Bilingual language processing is the set of cognitive operations through which a person recognizes, understands, retrieves, and produces language while having access to two or more linguistic systems. The central finding of modern psycholinguistics is that knowing two languages changes the conditions under which words are processed: information associated with one language can become active during the use of another. This cross-language activation is observable even when a speaker intends to use only one language. Successful bilingual communication therefore depends on the interaction of shared meaning, partially overlapping word forms, contextual expectations, and flexible language control. It does not require that a person consciously translate every word or alternate between languages. The extent of interaction varies with the person, the languages, the task, and the moment. Kroll et al. (2015); Lauro and Schwartz (2017).
What Does It Mean for Two Languages to Interact in One Mind?
Language knowledge is coordinated across several levels
An individual who uses Spanish and English has learned at least two sets of phonological patterns, word forms, grammatical regularities, and conventions for selecting expressions. Those sets are interrelated through meanings, situations, memories, and expectations. When the person hears cat, the meaning may be reached directly, while the Spanish translation gato can also become more accessible through semantic relationships. When the person reads piano, its similar spelling and meaning across languages can make recognition easier than recognition of a word without such overlap. Neither example implies that each bilingual has exactly the same mental representation or that translation is a compulsory intermediate step. It shows why a model of bilingual processing must explain interactions between representations as well as the ability to produce an appropriate response. Dijkstra and van Heuven (2002); Kroll et al. (2010).
Processing takes place over time. Hearing a word involves interpreting acoustic cues before its identity is certain; reading involves mapping letters onto lexical, phonological, and semantic representations; speaking requires choosing an intended concept, retrieving an expression, preparing sound structure, and executing articulation. Language cues guide each stage, but the consequences of knowing a second language can emerge at several stages. An experimental result about printed word recognition cannot automatically be generalized to speech production, conversational comprehension, or grammatical agreement. Research treats these as related mechanisms with partly different demands.
Activation, selection, and control are distinct questions
Activation concerns which candidate representations become accessible. Selection concerns which candidate is used to interpret input or produce output. Control concerns the regulation of goals, competing responses, and language use across contexts. A bilingual can activate a word from an unintended language and still choose the correct word with no observable hesitation. Theoretical disagreement often arises because different experiments study different stages. Strong evidence for nonselective activation does not by itself prove that every act of selection is competitive or relies on inhibition; those are additional hypotheses that require independent evidence. Kroll et al. (2008).
Bilingualism is a profile of experience
A bilingual person may have learned two languages from infancy, acquired a second language in school, become highly proficient through migration, or retained a heritage language used chiefly with family. Proficiency may differ across speaking, listening, reading, and writing; a person may know professional vocabulary in one language and family vocabulary in another. Relative language dominance is consequently a relationship among skills and circumstances rather than a permanent label attached to a language. Frequency of use, age and conditions of acquisition, language pair, literacy, and interlocutors influence performance. Studies that simply divide participants into 'bilingual' and 'monolingual' groups can conceal these dimensions. Treffers-Daller (2019).
The Core Finding: Both Languages Can Be Active at Once
Language-nonselective lexical access
Lexical access is the process through which incoming language activates candidate words and their meanings. In a strictly language-selective account, an instruction to read English would prevent Dutch or Spanish lexical competitors from becoming active. In a nonselective account, features of the stimulus activate compatible representations regardless of their language membership, after which linguistic context and response demands affect the outcome. Decades of experiments favor substantial nonselective activation, especially in earlier stages of word recognition, while showing that its size and duration are context sensitive. The conclusion concerns measurable processes, not a claim that an entire second vocabulary is continuously or consciously present. van Heuven et al. (1998); Lauro and Schwartz (2017).
A printed English word may resemble a word in Dutch, and an English speech sound sequence may resemble the beginning of a Russian word. Each language contributes candidates because the input has overlapping features. The cognitive system must use the available evidence to converge on a useful interpretation. Cross-language activation can facilitate recognition when representations reinforce one another or interfere when they support incompatible responses. These patterns are systematic consequences of learned representations and task design.
Cognates: when two languages support recognition
Cognates are words in different languages with related meaning and similar form. Many English and Spanish words, such as animal and animal, provide familiar illustrations, although the degree of similarity and pronunciation differs across pairs. In lexical-decision and reading experiments, cognates are often processed faster or more accurately than matched noncognates. This 'cognate facilitation' provides evidence that representations associated with both languages can contribute to recognition. The effect is conditional: word frequency, degree of form overlap, language dominance, sentence predictability, and the task all matter. Researchers must match frequency, length, semantic properties, and orthographic neighborhood before attributing a response difference to bilingual coactivation. Lauro and Schwartz (2017).
Cognate facilitation also has a theoretical limit. Faster naming or lexical decisions do not reveal a unique internal pathway. Facilitation might arise through overlapping orthographic codes, common phonological patterns, shared semantic activation, or interactions across levels. Converging evidence from different tasks is needed to distinguish mechanisms. A speaker can also benefit from a cognate strategically after identifying it, so timing is crucial.
Interlingual homographs: when form points to different meanings
An interlingual homograph has the same or similar written form in two languages but different meanings. English coin and French coin ('corner') illustrate the potential conflict. Under some laboratory conditions, homographs produce interference, and under others they produce little effect or even facilitation. These differences matter: they indicate that the outcome of cross-language activation cannot be reduced to a single universal 'bilingual interference' rule. In an eye-tracking study, Libben and Titone observed early evidence of cross-language activation in English sentences, whereas strongly constraining sentences reduced later interference and facilitation effects. Libben and Titone (2009).
The distinction between early and late measures explains how a reader can activate multiple candidates without ending in a confused interpretation. Context can change which meaning remains accessible as the sentence unfolds. The processing system is dynamic; initially broad activation can be followed by rapid resolution.
Cross-language orthographic neighborhoods
A written word's orthographic neighborhood consists of other words resembling it in spelling. For Dutch-English bilinguals, Dutch neighbors can influence how rapidly an English target is recognized even when the task uses English alone. The classic experiments by van Heuven, Dijkstra, and Grainger showed such cross-language neighborhood effects and provided support for a computational architecture in which word forms from both languages interact. The result is especially informative because the competing items need not be translation equivalents. Their influence is explained by visual-form similarity and learned lexical representations. van Heuven et al. (1998).
Spoken Language: Competition Begins Before the Word Is Complete
Hearing a sound sequence activates possible words
Speech unfolds incrementally. A listener usually has to identify words before the entire acoustic sequence has arrived, and many words share early sounds. In a bilingual mind, the pool of phonologically compatible candidates can include words from both known languages. This is a testable prediction: when listeners hear a target word, they may briefly inspect an object whose name resembles the target in the other language. Eye movements provide a measure of this competition without requiring an explicit translation judgment. Marian, Spivey, and Hirsch (2003)00535-7).
What the visual-world paradigm shows
In the visual-world paradigm, a participant views objects while following spoken instructions. The experiment measures where the eyes move as the acoustic signal develops. In work on Russian-English bilinguals, cross-language phonological competitors influenced visual attention even though the spoken instructions used only one language. The results support parallel activation in spoken-word comprehension. They also clarify a methodological point: the effect concerns competition among phonological forms evoked by the stimulus, not deliberate switching between conversational languages. Marian, Spivey, and Hirsch (2003)00535-7).
Spoken coactivation depends on proficiency and lexical characteristics. A person who seldom uses one language may show a smaller or later competitor effect than someone who uses both languages regularly; the effects also depend on how similar the candidate words are, the speed of the signal, and how the task distributes attention. Experiments that average across heterogeneous bilingual participants may miss consistent patterns at the individual level. Blumenfeld and Marian (2007).
Phonology, accent, and phonetic categories
Bilingual experience influences the learned relationships among speech sounds and phonological categories. An individual who distinguishes two contrasts in one language may rely on different acoustic cues in the other. A second language can also affect the perception and production of a first language over time. This interaction belongs to a broader speech-perception literature, but its relevance here is precise: changes in sound categories shape which lexical candidates are activated and how efficiently a bilingual recognizes words. A difference in accent or cue weighting is an expected aspect of language experience and is not, by itself, a communication disorder.
Reading Sentences: How Context Shapes Cross-Language Activation
Sentence meaning constrains competing interpretations
A word rarely appears outside a meaningful context in everyday reading. A sentence establishes topic, grammar, and expectations that can make one meaning much more likely than another. An influential hypothesis is that such context can constrain the influence of the non-target language. Early evidence has been interpreted in different ways because some measures capture initial activation while others capture later interpretation. A meta-analysis of cognate facilitation in sentence contexts found effects under both low and high semantic constraint, but smaller effects in high-constraint contexts. This supports graded contextual modulation more strongly than the idea that a meaningful sentence completely shuts the other language off. Lauro and Schwartz (2017).
Timing changes what an experiment establishes
First-fixation duration, gaze duration, regression patterns, and total reading time reflect overlapping but nonidentical aspects of comprehension. Early measures often reveal form-based competition; later measures reflect resolution, integration, and repair. A late absence of interference is compatible with initial nonselective access followed by successful contextual selection. Likewise, a late cognate advantage does not prove that all lexical decisions were made before sentence context became available. The strongest explanations specify when contextual information enters processing and how its effects are identified. Libben and Titone (2009).
Grammar and discourse introduce further constraints
When a bilingual understands a sentence, grammatical cues, word order, referents, and discourse expectations help establish the intended interpretation. Linguistic knowledge may transfer across languages in some circumstances, but effects depend on construction, proficiency, task, and language pair. Cross-language syntactic influence is a different research question from lexical competition; evidence for shared or interacting syntactic structures should be evaluated with structural priming, comprehension, and production studies rather than inferred directly from cognate experiments. Language mixing in real conversations introduces additional discourse cues, but those patterns also deserve an independent analysis of code-switching conventions.
Speech Production: Choosing a Word When Alternatives Are Available
From intention to articulation
To name an object, a speaker forms a conceptual intention and retrieves a linguistic expression, then plans its form and prepares articulation. A bilingual has more than one potential label for many concepts, yet usually produces a contextually appropriate expression. In picture-naming studies, even using a single language can activate lexical alternatives associated with another language. The central production problem is how the target form becomes selected, and to what extent unrelated or cross-language alternatives participate in competition. Kroll et al. (2008).
Does lexical selection require inhibition?
In competitive accounts, activated candidates interfere with one another and a speaker must resolve competition before speaking. Some versions propose inhibitory control over alternatives from the non-target language. In language-specific selection accounts, several languages may become active while selection rules privilege words belonging to the intended language. Both positions can accommodate some evidence of parallel activation, so activation alone cannot decide between them. Reaction-time patterns, error distributions, and the consequences of switching provide stronger tests. Reviews of bilingual production document evidence favoring inhibitory contributions while also showing why selection mechanisms remain disputed. Kroll et al. (2008).
The distinction also prevents a conceptual confusion. Language inhibition is a proposed mechanism of managing interference during a linguistic task. Its presence does not entail a general, transferable improvement in all forms of attention or cognitive control. That broader claim requires separate comparisons and cannot be established from word-naming experiments alone.
Translation equivalents and shared concepts
Many concepts are connected with expressions in both languages. Models differ in whether they assume direct links between translation words, shared conceptual representations, or distributed mappings involving language-specific aspects of meaning. The Revised Hierarchical Model historically emphasized differences between early second-language lexical-conceptual connections and more proficient use, while later theoretical criticism questioned rigid separations between lexicons and the strength of links from second-language words to meaning. Current research treats the model as a useful historical framework whose specific predictions must be tested against more integrated and dynamic accounts. Kroll et al. (2010); Brysbaert and Duyck (2010).
Word retrieval and the frequency-of-use problem
A bilingual's word-retrieval performance partly reflects opportunities to use each word. If vocabulary is distributed across two languages, the frequency of a particular lexical form in either language may be lower than the frequency of a comparable form in a monolingual's routine. Some bilingual samples therefore show slower picture naming or produce fewer items in single-language verbal-fluency tasks, even when their overall conceptual knowledge is extensive. These results should be interpreted in light of task language, vocabulary domain, frequency, and participant matching. Lower performance in a specific laboratory measure does not justify characterizing bilingual cognition globally as deficient. Gollan, Montoya, and Werner (2002).
Theoretical Architectures of the Bilingual Lexicon
Bilingual Interactive Activation and BIA+
The original Bilingual Interactive Activation model was developed primarily to explain visual word recognition. It describes the interaction of representations for letters and words belonging to more than one language, including competitive processes across candidate words. The later BIA+ framework extended the representational architecture to include phonology and semantics and separated lexical identification from the task-and-decision processes that determine the response. This distinction is especially important because a bilingual may show early cross-language activation while still giving a language-appropriate response. The framework predicts influences of input similarity and contextual information, but it does not settle every question about speech production, spoken comprehension, or learning. Dijkstra and van Heuven (2002).
BIA+ is a substantive theory rather than a catch-all label for every bilingual result. Its usefulness depends on whether it makes discriminating predictions about form overlap, word frequency, timing, and sentence context. The architecture has inspired further computational work, while evidence from diverse languages and modalities continues to challenge assumptions derived predominantly from alphabetic reading tasks.
Revised Hierarchical Model and changing L2 access
The Revised Hierarchical Model distinguishes links between lexical forms and concepts and historically predicted that, early in second-language learning, translation pathways through a dominant language may be stronger than direct conceptual access in the weaker language. As experience grows, direct L2-to-meaning connections may strengthen. The theory helped organize translation and naming experiments, but its simple asymmetries do not capture every learner or every task. Critical evaluations argue that evidence for strict separate lexicons or uniformly weak L2 conceptual access is limited and that more integrated accounts are necessary. Kroll et al. (2010); Brysbaert and Duyck (2010).
Language control models
Green's inhibitory-control perspective treats the regulation of competing language plans as a cognitive-control problem. Green and Abutalebi subsequently proposed the Adaptive Control Hypothesis, according to which recurrent language-use contexts place different demands on goal maintenance, interference management, cue detection, and switching. A household where languages are strongly separated across situations, a community where interlocutors require alternation, and a community where languages are densely mixed create different control environments. The theory offers explicit predictions, but its proposed long-term adaptations cannot be assumed from observing the existence of bilingual communication. They require targeted evidence that experience changes processing mechanisms. Green and Abutalebi (2013).
Language Switching and the Dynamics of Control
Switch costs in controlled naming experiments
A common laboratory task asks a bilingual to name pictures or numerals in the language indicated by a cue. Responses on switch trials can be slower than on trials that repeat the language; this difference is called a switch cost. Meuter and Allport reported an asymmetry in which switching back into the more dominant first language could be especially costly. One influential interpretation is that prior suppression of the dominant language carries over to the next trial. Other mechanisms, including cue processing and persistent task settings, can also contribute. The pattern establishes an effect in a specific experimental context, not a universal rule about conversational bilingualism. Meuter and Allport (1999).
Proficiency changes the pattern
Costa and Santesteban compared second-language learners with highly proficient bilinguals and found that switching patterns differed between groups. Some highly proficient bilinguals did not show the asymmetrical costs expected from learners with a strongly dominant first language. This finding limits theories that treat the same fixed inhibitory operation as a complete explanation at every proficiency level. Language selection may be reorganized through experience, and individual strategies, response cues, and linguistic contexts may contribute. Costa and Santesteban (2004).
Switching, mixing, and everyday conversation
Experimentally forced switching, voluntary alternation between languages, and socially meaningful code-switching are related but different situations. In conversation, listeners, topics, cultural conventions, and communicative goals provide information about which language to use. The Adaptive Control Hypothesis distinguishes contexts partly because communication practices shape control demands. Controlled naming results illuminate the costs of selecting responses, whereas naturally occurring code-switching also involves syntax, identity, interaction, and discourse. The specialized question of why bilingual speakers alternate languages belongs to a separate scientific account; here the focus is the processing architecture that makes context-sensitive selection possible. Green and Abutalebi (2013).
How Proficiency, Dominance, and Experience Change Processing
Dominance is multidimensional
A speaker's first-acquired language is often dominant, but not necessarily in every skill or life stage. Dominance can be evaluated through vocabulary, fluency, grammatical performance, comprehension, speed of retrieval, exposure, or distribution of language use. The measures are related but not interchangeable. A heritage-language speaker might be fast at informal conversation but less practiced at technical writing, while their second language supports education or professional reasoning. Any claim that one language is 'stronger' must therefore specify the domain and measurement. Treffers-Daller (2019).
Learning history influences representations
Age of acquisition, amount of exposure, literacy experience, frequency of active production, and social necessity shape which forms become entrenched. The comparison of an early simultaneous bilingual with an adult who learned another language through classroom instruction involves differences in learning history as well as proficiency. These differences can change lexical access and processing speed. Yet no single age threshold guarantees a particular profile of coactivation or language control; age effects interact with experience, practice, and the linguistic domain studied. General claims about critical periods and second-language learning belong to a broader acquisition literature.
Contextual language use matters
People choose language differently with family, friends, coworkers, classmates, and services. Such contexts change expectations and repetition rates, and potentially the degree of control needed. In the Adaptive Control Hypothesis, single-language and dual-language environments exert different recurring demands. Measuring those contexts rigorously is challenging: self-report is imperfect, people move across settings, and proficiency may influence chosen contexts as much as contexts influence proficiency. Correlations between language practices and cognitive behavior therefore cannot alone establish the direction of causation. Green and Abutalebi (2013); Treffers-Daller (2019).
Language pair and writing system
Cognate relationships and spelling-based overlap are abundant for some language pairs and rare or differently represented for others. Scripts, orthographic depth, word order, morphological complexity, and phoneme inventories all influence experimental design. Findings from Dutch-English or Spanish-Catalan participants cannot automatically characterize Arabic-English, Chinese-English, or bilingual sign-language processing. A credible account identifies which result follows from a general principle and which reflects the distinctive structure of the languages tested.
What the Brain Evidence Adds
Distributed language and control systems
Functional neuroimaging links bilingual processing to coordinated cortical and subcortical networks involved in language representation, selection, and action regulation. Meta-analytic work indicates substantial neural overlap between first- and second-language processing alongside differences associated with linguistic level, acquisition history, and control demands. These are group-level convergences rather than dedicated anatomical 'boxes' for individual languages. A brain region's increased activation does not, on its own, establish that a single psychological process occurred or that the participant used more effort. Bilingual language processing neuroimaging meta-analysis (2019).
Language control and general task control partly overlap
A neuroimaging meta-analysis comparing language switching with nonlinguistic task switching found both shared and differentiated activation patterns. The overlap is consistent with the recruitment of some domain-general control processes during bilingual language selection; the differences are consistent with language-specific processing demands. It does not follow that using two languages automatically enhances broad executive ability. Similar neural activation in two tasks can arise for multiple reasons, and the relationship between activation and behavioral performance requires independent tests. Meta-analysis of language and task switching (2022).
Recent evidence on language balance
A 2025 activation-likelihood-estimation meta-analysis examined whether relative balance between languages moderates convergence in bilingual language-control studies. It reported some differences in cortical activation patterns between balanced and unbalanced groups and emphasized the importance of language proficiency. This is an informative group-level synthesis, but meta-analytic spatial convergence is not a direct measure of individual inhibition, representational integration, or improved everyday functioning. Definitions of balance, task diversity, and the number of eligible experiments influence interpretation. Chen and colleagues (2025).
How Researchers Test Bilingual Language Processing
Lexical decision and word naming
Lexical-decision tasks require participants to determine whether a letter string is a word under specified language instructions. Naming tasks ask participants to say a displayed word or label a picture. Researchers manipulate cognate status, homography, frequency, orthographic neighbors, response language, and trial sequence while recording reaction time and accuracy. These tasks efficiently isolate measurable components of access, but participants may adopt decision strategies that are uncommon in natural conversation. A significant reaction-time difference identifies a behavioral contrast; its underlying process must be established through models and converging experiments. Dijkstra and van Heuven (2002).
Eye tracking during reading and listening
Eye movements provide temporally informative data about when competing words attract attention. During reading, early and late fixation measures can distinguish initial cross-language activation from contextual resolution. In spoken visual-world tasks, looks to phonological competitors show which candidate forms are momentarily relevant while a word unfolds. Eye movements are indirect measures: visual salience, display composition, and timing must be controlled before linguistic interpretations are warranted. Libben and Titone (2009); Marian, Spivey, and Hirsch (2003)00535-7).
EEG, event-related potentials, and neuroimaging
Electroencephalography can help identify the temporal course of processing, including responses associated with unexpected words or difficult integration. Functional magnetic resonance imaging estimates changes in blood oxygenation associated with task performance. These approaches answer different questions: millisecond-scale temporal information comes with uncertain spatial localization in scalp EEG, whereas fMRI maps spatially informative but temporally indirect signals. Neither technique can directly read a participant's mental lexicon or identify a psychological mechanism uniquely from a single pattern. The strongest studies combine neurophysiology with carefully designed behavioral contrasts. Kroll et al. (2015).
Computational modeling and discriminating predictions
Computational models specify representations, activation rules, competitive interactions, learning mechanisms, and output decisions. A model gains explanatory strength by predicting reaction-time patterns, cross-language neighborhood effects, and differences between experimental conditions, ideally including data not used to fit its parameters. Fitting a data set is necessary but not sufficient: alternative architectures may reproduce the same observations. Researchers need comparisons of model predictions, generalization to different language pairs, and robust sensitivity analyses. BIA+ illustrates how representational and decision components can be distinguished; competing language-control models illustrate why one observation can accommodate several mechanisms. Dijkstra and van Heuven (2002).
Study design, participant matching, and reproducibility
Bilingual experience varies continuously, but many studies use coarse group labels, convenience samples, or tasks with modest reliability. Investigators should document proficiency in each modality, age and context of acquisition, language use, literacy, educational background, socioeconomic conditions, and relevant linguistic characteristics. Trial-level analyses should account for participants and items; researchers should distinguish confirmatory contrasts from exploratory subgroup analyses, correct for multiple tests when appropriate, and report uncertainty around effect sizes. Pre-registration and independent replication improve the evidential value of contentious claims. A meta-analysis is stronger when it detects heterogeneity and publication bias instead of concealing them. Lehtonen et al. (2018).
Established Results, Competing Explanations, and Open Questions
Established: cross-language interaction is measurable
Converging evidence from printed-word recognition, spoken-language eye tracking, and sentence-reading experiments supports the claim that known languages can become coactive. The strongest inference is conditional and mechanistic: under identifiable circumstances, information in the non-target language changes behavioral processing of target-language material. Evidence is particularly persuasive when an experimental manipulation creates different predictions for selective and nonselective models. Findings also show that context can reduce the effect after initial activation. Lauro and Schwartz (2017); Marian, Spivey, and Hirsch (2003)00535-7).
Contested: how selection is implemented
Researchers disagree over the relative contributions of competitive lexical selection, language-specific response rules, inhibition, task-set maintenance, and contextual expectations. These mechanisms need not be mutually exclusive at every stage or proficiency level. A model can explain a switch cost without explaining cognate facilitation, and a model of visual recognition may not predict voluntary speech choices. The scientific task is to identify combinations of mechanisms that survive tests across tasks and populations rather than declare one verbal description sufficient. Kroll et al. (2008); Costa and Santesteban (2004).
Limited: causal adaptation across a lifetime
Theories proposing that language-use environments reshape control networks generate researchable hypotheses. Observational differences among bilingual groups may reflect selection into environments, language dominance, education, and many other influences. Neural correlates of bilingual experience provide potentially important constraints, yet cannot alone establish a causal training effect. Longitudinal and intervention studies with appropriate comparison groups are needed to determine how particular experiences affect particular mechanisms. Green and Abutalebi (2013).
Separate question: general cognitive advantage
The ability to manage two languages is a linguistic achievement, but it does not establish a general cognitive superiority across attention, memory, or reasoning. A large meta-analytic review of adult executive-function comparisons found that initial very small effects did not remain after adjustments for publication bias. A meta-analysis of children reported a marginal overall effect alongside signs of bias and heterogeneity. These findings leave room for specific context-dependent effects but do not support a broad guaranteed bilingual advantage. The causal and methodological controversy is the independent focus of research on bilingualism and executive function, not the basis for describing ordinary bilingual lexical access. Lehtonen et al. (2018); Gunnerud et al. (2020); de Bruin et al. (2015).
Development, Learning, and Clinical Boundaries
Young bilinguals develop language within different input environments
Children learning more than one language encounter varying distributions of speech sounds, vocabulary, constructions, and conversational contexts. A child's vocabulary may be distributed across languages, so a single-language test can underestimate what the child knows across both. Variation in relative proficiency and accent is not itself a disorder. Research on bilingual processing helps explain these profiles, while developmental milestones, instructional decisions, and language interventions require their own evidence and individual assessment. ASHA, Multilingual Service Delivery.
Language difference and language disorder require different evidence
A bilingual person may use a construction transferred from another language, pause while searching for a word, or be more comfortable with particular topics in one language. Such features cannot establish developmental language disorder, aphasia, dyslexia, stuttering, or another clinical condition. The American Speech-Language-Hearing Association emphasizes multilingual, culturally responsive assessment and the need to distinguish differences associated with second-language acquisition from communication disorders. Clinical conclusions may require information across languages, appropriate norms, samples, histories, and dynamic assessment. The present article explains general mechanisms and leaves diagnosis and treatment to the relevant clinical disciplines. ASHA, Multilingual Service Delivery.
Practical Implications of the Evidence
Everyday word search and mixed-language environments
Occasional retrieval difficulty can reflect frequency of use, the strength of a particular lexical form, fatigue, task context, and the demand to select an expression. A bilingual may know a concept well but have faster access to its name in the language most often used for that topic. Increasing opportunities to retrieve words in the desired language can support language-specific fluency, though no single training practice follows universally from coactivation findings. Matching a communication setting to the person's actual linguistic experience is more informative than assuming that one language should always dominate. Gollan, Montoya, and Werner (2002); Treffers-Daller (2019).
Language education and assessment
Teachers can recognize that similarities across languages sometimes facilitate recognition and that apparently familiar word forms can create misleading meanings. Cognates may be useful supports when their forms and meanings actually align; interlingual homographs require attention to context. Assessments should specify whether they are measuring access to a word in one language, conceptual knowledge distributed across languages, or broader comprehension. A single-language score should be interpreted as the performance demanded by that test, with the person's history and linguistic repertoire taken into account. Clinical interpretation requires validated procedures appropriate to the population. Lauro and Schwartz (2017); ASHA, Multilingual Service Delivery.
Communication technology and the Artificial Era
Speech recognition, translation systems, and large language models can be evaluated against bilingual processing tasks, but computational success at mapping strings across languages establishes performance on those tasks rather than human-like lexical access. Human bilingual experiments measure perception, retrieval, learning history, bodily speech production, reaction time, and neural responses within lived communicative contexts. A computational system may implement shared representations and context-sensitive language selection without reproducing those psychological mechanisms or providing evidence of subjective experience. Comparisons should identify the explanatory level being compared: word predictions, internal representations, behavior, or neural and experiential phenomena. This distinction preserves the scientific meaning of human bilingual language processing while allowing computational research to supply useful testable models.
Frequently Asked Questions
Are both languages active even when I speak only one?
Often some representations associated with the other language can become active, as shown by cross-language competitor, cognate, and homograph effects. The degree and duration depend on the words, language pair, proficiency, and task. Activation ordinarily remains outside awareness and does not imply that the speaker will use the wrong language. Kroll et al. (2015).
Does a bilingual translate every word internally?
No universal translation step is established. Proficient speakers can often connect expressions with meanings directly, and models allow several routes between lexical forms and concepts. Translation-mediated processing may occur more often in certain tasks or learning stages, but bilingual comprehension does not require conscious word-by-word translation. Kroll et al. (2010).
Why do words sometimes come to mind faster in one language?
A word's accessibility depends on frequency, recent use, proficiency, topic, and competing candidates. Language dominance can change across domains and over a person's life. Different retrieval speeds for equivalent labels are compatible with a person having rich conceptual knowledge across both languages. Treffers-Daller (2019).
Is switching languages evidence of confusion?
Language alternation can be a normal, purposeful behavior shaped by interlocutors and context. Laboratory switch costs concern experimental response selection and do not diagnose confusion in everyday conversation. The precise social and grammatical functions of code-switching form a separate research topic. Green and Abutalebi (2013).
Does bilingualism improve executive function?
Managing languages involves control processes, but strong claims about a general executive-function advantage are disputed. Large meta-analyses show small, inconsistent or bias-sensitive group differences. Linguistic proficiency and the ability to coordinate languages remain meaningful outcomes independently of any general advantage. Lehtonen et al. (2018); Gunnerud et al. (2020).
Can a multilingual person have a language disorder?
Yes. Multilingual experience and communication disorders can coexist. Proper evaluation considers performance and experience across languages and distinguishes differences caused by language learning from clinically significant difficulties. Assessment and intervention require qualified, linguistically responsive professionals. ASHA, Multilingual Service Delivery.
Scientific Synthesis
Bilingual processing reveals that human language representations are interconnected, context sensitive, and regulated in real time. Experimental work establishes that the language a person is not currently using can influence word recognition and speech planning. Facilitation, interference, and rapid contextual resolution are different outcomes of this interaction. Computational and control theories explain aspects of these findings, while debates remain over exactly how lexical selection, inhibition, and adaptation are implemented. The strongest account treats bilingualism as a multidimensional history of experience and tests specific mechanisms with convergent behavioral, neural, and computational evidence. This provides a foundation for studying language acquisition, multilingual conversation, and linguistic change without treating contested claims about domain-general cognitive advantages as settled facts.
Related Articles
• Cluster Language, Thought and Psycholinguistics — the integrated scientific chapter on language, thought, and psycholinguistic evidence.
• Word Recognition: How the Mind Identifies Spoken and Written Words — lexical-access mechanisms relevant to cross-language coactivation.
• Speech Perception: How the Mind Turns Sounds Into Words — acoustic cues and competition among spoken-word candidates.
• Sentence Processing: How We Understand Grammar in Real Time — the time course of linguistic integration and contextual constraint.
• Psycholinguistic Research Methods: How Scientists Study Language Processing — eye tracking, behavioral experiments, and neurocognitive measurement.
• Language Processing in the Brain: Cognitive Networks and Neural Systems — brain networks supporting language and cognitive control.
Code-Switching: Why Bilingual Speakers Alternate Languages — the psycholinguistic control and communication of language switching.
Bilingualism and Executive Function: What Research Actually Shows — the contested evidence concerning bilingual executive-function advantages.
Language Attrition: Why Language Skills Change With Disuse — how linguistic accessibility and proficiency change with language use.
References
• American Speech-Language-Hearing Association. (n.d.). Multilingual service delivery in audiology and speech-language pathology. https://www.asha.org/practice-portal/professional-issues/multilingual-service-delivery/
• Blumenfeld, H. K., and Marian, V. (2007). Constraints on parallel activation in bilingual spoken language processing: Examining proficiency and lexical status using eye-tracking. Language and Cognitive Processes, 22(5), 633–660. https://doi.org/10.1080/01690960601000746
• Brysbaert, M., and Duyck, W. (2010). Is it time to leave behind the Revised Hierarchical Model of bilingual language processing after fifteen years of service? Bilingualism: Language and Cognition, 13(3), 359–371. https://doi.org/10.1017/S1366728909990344
• Costa, A., and Santesteban, M. (2004). Lexical access in bilingual speech production: Evidence from language switching in highly proficient bilinguals and L2 learners. Journal of Memory and Language, 50(4), 491–511. https://doi.org/10.1016/j.jml.2004.02.002
• de Bruin, A., Treccani, B., and 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., and 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
• Gollan, T. H., Montoya, R. I., and Werner, G. A. (2002). Semantic and letter fluency in Spanish-English bilinguals. Neuropsychology, 16(4), 562–576. https://pubmed.ncbi.nlm.nih.gov/12382994/
• Green, D. W., and 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
• Gunnerud, H. L., Ten Braak, D., Reikerås, E. K. L., Donolato, E., and Melby-Lervåg, M. (2020). Is bilingualism related to a cognitive advantage in children? A systematic review and meta-analysis. Psychological Bulletin, 146(12), 1059–1083. https://doi.org/10.1037/bul0000301
• Jiao, L., Meng, N., Wang, Z., Schwieter, J. W., and Liu, C. (2022). Partially shared neural mechanisms of language control and executive control in bilinguals: Meta-analytic comparisons of language and task switching studies. Neuropsychologia, 172, 108273. https://doi.org/10.1016/j.neuropsychologia.2022.108273
• Kroll, J. F., Bobb, S. C., Misra, M., and Guo, T. (2008). Language selection in bilingual speech: Evidence for inhibitory processes. Acta Psychologica, 128(3), 416–430. https://doi.org/10.1016/j.actpsy.2008.02.001
• Kroll, J. F., Dussias, P. E., Bice, K., and 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., van Hell, J. G., Tokowicz, N., and Green, D. W. (2010). The Revised Hierarchical Model: A critical review and assessment. Bilingualism: Language and Cognition, 13(3), 373–381. https://doi.org/10.1017/S136672891000009X
• Lauro, J., and Schwartz, A. I. (2017). Bilingual non-selective lexical access in sentence contexts: A meta-analytic review. Journal of Memory and Language, 92, 217–233. https://doi.org/10.1016/j.jml.2016.06.010
• Lehtonen, M., Soveri, A., Laine, A., Järvenpää, J., de Bruin, A., and Antfolk, J. (2018). Is bilingualism associated with enhanced executive functioning in adults? A meta-analytic review. Psychological Bulletin, 144(4), 394–425. https://doi.org/10.1037/bul0000142
• Libben, M. R., and Titone, D. A. (2009). Bilingual lexical access in context: Evidence from eye movements during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(2), 381–390. https://doi.org/10.1037/a0014875
• Marian, V., Spivey, M., and Hirsch, J. (2003). Shared and separate systems in bilingual language processing: Converging evidence from eyetracking and brain imaging. Brain and Language, 86(1), 70–82. https://doi.org/10.1016/S0093-934X(02)00535-7
• Meuter, R. F. I., and Allport, A. (1999). Bilingual language switching in naming: Asymmetrical costs of language selection. Journal of Memory and Language, 40(1), 25–40. https://doi.org/10.1006/jmla.1998.2602
• Sulpizio, S., Del Maschio, N., Fedeli, D., and Abutalebi, J. (2020). Bilingual language processing: A meta-analysis of functional neuroimaging studies. Neuroscience and Biobehavioral Reviews, 108, 834–853. https://doi.org/10.1016/j.neubiorev.2019.12.014
• Treffers-Daller, J. (2019). What defines language dominance in bilinguals? Annual Review of Linguistics, 5, 375–393. https://doi.org/10.1146/annurev-linguistics-011817-045554
• van Heuven, W. J. B., Dijkstra, T., and Grainger, J. (1998). Orthographic neighborhood effects in bilingual word recognition. Journal of Memory and Language, 39(3), 458–483. https://doi.org/10.1006/jmla.1998.2584
• Wang, T., Yin, K., Zhou, Q., Hu, H., Chen, S., and Wang, M. (2025). An activation likelihood estimation meta-analysis of how language balance impacts the neural basis of bilingual language control. Brain Sciences, 15(8), 803. https://doi.org/10.3390/brainsci15080803