Jury-eligible subjects were randomly assigned to view one of four versions of a simulated capital penalty trial in which the race of the defendant (black or white) and the race of the victim (black or white) were varied orthogonally. Dependent measures included a sentencing verdict (either life without the possibility of parole or the death penalty), ratings of penalty phase evidence, and a test of instructional comprehension. Results indicated that instructional comprehension was poor overall. Although black defendants were treated only slightly more punitively than white defendants in general, discriminatory effects were concentrated among participants whose comprehension was poorest. In addition, the use of penalty phase evidence differed as a function of defendant race and whether the participant sentenced the defendant to life or death. Results suggest that racially biased and capricious death sentencing may be in part caused or exacerbated by the inability of jurors to comprehend penalty phase instructions. 56 references, 6 tables, and 1 figure