PURPOSE: To investigate the effect of different types of physical training on balance performance and whether improved balance correlates with improved walking performance. METHODS: Forty eight participants with chronic stroke were randomly assigned to aerobic training on cycle ergometer (AT-group), resistance training of the lower extremities (RT-group), or sham training of upper extremities (ST-group). Participants exercised 3 d/week for 12 weeks. Balance (Berg Balance Scale), peak oxygen uptake rate, isometric knee extensor strength, maximal gait speed, and 6 minute walk test were measured at baseline and after 12 weeks. RESULTS: Training specific effects were observed; the AT-group improved peak oxygen uptake rate by 15.5 (6.0-25.0)%, the RT-group improved non-paretic knee extensor strength by 35.1 (18.3-51.9)% and the ST-group improved non-paretic knee extensor strength by 8.9 (0.7-17.1)%. All groups improved balance (6.0 (95% CI: 3.2-8.8)%), maximal gait speed (10.2 (6.5-14.0)%), and 6 minute walk distance (12.4 (8.8-15.9)%) but balance improvements did not correlate with improvements in muscle strength, peak oxygen uptake rate, or walking. CONCLUSIONS: Physical exercise improves balance and walking performance, but improved balance is not a prerequisite for functional improvements in chronic stroke. Implications for Rehabilitation Aerobic training and progressive resistance training show small significant improvements in balance and walking, indicating a possible clinical relevance of these training modalities. Improvements in balance may not be a prerequisite for improvements in walking distance when assistive devices are allowed during walking tests.