Definition:
The bilingual lexicon refers to the mental representation and organization of vocabulary from two (or more) languages within a single cognitive system, encompassing how words from each language are stored, how they relate to each other and to shared conceptual representations, and how they are selectively or jointly activated during comprehension and production. Understanding the bilingual lexicon is central to both theoretical psycholinguistics and applied second language acquisition research.
Is There One Lexicon or Two?
A foundational question is whether bilinguals maintain two separate lexica (one per language) or a single integrated lexicon with language-tagged entries. The evidence overwhelmingly supports a single, integrated system: words from both languages are activated simultaneously during processing, even when only one language is being used.
Evidence includes:
- Cross-language priming: Translating “table” activates the L2 equivalent (e.g., mesa in Spanish) even in a task where Spanish is irrelevant
- Cross-language homograph effects: Words that look the same in two languages (e.g., gift = present in English, poison in German) create interference
- Cognate effects: Translation equivalents sharing form and meaning are processed faster than non-cognates
- Language non-selective lexical access: Presenting words in one language activates phonological neighbors in the other language
Theoretical Models
Revised Hierarchical Model (RHM; Kroll & Stewart, 1994)
The RHM proposes distinct representational levels for words (lexical level) and concepts (conceptual level), with asymmetric connections between L1 and L2:
- Strong direct connections from L2 words → L1 words (especially for less proficient bilinguals)
- Strong connections from L1 words → concepts
- Weaker direct connections from L2 words → concepts
Prediction: L2 learners initially access word meaning through L1 mediation; as proficiency increases, direct L2-to-concept links strengthen. This accounts for why less proficient bilinguals translate mentally when comprehending L2, while advanced bilinguals comprehend directly.
Bilingual Interactive Activation Model (BIA; Van Heuven et al., 1998)
Extended the interactive activation framework (from word recognition research) to bilingual contexts. Both languages exist in a single representational network, with all visually similar words (orthographic neighbors) activated in parallel regardless of language membership. A language node mechanism allows top-down inhibition to suppress the non-target language.
BIA+ (Dijkstra & Van Heuven, 2002)
Updated the BIA model to separate the identification system (non-selective activation, shared) from the task/decision system (where language goals and context apply). This cleanly predicted when language non-selectivity produces visible behavioral interference vs. not.
L1–L2 Asymmetries
| Feature | L1 | L2 (lower proficiency) | L2 (high proficiency) |
|---|---|---|---|
| Activation speed | Fast, automatic | Slow, effortful | Approaches L1 speed |
| L1 interference | — | High | Reduced |
| Direct concept links | Strong | Weak | Strengthening |
| Response to cognates | Baseline | Facilitation | Facilitation |
| Cross-language competition | Present | Very high | Present but managed |
Lexical Selection in Bilingual Production
During speech production in one language, words from both languages are simultaneously activated. Bilinguals must select the target-language word while suppressing competing L1 or L2 alternatives. The primary proposed mechanism is inhibitory control: the dominant language is actively suppressed to allow the weaker language to be selected. Evidence for this includes the language switch cost (slower production on language switch trials in bilingual experiments) and the counter-intuitive finding that bilinguals show stronger inhibition of the dominant L1 precisely when switching to it from a weaker L2.
Development of the Bilingual Lexicon
The bilingual lexicon is not static. As L2 proficiency increases:
- Direct L2-to-concept connections strengthen (less L1 mediation)
- Phonological representations for L2 words consolidate
- The mental lexicon becomes more elaborated with collocational, morphological, and pragmatic associations
- Susceptibility to cross-linguistic influence and tip-of-the-tongue states decreases
History
The modern study of the bilingual lexicon began in earnest in the 1980s, growing out of word recognition and spreading activation research. Kroll and Stewart’s Revised Hierarchical Model (1994) was the first comprehensive theoretical account of L1–L2 lexical organization, sparking two decades of empirical testing. The BIA and BIA+ models (Van Heuven, Dijkstra) brought connectionist computational modeling. Albert Costa’s work on language selection and inhibitory control mechanisms became influential in the 2000s. Contemporary work uses eye-tracking and ERP methods to trace real-time bilingual lexical activation with millisecond precision.
Common Misconceptions
- “Bilinguals keep their languages completely separate.” Extensive evidence shows both languages are always active to some degree, even when only one is being used. Complete separation is not achievable or necessary.
- “Bilinguals always think in one language.” Language of thought is fluid and context-dependent. Many bilinguals operate conceptually without mediation through any specific language form for highly familiar concepts.
- “Advanced bilinguals are equivalent to monolingual speakers.” Even highly advanced bilinguals show measurable differences in lexical access speed and cross-language interference compared to matched monolinguals.
Criticisms
The Revised Hierarchical Model has been criticized for underspecifying the mechanisms of L2 vocabulary learning and for treating L2-to-concept links as uniformly weak. More recent work shows that even early L2 learners acquire some direct conceptual mappings, especially for concrete nouns. Both the RHM and BIA models were primarily developed for European-language pairs and face challenges with typologically distant language pairs (e.g., Chinese–English). The inhibitory control model of bilingual language selection has been questioned by findings that show the switch cost can be symmetric rather than asymmetric.
Social Media Sentiment
Bilingual lexicon topics generate vibrant social media discussions, especially around code-switching, the experience of “thinking in L2,” and the frustration of L1 interference in production. Advanced learners celebrate milestone moments when they stop translating mentally — a subjective marker of the shift from L1-mediated to direct L2-concept access. Multilingual speakers often debate whether they have “different personalities” in each language, a claim that has some empirical basis in how cultural schemas are organized differently across languages.
Last updated: 2025-07
Practical Application
For L2 learners, the bilingual lexicon research translates into actionable guidance. The goal of vocabulary instruction is not merely knowing a translation equivalent, but building rich, direct conceptual associations in L2 — moving from L1-mediated access to direct access. This requires not just studying translation pairs but encountering words in rich, varied L2 contexts, in multiple modalities, and through active retrieval. Spaced repetition tools like Sakubo support the development of direct L2-to-concept links by systematically recycling words in varied contexts across intervals optimized to consolidate both semantic and phonological form access — exactly what the Revised Hierarchical Model predicts is needed for fluent bilingual lexical access.
Related Terms
- Psycholinguistics
- Mental Lexicon
- Lexical Access
- Word Recognition
- Spreading Activation
- Cross-Linguistic Influence
- Tip-of-the-Tongue
- Speech Production
- Spaced Repetition
- Second Language Acquisition
See Also
Research
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.
The paper proposing the Revised Hierarchical Model, demonstrating asymmetric translation times (L2→L1 faster than L1→L2 at lower proficiency) and providing the dominant theoretical framework for L1–L2 lexical organization.
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.
Proposed the BIA+ model, distinguishing language-non-selective identification processes from language-selective decision processes. Clarified when bilingual parallel activation produces observable interference and when it does not.
Kroll, J. F., Bobb, S. C., & Wodniecka, Z. (2006). Language selectivity is the exception, not the rule: Arguments against a fixed locus of language selection in bilingual speech. Bilingualism: Language and Cognition, 9(2), 119–135.
Reviews evidence that bilinguals do not selectively activate the target language at the lexical level — both languages compete — and that inhibitory control mechanisms operate at the decision stage. Influential in grounding the inhibitory control model of bilingual language use.