The present study tested the Supreme Court's assumption that jurors discount a coerced confession as unreliable and do not allow it to influence their decisions. In two experiments, subjects read a transcript of a trial in which testimony revealed that the defendant had confessed either on his own accord (no constraint), in response to an offer of leniency (positive constraint), in response to a threat of punishment (negative constraint), or not at all (control group). In Experiment 1, subjects discounted the negatively induced confession. However, their estimates of the probability that the defendant committed the crime were increased not only by the unconstrained confession, but by the positively constrained one as well. Experiment 2 essentially replicated this pattern for probability-of-commision estimates and verdicts despite the additional finding that the positively constrained confession was in fact perceived as involuntary. The potential danger of admitting the latter as evidence was noted and specific recommendations were made.