Sherman v. Smith

PUBLISHED UNITED STATES COURT OF APPEALS FOR THE FOURTH CIRCUIT TIMOTHY SCOTT SHERMAN, Petitioner-Appellant, v. WILLIAM L. SMITH, Warden, No. 94-6831 Maryland House of Correction- Annex; JOHN JOSEPH CURRAN, Attorney General for the State of Maryland, Respondents-Appellees. Appeal from the United States District Court for the District of Maryland, at Baltimore. William M. Nickerson, District Judge. (CA-91-2006) Argued: April 2, 1996 Decided: July 17, 1996 Before WILKINSON, Chief Judge, and RUSSELL, WIDENER, HALL, MURNAGHAN, ERVIN, WILKINS, NIEMEYER, HAMILTON, LUTTIG, WILLIAMS, MICHAEL, and MOTZ, Circuit Judges, sitting en banc. _________________________________________________________________ Affirmed by published opinion. Chief Judge Wilkinson wrote the majority opinion, in which Judges Russell, Widener, Hall, Wilkins, Niemeyer, Hamilton, Luttig, and Williams joined. Judge Murnaghan wrote a dissenting opinion, in which Judges Ervin and Michael joined. Judge Motz wrote a concurring and dissenting opinion. _________________________________________________________________ COUNSEL ARGUED: Andrew Lewis Frey, MAYER, BROWN & PLATT, Washington, D.C., for Appellant. Ann Norman Bosse, Assistant Attorney General, Criminal Appeals Division, OFFICE OF THE ATTORNEY GENERAL, Baltimore, Maryland, for Appellees. ON BRIEF: Roy T. Englert, Jr., James G. Duncan, MAYER, BROWN & PLATT, Washington, D.C.; Stuart J. Robinson, Bel Air, Maryland, for Appellant. J. Joseph Curran, Jr., Attorney General of Maryland, Criminal Appeals Division, OFFICE OF THE ATTORNEY GEN- ERAL, Baltimore, Maryland, for Appellees. _________________________________________________________________ OPINION WILKINSON, Chief Judge: We granted en banc review in this case to consider whether the dis- trict court properly denied Timothy Sherman's petition for a writ of habeas corpus. In 1988, a state jury convicted Sherman of killing his mother and adopted father, Ann and Stevenson Sherman, and sen- tenced him to two consecutive terms of life. Sherman asks this court to overturn his conviction due to one juror's unauthorized visit to the crime scene during the course of his trial. Because we find that the juror's excursion was not a structural error requiring a per se reversal of Sherman's conviction, and that the effect of the juror's visit was harmless, we affirm. I. Early in the morning of October 12, 1987, Ann and Stevenson Sherman were shot to death as they slept. Each died from a fatal shot- gun blast. When police arrived, Timothy Sherman, Ann and Steven- son's eighteen-year-old son, was in the house with his maternal grandfather, William Gibson. Timothy told the officers that he had heard gun shots and then had run to his grandfather's nearby home. The two of them returned to the Sherman house, where Gibson sum- moned the police. 2 Timothy was in his parents' house when the murders occurred. The house, which was equipped with an extensive alarm system, revealed no signs of forced entry or theft. The burglar alarm system was turned off. Police officers found two expended Remington 12-gauge shotgun shells in the hallway outside Ann and Stevenson's bedroom. In Timo- thy's bedroom, officers uncovered a box of Remington 12-gauge shotgun shells stuffed under the mattress of his bed. The box held five shells, and two were missing; the three remaining shells matched the expended shells found outside the Shermans' bedroom. Police deputies also found a 12-gauge shotgun, which belonged to the Shermans, lodged in a pine tree near Gibson's house. The gun was pushed inside the branches of the tree, with the barrel pointing down- ward toward the trunk of the tree and the butt pointing upward. Tests confirmed that the two shells that police found in the hallway outside the Shermans' bedroom had been fired from this shotgun. Three latent fingerprints matching those of Timothy Sherman were also found on the weapon, including a fingerprint above the trigger assembly. Timothy Sherman was charged and tried for the murder of his par- ents. At trial, the state relied heavily on the evidence gathered at the crime scene. The defense emphasized that Sherman had no apparent motive for the murders, no gunshot residue on his hands or clothing, and no pine needles or sap on his clothing even though he allegedly hid the weapon in the tree. After a lengthy trial, the jury convicted Sherman of two counts of first-degree murder. Sherman thereafter moved for a new trial based, in part, on allega- tions of juror misconduct during the course of the trial. Specifically, he alleged that one of the jurors, Blane Miller, had made an unautho- rized visit to the crime scene on the second or third day of the trial. Sherman argued that he was entitled to a new trial because the juror's visit infringed his Sixth Amendment rights. The trial judge held an evidentiary hearing on the motion for a new trial. At the hearing, juror Miller testified that one evening two or three days into the trial he and his wife drove to the crime scene, which was located in a development called Gibson Manor. As he explained, he "went to the Sherman house and then[ ] drove back the streets from the Sherman house, back to the entrance of Gibson 3 Manor in looking for a tree that was so involved in the case." Miller confirmed that he saw the tree and the house. He visited the site, he said, "so I could see the tree that was so much in question." At the close of the hearing, the trial judge rejected the motion for a new trial. Sherman appealed this ruling, and alleged nine other assignments of error. The Maryland Court of Special Appeals, how- ever, declined to grant relief. The Maryland Court of Appeals and the United States Supreme Court declined to review Sherman's convic- tion. Sherman v. Maryland, 498 U.S. 950 (1990). Sherman then filed a federal habeas petition alleging that the juror's site visit warranted reversal of his conviction. In August, 1992, the district court denied his petition. A panel of this court vacated the district court's judgment and remanded to the district court for a de novo review of the record to determine "the nature of the error" and whether it "influenced the jury's deliberations." Sherman v. Smith, 8 F.3d 820 (Table), No. 92-6947, slip op. at 7 (4th Cir. 1993) (per curiam). On remand, following a de novo review of the record, the district court again denied the petition, and Sherman filed the instant appeal. A panel of this court reversed and granted Sherman's petition. 70 F.3d 1263 (Table), No. 94-6831 (4th Cir. 1995) (per curiam). The court then voted to hear the case en banc. II. Sherman contends that juror Miller's unsupervised visit to the crime scene violated his Sixth Amendment rights to confront and cross-examine witnesses against him and to be judged by an impartial jury. We shall assume for purposes of argument that juror Miller's site visit amounted to a constitutional violation of Sherman's rights. This error, Sherman claims, constituted a structural error requiring automatic reversal of his conviction, rather than a trial error which can be "quantitatively assessed in the context of other evidence pres- ented" in order to determine whether its occurrence was harmless. Arizona v. Fulminante, 499 U.S. 279, 308 (1991). We disagree with Sherman's claim. An unsupervised juror site visit does not constitute structural error, but rather is subject to harmless error analysis. 4 A. Criminal defendants in this country are entitled to a fair, but not a perfect trial. "[G]iven the myriad safeguards provided to assure a fair trial, and taking into account the reality of the human fallibility of the participants, there can be no such thing as an error-free, perfect trial," and the Constitution does not demand one. United States v. Hasting, 461 U.S. 499, 508 (1983). This focus on fairness, rather than on per- fection, protects society from individuals who have been duly and fairly convicted of crimes, thereby promoting "public respect for the criminal process." Delaware v. Van Arsdall , 475 U.S. 673, 681 (1986). With this in mind, the Supreme Court has recognized that most errors do not automatically render a trial unfair and thus, can be harm- less. Fulminante, 499 U.S. at 306-07. Fulminante enumerated the wide variety of constitutional errors subject to harmless error analysis. They include improper admission of an involuntary confession, id. at 306-12; overbroad jury instructions at the sentencing stage of a capital case, Clemons v. Mississippi, 494 U.S. 738 (1990); improper admis- sion of evidence at the sentencing stage of a capital case, Satterwhite v. Texas, 486 U.S. 249 (1988); jury instructions containing erroneous conclusive or rebuttable presumptions, Carella v. California, 491 U.S. 263, 266-67 (1989) (per curiam); Rose v. Clark, 478 U.S. 570 (1986); erroneous exclusion of a defendant's testimony regarding the circumstances of a confession, Crane v. Kentucky , 476 U.S. 683, 691 (1986); improper restriction on a defendant's right to cross-examine witnesses for bias, Van Arsdall, 475 U.S. at 673; denial of a defen- dant's right to be present at trial, Rushen v. Spain, 464 U.S. 114, 117- 19 and n.2 (1983) (per curiam); improper comment on a defendant's silence at trial, Hasting, 461 U.S. at 499; improper prohibition on the provision of a lesser included offense instruction in a capital case, Hopper v. Evans, 456 U.S. 605 (1982); failure to instruct the jury on the presumption of innocence, Kentucky v. Whorton, 441 U.S. 786 (1979) (per curiam); improper admission of identification evidence, Moore v. Illinois, 434 U.S. 220, 232 (1977); erroneous admission of an out-of-court statement of a nontestifying codefendant, Brown v. United States, 411 U.S. 223, 231-32 (1973); improper admission of a confession made to an undercover officer, Milton v. Wainwright, 407 U.S. 371 (1972); admission of evidence obtained in violation of 5 the Fourth Amendment, Chambers v. Maroney, 399 U.S. 42, 52-3 (1970); and improper denial of counsel at a preliminary hearing, Coleman v. Alabama, 399 U.S. 1, 10-11 (1970). Indeed, "if the defen- dant had counsel and was tried by an impartial adjudicator, there is a strong presumption that any other errors that may have occurred are subject to harmless-error analysis." Rose, 478 U.S. at 579; see also United States v. Blevins, 960 F.2d 1252, 1261-62 (4th Cir. 1992). The Supreme Court has also recognized that certain structural errors are so severe as to render a trial inherently unfair and thus, should not be subject to harmless error analysis. Fulminante, 499 U.S. at 309-10; see, e.g., Sullivan v. Louisiana, 508 U.S. 275 (1993) (the denial of the right to a jury verdict of guilt beyond reasonable doubt); McKaskle v. Wiggins, 465 U.S. 168 (1984) (the denial of the right of self-representation at trial); Waller v. Georgia , 467 U.S. 39 (1984) (the denial of the right to public trial); Gideon v. Wainwright, 372 U.S. 335 (1963) (the total deprivation of the right to counsel); Tumey v. Ohio, 273 U.S. 510 (1927) (the presence of a biased judge). Unlike other errors, "[t]hese are structural defects in the constitution of the trial mechanism" and "`[w]ithout these basic protections, a criminal trial cannot reliably serve its function as a vehicle for determination of guilt or innocence, and no criminal punishment may be regarded as fundamentally fair.'" Fulminante, 499 U.S. at 309-10 (quoting Rose, 478 U.S. at 577-78). Structural errors affect the "entire conduct of the trial from beginning to end," and therefore cannot be harmless. Fulminante, 499 U.S. at 309. Correctly applied, harmless error and structural error analyses pro- duce identical results: unfair convictions are reversed while fair con- victions are affirmed. Expanding the list of structural errors, however, is not mere legal abstraction. It can also be a dangerous endeavor. There is always the risk that a sometimes-harmless error will be clas- sified as structural, thus resulting in the reversal of criminal convic- tions obtained pursuant to a fair trial. Given this risk, judges should be wary of prescribing new errors requiring automatic reversal. Indeed, before a court adds a new error to the list of structural errors (and thereby requires the reversal of every criminal conviction in which the error occurs), the court must be certain that the error's pres- ence would render every such trial unfair. See id. at 310. 6 Here, Sherman maintains that a juror site visit is so unfair that it must constitute a new addition to the short list of structural errors. But a juror site visit "does not compare with the kinds of errors that auto- matically require reversal of an otherwise valid conviction." Rose, 478 U.S. at 579. Unlike the complete denial of counsel and other structural errors, which affect the "entire conduct of the trial from beginning to end," juror site visits can be discrete moments in the course of an otherwise fair trial. Fulminante , 499 U.S. at 309. As it cannot be said with any certainty that a juror site visit renders every trial in which it occurs unfair, it would be a reckless undertaking to remove such errors from the rubric of harmless error analysis. After all, our criminal justice system represents a balance between the rights of accused persons and the need for public safety. This bal- ance is best expressed in the notion of a fair, but not a perfect, crimi- nal trial. When an error is misclassified as one requiring automatic reversal, the balance is upset, and proceedings that in reality are per- fectly fair are discarded in the name of an elusive systemic perfection. B. Sherman maintains, however, that unauthorized juror site visits constitute structural error because they "defy" harmless error analysis. That analysis requires a reviewing court to quantitatively assess the effect of the error "in the context of other evidence presented" at trial. Fulminante, 499 U.S. at 308. Sherman observes that the rules of evi- dence prevent examination of the jury's mental impressions formed during the deliberative process. See Tanner v. United States, 483 U.S. 107, 117 (1987). Without such evidence, he argues, a reviewing court could never determine the effect of a crime scene visit on the jury's verdict. We disagree. Alleged infringement of Sixth Amendment rights is no exception to the general rule that "most constitutional errors can be harmless." Fulminante, 499 U.S. at 306. The Supreme Court has "long since rejected the argument that, as a general matter, the Sixth Amendment prohibits the application of harmless-error analysis in determining whether constitutional error had a prejudicial impact on the outcome of a case." Sullivan, 508 U.S. at 282-83 (Rehnquist, C.J., concurring). 7 Sherman's argument is further undercut by the Supreme Court's application of harmless error analysis to claims of juror misconduct and bias, claims that are essentially indistinguishable from those Sher- man alleges. In Smith v. Phillips, 455 U.S. 209 (1982), the Court held that a juror's mid-trial application for employment in the District Attorney's office responsible for prosecuting the defendant did not require automatic reversal of the conviction. As the Court observed, the Constitution "does not require a new trial every time a juror has been placed in a potentially compromising situation . . . [because] it is virtually impossible to shield jurors from every contact or influence that might theoretically affect their vote." Id. at 217. In Remmer v. United States, 347 U.S. 227 (1954), an outsider apparently offered a juror a bribe in return for a favorable verdict; the trial judge ordered an investigation without disclosing the proffered bribe or investiga- tion to defense counsel. Even in these circumstances, the Court did not require automatic reversal of the conviction. Instead, it directed the trial court to conduct a hearing to determine the prejudicial impact of the developments. Id. at 230. Smith and Remmer thus contemplate the precise inquiry that Sherman characterizes as impossible: discern- ing the effect of juror misconduct or bias on the verdict without exam- ining the thought processes of the jury. Following the Supreme Court's lead, this court has repeatedly examined instances of juror misconduct and bias for harmlessness. See United States v. Seeright, 978 F.2d 842, 849-50 (4th Cir. 1992) (juror's independent investigation of evidence did not require a mis- trial when judge excused juror from further service and satisfied him- self that other jurors were not affected); Stockton v. Virginia, 852 F.2d 740, 743-46 (4th Cir. 1988) (jury's exposure to improper third-party contact examined to determine extent of prejudice), cert. denied, 489 U.S. 1071 (1989); United States v. Malloy, 758 F.2d 979, 982-83 (4th Cir.) (juror's previous service at trial of co-defendant did not require a new trial), cert. denied, 474 U.S. 1009 (1985); Miller v. Harvey, 566 F.2d 879, 881 (4th Cir. 1977) (jury's improper experiment, in which one juror bit another to observe the resulting bruises, did not violate due process and thus did not require granting a writ of habeas corpus),cert. denied, 439 U.S. 838 (1978). Likewise, other courts have applied harmless error analysis to such claims. See, e.g., Lawson v. Borg, 60 F.3d 608, 612-13 (9th Cir. 1995) (juror's comments about defendant's reputation for violence subject to harmless error analy- 8 sis); United States v. De La Vega, 913 F.2d 861, 869-71 (11th Cir. 1990) (jury foreman's actions of reading a book, showing book to other jurors, and organizing deliberations based on book subject to harmless error analysis), cert. denied, 500 U.S. 916 (1991); Marino v. Vasquez, 812 F.2d 499, 504-07 (9th Cir. 1987) (juror's out of court experiment, in which she attempted to fire a weapon while holding it in a position consistent with defense's version of shooting, subject to harmless error analysis). Abundant caselaw thus rejects any special rule of automatic rever- sal for unauthorized juror site visits. Such visits do not, as Sherman alleges, "defy" harmless error inquiry. In performing that inquiry, a court can look to the nature and extent of the juror's activity and assess how that activity fit into the context of the evidence presented at trial. See De La Vega, 913 F.2d at 870-71. The court can consider whether the juror learned information that was merely cumulative of other evidence or whether he unearthed new information not previ- ously presented to the jury. See Farese v. United States, 428 F.2d 178 (5th Cir. 1970) (jury finds substantial sum of money in shirt admitted into evidence; existence of money unknown to court or parties). In short, juror site visits may be casual or they may be intensive. They may reveal much or they may uncover little. It makes no sense, how- ever, to apply to this wide range of circumstances a single rule of automatic reversal. Rather, an assessment of the degrees of potential prejudice from a site visit is ideally suited for harmless error inquiry. The inquiry here is no more speculative than consideration of other errors that are examined for harmlessness. All harmless error analysis involves some level of indeterminacy because, "in the end no judge can know for certain what factors led to the jury's verdict." Sullivan, 508 U.S. at 284 (Rehnquist, C.J., concurring). Nonetheless, reviewing courts may capably judge the effect of the unconstitutional admission or exclusion of particular evidence, Crane, 476 U.S. at 691, including the admission of an involuntary confession, Fulminante, 499 U.S. at 306-12. And they are able to gauge the effect of a partial denial of a defendant's right to be present at trial, Rushen , 464 U.S. at 117-19 and n.2, and limitations on the right to cross-examine witnesses, Van Arsdall, 475 U.S. at 681-84. By comparison to these errors, assessing the effect of a juror's site visit does not involve an unacceptable level of conjecture. 9 Our conclusion that harmless error analysis applies here is rein- forced by the Supreme Court's treatment of errors that have a more direct influence on the jury's deliberative process. For instance, "the Court has subjected jury instructions plagued by constitutional error to harmless-error analysis." United States v. Gaudin, 115 S. Ct. 2310, 2321 (1995) (Rehnquist, C.J., concurring); see , e.g., Yates v. Evatt, 500 U.S. 391 (1991), disapproved in part on other grounds, Estelle v. McGuire, 502 U.S. 62, 72-3 n.4 (1991); Carella, 491 U.S. 263; Rose, 478 U.S. at 579-82. Juries receive instructions from the trial court, and that guidance carries the court's official sanction. Yates, 500 U.S. at 403. Yet the Supreme Court has held that even an errone- ous instruction carrying the court's imprimatur can be harmless if the faulty instruction is "unimportant in relation to everything else the jury considered on the issue in question, as revealed in the record." Id. at 403. The discrete error is viewed from the perspective of the total trial. If instructional errors that carry the trial court's seal of approval can be harmless, a juror site visit can also be harmless. Far from viewing such a visit as officially sanctioned, other jurors will know that the visit was improper. At a minimum, we cannot conclude that one juror's unauthorized site visit is a structural error that renders every trial inherently unfair. We conclude, to the contrary, that the issue here is amenable to the traditional tools of harmless error analy- sis. III. We must next address which harmless error standard applies to the error alleged in this case. In habeas proceedings, an error is harmful only if it "`had substantial and injurious effect or influence in deter- mining the jury's verdict.'" Brecht v. Abrahamson, 507 U.S. 619, 637 (1993) (quoting Kotteakos v. United States, 328 U.S. 750, 776 (1946)). We have consistently applied the Brecht standard on collat- eral review. Tuggle v. Netherland, 79 F.3d 1386, 1392-93 (4th Cir. 1996); Correll v. Thompson, 63 F.3d 1279, 1291 (4th Cir. 1995), cert. denied, 116 S. Ct. 688 (1996). Sherman argues, however, that the Brecht standard is inapplicable here. Relying on decisions of the Eighth Circuit, Sherman asserts that the stricter harmless error stan- dard of Chapman v. California, 386 U.S. 18 (1967), governs our 10 review because, unlike in Brecht, the Maryland courts allegedly failed to apply the Chapman standard on direct review. See Starr v. Lockhart, 23 F.3d 1280, 1292 (8th Cir.), cert. denied, 115 S. Ct. 499 (1994); Orndorff v. Lockhart, 998 F.2d 1426, 1430 (8th Cir. 1993), cert. denied, 114 S. Ct. 1631 (1994). Brecht cannot be so easily circumvented. That decision recognizes that a federal court's collateral review of state court convictions implicates the "State's interest in the finality of convictions that have survived direct review within the state court system." Brecht, 507 U.S. at 635; see also Barefoot v. Estelle, 463 U.S. 880, 887 (1983). Based on this weighty state interest, the Supreme Court has frequently applied different rules to direct and collateral review. Brecht, 507 U.S. at 633-35. And Brecht itself rested on this respect for the finality of state court convictions. As the Court noted,"[o]verturning final and presumptively correct convictions on collateral review because the State cannot prove that an error is harmless under Chapman undermines the States' interest in finality and infringes upon their sovereignty over criminal matters." Id. at 637. Consequently, it con- cluded that a less onerous harmless error standard was "better tailored to the nature and purpose of collateral review and more likely to pro- mote the considerations underlying" its habeas jurisprudence. Id. at 638. These principles of federalism, comity, and finality apply regard- less of the harmless error standard used by the state court. We have already recognized as much in Smith v. Dixon, 14 F.3d 956 (4th Cir.) (en banc), cert. denied, 115 S. Ct. 129 (1994). There, we applied the Brecht standard in a habeas proceeding where the state court (because it found no error) conducted no harmless error analysis. Id. at 974-81. And at least three other circuits have rejected the Eighth Circuit's rea- soning. Castro v. Oklahoma, 71 F.3d 1502, 1516 n.14 (10th Cir. 1995); Tyson v. Trigg, 50 F.3d 436, 446-47 (7th Cir. 1995), cert. denied, 116 S. Ct. 697 (1996); Horsley v. Alabama, 45 F.3d 1486, 1492 n.11 (11th Cir.), cert. denied, 116 S. Ct. 410 (1995). In the pres- ent case, Sherman received full consideration of his claim by the Maryland courts, including a post-trial evidentiary hearing in which juror Miller testified and was subject to cross-examination. See Brecht, 507 U.S. at 636. The trial itself occurred some eight years ago. And the federal district court was the fifth court asked to assess 11 the effect of juror Miller's site visit; we are the sixth. Collateral review in this case thus implicates the same interests of respect for final state convictions that animated the Brecht decision. Our conclusion is reinforced by the implications of the rule that Sherman suggests. In many habeas cases, a state court will have rejected the petitioner's claim of error, and thus will have had no opportunity to apply harmless error analysis. Tyson, 50 F.3d at 446; Smith, 14 F.3d at 979. Sherman would have us ignore Brecht's com- mands in these cases, a "limitation that would rob the decision of any general significance." Tyson, 50 F.3d at 446. Moreover, it is unwise to make our harmless error standard turn on a characterization of the state court's standard of review. Hinging a habeas court's standard on such an inquiry will inevitably lead to liti- gation over what methodology of review the state court applied. This in itself is intrusive. State courts have no obligation to use particular language in considering claims presented in their courts. See Coleman v. Thompson, 501 U.S. 722, 739 (1991). But the rule Sherman pro- poses will place an implicit obligation on state courts to do just that. Finally, inconsistent harmless error standards on collateral review will prove to be confusing and inequitable. We shall adhere consistently to the Brecht standard in collateral proceedings. IV. We turn finally to whether juror Miller's site visit had a "`substan- tial and injurious effect or influence in determining the jury's ver- dict.'" Brecht, 507 U.S. at 637 (citation omitted). We hold that it did not.1 _________________________________________________________________ 1 The Antiterrorism and Effective Death Penalty Act of 1996 does not address harmless error standards governing cases arising under 28 U.S.C. § 2254. See Pub. L. 104-132, 110 Stat. 1214 (1996). The Act, however, does require a federal habeas court to defer to state court legal determina- tions. Id., § 104. In view of the fact that we affirm the district court and deny habeas relief in all events, we have no occasion to address whatever additional hurdles Sherman might face under the Act. 12 At the post-trial hearing in state court, juror Miller testified that two or three days into the trial he drove to the crime scene with his wife. His visit seemed motivated by a simple sense of curiosity. Mil- ler acknowledged that he saw the Sherman house and the tree where police officers recovered the weapon. As he said, he made the visit "so I could see the tree that was so much in question." Based on juror Miller's testimony, it is not clear whether he ever left his car for a more detailed inspection of the area.2 We agree with the district court that Miller's site visit was harm- less. First, whatever juror Miller may have observed did not have a "substantial and injurious" influence on the jury's verdict because it was cumulative of the abundant evidence admitted at trial about the crime scene. See Brecht, 507 U.S. at 639. That evidence included numerous photographs, videotapes, and extensive testimony about the Shermans' neighborhood and the tree where police found the weapon. The introduction of this large amount of evidence persuaded the Maryland courts that Miller's visit would add little to what the jury already knew. When the state trial court rejected Sherman's motion for a new trial, it referred to its earlier ruling denying a request to have the jury "view" the crime scene: "I denied the view, not because I felt that any significant information would come about, but I felt it would be a waste of time and expense to take the jury out there because I felt the issue was fully covered. So that I think that [Miller's visit] is not of such a magnitude as to warrant a new trial or reversible appeal." Likewise, the Maryland appellate court remarked on the "nu- merous testimonial references, videotapes, and photographs relating to and depicting the crime scene." _________________________________________________________________ 2 In the state court hearing, Miller did not say -- nor was he asked -- whether he left his car, and for purposes of our review we need not assume that he merely observed the crime scene from his car. Miller's description of the visit, however, left the impression with at least two courts that he stayed in his car. The Maryland Court of Special Appeals said, without more, that Miller "drove to the neighborhood of the murder scene." And the panel of this court that remanded Sherman's petition to the district court for further consideration said that Miller "found the tree but did not leave his car [ ] to inspect it." Sherman v. Smith, No. 92-6947, slip op. at 5 (4th Cir. 1993) (per curiam). 13 The district court also concluded that Miller's site visit was cumu- lative of other evidence presented at trial. In particular, the court addressed the specific claim that Sherman presses most forcefully in his petition: that Miller's visit may have led him to conclude that Sherman, despite a slight build, had the ability, as well as the opportu- nity, to hide the weapon in the thick branches of the tree -- an issue Sherman contends was a key contested question at trial. In reviewing this claim, the court granted Sherman the benefit of several assump- tions: "that Miller examined the tree, told the other jurors that he dis- agreed with the photographs, and concluded that it was possible for Sherman to hide the gun in the tree in the condition in which it was found." Even under these assumptions, the district court found that the error was harmless because substantial other evidence indicated that Sherman hid the weapon in the tree. That evidence included pho- tographs showing that Sherman had an opportunity to hide the weapon in the tree, which was located on the route between the Sher- man and Gibson houses. Testimony introduced at trial also indicated that Sherman had previously used -- as a hiding spot -- the precise place in the tree where the gun was hidden.3 In this context, the dis- trict court appropriately concluded that the juror's site visit "was cumulative of the detailed evidence presented at trial" about the neighborhood where the murder took place and the hiding place for the weapon. _________________________________________________________________ 3 Sherman contends that Corporal Hopkins, who removed the weapon from the tree, cast serious doubt on Sherman's ability (because of his slight build) to wedge the weapon in the tree. Hopkins, who was six feet tall and 210 pounds, testified at trial that "with [his] size and stature" he "would have had difficulty putting the gun inside the tree." But Hopkins also testified that the person who put the weapon in the tree "would not [have] to be strong, but had to be forceful." Under questioning, Hopkins even left open the possibility that the weapon could have been thrown into the tree: as he said, he could not "put a figure on the force [used to hide the weapon] or if someone would have thrown it would it have landed in the same position; I can't answer that. It was deeply placed inside the tree." Finally, the state emphasized that the gun proved diffi- cult to remove because the officers did not want to disturb this key piece of evidence. In short, Hopkins' testimony is hardly the confirmation of reasonable doubt that Sherman makes it out to be. 14 Second, the state offered powerful evidence at trial from which the jury concluded that Sherman killed his mother and stepfather. The murder occurred in the middle of the night when Sherman was home; there was no indication of forced entry, and the house was equipped with an alarm system. The murder weapon was a 12-gauge shotgun that belonged to the Shermans. Police discovered Sherman's finger- prints above the weapon's trigger assembly and a box of 12-gauge shotgun shells under his mattress. The box itself contained only three shells, and police located two matching (the box held five) expended shells that experts concluded were fired from the murder weapon. Police found the shotgun lodged in the branches of a tree, where, as already noted, Sherman had previously hidden objects and which is located between his own house and that of his grandparents where he ran to report the murder. Sherman argues, in response, that the question of guilt in this case was a close one. He emphasizes that at trial the state could not point to a motive for the murders. And he notes that officers discovered no gunpowder residue, pine needles or sap on his hands or clothing. Given what he characterizes a close case, Sherman contends that Mil- ler's investigation of the crime scene, and in particular the tree, can- not be considered harmless because the characteristics of the tree were relevant to a crucial disputed point at trial: whether Sherman could have hidden the weapon there. Sherman's argument is simply overwhelmed by what we alluded to earlier, namely, the powerful array of evidence presented at trial that convinced the jury that Sherman was guilty. And, even if Miller's site visit generated conclusions damaging to Sherman about the tree, we already observed that there was substantial other evidence from which the jury could conclude that Sherman hid the weapon in the tree. In fact, considering the evidence that contributed to the jury's verdict, the juror's site visit is nothing more than a roundabout way for Sher- man to challenge the sufficiency of the evidence against him, an argu- ment that Sherman himself admits cannot succeed. In light of all the evidence presented at trial, we harbor no "grave doubt as to [the] harmlessness" of Miller's site visit. O'Neal v. McAninch, 115 S. Ct. 992, 995 (1995). We conclude that the unautho- rized excursion to the crime scene was harmless. A new trial -- some 15 eight years after the fact -- would not produce a fairer, more reliable, or more just verdict than the one already rendered. V. For the foregoing reasons, we affirm the judgment of the district court and deny Sherman's petition for a writ of habeas corpus. AFFIRMED MURNAGHAN, Circuit Judge, dissenting: The United States Constitution guarantees criminal defendants the right to "a fair trial in a fair tribunal" and a jury verdict "based upon the evidence developed at the trial." Irvin v. Dowd, 366 U.S. 717, 722 (1961) (citations omitted). "This is true, regardless of the heinousness of the crime charged, the apparent guilt of the offender or the station in life which he occupies." Id. In the matter before us, neither the trial nor the tribunal was fair because a juror made an unauthorized and unsupervised visit to the scene of the crime which, though discussed with other jurors, remained undisclosed to the parties and the court until after the guilty verdict had been returned. The majority decision upholding the conviction despite this fundamental flaw, therefore, strikes a blow to the foundations of American criminal jurisprudence and sacrifices the integrity of the means our nation has chosen for what the majority views to be the proper ends in this specific case. Even a wrong that appears slight becomes significant when it has the power to lessen public confidence in our guarantee that a jury trial will be a fair trial, and thus cannot be ignored. I. Here, Timothy Sherman's trial was not fairly conducted. At the start of it, jurors received a handbook to assist them in understanding their duties which cautioned them about discussing evidence among themselves and considering information from outside the courtroom. The handbook stated that "a juror should never, while a trial is in progress, inspect the scene of an accident or other event involved in the case."1 Handbook at 33. The defense asked that the jury be _________________________________________________________________ 1 While nothing in the handbook"is to be regarded by jurors as instruc- tions in law," the pamphlet makes clear what constitutes proper evidence and that an unsupervised or unpermitted inspection of the scene of the crime is wrong. 16 allowed to visit the crime scene, but the trial judge refused.2 During the presentation of evidence, two central issues arose: whether the defendant--6' 2" tall and weighing but 130 pounds--possessed the physical capability to lodge the shotgun deep into the branches of a thick pine tree as it was found; and whether the defendant could have hidden the gun in the tree without generating some outward signal of his efforts, such as a scratch from a branch or tree sap and pine nee- dles on his person. One juror, confused by the seemingly conflicting testimony on these issues and the unrevealing photographs of the tree, took it upon himself to see, i.e., to witness, "the tree that was so much in question."3 In disregard of the handbook and without the knowledge of the par- ties, their counsel or even the judge supervising the trial, the juror and his wife drove one evening after court ended for the day to the subdi- vision where the crime had occurred. Because of the sparse inquiry into the matter at a post-trial hearing, we do not have all of the facts concerning the juror's investigation.4 We do know that he located the Sherman house and then drove through the neighborhood "looking for a tree that was so involved in the case." We also know that he found the tree. We do not know, though, whether the juror touched the tree in an attempt to determine if there was exposed sap or loose needles, or even conducted his own experiment by trying to force something down among the branches. _________________________________________________________________ 2 The State has asserted that the jury knew about neither the request nor the denial. Those circumstances do not change the fact that a site visit was precluded both before the request and after the denial. 3 Photographs of the tree and the neighborhood had been introduced at that point in the trial, but the juror wanted a clearer view. 4 Federal and state evidentiary rules forbid inquiry into how outside influences might have affected the jury's deliberations and verdict, but permit questioning about whether extraneous prejudicial information was improperly brought to the attention of any juror. Fed. R. Evid. 606(b); Md. R. Evid. 5-606(b). Thus the parties could have asked for the details of the juror's visit. The trial judge prohibited the question of why the juror went to the site, but the juror explained his actions anyway, testify- ing at the post-trial hearing that "the reason why I went there was so I could see the tree that was so much in question." 17 Nor do we know whether he tried to gauge the distance between the two houses and the tree. The wayward juror told others on the panel of his outing.5 The trial judge and the parties learned of the impropriety only after the jury had rendered its verdict, when it was too late to take corrective action. Before deliberations began, therefore, the judge had no reason to remind the jurors to disregard outside viewing of a scene nor to direct them specifically to ignore any information gathered during the juror's visit to Sherman's neighborhood. See Jeffries v. Blodgett, 5 F.3d 1180, 1191 (9th Cir. 1993) (finding unfair prejudice where infor- mation that the defendant was a convicted criminal was conveyed to jurors outside the courtroom without the attorneys and the judge knowing or having an opportunity to correct), cert. denied, 114 S. Ct. 1294 (1994). We do not know what role the juror's independent investigation played in his own mind or in deliberations because we cannot inquire into such matters. See Mattox v. United States, 146 U.S. 140, 149 (1892) (holding inadmissible juror testimony regarding the motives and influences which affected the jury's deliberations). Nor do we have juror testimony regarding the presence or absence of prejudice resulting from the visit.6 See United States v. Seeright, 978 _________________________________________________________________ 5 Sherman has submitted the affidavit of an alternate juror, indicating that the juror told the others about his investigation and shared what he learned. Because that affidavit was not presented before a state court, however, there is some question as to whether it may properly be consid- ered on federal habeas review. See Boggs v. Bair, 892 F.2d 1193, 1199 n.2 (4th Cir. 1989) (prohibiting the introduction of an affidavit for the first time on federal habeas review after determining that it could have been presented at state court proceedings), cert. denied, 495 U.S. 940 (1990). We previously ruled in the present matter that the state court's determination that the unauthorized site visit did not warrant a new trial was not a finding of fact entitled to deference under 28 U.S.C. § 2254(d), but subject to de novo review in a federal habeas proceeding. Sherman v. Smith, 8 F.3d 820 (table), 1993 WL 433317, at *4 (4th Cir. Oct. 27, 1993) (unpublished) (No. 92-6947); see also Lawson v. Borg, 60 F.3d 608, 612 (9th Cir. 1995) (stating that the harmlessness of a constitutional error is not a factual determination entitled to§ 2254(d)'s presumption of correctness). It therefore would seem proper for the federal habeas court to consider the affidavit, as the district court here did, especially considering the time when the subject arose. 6 There was no questioning of individual jurors regarding the possibil- ity of taint or prejudice here, nor was there warning by the judge for the jurors to avoid it. 18 F.2d 842, 850 (4th Cir. 1992) (finding that a juror's personal investi- gation and sharing of result with rest of panel did not warrant a mis- trial because the trial judge dismissed the offending juror and questioned each of the other jurors individually and specifically until he was satisfied that they were neither tainted nor prejudiced).7 _________________________________________________________________ 7 Like the instant case, Crockett v. Commonwealth, 47 S.E.2d 377 (Va. 1948), concerned a prosecution for murder and an unauthorized, though innocent, jury view of the crime scene. There, though, the question of the burden of proof to show prejudice or lack thereof was at issue. Id. at 386. The Virginia Supreme Court of Appeals reversed the conviction and ordered a new trial. Id. at 387. Not deciding whether the Virginia rule requiring the defendant's presence on a jury view came into play in the matter before it, the court said a new trial was necessary to protect "the fundamental rights of an accused." Id. at 385. The court concluded: We cannot say that impressions the jury could have so obtained may not have been harmful to the accused. Though no wrongful action was intended . . . the happening casts a shadow of such magnitude over that part of the proceedings as to constitute such misconduct on their part as to impose upon the Commonwealth the burden of disproving any harmful result to the accused. That such proof, under the circumstances here presented, is difficult if not well-nigh impossible, does not render less imperative the necessity of its production. Id. at 386. Roberts v. United States, 60 F.2d 871 (4th Cir. 1932), also concerned an unauthorized visit by jurors to the scene of the crime involved in the case. There, however, the court found that the mere fact that an unautho- rized view took place caused no prejudice to the defendant, but said nothing about on whom the burden of proving prejudice fell. Id. at 873. Relying on Roberts, the court in People v. Kraus, 265 N.Y.S. 294 (Ct. Gen. Sess. 1933), found no prejudice shown to have resulted from an independent site visit by a juror who was unaware that his actions were improper. Id. at 297-300. The trial court admitted that it had failed to order the jurors not to visit the scene of the crime. Id. at 296. Another case invoking Roberts, Orenberg v. Thecker, 143 F.2d 375 (D.C. Cir. 1944), found that an unauthorized visit to the scene brought no extrane- ous influences to bear upon the jurors and therefore affirmed the trial court's denial of a new trial. Id. at 376. United States v. Kansas City, MO, 157 F.2d 459 (8th Cir. 1946), also relied on Roberts in holding that the verdict was not affected by an unauthorized viewing by some of the jurors of the property at issue in the case. Id. at 462-63. The court there 19 II. The juror's unauthorized and unsupervised visit clearly amounted to constitutional error as a violation of Sherman's Sixth and Four- teenth Amendment rights. The majority treats the error as a trial defect and holds it to be harmless. If harmlessness were the proper inquiry, I do not agree with the majority's finding. In my opinion, the personal viewing of the tree was harmful because it allowed the injec- tion of evidence without the defendant having an opportunity to object, confront or cross-examine.8 Actual, secret, physical inspection _________________________________________________________________ also made much of the fact that, if a request for permission to view had been requested, it "would doubtless have been granted." Id. at 462. Neither Roberts nor its progeny expressed any belief or concern that unauthorized viewing was unconstitutional. While they therefore did not address the question of whether the error was "structural" or "trial," that omission is understandable because they antedated the United States Supreme Court's decision that constitutional errors could be harmless in Chapman v. California, 386 U.S. 18 (1967). Most significantly, however, these cases did not even allude to the vital rights of confrontation and cross-examination of which the parties were deprived. Nor did they con- cern situations where a visit to the scene was not only unauthorized but specifically forbidden. In considering the matter before us, we cannot overlook the fact that evidentiary rules precluded the parties from questioning the jurors-- including the maker of the forbidden site visit--as to what effect the unauthorized viewing had on them and that the juror's actions directly violated both the handbook and the trial judge's denial of a jury visit to the scene. Despite the results in Roberts, the error in Sherman's case very clearly involves both the flawed structural character of the jury trial he was given and exhibits a behavior infraction that amounted to a funda- mental invasion of the defendant's rights. 8 Unauthorized juror viewing of the scene at issue in an ongoing trial is not generally acceptable conduct. Ruling in an 1888 case deemed "without precedent" and "of the greatest importance," Scottish judges found that a juror's unpermitted and unsupervised visit to the site of an injury led to "a gross miscarriage of justice." Sutherland v. Preston- grange Coal & Firebrick Co., 15 R. (Ct. of Sess.) 494, 495 (Scot. 1888). In setting aside the jury verdict, the judges viewed it as largely irrelevant 20 of the tree can in no way be construed as "cumulative" of the evidence presented at trial, which consisted of photographs, videotape and tes- timony. The juror presumably would not have made his outing if that evidence were sufficient for him to resolve the issues in dispute. Even the district court, accepting that there was error in the juror's actions but finding it harmless, acknowledged that the prosecution's most _________________________________________________________________ whether the juror had influenced his fellow panel members. The Lord President explained: The jury are empaneled and sworn to return a verdict according to the evidence led before them. They are not entitled to proceed upon anything else but the evidence given upon oath, and if they do proceed upon anything else they cannot return a verdict according to their oath. . . . I think therefore it is essential to the justice of the case that we should grant a new trial. Id. at 495-96. A second judge expressed his opinion as follows: It is impossible to feel satisfied that the verdict in this case was returned according to the evidence laid before the jury. It is impossible to believe that this gentleman . . . was not impressed by what he saw. He therefore gave his verdict upon evidence not obtained at the trial. Whether, or how far, he influenced others by stating the impression made upon himself it is impossible to state, but it is very probable that he did to some extent influence his co-jurors. Id. at 496 (opinion of Lord Adam). In Rex v. Ryan, 4 W.W.R. (N.S.) 32, 101 C.C.C. 101 (Brit. Col. 1951), the court required a new trial after conviction because the jury in a mur- der case made an out-of-court excursion without the trial judge's permis- sion. The Court of Appeal explained: The jury is a legal institution in which the people take great pride. Through it the people take their responsible part in the administration of legal justice. Removal of the jury from all out- side influences lies at the very foundation of the confidence that has been maintained in it. It is of the highest importance there- fore not only that no communication with outsiders shall actually in fact occur, but also that nothing shall seem to take place which may weaken respect for the jury in the public mind. 101 C.C.C. at 105 (internal citation omitted). 21 crucial photograph, a close-up of the tree, "does not show me any- thing on its face." It is therefore impossible to believe that the juror's investigation did not impress him and possibly others and impossible to say--as one must in finding harmlessness--that the error did not have "substantial and injurious effect or influence in determining the jury's verdict."9 See Brecht v. Abrahamson, 507 U.S. 619, 637 (1993). _________________________________________________________________ 9 The majority appears to have forgotten the Supreme Court's exhorta- tions not to focus on the guilt of the defendant but on the effect that the error had on the jury's verdict. "Harmless-error review looks . . . to the basis on which the jury actually rested its verdict . . . not whether, in a trial that occurred without the error, a guilty verdict would surely have been rendered." Sullivan v. Louisiana, 113 S. Ct. 2078, 2081 (1993) (citations omitted); see also O'Neal v. McAninch , 115 S. Ct. 992, 994 (1995) (advising appellate judges, when "in grave doubt" about whether an error was harmless, to treat the error not as harmless, but as though it affected the verdict). See generally Harry T. Edwards, To Err is Human, But Not Always Harmless: When Should Legal Error be Tolerated?, 70 N.Y.U. L. Rev. 1167 (1995). Chief Judge Edwards asserts: Indeed, Chief Justice Rehnquist says that, in any harmless-error review, the role of the appellate court is to "determine whether it is possible to say beyond a reasonable doubt that the error did not contribute to the jury's verdict." Only last term, the Supreme Court offered up its decision in O'Neal v. McAninch, [115 S. Ct. 992 (1995)] the crown jewel in the decisions moving away from guilt-based applications of the harmless-error doctrine. In O'Neal, the Court considered what action a federal habeas court must take when, upon review of a state-court judgment from a criminal trial, it finds itself left in "grave doubt" as to whether a constitutional error was harmless. The court, in an opinion by Justice Breyer, held that the appellate judge in such a case should treat the error not as harmless, but rather as though it affected the verdict. This conclusion, the Court stated, is consistent with the application of the Kotteakos [v. United States, 328 U.S. 750 (1946)] standard, which applies even to constitutional errors in habeas proceedings, and which admonishes that "`if [a reviewing court] is left in grave doubt [as to the harmlessness of an error], the conviction cannot stand.'" Id. at 1201-02 (emphasis in original; footnotes omitted). How can one properly regard as non-structural, or harmless, an error where the defen- 22 In any event, I believe that harmlessness review is not appropriate here because the error is fundamental and structural. The basic frame- work of our trial system requires that evidence be presented and _________________________________________________________________ dant is kept altogether ignorant of a witness he should have been allowed to cross-examine about a matter which played such a central part in the case? Chief Judge Edwards describes Justice Breyer's "common-sense view of harmless error focused not on artificial categories of cases, but on notions of fundamental fairness." Id. at 1202. He notes that O'Neal establishes that "the proper measure of harmlessness is whether the error `had substantial and injurious effect or influence in determining the jury's verdict,' not whether the record evidence is sufficient absent the error to warrant a verdict of guilt." Id. (Emphasis in original; citation omitted). In note 160, Judge Edwards continues: In Kyles v. Whitley, 115 S. Ct. 1555, 1566 (1995), the Court ruled that, in determining whether the defendant has been preju- diced by a violation of Brady, "[t]he question is not whether the defendant would more likely than not have received a different verdict with the evidence, but whether in its absence he received a fair trial, understood as a trial resulting in a verdict worthy of confidence." . . . According to the majority[of the Supreme Court] "[a] defendant need not demonstrate that after discounting the inculpatory evidence in light of the undisclosed evidence, there would not have been enough left to convict." . . . Thus, this case seems to return to Kotteakos's original focus on the severity of the error rather than the cumulative weight of the untainted evidence. Id. at 1203. The article describes the Supreme Court as making "it clear that `[a]n error may seriously affect the fairness, integrity or public repu- tation of judicial proceedings independent of the defendant's innocence." Id. at 1204 (emphasis in original). Judge Edwards concludes by saying that "[t]he mission of the appel- late courts in evaluating claims of harmless error should be to address significant errors and ensure fundamental fairness." Id. at 1209. The statement is reminiscent of the Supreme Court's admonition that a prose- cutor's interest in a criminal case "is not that it shall win a case but that justice shall be done." Berger v. United States, 295 U.S. 78, 88 (1935). The interest of the trial judge is at least as great. 23 tested in a public courtroom before the jury, the judge and the defen- dant. Turner v. Louisiana, 379 U.S. 466, 472-73 (1965) (stating that "[t]he requirement that a jury's verdict`must be based upon the evi- dence developed at the trial' goes to the fundamental integrity of all that is embraced in the constitutional concept of trial by jury" and that the Constitution requires "at the very least that the `evidence devel- oped' against a defendant shall come from the witness stand in a pub- lic courtroom where there is full judicial protection of the defendant's right of confrontation, of cross-examination, and of counsel"); see also Pointer v. Texas, 380 U.S. 400, 405 (1965) ("There are few sub- jects, perhaps, upon which this Court and other courts have been more nearly unanimous than in their expressions of belief that the right of confrontation and cross-examination is an essential and fundamental requirement for the kind of fair trial which is this country's constitu- tional goal. Indeed, we have expressly declared that to deprive an accused of the right to cross-examine the witnesses against him is a denial of the Fourteenth Amendment's guarantee of due process of law.") Independent and unauthorized jury viewing therefore is prohib- ited. The Ninth Circuit Court of Appeals reaffirmed these fundamental principles in a recent opinion. United States v. Noushfar, 78 F.3d 1442 (9th Cir. 1996). Reversing and remanding for a new trial, the court held that allowing the jury to listen to matters that had never been presented in open court requires reversal because the jury's con- sideration of evidence is "a stage of the trial at which the presence of the defendant is required." Id. at 1444. The court went on to say: The court completely abdicated control of the presentation of the evidence. . . . In cases where the error is so fundamen- tal and defies meaningful review, we have said that harmless or plain error analysis may not be applied. Instead, we find the error to be a structural error requiring automatic reversal . . . . [In an earlier case,] we said that structural error analy- sis was the correct approach where there was a "complete abdication of judicial control over the process." Id. at 1445.10 _________________________________________________________________ 10 The opinion of the majority in the instant case therefore seems to create a circuit split inviting the Supreme Court to grant certiorari. Per- haps then the majority opinion would be shown to have but a short life. 24 A mistake that violates basic trial structure and taints the entire process is most properly labelled "structural error." See Brecht, 507 U.S. at 629-30 (discussing difference between structural and trial error);11 Arizona v. Fulminante, 499 U.S. 279, 306-10 (1991) (same). The impact of a juror's personal gathering of evidence outside the courtroom by way of an unsupervised--indeed forbidden--viewing is difficult to isolate and assess. See Fulminante , 499 U.S. at 307-09 (explaining that trial error may be "quantitatively assessed . . . in order to determine whether its admission was harmless" while struc- tural error defies such analysis);12 see also Brecht, 507 U.S. at 629 _________________________________________________________________ 11 It should not be overlooked that the Supreme Court in Brecht relied very much on the fact that state courts had earlier found the error harm- less beyond a reasonable doubt in concluding that the federal court on habeas corpus review need not duplicate the analysis set forth in Chapman, 386 U.S. at 24. Brecht, 507 U.S. at 635-38. Here, there was never a state court finding pursuant to Chapman , so requiring that inquiry may not as easily be viewed duplicative and unnecessary. 12 Again, there was no opportunity in Sherman's case to determine whether the error could be cured and thus deemed harmless. Neither the trial judge nor Sherman knew of the site visit and resulting breach of Sherman's right of cross-examination until after the jury had rendered its verdict. Under different circumstances, the matter might well require a different outcome. The fact remains, however, that the error in Sher- man's case remained unaddressed and uncured, and thus violated the entire structure of the trial. Although it concerned somewhat different circumstances, State v. Magwood, 432 A.2d 446 (Md. 1981), is instructive. There, the trial judge permitted a jury hearing a criminal case to separate overnight despite the fact that it was in the midst of deliberations. The defendant's counsel agreed to the separation and the defendant did not object, and the judge gave the jury the ordinary admonition "not to discuss the case during their overnight separation or to consult outside references." Id. at 447. In analyzing the claim of error, the Maryland Court of Appeals emphasized the judge's instruction not to discuss the case and"not to consult outside sources." Id. at 451. Although it found no demonstration of "any viola- tion of this court instruction," the court of appeals declared that if a breach of the prohibition had occurred, the remedy would have been "to attack the jury verdict as improperly reached." Id. The court further found that counsel's acceptance of the jury separation waived the defen- dant's right to a sequestered jury, but noted that only the defendant could 25 (same). Its pervasive nature cannot be discounted in an attempt to consider whether it was harmless. Even if the information gleaned could be called harmless because it did not substantially influence the jury's verdict, the independent investigation undermined the integrity of the trial and thus the jury's decision.13 The harmful and truly the structural nature of the juror's unauthorized and undisclosed visit to the tree becomes clear with the realization that Sherman was, as a consequence, deprived of his right to cross-examine the undisclosed witness.14 Contained in the Sixth _________________________________________________________________ have waived his right to confront witnesses against him. Id. at 450. Here, of course, neither Sherman nor his counsel could waive his rights because they did not know of the unpermitted jury excursion. Further- more, the forbidden juror activity actually occurred. It is also helpful to consider State v. Collins , 288 A.2d 163 (Md. 1972). There, notice of a deposition was sent to, but not actually received by, the defendant. His counsel nevertheless appeared and cross-examined the witness. Id. at 168. The Maryland Court of Appeals reversed the convic- tion and remanded for a new trial, noting that it had made "assiduous efforts to protect the constitutional guarantee of confrontation at all stages of a trial" and that "these efforts would be meaningless if this invi- olable right could be circumvented by allowing the deposition of a wit- ness, taken while the accused is involuntarily absent, to be admitted in evidence at trial." Id. at 170. 13 In Sutherland, the Scottish court required a new trial although it could not be sure whether the improperly behaving juror influenced his fellow jurors or himself. 15 R. (Ct. of Sess.) at 495-96. The court essen- tially found that the jury's verdict was structurally deficient, ruling that it had not been returned "according to their oath." Id. at 495. 14 The outcome in Delaware v. Van Arsdall, 475 U.S. 673 (1986), illus- trates an essential distinction. Upon review of a reversal of a conviction because of a restriction on the defendant's ability to cross-examine a wit- ness which violated his Confrontation Clause rights, the Court remanded to allow harmlessness to be explored. Id. at 684. Unlike Sherman's situa- tion, the defense in Van Arsdall was fully aware of the bias issues it sought to explore with cross-examination, but was denied the opportu- nity. Hence, remand to examine whether prejudice or harm had occurred was appropriate, for the error was a trial error. On remand, the Supreme Court of Delaware held that the error violated the Delaware Constitution and was not harmless. Van Arsdall v. State, 524 A.2d 3, 6 (Del. 1987). 26 Amendment's guarantee that a defendant will be informed of the nature of the accusation and confronted with the witnesses against him, cross-examination is both a fundamental right and perhaps the most critical tool to be employed in the representation of a criminal defendant. Pointer, 380 U.S. at 404-05. Allowing a juror to conduct his own investigation outside the courtroom in an attempt to answer questions that arose in the criminal matter before him certainly "af- fects the framework within which the trial proceeds" and justifies automatic reversal of the conviction. Fulminante , 499 U.S. at 309-10. I therefore do not believe that we can say that Sherman's trial "reli- ably serve[d] its function as a vehicle for determination of guilt or innocence, and that [the] criminal punishment may be regarded as fundamentally fair." Id. at 310 (quoting Rose v. Clark, 478 U.S. 570, 577-78 (1986)). It is not fundamentally fair to the parties for a juror, in effect, to conduct privately his own trial. A juror's reliance upon extrinsic information, even if that informa- tion is accurate and useful, "would tend to obstruct the administration of justice, because even a correct conclusion is not to be reached or helped in that way, if our system of trials is to be maintained."15 _________________________________________________________________ Sherman, though he would be aware of how vital cross-examination of the errant juror would be if he testified, was totally ignorant that the juror had in essence become a witness subject to cross-examination. Thus, harmlessness could not possibly be found. Under the circum- stances of Van Arsdall, knowledge of the witness's presence and the pos- sibility of exploration by cross-examination perhaps allowed the error to be treated as a "trial" error. In Sherman's case, however, the error was clearly "structural" from the moment it occurred. Sherman, altogether unaware that he was being denied the right of confrontation, cannot be faulted for not attempting to cross-examine. The artificial category of cases labelled "Confrontation Clause" cases and deemed amenable to harmless-error review should not be permitted to eradicate "notions of fundamental fairness." 15 Again, the principle is fundamental to Anglo justice systems. In ordering a new trial after jurors had improperly asked questions of a boatman taking them on an authorized view, an Australian court stated: It is quite clear that a jury, sworn to find their verdict according to the evidence, cannot have any evidence before them except 27 Patterson v. Colorado, 205 U.S. 454, 462 (1907)."The theory of our system is that the conclusions to be reached in a case will be induced only by evidence and argument in open court, and not by any outside influence." Id. The majority here ignores that fundamental precept and suggests that, in some cases, jurors may ignore a judge's orders and seek whatever information they like about a case from whatever source they wish.16 It overlooks the structural character of the error, as well as the destructive effect on the public's trust, to allow a jury trial to be conducted with such blatant violations of governing rules. III. Some errors are so egregious that they must be corrected, even if the result is to overturn a guilty verdict. "Surely no fair-minded per- son will contend that those who have been deprived of their liberty without due process of law ought nevertheless to languish in prison. . . . For such anomalies, such affronts to the conscience of a civilized society, habeas corpus is predestined by its historical role in the strug- gle for personal liberty to be the ultimate remedy." Fay v. Noia, 372 U.S. 391, 441 (1963). The result of granting habeas relief in this case would not necessar- ily be to let a convicted killer go free. The State of Maryland would have an opportunity to retry Sherman properly in an attempt to secure _________________________________________________________________ such as is adduced in open Court, and if such evidence, whether it be oral or in writing, come to them the verdict may be avoided . . . . Nor is the case altered by the fact that the evidence so given out of Court may be true. Smith v. Neild, 10 N.S.W.L.R. 171, 173 (Aus. 1889) (citations omitted). 16 There is no indication that the misbehaving juror here was sanc- tioned, or otherwise rebuked or punished, once the error he committed became known. The lack of such a response suggests that any juror may so violate a court's directions with impunity. The majority underscores the message by allowing the error to stand uncorrected. Its decision con- trols not only in Harford County, Maryland, but in every city and county in the five states comprising the Fourth Circuit. Hereafter jurors will feel free to disregard instructions not to perform unsupervised viewing of the scene of the crime. 28 a conviction from an untainted jury. Here, "[p]ublic confidence in the fair and honorable administration of justice, upon which ultimately depends the rule of law, is the transcending value at stake." Sherman v. United States, 356 U.S. 369, 380 (1958) (Frankfurter, J., concur- ring). Sherman may well have killed his parents, but we are a less civ- ilized nation if we do not require the State to prove in a fair trial that he did so. It is not enough merely that a defendant committed a crime. The fundamental tenets of our justice system require that he be con- victed properly. IV. For the above reasons and those contained in the earlier majority opinion of the panel, Sherman v. Smith, 70 F.3d 1263 (4th Cir. 1995) (unpublished) (per curiam), vacated and reh'g en banc granted (4th Cir. Jan. 18, 1996), I respectfully dissent. I would favor the grant of the writ of habeas corpus. Judge Ervin and Judge Michael join in this dissent. DIANA GRIBBON MOTZ, Circuit Judge, concurring in part and dis- senting in part: As the Supreme Court has specifically recognized, the right at stake here, the right to have "a jury's verdict .. . based upon the evi- dence developed at trial[,] goes to the fundamental integrity of all that is embraced in the constitutional concept of trial by jury." Turner v. Louisiana, 379 U.S. 466, 472 (1965). Unfortunately, the atrocious crimes committed and abundant circumstantial evidence pointing to Sherman as the perpetrator have led the majority to disregard this bedrock principle. Regardless of how much sympathy we have for the victims or how certain we may be of Sherman's guilt, we must follow the Constitution. Adherence to it does not require that Sherman be released but it does require that he be provided a new trial. Accord- ingly, although I agree with some of the majority's conclusions, I must respectfully dissent from its holding. I. Most constitutional errors are trial errors and can be harmless, but some "will always invalidate the conviction." Sullivan v. Louisiana, 29 508 U.S. 275, 279 (1993). Such "structural defects in the constitution of the trial mechanism . . . defy analysis by `harmless-error' stan- dards." Arizona v. Fulminante, 499 U.S. 279, 309 (1990). Structural errors are those that affect "the framework within which the trial pro- ceeds." Id. The right at issue in this case -- the right to be convicted solely on the basis of evidence presented at trial-- seems to me to fit squarely within this definition. Thus, an error involving this right, like errors involving the related right to a public trial or the right to be represented by counsel at trial, should be deemed structural. If we were writing on the proverbial clean slate, I would so hold. After all, of what consequence is the right to counsel, to an impar- tial judge, to a public trial, to a correct reasonable doubt jury instruction,1 or to a criminal trial at all, if a defendant can be convicted based on evidence not presented at trial. The right to have a conviction based only on the evidence presented at trial, like other rights whose depri- vation constitutes structural error, is one of those protections without which "a criminal trial cannot reliably serve its function as a vehicle for determination of guilt or innocence, and no criminal punishment may be regarded as fundamentally fair." Rose v. Clark, 478 U.S. 570, 577-78 (1986) (citation omitted). Permitting a verdict to be based on evidence other than that pres- ented at trial, like other structural errors, has repercussions that are "necessarily unquantifiable and indeterminate, unquestionably qualif- [ying it] as `structural error.'" Sullivan, 508 U.S. at 282. This is par- ticularly true in light of the rules of evidence and the restrictions they quite legitimately place on any inquiry into jury deliberations. See Tanner v. United States, 483 U.S. 107, 117-125 (1987). Because they prohibit any inquiry into the effect that additional, unopposable, and possibly inadmissible evidence might have on a jury, a juror's expo- sure to such evidence is necessarily unquantifiable and indeterminate. See United States v. Bagley, 473 U.S. 667, 693 (1985) ("The private whys and wherefores of jury deliberations pose an impenetrable bar- _________________________________________________________________ 1 Deprivations of each of these rights has been held to be structural error. See Sullivan v. Louisiana, 508 U.S. 275 (1993) (reasonable doubt instruction); Waller v. Georgia, 467 U.S. 39, 49 n.9 (1984) (public trial); Gideon v. Wainwright, 372 U.S. 335 (1963) (counsel at trial); Tumey v. Ohio, 273 U.S. 510 (1927) (impartial judge). 30 rier to our ability to know which piece of information might make, or might have made, a difference") (Marshall, J., dissenting). "In the constitutional sense, trial by jury in a criminal case neces- sarily implies at the very least that the evidence developed against a defendant shall come from the witness stand in a public courtroom where there is full judicial protection of a defendant's right of con- frontation, of cross-examination, and of counsel." Turner, 379 U.S. at 472-73.2 Consequently, just as when a jury is given an improper rea- sonable doubt instruction, when a verdict is not based solely on evi- dence presented at trial, "there has been no jury verdict within the meaning of the Sixth Amendment" and so "no object . . . upon which harmless error scrutiny can operate." Sullivan, 508 U.S. at 280. The most an appellate court can conclude in the case at hand is that a jury would surely have found the defendant guilty based on the evidence presented at trial. A court cannot conclude anything about the jury's actual finding of guilt, because that finding was based to some unknown extent on evidence that was never presented at trial. Cf. id. at 279-80. The majority rightly recognizes that an error should not be deemed structural unless its presence renders "unfair" every criminal convic- tion in which the error occurs. See Maj. Op. at 6. Structural error, however, is not merely a "shorthand" form of harmless error analysis, rather it involves the broader question of the fundamental fairness of the trial process. Accordingly, an error that so infects a trial as to ren- der the process fundamentally unfair is structural, regardless of whether, looking narrowly at its "actual impact" on the verdict, the error might sometimes be deemed harmless. For example, the Supreme Court has held that the denial of the right to a public trial is structural error, Waller, 467 U.S. at 49 n.9, because it renders the entire process unfair, not because such an error can never be "harmless" in terms of its impact on the jury's verdict _________________________________________________________________ 2 Significantly, as noted above, the deprivation of a defendant's right to a public trial or to counsel is considered structural error. The error in this case deprived Sherman of the remaining three components the Court recognized in Turner as essential to the fundamental right to trial by jury. Surely this error should also be considered structural. 31 -- the harm is to the trial process, not necessarily to the specific ver- dict reached. Similarly, because a juror's unsupervised and unautho- rized site visit, in a case in which the physical characteristics of that site are critical to the case, deprives the defendant of the right to be convicted solely on the basis of evidence presented at trial, it too ren- ders the trial process fundamentally unfair. Such an error would seem to me therefore to be structural error, whether or not one can imagine factual scenarios in which the information obtained by the juror has no impact on the jury's deliberations or verdict. Nor would such a holding require reversal in every case in which a juror engaged in an unauthorized site visit. The majority convinc- ingly explains the difficulties of a holding that would lead to this result. However, the majority has created and then defeated a straw man. Sherman specifically disavows any request for such a broad rul- ing. Instead, he maintains that unauthorized site visits amount to con- stitutional error only in cases in which"significant issues were raised at trial concerning the physical aspects of the areas visited by the juror." Brief of Appellant at 21 (quoting Commonwealth v. Price, 344 A.2d 493, 494 (Pa. 1975)). An analysis that results in finding constitutional error in some cir- cumstances but not in others is unusual, but not unprecedented. Indeed, as Sherman points out, the Supreme Court has adopted a simi- lar approach in examining court orders forbidding criminal defendants from consulting with counsel. An order preventing a defendant from consulting with counsel during an overnight recess interferes with his Sixth Amendment right to counsel, and reversal is required; the error is not examined for harmlessness and no proof of prejudice is neces- sary. See Geders v. United States, 425 U.S. 80 (1976). In contrast, an order preventing a defendant from consulting with counsel briefly during his testimony is regarded as so de minimis as not to be consti- tutional error at all. Perry v. Leeke, 488 U.S. 272, 280-84 (1989). Between these two extremes, there is "a line of constitutional dimen- sion." Id. at 278-80. In the same way, unauthorized site visits in cases in which the physical aspects of the site are not at issue should be regarded as de minimis -- not constitutional error at all. This approach has much to 32 recommend it. It protects the fundamental right involved here, yet avoids trivialization of the structural error inquiry. Because the Supreme Court has never dealt with a case involving juror's unauthorized site visit -- let alone such a visit in a case, like this, where the physical features of the site are of critical importance, no direct precedent prohibits the above approach. However, in those cases involving what I regard as the most similar constitutional errors -- unauthorized private contacts with jurors during trial -- the Supreme Court seems to have applied a harmless error analysis, hold- ing that such contacts require reversal of a conviction only if the gov- ernment fails to "establish, after notice to and hearing of the defendant that such contact was harmless to the defendant." Remmer v. United States, 347 U.S. 227, 229 (1954) (emphasis added).3 See also Mattox v. United States, 146 U.S. 140, 150 (1897) ("Private com- munications, possibly prejudicial, between jurors and third persons . . . are absolutely forbidden, and invalidate the verdict, at least unless their harmlessness is made to appear.") (emphasis added). Both Remmer and Mattox were decided well before 1991, when the Supreme Court first began analyzing constitutional error in terms of structural error, which is not subject to harmless error analysis, and trial error, which is. See Arizona v. Fulminante , 499 U.S. 279 (1991). Indeed, Remmer and Mattox predate many of the Court's landmark decisions regarding the constitutional protections due to criminal _________________________________________________________________ 3 The majority characterizes the instant case as "virtually indistinguish- able" from both Remmer and Smith v. Phillips, 455 U.S. 209 (1982). I agree that the error involved in this case is very similar to the error in Remmer. However, as this court recognized in Stockton v. Virginia, 852 F.2d 740, 744 (4th Cir. 1988), cert. denied, 489 U.S. 1071 (1989), Phillips involves clearly distinguishable concerns. As we explained in Stockton, when, as in Phillips, "some external manifestation of a juror's predisposition subsequently calls the juror's impartiality into question, the defendant is afforded the opportunity to establish the juror's actual bias." 852 F.2d at 744 (emphasis added). On the other hand, where, as in the case at hand and Remmer, "the danger is not one of juror impair- ment or predisposition, but rather the effect of an extraneous communi- cation upon the deliberative process of the jury, the defendant's right to an impartial jury requires that the government bear the burden of estab- lishing the nonprejudicial character of the contact." Id. (emphasis added). 33 defendants. E.g., Edwards v. Arizona, 451 U.S. 477 (1981); Miranda v. Arizona, 384 U.S. 436 (1966); Escobedo v. Illinois, 378 U.S. 478 (1964). Consequently, if Remmer or Mattox were decided today, the Court might well regard the errors in them as structural. Alternatively, the Court might conclude that even if harmless error analysis applies to unauthorized contacts with jurors, a juror's uns- upervised and unauthorized visit to examine for himself the disputed physical aspects of a site presents a more fundamental problem merit- ing treatment as structural error. A juror's private fact-finding mis- sion, which results in new facts being presented to the jury without the benefit of cross-examination by (or even the knowledge of) the defendant, certainly presents more serious Confrontation Clause prob- lems than juror contacts having nothing to do with the facts of the case. Thus, when the Supreme Court is presented with an appropriate case, it may well conclude, as my dissenting colleagues do, that the error involved here is indeed structural. However, in view of Remmer and Mattox , the Supreme Court's reluctance to classify errors as structural, and the decisions of this court extending Remmer to cases involving a jury's exposure to unad- mitted evidence, see United States v. Barnes, 747 F.2d 246, 251 (4th Cir. 1984); see also Hinkle v. City of Clarksburg, 81 F.3d 416, 427 (4th Cir. 1996) (following Barnes without citing Remmer), I believe that I am bound to treat the error here as trial error, subject to harm- less error analysis. I therefore turn to that analysis. II. On collateral review, we may set aside a conviction only if con- vinced that the asserted error "`had substantial and injurious effect or influence in determining the jury's verdict.'" Brecht v. Abrahamson, 507 U.S. 619, 623 (1993) (quoting Kotteakos v. United States, 328 U.S. 750, 776 (1946)).4 If we find the issue so close as to leave us "in _________________________________________________________________ 4 Like the majority, I believe that our decision in Smith v. Dixon, 14 F.3d 956 (4th Cir.) (en banc), cert. denied, ___ U.S. ___, 115 S. Ct. 129 (1994), established that the Brecht standard applies on collateral review, even when the state court has failed to find an error harmless beyond a reasonable doubt. 34 virtual equipoise as to the harmlessness of the error," we must resolve the question in favor of the petitioner and grant habeas relief. O'Neal v. McAninch, ___ U.S. ___, 115 S. Ct. 992, 994 (1995). In the instant case, the state trial court held a hearing on Sherman's motion for a new trial, and juror Miller was called to testify on Sher- man's behalf. On direct examination, the juror testified that two or three days into the trial, he and his wife drove to Gibson Manor, "looking for [the] tree that was so involved in the case." He stated that he did, in fact, find the tree as well as the house in which the murders occurred. Apparently wary of violating the secrecy of the jury's delib- erations, and consistent with the State's objections, the court did not permit Sherman's counsel to ask the juror why he visited the site. On cross-examination by the State, the juror testified that by the time he visited the site, he had already seen (in court) the videotape of the outside of the Sherman house, but was not sure if he had yet seen the aerial photographs of the neighborhood. When asked whether, at the time of his site visit, several photographs of the tree had been introduced, the juror responded, "Yes, the tree had been in question, and that's one of the reasons I went there." Finally, the State asked the juror whether other photographs of the scene had been introduced by the time he made his visit, to which he responded, "I'm not sure, sir. I'm not sure what sequence -- the reason why I went there was so I could see the tree that was so much in question." The State made no further inquiry into the facts and circumstances sur- rounding the juror's site visit. All told, the record regarding the asserted error provides the fol- lowing information: (1) Two or three days into the trial, the juror drove to the neighborhood in which the murders occurred, looking for the tree in which the murder weapon was found. (2) The juror saw the Sherman house and the tree. (3) This visit occurred after a videotape of the outside of the Sherman house was played for the jury, and after the jury saw several photographs of the tree. 35 (4) The juror's stated purpose for this site visit was to "see the tree that was so much in question." To this day, no details of the juror's site visit are known. We do not know, nor did the state trial court know, for example, the length of the juror's site visit; how close he came to the tree; whether he got out of his car and walked to the tree or simply drove past; whether he performed "experiments" based on the testimony about the tree or merely looked at it to ascertain its size and location. Despite the spar- sity of the record, the state trial court determined that Sherman suf- fered no prejudice as a result of the juror's site visit.5 I cannot agree. Based on the extremely limited information before us regarding the juror's site visit, it is simply impossible to ascertain whether and to what extent Sherman was prejudiced by the visit. The juror's testi- mony is consistent with a visit in which he simply drove through Gib- son Manor and looked at the house and the tree as he drove by. Such a visit, if properly subject to harmless error analysis, might well be found to be harmless.6 However, the juror's testimony would be equally consistent with a scenario in which, dissatisfied with the evidence presented at trial, he set out to conduct his own investigation, including taking measure- ments of the tree and its distance from the house and attempting to hide items in the tree in the manner in which the murder weapon was found. Such a visit would certainly be prejudicial. Contrary to the majority's suggestion, Maj. Op. at 13, it would not be merely "cumu- lative" of the other evidence at trial, as the proper interpretation of that evidence was disputed. For example, perhaps confused by the parties' conflicting interpretations of Officer Hopkins' testimony, the _________________________________________________________________ 5 In my view, the Antiterrorism and Effective Death Penalty Act of 1996, Pub. L. 104-132, 110 Stat. 1214 (1996), does not apply retroac- tively to cases like the one at hand. See Landgraf v. USI Films Prods., ___ U.S. ___, 114 S.Ct. 1483, 1505 (1994). 6 As noted above, this does not preclude characterization of that error as structural. The harmless error inquiry focuses on the impact of an error on the verdict rendered. Structural error is concerned with the integ- rity and fairness of the trial process, not the impact of a particular error on the actual verdict. 36 juror sought to resolve for himself whether someone of Sherman's height and build could have wedged the gun in the tree in the manner in which it was found. Alternatively, perhaps the juror wanted to test whether it was possible to hide something in the tree without getting sap and pine needles on one's clothing, or wanted to see whether it made sense for Sherman to have stopped at the tree en route to his grandparents' house. Each of these scenarios would result in the dis- covery of new, rather than cumulative, information. The State of Maryland seems to concede, as it must, that the record in this case with respect to the juror's site visit leaves much to be desired. The State argues, however, that "Sherman had the burden to show that an error warranting new trial relief had occurred," Brief of Appellees at 28, and thus should bear the consequences of the insuffi- cient record. This argument is foreclosed by Remmer, which man- dates that any private communication with a juror about the matter pending before the jury is "presumptively prejudicial," and the government bears the heavy burden of rebutting that presumption. 347 U.S. at 229. Moreover, we have recognized as recently as a few months ago, in a case that the majority neither overrules nor acknowl- edges, that this presumption governs when jurors consider evidence not admitted at trial. See Hinkle, 81 F.3d at 427. See also United States v. Brooks, 957 F.2d 1138, 1142 (4th Cir.), cert. denied, 505 U.S. 1228 (1992); United States v. Greene, 834 F.2d 86, 88 (4th Cir. 1987); Barnes, 747 F.2d at 250-51. Accordingly, while Sherman bore the initial burden of proving that the site visit occurred,7 once that fact was established, the burden shifted to the State to"demonstrat[e] the absence of prejudice." Stockton, 852 F.2d at 743. _________________________________________________________________ 7 Remmer requires that the improper jury contact be "about the matter pending before the jury." 347 U.S. at 229. Stockton likewise requires that the defendant establish that the contact "was of such a character as to rea- sonably draw into question the integrity of the verdict." 852 F.2d at 743. Our cases applying the Remmer presumption where the jury has been exposed to evidence not admitted at trial suggest that, in such cases, the defendant need only establish that the jury has, in fact, been exposed to the unadmitted evidence to trigger the presumption. See, e.g., Hinkle, 81 F.3d at 427. In any event, in this case, by establishing that a juror had engaged in an unauthorized site visit and that the physical characteristics of that site were of critical importance, Sherman satisfied the additional requirements described in Remmer and Stockton. 37 Concededly, due to the significant restrictions the rules of evidence place on questioning of jurors, it would have been difficult, if not impossible, for the State to prove by direct evidence that the juror's site visit did not prejudice Sherman. We have suggested that "the state may rebut the presumption of prejudice through whatever circumstan- tial evidence is available, including juror testimony on the facts and circumstances surrounding the extraneous communication [or receipt of unadmitted evidence]." Id. at 744. Ultimately, however, because "[t]he right to an impartial jury belongs to the defendant," the risk of being unable to prove the impact of an improper jury contact or a jury's exposure to unadmitted evidence is properly borne by the State. Id. at 743-44. In this case, the State simply failed to meet its burden. The only circumstantial evidence the State presented to meet its heavy burden was the juror's testimony that at the time he made the visit, the jury had already seen a videotape and some photographs of the area in question. While the State argues (and a majority of this court finds) that the site visit was merely cumulative of the other evi- dence presented at trial regarding the tree, the juror's testimony in response to the State's questions indicates otherwise. When asked whether he had seen the photographs of the tree prior to his visit, the juror responded that that was why he went to the site -- because he wanted to see the tree "in question." This implies a causal relationship between the photographs and the visit; the juror was not satisfied with the photographs and visited the site to obtain additional information about the tree. As my dissenting colleagues note, the district court, which nonetheless found the site visit to be harmless error, remarked that the state's close-up photograph of the tree"does not show me anything on its face." The majority makes much of the "powerful array of evidence pres- ented at trial" that supports the jury's guilty verdict. This miscon- ceives the appropriate harmless error inquiry, which focuses not on the sufficiency of the evidence absent the error, but rather on the impact of the error on the jury's verdict. See Sullivan, 508 U.S. at 279 ("Harmless error review looks . . . to the basis on which the jury actually rested its verdict. . . . not [to] whether, in a trial that occurred without the error, a guilty verdict would surely have been rendered.") (internal citation and quotation omitted). Accordingly, the majority's characterization of Sherman's argument as merely a"roundabout" 38 attack on the sufficiency of the evidence against him is off the mark. Sherman does not challenge the sufficiency of the evidence presented at trial, he challenges the fairness of the trial and "the basis on which the jury actually rested its verdict." Id. In summary, the record in this case reveals that a juror in Sher- man's capital murder trial made an unsupervised and unauthorized visit to a site whose physical characteristics were disputed and critical to the State's case. Unfortunately, the record does not reveal much else about that visit. At the post-trial hearing, the burden was on the State to rebut the presumption of prejudice raised by the juror's site visit, and the State must therefore bear the consequences of the spar- sity of the record. Given the lack of detail as to the juror's site visit and the wide range of conduct that his testimony might describe, the majority's conclusion that the site visit was harmless is tantamount to a conclusion that a juror's unauthorized site visit can never be preju- dicial error. I cannot agree with this conclusion. While some of the possibilities encompassed by the juror's ambiguous testimony might be harmless, many of them would certainly be prejudicial. Without any evidence as to the details of the juror's site visit, I find myself "in virtual equipoise as to the harmlessness of the error." O'Neal, 115 S. Ct. at 992. My "grave doubt as to the harmlessness of [the] error" requires me to resolve the issue in Sherman's favor. Id. at 999. The "obviousness" of Sherman's guilt to a reviewing court, or to the public for that matter, is irrelevant absent a proper jury verdict. I would, therefore, grant the writ of habeas corpus, so that Sherman could receive a new trial in which the jury returns a proper verdict, based solely on the evidence presented at trial. The Constitution requires nothing less. 39