Semantic control across Alzheimer’s disease phenotypes: cognitive mechanisms and neural correlates

Introduction – Semantic cognition, our ability to understand meaning, relies on two distinct but interacting systems: one involved in knowledge representation and another in control processes. This control system is thought to comprise at least two mechanisms: controlled retrieval, which facilitates access to information relevant to a specific goal or context, and selection, which is engaged to resolve competition among multiple simultaneously active semantic representations. The present study aimed at understanding if, and how, controlled semantic cognition is impaired in Alzheimer’s disease (AD), and the neuroanatomical correlates of these deficits. We supposed that in AD, semantic deficits could reflect damage to the retrieval and/or the selection mechanism and could differ according to the clinical phenotype of AD. Methods – Our sample included the typical amnestic (ADtyp, N = 11) clinical presentation as well as the atypical language-dominant variant (logopenic variant primary progressive aphasia, lvPPA, N = 17). 16 healthy controls also took part. We examined object use knowledge in a task that manipulated the levels of both controlled-retrieval and selection semantic control mechanisms. Results – Increased semantic control demands worsened performance in all groups including controls. Both patient groups were impaired in controlled retrieval, but not selection, compared to controls. Furthermore, lvPPA scored consistently lower than controls and ADtyp in all conditions of the experimental task, even the condition with the lowest semantic control demands, pointing to broader semantic difficulties beyond controlled-retrieval and selection… Exploratory analyses suggested the two mechanisms were linked to partially distinct predominant cognitive associations, controlled-retrieval with tests of semantic-knowledge and selection with executive function. Neuroimaging showed partially overlapping but distinct correlates for the two mechanisms, potentially reflecting differing reliance on semantic versus domain-general control brain networks. Conclusion – Our results refine current models of semantic dysfunction in AD, highlighting involvement of both domain-general and more specialized control mechanisms, with potential implications for differential diagnosis, symptom management, and rehabilitation planning.