Graduation date: 2007Falls, particularly those due to slips, constitute a major health risk for older adults. Past research has found that individuals can learn not to fall through repeated exposure to a slipping perturbation. Slips can be unpredictable, however. The goal of this study was to determine whether similar adaptations would occur for repeated slips in unpredictable directions. A sliding platform was used to cause 30 apparently healthy young adults to lose their balance while rising from a semi-squatted position during a lifting task. A series of 27 perturbations were administered in unpredictable directions, forcing participants to either fall or step to recover. Participants were grouped based on whether they fell or successfully recovered in response to the first perturbation. Twenty-two variables quantifying proactive and reactive behavior were derived from motion capture data and compared between groups and across the first, second, and fifth perturbations that caused a backward balance loss, as well as a final “predictable” perturbation. Eight participants fell upon first exposure to the perturbation. Fallers had a more rearward center of mass at recovery step liftoff and lower hip height at step touchdown than those who recovered. The fallers’ hips were also dropping much faster at step liftoff and touchdown than those who recovered. Hip height at step touchdown was able to predict 100% of falls upon the first perturbation exposure in a stepwise, logistic regression model. All of the fallers adapted their reactive responses, despite the unpredictability of the perturbation direction, leading to successful recoveries in all trials after the first. By the second backward balance loss, the fallers made improvements to each variable associated with falling to become similar to those who recovered on the first exposure. Over the course of the perturbations, the fallers also increased their hip height at step liftoff and maximum hip height during the initial recovery step. Those who recovered on the first perturbation also made adjustments across perturbations. Both groups decreased their response time to step liftoff, placed their stepping foot further behind the center of mass, and increased the lateral component of their initial recovery step. The “predictable” case differed primarily by the presence of proactive adjustments. Repeated exposures to this unpredictable perturbation produced evidence of learning and adaptation of the reflexive response for both those who fell and those who recovered as a result of their first exposure. The fact that learning does occur, even in unpredictable situations, may have application to the design of fall prevention programs.