To quantify the energetic cost of maintaining lateral balance during human running, we provided external lateral stabilization (LS) while running with and without arm swing and measured changes in energetic cost and step width variability (indicator of lateral balance). We hypothesized that external LS would reduce energetic cost and step width variability of running (3.0 m/s), both with and without arm swing. We further hypothesized that the reduction in energetic cost and step width variability would be greater when running without arm swing compared with running with arm swing. We controlled for step width by having subjects run along a single line (zero target step width), which eliminated any interaction effects of step width and arm swing. We implemented a repeated-measures ANOVA with two within-subjects fixed factors (external LS and arm swing) to evaluate main and interaction effects. When provided with external LS (main effect), subjects reduced net metabolic power by 2.0% (P = 0.032) and step width variability by 12.3% (P = 0.005). Eliminating arm swing (main effect) increased net metabolic power by 7.6% (P < 0.001) but did not change step width variability (P = 0.975). We did not detect a significant interaction effect between external LS and arm swing. Thus, when comparing conditions of running with or without arm swing, external LS resulted in a similar reduction in net metabolic power and step width variability. We infer that the 2% reduction in the net energetic cost of running with external LS reflects the energetic cost of maintaining lateral balance. Furthermore, while eliminating arm swing increased the energetic cost of running overall, arm swing does not appear to assist with lateral balance. Our data suggest that humans use step width adjustments as the primary mechanism to maintain lateral balance during running.