Age-related Cognitive Decline Leads to Greater Entropic Descriptiveness as Measured in Connected Speech Tasks Within Biomarker-negative Cognitively Normal Individuals
Objective:
Speech production is complex and requires an equally complex array of cognitive processes to work in concert, resulting in age-related difficulties as these processes slow. We aim to use speech behavior in connected speech tasks, including picture descriptions and open-ended responses, to measure the continuous trajectory of decline and examine language changes due to aging.
Background:
Measures of connected speech have been used to identify patterns of age-related decline, including increased disfluencies (especially “uh” utterances), and altered diversity and part-of-speech usage. Most studies, however, determine these effects through categorical analysis, assigning categories such as “young-old” and “old-old” to inconsistent age ranges. As a result, conflicting findings emerge; for example, some find decreased diversity due to age, while others report increased diversity. By examining changes continuously across age ranges, we hope to address this problem.
Design/Methods:
53 healthy participants (age range: 55 to 85, mean: 70.3±8.1) were administered three connected speech tasks: picture descriptions, narrative recall, and open-ended responses. Transcribed speech was analyzed to produce measures of lexical complexity, lexical diversity, lexical frequency, parts of speech ratios, entropy, disfluencies, and speech rate. Each measure was examined continuously using linear regressions and correlations.
Results:
Increases in entropy and filled disfluency rate and decreases in closed-to-open class word ratios were found across all tasks, with increases in adjective-to-verb ratio differences found in picture description tasks. Post-disfluency latencies and open-class word use also increased continuously.
Conclusions:
Increasing adjective-to-verb ratios and entropy suggest the production of extraneous descriptions of each content unit, and decreasing closed-to-open class ratios indicate a tendency towards less complex language. Increased filled and longer disfluencies across ages reflect increased searching behavior, especially for content (open-class) words. Overall, we observe that increased age leads to a decline in conciseness with a tendency to over describe, and an increase in processing time when searching for these descriptions.