Hemiparetic and amputee walking often has asymmetrical step lengths and step times, and it is metabolically costlier than symmetrical able-bodied walking. Consequently, asymmetry has been suggested to account for the greater energy expenditure, but the metabolic cost of asymmetrical walking is poorly understood. Conversely, even though symmetry is metabolically optimal in able-bodied walking, it is also possible that asymmetrical gait parameters may be selected if they are optimal under imposed constraints. First, to understand the metabolic cost of asymmetry, we performed experiment 1 in which we recruited 10 able-bodied subjects to walk with a range of different combinations of asymmetrical step lengths and step times on a normal treadmill at 1.25 m s‚àí1. We found that the metabolic cost of step time asymmetry was more than twice the cost of step length asymmetry, but that the costs were not additive. Rather, the metabolic cost of walking with concurrent asymmetry in step lengths and step times was best explained by the cost of step time asymmetry alone. Second, to understand if asymmetrical gait parameters may be selected if they are energetically optimal, we performed experiment 2 in which we recruited 10 able-bodied subjects to walk with a range of different combinations of asymmetrical step lengths and step times on a split-belt treadmill in 3 conditions with speed-differences of 0.5 m s‚àí1, 1.0 m s‚àí1 and 1.5 m s‚àí1 at an average speed of 1.25 m s‚àí1. Across all speed-difference conditions, we found that subjects preferred to use asymmetrical step times that were energetically optimal, but that the preferred asymmetry in step lengths was sub-optimal. Overall, our results suggest that step time asymmetry is more effective than step length asymmetry to induce changes to metabolic cost and that symmetry is not necessarily energetically optimal. Instead, asymmetry in step times may be preferred when it is energetically optimal. Our results are based on asymmetry imposed on able-bodied walking and future research may test how the results generalize to other types of gait asymmetry in order to understand the implications for gait rehabilitation in which the goal often is to achieve a more symmetrical walking pattern.