The liberation hypothesis argues that the effects of extra-legal factors such as victim and/or offender race on sentencing outcomes are conditioned by legally relevant factors, particularly the severity or the strength of the case. Where the evidence is weak or contradictory or the offense is less severe, decision makers are most liberated to use extra-legal factors in reaching their decisions. This study uses data on a large sample of capital murder trials in North Carolina from 1977 to 2009 to test this hypothesis. The results show that the effects of extra-legal factors (specifically, the race of offender–race of victim dyad) vary across levels of offense severity, but in a complex manner. Most notably, Black defendant–White victim dyads demonstrated an increased probability of death sentences at high levels of severity, but decreased probabilities at lower levels of severity.