Voice mismatch between conjuncts is impossible in the gapping construction. Some recent studies explained this effect by analyzing gapping as involving the ellipsis of a category at least as large as VoiceP. One prediction this analysis makes is that mismatch of any head structurally lower than Voice should not be possible in gapping. In this study, through a series of acceptability judgment experiments examining argument structure mismatches in gapping, we provide empirical observations that challenge this prediction.