Word-Form and Lemma Dependency Networks of Connected Speech in Probable Alzheimer's Disease: Morphological Impacts on Syntactic Topological Metrics.

BACKGROUND: Alzheimer's disease causes progressive cognitive and linguistic deterioration. Few studies investigate how morphological inflections shape syntactic network topology by comparing word-form and lemma networks. METHOD: Spontaneous language samples were collected by the Cookie Theft picture description task from 68 individuals with probable Alzheimer's disease (PA) and 68 healthy controls. RESULTS: In both groups, lemma networks had fewer nodes/edges, higher average degree, density, and clustering coefficient, and shorter average path length than word-form networks. All networks showed small-world properties; only healthy networks obeyed scale-free rules, whereas patient networks deviated slightly. Controls had more nodes and edges, while patients demonstrated higher density and clustering coefficients. Intergroup average degree differences were limited to lemma networks (higher in controls). Average path length and diameter were similar across groups. Significant topological differences were accompanied by modest effect sizes (r = 0.21-0.29), likely arising from variability in spontaneous speech. Function words and basic verbs formed core nodes in both network types. CONCLUSION: This study identifies quantitative syntactic network biomarkers for linguistic impairments in early PA. Such topological measures may combine with clinical indicators for risk prediction and inform targeted early cognitive interventions.