Primary progressive aphasia (PPA) is a dementia syndrome characterized by initial development of progressive language deficits in the absence of impairment in other cognitive domains. It has historically been difficult to assess the presence or nature of true memory deficits in this population due to interference from language disturbance on task performance. The Three Words Three Shapes test (3W3S) is a relatively easy memory task that evaluates both verbal and nonverbal memory within the same modality and assesses different aspects of memory, including incidental encoding, effortful encoding, delayed recall, and recognition. Persons with PPA show a material-specific dissociation in performance on 3W3S; specifically, deficits in incidental encoding and recall are limited to verbal, not nonverbal material, in PPA, with preserved recognition of both types of information. However, it is unknown whether this pattern persists over time as the disease progresses.

Participants were 73 participants enrolled in an observational PPA research study at the Mesulam Center for Cognitive Neurology and Alzheimer’s Disease (Mage = 66.75 years, SD = 6.77; Meducation = 16.11 years, SD = 2.38; 51% female). Participants were subtyped as semantic (n = 15), logopenic (n = 27), or agrammatic PPA (n = 31) based on Gorno-Tempini et al., 2011, using 3W3S and other neuropsychological measures as described previously. Participants were followed at 2-year intervals and tests were administered longitudinally. All participants in the current study had 3W3S scores from at least two research visits collected between September 2012 and September 2022.

There were no significant baseline group differences on 3W3S performance, except for better incidental encoding in the logopenic than the semantic group for shapes (p = .040) and words (p = .043). We then conducted a mixed measures ANOVAs to determine baseline within-person comparisons between words vs shapes. Within individuals, performance on incidental encoding, effortful encoding, and recognition was worse for words than shapes (ps < .01). There was an interaction between material and group for delayed recall (p < .001) such that there was a significantly larger discrepancy between word and shape recall in the semantic (Mdiff = -9.14) compared to logopenic (Mdiff = -3.07) and agrammatic groups (Mdiff = -2.13). Repeated measures ANOVAs determined changes in scores over time collapsed across PPA subtypes. Incidental encoding (ps = <.01), effortful encoding (ps < .05), and delayed recall (ps < .01) declined for both words and shapes over time. Copy and recognition of words (ps < .05), but not shapes declined over time.

The current results are consistent with prior findings of relative preservation of memory for nonverbal compared to verbal material in PPA as measured by 3W3S, especially in the semantic subtype. Learning and recall of words and shapes declined over time in all groups, whereas there was selective decline in copy and recognition of words compared to shapes. These results provide evidence of differential patterns of decline in certain aspects of memory over time in PPA and highlight the relative preservation of memory in this language-focused dementia even over time.