Few studies have reported the effect of exercise intervention for improving postural control deficit in older adults at high risk of falling. We have developed a "Dual-task Switch Exercise (DSE)" program that focuses on gait initiation performance under the dual-task condition. The purpose of this study was to evaluate whether gait initiation performance could be improved by a specific exercise intervention. Eighteen participants were randomly assigned to either DSE or control groups. The DSE group received focused training to improve the ability to initiate movements quickly under the dual-task condition. The control group received steady-state walking training. After 30-min of seated training sessions, participants received 5-min individualized training sessions once a week for 24 weeks. In the pre- and post-training period, performance of the steady-state gait (10-m walking time) and gait initiation (reaction time, backward center of pressure (COP) displacement) were measured under the single-and dual-task conditions. The results of a randomized clinical trial showed that both groups showed improvement of steady-state walking time under the dual-task condition (main effect of time; p = 0.018). However, DSE was more effective in improving both the reaction time and backward COP displacement during gait initiation under the dual-task condition than control (interaction effect of time x group; reaction time, p = 0.015; COP displacement, p = 0.011). There were no significant differences in steady-state gait and gait initiation performance under the single-task condition between pre- and post-training in both groups. Only the specific exercise intervention improved gait initiation performance under the dual-task condition. (C) 2012 Elsevier Ireland Ltd. All rights reserved.