Graduation Year

2020

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Communication Sciences and Disorders

Major Professor

Jennifer Jones Lister, Ph.D.

Co-Major Professor

Jerri D. Edwards, Ph.D.

Committee Member

Supraja Anand, Ph.D.

Committee Member

Theresa Chisolm, Ph.D.

Keywords

Aging, Cognitive-Linguistic, Linguistic Features

Abstract

The research purpose of the present study was to (1) examine cognitive-linguistic features related to processing and production across a series of tasks that are representative of everyday discourse and (2) compare older adults with and without mild cognitive impairment (MCI) across linguistic features. Twenty-seven participants, including 12 individuals with- and 15 individuals without MCI, were enrolled from a larger study (Hudak et al., 2019). Cognitive status was initially assessed as part of the larger study using the Montreal Cognitive Assessment (MoCA; Nasreddine et al., 2005). Participants who scored ≤ 25 on the MoCA received a standardized neuropsychological evaluation and a physician’s examination to confirm or exclude a diagnosis of MCI.

For the current study, participants were additionally administered cognitive-linguistic measures. Measures consisted of obtaining a severity rating and complexity index from the Boston Diagnostic Aphasia Examination (BDAE) and extrapolating linguistic features from a series of speech-language samples (i.e., Semi-Structured Interview/Free Conversation adapted from the BDAE, Picture Description adapted from the BDAE, Story Narration task). For the purposes of this dissertation, the analytic sample for the BDAE and speech-language measures consisted of 16 participants (n=8 MCI, n= 8 Non-MCI) of the total sample of 27 older adults. This study also examined Lexical Decision-Making from the larger sample of 27 older adults with and without MCI.

Descriptive analysis of 98 linguistic features extrapolated across the three speech-language sample types was completed to identify potentially promising variables for further analyses. Thirty-eight variables had zero variance and were not further analyzed. Correlation analysis was conducted to assess the relationships among the remaining 60 linguistic features within type of speech-language sample. To further reduce the number of dependent variables for subsequent analyses, composites were created by combining linguistic variables that were highly correlated (i.e., r ≥ .6), as they are likely assessing the same skill. Next, correlation analysis of the linguistic variables to MoCA scores was completed. Results indicated seven linguistic variables with medium-strong correlations (r ≥ .45) with MoCA: these variables were examined in subsequent analyses. These seven variables included two composites related to (1) empty utterances, phonemic errors, repetitions and (2) filled pauses and indefinite terms and five individual linguistic variables related to agrammatic deletions, mean length of utterances (in both free conversation and story narration task), repetitions, and correct informational units. It was also noted that four of the seven variables correlated to MoCA performance were from the story narration-wordless picture book task.

Next, a multivariate analysis of variance (MANOVA) was conducted to examine group performance on the seven dependent variables (two composites and five individual linguistic variables) that were correlated with MoCA at r ≥ .45. Overall, individuals with and without MCI did not differ significantly across linguistic features, Wilk's Λ = .484, F (7, 8) = 1.218, p=.391, partial η 2 = .516. Results of follow-up univariate analysis of variance (ANOVA) indicated significant differences between older adults with and without MCI on linguistic features extrapolated from the wordless picture book - story narration task, related to agrammatic deletions, p=.020 and mean length of utterances, p=.020. Those with MCI had more agrammatic deletions and shorter mean length of utterances than those without MCI.

Finally, to examine group differences in Lexical Decision-Making average accuracy rate, a two (high versus low-density) x three (word, filler, pseudo) repeated measures ANOVA was completed. Results indicated no significant differences between those with and without MCI on the Lexical Decision-Making task, Wilk's Λ = .992, F(3, 23) = .065, p = .978, partial η2 = .008. Findings additionally indicated statistically significant effects for high versus low-density conditions, Wilk's Λ = .358, F (3, 23) = 13.750, p < .001, partial η2 = .642, but no significant interaction between high versus low-density conditions and MCI group, Wilk's Λ = .950, F (3, 23) = .403, p = .752, partial η 2 = .050. Average accuracy rates were better in the low-density condition for the word- and filler stimuli and better in the high-density condition for the pseudo stimuli.

Overall, findings indicate that linguistic features extrapolated from connected speech-language samples are useful in identifying cognitive-linguistic performance that is correlated to MoCA and shows group differences between persons with and without MCI. Specifically, the composite linguistic features correlated to MoCA score were related to (1) empty utterances, phonemic errors, repetitions and (2) filled pauses and indefinite terms. Individual linguistic variables correlated to MoCA were related to agrammatic deletions, repetitions, mean length of utterances, and correct informational units. The linguistic features that differed by MCI status were related to agrammatic deletions and mean length of utterances. These features affect the quality of syntax and speech-language processing and production. Deficits in these areas often result in increased communication breakdowns, losing their communicative turns, significant impacts on interpersonal relationships, and as a result have increased social isolation (M. Johnson & Lin, 2014; Mueller, Hermann, Mecollari, & Turkstra, 2018). Findings further support the importance of utilizing higher-level discourse tasks that facilitate observation of natural interactions of lexical units at the sentence or conversational level to detect cognitive-linguistic deficits. Specifically, the current findings indicated that the story narration – wordless picture book task was more effective in eliciting cognitive-linguistic features that significantly differ between persons with and without MCI. Additionally, the present study supports previous research highlighting the importance of developing optimal manual and/or automated ways to measure and analyze cognitive-linguistic features.

Download

Share

COinS