© 2016 Elsevier B.V. The purpose of this study was to compare the efficacy of four different home-based interventions on dual-task balance performance and to determine the generalizability of the four trainings to untrained tasks. Sixty older adults, aged 65 and older, were randomly assigned to one of four home-based interventions: single-task motor training, single-task cognitive training, dual-task motor-cognitive training, and dual-task cognitive–cognitive training. Participants received 60-min individualized training sessions, 3 times a week for 4 weeks. Prior to and following the training program, participants were asked to walk under two single-task conditions (i.e. narrow walking and obstacle crossing) and two dual-task conditions (i.e. a trained narrow walking while performing verbal fluency task and an untrained obstacle crossing while counting backward by 3 s task). A nine-camera motion capture system was used to collect the trajectories of 32 reflective markers placed on bony landmarks of participants. Three-dimensional kinematics of the whole body center of mass and base of support were computed. Results from the extrapolated center of mass displacement indicated that motor-cognitive training was more effective than the single-task motor training to improve dual-task balance performance (p = 0.04, ES = 0.11). Interestingly, balance performance under both single-task and dual-task conditions can also be improved through a non-motor, single-task cognitive training program (p = 0.01, ES = 0.13, and p = 0.01, ES = 0.11, respectively). However, improved dual-task processing skills during training were not transferred to the novel dual task (p = 0.15, ES = 0.09). This is the first study demonstrating that home-based dual-task training can be effectively implemented to improve balance performance during gait in older adults.