When Dementia Speaks Another Language: The Fight to Make Aphasia Diagnosis Fair for the World’s 7,000 Tongues

Imagine watching your mother slowly lose her words—not all at once, but steadily, over years, as a disease erodes the very language centers of her brain. Now imagine that when you finally bring her to a clinic, the specialist hands her a test written in a language she barely speaks, built on grammar rules her native tongue doesn’t even possess. This is not a hypothetical scenario. It is the daily reality for millions of people worldwide who develop primary progressive aphasia (PPA), a devastating neurodegenerative condition, yet speak languages that the world’s diagnostic machinery was never designed to understand.

PPA is a rare but profound disorder in which language deteriorates progressively due to the death of brain cells in regions responsible for speech and comprehension, typically without the memory collapse seen early in classic Alzheimer’s disease. Since the landmark consensus criteria published by Gorno-Tempini and colleagues in 2011, clinicians worldwide have recognized three canonical variants: the non-fluent or agrammatic form (nfvPPA), marked by halting, effortful speech and broken grammar; the semantic variant (svPPA), in which word meaning itself dissolves; and the logopenic variant (lvPPA), characterized by impaired word retrieval and repetition. But a provocative new perspective article, published in Frontiers in Language Sciences by Mohamed Taiebine of the Euro-Mediterranean University of Fez in Morocco, argues that this entire diagnostic edifice rests on a hidden assumption—one that quietly fails the vast majority of humanity.

The problem, Taiebine contends, is that the world hosts approximately 7,168 living languages, yet English has become the absolute lingua franca of scientific research. Diagnostic tools, brain-imaging norms, and linguistic benchmarks have been built almost exclusively around English speakers—often from Western, Educated, Industrialized, Rich, and Democratic (WEIRD) populations. When these instruments are merely translated into other languages, they frequently lose their validity, because translation cannot capture the unique psycholinguistic, sociolinguistic, and cultural properties of the target language. A test designed to probe English syntax may be meaningless for a speaker of Arabic, Turkish, or an agglutinative language where meaning is packed into morphological suffixes rather than word order.

The consequences are not abstract. Screening tools rarely account for the nuances of non-dominant languages in bilingual and multilingual populations, which means early signs of decline can be missed entirely. The first symptoms of PPA often surface in the most deeply embedded, natural registers of a patient’s speech—the home dialect, the mother tongue—precisely the structures that standardized tests are least equipped to examine. In diglossic societies, where people use one variety of a language at home and another in formal settings, the disease may strike these registers disproportionately, eroding one while sparing the other. A patient might retain their first language while losing a second, or vice versa, complicating any assessment that assumes a monolingual baseline.

To address this, Taiebine proposes a radically multidimensional framework that maps PPA across six interconnected pillars: typological (the structural diversity of the patient’s language), topological (the specific brain regions undergoing atrophy), hodotopic (the location of language functions within brain networks), hodological (the integrity of white-matter tracts connecting those regions), biological (the underlying proteinopathies such as tau or TDP-43), and psychosocial (the environmental and cultural context). The idea borrows from the Research Domain Criteria (RDoC) model originally developed for psychiatry, which abandons rigid diagnostic categories in favor of dimensions that span molecules, circuits, and behavior. Applied to PPA, this means a diagnosis would integrate not just which brain areas are shrinking, but how the patient’s particular linguistic system is wired into those areas.

This matters because the brain’s language architecture is not language-neutral. The hodological level—the wiring—illustrates the point vividly. In classic conduction aphasia and in the logopenic variant of PPA, damage often involves the dorsal pathway, including the superior longitudinal fasciculus and the arcuate fasciculus, the fiber bundles connecting speech production and comprehension regions. But for a bilingual speaker of an understudied language, Taiebine hypothesizes, a repetition deficit might reflect network interference between languages rather than the classic arcuate fasciculus breakdown seen in monolinguals. He is careful to flag that such neuroanatomical correlations remain speculative and require far more investigation in speakers of understudied languages—yet the very possibility exposes how Anglo-centric assumptions could lead clinicians astray, misattributing dialect-specific vocabulary loss to general memory decline rather than focal language breakdown.

The framework also intersects with a broader scientific revolution already underway. Artificial intelligence and machine learning are transforming PPA diagnosis by analyzing connected speech—the natural, flowing language patients produce—rather than relying solely on structured tests. Large language models can identify natural clusters within patient speech that align with independent clinical diagnoses, and a machine-learning system called NeuroScreen has achieved up to 91 percent accuracy in classifying PPA and its variants by differentiating neurological conditions through language analysis. AI is even being applied to neural assessment through EEG and cortical speech envelope tracking, offering objective measures of language network plasticity. Crucially, Taiebine argues that this same technology could become the great equalizer: rather than translating thousands of tests into thousands of languages, AI and large language models could offer a scalable, low-cost pathway to map dialectal nuances directly, detecting subtle acoustic markers of motor-speech disorders across multilingual clinical settings.

Existing tools point the way forward, even as they reveal the scale of the gap. The Mini Linguistic State Examination (MLSE) has emerged as a pivotal brief screening instrument, evaluating five core domains—motor speech, phonology, semantics, syntax, and verbal working memory—through eleven subtests in roughly fifteen to twenty minutes, with cross-linguistic validity demonstrated in more than twenty languages. A newly developed web-based version promises automated scoring, potentially reducing clinician burden and standardizing diagnostic profiling across international trials. Yet its automated scoring currently remains restricted to English. In parallel, the Dépistage Cognitif de Québec (DCQ), validated in French, English, and Spanish, screens for atypical dementias including PPA and covers social cognition alongside language—a notable exception that still reaches only a fraction of global language families. As an interim measure, Taiebine suggests pairing these tools with the Bilingual Aphasia Test (BAT), which can document a patient’s sociolinguistic history, age of language acquisition, and dialect usage patterns before any standardized assessment is administered, thereby attenuating bias at the intake stage.

The article is equally candid about the limits of its own proposal. The interaction between specific linguistic structures and PPA pathology is strongly motivated by theoretical linguistics, but empirical evidence across all minority language populations is still emerging. Establishing typologically informed frameworks—prioritizing, for example, tonal variance in tonal languages or agglutinative morphology in Arabic rather than literal translations of English syntactic structures—will require rigorous, ongoing cross-linguistic normative research. The author calls for shared, open-access repositories of diverse speech samples and normative cross-linguistic data, so that clinicians and researchers worldwide can build assessments grounded in the actual linguistic reality of their patients rather than in borrowed norms.

What elevates this argument beyond technical reform is its framing as a matter of cognitive rights. Linguistically fair screening, Taiebine writes, is not merely a clinical improvement but a profound commitment to the dignity of diverse populations—a blueprint for an aphasiology in which the depth of a patient’s linguistic identity enhances, rather than hinders, the accuracy of their diagnosis and their quality of life. As dementia rates rise globally and populations age across every continent, the question is no longer whether the world’s languages deserve a place in the neurology clinic, but whether the science of the brain can finally catch up to the dazzling diversity of the minds it seeks to heal.

Subject of Research: Linguistically fair screening and transdiagnostic assessment of primary progressive aphasia in understudied and minority languages

Article Title: Toward linguistically-fair screening and assessment of primary progressive aphasia: a transdiagnostic approach for understudied and minority languages

Article References: Taiebine, M. (2026). Toward linguistically-fair screening and assessment of primary progressive aphasia: a transdiagnostic approach for understudied and minority languages. Frontiers in Language Sciences, 5, Article 1782730. https://doi.org/10.3389/flang.2026.1782730

Image Credits: AI Generated

DOI: 10.3389/flang.2026.1782730

Keywords: primary progressive aphasia, aphasia screening, minority languages, transdiagnostic framework, Research Domain Criteria, bilingualism, diglossia, artificial intelligence, machine learning, neurodegeneration, cross-linguistic assessment, health equity