A total of 85 introductory psychology students served as mock jurors and read transcripts of a murder trial containing a confession that was elicited in a high- or low-pressure interrogation and ruled admissible or inadmissible by the judge. No-confession control groups were also included. As prescribed by law, jurors saw the high-pressure confession as less voluntary, correctly recalled the judge's ruling, and reported the confession had less influence on their decisions. On verdicts, however, the confession increased the conviction rate--even when it was seen as coerced, even when it was stricken from the record, and even when jurors said it had no influence. The results suggest that appellate courts should exercise caution in applying the harmless error rule to the admission of coerced confessions. 50 references, 6 tables, and 1 figure