Systems contain normal noises even in a stable state, but larger noises in an uncontrollable state. In this study, we investigated different responses between experts and novices in controllability judgment. A half participant was required to discriminate variance magnitude of two sound tone sequences (standard vs. comparison). The sequence consisted of 16 tones whose pitch contained Gaussian noises. They continued to train the variance discrimination task up to a criterion level, and were regarded as experts. Then, the experts and novices participated in the controllability judgment task (i.e. a kind of the risk judgment task) with use of similar stimuli in the discrimination task. They were allowed to continue to judge for gain whether they were in an uncontrollably higher risk state with larger variances. They could stop the trial to make smaller costs. Our results showed experts in the discrimination task increased avoidance responses more than novices.