Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Second-Language Acquisition and Cross-Linguistic Transfer

Abstract. This fundamental research monograph investigates the acquisition of additional linguistic systems and the influence of previously acquired languages on emerging representations, processing routines, and proficiency. A historical and conceptual reconstruction of interlanguage and language transfer is integrated with evidence from cross-language priming, speech perception, syntactic parsing, lexical access, longitudinal development, age-related attainment, educational interventions, and recent quantitative syntheses. It compares competing accounts of cue learning, transfer, representation, processing efficiency, and maturational change; evaluates research designs and the limits of causal interpretation; and distinguishes short-term bilingual language control from developmental second-language acquisition. The synthesis considers learners across ages, languages, literacy backgrounds, and learning contexts, and examines experimental methods, instructional consequences, and human–artificial learning environments. It is a critical narrative synthesis of independently published scholarship, not a preregistered systematic review, original experimental report, or claim of a novel psychological theory.

Keywords: second-language acquisition; cross-linguistic transfer; cross-linguistic influence; interlanguage; language learning; second-language speech perception; L2 grammar; bilingual lexical access; age of acquisition; critical period hypothesis; syntactic priming; sentence parsing; language transfer; second-language instruction; implicit and explicit learning; language proficiency; psycholinguistic methods; Artificial Era

Scientific Scope and Problems of Acquisition

Second-language acquisition concerns the development of linguistic knowledge and use after a learner has already acquired one or more languages. The additional language may be encountered in childhood, adulthood, schooling, migration, work, or sustained multilingual community life. Research therefore studies changes in perception, word recognition, grammatical representation, sentence interpretation, speaking, and communicative participation across time. A psychologically adequate account connects these outcomes to exposure history, prior linguistic experience, practice, motivation, developmental opportunity, and the demands of particular tasks. Odlin’s synthesis of language transfer locates earlier linguistic knowledge among the causes of additional-language performance, while Selinker’s interlanguage proposal treats the learner’s developing system as an object of scientific investigation in its own right. Odlin (1989) Selinker (1972)

The central question is how an existing linguistic system changes the formation and deployment of a further one. Transfer may accelerate a familiar mapping, prompt a competing interpretation, change which acoustic cues attract attention, or alter the probability of selecting a structure in production. Such outcomes need not occur together. Accurate grammaticality judgments can coexist with slow parsing; conversational fluency can develop before certain morphosyntactic contrasts become reliable; and a learner may recognize a sound contrast more consistently than they produce it. Research must therefore distinguish the state of learning from a particular observed performance, and a short-term experimental effect from durable development. Psycholinguistic models of shared lexical representation and online parsing supply methods for making these distinctions observable. Kroll et al. (2015) Clahsen & Felser (2006)

Additional-language learning is distributed across timescales. On a trial, a learner may use an available cognate or a recent syntactic frame; across weeks, repeated input changes cue expectations; across years, accumulated participation changes habitual comprehension and production routines. Interpreting a laboratory priming effect as evidence for enduring acquisition requires further observation, just as interpreting adult performance as evidence for an immutable age limit requires information about exposure and testing. The synthesis in this dossier examines conceptual histories, original experimental paradigms, longitudinal evidence, quantitative reviews, and educational interventions. It treats age, linguistic similarity, learning context, and methodological design as interacting dimensions of the research object rather than isolated explanations. Hartsuiker et al. (2004) Hartshorne et al. (2018) Norris & Ortega (2000)

What Counts as Cross-Linguistic Transfer

Cross-linguistic transfer, also called cross-linguistic influence, describes a difference in learning or language use attributable to experience with another language. Inference ordinarily depends on comparison: learners with different linguistic backgrounds, structures with different cross-language correspondences, or the same learners under conditions that alter cross-language activation. A similarity in two languages can facilitate generalization when the mapping is relevant, while a superficially similar pattern can increase misanalysis when its distribution or meaning differs. Odlin’s account extends transfer beyond straightforward substitution of a native-language rule, including avoidance, differential learning rates, and consequences of perception and meaning. Transfer is thus a family of mechanisms requiring evidence at the level of the specific contrast studied. Odlin (1989) Ellis (2006)

The direction of influence is also an empirical matter. A first language can shape the second through established categories and entrenched cue associations, while sustained acquisition and use of the second can reshape lexical access, sentence interpretation, and even first-language choices. Cross-language activation in a bilingual lexical-recognition task establishes simultaneous availability of competing candidates; it does not by itself identify when or how a developmental change occurred. Dijkstra and van Heuven’s account of bilingual visual word recognition and Kroll and colleagues’ review connect joint activation to architectures of lexical processing. Findings of online interaction must be combined with learner histories and repeated measures before they explain acquisition trajectories. Dijkstra & van Heuven (2002) Kroll et al. (2015)

Several neighboring phenomena should be distinguished analytically. Code-switching is a contextually organized choice between linguistic resources; cross-language priming is temporary facilitation following a preceding item or structure; transfer in acquisition concerns an effect of earlier experience on subsequent learning or performance. These phenomena can share representational resources without being interchangeable. Kellerman’s analysis of perceived transferability shows that learners are selective about which properties they project across languages, underscoring the role of perceived similarity and conceptual interpretation. A convincing transfer claim identifies the predictor, dependent measure, exposure conditions, plausible alternatives, and whether the result generalizes beyond a single stimulus set or group. Kellerman (1979) Hartsuiker et al. (2004) Odlin (1989)

Historical Formation of the Research Program

Early studies of additional-language learning often classified differences between first- and second-language forms as possible sources of learner difficulty. Contrastive analysis motivated valuable comparisons of phonological, lexical, and grammatical systems, yet a simple inventory of structural differences could not predict which errors would occur, how long they would persist, or why apparently similar structures were sometimes difficult. The later development of transfer research shifted attention toward learner perception, developmental timing, processing mechanisms, and the observed distribution of forms in input. Odlin’s historical synthesis documents this widening explanatory field and the necessity of taking both facilitative and interfering consequences seriously. Odlin (1989) Kellerman (1979)

Selinker’s 1972 account of interlanguage changed the unit of analysis by identifying a learner-language system with its own regularities and developmental trajectory. This made persistent patterns, variable performance, and restructuring legitimate objects of explanation. An interlanguage analysis asks what generalization the learner has constructed, how that generalization changes, and what evidence supports an inferred internal representation. A learner’s non-target form can express a principled hypothesis about meaning or grammatical distribution, rather than an isolated failure to reproduce a target expression. Modern computational and cue-based approaches extend this program by examining how attention, frequency, and competition help determine which generalizations are acquired. Selinker (1972) Ellis (2006) MacWhinney (2005)

The history of second-language theory also includes disagreements about conscious awareness, age, and the nature of grammatical representations. Schmidt’s treatment of noticing helped formalize a question about which features of input enter learning, while research on age of first substantial exposure connected developmental constraints with decades of accumulated experience. Later experiments added time-sensitive comprehension measures and structural priming to traditional learner judgments and elicited production. No one methodological tradition resolves all of these questions: observational age comparisons constrain possible developmental theories, and controlled manipulations help distinguish mechanisms at shorter timescales. The history is best understood as a progressive differentiation of constructs and research designs. Schmidt (1990) Johnson & Newport (1989) Dussias & Sagarra (2007)

Interlanguage and Developmental Pathways

An interlanguage is a developing linguistic organization whose patterns can be internally systematic even when they depart from conventions in the community language. Its study requires more than counting errors against a fixed standard. Researchers examine the learner’s available contrasts, the range of contexts in which an expression appears, alternations under processing pressure, and shifts following exposure or feedback. A grammatical form can become accurate first in practiced constructions and only later generalize across lexical items, modalities, and discourse contexts. This variability is theoretically important because it permits competing accounts of representation, attention, and practice to yield distinct predictions. Selinker (1972) MacWhinney (2005)

Transfer can influence the starting configuration of an interlanguage without specifying its ultimate organization. Learners may recruit a familiar semantic distinction, interpret grammatical cues through previously learned reliability patterns, or avoid a construction whose cross-language mapping feels uncertain. The Competition Model treats interpretation as the outcome of cues whose relative validity and availability differ between linguistic environments. Ellis extends this family of explanations through selective attention, cue competition, blocking, and salience. These models predict that transfer effects should vary with the structures presented, the input distribution, and the learner’s history of attending to a contrast; they invite experiments that change one of these conditions rather than merely labeling an error as transfer. MacWhinney (2005) Ellis (2006)

Development can also involve reorganization rather than simple accumulation. Learners may move from item-based expressions to broader patterns, revise overgeneralizations, and discover constructions whose meanings overlap imperfectly with those of their first language. Because training on one construction can produce both improvement and temporary interference elsewhere, a complete evaluation should track several dependent variables. Longitudinal sampling, novel-item generalization, and repeated sentence-processing measures are especially useful for following whether a pattern becomes productive and automatic. Classroom and interaction studies complement laboratory observations by showing which inputs and opportunities are encountered in actual learning settings. Mackey (1999) Norris & Ortega (2000) Hartsuiker et al. (2004)

Research Designs and Inference

Claims about second-language acquisition are produced through several kinds of evidence. Cross-sectional proficiency studies compare learners at different observed levels; longitudinal designs follow the same individuals; controlled training studies manipulate the amount or character of input; and experimental psycholinguistics measures recognition, judgments, eye movements, reaction times, and structural choices under defined conditions. Each design answers a different causal question. A cross-sectional relationship between age of arrival and grammatical performance cannot isolate maturation from education, social opportunity, or years of use. A randomized learning manipulation provides a clearer estimate of a training effect within its tasks while leaving long-term retention and everyday generalization to be tested separately. Johnson & Newport (1989) Hartshorne et al. (2018) Norris & Ortega (2000)

Operationalization determines what an outcome means. A grammaticality judgment reflects decisions about the acceptability of selected forms, often with time for reflection; reading times index online processing costs but are affected by lexical predictability and task strategy; structural priming reflects a tendency to reuse a recently encountered configuration. Accent ratings combine pronunciation patterns with listeners’ perceptual expectations, and vocabulary tests may assess recognition, recall, or production. A proposal about general linguistic competence must explain how these measures converge or diverge. Hopp’s work on advanced learners and Clahsen and Felser’s processing analysis illustrate the importance of asking whether end-state accuracy and real-time mechanisms tell the same story. Hopp (2010) Clahsen & Felser (2006)

Evidence synthesis requires attention to sampling, language pairs, item-level variance, attrition, outcome comparability, and potential publication bias. Reviews and meta-analyses integrate studies under explicit inclusion and coding rules, but their pooled results remain conditional on the studies and constructs represented. Norris and Ortega’s instructional synthesis and Spada and Tomita’s comparison of grammatical instruction provide aggregate evidence while also motivating closer consideration of instructional treatments and linguistic targets. Publication-bias demonstrations in the neighboring bilingual-cognition literature reinforce the importance of searching for unreported or null findings. This dossier is a critical narrative synthesis that interprets available studies, not an exhaustive registered systematic review. Norris & Ortega (2000) Spada & Tomita (2010) de Bruin et al. (2015)

Shared and Distinct Lexical Representations

Acquiring an additional language requires learning how unfamiliar word forms relate to concepts, grammatical classes, and situations of use. Lexical knowledge has at least three separable dimensions: recognizing a form, retrieving its meaning, and selecting the form during production. An apparently familiar translation equivalent may be accessed efficiently in recognition while remaining difficult to produce under time pressure. Bilingual word-recognition accounts therefore model activation across two lexicons together with decisions appropriate to the current task. Dijkstra and van Heuven’s BIA+ framework specifies how lexical candidates and decision processes can interact while retaining a distinction between language-independent recognition and task-dependent response demands. Dijkstra & van Heuven (2002) Kroll et al. (2015)

A key theoretical question is whether words from different languages are stored in separate compartments or draw on partially shared representations. Findings that orthographic and semantic competitors from a currently irrelevant language influence processing provide evidence against a rigidly selective access assumption in some bilingual tasks. Their magnitude varies with scripts, spelling overlap, frequency, proficiency, and the experimental demand to make a language-specific decision. Shared activation should not be equated with identical lexical representations: phonological detail, semantic range, pragmatic use, and collocational information can differ across translation equivalents. Developmental research must establish which dimensions of lexical knowledge change with exposure and which observed asymmetries reflect retrieval practice. Dijkstra & van Heuven (2002) Kroll & Stewart (1994)

Lexical learning supplies a clear example of productive transfer and of its limits. Cognates can offer reusable form–meaning correspondences, and prior conceptual knowledge reduces some burdens of semantic learning, yet a cognate can invite inappropriate pronunciation or an overextended sense. The direction and strength of these effects depend on similarity as experienced by the learner, not solely on formal linguistic classification. Kroll and colleagues’ review emphasizes the interactive consequences of bilingual experience for comprehension, production, and cognitive control. A strong empirical analysis compares several language pairs and uses tasks that separate recognition from naming, rather than assuming that a speed advantage demonstrates enduring acquisition. Kroll et al. (2015) Odlin (1989) Kellerman (1979)

Asymmetric Translation and Conceptual Access

Kroll and Stewart’s translation and picture-naming experiments helped establish an influential account of asymmetric links between lexical forms and conceptual representations during bilingual development. In the Revised Hierarchical Model, an additional-language word may initially receive substantial support from its corresponding first-language form, while direct conceptual mediation becomes more effective with greater proficiency. The account predicts different patterns for translation in each direction and for tasks requiring a conceptual decision. Such predictions can be evaluated through speed, accuracy, category interference, and the relative effects of distractors rather than relying only on learners’ descriptions of how they translate. Kroll & Stewart (1994)

This asymmetry is a model of processing tendencies under specified developmental conditions, rather than a universal two-stage sequence that every bilingual must follow. Translation can be shaped by the frequency of particular word pairs, the learner’s context of use, the number of languages known, and whether the two expressions differ in sense or register. An automatic response to an isolated word also differs from comprehension within a meaningful sentence, where conceptual and discourse constraints are available. Bilingual lexical research therefore benefits from convergent measures that include picture naming, semantic categorization, translation recognition, and naturalistic language use. The BIA+ framework further demonstrates why access and final response decisions need separate explanation. Kroll & Stewart (1994) Dijkstra & van Heuven (2002)

The practical implication is that word learning involves multiple representational connections. Activities that repeatedly match isolated forms to first-language translations may improve one retrieval route, while sentence interpretation, picture-based retrieval, conversation, and varied contextual examples create additional opportunities to link forms to meanings and uses. These proposed differences require experimental tests with equivalent exposure time and delayed outcome measures. Vocabulary meta-analyses provide evidence about the effectiveness of intentional activities, but their results must be mapped to the exact learning conditions and tests involved. The strongest synthesis combines models of bilingual memory with demonstrations that different forms of practice have distinguishable learning consequences. Webb et al. (2020) Kroll & Stewart (1994) Norris & Ortega (2000)

Phonetic Categories and Perceptual Assimilation

Second-language speech perception poses a fundamental category-learning problem. A listener arrives with phonetic categories whose boundaries and cue weights were shaped by earlier linguistic input. A new speech contrast may be heard as two instances of a familiar category or as a distinction that is more clearly separated by prior experience. Best and Tyler’s discussion of nonnative and second-language speech perception emphasizes relations between existing categories, perceived similarity, and learning context. The resulting predictions concern patterns of discrimination and identification across specific contrasts rather than the generic difficulty of hearing foreign speech. Best & Tyler (2007)

Perceptual assimilation and category learning also distinguish early auditory sensitivity from language-specific interpretation. Learners may perceive acoustic differences in a nonlinguistic discrimination task yet fail to use those differences consistently for word recognition. Conversely, successful identification in a training set may reflect memorized acoustic distributions without generalization to new speakers or conversational environments. A suitable experiment measures categorization on continua, discrimination, and performance on untrained items; adding multiple talkers helps evaluate robustness against incidental voice characteristics. The research question is whether experience produces a reusable phonological distinction and an appropriate allocation of attention to its diagnostic features. Best & Tyler (2007) Ellis (2006)

Cross-language differences in perceived phonetic similarity are specific to the sounds and phonological contexts studied. Even within a single language pair, some contrasts are discriminated readily and others resist learning. The same learner can change perceptual strategies with literacy, prior multilingual experience, and the distribution of exposure. Models emphasizing selective attention predict that highly familiar but weakly informative cues can dominate initially, whereas repeated diagnostic evidence can shift perceptual weights. This makes transfer a testable hypothesis about the allocation of learning resources: the researcher should identify the cues, specify the prior system, and predict conditions under which the influence will diminish or reverse. Ellis (2006) Flege & Bohn (2021) Best & Tyler (2007)

Speech Learning and Input Quality

The Speech Learning Model and its revised formulation explain additional-language pronunciation and perception in relation to the speaker’s evolving phonetic system. Flege and Bohn’s revised model treats cross-language category relations and the quality and quantity of phonetic input as central to how new distinctions develop. A learner may establish a distinct category when the new sound is sufficiently discriminable from established categories, whereas perceived similarity can promote category equivalence and slow differentiation. The framework links perception with production while recognizing that articulatory routines, auditory feedback, and the experience of individual speakers contribute to observed performance. Flege & Bohn (2021)

Training results offer a way to evaluate these mechanisms beyond retrospective comparisons. High-variability phonetic training exposes learners to multiple examples of a contrast, often from different speakers, with the goal of encouraging attention to invariant category information. Reviews of pronunciation interventions and analyses of high-variability training report benefits under studied conditions, while highlighting uneven transfer from identification exercises to production and spontaneous speech. The relationship between perception and articulation is therefore an empirical outcome, not a correspondence to be inferred from an improved laboratory score. Assessment should sample trained and untrained tokens, generalize across speakers, and include delayed retention when possible. Saito (2012) Uchihara et al. (2024)

Input quality includes the distribution of phonetic tokens, signal clarity, contextual predictability, opportunities for corrective information, and the social range of speakers encountered. Frequency alone does not determine learning when repeated examples are uninformative about the contrast, and brief training may yield fragile gains without continued practice. Recent quantitative syntheses of phonetic training are informative about average effects but encompass diverse learner populations, contrasts, procedures, and dependent measures. Their implications for classroom instruction should therefore be expressed in terms of replicable conditions: what auditory distinctions were trained, what production measures improved, and how much generalization was observed. Yao et al. (2025) Uchihara et al. (2024) Flege & Bohn (2021)

Bidirectional Influence and Dynamic Systems

Cross-linguistic influence becomes especially clear when language experience is treated as cumulative and bidirectional. Using an additional language can increase the availability of its lexical forms and alter which language is favored in a particular context, while sustained reduced use of another can affect ease of retrieval. The weaker-links account interprets some bilingual word-retrieval differences through the frequency of use of each language’s lexical representations. Gollan and colleagues’ investigation of frequency effects helps separate this account from explanations that invoke language switching or inhibitory control in every case. The same observed naming cost can have multiple sources. Gollan et al. (2008)

Adaptive control accounts examine how communicative demands regulate activation and selection when bilinguals encounter single-language, dual-language, and switching contexts. Green and Abutalebi propose that language-control requirements adapt to the interactional environment. This concerns the management of learned languages during use and should be connected cautiously to the longer-term formation of new linguistic knowledge. When the opportunity to practice particular forms changes, shifts in retrieval speed or language dominance can occur without a complete restructuring of grammar. Conversely, lasting acquisition may appear in transfer to novel constructions after temporary activation effects have disappeared. Green & Abutalebi (2013) Gollan et al. (2008)

A dynamic account therefore measures change at multiple levels rather than assuming a single language-proficiency variable explains all outcomes. Phonetic categories can reorganize alongside lexical access; grammatical interpretation can remain sensitive to previous-language experience despite substantial fluency; and different home, school, or workplace contexts can favor different resources. Repeated measures within individuals are especially valuable because they distinguish developmental trajectories from differences between cohorts. They also clarify the scope of claims about bidirectional influence: a shift in response latency is evidence about retrieval, whereas change in generalization to unfamiliar forms bears more directly on a learned representation. Kroll et al. (2015) Dussias & Sagarra (2007) Green & Abutalebi (2013)

Grammar Transfer as Cue Reweighting

Grammatical transfer is often described as the projection of an established syntactic rule onto an additional language. Cue-based models give a more differentiated explanation by asking which observable properties a listener uses to determine thematic roles, reference, grammatical relations, and sentence structure. Word order, agreement, case marking, and semantic plausibility vary in predictive reliability across languages. A learner accustomed to a strongly informative word-order pattern may initially underweight a morphological marker that becomes more diagnostic in the new language. MacWhinney’s Competition Model makes these relations experimentally tractable by comparing responses to sentences containing conflicting or converging cues. MacWhinney (2005)

Experience-dependent cue weighting creates opportunities for facilitation as well as interference. Previously acquired knowledge can provide a useful representation when grammatical functions align, while repeated reliance on an inappropriate cue can delay acquisition of a contrast. Ellis’s account of selective attention and cue competition predicts that learning will depend on whether input makes a relevant cue noticeable, informative, and available across enough contexts to overcome earlier expectations. This explanation can be tested with training materials whose cue validity is manipulated while the lexical and semantic content is controlled. Such experiments can distinguish a failure to represent a category from a failure to attend to evidence about its distribution. Ellis (2006) MacWhinney (2005)

Grammatical similarity is not uniform across constructions. Two languages may share a basic word order yet differ in article use, relative-clause attachment, aspectual marking, or the relation between form and discourse prominence. A theory that predicts transfer from a language-level typological label alone is too coarse to account for this variation. Detailed structural comparison should be combined with proficiency-sensitive measures and exposure histories. When a learner’s performance changes after encountering reliable counterevidence, the result bears on the capacity for cue revision; when the pattern persists across contexts, alternative explanations include processing cost, insufficient informative input, and the strength of previously learned mappings. Odlin (1989) Kellerman (1979) Ellis (2006)

Online L2 Parsing and Structural Attainment

Online sentence parsing investigates how an additional-language user constructs interpretations while a sentence unfolds. Self-paced reading, eye movements, timed judgments, and comprehension responses can distinguish early commitments from later revision. Clahsen and Felser’s Shallow Structure Hypothesis proposed that some second-language processing differs from native-language parsing in the relative use of detailed syntactic structure and other information. This is an explanatory hypothesis whose applicability may vary by language pair, construction, proficiency, and task. Evidence of altered reading times alone does not establish that a learner lacks a grammatical representation; performance also depends on memory, lexical access, and processing efficiency. Clahsen & Felser (2006)

Experience can affect processing preferences when speakers learn to interpret structures under a different statistical distribution. Dussias and Sagarra’s work on Spanish–English bilinguals links language exposure with relative-clause attachment preferences. The importance of such experiments lies in connecting linguistic input history to a real-time interpretation measure, rather than inferring transfer merely from a final acceptability judgment. Alternative explanations must be evaluated, including differences in lexical materials, discourse context, participant population, and the effects of the testing language. Observing a systematic shift with exposure strengthens a learning account, although broader causal claims require longitudinal or controlled manipulation of the relevant input. Dussias & Sagarra (2007)

Advanced attainment presents a separate challenge. Highly proficient learners may reach levels of performance that approximate those of a comparison group on some constructions, while specific timed tasks continue to reveal processing differences. Hopp’s studies of inflection and grammatical processing support evaluating attainment at the level of particular representations and real-time measures. Native-speaker comparison groups themselves contain individual variation; a single cutoff for native-likeness risks obscuring this distribution. The proper question is whether different populations rely on comparable information and show similar patterns under multiple processing demands, with sensitivity to the reliability and difficulty of each task. Hopp (2010) Clahsen & Felser (2006)

Cross-Language Syntactic Priming

Syntactic priming occurs when recently processing or producing a structure increases its subsequent use or accessibility. In bilingual experiments, a prime in one language can influence production in another, providing evidence relevant to the sharing of structural representations across languages. Hartsuiker, Pickering, and Veltkamp used Spanish–English priming paradigms to investigate whether structural information is wholly language-specific. The theoretical inference depends on the controlled relationship between prime and target, on how structural alternatives are measured, and on whether conceptual or lexical repetition can explain the effect. Priming is a valuable probe of accessible representations because it manipulates a temporally local cause. Hartsuiker et al. (2004)

Shared priming does not imply that the entire grammar of two languages is represented identically. Structural overlap is partial; constructions can differ in word order, argument realization, morphological requirements, and discourse functions. Moreover, priming may involve several mechanisms, including residual activation, implicit adaptation to experienced distributions, and strategic alignment with a conversational partner. Experiments that vary verb repetition, structural compatibility, prime language, and retention interval can discriminate some of these explanations. A transfer account becomes stronger when priming effects predict learning or generalization over later sessions, rather than being confined to immediate repetition of familiar alternatives. Hartsuiker et al. (2004) MacWhinney (2005)

Structural priming also creates a bridge between psycholinguistic theory and educational practice. If exposure to a construction increases its subsequent use, repeated communicative opportunities may help stabilize a new pattern. Yet the effect of an immediate prime should not be described as instructional acquisition unless learners later use the construction without the original cue and across novel lexical contexts. Interaction studies provide complementary evidence by showing changes in question formation and related linguistic outcomes following communicative engagement. The research design must preserve a distinction among elicited repetition, task-specific practice, and transferable knowledge that endures after the episode. Mackey (1999) Hartsuiker et al. (2004)

Morphosyntactic Accuracy and Processing Efficiency

Morphosyntax reveals how final-response accuracy and online efficiency can diverge. Agreement endings, tense or aspect forms, and case morphology often occur with low perceptual prominence while carrying substantial grammatical information. Learners can sometimes choose an appropriate form in a reflective test but fail to use it rapidly during fluent comprehension or speech. This discrepancy motivates task combinations that independently measure grammatical sensitivity, processing speed, and production under cognitive load. Hopp’s research on advanced second-language inflection illustrates that some learners attain strong performance across demanding tasks, challenging explanations that equate all late learning with a permanently separate grammatical competence. Hopp (2010)

The distribution of evidence matters particularly for inflection. A frequent form may be learned as a lexical sequence without a productive rule, whereas learning that generalizes to unfamiliar stems suggests more abstract knowledge. At the same time, low-frequency forms may elicit failures of retrieval rather than failures of grammatical understanding. Experimental designs can compare trained and novel items, vary time pressure, and test sensitivity to agreement violations across several lexical classes. The Competition Model interprets some outcomes through reliability and competition among grammatical cues; a processing approach asks how quickly the needed information becomes available. These accounts can generate complementary predictions when carefully operationalized. MacWhinney (2005) Clahsen & Felser (2006)

Accuracy on a limited set of sentence materials is insufficient to establish the absence or presence of a universal age-dependent grammatical ceiling. To compare groups fairly, researchers need broadly sampled constructions, tests that avoid ceiling and floor effects, and appropriate adjustments for proficiency, literacy, and exposure. When learner differences arise, the analysis should report uncertainty and describe how alternative sampling decisions affect the result. The same principle applies to claims that explicit rule teaching changes implicit processing: improvement on a rule-focused test is credible evidence for performance on that test, while automatic generalization requires independent measures. Hopp (2010) Spada & Tomita (2010) Hartshorne et al. (2018)

Competing Accounts of Linguistic Representation

Several research programs explain transfer using different units of analysis. Representational accounts ask which lexical, syntactic, and phonological structures are shared across languages. Cue-learning accounts ask how attention and predictive reliability determine interpretation. Processing accounts examine time-dependent retrieval and integration; interaction accounts relate communicative feedback to learning opportunities; maturational accounts consider how developmental change constrains the system. These explanations are potentially complementary, but their claims must be matched to evidence. A theory of a processing difference cannot substitute for an account of long-term change, and an observed longitudinal change alone does not identify the microscopic learning mechanism. Kroll et al. (2015) Ellis (2006) Clahsen & Felser (2006)

Competing predictions become clearer when studies manipulate the proposed causes. Cross-language priming tests representational overlap under controlled structural conditions; cue-conflict tasks assess which signals guide interpretation; perceptual training tests whether distributional exposure reorganizes categories; and age comparisons assess patterns of eventual attainment. When two theories predict the same average accuracy, response times, transfer to novel stimuli, retention, or training sensitivity may discriminate them. An evidence hierarchy should attend to directness of manipulation and replication, while keeping the inferential limits of each paradigm visible. Hartsuiker et al. (2004) MacWhinney (2005) Flege & Bohn (2021) Johnson & Newport (1989)

Theoretical comparison also requires a consistent definition of what is being learned. A representation can become more detailed without becoming faster to retrieve; a strategic response can improve without a change in underlying knowledge; and a grammatical generalization can become productive before it is reliably automated. Where available, convergence among behavioral, linguistic, longitudinal, and neural measures is more persuasive than any one signature. The strongest explanatory models identify the expected boundary conditions and the observations that could falsify their claims. In this synthesis, transfer is treated as a set of measurable relations between prior language experience and subsequent outcomes rather than as a single mechanism guaranteed to operate identically across linguistic domains. Odlin (1989) Kroll et al. (2015) Norris & Ortega (2000)

Age of Onset and the Critical-Period Debate

The relationship between age of first substantial second-language exposure and later proficiency is one of the most influential and contested topics in acquisition research. Johnson and Newport’s investigation of immigrants’ English grammatical judgments found an association between age of arrival and performance within the population and tasks they studied. Such an association is informative about developmental patterns, yet age at arrival also correlates with social and educational circumstances. A critical-period hypothesis adds a particular causal claim: maturational changes create constraints on certain learning capacities. Distinguishing that hypothesis from effects of cumulative experience requires models that consider the temporal shape of learning and the distribution of opportunity. Johnson & Newport (1989)

Hartshorne, Tenenbaum, and Pinker analyzed a very large sample of English learners and used formal modeling to investigate age-related change in grammar-learning ability. Their analysis distinguished the age at which learning capacity might change from the age at which a learner begins using English. This distinction matters because eventual proficiency reflects both initial learning conditions and the time available for development. The impressive survey scale increases precision for some modeled patterns, while recruitment, retrospective self-reports, online testing, cohort composition, and assumptions of the fitted model remain relevant to interpretation. A model-based critical-period estimate should be described as conditional on its measurement and structural assumptions, not as a direct physiological observation. Hartshorne et al. (2018)

Age effects also differ across linguistic outcomes. Pronunciation, lexical retrieval, comprehension, morphological judgments, and pragmatic participation draw on partially different learning and processing demands. Age of onset cannot be treated as interchangeable with duration of exposure, years of formal instruction, or age at testing. A rigorous comparison therefore includes those measures, represents nonlinear trajectories, and evaluates whether conclusions survive alternative definitions of proficiency. The scientific question is which components of acquisition change with development, the causal mechanisms of such changes, and the degree to which sustained learning opportunities permit later attainment. Abrahamsson & Hyltenstam (2009) Flege & Bohn (2021) Johnson & Newport (1989)

Ultimate Attainment and Relative Nativeness

Studies of ultimate attainment ask how far highly experienced additional-language users develop in relation to distributions of performance among people who learned the language from early childhood. The outcome depends strongly on what counts as attainment. Accent ratings, detailed phonetic analyses, timed grammatical processing, lexical decisions, and spontaneous communication can yield different comparisons. Abrahamsson and Hyltenstam distinguish listener perceptions of native-likeness from closer linguistic scrutiny, illustrating why a socially plausible accent cannot be treated as a comprehensive estimate of linguistic knowledge. Comparisons require the properties of the task and comparison sample to be stated explicitly. Abrahamsson & Hyltenstam (2009)

Native speakers themselves vary in education, exposure, dialect, literacy, memory, and speed of responding. Consequently, a strict criterion requiring an additional-language learner to fall at a particular point on every measure can turn ordinary population variability into an apparent developmental impossibility. Hopp’s investigation of advanced inflectional performance demonstrates why direct processing evidence and multiple measures are important when discussing the upper range of attainment. Individual cases may establish that a particular degree of performance is possible under particular conditions, even when they do not determine its population frequency. Scientific conclusions should differentiate these logically distinct claims. Hopp (2010) Abrahamsson & Hyltenstam (2009)

Attainment is also shaped by the range of situations in which a language is actually used. A learner may develop sophisticated academic reading while rarely using informal phonology, or acquire fluent workplace interaction with limited exposure to written morphology. These patterns reflect experience, not simply a single rank order of linguistic ability. For longitudinal research, outcome batteries should sample comprehending and producing language, trained and unfamiliar materials, and levels of time pressure. Age-related models become more meaningful when their predicted outcomes are tied to specified tasks and measured histories of input. This approach accommodates developmental constraints while making room for evidence about sustained learning and successful specialization. Hartshorne et al. (2018) Johnson & Newport (1989) Norris & Ortega (2000)

Individual Differences and Learning Opportunity

Second-language acquisition trajectories vary among individuals exposed to seemingly similar curricula. Some variability arises from differences in actual input: attention during class, opportunities for conversation, time spent reading, exposure to relevant structures, and feedback after misunderstandings. Other potentially important dimensions include perceptual discrimination, prior language learning, literacy, motivation, and habits of sustained practice. These constructs need operational definitions if they are to explain outcomes. For example, a general self-rating of motivation may correlate with proficiency but does not reveal whether the mechanism is longer exposure, more intensive practice, or greater willingness to seek corrective information. Norris & Ortega (2000) Mackey (1999)

Methodologically, individual differences should be evaluated as distributions rather than used to produce simplistic profiles of good and poor language learners. A study of predictive factors needs reliable measures, sufficient variation in outcomes, and prospective data where feasible. Because learning opportunities change with proficiency, a correlation between confidence and attainment may reflect effects in both directions. Randomized instruction can test specific pedagogical mechanisms, whereas longitudinal observations can clarify developmental sequences and the persistence of gains. The same evidence base warns against treating classroom averages as an adequate description of students whose home languages or educational histories differ substantially. Spada & Tomita (2010) Norris & Ortega (2000)

Prior multilingual experience may also alter how learners approach another language. Knowledge of multiple phonological inventories and grammatical mappings can supply alternative comparisons and may increase awareness that a familiar form has different functions across languages. These are plausible learning pathways, but a researcher must distinguish the number of languages known from proficiency, literacy, educational access, and frequency of use. Transfer patterns should be tested with the actual linguistic features shared by the learner’s known systems. An individualized instruction program can then address observed difficulties through diagnostic evidence rather than assuming that all speakers of a particular language will make the same errors. Odlin (1989) Kellerman (1979) Ellis (2006)

Lexical Frequency, Proficiency, and Practice

Lexical frequency influences the speed and reliability of access, but frequency effects in bilingual populations can have several explanations. The weaker-links hypothesis emphasizes that lexical representations are used less often when experience is distributed across languages. Gollan and colleagues investigated how frequency effects vary with bilingualism and aging, supplying a retrieval-based account of some observed differences. This perspective is important for acquisition because a slow naming response to a known word need not indicate a weak conceptual representation. Testing recognition, naming, and translation separately permits a more precise description of the difficulty. Gollan et al. (2008)

Practice affects the accessibility of particular linguistic forms and their integration into larger patterns. Frequent encounters can strengthen familiar sequences, while varied contextual examples may be necessary for flexible semantic and grammatical use. Vocabulary interventions differ in whether learners recognize, recall, generate, or use target words in meaning-bearing activity. Webb and colleagues’ meta-analytic treatment of intentional vocabulary learning underscores the importance of defining the outcome and comparing activities within their experimental conditions. An apparent benefit measured immediately after instruction may change when tested after a delay or on a previously untrained context. Webb et al. (2020)

Proficiency is often used as a predictor in studies of cross-language interaction, although common proficiency measures aggregate heterogeneous abilities. High vocabulary scores need not imply native-like accent or efficient syntactic revision, and a fluent conversation does not demonstrate knowledge of every morphological distinction. Such variation is predicted when exposure and retrieval histories differ across domains. A better developmental account represents proficiency as a profile of measured capabilities and links each outcome to the amount, type, and context of relevant practice. This is especially necessary when interpreting age comparisons in which younger and older learners have received different educational experiences. Kroll & Stewart (1994) Hartshorne et al. (2018) Hopp (2010)

Multilingual Development and Heritage Language Contexts

Transfer takes place in multilingual societies where linguistic exposure is distributed unevenly across home, school, peers, and institutions. A bilingual child’s development cannot be interpreted simply by comparing them with separate monolingual norms in each language. Knowledge of a phonological distinction or morphological pattern may support related learning across languages, while differences in educational opportunity shape which skills receive explicit practice. Longitudinal investigation is particularly valuable because the direction of influence can change with schooling and shifts in language dominance. Research on bilingual children’s phonological and morphological awareness provides a framework for evaluating which cross-language relations predict later literacy-related outcomes. Zhang et al. (2025)

Heritage-language experience also illustrates the importance of separating linguistic competence from opportunities to demonstrate it. A child may converse fluently in a home language while receiving nearly all formal literacy instruction in the school language. Testing only written vocabulary in the heritage language would confound educational access with broad linguistic ability. Studies of transfer should report where a construction is encountered, in which register, and how the relevant task relates to habitual use. When children possess multiple linguistic resources, cross-language relations may be mutually supportive even when observed developmental rates differ by setting. Zhang et al. (2025) Odlin (1989)

Research on multilingual development carries methodological and ethical responsibilities. Language-background questionnaires need to capture more than binary native and nonnative labels; community patterns of switching, dialect variation, and literacy opportunities can change the interpretation of results. Comparisons require care when the instruments were developed for populations whose educational histories differ. A finding that two groups perform differently on a specific language task should be attributed to the measured task and supported predictors, while cultural or linguistic stereotypes add no explanatory value. The long-term goal is a model of how multiple linguistic systems develop through participation, practice, and interacting learning conditions. Kroll et al. (2015) Zhang et al. (2025) Norris & Ortega (2000)

Implicit Learning, Noticing, and Feedback

Learning an additional language involves incidental adaptation to encountered linguistic distributions as well as deliberate study of linguistic rules. These descriptions refer to learning conditions and psychological processes that require independent measurement. A learner can improve after exposure without being able to articulate a rule, while explicit knowledge may support editing under sufficient time. Schmidt’s account of noticing assigns an important role to conscious registration of relevant input features, particularly when features are low in perceptual salience. Subsequent empirical work has debated how awareness should be operationalized and how much learning can occur under different conditions of attention. Schmidt (1990) Ellis (2006)

Feedback adds information about the relation between a learner’s present system and the forms encountered in communication. It may come through explicit correction, reformulation, clarification requests, or changes in the success of interaction. Mackey’s experimental investigation of interaction and English question formation contributed evidence that communicative exchanges can affect development in specifically measured structures. Nevertheless, an immediate corrected response does not automatically demonstrate an enduring modification of grammar. Studies should include delayed measures and opportunities to produce new examples without the original feedback. Distinguishing what learners noticed from what they subsequently generalized remains essential to explaining the mechanism. Mackey (1999) Schmidt (1990)

Selective-attention models help connect noticing with transfer. When a familiar grammatical cue is dominant, learners may overlook a less salient signal even after encountering several counterexamples. Training that makes the neglected cue informative or that directs attention to it offers a way to test whether earlier experience created a learning bottleneck. Ellis discusses blocking, overshadowing, cue competition, and perceptual learning as mechanisms relevant to these situations. Empirically, the design should avoid conflating additional exposure with increased attention, and should distinguish explicit recall of a rule from automatic sensitivity during comprehension. Ellis (2006) MacWhinney (2005)

Instructional Effectiveness and Design

Second-language instruction can be evaluated through controlled comparisons of instructional conditions, linguistic features, learner groups, and measured outcomes. Norris and Ortega’s research synthesis and meta-analysis reported broad evidence that instruction can improve measured language-learning performance, while drawing attention to differences in study designs and outcome categories. A credible educational inference preserves these distinctions: the effects of a short, intensive intervention on a specific grammatical test may not generalize unchanged to conversational fluency or year-long retention. Research should identify what instruction taught, how the comparison group learned, and whether the test reflected practiced items or transferable competence. Norris & Ortega (2000)

Explicit and implicit instructional conditions are often compared through tasks that differ in opportunities to focus on form and verbalize rules. Spada and Tomita’s meta-analysis examined interactions between instructional type and language-feature type, making an important contribution to the debate over whether instruction should be matched to structural complexity. Yet broad labels such as explicit and implicit contain diverse procedures: explanation, recasts, input enhancement, production practice, and distributional exposure are not equivalent experimental treatments. Educators and researchers must interpret aggregate effects with attention to how each study defined instruction, the contrast targeted, and the validity of its tests. Spada & Tomita (2010) Norris & Ortega (2000)

An evidence-based approach to instruction follows a chain of reasoning from diagnosed learning difficulty to a proposed learning mechanism and an outcome measure. If the problem is weak use of a morphological cue, the training can increase attention to informative morphology; if the problem is a phonetic category, varied auditory exemplars may be more appropriate than written rule explanation alone. Delayed testing, novel-item generalization, learner diversity, and resource constraints determine practical significance. No intervention should be described as universally optimal across phonetics, grammar, vocabulary, and interaction, because these domains contain different learning problems and assessments. Ellis (2006) Saito (2012) Webb et al. (2020)

Phonetic Training and Generalization

Pronunciation instruction provides a comparatively direct way to examine perceptual learning and its transfer to production. Saito’s synthesis of quasi-experimental interventions investigated whether instruction could improve second-language pronunciation, while later analyses of high-variability phonetic training examined relations between trained perception and production outcomes. The research questions concern what changed, which materials elicited the improvement, and whether learners used the contrast with unfamiliar voices or spontaneous speech. A gain in discrimination for two trained syllables is narrower evidence than improvement in intelligibility across natural conversational contexts. Saito (2012) Uchihara et al. (2024)

High variability can be valuable because it exposes listeners to distinctions across different talkers, acoustic realizations, and linguistic contexts, reducing the chance that a learner memorizes accidental details of one stimulus set. Even so, variability is a design feature whose optimal amount and distribution depend on the contrast and learner. Quantitative syntheses report that effects and moderators differ across studies, and some claims about transfer are sensitive to the choice of perceptual or production measure. Studies should therefore report exposure schedules, sound inventories, participant backgrounds, assessment intervals, and whether scoring was performed blind to training condition. Uchihara et al. (2024) Yao et al. (2025)

Phonetic instruction illustrates a broader scientific principle: perceptual category formation, articulate production, and communicative intelligibility are related but separable outcomes. Training can improve the recognition of a contrast before articulatory coordination catches up, and accent judgments can reflect listener expectations in addition to acoustic production. Best and Tyler’s perceptual account and Flege and Bohn’s revised speech-learning framework offer hypotheses for why certain cross-language mappings resist differentiation. These hypotheses become educationally useful when they predict which contrasts require targeted exposure and which gains are likely to survive transfer to novel voices, words, and communicative settings. Best & Tyler (2007) Flege & Bohn (2021) Saito (2012)

Vocabulary Learning and Retention

Vocabulary learning concerns the formation of mappings among word forms, meanings, contexts, and usage constraints. Recognition of a written word, comprehension of it in a sentence, and retrieval during speaking can develop at different rates. Intentional learning activities attempt to change these processes through study, retrieval, meaning-focused engagement, or repeated exposure. Webb, Yanagisawa, and Uchihara’s meta-analytic work provides an empirical basis for evaluating such activities, although the diversity of interventions and tests requires careful interpretation. An outcome on an immediate multiple-choice recognition measure establishes a different learning achievement from the ability to recall a word weeks later. Webb et al. (2020)

Cross-linguistic transfer affects vocabulary development through cognates, translation equivalents, semantic categories, and the organization of lexical retrieval. Prior knowledge may reduce the burden of learning a concept but can also create a misleading expectation about collocations or semantic boundaries. Translation practice and picture-based naming emphasize partially different links between forms and concepts, as illustrated by Kroll and Stewart’s experiments. Instructional comparisons should measure proficiency at baseline and sample both words with transparent correspondences and words whose form or meaning is less readily transferable. This permits stronger attribution of improvement to the intended learning mechanism. Kroll & Stewart (1994) Odlin (1989)

Retention is critical because many short-term acquisition effects decay or change when practice stops. Experiments should specify whether they measure productive recall, receptive recognition, use in meaningful sentences, or fluent integration in discourse. Delayed tests and unfamiliar-item transfer provide evidence about more durable learning than repeated tests of memorized materials alone. Where possible, analysis should consider the confounding relationship between how frequently a form is encountered and how often it is selected for production. The weaker-links hypothesis offers one account of retrieval-frequency effects, complementing studies that focus on the establishment of semantic knowledge during learning. Webb et al. (2020) Gollan et al. (2008)

Longitudinal Methods and Evidence Quality

Longitudinal research is central to the acquisition question because it directly observes variation within the same learners over time. Repeated measurements can reveal changes in phonological discrimination, construction use, word access, grammatical judgments, and comprehension strategy. Yet a time series does not automatically establish a causal mechanism. Changes may coincide with classroom instruction, migration, changes in social network, reading intensity, or measurement familiarity. The design should collect relevant exposure information, represent missing observations explicitly, and examine whether progress differs across linguistic targets. The strongest longitudinal evidence combines repeated assessment with theoretically relevant variation in learning conditions. Dussias & Sagarra (2007) Zhang et al. (2025)

Experimental evidence likewise requires attention to item sampling and analytic flexibility. A language pair can contain dozens of relevant structures, and results from one selected contrast may not apply to another. Research reports should specify participant criteria, the operational definition of bilingualism or proficiency, the unit of randomization, exclusion decisions, the role of lexical items, and uncertainty around estimates. When a study compares different teaching techniques, adequate controls must make clear whether an observed effect reflects the technique itself or additional exposure, time, or attention. Methodological transparency increases the value of both positive and null findings for cumulative theory building. Norris & Ortega (2000) Spada & Tomita (2010)

Meta-analyses and systematic reviews support aggregate inference when their searches, eligibility rules, coding decisions, and heterogeneity assessments are transparent. They cannot eliminate weaknesses in the underlying primary studies, and broad average effects can conceal important moderators. Work identifying publication bias in the bilingual-cognition literature is a cautionary methodological analogue: published results need not represent the full distribution of conducted research. For second-language transfer, stronger synthesis requires multilingual sample diversity, preregistered predictions where appropriate, open materials for stimuli and tasks, and direct replication across language pairs and educational settings. de Bruin et al. (2015) Yao et al. (2025) Uchihara et al. (2024)

Neural and Cognitive Accounts

The cognitive architecture of additional-language learning links language-specific representations to learning, memory, attention, and action selection. An individual may recognize two languages’ words in overlapping processing pathways while controlling which form is appropriate to a particular communicative task. Kroll and colleagues examine how bilingual experience informs questions about language, mind, and brain, and Green and Abutalebi’s adaptive-control account describes how different interactional contexts place different demands on control. Such frameworks motivate hypotheses about the consequences of sustained multilingual experience; the cognitive measures used to test them must be distinguished from direct evidence of progress in learning phonological or grammatical material. Kroll et al. (2015) Green & Abutalebi (2013)

Claims about broad bilingual cognitive benefits require special scrutiny. A difference between bilingual and monolingual samples on a laboratory executive task does not identify a learning-induced brain change without adequate control for sampling and other influences. De Bruin and colleagues’ analysis of publication bias in the bilingual-advantage debate illustrates the danger of interpreting a selective published literature as a uniform population effect. This caution does not diminish the established fact that bilinguals coordinate linguistic alternatives during use; it clarifies that broader transfer to executive functioning constitutes a different empirical hypothesis and requires its own evidence. de Bruin et al. (2015) Green & Abutalebi (2013)

Neural observations may complement behavioral learning measures by showing changes associated with proficiency or experience, but they rarely reveal a unique causal history by themselves. A change in activation can reflect greater efficiency, altered strategy, task difficulty, or familiarity; interpretation depends on the experimental contrast and convergent evidence. The acquisition field benefits from research that integrates longitudinal language behavior with neural and cognitive measures while retaining the distinction among representation, access, control, and learning. Computationally explicit models are particularly useful when they translate these concepts into predictions for novel word forms, cue-conflict sentences, and repeated exposure. Kroll et al. (2015) MacWhinney (2005) Dijkstra & van Heuven (2002)

Social Interaction, Identity, and Educational Equity

An additional language is learned and used within social relationships. Interaction supplies examples of linguistic choices, opportunities to test intended meanings, and feedback when communication fails. Mackey’s study of interaction and question formation provides experimentally relevant evidence for the role of communicative episodes in measured development. Yet individuals’ opportunities to participate are distributed unevenly. Education, workplace access, migration experience, community attitudes, and social confidence can influence both the frequency of practice and which varieties of language are rewarded. Research that attributes these differences solely to age or inherited linguistic background loses explanatory information. Mackey (1999) Johnson & Newport (1989)

Social context also determines what counts as successful proficiency. A highly effective multilingual speaker may adapt to several communities, dialects, and registers rather than approximate a single idealized native-speaker model. Intelligibility, grammatical precision, discourse competence, literacy, and identity-related choices should therefore be assessed in relation to their actual functions. Accent can remain a salient social signal even where the speaker’s message is clearly understood, a distinction that matters when interpreting listener ratings and claims about ultimate attainment. Scientific assessment should avoid confusing linguistic performance with social judgments about whose speech is considered legitimate. Abrahamsson & Hyltenstam (2009) Saito (2012)

Equitable educational evaluation follows from this construct distinction. Students who acquired a language at home may display oral knowledge that conventional literacy tests underestimate; those who study a language in classrooms may know formal rules without having frequent opportunities for spontaneous conversation. Multi-method assessment should record the contexts in which each ability is practiced, use appropriate comparison populations, and make the stakes of each test explicit. Research across bilingual children’s phonological and morphological development shows why cross-language knowledge can be relevant to educational trajectories. An effective teaching response targets the observed linguistic need while preserving the learner’s existing repertoire as a resource. Zhang et al. (2025) Odlin (1989)

Artificial Tutoring and Human–Artificial Language Practice

Artificial conversational systems introduce a distinct environment for second-language practice. They can generate examples, sustain repeated dialogue, offer alternative phrasing, and simulate communicative tasks at adjustable levels of complexity. The scientific question is whether these affordances improve durable language learning, for whom, and under what conditions. A review of educational chatbot research by Wiboolyasarin and colleagues provides a contemporary entry point, while earlier interaction research supplies a framework for evaluating negotiation of meaning and feedback. Technology itself is not a learning outcome: improvements must be measured in independently assessed phonology, vocabulary, grammar, comprehension, or communicative participation. Wiboolyasarin et al. (2025) Mackey (1999)

An artificial tutor may provide immediate corrective explanations and extensive individualized practice, yet generated feedback can be inconsistent or linguistically unsuitable. Scientific evaluations should report the system version, the tasks used, whether outputs were checked for accuracy, and the extent of teacher or human interlocutor involvement. Randomized or carefully matched comparisons can test the effect of feedback type and practice opportunity; delayed assessments are needed to determine whether gains outlast the interaction. Outcomes should include unfamiliar materials and, where relevant, conversations with people so that apparent improvement in responding to a particular system is not mistaken for general communicative competence. Wiboolyasarin et al. (2025) Norris & Ortega (2000) Webb et al. (2020)

In the Artificial Era, additional-language learning also becomes a setting for studying cognitive partnership. Human learners may delegate example generation, translation, vocabulary retrieval, or form correction to artificial systems while retaining their own goals and linguistic judgments. This distribution of tasks raises experimentally tractable questions about cognitive offloading: whether assistance increases independent later performance, changes metalinguistic attention, or creates dependency on readily available suggestions. These are prospective interdisciplinary hypotheses rather than conclusions demonstrated by the established transfer literature. They can be studied with conditions that vary access to artificial assistance during practice and during subsequent unaided testing. Wiboolyasarin et al. (2025) Schmidt (1990) Kroll et al. (2015)

Integration of Transfer Across Levels

The evidence can be organized across four connected levels. At the representational level, familiar categories and lexical or syntactic relations supply starting structures for new learning. At the processing level, both languages can contribute competitors and facilitatory information to comprehension and production. At the developmental level, repeated exposure and communicative activity change which cues and constructions are available and how reliably they are used. At the social level, interactional settings determine opportunities for practice and the functions of successful language use. These levels help explain why a single result, such as a reaction-time difference, cannot settle every question about transfer. Odlin (1989) Dijkstra & van Heuven (2002) MacWhinney (2005)

Across linguistic domains, the clearest transfer claims identify a specific prediction about previously learned knowledge and compare it with plausible alternative accounts. In perception, the prediction may concern how an established category assimilates a new contrast; in syntax, it may concern the weighting of word order against case; in lexical processing, it may concern cross-language competition or the strength of translation links. Each prediction requires a dependent measure matched to the proposed mechanism. A useful synthesis therefore preserves domain-specific detail while connecting findings through recurring principles of exposure, competition, generalization, and task demands. Best & Tyler (2007) Hartsuiker et al. (2004) Kroll & Stewart (1994)

The unresolved research agenda is substantial. Comparative longitudinal studies should examine more language combinations and learning environments, including heritage-language and multimodal histories. Experiments need better separation between temporary activation and long-term restructuring, between reflective knowledge and fluent processing, and between benefits on practiced test items and transfer to authentic communication. Age-related studies must distinguish maturational hypotheses from correlated experiences with more precise developmental histories. Educational research should report interventions sufficiently well for independent replication and investigate variation rather than relying only on pooled averages. These directions follow directly from the conceptual and methodological gaps identified in the available research. Hartshorne et al. (2018) Spada & Tomita (2010) Zhang et al. (2025)

Conclusion

Second-language acquisition is the progressive development of additional linguistic knowledge and use through experience with an already organized linguistic repertoire. Cross-linguistic transfer is a measurable relation between that prior experience and subsequent perception, interpretation, learning, or production. The historical movement from error comparison toward interlanguage, cue learning, shared representations, and online experimental processing has established a diverse set of testable explanations. Prior linguistic knowledge can facilitate acquisition, bias attention toward unhelpful signals, shape interpretation, or alter retrieval probabilities. Which result occurs depends on the specific structure, the learner’s experience, and the method of measurement. Selinker (1972) Odlin (1989) Ellis (2006)

Empirical findings are strongest when evidence from complementary methods converges. Controlled phonetic training can evaluate category differentiation and generalization; cross-language priming probes structural availability; longitudinal designs trace within-person change; age-related samples constrain theories of developmental attainment; and instructional comparisons identify the effects of defined teaching conditions. The limitations are equally instructive. Accuracy does not specify processing mechanisms, online facilitation does not by itself demonstrate lasting acquisition, and correlations between age and eventual proficiency cannot independently isolate biological maturation. Reviews and meta-analyses sharpen aggregate inference while retaining the heterogeneity and design limitations of the underlying studies. Flege & Bohn (2021) Hartsuiker et al. (2004) Hartshorne et al. (2018) Norris & Ortega (2000)

The resulting scientific synthesis treats linguistic transfer as multilevel, selective, context-sensitive, and open to change. Research advances when representations, processing, developmental opportunity, and social participation are described with distinct operational constructs and then integrated through experimentally discriminating predictions. The growth of artificial language-learning environments expands the possible settings for practice and introduces new questions about feedback reliability, cognitive partnership, and independent retention. The enduring standard for evaluating these innovations remains the same: clearly defined outcomes, authentic source-based evidence, appropriate comparisons, and reproducible procedures that distinguish learning from immediate task performance. Kroll et al. (2015) Wiboolyasarin et al. (2025) de Bruin et al. (2015)

References

  • Abrahamsson, N., & Hyltenstam, K. (2009). Age of onset and nativelikeness in a second language: Listener perception versus linguistic scrutiny. Language Learning, 59(2), 249–306. https://doi.org/10.1111/j.1467-9922.2009.00507.x
  • Best, C. T., & Tyler, M. D. (2007). Nonnative and second-language speech perception: Commonalities and complementarities. In O.-S. Bohn & M. J. Munro (Eds.), Language experience in second language speech learning: In honor of James Emil Flege (pp. 13–34). John Benjamins. https://doi.org/10.1075/lllt.17.07bes
  • Clahsen, H., & Felser, C. (2006). How native-like is non-native language processing? Trends in Cognitive Sciences, 10(12), 564–570. https://doi.org/10.1016/j.tics.2006.10.002
  • de Bruin, A., Treccani, B., & Della Sala, S. (2015). Cognitive advantage in bilingualism: An example of publication bias? Psychological Science, 26(1), 99–107. https://doi.org/10.1177/0956797614557866
  • Dijkstra, T., & van Heuven, W. J. B. (2002). The architecture of the bilingual word recognition system: From identification to decision. Bilingualism: Language and Cognition, 5(3), 175–197. https://doi.org/10.1017/S1366728902003012
  • Dussias, P. E., & Sagarra, N. (2007). The effect of exposure on syntactic parsing in Spanish–English bilinguals. Bilingualism: Language and Cognition, 10(1), 101–116. https://doi.org/10.1017/S1366728906002847
  • Ellis, N. C. (2006). Selective attention and transfer phenomena in L2 acquisition: Contingency, cue competition, salience, interference, overshadowing, blocking, and perceptual learning. Applied Linguistics, 27(2), 164–194. https://doi.org/10.1093/applin/aml015
  • Flege, J. E., & Bohn, O.-S. (2021). The revised Speech Learning Model (SLM-r). In R. Wayland (Ed.), Second language speech learning: Theoretical and empirical progress (pp. 3–83). Cambridge University Press. https://doi.org/10.1017/9781108886901.002
  • Gollan, T. H., Montoya, R. I., Cera, C., & Sandoval, T. C. (2008). More use almost always means a smaller frequency effect: Aging, bilingualism, and the weaker links hypothesis. Journal of Memory and Language, 58(3), 787–814. https://doi.org/10.1016/j.jml.2007.07.001
  • Green, D. W., & Abutalebi, J. (2013). Language control in bilinguals: The adaptive control hypothesis. Journal of Cognitive Psychology, 25(5), 515–530. https://doi.org/10.1080/20445911.2013.796377
  • Hartshorne, J. K., Tenenbaum, J. B., & Pinker, S. (2018). A critical period for second language acquisition: Evidence from 2/3 million English speakers. Cognition, 177, 263–277. https://doi.org/10.1016/j.cognition.2018.04.007
  • Hartsuiker, R. J., Pickering, M. J., & Veltkamp, E. (2004). Is syntax separate or shared between languages? Cross-linguistic syntactic priming in Spanish-English bilinguals. Psychological Science, 15(6), 409–414. https://doi.org/10.1111/j.0956-7976.2004.00693.x
  • Hopp, H. (2010). Ultimate attainment in L2 inflection: Performance similarities between non-native and native speakers. Lingua, 120(4), 901–931. https://doi.org/10.1016/j.lingua.2009.06.004
  • Johnson, J. S., & Newport, E. L. (1989). Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology, 21(1), 60–99. https://doi.org/10.1016/0010-0285(89)90003-0
  • Kellerman, E. (1979). Transfer and non-transfer: Where we are now. Studies in Second Language Acquisition, 2(1), 37–57. https://doi.org/10.1017/S0272263100000942
  • Kroll, J. F., Dussias, P. E., Bice, K., & Perrotti, L. (2015). Bilingualism, mind, and brain. Annual Review of Linguistics, 1, 377–394. https://doi.org/10.1146/annurev-linguist-030514-124937
  • Kroll, J. F., & Stewart, E. (1994). Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and Language, 33(2), 149–174. https://doi.org/10.1006/jmla.1994.1008
  • Mackey, A. (1999). Input, interaction, and second language development: An empirical study of question formation in ESL. Studies in Second Language Acquisition, 21(4), 557–587. https://doi.org/10.1017/S0272263199004027
  • MacWhinney, B. (2005). Extending the Competition Model. International Journal of Bilingualism, 9(1), 69–84. https://doi.org/10.1177/13670069050090010501
  • Norris, J. M., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis and quantitative meta-analysis. Language Learning, 50(3), 417–528. https://doi.org/10.1111/0023-8333.00136
  • Odlin, T. (1989). Language transfer: Cross-linguistic influence in language learning. Cambridge University Press. https://doi.org/10.1017/CBO9781139524537
  • Saito, K. (2012). Effects of instruction on L2 pronunciation development: A synthesis of 15 quasi-experimental intervention studies. TESOL Quarterly, 46(4), 842–854. https://doi.org/10.1002/tesq.67
  • Schmidt, R. W. (1990). The role of consciousness in second language learning. Applied Linguistics, 11(2), 129–158. https://doi.org/10.1093/applin/11.2.129
  • Selinker, L. (1972). Interlanguage. International Review of Applied Linguistics in Language Teaching, 10(1–4), 209–232. https://doi.org/10.1515/iral.1972.10.1-4.209
  • Spada, N., & Tomita, Y. (2010). Interactions between type of instruction and type of language feature: A meta-analysis. Language Learning, 60(2), 263–308. https://doi.org/10.1111/j.1467-9922.2010.00562.x
  • Uchihara, T., Karas, M., & Thomson, R. I. (2024). Does perceptual high variability phonetic training improve L2 speech production? A meta-analysis of perception-production connection. Applied Psycholinguistics, 45(4), 591–623. https://doi.org/10.1017/S0142716424000195
  • Webb, S., Yanagisawa, A., & Uchihara, T. (2020). How effective are intentional vocabulary-learning activities? A meta-analysis. The Modern Language Journal, 104(4), 715–738. https://doi.org/10.1111/modl.12671
  • Wiboolyasarin, W., Wiboolyasarin, K., Tiranant, P., Jinowat, N., & Boonyakitanont, P. (2025). AI-driven chatbots in second language education: A systematic review of their efficacy and pedagogical implications. Ampersand, 14, Article 100224. https://doi.org/10.1016/j.amper.2025.100224
  • Yao, Y., He, M., Chen, F., & Zhu, J. (2025). A meta-analysis of second language phonetic training: Exploring overall effect and moderating factors. Journal of Speech, Language, and Hearing Research, 68(4), 1784–1802. https://doi.org/10.1044/2024_JSLHR-24-00432
  • Zhang, K., Sun, X., Flores-Gaona, Z., Yu, C.-L., Eggleston, R. L., Nickerson, N., Caruso, V. C., Tardif, T., & Kovelman, I. (2025). Cross-linguistic transfer in bilingual children’s phonological and morphological awareness skills: A longitudinal perspective. Bilingualism: Language and Cognition, 28(2), 327–342. https://doi.org/10.1017/S1366728924000439