AIM: To examine test-retest reliability of time and frequency domain heart rate variability (HRV) in patients 1 month after stroke during rest, paced breathing and light-to-moderate physical activity. METHODS: Fifteen patients up to 1 month after stroke underwent two measurements of HRV, with the measurements 4 days apart. Measurements took place under three conditions while sitting: (1) at rest with self-select breathing frequency, (2) paced breathing and (3) cycling while sitting. Reliability was assessed statistically by calculating intraclass correlation coefficients (ICC), standard error of measurement and coefficient of variance (CV). RESULTS: The relative reliability was found to be good-to-excellent for SDNN (ICC: 0.86-0.91), RMSSD (ICC: 0.81-0.87) and HF (ICC: 0.91-0.94) in all three conditions and poor for LF at rest and paced breathing (ICC: 0.43-0.47). The absolute reliability for all measures was found to be poor (CV >15%). CONCLUSIONS: HRV can be reliably assessed at rest, paced breathing and light-to-moderate physical activity for identifying differences between patients, while individual changes in autonomic functioning exhibited large random variations between test-retest measurements.