Primary progressive aphasia (PPA) is an age-related neurodegenerative syndrome characterized by isolated language impairment. Among its variants, the logopenic variant (lvPPA) and nonfluent/agrammatic variant (naPPA) present diagnostic challenges due to overlapping speech error profiles.This project explores whether specific phonemic features and language production errors pattern to these two variants, with the goal of improving diagnostic accuracy and informing neural stimulation treatment targets.

We analyzed lvPPA and naPPA patient speech data from the Western Aphasia Battery Object Naming Task. Productions were phonemically transcribed, time-aligned, and segmented in Praat textgrids to quantify phonological properties. Generalized linear models assessed accuracy and error frequency, while mixed effects models evaluated word attempts based on diagnosis and number of phonemes in the target word.

Findings indicated that lvPPA patients predominantly exhibited phonological errors, whereas naPPA patients exhibited more articulatory errors. The two variants struggled with place and manner of articulation to a similar extent, but lvPPA patients demonstrated higher accuracy in voicing. While both variants produced more phonemes as word length increased, lvPPA patients produced fewer phonemes overall and their production plateaued as words got longer. Additionally, lvPPA patients made significantly more word attempts than naPPA patients for longer words.

Future research will investigate how these error patterns influence differential responses to neural stimulation treatments, with the potential to guide targeted therapeutic interventions for PPA patients.