dissenting.
Because five of the jurors who convicted Lawrence DeLisle had read or heard about the suppressed statement of DeLisle, I would reverse the district court’s denial of habeas relief. These jurors were never told that the statement was inherently unreliable, and one testified that she believed coerced statements were likely to be true. Joint Appendix at 562a (Voir Dire of Juror Stevenson). Confessions “ ‘have profound impact on the jury, so much so that we may justifiably doubt its ability to put them out of mind even if told to do so.’ ” Arizona v. Fulminante, 499 U.S. 279, 296, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991) (quoting Bruton v. United States, 391 U.S. 123, 139-40, 88 S.Ct. 1620, 20 L.Ed.2d *397476 (1968) (White, J., dissenting)) (holding that admission of coerced confession was not harmless error). Given the circumstantial nature of the state’s case, I believe that the jurors may well have relied on their knowledge of DeLisle’s “confession” to conclude that he acted with intent to kill.
Although public scrutiny of court proceedings is an essential feature of our system of government, and serves in part to protect the accused, the Michigan judiciary was obliged to minimize the prejudicial effects of pre-trial publicity on DeLisle’s trial. See Gannett Co. v. DePasquale, 443 U.S. 368, 378, 99 S.Ct. 2898, 61 L.Ed.2d 608 (1979). Instead, on the basis of First Amendment arguments which the respondent does not attempt to defend, the Michigan judiciary contributed to those effects. The fact that the source of DeLisle’s coerced statement was the judiciary lent the statement greater credibility, making it less likely that the jurors would disregard what they had heard. This fact also heightened the judiciary’s obligation to ensure that De-Lisle’s jury was not tainted by knowledge of the statement. The court was able to find seven jurors who had read or heard general reports about the case but did not know that DeLisle had supposedly confessed. To ensure a fair trial, it should have found five more.