The authors investigated metabolic and attentional energy costs as participants (N = 6) practiced in-phase, antiphase, and 90 degrees-phase cycling (order counterbalanced) on independent bicycle ergometers, with resistance (40 W/ergometer) and frequency (40 rpm) held constant. Coordination stabilized and became more accurate for all 3 cycling modes, as shown by measures of relative phase, but that collective variable could not account for other relevant attributes of the multifaceted motor behavior observed across the 3 coordination modes. In-phase and antiphase cycling were similar in stability and accuracy, but antiphase had the lowest metabolic and attentional energy costs. Because both homologous muscle action and perceptually symmetrical oscillations coincided in the in-phase mode, the absence of predominance of the in-phase pattern showed that neither of those musculoskeletal and perceptual factors exclusively determined the strongest attractor of the coordination dynamics. Both metabolic and attentional costs declined with practice, consistent with the hypothesis that adaptive motor behavior is guided by sensory information concerning the energy demands of the task. Attentional cost was influenced not only by the information-processing demands of kinematic stability but also by the metabolic energy demands. Metabolic energy cost appeared to be the crucial determinant of the preferred solution for this coordination task.