Disentangling phonology from phonological short-term memory in Alzheimer's disease phenotypes

Abstract

Background: Impaired phonological short-term memory is a core feature of the logopenic variant of primary progressive aphasia (lvPPA), but it is not clear whether a core phonological processing deficit is also present.

Methods: We asked three questions: (i) beyond short-term memory impairment, do lvPPA patients have an impairment within phonology itself?; (ii) is their performance in working memory and naming reflective of this phonological impairment?; and (iii) is their repetition performance related to structural and functional differences in key language-dominant regions? We compared non-word and word repetition and short-term memory performance in patients with typical Alzheimer's disease (tAD), lvPPA per consensus criteria, and others who previously satisfied definitions of lvPPA but had progressed with multi-domain cognitive impairments (lvPPA+).

Results: Bayesian analyses revealed no group differences in phonological tasks of word and non-word repetition. We found very strong evidence for an effect of self-reported hearing loss on word and non-word repetition, but not multi-syllabic word/phrase repetition. A comparison of phonological versus working memory and naming tasks produced either no evidence or evidence for no correlation. Beyond the expected grey matter reductions in patients relative to controls, there was anecdotal evidence for an association between non-word repetition and functional connectivity between dorsal premotor and posterior superior temporal gyrus regions in patients.

Conclusions: Our results indicated that, in the absence of self-reported hearing loss, patients did not exhibit impairments in tasks tapping "pure" phonological processing. Our results suggest that instead of having a core phonological impairment, lvPPA patients have a working memory/buffering impairment.

Keywords: Information buffering impairment; Logopenic variant of primary progressive aphasia; Non-word repetition; Phonological processing; Verbal working memory.

Fig. 1

Effect of self-reported hearing loss on word (left), non-word (middle) and multisyllabic word/phrase (right) repetition performance. Bayesian group x hearing loss ANOVA revealed moderate evidence for an effect of self-reported hearing status (F(1,31) = 14.42, P < 0.001, BF = 9.46) and group (F(3,31) = 10.10, P < 0.001, BF = 4.90), and strong evidence for a hearing-by-group interaction (F(3,31) = 8.04, P < 0.001, BF = 11.84) on word repetition. Post hoc Bayesian t-test revealed moderate evidence for a difference in those with and others without hearing loss in the lvPPA + group only (t = 3.18, P = 0.02, BF = 3.25). The letter ‘a’ indicates the level of Bayesian evidence: a = moderate (3 < BF < 10). For non-word repetition, Bayesian group x hearing loss ANOVA revealed very strong evidence for an effect of self-reported hearing status only (F(1,31) = 18.25, P < 0.001, BF = 49.30). For ACE-R multisyllabic word/phrase repetition, there was no evidence for an effect of hearing status or hearing-by-group interaction (0.33 < BF < 1). BF, Bayes factor; lvPPA, logopenic variant primary progressive aphasia; tAD, typical Alzheimer’s disease

Fig. 2

Associations between PALPA non-word and ACE-R multisyllabic word/phrase repetition and measures of phonology and verbal working memory. Panels indicate correlations between (A) non-word repetition and digit span forward (left) and backward (right) in the whole group, (B) non-word repetition and the additional items correct with phonemic cues on the Boston Naming Test (BNT) in the lvPPA/lvPPA + combined group, (C) non-word repetition and the number of phonological paraphasias per 100 words during connected speech in the lvPPA + group (left) and all groups (right), (D) ACE-R multisyllabic word/phrase word repetition and digit span forward (left) and backward (right) in the whole group, (E) ACE-R multisyllabic word/phrase word repetition and the additional items correct with phonemic cues on the BNT in the lvPPA/lvPPA + combined group, and (F) ACE-R multisyllabic word/phrase word repetition and the number of phonological paraphasias per 100 words during connected speech in the lvPPA + group (left) and all groups (right). In the scatterplots for the whole group, the dots indicate the number of participants (e.g., larger dots portray more participants who have the same scores). In each scatterplot, the Bayesian Pearson’s correlation values, their associated Bayes factors (BFs), and Frequentist P values are shown. ACE-R, Addenbrooke’s Cognitive Examination – Revised; lvPPA, logopenic variant primary progressive aphasia

Fig. 3

Associations between non-word and multisyllabic word/phrase repetition scores and region of interest (ROI) grey matter volumes and functional connectivity. Panel (A) shows the association between ACE-R multisyllabic word/phrase repetition and grey matter volumes in the left posterior superior temporal gyrus (pSTG; left), anterior temporal lobe (ATL; middle), and IFG (right) in the whole group. Panel (B) shows the association between ACE-R multisyllabic word/phrase repetition and grey matter volumes in pSTG (left), ATL (middle), and the association between non-word repetition and pSTG-dorsal premotor (dPM) resting state functional connectivity after controlling for grey matter volumes in the pSTG and dPM (right) in patients only. Each scatterplot shows the results of a Bayesian multiple regression, including the beta coefficients, associated Bayes factors (BFs), and Frequentist P values. ACE-R, Addenbrooke’s Cognitive Examination – Revised; FC, functional connectivity

Fig. 4

Cognitive scores over time in six patient participants. The x-axis represents the time from to initial to follow-up assessment in months. The filled colour dots represent the group average scores on each test. The connected lines indicate the scores of each participant who completed the same test at two time-points. Note: The lines merely connect the same participants across two time-points and do not suggest a linear decline over time