The purpose of this study was to characterize test-retest reliability of discourse measures across a battery of common tasks in individuals with aphasia and prospectively matched adults without brain damage. We collected spoken discourse during five monologue tasks at two timepoints (test and retest; within 2 weeks apart) in an aphasia group (n = 23) and a peer group with no brain damage (n = 24). We evaluated test-retest reliability for percentage of correct information units, correct information units per minute, mean length of utterance, verbs per utterance, noun/verb ratio, open/closed class word ratio, tokens, sample duration (seconds), propositional idea density, type-token ratio, and words per minute. We explored reliability's relationship with sample length and aphasia severity. Rater reliability was excellent. Across tasks, both groups demonstrated discourse measures with poor, moderate, and good reliability, with the aphasia group having measures demonstrating excellent test-retest reliability. When evaluating measures within each task, test-retest reliability again ranged from poor to excellent for both groups. Across groups and task, measures that appeared most reliable appeared to reflect lexical, informativeness, or fluency information. Sample length and aphasia severity impacted reliability, and this differed across and by task. We identified several discourse measures that were reliable across and within tasks. Test-retest statistics are intimately linked to the specific sample, emphasizing the importance of multiple baseline studies. Task itself should be considered an important variable, and it should not be assumed that discourse measures found to be reliable across several tasks (averaged) are likewise reliable for a single task. https://doi.org/10.23641/asha.23298032.
Read full abstract