Lateral stability is important to walking and critical to prevent falls. During normal (unrestrained and unperturbed) walking, humans prioritize corrections in step width more so than absolute lateral position in the frontal plane. The experimentally observed step-to-step dynamics resulting from these strategies is predicted from parsimonious multi-objective computational models. However, as the real-world environment is more complex, walking under different task conditions will introduce new or modified task goals, and it is expected that different combinations of lateral stepping regulation will arise. In this experiment, we investigated the extent to which humans can voluntarily modulate their lateral stepping regulation. Based on model predictions, we hypothesized that to maintain a constant value of one prescribed lateral stepping variable, young healthy people will exhibit inversely proportional shifts in the degree of regulation between that prescribed lateral stepping variable, and the complimentary variable. Twenty-four healthy adults walked on a treadmill in a virtual-reality environment for each of four conditions:​ normal walking, and while trying to maintain absolute lateral position, heading (i.e., direction), or step width. Real-time visual feedback of the participant’s error with respect to the desired goal value was given at each step. Time series of the lateral stepping variables were extracted and variability and statistical persistence (reflecting step-to-step regulation) were quantified. Participants exhibited less variability of the prescribed lateral stepping variable compared to normal walking during each of the three feedback conditions. To maintain either position or step width, participants increased regulation of the prescribed variable, and decreased regulation of its compliment. Therefore, specific task goals determined how lateral foot placements were regulated in predictable and systematic ways.