dissenting:
The majority holds that St. Luke Gregory’s conviction for taking sexual advantage of his daughter must be overturned because one of two virtually identical hearsay statements identifying him as the assailant was, in its view, erroneously admitted into evidence. At bottom, the majority doubts the reliability of a jury verdict in a case where a father and his three-year-old daughter had simultaneously been diagnosed with gonorrhea, a sexually transmitted disease. If this verdict is unreliable, then I fear reliability in prosecutions for the sexual abuse of very young children is beyond our grasp. Because the evidence supports this conviction beyond any reasonable doubt, I would hold any error in the admission of the statement harmless and reverse the grant of habeas corpus by the district court.
The pervasive error the majority commits is plain. It will not permit the 1982 diagnosis of gonorrhea to support the admission of the June 1984 hearsay statement because the diagnosis and the statement were not “contemporaneous.” It holds that, un*711der Ohio v. Roberts, 448 U.S. 56, 66, 100 S.Ct. 2531, 2539, 65 L.Ed.2d 597 (1980), the “particularized guarantees of trustworthiness” for hearsay statements must be contemporaneous as well. The Roberts Court did not embrace that narrow view, and I dissent from its adoption by the majority. Under the majority’s analysis, no evidence of prior sexual abuse could support the trustworthiness of a child’s later statement. The Eighth Circuit took a contrary position in United States v. Dorian, 803 F.2d 1439, 1444-47 (1986), and I agree with its reasoning.
I.
At issue here are two statements made by three-year-old LaTonya in June, 1984 and on September 7, 1984 as reported by her grandmother, Doris Griffin. One morning in June, immediately after her father dropped her off at Griffin’s house, LaTonya stated, “My daddy pooted in my butt.” Her grandmother told the child to “Shut up,” and notified LaTonya’s mother of the statement. Similarly, on September 7, again after her father had delivered her to Griffin’s house, LaTonya stated, “My daddy put it in my butt.” On that occasion, LaTonya also was found by her grandmother to have a thick, white discharge between her legs and she complained of pain in her genital area. The physician who examined LaTonya that afternoon in connection with these complaints found the presence of redness, irritation and pus-like discharge in her vaginal area. He diagnosed the condition as vaginitis, which he testified “fully supported” the child’s statements describing sexual abuse. The district court held that the September statement was reliable.
This case is paradigmatic of situations where there may have been repeated instances of child abuse and the child may well have made statements about a number of them. See, e.g., Morgan v. Foretich, 846 F.2d 941, 945-46 (4th Cir.1988). The majority bases its decision, in part, on the view that LaTonya was too young to give reliable testimony, but if that were the end of the matter, the protection offered by law to these most vulnerable of beings would be scant. In this case, both statements were more supportable than the majority believes, and any error in the admission of the June statement was clearly harmless.
II.
The harmlessness of any error here should be apparent. See Delaware v. Van Arsdall, 475 U.S. 673, 684, 106 S.Ct. 1431, 1438, 89 L.Ed.2d 674 (1986). The June statement was unimportant in light of the state’s otherwise strong ease against Gregory. See United States v. Crockett, 813 F.2d 1310, 1315-16 (4th Cir.1987). Gregory was not convicted in connection with any events surrounding the June statement. The September statement already identified Gregory and implicated him as the perpetrator of the September 7 crimes for which he was found guilty. The majority expresses concern that the jury may have taken the June statement as “sound corroboration” for LaTonya’s identification of her father in the September statement. However, the September statement was soundly corroborated by no less than twelve separate factors enumerated by the district court. I summarize them briefly here: (1) LaTonya was in the physical custody of her father immediately before the September 7 statement was made; (2) Prior to arriving at her grandmother’s house at 6:05 A.M. on September 7, LaTonya would have been at home in bed, a fact compatible with the child’s statement that her father molested her in bed while her mother was asleep; (3) LaTonya’s grandmother inferred that sexual abuse had occurred because she found an undried substance on the child’s underpants which subsequent medical tests determined to be the result of vaginitis; (4) A medical examination of LaTonya on the afternoon of September 7 revealed inflammation and irritation of the child’s vagina and was consistent with LaTonya’s statement; (5) LaTonya had made a similar statement in June 1984; (6) In 1982, LaTon-ya and her father simultaneously had been diagnosed with gonorrhea, a sexually *712transmitted disease;1 (7) LaTonya’s statements to the doctor outside of the presence of her grandmother that her daddy had “unzipped his pants” and “told me do my legs apart” and “hurt me here [pointing to between her legs]” corroborated the statement; (8) LaTonya would be motivated to be truthful in her statements to the doctor because she perceived him as trying to help her; (9) LaTonya’s version of the events had no internal inconsistencies; (10) A child of age three would be unlikely to spin a fabrication out of self-interest; (11) There is no evidence to suggest that LaTonya was influenced or coerced to identify her father as the perpetrator of sexual abuse; (12) Since LaTonya was within the parental control of her father and her negative statements could have made him angry, they could be regarded as statements against her personal interest.
The June statement was at most cumulative of the September statement and the “other overwhelming and largely uncontro-verted evidence properly before the jury.” Brown v. United States, 411 U.S. 223, 231, 93 S.Ct. 1565, 1570, 36 L.Ed.2d 208 (1973). The June and September statements were virtually identical except for the substitution of the words “pooted in" for “put it in,” and the implication of sexual molestation implicit in each is essentially the same. The June statement did not contradict either the later statement or any other evidence offered in the case. At most it tended to corroborate other, more detailed and admissible evidence. See Schneble v. Florida, 405 U.S. 427, 431, 92 S.Ct. 1056, 1059, 31 L.Ed.2d 340 (1972).
It was clear to the North Carolina Court of Appeals that any error here was nonprejudicial. See North Carolina v. Gregory, 78 N.C.App. 565, 338 S.E.2d 110, 113 (1985) (“in the light of all the circumstances of this case, the admission of the statement is at worst non-prejudicial error”). Why the majority rejects this conclusion is beyond me.2 Neither protection of the innocent nor respect for the criminal process is furthered by focusing on the “virtually inevitable presence of immaterial error” rather than upon “the underlying fairness of the trial.” Van Arsdall, 475 U.S. at 681, 106 S.Ct. at 1436. The jury believed this child was sexually molested by her father, and the majority offers no basis for disbelief. Retrials, which may be less contemporaneous, more rehearsed, and less reliable, should be reserved for situations where it is clear that “there is a reasonable possibility that the improperly admitted evidence contributed to the conviction.” Schneble, 405 U.S. at 432, 92 S.Ct. at 1059.
III.
The majority also refuses to consider the 1982 diagnosis of gonorrhea as part of the totality of circumstances that support the trustworthiness of LaTonya’s June statement. The majority apparently concludes that only “contemporaneous” circumstances may be considered as “particularized guarantees of trustworthiness” under Ohio v. Roberts, 448 U.S. 56, 66, 100 S.Ct. 2531, 2539, 65 L.Ed.2d 597 (1980). Neither the *713language in Ohio v. Roberts, nor the analogous language of Fed.R.Evid. 803(24) and 804(b)(5), nor the cases cited by the majority construing these rules refer to any contemporaneity requirement. Even if I agreed with the majority’s conclusion that only contemporaneous circumstances may be considered under the above hearsay exceptions, the Confrontation Clause, which governs here, is not merely a “codification of the rules of hearsay,” California v. Green, 399 U.S. 149, 155-56, 90 S.Ct. 1930, 1933-34, 26 L.Ed.2d 489 (1970), and Ohio v. Roberts makes no reference to the tempo-rality of the required “particularized guarantees of trustworthiness.”
Confrontation Clause analysis has always required an evaluation of each case on its facts rather than the application of a mechanical test. Barker v. Morris, 761 F.2d 1396, 1400 (9th Cir.1985). The test is “whether the factors surrounding the making of the out-of-court statement, taken as a whole, indicate trustworthiness.” Id. at 1403. Contemporaneous corroboration is one, but only one, such factor. For example, in United States v. Dorian, 803 F.2d 1439, 1444-47 (8th Cir.1986), the court considered, under both Fed.R.Evid. 803(24) and Ohio v. Roberts, the trustworthiness of hearsay statements made by a five-year-old sexual abuse victim nearly one month after the alleged abuse could have occurred. The court concluded that the statement was trustworthy under both the Federal Rules and the Confrontation Clause based upon evidence spanning the period between the abuse and the statement, including the child’s behavior on various occasions, statements made by her mother, inculpatory actions by the defendant, and medical evidence obtained five days after the alleged abuse and three weeks before the statement. Under the majority’s analysis, it is doubtful that any of these non-contemporaneous factors could have been considered. See also Nelson v. Farrey, 874 F.2d 1222, 1234 (7th Cir.1989) (Flaum, J., concurring) (particularized guarantees of trustworthiness included child’s behavior at the prospect of seeing her allegedly abusive father approximately three to eleven months before the hearsay statements were made).
To be sure, circumstantial guarantees of trustworthiness will often be contemporaneous. However, circumstantial evidence as a category includes many kinds of indirect evidence — both contemporaneous and non-contemporaneous — from which to evaluate the reliability of a statement. The importance of recognizing non-contemporaneous factors may be particularly great in child abuse prosecutions because “significant delays in reporting this abuse may occur because of confusion, guilt, and fear on the part of the child.” Morgan, 846 F.2d at 947. See Dorian, 803 F.2d at 1444 (“it frequently takes a long time for children to share what is really going on and they may then do so in stages, telling a little more each time”).
Courts should ask simply whether a circumstantial guarantee of trustworthiness is a strong one, not whether it does or does not fit into some judge-made category of “contemporaneity.” Appellate judges buy nothing but trouble when they attempt to prescribe evidentiary litmus tests as a substitute for the reliability determinations of the presiding judge at trial. The drafters of the Federal Rules of Evidence and the Supreme Court majority in Ohio v. Roberts avoided this pitfall, and I am sorry to see the majority slip into it. The majority’s position will inevitably lead to uncertainty as to how close in time to the statement a guarantee of trustworthiness must be to qualify as a contemporaneous one.
The artificiality of the majority’s focus on contemporaneity is amply illustrated by its discussion of the June statement. The trustworthiness of LaTonya’s earlier statement is not merely a matter of “bootstrapping” it by the September events into a category of reliability, as the majority claims. Nor is her statement the likely product of a childish fantasy or some televised depiction. No fewer than seven out of the twelve factors which the district court found to be sound corroboration for the September statement apply equally to the June statement, including the earlier “non-contemporaneous” diagnosis of gonorrhea. The majority presumably holds this diagnosis to be irrelevant because it *714occurred in 1982, but I would consider it, albeit cautiously, as evidence of trustworthiness because the subject of the simultaneous diagnosis is the very same person about whom LaTonya later voiced her complaint and because there was no explanation, other than sexual contact with her father, offered for this child’s venereal infection, Given the well-corroborated September statement and the circumstantial evidence that does support the trustworthiness of the June statement, any error in its admission was of the most marginal sort.
IV.
The majority expresses an interest in “contemporaneity” as a guarantor of reliability, but the irony of its position apparently escapes it. The result of its holding is the need for a retrial, if the witnesses can be located and remain in health (if witnesses are unavailable, the retrial may compound the very problem of hearsay evidence that gave rise to this case). Nothing about the retrial will be in any way contemporaneous: the passage of time only renders evidence more suspect, not less.
Writs of habeas corpus ring hollow at this belated hour and after careful state court consideration of the case. The seeming flaw we have detected after flyspecking state process does not warrant resurrection of what has long been laid to rest. I respect the majority’s concern for the perils inherent in child abuse prosecutions which may involve the most baseless and even vengeful accusations. Nothing of that sort is hinted here, however. Harmless-error doctrine recognizes that trial is an ordeal to which citizens should not lightly be subjected twice. By declining to apply the doctrine, the majority renders an event six years in this family’s past one it may prove unable to forget.
. The majority suggests that the diagnosis does not necessarily imply a sexual contact between LaTonya and her father, noting the fact that LaTonya's uncle, Gideon Griffin, also tested positive for gonorrhea in 1982. However, there is no evidence in the record that Gideon Griffin was ever implicated by LaTonya or anyone else as having had sexual contact with her or that he ever had sole physical custody of the child. At Gregory's trial the parties stipulated to the fact of Gideon Griffin's 1982 diagnosis, but no exculpatory use was ever made of it, and the district court noted but attached no significance to it in its own analysis. I fail to see why the simultaneous diagnosis of gonorrhea in LaTonya and her father is somehow less significant or trustworthy because another individual, against whom no evidence has ever been offered, con-traded a venereal disease. Indeed, Gregory himself makes no such argument in his brief on appeal.
. The majority is properly worried about resting its analysis of the harmless error point solely on the grounds of waiver. This court has, for obvious reasons, always been cautious about saddling parties with "concessions” made at oral argument. The appellee vigorously argued the issue of harmless error in its brief, presumably to counter the government's presentation of the case against Gregory apart from the June statement. Addressing the harmless error question directly is preferable to finding a concession at argument which frankly seems at odds with much of the government’s position in its brief.