BACKGROUND: Balance disability measurements routinely used to identify fall risks in frail populations have limited value in the early detection of postural stability deficits in community-living older adults. The objectives of the study were to 1) measure performance-based limits of stability (LOS) in community-living older adults and compare them to theoretical LOS computed from data proposed by the Balance Master system, 2) explore the feasibility of a new measurement approach based on the assessment of postural stability during weight-shifting tasks at performance-based LOS, 3) quantify intra-session performance variability during multiple trials using the performance-based LOS paradigm. METHODS: Twenty-four healthy community-living older adults (10 men, 14 women) aged between 62 to 85 (mean age +/- sd, 71.5 +/- 6 yrs) participated in the study. Subjects' performance-based LOS were established by asking them to transfer their body weight as far as possible in three directions (forward, right and left) without changing their base of support. LOS were computed as the maximal excursion of the COP in each direction among three trials. Participants then performed two experimental tasks that consisted in controlling, with the assistance of visual feedback, their centre of pressure (COP) within two predefined targets set at 100% of their performance-based LOS. For each tasks 8 trials were performed. Ground reaction forces and torques during performance-based LOS evaluation and experimental tasks were recorded with a force plate. Sway area and medio-lateral mean COP displacement speed variables were extracted from force plate recordings. RESULTS: Significant differences between theoretical LOS computed from maximum leaning angles derived from anthropometric characteristics and performance-based LOS were observed. Results showed that a motor learning effect was present as the participants optimized their weight-shifting strategy through the first three trials of each task using the visual biofeedback provided on their COP. Reliable measures of control of postural stability at performance-based LOS can be obtained after two additional trials after the learning phase (0.69 > ICC > 1.0). CONCLUSION: Establishing performance-based LOS instead of relying on estimations of theoretical LOS offers a more individualized and realistic insight on the true LOS of an individual. Performance-based LOS can be used as targets during weight-shifting postural tasks with real time visual feedback of the COP displacement to assess postural stability of community-living older adults. In order to obtain reliable results, a learning phase allowing subjects to learn how to control their COP displacement is needed.