In this study, psychoacoustic experiments are performed with sound stimuli simulating children running in order to investigate the psychoacoustic effect of the duration, number of impacts, and level variation of the impact sound. For the sound stimuli, the jumping sound of an eight-year-old child is recorded in a test building similar to a real apartment building. Using the recorded single impact sound (SIS), eight different multi-impact sources are produced, which simulate the sound of children running by varying the duration, number of impacts, and level variation. Additionally, a single-impact source with different A-weighted maximum sound pressure levels (57, 62, 67, 72, 77 dB in LiA,Fmax) and multi-impact source (57 dB in LiA,Fmax) are randomly paired. In the psychoacoustic experiment, subjects are asked to select the more annoying sound stimulus by comparing the multi-impact source and single-impact source. From the experiment, we note that for longer duration and larger number of repetitions of the impact sound, the sound is perceived as more annoying by the subjects. In the case of the level variation of the impact sound, the annoyance is determined by the total energy level rather than the maximum sound pressure level. These results revealed that findings in the previous studies using artificial noise or tone can be applied to impact sound generated in a real situation. This study suggests that rather than maximum level, evaluation indices that can reflect the duration and energy level of the impact sound are appropriate for impact sounds which exhibit random temporal patterns.