Brain-computer interface (BCI) provides a novel communication channel between human and external devices, which translates the brain activities directly to the computer commands. In this paper, an event-related potential (P300) based BCI paradigm was constructed to seek the effects of the continuous movement of the stimulus targets on the classification accuracy in virtual environment, which was seldom noticed previously. Two methods including fixed positions, broken-line path, with an origin stimulus' ring arrangement were set to look for performances differences. Experiment results show that the accuracy of motion type is not always lower than that for the conventional non-moving paradigms. This approach may help people train their ability focus on target.