PROSODIC CHUNKING IN OUT-LOUD READING: A WINDOW INTO COGNITIVE FUNCTION IN MILD-TO-MODERATE ALZHEIMER’S DISEASE

Abstract

Recent psycholinguistic research on healthy individuals reported that speech production planning in reading is associated with cognitive functioning (CF). Speakers with smaller working memory (WM) spans tend to split their sentences into multiple speech units (prosodic chunking) as they may not gather as much advance planning information as high WM speakers prior to speaking (Swets et al, 2014; Petrone et al, 2011; Becic et al, 2003). We investigate herein prosodic chunking in out-loud reading as a function of cognitive capacity in mild-to-moderate Alzheimer's disease (AD), particularly in relation to difficulties with basic attention processes, processing speed and WM. Twenty adults with mild-to-moderate AD and twenty healthy volunteers matched in gender, age and education level (Table 1) were administered the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS), the Montreal Cognitive Assessment (MoCA) and the Sustained Attention to Response Task (SART). They performed two speech tasks: reading isolated sentences varying in length and syntactic complexity, and engaging in a collaborative game with caregivers. Several speech features (Table 2) were extracted. Thirteen participants in each group also underwent MRI/fMRI scans. Mann-Whitney tests showed significant differences between the two groups in all the neuropsychological scores (Table 1) and most read speech measures (Table 2). AD read speech is marked by a higher number of speech units, pauses and dysfluences, along with shorter units, longer sentences and slower articulation rates. Preliminary analyses also showed significant correlations between the number of speech units, pauses and hesitations, and the RBANS and MOCA scores and some of the attention and WM scores (rSpearman = 0.61 − 0.81). Initial results suggest that prosodic chunking may be associated with CF. This corroborates current literature for different neuropathologies including AD (Dodge et al, 2015; König et al 2015; De Looze et al; in revision). Shorter speech units suggest a narrowing down of the scope of planning with greater WM deficits. Prosodic chunking may provide valuable information for clinicians towards the assessment of CI in AD. Current analysis is exploring the relationship between speech features and structural neuroimaging metrics to probe the neuroscientific basis of such deficits.