Bilingual/Multilingual Aphasia (BWA) is the loss of language comprehension and expression abilities that occurs after brain damage in bilingual or multilingual individuals. Impairment occurs in two or more languages, but the degree of impairment may differ for each language.
Epidemiology:
Historical background: Aphasia research has traditionally been monolingual-oriented, but in recent years research on bilingual aphasia has advanced rapidly1).
In the United States, approximately 300,000 cases of aphasia occur annually, of which about 45,000 are estimated to be bilingual aphasia. It is also known that about one-third of stroke patients develop aphasia.
In bilingual aphasia, language impairment is the main feature, but depending on the location of the brain lesion, visual field defects may also occur.
The pattern of visual field impairment varies depending on the location of the brain lesion.
Occipital Lobe Disorder
Homonymous hemianopia: The most common responsible lesion for homonymous hemianopia. Mainly caused by posterior cerebral artery ischemia.
Macular sparing: The macula may be spared from visual field loss.
Homonymous scotoma: Scotoma occurs only in part of the visual field.
Alexia: Caused by lesions in the left occipital lobe and splenium of the corpus callosum. Writing ability is preserved.
Temporal and Parietal Lobe Disorders
Upper visual field defect: In temporal lobe lesions, damage to Meyer’s loop results in a superior visual field defect.
Lower visual field defect: In parietal lobe lesions, damage results in an inferior homonymous visual field defect.
Higher visual function disorders: Visual agnosia, hemispatial neglect, Balint syndrome, etc.
Hemiparesis: In middle cerebral artery territory infarction, hemiparesis often occurs on the same side as the hemianopia.
Slit-lamp examination usually reveals no abnormalities. Dominant hemisphere (usually left hemisphere) lesions may cause language disorders. Ophthalmological evaluation of visual acuity, visual field, and pupillary reflex is necessary.
The occipital lobe functions as the visual starting point of language pathways, and when brain lesions extend to the occipital lobe, visual field defects occur. Homonymous hemianopia due to posterior cerebral artery ischemia is typical, and left occipital lobe damage may be accompanied by alexia in addition to homonymous hemianopia.
The main cause of bilingual aphasia is cerebral infarction, with about one-third of stroke patients developing aphasia 1). Other causes include trauma, epileptic seizures, and toxic metabolites.
The main types of cerebral infarction are as follows:
| Type | Characteristics | Main Risk Factors |
|---|---|---|
| Atherothrombotic infarction | Common in middle-aged and elderly, slow progression | Hypertension, diabetes, dyslipidemia |
| Cardioembolic stroke | Sudden onset, large infarct | Atrial fibrillation, valvular heart disease |
| Lacunar infarction | Small infarction (<15 mm) of a perforating artery | Hypertension (common in elderly) |
Main risk factors:
Evaluation of bilingual aphasia requires comprehensive assessment covering all languages spoken by the patient.
A comparison of major assessment tools is shown below.
| Assessment Tool | Target/Features | Structure |
|---|---|---|
| BAT (Bilingual Aphasia Test) | Parallel assessment of multilingual history and each language ability | Three parts: Part A (50 items), B (472 items), C (58 items) |
| CAT (Comprehensive Aphasia Test) | Multidimensional assessment of language, cognition, and QoL | Provides a comprehensive aphasia severity score |
| WAB / BNT | Multilingual versions available | Direct translation from English; may not reflect cultural and linguistic differences |
Three temporal phases of assessment (Appendix C):
Visual acuity, visual field testing, pupillary reflex, and slit-lamp microscopy are performed.
Treatment for bilingual aphasia is based on a multidisciplinary approach, with speech therapy as the core 1).
Semantic Feature Analysis (SFA)
Target: Word retrieval difficulty.
Method: Systematically generate and review semantic features of the target word, such as category, use, and physical characteristics.
Evidence: A meta-analysis in bilingual patients showed a moderate effect size for trained words (L1: TE=8.36).
Verb Network Strengthening Treatment (VNeST)
Target: Difficulty recalling verbs.
Method: Systematically generate and discuss the thematic roles (agent-patient) of verbs.
Features: Strengthens the entire semantic network of verbs.
Phonological Component Analysis (PCA)
Target: Phonological network impairment.
Method: Strengthen networks based on phonological characteristics of words.
Report: After 15 hours of therapy, bilingual patients showed better performance than monolingual patients.
This method is useful when multilingual rehabilitation resources are scarce. It utilizes cross-linguistic generalization (CLG), where treatment effects in one language transfer to another language 1).
This method recognizes language mixing as an adaptive strategy rather than pathological, and evaluates responses regardless of language. Spontaneous language choice has been reported to yield 84.9% correct responses, outperforming monolingual conditions (Dutch 79.7%, English 73.1%) 1).
It is considered effective for patients with visual field deficits.
Cross-linguistic generalization, where treatment effects in one language transfer to another, can be expected. However, meta-analyses show a limited effect size of g=0.14 compared to within-language effects (g=0.36). Age of acquisition (AoA) is the strongest predictor, with higher effect sizes for languages learned in adulthood.
Regarding language organization in the bilingual brain, the amalgamated theory is currently mainstream. Following early separate localization and common language representation theories, it is now understood that language is represented in both unique and shared brain regions1).
Anterior expressive area
Location: Motor-related area of the frontal lobe (Broca’s area).
Feature: Includes common areas for L1 and L2.
Effects of damage: Damage to the anterior expressive area tends to cause multilingual impairments.
Posterior Receptive Area
Location: Language receptive area of the temporal lobe (Wernicke’s area).
Characteristics: Includes areas specific to L2.
Effects of damage: Damage to the posterior receptive area tends to cause L2-selective impairments.
Broca-Wernicke-Geschwind model:
The basic pathway is: visual cortex → Wernicke’s area (comprehension) → arcuate fasciculus → Broca’s area (speech production) → motor cortex, with the Geschwind territory (inferior parietal lobule: supramarginal gyrus + angular gyrus) responsible for multimodal integration.
Concept of hybrid architecture:
Intraoperative cortical stimulation studies show that the same cortical site supports both languages. In BWA patients who have shifted to L2 dominance, increased activation in the left frontal and anterior cingulate areas is observed during weak L1 processing. This is explained by population encoding theory (the theory that the same cortical area supports multiple functions through different distributed activation patterns)1).
Language control mechanism: The dorsolateral prefrontal cortex (DLPFC), anterior cingulate cortex (ACC), and basal ganglia mediate language selection, inhibition, and set-shifting1).
The occipital lobe handles visual input as the starting point of language pathways. Posterior cerebral artery ischemia causes homonymous hemianopia and higher-order visual dysfunction, and damage to the left occipital lobe and splenium of the corpus callosum results in alexia.
Mechanism of cross-language generalization (CLG):
It is based on the sharing of conceptual and semantic representations. Strengthening one language spreads to the other language through shared neural encoding via spreading activation 1).
According to the current “mixed model,” languages are represented in both common and unique brain regions. In anterior expressive areas, L1 and L2 share common regions, while posterior receptive areas include regions specific to L2. This structure explains why lesion location leads to different patterns of impairment across languages.
Marte et al. (2025) conducted machine learning-based cross-language generalization prediction in a cohort of 48 Spanish-English bilingual aphasia patients, achieving an F1 score of 0.790. The WAB-R AQ score in the untreated language and cognitive function were the strongest predictors 1).
In the BiLex computational model, prediction accuracy with R² of 0.54 to 0.82 was achieved at the fourth treatment session. For CLG detection, specificity of 100% (7/7 cases) and sensitivity of approximately 80% (4/5 cases) have been reported 1).
ASR-based (automatic speech recognition) aphasia detection reported an F1 score of 0.99, and subtype classification an F1 score of 0.91 (Wagner et al.) 1).
Using GPT-based embeddings to identify three variants of primary progressive aphasia (PPA) showed a diagnostic agreement rate of 88.5% and classification accuracy of 97.9% (Rezaii et al.) 1).
In cross-language automatic evaluation, a model trained on English achieved 78% accuracy in French and 74% in Greek (Chatzoudis et al.) 1).
In bilingual aphasia patients, a shortening of mismatch negativity (MMN) latency is observed, whereas monolingual aphasia patients show prolonged latency (De Letter et al.). This suggests that bilingualism may confer cognitive reserve 1).
Applications of TMS (transcranial magnetic stimulation), tDCS (transcranial direct current stimulation), and tACS are advancing.
The combination of cerebellar tDCS plus behavioral therapy has shown improvements in trained and untrained words and cross-language generalization in L2 treatment (Coemans et al.)1).
Meta-analyses have shown that combining NIBS with language therapy yields superior effects compared to language therapy alone (Chai et al.)1).
Due to its non-invasive and portable nature, it is suitable for monitoring brain activity in clinical and natural settings, and is expected to be applied in the assessment and treatment outcome measurement of bilingual aphasia1).
