Walking humans respond to pulls or pushes on their upper body by changing where they place their foot on the next step. Usually, they place their foot further along the direction of the upper body perturbation. Here, we examine how this foot placement response is affected by the average step width during walking. We performed experiments with humans walking on a treadmill, both normally and at five different prescribed step widths. We prescribed step widths by requiring subjects to step on lines drawn on the treadmill belt. We inferred a linear model between the torso marker state at mid-stance and the next foot position. The coefficients in this linear model (which are analogous to feedback gains for foot placement) changed with increasing step width as follows. The sideways foot placement response to a given sideways torso deviation decreased. The fore–aft foot placement response to a given fore–aft torso deviation also decreased. Coupling between fore–aft foot placement and sideways torso deviations increased. These changes in foot placement feedback gains did not significantly affect walking stability as quantified by Floquet multipliers (which estimate how quickly the system corrects a small perturbation), despite increasing foot placement variance and upper body motion variance (kinematic variability).