Purpose: This study examined the test-retest stability of select word-retrieval measures in the discourses of people with aphasia who completed a 5-stimulus discourse task. Method: Discourse samples across 3 sessions from 12 individuals with aphasia were analyzed for the stability of measures of informativeness, efficiency, main concepts, noun and verb retrieval, word-finding difficulty, and lexical diversity. Values for correlation coefficients and the minimal detectable change score were used to assess stability for research and clinical decision making. Results: Measures stable enough to use in group research studies included the number of words; the number of correct information units (CIUs); the number of accurate-complete, accurate-incomplete, and absent main concepts; the percentage of T-units that had word-finding behaviors of any kind; the percentage of T-units that contained empty words; and a lexical diversity measure. Words per minute, CIUs per minute, and the percentage of T-units that contained time fillers or delays were sufficiently stable to use when making clinical decisions about an individual. Conclusion: Although several of the measures demonstrated acceptable stability for group research studies, relatively few were sufficiently stable for making clinical decisions about individuals on the basis of a single administration.