology
to tailor foreign language learning paths to individuals, using to support a view of language learning, including different
real-time analysis of learner responses to provide practice that is complex stages, rather than just the artificial dichotomies of
neither too easy nor too hard, but optimally challenging for that learning and acquisition, as a process of continual coordination
particular learner at that moment (Xia et al., 2024; Paladines and between the learner’s neural systems and the environment
Ramirez, 2020; Lin et al., 2023). The results have been impressive: where learning takes place. Neurolinguistic research demonstrates
in case studies, classes using such adaptive systems saw significant that experience-dependent plasticity, interaction, feedback, and
improvements in student engagement and comprehension, with embodied activity contribute to the neural changes that sustain
data showing faster mastery of language concepts than traditional language learning. Ecological scholarship shows that learners
methods (Paladines and Ramirez, 2020; Lin et al., 2023). perceive and act on affordances, the opportunities for action and

Frontiers in Psychology 06 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

meaning that emerge within social and material settings. Taken adaptive support, fundamentally different from Krashen’s one-
together, these findings define a neuro-ecological perspective in way input model (Paladines and Ramirez, 2020; Lin et al.,
which language development reflects the gradual alignment of 2023). Whereas comprehensible input, as initially conceived, was
neural adaptation with environmental affordances. In this view, largely unidirectional (teacher talks, student listens), AI tutors
repeated interaction and feedback strengthen neural circuits while respond to the learner in real time. Interactive, responsive systems
learners simultaneously detect and exploit the social, material, inherently personalize the experience, ensuring the learner is
and technological opportunities for action that the environment neither bored nor lost for long (Godwin-Jones, 2019; Paladines
offers. As neural representations become more efficient, learners and Ramirez, 2020; Lin et al., 2023). Early research indicates
can perceive finer distinctions in the environment, which in that such systems improve learning outcomes and greater learner
turn invites richer interaction and further neural change. This autonomy (Godwin-Jones, 2019; Paladines and Ramirez, 2020; Lin
theoretical integration provides the foundation for the empirical et al., 2023). As we incorporate these technologies, the notion of a
data presented in the next section. fixed “i+1” fades away, replaced by the concept of a continuously
calibrated zone for each learner, responsive to their unique progress
and needs.
Empirical support and research In sum, while i+1 was a helpful reminder not to overwhelm
critique in the age of personalization learners with incomprehensible input, it is an oversimplified and
outdated guide for instruction, as there is very limited research
Following the theoretical arguments of the previous part, to validate this hypothesis until now. Also, it fails to account for
the following section seeks to provide empirical studies for the individual differences, the non-linearity of language learning, and
adoption of another perspective on language learning. At the the proven efficacy of adaptive, interactive pedagogies. Modern
point this article is written, Krashen’s i+1 concept has never been personalized learning paradigms, informed by AI and data,
fully validated empirically in the same way that modern adaptive offer a more powerful and precise toolkit to achieve what i+1
learning algorithms are tested through A/B experiments and large- aspired to: keeping learners in that productive, challenged-but-
scale learning analytics. It remains a theoretical proposal derived comprehending zone. As such technologies and approaches gain
from observation rather than a falsifiable model. Studies attempting prominence, clinging to the simplistic i+1 formula becomes
to show that input exactly at i+1 produces superior learning have unnecessary and a hindrance to language teaching innovation.
yielded mixed results. For example, Namaziandost et al. (2019)
compared Iranian EFL learners exposed to i+1, i-1, and mixed-
difficulty input and found that while i+1 groups improved reading Empirical research in three decades:
comprehension and motivation, the advantage over mixed input
was not decisive. Research on form-focused instruction likewise
evidence for a neuro-ecological
shows that a balanced combination of input types often yields the approach
most tremendous gains. Ghorbani and Atai (2012) reported that
explicit form-focused instruction led to more substantial gains in On the other hand, a growing body of empirical research
both explicit and implicit grammar knowledge than purely implicit strongly and comprehensively supports the theoretical critiques
treatment. A large meta-analysis by Spada and Tomita (2010) above in the past decade. Studies in neurolinguistics, ecological
confirmed that explicit instruction produces significant accuracy second language learning, and adaptive learning conclude
benefits, especially for more complex language features, and that that effective language learning involves much more than
a mix of implicit and explicit activities can optimize fluency understanding level-appropriate input. This article highlights key
and confidence. Norris and Ortega (2000) seminal meta-analysis findings from recent research that reinforce the shortcomings
likewise demonstrated robust overall effects of instruction on L2 of the Comprehensible Input hypothesis and point toward more
grammatical accuracy, with explicit approaches outperforming comprehensive frameworks.
input-only exposure. These findings align with contemporary data-
driven approaches in which learning analytics reveal individualized
patterns of difficulty. For instance, an AI tutor can detect that a Neurolinguistic and cognitive evidence
learner retains vocabulary best when new words are introduced
in a moderately challenging context (perhaps beyond i+1) and Multiple recent studies conclude that social interaction,
then recycled in easier contexts. Such adaptive, mixed-difficulty embodiment, and feedback drive measurable neurocognitive
sequencing is now standard in intelligent tutoring systems, whereas change in second language learning. Li and Jeong (2020),
the Input Hypothesis remains static and offers no mechanism for synthesizing research across child language, adult SLA, and
personalized adjustment. cognitive science, show that interactive learning produces
The rise of AI-supported learning environments provides a enhanced behavioral outcomes and increased functional
direct counterpoint to the comprehensible classroom input. In connectivity in brain language networks, whereas non-interactive
an AI language tutor scenario, the system can play multiple listening does not. One experiment they reviewed found that
roles: conversational partner (providing interactive input and adults who engaged in face-to-face interactive sessions, complete
output practice), assessor (gauging the learner’s responses), and with turn-taking and gestures, showed stronger post-test gains in
assistant (offering hints, translations, or simpler rephrasings speaking and comprehension and greater connectivity in language-
as needed). This dynamic interplay enables personalized and related neural networks than peers who spent the same amount

Frontiers in Psychology 07 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

of time on non-interactive listening. Reggin et al. (2023) likewise AI-enhanced CALL platforms, produce neural and behavioral
provide consensus evidence for embodied language learning, benefits that comprehensible input alone cannot match.
demonstrating that sensorimotor engagement and affective
resonance facilitate retention and transfer of new forms. They
review studies in which children who learned action verbs through Ecological and affordance-based research
physical enactment (e.g., learning push by actually pushing)
retained and generalized those words better than children who In language education, there has been an apparent uptick
learned through static definitions, and they note similar embodied in studies adopting ecological frameworks to address modern
effects in adult learners. learning contexts. Lai (2015) and Rokita-Jaskow (2024) provide
Another example is the rapidly expanding body of research an overview of recent research trends from an ecological
that offers convergent neurolinguistic evidence using Verbotonal- perspective, highlighting topics like the spread of technology,
based methods. Experiments with Chinese EFL learners show out-of-school learning, and linguistically diverse classrooms. They
that dichotic listening and filtered/unfiltered input manipulation argue that ecological approaches are well-suited to describing
can induce measurable shifts in cerebral lateralization for current language education, characterized by dynamic change and
speech processing (Lian et al., 2020), a neural marker of unpredictability. For example, one strand of research examines
phonological learning that signals reorganization of auditory informal digital language learning, how learners pick up languages
pathways. Training that combines filtered auditory input with through social media, online gaming, or YouTube (Sockett, 2014;
rhythmic body movement and immediate feedback, as in Yang Lai and Zheng, 2018). These studies often find that learners
et al. (2023), stimulates auditory—motor integration networks acquire a great deal incidentally by navigating these environments,
(superior temporal gyrus, premotor cortex), directly supporting leveraging affordances like subtitles, translations, peer assistance,
the perception—action loops emphasized in embodied cognition. and multimedia cues. In such settings, the input is not graded
Complementary studies demonstrate the role of corrective to i+1 at all; it can range from very easy to far beyond the
feedback: Li and Lian (2022) and Wen et al. (2020) document learner’s level. However, learners acquire phrases and even complex
significant improvements in English intonation and pronunciation constructions because they engage with meaningful content and
when learners receive targeted corrective feedback during training, context. This supports the ecological claim that rich context
while Liu et al. (2021) and Guo et al. (2020) show that web- can compensate for linguistic complexity (comprehensibility from
based autonomous listening systems promote self-regulation and multimodal clues rather than simplified code) and that learner
adaptive engagement. Such tasks recruit executive functions agency drives learning (Lai, 2015).
(working memory, attention shifting, inhibitory control) that An illustrative empirical study on immigrant language-
predict long-term L2 attainment. Complementing these findings, minority students adapting to a new language environment found
Casillas (2020) tracked absolute-beginner English speakers in a 7- that the richness of the ecological environment (availability of
week domestic immersion program and found robust reductions in community interactions, extracurricular activities, etc.) was a
Spanish stop voice-onset-time (VOT) after only 21 days of exclusive strong predictor of language growth (Portes and Hao, 2002).
L2 use. Bayesian growth-curve analyses showed non-linear, Inferentially, students with more varied affordances, e.g., joining
segment-specific learning trajectories, providing direct evidence clubs, interacting with locals, and consuming local media,
that new phonetic categories can emerge rapidly under high- progressed faster, even if much of the input was beyond their level.
input, interaction-rich conditions, precisely the type of brain— The researchers conclude that fostering ecological connectivity
environment coupling predicted by neuro-ecological models. (tying classroom learning to real-life use opportunities) is
These findings align with a large body of interactional- key to accelerating SLA. This kind of evidence reinforces
feedback research. Meta-analyses show that learners who receive that predetermined comprehensible input, while helpful, may
recasts, prompts, or other interactional feedback make significantly be neither sufficient nor as potent as immersing learners in
greater gains in grammatical accuracy than those who do not (Li, vibrant, context-rich environments where they must actively
2010; Lyster and Saito, 2010). Neurolinguistic studies using EEG negotiate meaning.
reveal Error-Related Negativity (ERN) and P600 responses when Another domain of ecological research is classroom interaction
learners encounter corrective feedback, indicating that the brain is patterns. Studies using conversation analysis (e.g., Waring, 2016;
actively updating linguistic representations (Morgan-Short et al., van Lier, 2004) show that when teachers shift from monologic
2012). Therefore, the behavioral improvements reported in the input delivery to dialogic interaction, student language production
Chinese Verbotonal studies imply underlying neural adjustments and uptake increase. Even small changes, like asking more open-
consistent with these ERP signatures. This multi-regional evidence ended questions or allowing students to control topics, create
demonstrates that language development is not a process of passive more affordances for language use. Empirical evidence from these
input absorption but a brain—body—environment coupling studies indicates that classes with high interactional involvement
in which social interaction, multimodal sensory stimulation, see better vocabulary acquisition and pragmatic development than
and adaptive feedback jointly drive neuroplastic change. Far classes where the teacher does most of the talking (even when
from supporting a static Input Hypothesis, these findings the teacher’s talk is “comprehensible”). Learners benefit from
show that optimized auditory signals, sensorimotor engagement, being protagonists in the language environment rather than just
and feedback loops, whether delivered face-to-face or through an audience.

Frontiers in Psychology 08 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

CEFR A2 benchmarks, often exceeding expectations for reading

and listening (Jiang et al., 2024). Other research leveraging billions
of Duolingo activity logs has shown that machine-learning models
can predict which items an individual learner is most likely to
forget and schedule reviews accordingly, enabling more efficient
retention than static word lists (Portnoff et al., 2021; Cui and
Sachan, 2023). Another independent evaluation of Duolingo’s
Spanish for English speakers course by Vesselinov and Grego
(2012) reported that beginners required about 34 h of Duolingo
study to cover the material of one semester of U.S. college Spanish.
In a separate large-scale analysis of learner interaction logs, Settles
and Meeder (2016) estimated an average gain of 8.1 WebCAPE
points per hour of Duolingo study. These findings apply specifically
to beginning learners of Spanish under the testing conditions of
those studies and should not be interpreted as a direct, randomized
comparison with traditional classroom instruction. This kind of
individual optimization goes well beyond i+1. Students might get
i+0 repetition on one concept they struggle with, i+2 stretch on
another concept they grasp quickly, and even 1 on something
forgotten, all determined by their interaction data. Notably, the
FIGURE 1 personalized system often gives users material that is below or at
Neuro-ecological learning umwelt.
their level for review, and occasionally jumps ahead to challenge
them on upcoming material if they are doing well. This mix was
more effective than strictly maintaining an i+1 cadence. Such
Personalized and adaptive learning real-world data at scale makes a compelling case that adaptive
research algorithms can learn more efficiently than static input staging.
In summary, both theoretical developments and empirical
The integration of AI in language education has also yielded research of the last decade comprehensively support a shift
empirical studies that measure learning gains with and without away from the narrow Comprehensible Input paradigm toward
adaptive personalization. One recent review by Xu (2024) a more integrative, neuro-ecological approach. This approach
highlights numerous instances where AI-driven systems improved recognizes that:
learning outcomes. For example, an adaptive vocabulary learning (1) Active engagement and social interaction yield better neural
app that used machine learning through speech recognition to and learning outcomes than passive input alone,
predict which words an individual learner was likely to forget, (2) Embodied, context-rich learning leads to deeper learning
and scheduled reviews accordingly, led to higher long-term than disembodied, simplified input,
retention than a non-adaptive, one-size-fits-all word list approach (3) Learners thrive on affordances in a rich environment, even
(Wilschut et al., 2024). Another study compared a chatbot- if the input is not perfectly leveled, and
facilitated project-based learning (where the chatbot adjusted (4) Personalized, adaptive instruction outperforms one-size-
its language complexity based on the learner’s responses) to a fits-all input delivery by tailoring challenges and support to
static script conversation; learners using the adaptive chatbot individual needs.
showed greater increases in autonomy and transformative changes Together, these findings validate the critiques of Krashen’s
in efficacy (Nguyen and Doan, 2025b). These findings directly hypothesis and provide the foundation for new neuro-ecological
support the argument that individualized pathways outperform paradigms in language education.
the fixed incremental approach. Moreover, research in Intelligent
Tutoring Systems (ITS) for language learning has demonstrated
that students using an ITS that gives immediate, personalized
Conclusion: toward a
learning feedback (essentially providing comprehensible output
in context) learn specific linguistic constructions faster than neuro-ecological framework for
those who only receive additional input exposure to those language education
constructions (Godwin-Jones, 2019). This aligns with the idea
that hypothesis testing and feedback (enabled by adaptive Four decades after Krashen’s Comprehensible Input hypothesis,
systems) accelerate learning, a dynamic absent from pure the field of language education stands at a crossroads shaped
input-based methods. by advances in brain science, a richer appreciation of context
Finally, data analyses from large language-learning platforms and interaction, and the emergence of powerful technological
provide quantitative backing for adaptive learning. For example, tools. In this new era, persisting with the CI hypothesis as the
a Duolingo study analyzing millions of learner interactions found primary compass is neither conceptually sufficient nor empirically
that users who completed the full beginner and intermediate justified. Rather than viewing language learning as a linear, input-
English courses achieved STAMP 4S proficiency scores at or above driven process, current evidence and theory compel us to embrace

Frontiers in Psychology 09 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

frameworks that account for acquisition’s dynamic, interconnected, recalibrate norms, values, and strategies. The Five-Dimensional
and personalized nature. Affordance Framework, including perceptibility, learning valence,
Affordance-based learning foregrounds the opportunities for compositionality, normativity, and intentionality (as suggested in
action in learners’ material, social, and technological environments. Nguyen, 2025a,b), provides a critical analytic lens for studying
In this model, language development arises through learners’ active these next-generation systems, ensuring that future designs remain
engagement with rich, multimodal contexts that provide many grounded in the relational dynamics of brain-body-environment
meaningful cues and possibilities for communication. The teacher’s coupling rather than regressing to static models of input.
role shifts from input transmitter to designer of environments, The task for educators and researchers is to design and
scaffolding, and tasks that encourage learners to perceive, exploit, facilitate learning ecologies that are rich in affordances, networked
and even create new affordances. Besides, connectivism (Siemens, in their resources, and rhizomatic in their flexibility. Learners
2005) situates knowledge within networks of people, artifacts, are not passive consumers of input, but active participants,
and digital resources. Language learners today navigate a complex explorers, and creators, within an ever-evolving landscape of
web of social and technological connections, learning not only linguistic possibility with fine-tuned neurological support from
from teachers and textbooks but from global communities, AI- the technological/AI environment to enhance humans-ecologies
driven platforms, and participatory digital cultures. It is important perception and interactions as demonstrated in Figure 1. It
to note that Connectivism may have overstated its claim in is important to note that these interactions can be highly
suggesting that knowledge can be stored in networks, as digital complex, adaptive, emergent, and sophisticated in reality. By
systems today primarily store and process information rather embracing these contemporary paradigms, language education
than embody knowledge. Nevertheless, the rapid development of becomes more relevant, practical, and empowering, preparing
artificial intelligence suggests a possible future in which machines learners for the complexities of communication in a rapidly
and socio-technical systems may increasingly approximate or even changing world.
participate in processes of “knowing.” Effective language education
now means equipping learners with the skills to construct, curate,
and traverse these networks, becoming autonomous agents in their Author contributions
own learning trajectories. Additionally, rhizomatic learning (Lian,
2004, 2011; Lian and Sangarun, 2023) offers a non-linear, non- QN: Conceptualization, Writing – review & editing, Writing
hierarchical metaphor for knowledge growth. In this view, language – original draft. DD: Writing – original draft, Writing – review
development resembles a rhizome’s sprawling, interconnected & editing.
roots, unpredictable, emergent, and uniquely individualized.
Learning pathways are forged through exploration, adaptation, and
continual reconfiguration, rather than predetermined sequences.
Bringing these models together, underpinned by advances in
Funding
neuroscience, cognitive science, and educational technology, moves
The author declares that no financial support was received for
us decisively beyond the limitations of Comprehensible Input.
the research and/or publication of this article.
Modern language classrooms can harness adaptive platforms, AI
tutors, immersive digital environments, and authentic, multimodal
resources, allowing each learner’s needs, interests, and context to
shape their experience. The neuro-ecological language learning Acknowledgments
models we have implemented so far, such as ChatGPT-mediated
self-regulated learning (Nguyen and Doan, 2025b), idiographic The authors wish to express their deepest gratitude to Professor
affordance-uptake tracking (Nguyen and Doan, 2025a), and Andrew Peter Lian for his invaluable contributions in laying
verbotonal–CALL integrations that optimize auditory input and the conceptual foundations that have profoundly shaped the first
corrective feedback, represent only the earliest generation of author’s intellectual development and the theoretical orientations
brain—body—environment designs. These projects already display of this study.
the key features of a neuro-ecology: learners engage in real-
time prediction and error monitoring, adaptive algorithms fine-
tune input, and affordances emerge from the interaction of Conflict of interest
cognitive, social, and technological forces. Yet these systems
remain bounded by current platforms, institutional curricula, and The authors declare that the research was conducted in the
human facilitation, functioning as carefully curated pilots rather absence of any commercial or financial relationships that could be
than fully self-organizing ecologies. Looking ahead, advances in construed as a potential conflict of interest.
large-scale sensing, generative AI, and decentralized networks are
poised to move beyond today’s adaptive tutors toward emergent
learning ecosystems in which affordances are not merely detected Generative AI statement
but dynamically co-created and recombined across distributed
communities. In such environments, AI agents may act as co- The author(s) declare that Gen AI was used in the creation
learners that negotiate goals, generate novel action possibilities, and of this manuscript. The author(s) utilized artificial intelligence
evolve alongside human participants, while learners continuously (AI) tools (including ChatGPT, Grammarly) for proofreading and

Frontiers in Psychology 10 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

language refinement during the preparation of this manuscript. Publisher’s note

All intellectual content and interpretations remain the sole
responsibility of the author(s). All claims expressed in this article are solely those of the
Any alternative text (alt text) provided alongside figures in authors and do not necessarily represent those of their affiliated
this article has been generated by Frontiers with the support of organizations, or those of the publisher, the editors and the
artificial intelligence and reasonable efforts have been made to reviewers. Any product that may be evaluated in this article, or
ensure accuracy, including review by the authors wherever possible. claim that may be made by its manufacturer, is not guaranteed or
If you identify any issues, please contact us. endorsed by the publisher.

References
Barsalou, L. W. (2008). Grounded cognition. Ann. Rev. Psychol. 59, 617–645. Larsen-Freeman, D. (2018). Looking ahead: future directions in, and future
doi: 10.1146/[Link].59.103006.093639 research into, second language acquisition. Foreign Lang. Ann. 51, 55–72.
doi: 10.1111/flan.12314
Brown, H. D. (2000). Principles of Language Learning and Teaching, 4th Edn. New
York, NY: Pearson Education. Li, P., and Jeong, H. (2020). The social brain of language: grounding
second language learning in social interaction. NPJ Sci. Learn. 5, 1–13.
Cai, X., Lian, A.-P., Puakpong, N., Shi, Y., Chen, H., Zeng, Y., et al. (2021).
doi: 10.1038/s41539-020-0068-7
Optimizing auditory input for foreign language learners through a verbotonal-
based dichotic listening approach. Asian Pacific J. Second Foreign Lang. Educ. 6:14. Li, S. (2010). Li, S. (2010). The effectiveness of corrective feedback in SLA: a
doi: 10.1186/s40862-021-00119-0 meta-analysis. Lang. Learn. 60, 309–365. doi: 10.1111/j.1467-9922.2010.00561.x
Casillas, J. V. (2020). The longitudinal development of fine-phonetic detail: Li, Z., and Lian, A. (2022). Achieving self-imitation for english intonation
stop production in a domestic immersion program. Lang. Learn. 70, 768–806. learning: the role of corrective feedback. Chinese J. Appl. Ling. 45, 106–125.
doi: 10.1111/lang.12392 doi: 10.1515/CJAL-2022-0108
Chemero, A. (2009). Radical Embodied Cognitive Science. Cambridge, MA: MIT Lian, A.-P., Cai, X., Chen, H., Ou, J., and Zheng, W. (2020). Cerebral lateralization
Press. induced by dichotic listening to filtered and unfiltered stimuli: optimizing auditory
input for foreign language learners. J. Crit. Rev. 7:4608.
Cui, Z., and Sachan, M. (2023). “Adaptive and personalized exercise generation for
online language learning,” in Proceedings of the 61st Annual Meeting of the Association Lian, A.-P., and Sangarun, P. (2023). “Rhizomatic learning systems and precision
for Computational Linguistics, Vol. 1, 10184–10198. doi: 10.18653/v1/[Link]-long.567 language education: a perfect match,” in New Directions in Rhizomatic Learning: from
Poststructural Thinking to Nomadic Pedagogy, ed. M. S. Khine (Routledge), 119–141.
Ellis, R. (2008). The Study of Second Language Acquisition, 2nd Edn. Oxford: Oxford
doi: 10.4324/9781003376378-11
University Press.
Lian, A. P. (2004). “Technology-enhanced language-learning environments: a
Gass, S. M., and Selinker, L. (2008). Second Language Acquisition: An Introductory
rhizomatic approach,” in Computer-Assisted Language Learning: Concepts, Contexts
Course, 3rd Edn. New York, NY: Routledge. doi: 10.4324/9780203932841
and Practices, ed. J.-B. Son (iUniverse), 1–20.
Ghorbani, M., and Atai, M. R. (2012). Evaluating the effectiveness of explicit and
Lian, A. P. (2011). Reflections on language-learning in the 21st century: the rhizome
implicit form-focused instruction on explicit and implicit knowledge of EFL learners.
at work. Rangsit J. Arts Sci. 1, 5–16.
Issues Lang. Teach. 1, 23–50.
Lightbown, P. M. (2019). Perfecting practice. Mod. Lang. J. 103, 703–712.
Gibson, J. J. (1979). The Ecological Approach to Visual Perception. Boston, MA: HM.
doi: 10.1111/modl.12588
Godwin-Jones, R. (2019). Riding the digital wilds: learner autonomy and informal
Lin, C. C., Huang, A. Y., and Lu, O. H. (2023). Artificial intelligence in intelligent
language learning. Lang. Learn. Technol. 23, 8–25. doi: 10.64152/10125/44667
tutoring systems toward sustainable education: a systematic review. Smart Learn.
Gregg, K. R. (1984). Krashen’s monitor and Occam’s razor. Appl. Ling. 5, 79–100. Environ. 10:41. doi: 10.1186/s40561-023-00260-y
doi: 10.1093/applin/5.2.79
Lisman, J. E., Grace, A. A., and Duzel, E. (2011). A neoHebbian framework for
Gullberg, M. (2010). “Gestures and second language acquisition,” in Theories in episodic memory; role of dopamine-dependent late LTP. Trends Neurosci. 34, 536–547.
second language acquisition, eds. R. Van Patten and J. Williams (Routledge), 204–223. doi: 10.1016/[Link].2011.07.006
Guo, L., Lian, A., and Puakpong, N. (2020). “Development of a web-based Liu, K., Lian, A.-P., and Yodkamlue, B. (2021). “Integrating information literacy
autonomous listening comprehension for Chinese high school students,” in IVCST training in an English-speaking course in the Chinese context,” in Proceedings
Conference Proceedings (Nakhon Ratchasima), 605–612. of the 17th International Conference of the Asia Association of Computer-Assisted
Language Learning AsiaCALL 2021 (Ho Chi Minh City: Atlantis Press), 29–39.
Indefrey, P., and Levelt, W. J. M. (2004). The spatial and temporal signatures of word
doi: 10.2991/assehr.k.210226.004
production components. Cognition 92, 101–144. doi: 10.1016/[Link].2002.06.001
Long, M. (1996). “The role of the linguistic environment in second
Jiang, X., Peters, R., Plonsky, L., and Pajak, B. (2024). The effectiveness of duolingo
language acquisition,” in Handbook of Second Language Acquisition, eds. W.
english courses in developing reading and listening proficiency. CALICO J. 41,
C. Ritchie, and T. K. Bhatia (Cambridge, MA: Academic Press), 413-468.
249–272. doi: 10.1558/cj.26704
doi: 10.1016/B978-012589042-7/50015-3
Krashen, S. D. (1982). Principles and Practice in Second Language
Lyster, R., and Saito, K. (2010). Oral feedback in classroom SLA: a meta-analysis.
Acquisition. Oxford: Pergamon.
Stud. SLA 32, 265–302. doi: 10.1017/S0272263109990520
Krashen, S. D. (1985). The Input Hypothesis: Issues and implications. New York, NY:
Macedonia, M., and Von Kriegstein, K. (2012). Gestures enhance foreign language
Longman.
learning. Lang. Learn. 62, 271–290. doi: 10.5964/bioling.8931
Krashen, S. D. (1996). The case for narrow listening. System 24, 97–100.
McLaughlin, B. (1978). The monitor model: some methodological considerations.
doi: 10.1016/0346-251X(95)00054-N
Lang. Learn. 28, 309–332. doi: 10.1111/j.1467-1770.1978.tb00137.x
Krashen, S. D., and Terrell, T. D. (1983). The Natural Approach: Language
McLaughlin, B. (1987). Theories of Second-Language Learning. London: Edward
Acquisition in the Classroom. San Francisco, CA: The Alemany Press.
Arnold.
Lai, C. (2015). Perceiving and traversing in-class and out-of-class learning: accounts
Meyer, L., Obleser, J., and Friederici, A. D. (2013). Left parietal alpha
from foreign language learners in Hong Kong. Innov. Lang. Learn. Teach. 9, 265–284.
enhancement during working memory-intensive sentence processing. Cortex 56,
doi: 10.1080/17501229.2014.918982
142–149. doi: 10.1016/[Link].2012.03.006
Lai, C., and Zheng, D. (2018). Self-directed use of mobile devices for language
Morgan-Short, K., Steinhauer, K., Sanz, C., and Ullman, M. T. (2012). Explicit
learning beyond the classroom. ReCALL 30, 299–318. doi: 10.1017/S0958344017000258
and implicit second language training differentially affect the achievement of native-
Lai, C., Zhu, W., and Gong, G. (2015). Understanding the quality of out-of-class like brain activation patterns. J. Cogn. Neurosci. 24, 933–947. doi: 10.1162/jocn_a_00
English learning. TESOL Q. 49, 278–308. doi: 10.1002/tesq.171 119

Frontiers in Psychology 11 [Link]

Nguyen and Doan 10.3389/fpsyg.2025.1636777

Muntaha, M., Chen, J., and Dobinson, T. (2023). Exploring students’ experiences Rietveld, E., and Kiverstein, J. (2014). A rich landscape of affordances. Environ.
of using multimodal CMC tasks for English communication. Educ. Technol. Soc. Psychol. 26, 325–352. doi: 10.1080/10407413.2014.958035
26, 69–83.
Rokita-Jaskow, J. (2024). An ecological perspective on the challenges in
Namaziandost, E., Nasri, M., and Ziafar, M. (2019). Comparing the impacts language education: focus on agency and affordances. Angl. Wratisl. 62, 13–26.
of various inputs (i + 1 and i – 1) on pre-intermediate EFL learners’ reading doi: 10.19195/0301-7966.62.1.2
comprehension and reading motivation: the case of Ahvazi learners. Asian Pacific J.
Settles, B., and Meeder, B. (2016). “A trainable spaced repetition model for language
Second Foreign Lang. Educ. 4:13. doi: 10.1186/s40862-019-0079-1
learning,” Proceedings of the 54th Annual Meeting of the Association for Computational
Nguyen, Q. N. (2022a). “Affordance-based learning trajectories: a deviation Linguistics, eds. K. Erk, and N. A. Smith (Association for Computational Linguistics),
from the teachers-should-know-all model in language education,” in Proceedings 1st 1, 1848–1858. doi: 10.18653/v1/P16-1174
International Conference On Education 2022 (Muen Wai), 132:131.
Siemens, G. (2005). Connectivism: a learning theory for the digital age. Int. J.
Nguyen, Q. N. (2022b). Postlesson affordance-based reflective discussion in ELT Instruct. Technol. Dist. Learn. 2, 3–10. Available online at: [Link]
classes. TESOL J. 13:e677. doi: 10.1002/tesj.677 Jan_05/[Link]
Nguyen, Q. N. (2025a). Dogme 2.0: Da.y Ho.c. . . à ra thế! (ISBN: 978-632-604-709-7). Snow, C. E. (1977). “Mothers’ speech research: from input to interaction,” in Talking
Alphabooks: Thegioipublisher. to Children: Language Input and Acquisition, eds. C. E. Snow and C. A. Ferguson
(Cambridge University Press), 31–49.
Nguyen, Q. N. (2025b). Affordance Theory in Language Education: A
Multidimensional Framework. TESL-EJ. Advance online publication. Sockett, G. (2014). The Online Informal Learning of English. London: Palgrave
Macmillan. doi: 10.1057/9781137414885
Nguyen, Q. N., and Doan, D. T. H. (2025a). Towards a Five-Dimensional
Affordance-Based Framework for Social Injustice Research in Language Education. Spada, N., and Tomita, Y. (2010). Interactions between type of instruction
Available online at: [Link] or [Link] and type of language feature: a meta-analysis. Lang. Learn. 60, 263–308.
ssrn.5311420 (Accessed March, 26, 2025). doi: 10.1111/j.1467-9922.2010.00562.x
Nguyen, Q. N., and Doan, D. T. H. (2025b). ChatGPT’s effects Tomasello, M. (2019). Becoming Human: A Theory of Ontogeny. Cambridge, MA:
on EFL learners’ autonomy and self-efficacy in self-regulated learning Harvard University Press. doi: 10.4159/9780674988651
contexts. JALT CALL J. 21:102968. doi: 10.29140/jaltcall.v21n2. Ullman, M. T. (2016). “The declarative/procedural model: a neurobiological
102968 model of language learning, knowledge, and use,” in The Handbook of
Norris, J. M., and Ortega, L. (2000). Effectiveness of L2 instruction: a Language Emergence, eds. B. MacWhinney and W. O’Grady (Wiley), 259–280.
research synthesis and quantitative meta-analysis. Lang. Learn. 50, 417–528. doi: 10.1016/B978-0-12-407794-2.00076-6
doi: 10.1111/0023-8333.00136 van Lier, L. (2000). “From input to affordance: social-interactive learning from an
Paladines, J., and Ramirez, J. (2020). A systematic literature review of intelligent ecological perspective,” in Sociocultural Theory and Second Language Learning, ed. J. P.
tutoring systems with dialogue in natural language. IEEE Access 8, 164246–164267. Lantolf (Oxford: Oxford University Press), 245–259.
doi: 10.1109/ACCESS.2020.3021383 van Lier, L. (2004). The Ecology and Semiotics of Language Learning: a Sociocultural
Ping, R. M., Goldin-Meadow, S., and Beilock, S. L. (2014). Understanding Perspective. New York, NY: Springer. doi: 10.1007/1-4020-7912-5
gesture: is the listener’s motor system involved? J. Exp. Psychol. Gen. 143, 195–204. Vesselinov, R., and Grego, J. (2012). Duolingo Effectiveness Study. New York, NY:
doi: 10.1037/a0032246 City University of New York.
Portes, A., and Hao, L. (2002). The price of uniformity: language, Waring, H. Z. (2016). Theorizing Pedagogical Interaction: Insights from Conversation
family and personality adjustment in the immigrant second generation. Analysis. New York, NY: Routledge. doi: 10.4324/9781315751351
Ethn. Racial Stud. 25, 889–912. doi: 10.1080/01419870220000
09368 Wen, F., Lian, A.-P., and Sangarun, P. (2020). Determination of corrective
optimals for Chinese university learners of English. Speech 37, 3–28.
Portnoff, R. S., Settles, B., Hagiwara, M., and Tetreault, J. (2021). “Methods for doi: 10.22210/govor.2020.37.01
language learning assessment at scale: Duolingo case study,” in Proceedings of the
14th International Conference on Educational Data Mining, 688–693. Available online White, L. (1987). Against comprehensible input: the input hypothesis
at: [Link] and the development of second-language competence. Appl. Ling. 8, 95–110.
doi: 10.1093/applin/8.2.95
Price, C. J. (2012). A review and synthesis of the first 20 years of PET and fMRI
studies of heard speech, spoken language and reading. Neuroimage 62, 816–847. Wilschut, T., Sense, F., and van Rijn, H. (2024). Speaking to remember: model-based
doi: 10.1016/[Link].2012.04.062 adaptive vocabulary learning using automatic speech recognition. Comput. Speech
Lang. 84:101578. doi: 10.1016/[Link].2023.101578
Pulvermüller, F. (2018). Neural reuse of action perception circuits
for language, concepts and communication. Prog. Neurobiol. 160, 1–44. Xia, Y., Shin, S. Y., and Shin, K. S. (2024). Designing personalized learning paths for
doi: 10.1016/[Link].2017.07.001 foreign language acquisition using big data: theoretical and empirical analysis. Appl.
Sci. 14:9506. doi: 10.3390/app14209506
Reggin, L. D., Franco, L. E. G., Horchak, O. V., Labrecque, D., Lana, N., Rio, L.,
et al. (2023). Consensus paper: Situated and embodied language acquisition. J. Cogn. Xu, Z. (2024). AI in education: Enhancing learning experiences and student
6:63. doi: 10.5334/joc.308 outcomes. Appl. Comput. Eng. 51, 104–111.
Richards, J. C., and Rodgers, T. S. (2014). Approaches and Methods Yang, Y., Wannaruk, A., and Andrew-Peter, L. (2023). Improving the English-
in Language Teaching, 3rd Edn. Cambridge: Cambridge University Press. speaking skills of Chinese primary EFL learners with a verbotonal approach. J. Curr.
doi: 10.1017/9781009024532 Sci. Technol. 7, 141–156. doi: 10.14456/RJAS.2017.5

Frontiers in Psychology 12 [Link]