The purpose of this study was to propose a task-related circuits program for stroke patients and to test the difference in functional improvements between patients undergoing conventional physical therapy and those participating in a task-related circuits exercise program. The subjects were 10 stroke in-patients of the Korea National Rehabilitation Center in Seoul. We measured the following variables: Motor Assessment Scale (MAS), Berg Balance Scale (BBS), Tone Assessment Scale (TAS), speed of gait, rate of step, physiological costs index, age, weight, height, site of lesion, onset day and whether the subject participated in an exercise program. Collected data were statistically analyzed by SPSS 10.0/PC using descriptive statistics, Mann-Whitney U test, Wilcoxon rank sum test and Spearman's correlation. The results of the experiment were as follows: (1) In the pre-test and post-test for function, there was not a statistical significance between the group partaking in a task-related circuits program and the group of conventional physical therapy (p>.05). (2) In the MAS, BBS and speed of gait test, the group undergoing conventional physical therapy showed a statistical significance (p<.05). (3) In the MAS, BBS, speed of gait, PCI, TAS (passive, associated reaction, TAS total score), the group of task-related circuits program showed a statistical significance (p<.05). As a result, the group participating in a task-related circuits program had a more functional improvement than the group participating in conventional physical therapy. Therefore, an intervention recommended for a stroke patient would be a task-related circuits program consisting of a longer session of each task for a more improved functional recovery.