Neurolinguistic Approach to Translation
The same four levels of analysis as the cognitive approach — word, sentence, text, translation — read this time through what the bilingual brain actually does: how words are accessed, how sentences are processed and predicted in real time, how a discourse becomes a mental model, and what each translation strategy costs the mind that executes it.
Part of the COGNILANG international research programme directed by Prof. Mathieu Guidère.
From lexical access to the translator's brain
Lexical access in the bilingual mind
Where the cognitive approach asks what baggage a word carries, the neurolinguistic approach asks how the bilingual brain retrieves it. A translator reading a source word does not consult one dictionary: both languages activate in parallel, and the word's emotional and cultural charge is part of what gets activated — measurably, and before any conscious decision.
| Phenomenon to detect | What it means for translation |
|---|---|
| Parallel co-activation | Both languages of a bilingual are active even in monolingual reading; translation equivalents and false friends compete automatically. A "wrong" first equivalent is often the brain's fastest, not the text's best. |
| Cognate effects | Form-similar words across languages are accessed faster — which makes cognate translations tempting even when their connotations diverge (régime/regime, versatile/versatile). |
| Emotional resonance | Affectively loaded words engage emotion circuits more strongly in a speaker's dominant language; the same word may feel weaker in L2. A translator can under- or over-dose a text's charge simply because of language dominance. |
| Frequency & familiarity | Rare or culture-specific words cost more to access and to hold in working memory — cost that accumulates over a text and shapes which equivalents a tired translator reaches for. |
Real-time processing and prediction
On the sentence level, the brain does not wait for the full stop: it parses and predicts incrementally, word by word. The biases the cognitive approach detects in a sentence have neural signatures — the shortcuts are visible in how easily or laboriously the brain integrates each incoming word.
| Phenomenon to detect | What it means for translation |
|---|---|
| Semantic expectation (N400 family) | Words that fit the built-up expectation integrate cheaply; unexpected ones cost more. A sentence engineered so that a loaded conclusion feels "expected" is doing bias work the reader never notices — and a translation can inherit or dismantle that engineering. |
| Structural reanalysis (P600 family) | Garden-path constructions, deep embeddings, and heavy passives force the parser to backtrack. Syntactic complexity is not a style trait; it is a measurable processing cost the target reader will pay. |
| Working-memory load | Distance between dependent elements (subject and verb, pronoun and referent) taxes memory. Languages differ in what they make readers hold; reorganising a sentence redistributes that burden. |
| Anticipation | Skilled readers — and, in the extreme, simultaneous interpreters — survive by predicting what comes next. Sentence-level word order decides how early a prediction can be formed, which is why the same content can be easy in one language and punishing in another. |
From discourse to situation model
On the text level, the brain builds more than a record of the sentences: it builds a situation model — a running mental simulation of the world the text describes, with its actors, causes, places, and stakes. The cognitive framing detected in the other approach is, neurally, the shape this model is pushed into; once built, it filters everything read next.
Model construction
Readers integrate each sentence into the evolving model, updating actors and causal links. What the text makes easy to integrate becomes "what happened"; what it makes hard tends to be dropped from memory.
Inference generation
Coherence gaps are filled automatically with the reader's own knowledge. Two audiences with different background knowledge build different worlds from the same words — the central problem of translated texts.
Memory for gist
Days later, readers keep the model, not the sentences: the frame, the agents, the causal story. Whatever survived into the situation model is what the text will have "said" — regardless of its exact wording.
Strategies as neural routes
Translation recruits, on top of comprehension and production, a language control system: selecting one language, inhibiting the other, switching under constraint. The five strategies of the cognitive approach are, from this angle, different distributions of effort between the translator's brain and the reader's:
Literal
The low-control route: source structures are transposed with minimal restructuring. Cheap for the translator, but the processing costs the source syntax carried are exported, unexamined, to the target reader.
Domesticating
Maximal restructuring toward the target reader's expectations: predictions land, load drops, fluency rises. The translator pays in inhibition and reformulation effort what the reader saves in processing.
Foreignising
Deliberately preserved foreignness keeps the reader's prediction machinery slightly off balance — sustained, measurable extra effort that is precisely the point: the text keeps signalling that another mind, in another language, produced it.
Explicitating
The translator runs the inference for the reader and writes down the result. Coherence gaps close, situation models converge across audiences — at the price of longer texts and of substituting the translator's inference for the reader's own.
Neutralising
Affectively loaded items are swapped for low-arousal equivalents; emotion circuits stay quiet where the source engaged them. The most invisible strategy on the page and the most consequential for what the reader will feel — and remember.
Try our tool
Our pedagogical application walks a text through all four levels of the neurolinguistic approach — lexical access at word level, real-time processing at sentence level, discourse integration at text level, and the strategies at translation level — for a pedagogical and didactic aim: training translation students to reason about what a text does to the brain of its reader.
Opens in a new window. For teaching and self-training purposes; the research-grade pipeline is described on the Methodology page.