Written by This article examines the ongoing debate in second language acquisition (SLA) research regarding the relative effectiveness of explicit grammar instruction versus implicit language acquisition. Drawing on nine primary sources spanning from foundational theoretical work to recent empirical studies, this review synthesizes evidence across theoretical, psychometric, pedagogical, and methodological dimensions. The sources reviewed include Krashen’s (1982) acquisition-learning distinction, Ellis’s (2005) psychometric framework for measuring implicit and explicit knowledge, Ferreira’s (2017) domain-specific analysis, Bowles and Montrul’s (2008) study of computer-assisted explicit instruction, Akbana’s (2018) form-focused instructional design, Alahmad’s (2019) input enhancement study, Rohollahzadeh Ebadi et al.’s (2014) investigation of metalinguistic feedback, Stratton’s (2023) learner preference study, and Shin’s (2010) methodological critique of Norris and Ortega’s (2000) landmark meta-analysis. Taken together, the evidence suggests that while implicit input provides necessary conditions for language acquisition, explicit instruction—particularly form-focused approaches—offers measurable advantages across most linguistic domains, with learner variables, structural complexity, and instructional context serving as critical moderating factors. Methodological limitations in the broader research base, however, warrant caution in drawing sweeping pedagogical conclusions.
Keywords: explicit instruction, implicit acquisition, form-focused instruction, second language acquisition, metalinguistic knowledge, noticing hypothesis, interface debate
Few questions have generated more sustained controversy in second language acquisition (SLA) research than whether learners benefit from explicit instruction in grammar or whether language is best acquired through implicit exposure to comprehensible input. This debate has profound pedagogical implications: it shapes curriculum design, classroom methodology, teacher training, and learner expectations worldwide. Despite decades of investigation, no settled consensus has emerged, partly because the question is not unitary. The optimal instructional approach may vary by learner age and proficiency, the linguistic domain targeted, the task type used to measure outcomes, the instructional context, and the theoretical framework adopted.
The literature reviewed here spans foundational theoretical positions, psychometric tools for measuring knowledge types, domain-specific analyses, empirical classroom studies, and critical methodological commentary. Together, these sources map the contours of an ongoing and unresolved conversation, one that has grown increasingly nuanced as researchers have refined their constructs, measurement instruments, and analytical methods. This review is organized thematically, moving from the foundational theoretical framework that established the terms of the debate, through the psychometric and empirical evidence bearing on it, to the pedagogical implications and methodological critiques that complicate any straightforward reading of the research.
Theoretical Foundations: The Acquisition-Learning Distinction
The theoretical architecture underlying this debate was most influentially articulated by Krashen (1982), whose Input Hypothesis framework proposed a strict separation between acquisition and learning in the development of second language competence. For Krashen, acquisition is a subconscious process analogous to first language development, driven by exposure to comprehensible input pitched slightly above the learner’s current level (i+1). Learning, by contrast, is the conscious internalization of grammatical rules, which Krashen argued functions only as a Monitor—an editor capable of checking output after the fact, provided the learner has time, attends to form, and knows the relevant rule. Crucially, Krashen’s No Interface Hypothesis holds that learned knowledge cannot become acquired knowledge; the two systems are entirely separate, and classroom grammar instruction therefore contributes nothing to the underlying competence that drives spontaneous, fluent language use.
Krashen’s framework further holds that affective variables—motivation, self-confidence, and anxiety—operate as a filter on the uptake of comprehensible input, such that an unfavorable affective environment can block acquisition regardless of input quality. This Affective Filter Hypothesis has direct pedagogical implications, suggesting that meaning-focused, low-anxiety classrooms are preferable to form-focused environments that might raise learner anxiety. Krashen also proposed a Natural Order Hypothesis, arguing that grammatical structures are acquired in a predictable developmental sequence that instruction cannot substantially accelerate.
Though enormously influential, Krashen’s framework has been substantially challenged by subsequent theoretical and empirical work. Most relevant is Schmidt’s (1990) Noticing Hypothesis, which holds that conscious attention to form is a necessary condition for input to become intake. Where Krashen treats subconscious acquisition as sufficient and instruction as irrelevant, Schmidt argues that noticing—at minimum awareness of a form in the input—is what converts input into data the learner’s system can process. This theoretical reorientation provided the foundation for form-focused instructional approaches and for a generation of empirical work seeking to demonstrate that explicit attention to form enhances acquisition.
The Implicit-Explicit Knowledge Distinction: Psychometric Evidence
A precondition for evaluating instructional effects is the availability of valid instruments to distinguish implicit from explicit grammatical knowledge. Ellis (2005) addressed this gap directly, administering a battery of five tests to 91 native speakers and 106 non-native speakers of English across 17 grammatical structures. The five instruments included an Oral Imitation Test, an Oral Narration Test, and a Timed Grammaticality Judgment Test (GJT) as measures of implicit knowledge—tasks designed to invoke automatized, spontaneous processing—alongside an Untimed GJT and a Metalinguistic Knowledge Test as measures of explicit knowledge, which allow learners time to consciously apply rules.
Principal component factor analysis of the test scores yielded two clearly distinct factors corresponding to the implicit and explicit measures, providing strong psychometric evidence that these represent separable constructs rather than a single underlying competence. A particularly important finding was the hybrid nature of GJTs: timed versions reflect implicit knowledge, while untimed versions reflect explicit knowledge. This distinction is methodologically consequential, as Ellis (2005) noted that most prior instructional research—including Norris and Ortega’s (2000) widely cited meta-analysis—relied predominantly on controlled or metalinguistic outcome measures that favor explicit knowledge, potentially inflating estimates of instructional effectiveness.
Rohollahzadeh Ebadi, Mohd Saad, and Abedalaziz (2014) built directly on Ellis’s (2005) psychometric framework in a study that employed all four of the adapted instruments—the Elicited Oral Imitation Test, Timed GJT, Untimed GJT, and Metalinguistic Knowledge Test—to assess the effects of explicit form-focused instruction (FFI) on both knowledge types in 91 adult ESL learners in a Malaysian context. Their pretest-treatment-posttest design with a control group found that the experimental group, which received metalinguistic corrective feedback during communicative tasks, significantly outperformed controls on both implicit and explicit knowledge measures, with large effect sizes (η² = .74 and .78, respectively). The authors interpreted these results as consistent with the weak interface position, drawing on Ellis’s (2006) theoretical framework: explicit corrective feedback, by triggering noticing and directing attention to form-meaning mappings, may create conditions under which explicitly learned knowledge can eventually be proceduralized into implicit competence. While the absence of a delayed posttest limits conclusions about long-term retention, and the very large effect sizes invite scrutiny regarding possible demand characteristics, this study is notable for being among the relatively few to measure instructional effects on implicit as well as explicit knowledge.
Form-Focused Instruction as a Middle Ground
Between the poles of purely implicit acquisition and overt grammar-translation instruction, a substantial body of research has investigated form-focused instruction (FFI)—approaches that embed attention to form within meaningful communicative activity, aiming to reconcile the communicative orientation of modern pedagogy with the demonstrated benefits of explicit attention to language. This tradition draws heavily on Long’s (1991) distinction between Focus on FormS (FonFS), which refers to pre-specified grammar syllabi, and Focus on Form (FonF), which refers to reactive attention to form arising incidentally during meaning-focused activity, as well as on Schmidt’s (1990) Noticing Hypothesis.
Alahmad (2019) investigated a specific FonF technique—typographical input enhancement—in a study of 60 intermediate EFL learners at an Iranian language institute. Learners in the experimental condition read texts in which target conditional structures were made visually salient through bolding, underlining, or italics, while the control group read identical texts without enhancement. Post-test results showed that the enhanced input group significantly outperformed controls on both multiple-choice and cloze measures, leading Alahmad to conclude that the typographical enhancement successfully triggered noticing of the target forms without interrupting the flow of meaning-focused reading. This study positions FonF input enhancement as an efficient instructional middle ground, consistent with Krashen’s emphasis on meaningful input while incorporating Schmidt’s insistence that form must be noticed. The contrast between the experimental group’s superior performance and the control group’s more modest gains offers empirical pushback against a pure input hypothesis, suggesting that flooding learners with meaningful input—without guiding their attention—is less effective than designed enhancement.
Akbana (2018) examined the effects of a broader FFI instructional design in a Turkish EFL context, employing a 12-week mixed-methods study with 40 university preparatory students. The experimental FFI condition incorporated structured input, input flooding, noticing tasks, interpretation activities, and consciousness-raising techniques applied across six grammatical domains. Quantitative findings indicated a modest but positive FFI advantage in overall language proficiency and a more substantial effect on L2 English writing skills, though the experimental group’s superiority in syntactic complexity was not as pronounced as hypothesized, and delayed post-test differences were not statistically significant. The qualitative data, however, told a richer story: learners in the FFI condition reported increased autonomy, more deliberate engagement with linguistic forms, and a sense of discovery in their language learning. Akbana (2018) concluded that FFI’s cognitive and affective benefits may be substantial even when quantitative performance differences are modest, and recommended that instructors vary FFI input types and encourage learners to discover grammar patterns rather than receive them passively.
Domain-Specific Evidence: Where Explicit Instruction Matters Most
One resolution to the debate between implicit and explicit instruction is that the optimal approach is not uniform across all aspects of language but varies systematically by linguistic domain. Ferreira (2017) conducted a literature review of post-2005 SLA research to map this variation, drawing on Ullman’s Declarative/Procedural (DP) model—which posits that lexical knowledge depends on declarative memory while grammatical knowledge draws on procedural memory—and Paradis’s neurolinguistic distinction between implicit linguistic competence and metalinguistic knowledge. Ferreira’s analysis found that syntax tends to rely primarily on implicit knowledge, particularly for rapid acquisition of complex structures, though explicit instruction can assist when L2 structures diverge substantially from L1. Vocabulary learning, by contrast, showed a consistent advantage for explicit instruction, with wordlists outperforming purely incidental reading for long-term retention. Pragmatic competence demonstrated a strong advantage for explicit instruction, particularly where meta-pragmatic information about cultural or social context was required. Phonetics showed a high degree of benefit from explicit awareness, especially for sounds without L1 equivalents that learners otherwise tend to assimilate to familiar categories. Morphological learning favored explicit instruction for structures unique to the L2, while input flooding was effective for structures shared with the L1. Semantic knowledge showed an advantage for explicit instruction, with evidence that explicit teaching can simultaneously build both implicit and explicit semantic knowledge.
Bowles and Montrul (2008) provide targeted empirical evidence consistent with Ferreira’s (2017) analysis, focusing on the acquisition of the Spanish personal a—a preposition used before human direct objects that has no English equivalent and is notoriously difficult for English-speaking learners to acquire through exposure alone. Using a pretest-posttest-delayed posttest design with 51 intermediate-level university students, Bowles and Montrul found that the explicit instruction group, which received formal rule presentation and computerized practice with immediate right-wrong feedback, significantly outperformed a no-instruction control group on both a recognition task and a controlled production task. Gains were maintained at the delayed posttest one week later, suggesting at least short-term durability. The authors interpret these results as evidence that for low-salience L2 features—those that are infrequent, semantically redundant, or cross-linguistically novel—comprehensible input alone is unlikely to generate sufficient noticing for acquisition to proceed, and some form of explicit attention is necessary. Bowles and Montrul (2008) also contribute a practical demonstration that explicit instruction need not be teacher-fronted: computer-delivered rule presentation and feedback proved effective, with implications for technology-mediated language learning environments.
Learner Perspectives and the Affective Dimension
While most research in this area focuses on learning outcomes measured by performance tests, Stratton (2023) took a complementary approach by investigating how learners themselves perceive and prefer different instructional modalities. In a mixed-methods study with 64 third-semester university learners of German in a North American context, Stratton administered an exit survey and conducted semi-structured interviews following a semester-long pre/post/delayed-posttest study in which learners had been assigned to explicit or implicit instruction in either pronunciation or vocabulary.
The learner preference data aligned strongly with the objective outcomes. Among pronunciation learners, 94% of those in the explicit condition reported that instruction improved their pronunciation, compared with 54% in the implicit condition—a statistically significant difference. In the vocabulary domain, 100% of explicit learners reported improvement versus 70% of implicit learners. Qualitative data revealed that learners in implicit conditions frequently experienced recasts and other indirect feedback as imperceptible, frustrating, and opaque. By contrast, explicit instruction was described as anxiety-reducing, because it gave learners clear rules to apply, and as appropriate to the higher-education context, with several students articulating an expectation of declarative rule-based instruction as part of university-level study. Notably, 80% of implicit vocabulary learners reported improvement despite showing no significant objective gains in the quantitative component of Stratton’s research, a finding that illustrates a potentially misleading disconnect between self-perceived and actual learning under implicit conditions.
Stratton’s (2023) findings bear directly on Krashen’s (1982) Affective Filter Hypothesis, which anticipates that anxiety-inducing instruction will impede acquisition. In Stratton’s study, it was the implicit, communicative condition—consistent with Communicative Language Teaching ideology—that learners found more stressful and less productive, while the explicit condition was experienced as less anxiety-inducing. This inversion of the expected affective pattern suggests that the relationship between instruction type, anxiety, and learning is more context-dependent than Krashen’s framework implies, and that learner expectations, educational culture, and the transparency of instructional feedback all mediate affective responses to instruction. The study’s limitations include small sample sizes, a potential confound from different instructional languages across conditions, and the fact that implicit learners had more prior German exposure—all of which Stratton acknowledges.
Methodological Considerations and the Limits of Meta-Analytic Evidence
The broader evidence base on instructional effectiveness in SLA has been shaped, and arguably distorted, by methodological limitations documented with particular rigor by Shin (2010) in a critical review of Norris and Ortega’s (2000) landmark meta-analysis. Norris and Ortega synthesized experimental and quasi-experimental studies to conclude that explicit instruction outperforms implicit instruction with a mean effect size of d = 0.96, that Focus on Form and Focus on FormS are equally effective, and that instructional effects are durable. These conclusions have been enormously influential, and the meta-analytic methods Norris and Ortega employed became a template for subsequent research synthesis in applied linguistics.
Shin (2010) identifies three categories of methodological problems that undermine confidence in these conclusions. First, data collection problems: 47% of the 49 primary study samples lacked random assignment, threatening internal validity; 88.5% had sample sizes of N ≤ 30, undermining the utility of what randomization existed; instrument validity was questionable in multiple studies due to very few test items or recycled items across pre- and posttest; publication bias was not addressed; and an inclusive approach with no quality filters likely inflated average effect sizes. Second, coding problems: experimental and quasi-experimental designs were pooled without distinction; learner characteristics were not systematically differentiated; the FonF/FonFS coding was oversimplified; and moderating variables including proficiency, age, context, and retention interval were largely ignored. Third, statistical problems: Cohen’s d was computed as an unweighted average across studies with very different sample sizes, ignoring the well-documented tendency for small-sample studies to produce larger and more variable effect sizes, and the resulting effect size of d = 0.96 is therefore likely an overestimate. Shin recommends the adoption of Hedges’ adjusted g, which corrects for small-sample upward bias, or Hierarchical Linear Modeling (HLM) to properly account for nested data structures and moderating variables, alongside a best evidence synthesis approach that combines quantitative and qualitative methods and applies stricter quality criteria.
Shin’s (2010) critique does not argue that explicit instruction is ineffective; it argues that the magnitude of the reported advantage is uncertain, and that the conditions under which explicit instruction is most effective—learner age, proficiency, target structure, task type—have not been adequately specified by the existing meta-analytic record. This methodological argument resonates throughout the literature reviewed here. Rohollahzadeh Ebadi et al.’s (2014) very large effect sizes invite similar scrutiny. Ellis’s (2005) observation that most instructional research uses explicit knowledge measures alone—thereby potentially confounding the effects of instruction with test characteristics—highlights a pervasive validity threat. And Stratton’s (2023) finding that learner-perceived improvement can diverge substantially from objective performance underscores the importance of triangulating across multiple outcome measures.
Synthesis and Pedagogical Implications
The literature reviewed here does not yield a simple conclusion in favor of one instructional mode over the other, but it does permit several qualified generalizations. First, Krashen’s (1982) No Interface Hypothesis, in its strong form, is not well supported by the evidence: explicit instruction does produce measurable gains across a wide range of linguistic domains, and there is theoretical and some empirical support for the weak interface position, which holds that explicit knowledge can, under appropriate conditions, contribute to the development of implicit competence (Ellis, 2006; Rohollahzadeh Ebadi et al., 2014). Second, the value of explicit instruction appears to be domain-dependent: it is particularly important for low-salience, cross-linguistically novel features (Bowles & Montrul, 2008; Ferreira, 2017), for pragmatic and phonological competence, and for morphological structures that diverge from the L1. Third, form-focused instruction—which embeds attention to form within meaningful activity rather than treating grammar as an end in itself—offers a pedagogically viable synthesis that is supported by evidence of both objective and affective benefits (Akbana, 2018; Alahmad, 2019). Fourth, learner variables matter: proficiency level, L1 background, educational context, and learner expectations about what instruction should look like all moderate instructional effects (Stratton, 2023; Ferreira, 2017).
This distinction between explicit and implicit knowledge is not merely theoretical. In my own classroom experience teaching B1–C1 professionals in Colombia, I observed a student who consistently scored the highest marks on grammar assessments — demonstrating near-perfect explicit knowledge of English structures — yet was functionally unable to produce fluent speech in real-time conversation. At work he spent his days exclusively among Spanish-speaking coworkers. His colleagues in the office, who regularly interacted with native English speakers, had lower test scores but could read, write, and converse in English with relative ease. The difference was not knowledge — it was acquisition through meaningful, repeated exposure. He had learned the rules. They had internalized the language. This is precisely the gap Ellis (2005) operationalizes: explicit knowledge measurable under controlled conditions, and implicit knowledge that only emerges through authentic, meaning-driven contact with the target language.
Fifth, and critically, the methodological quality of the research base constrains the confidence with which any of these conclusions can be held. The reliance on explicit knowledge measures in most instructional studies, documented by Ellis (2005) and reinforced by Rohollahzadeh Ebadi et al. (2014), means that measured instructional effects may reflect the relative accessibility of explicit knowledge in controlled-task conditions rather than genuine acquisition of implicit, automatized competence. The effect size inflation documented by Shin (2010) means that quantitative estimates of instructional advantage should be treated with caution pending more rigorous meta-analytic synthesis. And the predominance of short-term designs limits conclusions about the durability of gains.
For practitioners, the most defensible pedagogical implication is not a wholesale endorsement of either approach but an argument for principled eclecticism: instruction that uses meaningful, comprehensible input as its foundation while strategically incorporating explicit attention to forms that are difficult to acquire implicitly, especially those that are infrequent, cross-linguistically novel, or pragmatically nuanced. This recommendation aligns with the FonF tradition (Long, 1991) as operationalized in studies like Alahmad (2019) and Akbana (2018), and is consistent with Ferreira’s (2017) call to match instructional modality to linguistic domain. It also takes seriously Stratton’s (2023) finding that learner expectations and affective responses to instruction are themselves pedagogically consequential and should inform instructional design.
Conclusion
The question of whether explicit grammar instruction or implicit acquisition produces better second language outcomes resists a definitive answer, but the weight of the evidence reviewed here suggests that explicit instruction, particularly in its form-focused variants, offers advantages that a purely input-based approach cannot replicate. These advantages are most pronounced for linguistically complex, cross-linguistically novel, or pragmatically subtle features; for learners who have metalinguistic expectations aligned with explicit instruction; and for outcomes measured over shorter time frames with tasks that tap explicit knowledge. Implicit input remains a necessary, and in some domains sufficient, condition for acquisition. The interface between these systems—how and whether explicit knowledge contributes to implicit competence over time—remains theoretically contested and empirically underspecified.
Methodological progress, including the adoption of validated measures for both knowledge types (Ellis, 2005), more rigorous meta-analytic procedures (Shin, 2010), and designs that track both short- and long-term effects across learner populations, is needed before the field can claim to have produced cumulative, trustworthy evidence on these questions. In the meantime, the emerging consensus favors integrating explicit and implicit approaches within a form-focused instructional framework that is responsive to learner variables, linguistic domain, and instructional context—a conclusion modest enough to reflect genuine uncertainty, and specific enough to be of practical use.
References
Akbana, Y. E. (2018). L2 English grammar through form focused instructional design. Unpublished master’s thesis.
Alahmad, M. (2019). The role of form-focused instruction on foreign language learners. Unpublished manuscript, Iranian EFL context.
Bowles, M., & Montrul, S. (2008). The role of explicit instruction in the L2 acquisition of the a-personal. University of Illinois at Urbana-Champaign.
Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition, 27(2), 141–172.
Ferreira, A. L. (2017). Implications of implicit and explicit instruction for second language acquisition. Unpublished review paper.
Krashen, S. D. (1982). Principles and practice in second language acquisition. Pergamon Press.
Long, M. H. (1991). Focus on form: A design feature in language teaching methodology. In K. de Bot, R. Ginsberg, & C. Kramsch (Eds.), Foreign language research in cross-cultural perspective (pp. 39–52). John Benjamins.
Norris, J. M., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis and quantitative meta-analysis. Language Learning, 50(3), 417–528.
Rohollahzadeh Ebadi, M., Mohd Saad, M. R., & Abedalaziz, N. (2014). Explicit form focus instruction: The effects on implicit and explicit knowledge of ESL learners. Malaysian Online Journal of Educational Science, 2(4), 25–34.
Schmidt, R. W. (1990). The role of consciousness in second language learning. Applied Linguistics, 11(2), 129–158.
Shin, H. W. (2010). Another look at Norris and Ortega (2000). Teachers College, Columbia University Working Papers in TESOL & Applied Linguistics, 10(1), 15–38.
Stratton, J. M. (2023). Implicit and explicit instruction in the second language classroom: A study of learner preferences in higher education. Die Unterrichtspraxis/Teaching German, 56, 103–117.
