St. Luke Gregory, imprisoned by North Carolina for taking sexual advantage of his young daughter, brought this Section 2254 case challenging the admission of two hearsay statements identifying him as the assailant. Gregory argues that the statements were so untrustworthy as to have violated his Sixth Amendment right to confront his accuser. The district court agreed that statements the child made several months before the crimes occurred were too unreliable to support the conviction, and ordered relief in the form of a new trial. We affirm.
I.
A jury concluded that Gregory had sexually assaulted his daughter LaTonya, then three-and-a-half years old, in September, 1984. Some of the most damning evidence against Gregory came in testimony from LaTonya’s maternal grandmother, Doris Griffin, who often babysat the child while Gregory and his wife, LaTonya’s mother, worked.1
In June, 1984, and again on September 7, 1984, Gregory dropped LaTonya off at Griffin’s home and proceeded to work. During both visits, LaTonya exclaimed something like “Daddy put it in my butt”, a remark the parties agree connotes a sexual *706contact between father and daughter.2
The September remark finds corroboration in Griffin’s discovery of a substance, apparently pus, on LaTonya’s panties and testimony that LaTonya would not allow her to touch her crotch because it hurt. Dr. Phillip Greene, who examined LaTonya later that day, testified that “[LaTonya] said that her daddy unzipped his pants and “told me do my legs apart,” and she showed me by the spreading of her legs, and he — those are my words. These are her words: “Wanted to get close to me,” and the fourth part of this continuous sentence was that “that he hurt me here,” and she pointed to between her legs.”3
No contemporaneous evidence corroborates LaTonya’s June statement. Griffin noticed nothing in LaTonya’s appearance or conduct that triggered suspicion, and seems to have dismissed the remark as a naive impertinence.
A Dr. Beals testified that on September 13, 1982, he had diagnosed gonorrhea in the seventeen-month-old LaTonya. Beals observed that sexual congress is the means of transmitting gonorrhea “almost in every case known”, and that the probability of infection by other means is very slight. Other evidence indicated that Gregory had been treated for gonorrhea at about the same time.4
*707The district court accepted the magistrate’s recommendation that it rule the June hearsay inadmissible, finding “absolutely no objective corroborative evidence of the abuse reported at that time.” The court viewed LaTonya’s September statements to Griffin and Greene in a different light, rejecting the magistrate’s counsel that it hold these two inadmissible also. Identifying a dozen factors corroborating the September statements, the district court concluded that while it “is on the cutting edge of the issue of relaxation of confrontation clause requirements”, the in-dicia of reliability associated with the statements warranted admission.5
II.
The issue for our review is one arising with depressing frequency in our nation’s courts, one that, as a sister Circuit has observed, “place[s] a strain on traditional notions of procedural justice.” Nelson v. Farrey, 874 F.2d 1222, 1224 (7th Cir.1989), cert. denied, — U.S. -, 110 S.Ct. 835, 107 L.Ed.2d 831 (1990) (citations omitted). The victim and principal witness to the crime charged is, by virtue of the youth that made her so vulnerable, unable to give reliable testimony. Her absence deprives the accused of the opportunity to cross-examine his accuser, an opportunity that, though some aver it a poor way to elicit truth from a very young child, our system holds dear. Id. at 1230; The Supreme Court, 1987 Term: Leading Cases, 102 Harv.L.Rev. 143, 157 (1988).
In Ohio v. Roberts, 448 U.S. 56, 100 S.Ct. 2531, 65 L.Ed.2d 597 (1980), the Court promulgated a rule mediating between the accused’s interest in confrontation, [and broadly in the best exposition of the truth,] and the victim’s and society’s interest in the effective prosecution of child abuse. The rule allows the admission of hearsay evidence from an unavailable declarant if, though not within any well-recognized exception to hearsay exclusion, the evidence bears “particularized guarantees of trustworthiness.” 6 Id. at 66, 100 S.Ct. at 2539. It is the district court’s Roberts analysis that we review today.
In this case, the issue is whether we may look a considerable distance backward and forward from June of 1984 — principally to the 1982 diagnosis of gonorrhea and the events of September of 1984 — for corroborating factors. This issue is novel to us. The district court, referring to a portion of the magistrate’s decision citing cases from two other courts of appeals, stated that it would address the statements’ reliability in light of contemporaneous circumstances. See United States v. Renville, 779 F.2d 430, 440 (8th Cir.1985) (“[T]he reliability of the declaration is assessed in light of the circumstances at the time of the declara*708tion and the credibility of the declarant.”) (citing Huff v. White Motor Corp., 609 F.2d 286, 292 (7th Cir.1979)); see also United States v. Dorian, 803 F.2d 1439, 1444 (8th Cir.1986) (quoting the Renville contemporaneity requirement); United States v. Love, 592 F.2d 1022, 1027 (8th Cir.1979). As we shall explain, even if we disagreed with the well-reasoned requirement of contemporaneous corroboration, a requirement implicit in the Roberts insistence on “particularized” guarantees of reliability, we nonetheless would hold the chronologically distant evidence presented in this case to be a logically insufficient index of the reliability of the June hearsay.
No doubt because of the relative novelty of the Roberts exception and the fact-specific examination each Roberts inquiry demands, the cases, rather than publishing a comprehensive list of possible corroborating factors, tend to consider what, in the declarant’s condition or otherwise, suggests truthfulness or a lack thereof. For example, in Ellison, the district court referred to the many discrepancies among the five-year-old sexual assault victim’s descriptions of her assailant and the place of her assault in concluding that admission of the descriptions as hearsay violated the accused’s Sixth Amendment rights. Ellison v. Sachs, 583 F.Supp. 1241, 1249 (D.Md.1984), aff'd, 769 F.2d 955, 957 (4th Cir.1985) (“[A]s the district court documented in careful detail, there are serious discrepancies which indicate that the victim’s out-of-court statements and identification of Ellison were not at all reliable.”)
In Morgan v. Foretich, 846 F.2d 941 (4th Cir.1988), one issue before the panel was whether the district court had abused its discretion in ruling five statements made by a young girl to her mother, all suggesting sexual abuse by the girl’s father, inadmissible under the excited utterance exception to the hearsay exclusion.7 Fed.R.Evid. 803(2). We noted “substantial physical evidence and doctors’ testimony” corroborating the girl’s tales of sexual assaults, which she described in childish terms but with an “extensive knowledge of sexual activities,” both characteristics belying an attempt at fabrication. 846 F.2d at 948. The girl’s near-hysterical condition left “little doubt ... that she was acting under the stress of the situation.” Id.
We find it impossible to weave the factors influential in Ellison and Morgan, or the principles informing the factors, into a justification for the admission of the June hearsay. As we have said, there is no corroborating evidence peculiar to the June statement. The September developments alone cannot bootstrap the earlier statement into a category of reliable hearsay. It is a considerable leap from the varied evidence that Gregory sexually abused La-Tonya in September to the conclusion that LaTonya’s tantalizing hint of an earlier assault must also be true. In this case, at least, where the trial court ruled the declar-ant incompetent because of her proclivity to “make up things,” including things about her father, and inability to appreciate the need for truthfulness, we are unable to see how the corroborated September statement relates favorably enough to the truthfulness of the June statement to overcome the complete absence of contemporaneous corroboration for the latter. See Dorian, 803 F.2d at 1444 (emphasizing the declar-ant’s credibility as a factor in evaluating reliability); Renville, 779 F.2d at 440 (same).
We consider first the substance of the statement itself, then consider whether evidence extrinsic to the statement reflects favorably on its truth.8 Without making *709light of what certainly happened to LaTon-ya in September, we can imagine a very young child, particularly one so bright, blithely making a sexually charged reference based on something she encountered on television or elsewhere.9 The reference might improperly describe an innocent contact by the father. Not a detailed account of an event ordinarily beyond a child’s experience, LaTonya’s remark is less credible per se than the tales of abuse recounted in Ellison and Morgan. Cf. Ellison, 769 F.2d at 957 (“[S]everal courts and commentators have observed that a young child’s description of a sexual assault may, in particular circumstances, contain its own inherent verity.”) (citations omitted). The June remark did not indicate when the contact had occurred, leaving unclear whether Gregory might have had the opportunity to commit the assault.
At least as troubling as the problematic significance of the June remark in se is the lack of any extrinsic evidence indicating LaTonya had undergone a recent sexual assault. LaTonya made her remark only minutes after Gregory left, when the excitement and fear induced by her assailant’s presence would likely persist. Griffin, however, familiar as she was with La-Tonya’s personality and solicitous, as we infer, of her welfare, read nothing in her manner to lend credence to the remark.10
Reflection on the corroborative value of the 1982 diagnosis of gonorrhea reveals many of the problems we associate with corroboration by the September evidence. The conclusions we must make to link the diagnosis and the June statement are these: the diagnosis implies a sexual contact between Gregory and LaTonya, and the statement reflects either that contact or one closer in time to June, 1984. We will assume the truth of the first and consider in turn the reliability of the second and third conclusions.11
We cannot confidently conclude that an instance of abuse in 1982 inspired the June statement. We must resist the temptation to decide that, because we would likely find the diagnosis sufficient corroboration for the remark if they had occurred at about the same time, the two years intervening between the events is insignificant. On the contrary, the time is likely more significant in the context of a very young declarant. We are not sure, and the record holds no evidence, that LaTonya would have remembered a sexual incident occurring at seventeen months, at least in the terms she used in the June statement. LaTonya could not have made the statement under the excitement of the two-year-old incident, at least not in any sense of excitement yet recognized under 803(2). Absent any evidence why LaTonya would have kept quiet about the incident for so long, or by contrast why she would have chosen June, 1984, to speak of it, the time between diagnosis and statement gives ample reason to doubt a causal relationship.
The circularity of the second conclusion, that the remark suggests a pattern of abuse, is clear. While we assume sexual abuse in 1982, and are confident of abuse in September, 1984, only by assuming as well what we are charged with deciding— that the June, 1984, remark reveals a third incident — may we conclude that the remark *710is reliable in indicating a pattern corroborated by the evidence of a 1982 contact. The remark is, on the evidence, as consistent with two discrete incidents as with a pattern of which these corroborated incidents are but a part. Without some evidence suggesting abuse between 1982 and September, 1984, then, there is no sound basis to extrapolate an intervening incident. This tautology, and the temptation to ignore it, confident that a young child would not suggest sexual abuse if it had not occurred, point up the wisdom of a rule requiring at least some contemporaneous corroboration.
III.
We now consider whether admission of the June hearsay was harmless beyond a reasonable doubt.12 The district court agreed with the magistrate’s conclusion that the error was not harmless, “inasmuch as [LaTonya’s] identification of [Gregory] was the sole source of identification as to the perpetrator of the crime” and notwithstanding its decision that the September hearsay was admissible.
We similarly believe that the jury might well have taken LaTonya’s June statement, phrased so similarly to her September exclamation to Griffin, as sound corroboration for LaTonya’s September identification of Gregory as her assailant. Without the June statement, and with some benign means of contracting gonorrhea a possibility, the jury might have had less confidence in LaTonya’s identification.13 Uncertain whether the June hearsay was significant to the jury, we must hold admission of the hearsay harmful error warranting habeas relief.14
III.
For the reasons presented in the preceding sections, we affirm the district court’s decision finding a violation of Gregory’s rights under the Confrontation Clause and ordering relief in the form of a new trial.
AFFIRMED.
. LaTonya did not testify. N.C. Superior Court Judge Small presided over Gregory’s trial and the voir dire hearing on Gregory’s motion to suppress LaTonya’s testimony. The judge observed that LaTonya, while brighter than average, could not intelligibly express what had happened, and could not appreciate her obligation to tell the truth. Though physically available and, apparently, present for at least some of the trial, LaTonya was declared an incompetent witness, leaving hearsay the only route by which her story could come before the jury.
.What Griffin reports LaTonya to have said in June was "Grandmama, my daddy pooted in my butt.” We believe, and the parties impliedly agree, that this statement imported to the jury the same message as the less ambiguous September remark "Daddy put it in my butt.”
Griffin's testimony follows verbatim. After describing her discovery of what looked like pus between LaTonya’s legs on September 7, the exchange between Griffin and the prosecutor was as follows:
Q: First, did she [i.e. LaTonya] tell you something?
A: Yeah. She told me her daddy put it in her butt.
******
Q: Okay. And did she point or anything like that at that time?
A: Yeah. She pointed down there.
Q: All right. Where did she point to?
A: In front. Below.
A: Okay. After she said that "My daddy put it in my — ” Were those her words, "My daddy put it in my butt?”
A: Yes.
Griffin subsequently related what passed between her and LaTonya several months earlier.
Q: Okay. How long — how much time was it before the September 7th incident?
A: About three months.
******
Q: Where was it that [LaTonya] spoke to you?
A: In the house. As you enter the house.
Q: About what time was it?
A: About six or a little after six.
Q: In the early morning?
A: Yes.
Q: And how had she gotten to your house?
A: I think her father brought her.
Q: Okay. And what was she doing at your house?
A: I kept her days.
Q: You were babysitting then?
A: Yes.
Q: And what, if anything, — okay. How long had she been in the house when she spoke to you?
A: She just spoke and then she told me that.
Q: Okay. What was the first thing she said to you?
******
A: The first thing she said to me, she said, "Grandmama, my daddy pooted in my butt.”
Q: And what did you say to her?
A: I told her, "Shut up." I would tell her mother when she came in from work. ******
Q: Okay. From that day up until September the 7th, 1984, at any time did she ever discuss the — did you ever discuss the matter with LaTonya again?
A: No, — no.
. The trial court allowed Greene to relate La-Tonya’s statements under N.C. R.Evid. 803(4), which allows for the admission of hearsay statements given for purposes of medical diagnosis. The North Carolina Court of Appeals, in its decision reversing Gregory’s conviction in part, affirmed the application of 803(4) to Greene’s testimony, and held the provision applicable as well to Griffin's hearsay testimony about what LaTonya told her on September 7. State v. Gregory, 78 N.C.App. 565, 338 S.E.2d 110 (1985).
The Court of Appeals also suggested that Griffin’s hearsay testimony about LaTonya’s June statements, admitted under the excited utterance exception of N.C.R.Evid. 803(2), might have come in under 803(4) as well. Holding the hearsay at worst harmless error, the court found it unnecessary to speak firmly to the admissibility of the September 7 hearsay. The court similarly held the admission of the gonorrhea test results no more than harmless error.
. The trial court admitted proof of LaTonya’s and Gregory’s treatment for gonorrhea under N.C.R.Evid. 404(b), as evidence of a prior bad *707act showing identity. Griffin testified that she had seen the pus-like substance she discovered on September 7 once before, when LaTonya was seventeen months old, but did not state to the jury that the earlier observation signaled gonorrhea.
. Gregory has not cross-appealed from the portion of the district court's order condoning admission of the September hearsay through the testimony of Griffin and Greene. Our analysis accordingly treats only the admissibility of the June hearsay.
. The threshold test for admissibility under Roberts is whether the hearsay declarant is "unavailable" to testify. 448 U.S. at 66, 100 S.Ct. at 2539. Roberts left unclear whether the unavailability requirement applies with equal force in all cases, suggesting as it did the option of compressing its two-part test into a single inquiry into reliability when the utility of confrontation is shown to be remote, as when the declarant is too immature for meaningful cross-examination. 448 U.S. at 65 n. 7, 100 S.Ct. at 2538 n. 7 (citing Dutton v. Evans, 400 U.S. 74, 91 S.Ct. 210, 27 L.Ed.2d 213 (1970)); see also United States v. Inadi, 475 U.S. 387, 396, 106 S.Ct. 1121, 1126, 89 L.Ed.2d 390 (1986).
We have, like other circuits, as a general matter held child abuse victims found incompetent to be "unavailable” under Roberts. Ellison v. Sachs, 769 F.2d 955, 957 n. 4 (4th Cir.1985); see also United States v. Dorian, 803 F.2d 1439, 1446-47 (8th Cir.1986); Haggins v. Warden, 715 F.2d 1050, 1055 (6th Cir.1983); United States v. Iron Shell, 633 F.2d 77, 87 (8th Cir.1980), cert. denied, 450 U.S. 1001, 101 S.Ct. 1709, 68 L.Ed.2d 203 (1981); United States v. Nick, 604 F.2d 1199, 1202 (9th Cir.1979). We agree with the district court that LaTonya, incompetent due to immaturity, was "unavailable” and pass to the knottier question of whether her June statement bore sufficient indicia of reliability to have comported with Gregory's Sixth Amendment right to confront her.
. Morgan arose as an appeal from a jury verdict for the defendants in a suit by mother and daughter against the father and his parents for damages arising from the alleged sexual abuse. The contested evidentiary rulings having run in defendants' favor, the panel had no occasion to address an argument that admitted hearsay contravened the Roberts minima for reliability. Morgan discusses "factors of trustworthiness" in its 803(2) analysis, however, and so is instructive in our analysis.
. As we have said, there is no contemporaneous corroboration for the June statement in the usual sense, that is in evidence other than the statement itself. Most courts examining such cases, however, include within the definition of corroborating evidence characteristics of the statement making it more likely reported fact *709than fiction. See, e.g. Nick, 604 F.2d at 1204 (certain statements of an alleged child abuse victim may have "the ring of verity”).
. Several sources have noted the unfortunate irony that increased education about child sexual abuse may increase the number of false alarms raised by children who mistake innocent contacts or overtures for malignant advances. See Note, The Testimony of Child Victims in Sex Abuse Prosecutions: Two Legislative Innovations, 98 Harv.L.Rev. 806, 820 & n. 94 (1985).
. Griffin testified that her daily routine when tending LaTonya included bathing the child. Unlike her account of LaTonya’s state on September 7, Griffin did not testify to having observed anything in the child’s physical condition suggesting recent sexual abuse.
.That the 1982 diagnoses reveal a sexual contact between Gregory and LaTonya is by no means a settled fact, although the conclusion seems indispensable to use of the diagnoses as corroboration for the June remark. Also, this evidence of gonorrhea in 1982, which is stressed in the dissent, loses much of its impact when it is recalled that Gideon Griffin, LaTonya’s uncle who lived across the street, was also diagnosed as having gonorrhea on August 19, 1982.
. Coy v. Iowa, 487 U.S. 1012, 1021, 108 S.Ct. 2798, 2803, 101 L.Ed.2d 857, 867 (1988) makes plain that the harmless error standard of Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967) applies in cases revealing violations of the Confrontation Clause.
. The doubts about LaTonya’s credibility that led to the incompetency ruling are apposite to this as well as the Sixth Amendment portion of our analysis. Although the jury was not aware of the information in the same fashion, we have in mind also LaTonya's testimony at the voir dire hearing that it was fun to make things up, including things about her father, that she would tell things that did not happen, and that her mother and grandmother had told her what to say at the hearing.
.In addition to our conviction that this is not a "harmless error’’ case, we are unwilling to use this analysis where the State does not rely upon it. Despite the North Carolina Court of Appeal’s lead, the State did not argue it to the district court, nor present it in its briefs or argument to us. On the contrary, the State’s representative, during oral argument, declared "I did not argue harmless error," and declined to adopt it on appeal despite several opportunities to do so.