The initiation of human walking involves postural motor actions for body orientation and balance stabilization that must be effectively integrated with locomotion to allow safe and efficient transport. Our ability to coordinately adapt these functions to environmental or bodily changes through error-based motor learning is essential to effective performance. Predictive compensations for postural perturbations through anticipatory postural adjustments (APAs) that stabilize mediolateral (ML) standing balance normally precede and accompany stepping. The temporal sequencing between these events may involve neural processes that suppress stepping until the expected stability conditions are achieved. If so, then an unexpected perturbation that disrupts the ML APAs should delay locomotion. This study investigated how the central nervous system (CNS) adapts posture and locomotion to perturbations of ML standing balance. Healthy human adults initiated locomotion while a resistance force was applied at the pelvis to perturb posture. In experiment 1, using random perturbations, step onset timing was delayed relative to the APA onset indicating that locomotion was withheld until expected stability conditions occurred. Furthermore, stepping parameters were adapted with the APAs indicating that motor prediction of the consequences of the postural changes likely modified the step motor command. In experiment 2, repetitive postural perturbations induced sustained locomotor aftereffects in some parameters (i.e., step height), immediate but rapidly readapted aftereffects in others, or had no aftereffects. These results indicated both rapid but transient reactive adaptations in the posture and gait assembly and more durable practice-dependent changes suggesting feedforward adaptation of locomotion in response to the prevailing postural conditions.