Voluntary responses (VR) in human classical eyelid conditioning were analyzed in this study. In Exp. I, where conditioning under a neutral instruction was compared with that under a facilitatory instruction, it was suggested that VRs were not caused by any learning factor, and that the neutral-instruction conditioning contained a great number of VRs. In Exp. II, where puff UCS conditioning was compared with noise-UCS conditioning, it was found that the responses in the puff UCS conditioning attenuated the aversiveness of the UCS, but not the noise-UCS conditioning. Any clear-cut conclusion on VRs in conditioning is not derived from this study, but these results would place some serious limitation upon this kind of conditioning.