BACKGROUND: Most human movements are executed while in a state of postural instability. For this reason, instability training is a highly-specific method that is intended to improve balance and postural control. This research aimed to determine the effect of instability training on the balance of individuals with similar baseline abilities, who initiated training within different stability conditions. RESEARCH QUESTION: Does the level of difficulty with which instability training is performed determine improvements in balance? METHOD: A two-arm randomized trial was undertaken, for which 22 and 21 participants were included in the experimental and control groups, respectively. The experimental group performed balance training on unstable surfaces and the control group implemented the same training on stable ground. The primary outcome was the Y-Balance Test (YBT); the Emery test, Functional Reach test, and platform measures were the secondary outcomes. Confidence intervals were set at 95% and Cohen's f statistic was used to estimate effect size. RESULTS: Dynamic balance, as measured by the YBT, showed significant Time improvements in both groups for right (p <  0.001, f = 0.53) and left (p =  0.005, f=0.33) limbs. Similar results were found in the Emery test and Functional Reach test. No statistical Group and Time*Group interactions were found. None of the proposals modified the center of pressure excursions. SIGNIFICANCE: Instability training is a safe and effective approach to enhance balance. However, the findings deduced that the level of difficulty (instability) with which participants with similar abilities initiate the training is not a determining factor in the balance improvements achieved. It follows that instability training may be either used as an alternative or as an additional method to train for balance on a stable ground.