BACKGROUND: Many ambulatory stroke survivors are discharged to community settings where they will confront highly attention-demanding mobility situations. Very little is known about cognitive-motor interference during walking in acutely rehabilitating stroke survivors. AIM: To examine the magnitude and patterns of cognitive-motor dual-task interference at hospital discharge post stroke and explore the characteristics associated with distinct interference patterns. DESIGN: Observational study. SETTING: Acute care and inpatient rehabilitation centers. POPULATION: Forty-seven adults with hospital admission for stroke who were being discharged to home. Mean age was 59.5 years (SD=11.7) and median days post stroke was 14 (IQR=7-21). METHODS: Gait and cognitive (category naming task) performance were assessed under single- and dual-task conditions at hospital discharge. Dependent variables were gait speed, stride duration, stride duration variability, stride length, cadence, and correct response rate. Single and dual-task values were compared to assess the effects of dual-tasking on gait and category naming. Relative dual-task effects on gait speed and cognitive performance were plotted to identify patterns of dual-task interference. Exploratory analysis compared clinical characteristics between subgroups defined by pattern of interference. RESULTS: There were significant dual-task declines in gait speed, with corresponding dual-task effects on stride length, cadence, and stride duration, but no dual-task effects on stride duration variability or correct response rate. Dual-task effects on the category naming task were not significant due to large between-subject variability. Three predominant patterns of cognitive-motor interference were evident: mutual interference (37% of participants), gait interference (30% of participants), and cognitive-priority trade-off (22% of participants) - these patterns reflect the consistently observed negative dual-task effect on gait speed with positive, negative, or null effects on category naming. Participants who demonstrated cognitive-priority trade-off pattern of interference had significantly worse single-task category-naming performance, while those with mutual interference had greater overall stroke severity. CONCLUSIONS: Cognitive-motor dual-task interference on gait speed is highly prevalent in ambulatory stroke survivors with mild cognitive-linguistic impairments at hospital discharge. Variability in cognitive-task performance under dual-task conditions has implications for the reliability of dual-task assessment after stroke. CLINICAL REHABILITATION IMPACT: Assessment of dual-task walking is feasible as a predischarge evaluation of attention-demanding mobility function after stroke.