Opinion by Judge THOMAS; Concurrence by Chief Judge KOZINSKI.
OPINION
THOMAS, Circuit Judge:This appeal presents the question of whether a credible showing of “actual innocence” under Schlup v. Delo, 513 U.S. 298, 115 S.Ct. 851, 130 L.Ed.2d 808 (1995), excuses the statute of limitations period established by the Antiterrorism and Effective Death Penalty Act of 1996 (“AED-PA”), 28 U.S.C. § 2241 et seq. We conclude that it does, but that the petitioner failed to present sufficient evidence of actual innocence to permit review of his constitutional claims on the merits. We reverse the judgment of the district court.
I
An Oregon jury convicted Richard Lee of two counts of first degree sex abuse and two counts of first degree sodomy, for which he received a sentence of 170 months in prison. Although Lee’s trial attorney provided appellate counsel with a list of potential trial errors, appellate counsel filed a “Balfour brief’ with the Oregon Court of Appeals, indicating there were no meritorious issues for appeal.1 The Oregon Court of Appeals affirmed his conviction without opinion. Lee did not file a petition for review to the Oregon Supreme Court.
Lee petitioned for state postconviction relief, alleging ineffective assistance of counsel at trial and on appeal.2 The state trial court denied the petition; the Oregon *932Court of Appeals affirmed; and the Oregon Supreme Court denied review. The appellate judgment became final on September 24, 2001.
Lee filed a federal habeas petition on March 11, 2002. The district court dismissed the petition as untimely, but we reversed this determination on appeal and the petition was reinstated.3 Lee filed his first amended petition on July 25, 2005, and filed a memorandum and exhibits almost two years later.
The district court granted Lee’s petition. Lee v. Lampert, 607 F.Supp.2d 1204 (D.Or.2009). The court held that a showing of actual innocence tolls AEDPA’s limitations period and concluded that Lee made the requisite showing. Id. at 1216-22. On the merits, the court held that Lee established his claim for ineffective assistance of counsel. Id. at 1226. The court therefore granted Lee’s petition, vacated his conviction and sentence, and allowed the State of Oregon 120 days to retry or release him. Id.
The State timely appealed, and a motions panel stayed the district court order pending appeal. A panel of this court reversed the district court. Lee v. Lampert, 610 F.3d 1125 (9th Cir.2010). As a matter of first impression, the panel held that there is no actual innocence exception to override AEDPA’s statute of limitations, and dismissed Lee’s petition as time-barred. Id. at 1133-34. We granted rehearing en banc. Lee v. Lampert, 633 F.3d 1176 (9th Cir.2011).
We review de novo the grant of a petition for habeas corpus. Paulino v. Harrison, 542 F.3d 692, 698 (9th Cir.2008).
II
Lee concedes that he did not file his federal habeas petition within AEDPA’s one-year statute of limitations period.4 He argues that a federal court nevertheless may review his first habeas petition upon a showing of “actual innocence” under Schlup. We agree.
We hold that a credible claim of actual innocence constitutes an equitable exception to AEDPA’s limitations period, and a petitioner who makes such a showing may pass through the Schlup gateway and have his otherwise time-barred claims heard on the merits. In recognizing an equitable exception based on a credible showing of actual innocence, we join three of our sister circuits on an issue that has divided the courts of appeal.5 See Souter v. Jones, 395 *933F.3d 577, 602 (6th Cir.2005); Lopez v. Trani, 628 F.3d 1228, 1230-31 (10th Cir.2010); San Martin v. McNeil, 633 F.3d 1257, 1267-68 (11th Cir.2011).6
A
AEDPA provides that “[a] 1-year period of limitation shall apply to an application for a writ of habeas corpus by a person in custody pursuant to the judgment of a State court.” 28 U.S.C. § 2244(d)(1). Section 2244(d)(1) “con-taints] multiple provisions relating to the events that trigger its running.” Holland v. Florida, — U.S. —, 130 S.Ct. 2549, 2561, 177 L.Ed.2d 130 (2010). The triggering events are the dates on which: direct review becomes final, an unlawful state-created impediment to filing is removed, a new constitutional right is made retroactively available, or the factual predicate of the claim(s) presented could have been discovered with “due diligence.” 28 U.S.C. § 2244(d)(l)(A)-(D).
1
As the Supreme Court has instructed us, AEDPA’s statute of limitations is subject to equitable exceptions “in appropriate cases.” Holland, 130 S.Ct. at 2560; Porter v. Ollison, 620 F.3d 952, 959 (9th Cir.2010). Because § 2244(d) is not jurisdictional, it is “subject to a ‘rebuttable presumption’ in favor ‘of equitable tolling.’ ” Holland, 130 S.Ct. at 2560 (quoting Irwin v. Dep’t of Veterans Affairs, 498 U.S. 89, 95-96, 111 S.Ct. 453, 112 L.Ed.2d 435 (1990)).7
That presumption applies with particular force to AEDPA, the Court has explained, for two reasons. First, “equitable principles have traditionally governed the substantive law of habeas corpus,” and federal courts “will not construe a statute to displace courts’ traditional equitable authority absent the clearest command.” Id. (citing Munaf v. Geren, 553 U.S. 674, 693, 128 S.Ct. 2207, 171 L.Ed.2d 1 (2008); Miller v. French, 530 U.S. 327, 340, 120 S.Ct. 2246, 147 L.Ed.2d 326 (2000)) (internal quotation marks omitted). Second, Congress enacted AEDPA when the presumption in favor of equitable tolling was well-established in case law, and therefore Congress was aware that courts would apply the presumption when interpreting § 2244(d). Id. at 2561. Thus, notwithstanding § 2244(d)’s silence as to an equitable exception and its express provision for statutory tolling, the Court has held that neither AEDPA’s text nor its purposes “rebut” the presumption in favor of equitable tolling. Id. at 2561-62.
At the time of AEDPA’s passage, federal courts had equitable discretion to hear *934the merits of procedurally-defaulted habeas claims where the failure to do so would result in a “fundamental miscarriage of justice,” such as the conviction of an actually innocent person. See McCleskey v. Zcmt, 499 U.S. 467, 502, 111 S.Ct. 1454, 113 L.Ed.2d 517 (1991); Schlup, 513 U.S. at 320-321, 115 S.Ct. 851; Souter, 395 F.3d at 598-99.
The actual innocence exception “serves as ‘an additional safeguard against compelling an innocent man to suffer an unconstitutional loss of liberty,’ guaranteeing that the ends of justice will be served in full.” McCleskey, 499 U.S. at 495, 111 S.Ct. 1454 (quoting Stone v. Powell, 428 U.S. 465, 492-93, n. 31, 96 S.Ct. 3037, 49 L.Ed.2d 1067 (1976)). It gained recognition in numerous contexts, Schlup, 513 U.S. at 320, 115 S.Ct. 851,8 including where a procedural default resulted from untimely filing in state postconviction proceedings, see Souter, 395 F.3d at 599 (discussing Coleman v. Thompson, 501 U.S. 722, 750, 111 S.Ct. 2546, 115 L.Ed.2d 640 (1991)). We agree with the Sixth Circuit that “[ajbsent evidence of Congress’s contrary intent, there is no articulable reason for treating habeas claims barred by the federal statute of limitations differently.” Id.
As with equitable tolling based on diligence and extraordinary circumstances, see Holland, 130 S.Ct. at 2562, we conclude that Congress intended for the actual innocence exception to apply to AEDPA’s statute of limitations, see San Martin, 633 F.3d at 1267-68; Lopez, 628 F.3d at 1230-31; Souter, 395 F.3d at 599, 602.9 We presume that Congress knew of the exception when it drafted AEDPA, and “absent the ‘clearest command,’ ” we will not construe the statute to displace that equitable authority. Holland, 130 S.Ct. at 2560-61 (quoting Miller, 530 U.S. at 340, 120 S.Ct. 2246). As the Court warned in Holland:
The importance of the Great Writ, the only writ explicitly protected by the Constitution, along with congressional efforts to harmonize the new statute with prior law, counsels hesitancy before interpreting AEDPA’s statutory silence as indicating a congressional intent to close courthouse doors that a strong equitable claim would ordinarily keep open.
Id. at 2562 (internal citation omitted). It is difficult to imagine a stronger equitable claim for keeping open the courthouse doors than one of actual innocence, “the *935ultimate equity on the prisoner’s side.”10 Withrow v. Williams, 507 U.S. 680, 700, 113 S.Ct. 1745, 123 L.Ed.2d 407 (1993) (O’Connor, J., concurring in part and dissenting in part) (noting that the Supreme Court “continuously has recognized that ... a sufficient showing of actual innocence” is normally enough, “standing alone, to outweigh other concerns and justify adjudication of the prisoner’s constitutional claim”). Indeed, “the individual interest in avoiding injustice is most compelling in the context of actual innocence.” Schlup, 513 U.S. at 324, 115 S.Ct. 851.11
2
An actual innocence exception to the limitations period is consistent with AED-PA’s underlying principles. To the extent the statute aims to eliminate delay in federal habeas review, the Supreme Court’s Holland decision refutes the notion that an actual Innocence exception undermines AEDPA:
We recognize that AEDPA seeks to eliminate delays in the federal habeas review process. But AEDPA seeks to do so without undermining basic habeas corpus principles and while seeking to harmonize the new statute with prior law, under which a petition’s timeliness was always determined under equitable principles. When Congress codified new rules governing this previously judicially managed area of law, it did so without losing sight of the fact that the “writ of habeas corpus plays a vital role in protecting constitutional rights.” It did not seek to end every possible delay at all costs.
130 S.Ct. at 2562 (quoting Slack v. McDaniel, 529 U.S. 473, 483, 120 S.Ct. 1595, 146 L.Ed.2d 542 (2000)) (internal citations omitted). Nor does our holding interfere with AEDPA’s other purposes, such as “curbing] the abuse of the statutory writ of habeas.” H.R. Conf. Rep. No. 104-518, at 111 (1996), reprinted in 1996 U.S.C.C.A.N. 944, 944. An actual innocence exception to the limitations provisions does not foster abuse or delay, but instead recognizes that in extraordinary cases, the societal interests of finality, comity, and conserving judicial resources “must yield to the imperative of correcting a fundamentally unjust incarceration.” Murray v. Carrier, 477 U.S. 478, 495, 106 S.Ct. 2639, 91 L.Ed.2d 397 (1986) (internal quotation marks omitted).
In Schlup, the Supreme Court explained that “the fundamental miscarriage of justice exception seeks to balance the societal interests in finality, comity, and conservation of scarce judicial resources with the individual interest in justice that arises in the extraordinary case.” 513 U.S. at 324, *936115 S.Ct. 851. The exception “is altogether consistent ... with AEDPA’s central concern that the merits of concluded criminal proceedings not be revisited in the absence of a strong showing of actual innocence.” Calderon v. Thompson, 523 U.S. 538, 558, 118 S.Ct. 1489, 140 L.Ed.2d 728 (1998).
3
Finally, the doctrine of constitutional avoidance requires us to construe the statute, if possible, to avoid a serious constitutional question. See Pub. Citizen v. U.S. Dep’t of Justice, 491 U.S. 440, 466, 109 S.Ct. 2558, 105 L.Ed.2d 377 (1989) (“It has long been an axiom of statutory interpretation that where an otherwise acceptable construction of a statute would raise serious constitutional problems, the Court will construe the statute to avoid such problems unless such construction is plainly contrary to the intent of Congress.” (internal quotation marks omitted)).
As our sister circuits have observed, denying federal habeas relief from an actually innocent petitioner would be “constitutionally problematic.” Souter, 395 F.3d at 601-02 (collecting cases); see, e.g., Wyzykowski v. Dep’t of Corr., 226 F.3d 1213, 1218 (11th Cir.2000) (noting that barring a habeas petitioner who can show actual innocence, but who filed after AEDPA’s limitations period, “raises concerns because of the inherent injustice that results from the conviction of an innocent person, and the technological advances that can provide compelling evidence of a person’s innocence” (footnotes omitted)).
In Ferguson v. Palmateer we held that AEDPA’s limitations period “is not a per se violation of the Suspension Clause” and “does not render federal habeas an inadequate or ineffective remedy.” 321 F.3d 820, 823 (9th Cir.2003) (emphasis added). Ferguson did not concern a successful Schlup claim, which presents individual interests of “overriding importance.” Schlup, 513 U.S. at 325, 115 S.Ct. 851. Indeed, in Schlup, the Supreme Court noted:
[Cjoncern about the injustice that results from the conviction of an innocent person has long been at the core of our criminal justice system. That concern is reflected ... in the “fundamental value determination of our society that it is far worse to convict an innocent man than to let a guilty man go free.”
513 U.S. at 325, 115 S.Ct. 851 (quoting In re Winship, 397 U.S. 358, 372, 90 S.Ct. 1068, 25 L.Ed.2d 368 (1970) (Harlan, J., concurring)). Given our holding, we need not revisit Ferguson’s ruling on the Suspension Clause or decide what other “grave constitutional concerns” would be presented by the denial of federal habeas relief to an innocent prisoner. See Souter, 395 F.3d at 602.
4
In short, Congress did not remove the equitable power of federal courts in habeas proceedings. Nothing in the text or structure of AEDPA indicates a contrary conclusion. Recognition of an actual innocence exception to AEDPA’s statute of limitations is entirely consistent with AEDPA’s underlying principles. The doctrine of constitutional avoidance further buttresses our conclusion.
B
Neither of the State’s arguments to the contrary is persuasive. First, the State points out that AEDPA’s limitations provisions say nothing of actual innocence, while other parts of the Act expressly refer to it. True, AEDPA contains a more exacting actual innocence exception for successive petitions and evidentiary hearings, requiring that the factual predicate of a claim could not have been discovered *937earlier and a showing “by clear and convincing evidence” that “but for constitutional error, no reasonable factfinder would have found the applicant guilty of the underlying offense.” 28 U.S.C. §§ 2244(b)(2)(B); 2254(e)(2).12 That Congress adopted a more stringent actual innocence exception for successive petitions and evidentiary hearings does not evince a desire to exclude the equitable exception in § 2244(d) for a first habeas petition. Souter, 395 F.3d at 598-99. Rather, the absence of an actual innocence trigger to § 2244(d)(1) suggests that Congress intended not to alter the existing jurisprudential framework that allowed for a showing of actual innocence to overcome a procedural default. Id. at 599.
The State also suggests that recognizing an equitable exception on the basis of actual innocence would conflict with our case law, which sets a “very high” threshold for triggering equitable exceptions, “lest the exceptions swallow the rule.” WaldronRamsey v. Pacholke, 556 F.3d 1008, 1011 (9th Cir.2009) (internal quotation marks omitted). Yet, today’s holding is entirely “consistent with our sparing application of the doctrine of equitable tolling.” Id. As we discuss in more detail later, and as the Supreme Court has stressed, habeas corpus petitions advancing a credible claim of actual innocence are “extremely rare.” Schlup, 513 U.S. at 321, 115 S.Ct. 851; see also House, 547 U.S. at 538, 126 S.Ct. 2064; Souter, 395 F.3d at 600. The Schlup exception involves a “narrow class of cases ... implicating a fundamental miscarriage of justice” because “a constitutional violation has probably resulted in the conviction of one who is actually innocent.” Schlup, 513 U.S. at 314-15, 321, 115 S.Ct. 851 (internal quotation marks omitted). Given that the exception is confined to these extraordinary cases, there is little danger of it swallowing the rule. The exacting requirements we have imposed for the application of equitable tolling remain intact and are not in conflict with an actual innocence exception.
C
Thus, we join the Sixth, Tenth, and Eleventh Circuits in holding that where an otherwise time-barred habeas petitioner demonstrates that it is more likely than not that no reasonable juror would have found him guilty beyond a reasonable doubt, the petitioner may pass through the Schlup gateway and have his constitutional claims heard on the merits.
Ill
We now consider whether Lee presents sufficient new evidence of actual innocence to excuse his untimely filing and permit review of his constitutional claims on the merits.
A
In order to present otherwise time-barred claims to a federal habeas court under Schlup, a petitioner must produce sufficient proof of his actual innocence to bring him “within the ‘narrow class of cases ... implicating a fundamental miscarriage of justice.’ ” 513 U.S. at 314-15, 115 S.Ct. 851 (quoting McCleskey, 499 U.S. at 494, 111 S.Ct. 1454). The evidence of innocence must be “so strong that a court cannot have confidence in the outcome of the trial unless the court is also satisfied that the trial was free of non-*938harmless constitutional error.” Id. at 316, 115 S.Ct. 851.
To pass through the Schlwp gateway, a “petitioner must show that it is more likely than not that no reasonable juror would have convicted him in the light of the new evidence.” Id. at 327, 115 S.Ct. 851; House, 547 U.S. at 538, 126 S.Ct. 2064. This exacting standard “permits review only in the ‘extraordinary’ case,” but it “does not require absolute certainty about the petitioner’s guilt or innocence.” House, 547 U.S. at 538, 126 S.Ct. 2064 (quoting Schlup, 513 U.S. at 327, 115 S.Ct. 851). As we have previously said, “where post-conviction evidence easts doubt on the conviction by undercutting the reliability of the proof of guilt, but not by affirmatively proving innocence, that can be enough to pass through the Schlup gateway to allow consideration of otherwise barred claims.” Sistrunk v. Armenakis, 292 F.3d 669, 673 (9th Cir.2002) (en banc) (citing Carriger v. Stewart, 132 F.3d 463, 478-79 (9th Cir.1997) (en banc)).13
Schlup requires a petitioner “to support his allegations of constitutional error with new reliable evidence — whether it be exculpatory scientific evidence, trustworthy eyewitness accounts, or critical physical evidence — that was not presented at trial.” Schlup, 513 U.S. at 324, 115 S.Ct. 851. The habeas court then “consider[s] all the evidence, old and new, incriminating and exculpatory,” admissible at trial or not. House, 547 U.S. at 538, 126 S.Ct. 2064 (internal quotation marks omitted); Carriger, 132 F.3d at 477-78. On this complete record, the court makes a “ ‘probabilistic determination about what reasonable, properly instructed jurors would do.’” House, 547 U.S. at 538, 126 S.Ct. 2064 (quoting Schlup, 513 U.S. at 329, 115 S.Ct. 851).
B
With these principles in mind, we turn to the claims in this case. In order to assess whether Lee has satisfied the stringent requirements of Schlup, we must first examine the underlying facts in some detail.
1
Daniel Hendricks and his wife hired Cheryll Lee to babysit their three children. Cheryll watched the children in her apartment, where three men spent time. Cheryll’s boyfriend, Robert Nachand, lived with Cheryll. Her ex-husband, Larry Lee, and his brother, petitioner Richard Lee,14 occasionally visited. Robert and Larry *939both had been previously indicted for molesting children.
On several occasions, while Cheryll left the children at the apartment, Robert and Richard sexually abused four-year old Matthew Hendricks.
Matthew began wetting his bed, throwing temper tantrums, and crying that he did not want to go to Cheryll’s. Matthew’s parents asked Cheryll to babysit the children in their home rather than at her apartment. Matthew told his mother that on several occasions while Cheryll left the Hendricks children at her apartment, Robert had fondled and fellated Matthew, and vice versa. His mother called the police and fired Cheryll.
Detective Chris Carter questioned Matthew with the aid of anatomically correct drawings. Matthew recounted four to six instances of abuse by Robert. According to Carter’s report, which describes Matthew’s statements, “Matthew gave no further disclosure of any other sexual abuse that occurred to him or his sisters.”
Robert eventually entered into a plea bargain, pleading guilty to sexual abuse and accepting a 28-month prison sentence. In exchange, the prosecutor dropped a sodomy charge.
Several months after Matthew reported Robert’s abuse, Matthew’s father Daniel saw Richard walking in the neighborhood and asked Matthew whether Richard also had molested him. Matthew affirmed that he had. Daniel brought Matthew to the police station for an interview with Detective John Fowler. Fowler used three anatomically correct dolls — designated “Robert,” “Richard,” and “Matthew,” respectively — so that Matthew could demonstrate the sex acts that had occurred between him and each man. Fowler photographed the dolls, but did not otherwise record the interview.
Detective Fowler later questioned Cheryll about the allegations against Richard. After an initial denial, Cheryll told Fowler she left Richard with the children on at least three occasions. Following one such occasion, Cheryll recounted, Matthew told her that Richard had touched his “pee-pee.” She confronted Richard about Matthew’s statement; he said the touching must have occurred accidentally while he wrestled with Matthew. Richard denied ever caring for the Hendricks children or talking with Cheryll about Matthew’s statement.
2
Richard was indicted on two counts of first degree sexual abuse and two counts of first degree sodomy. Before trial, the defense moved to admit evidence of abuse by Robert (who, by that time, had pled guilty) under Oregon Rule of Evidence 412, which prohibits the admission of a victim’s past sexual behavior unless such evidence falls into one of the Rule’s narrow exceptions. The defense argued that Robert’s sexual acts were relevant to Richard’s defense of mistaken identity, and that it would violate Richard’s confrontation rights if he were not able to cross-examine witnesses about Robert. The trial court denied the motion because Matthew was well-acquainted with the two men, Chéryll’s statement placed Richard in Matthew’s presence at her apartment, and Matthew’s statements to Cheryll tended to show that the conduct occurred. Thus, the court ruled, Lee made an insufficient predicate showing of confusion to warrant breach of Rule 412’s rape-shield protections.
3
At trial, the prosecution called five witnesses — Matthew Hendricks, Cheryll Lee, Detective John Fowler, Katherine Walling, and Detective Jonathan Strong — and ad*940mitted as exhibits drawings of human bodies and photos of the Fowler interview. The defense called Daniel Hendricks as a witness, asked Richard to show the jury tattoos on his arm and shoulder, and admitted as an exhibit Robert’s sentencing judgment.
On direct examination, Matthew initially stumbled in identifying petitioner, calling him “Robert — I mean, Richard.” He testified that Richard had been at Cheryll’s house, and, with the aid of human drawings, Matthew revealed how Richard had touched him. He circled both the “pee-pee” and “bottom” on the drawings and stated that, while Cheryll was away, Richard had touched him in both places with his hand. The touching occurred when Matthew was naked and Richard’s shirt was removed. Matthew recalled telling his father, Cheryll, and Detective Fowler about what Richard did. He recognized photos taken during the Fowler interview and identified himself and “Richard” in a photo, but would not say what was happening. Matthew said that when Richard touched him, he could see Richard’s “stomach.”
The defense began cross-examination with questions about Cheryll’s “friend named Robert.” Matthew remembered saying “Robert — I mean, Richard” on direct and agreed that he was getting the two “mixed up.” At that point, the prosecution objected, and with the jury and Matthew removed from the courtroom, the trial judge ruled that Rule 412 did not prevent the defense from using Robert’s name to distinguish what Richard did or did not do. When Matthew returned, the defense asked, “you said — you said that’s Robert?” (A: “[yjeah”) and whether Matthew “notieefdj anything about his body?” or “his arms or his stomach” (A: “[njo”).
On redirect, the prosecution asked, “Matthew, is this Robert?” (A: “[njo”); “Who is this?” (A: “Richard”); and “is this the person that did those things that you’ve drawn for us today?” (A: “[yjeah”). On recross, the defense asked “you said it was Robert; right? You said it was Robert?” (A: “No. Richard.”), and whether Matthew had talked with his parents while out of the courtroom (A: “[yjeah”).
Cheryll Lee confirmed that Richard came to her apartment a few times a month, before moving in with her. She left Richard and Robert with the children at least three times; Robert should have been sleeping at these times because he worked the night shift. Two days after one of these occasions, Matthew told Cheryll that Richard “touched” him on his “pee-pee.” Cheryll told Matthew to go play, and later that day, she confronted Richard about Matthew’s statement. Richard explained that it “must have happened the other day when I was wrestling with him when [she] went to the store.” They did not discuss it further; Cheryll believed him.
On cross-examination, Cheryll confirmed that each time she left Richard with the kids, Robert was there, too. The defense also asked whether Matthew “said, ‘Richard did it’?” (A: “[yjes”) and if Cheryll “[knew] if he meant Richard or Robert?” (A: “he stated ‘Richard’ ”). Cheryll confirmed that Matthew said only that he was touched, pointed downward, and when asked, “Where were you touched?”, said, “Right here. My pee-pee.”
Detective John Fowler’s testimony began with a review of his investigation of Matthew’s case. He briefly spoke with Matthew in a “soft interview room,” where Matthew pointed to his groin and said Richard had touched him “where he wasn’t supposed to.” Based on this disclosure, Fowler asked Matthew and his father to come with him to the Children’s Services Division, where Fowler could use anatomi*941cally correct dolls. Once they relocated, Fowler interviewed Matthew, with his father present.15 Fowler documented the interview in writing and with photographs.
As direct examination continued, Fowler described using “an adult and a child doll,” which Matthew used to show Fowler how Richard had touched him. Fowler recounted Matthew’s demonstration with the dolls and his statements — specifically, Matthew: took the adult male’s hand and placed it on the child doll’s penis and buttocks; put the adult doll’s mouth to the penis of the child doll; said that Richard had touched Matthew’s bottom with his mouth; said that Richard had put his penis in Matthew’s mouth; said that Richard had rubbed his penis on the “outside” of Matthew’s bottom; said that Richard had rubbed his bottom with a “cup” and a “toy”; said that when sexual touching happened “water” came out of Richard’s “pee-pee”; and said that when Richard touched his “pee-pee” it made Matthew “feel like he had to go to the bathroom” and made him cry because he felt “sad.”
On cross-examination, the defense elicited that Fowler had spoken with Matthew’s father a few times before the doll interview. Matthew’s father wished to report disclosures his son made to him. Next, the defense elicited that Fowler had not recorded the interview by audio or video because it was “discouraged” in Linn County.
The defense then asked about how Fowler phrased questions during the doll interview. The defense inquired whether Fowler asked Matthew if the “water” described by Matthew came out of “Richard’s pee-pee or Robert’s pee-pee?” (A: “[correct”). The defense again asked “Robert’s pee-pee?” (A: “Richard’s or Robert’s”). At that point, the prosecution objected on Rule 412 grounds. Outside the presence of the jury, the defense explained that questioning Fowler about his use of three dolls (Richard, Robert, and Matthew) during the interview required clarification. The trial court agreed and permitted questions about dolls depicting Robert as well as Richard. When questioning resumed, the defense elicited that there was a third Robert doll and that Fowler sometimes asked Matthew questions about “Richard or Robert” or “Richard and Robert.” Fowler said that Matthew did not demonstrate the same “specific sexual acts” with both adult dolls, but affirmed that some of Matthew’s answers to questions referred to both men.
On redirect, Fowler stated that Matthew identified the dolls by their physical appearance (e.g. Richard’s doll was “darker complected”) and attributed certain actions to Richard and others to Robert. Matthew consistently indicated the things Richard had done.
On reeross, Fowler agreed that Matthew had not said anything about oral sex in the “soft interview room.”
On further redirect and recross, the prosecution elicited that Robert had been prosecuted and imprisoned for molesting Matthew, and that Matthew had made an “additional disclosure” about Richard. According to Fowler, Matthew was not confused about what Robert or Richard had done to him; “[h]e made specific disclosures about what Richard had done that he did not talk about that had happened with Robert.” The defense could not inquire into Matthew’s “very specific disclosures” about Robert because Fowler lacked personal knowledge. The prosecution elicited that Matthew made allegations specific to Richard, including contact between Rich*942ard’s mouth and Matthew’s “bottom” and contact with the cup and toy. Fowler never found the toy, nor did he figure out to what toy Matthew referred.
Katherine Walling testified that she also took her children to Cheryll’s for babysitting, and saw Matthew and Richard there.
Detective Jonathan Strong testified to his interview with Richard. In that interview, Richard said that he visited Cheryll during the period of the alleged abuse— while Robert lived there — and later stayed with her. He knew where Matthew’s parents lived and that Cheryll babysat the Hendricks children there, but could not recall Cheryll bringing children to her apartment. Richard also denied that he: could recognize Matthew; had cared for the Hendricks children three times in Cheryll’s absence; spoke to Cheryll about Matthew’s disclosure and his statement that they had wrestled; and had sexual contact with Matthew.
On direct examination of Matthew’s father, Daniel Hendricks, the defense elicited that Daniel had been convicted of several past crimes of dishonesty, and that he contacted the police regarding Matthew’s allegations. The defense then asked about Matthew’s “very specific descriptions” of Robert’s acts. Daniel stated that in the case against Robert, Matthew talked about oral sex being performed both ways, and that the case against Robert had been going for at least three months before Daniel called the police about Richard. Daniel also recounted his son’s “very specific” allegations about Richard, including that Richard had sodomized or “stuck it up inside of’ Matthew. The defense elicited that Matthew had never said anything about Richard prior to his June or July report, and that Daniel never sought a medical investigation of Matthew.
On cross-examination, Daniel testified to Matthew’s changes in behavior prior to his report about Robert. The prosecution then asked about Richard. In Matthew’s first disclosure, which took place outside their home, Daniel saw Richard and asked Matthew if Richard was staying at Cheryll’s. Matthew answered “yes,” and Daniel “asked him if there had any — ever been any kind of sexual contact,” without specifying, and Matthew affirmed there had been. Daniel asked Matthew to “be honest with Dad and tell [him] everything.” Matthew told his father that Robert and Richard both had sex with him together, that Matthew had performed oral sex on Richard, and that Richard had “touched” Matthew’s genitals and Matthew’s “bottom had been hurt.” Daniel tested Matthew’s perceptions of the two men, asking Matthew “[i]s this Richard or was this Robert?” and adding that “Daddy really needs to know, because we don’t want to cause any undue problems.” Matthew answered, “Well, yes, Daddy. This was Richard.” The prosecution also elicited that Robert went to prison.
On redirect, defense counsel pressed Daniel on whether he — not Matthew— brought up the subject of sexual contact with Richard (A: “[y]es”) and whether Matthew volunteered any information pri- or (A: “[n]o”).
On recross, Daniel made clear that Matthew had alleged oral sex with Richard two times; sodomy two times; and a separate occasion where Robert and Richard had sex with Matthew simultaneously.
During closing argument, the prosecution rebutted Lee’s defense that Matthew could not have been molested by two people or that he was confused. The prosecution recounted the following evidence: Matthew’s initial disclosure about Richard to Cheryll; Cheryll’s testimony that both Robert and Richard frequented her apartment; the investigation of Robert; Matthew’s disclosure about Richard to his fa*943ther; Detective Strong’s testimony about Richard’s denials of knowing or watching Matthew at Cheryll’s; the pictures Matthew drew on the witness stand; and the photographs of the dolls.
The defense reminded the jury of Robert’s guilty plea and stressed that in March 1994, Matthew gave “extensive descriptions” of Robert’s abuse without so much as mentioning Richard. Counsel portrayed the disclosure about Richard as originating with Matthew’s father, reminding the jury of Daniel’s past convictions for crimes of dishonesty and questioning how Daniel’s testimony could be true without medical or physical verification. Counsel asserted that “children are impressionable,” that “reality is often defined by their parents and by adults.” The defense then recounted ways Matthew exhibited confusion or gave vague testimony, and the differences between Matthew’s testimony (e.g., that Richard had his pants on) and Fowler’s (e.g., where both dolls are naked). He noted that Matthew did not remember Richard’s tattoos and testified “with his fingers crossed.” Finally, he reminded the jury that the touching must have been for purposes of sexual arousal or gratification — as opposed to wrestling or horseplay.
The jury found Lee guilty on all counts.
C
Lee contends that three new items of evidence justify his Schlwp claim: (1) an expert opinion concerning the reliability of some of the trial testimony, (2) a police report about Robert Nachand, and (3) other evidence about Cheryll and Larry Lee. He urges that in light of the new evidence, it is more likely than not that no reasonable juror would have convicted him of sexual abuse and sodomy. We disagree.
1
Dr. Maggie Bruck, a Professor in the Division of Child and Adolescent Psychiatry at Johns Hopkins University School of Medicine, offered an expert opinion on the reliability of statements by key witnesses at Lee’s trial. Evidence “undercutting the reliability of the proof of guilt ... can be enough to pass through the Schlup gateway.” Sistrunk, 292 F.3d at 673.
Lee maintains that three aspects of Brack’s report establish areas of unreliability not heard by the jury. First, Bruck emphasizes the relative reliability of a child’s initial report and posits that in Matthew’s “first spontaneous[ ]” disclosure about Robert in February 1994, Matthew detailed abuse by Robert without mentioning Richard. However, months before Matthew’s disclosures about Robert to his mother or his interviews about Robert or Richard, Matthew tried to report Richard’s molestation to his baby-sitter Cheryll. According to Bruck, a child’s statement is “untainted” if “spontaneous, unprompted, and made in the absence of any previous suggestive elements.” Matthew’s early, spontaneous disclosure about Lee fits that description, and Brack’s reasons for discrediting it ring unpersuasive.16
Second, Lee points to Brack’s testimony about the importance of video recording interviews with child victims. Detective Fowler’s failure to video record his interview with Matthew was presented to the jury, however. More importantly, the lack of a video tape indicates a deficiency in the investigation, but it does not affirmatively *944establish suggestive interview tactics on Fowler’s part.
Finally, Lee relies on Bruck’s criticism of the use of anatomical dolls with children who have experienced prior sexual abuse. According to Brack, the use of dolls can elicit “highly unreliable information” and is inherently suggestive. Lee’s jury lacked this context. Detective Fowler referred to dolls as “tools,” and testified to his training and experience interviewing child victims of sexual abuse. Yet, while Bruck’s testimony somewhat undermines the reliability of Matthew’s statements to Detective Fowler, the jury heard evidence of at least two disclosures by Matthew implicating Richard before Fowler’s interview — namely, Matthew’s statements to Cheryll about Richard touching his groin and to his father about Richard’s touching and sodomy.
Brack’s conclusions hinge on her speculation from the trial record. For example, she suggests that Matthew’s father may have held prior beliefs about Richard, coerced Matthew into confirming those beliefs, and “tainted” Matthew’s testimony before the police interview. “If this is correct,” Brack wrote, “then Matthew’s statements to Detective Fowler were unreliable.” (Emphasis added.) Similarly, Brack notes that Matthew’s father’s “nonverbal cues may have prompted Matthew to say certain things” and “if the father had suggestively interviewed Matthew ... then the dolls would simply give Matthew an opportunity to act out these previous unreliable statements.” (Emphasis added.) Without a videotape, she explains, Brack could not evaluate the use of the dolls, questions asked, or answers given, and as a result, the reliability of Matthew’s reported statements is “unknowable.” Yet, she opines that regardless of interview technique, “Matthew’s statements were tainted by the time the interview began.”
Given these limitations of Bruck’s analysis, we cannot say “it is more likely than not that no reasonable juror would have convicted [Lee] in the light of the new evidence.” Schlup, 513 U.S. at 327, 115 S.Ct. 851. Presented with her expert testimony, a reasonable juror may well have rejected Bruck’s conclusions or convicted Lee on the basis of Matthew’s statements to Cheryll and his father.
2
According to Lee, the police report about Robert Nachand’s abuse would have revealed the following “key” facts: that Matthew was subjected to sexual abuse and sodomy by Robert in fall 1993; that Matthew described Robert’s acts in detail, in contrast to his “vague and ambivalent” trial testimony; and that Matthew initially reported no abuse other than by Robert.
Much of this evidence was presented to Lee’s trial jury. Matthew was questioned repeatedly about transposing the names “Richard” and “Robert”; Cheryll testified that both Robert and Richard were at home when she left the children there; Detective Fowler acknowledged that he questioned Matthew about both men, and he confirmed the charges against Robert; Matthew’s father testified to his son’s complaints about Robert, including two kinds of oral sex; and the defense introduced a copy of Robert’s judgment, which showed his conviction for abusing Matthew and a dismissed sodomy count. At closing argument, moreover, the defense referred to Robert’s conviction; Matthew’s confusion about the two men; that Matthew had not disclosed Richard’s abuse in his first report; and that no medical examination had been performed. The prosecution referred extensively to “Robert” or “Nachand.”
*945Even assuming, arguendo, that the police report constituted “new reliable evidence ... that was not presented at trial,” Schlup, 513 U.S. at 324, 115 S.Ct. 851; Sistrunk, 292 F.3d at 673 n. 4, we must assess its likely impact on reasonable jurors in light of the complete record, House, 547 U.S. at 538, 126 S.Ct. 2064. Given all that Lee’s jury heard about Robert at trial, the introduction of the police report would not make it “more likely than not that no reasonable juror viewing the record as a whole would lack reasonable doubt” of his guilt. Id. at 554, 126 S.Ct. 2064.
3
Lee’s remaining evidence only thinly supports his innocence claim. Cheryll Lee’s initial denial that Richard Lee was present around children might somewhat discredit her testimony about Matthew’s disclosure, but Lee’s jury also heard of that disclosure from Matthew and Detective Fowler. Information about Larry Lee, on the other hand, has little relevance to this case. Matthew never alleged abuse by Larry, nor has petitioner seriously suggested that his brother was a culprit.
4
Under Schlup, we must “assess how reasonable jurors would react to the overall, newly supplemented record,” including all the evidence Lee now proffers. House, 547 U.S. at 538, 126 S.Ct. 2064. Having done so, we cannot conclude that “it is more likely than not that no reasonable juror would have convicted him.” Schlup, 513 U.S. at 327, 115 S.Ct. 851.
IV
In sum, we hold that a petitioner is not barred by the AEDPA statute of limitations from filing an otherwise untimely habeas petition if the petitioner makes a credible showing of “actual innocence” under Schlup v. Delo. This construction continues the traditional equitable rule that Congress did not disturb in passing AED-PA, is consistent with AEDPA’s underlying philosophy, and avoids serious constitutional problems inherent in a contrary statutory interpretation.
However, in this case, because the petitioner did not present sufficient evidence of actual innocence to permit him to pass through the Schlup gateway and to argue the merits of his time-barred claims, we reverse the judgment of the district court and remand with instructions to dismiss Lee’s petition as untimely. We need not, and do not, reach any other question urged by the parties.
REVERSED AND REMANDED WITH INSTRUCTIONS.
. A Balfour brief is the Oregon variant of the briefing system established by the Supreme Court in Anders v. California, 386 U.S. 738, 87 S.Ct. 1396, 18 L.Ed.2d 493 (1967). See Reese v. Baldwin, 282 F.3d 1184, 1187 (9th Cir.2002), rev’d on other grounds, 541 U.S. 27, 124 S.Ct. 1347, 158 L.Ed.2d 64 (2004). In State v. Balfour, 311 Or. 434, 814 P.2d 1069 (1991), the Oregon Supreme Court established a procedure by which counsel can notify the court that he or she believes the appeal lacks merit without withdrawing from representation of the appellant. Id. at 1078-79. Unlike the Anders procedure, the court is not required to review the record independently for error. Id. at 1080.
. Lee alleged that his trial counsel failed to adequately investigate, to object to certain *932evidence, and io call an expert on the reliability of child testimony. He further alleged that trial counsel inadequately cross-examined witnesses, called a prejudicial witness, failed to move for a mistrial or judgment of acquittal, and did not adequately object to the denial of Lee’s motion under Oregon Rule of Evidence 412. He alleged that his appellate counsel failed to raise both preserved and plain error on appeal.
. A magistrate judge held that the petition was not filed within AEDPA's one-year limitations period, and the district court adopted the magistrate judge’s recommendation to dismiss. We reversed and remanded, accepting Oregon’s concession that the district court erred by dismissing the petition as untimely on its face, without opportunity to establish circumstances that might toll the limitations period. Lee v. Lampert, 92 Fed.Appx. 532, 532-33 (9th Cir.2004).
. Lee filed for state postconviction relief on February 23, 1998, more than 16 months after direct review of his conviction became final on September 30, 1996. Then, after his state postconviction relief proceedings ended, during the pendency of which the federal limitations period is tolled, see 28 U.S.C. § 2244(d)(2), another six months passed before Lee filed for federal habeas relief. All told, Lee filed for federal habeas relief well after the one-year statute of limitations had expired.
. We note that, in many cases, the phrase "equitable tolling” is used in describing the use of equitable power to allow the untimely filing of a habeas petition in an actual inno*933cence case. The more accurate characterization is "equitable exception,” because equitable tolling involves different theoretical underpinnings.
. The First, Fifth, and Seventh Circuits do not recognize an actual innocence exception under any circumstances. See David v. Hall, 318 F.3d 343, 347 (1st Cir.2003); Cousin v. Lensing, 310 F.3d 843, 849 (5th Cir.2002); Escamilla v. Jungwirth, 426 F.3d 868, 871-72 (7th Cir.2005). But see Riva v. Ficco, 615 F.3d 35, 44 n. 4 (1st Cir.2010) (characterizing David as merely "expressing skepticism” about whether an actual innocence claim can excuse AEDPA's limitations period, and remanding to the district court to consider the issue in the first instance); Gildon v. Bowen, 384 F.3d 883, 887 (7th Cir.2004) (adopting the Eighth Circuit’s approach in Flanders v. Graves, 299 F.3d 974, 976-78 (8th Cir.2002)). The Eighth Circuit has left open the possibility that actual innocence could bear on a claim for equitable tolling. See Flanders, 299 F.3d at 976-78.
. Historically, of course, there were no statutes of limitations applicable to federal writs of habeas corpus. As Justice Jackson explained "habeas corpus provides a remedy for jurisdictional and constitutional errors at the trial without limit of time.” United States v. Smith, 331 U.S. 469, 475, 67 S.Ct. 1330, 91 L.Ed. 1610 (1947).
. See Kuhlmann v. Wilson, 477 U.S. 436, 106 S.Ct. 2616, 91 L.Ed.2d 364 (1986) (filing of "successive” petitions raising grounds identical to those raised and rejected on the merits in prior federal petitions); Murray v. Carrier, 477 U.S. 478, 106 S.Ct. 2639, 91 L.Ed.2d 397 (1986) (failure to raise issues on direct appeal); Smith v. Murray, 477 U.S. 527, 106 S.Ct. 2661, 91 L.Ed.2d 434 (1986) (same); McCleskey, 499 U.S. 467, 111 S.Ct. 1454 (filing of "abusive” petitions raising grounds available but not relied upon in an earlier federal petition); Coleman v. Thompson, 501 U.S. 722, 111 S.Ct. 2546, 115 L.Ed.2d 640 (1991) (failure to comply with filing deadlines in state postconviction proceedings). As the Court explained in Schlup, these cases illustrate the "interplay between statutory language and judicially managed equitable considerations in the development of habeas corpus jurisprudence.” 513 U.S. at 319 n. 35, 115 S.Ct. 851.
. Because this case does not present the question, we need not — and do not — decide what diligence, if any, a petitioner must demonstrate in order to qualify for the actual innocence exception recognized in this opinion. Compare, e.g., Lopez, 628 F.3d at 1231 (requiring no showing of diligence by a petitioner seeking equitable tolling on actual innocence grounds), with Flanders, 299 F.3d at 978 (requiring a petitioner seeking equitable tolling on actual innocence grounds to show either that a state-created barrier prevented his timely discovery of relevant facts or that a "reasonably diligent petitioner” could not have discovered such facts in time to file within the limitations period).
. It would seem odd indeed that, under Holland, a petitioner could receive the benefit of equitable tolling for attorney error and yet be denied habeas review upon making a credible showing of actual innocence. 130 S.Ct. at 2564; see also Spitsyn v. Moore, 345 F.3d 796, 801 (9th Cir.2003). Notably, we already apply equitable principles to AEDPA’s statute of limitations in a variety of contexts. See, e.g., Miles v. Prunty, 187 F.3d 1104, 1107 (9th Cir.1999) (prison officials failed to timely prepare a check for the court’s filing fee); Corjasso v. Ayers, 278 F.3d 874, 878 (9th Cir.2002) (district court improperly dismissed, and then lost, petitioner’s § 2254 motion); Laws v. Lamarque, 351 F.3d 919, 923 (9th Cir.2003) (petitioner’s mental incompetence prevented him from timely filing). Surely if equitable tolling is appropriate in these contexts, untimely filings by those whose claims invoke the “ultimate equity,’’ actual innocence, should be excused.
. We also bear in mind that this case concerns a first federal habeas petition, the dismissal of which “is a particularly serious matter, for that dismissal denies the petitioner the protections of the Great Writ entirely, risking injury to an important interest in human liberty.” Lonchar v. Thomas, 517 U.S. 314, 324, 116 S.Ct. 1293, 134 L.Ed.2d 440 (1996).
. The "clear and convincing” standard mirrors the test from Sawyer v. Whitley, 505 U.S. 333, 336, 112 S.Ct. 2514, 120 L.Ed.2d 269 (1992), which adopted a stricter standard than Schlup’s "more likely than not” test. See House v. Bell, 547 U.S. 518, 539, 126 S.Ct. 2064, 165 L.Ed.2d 1 (2006); see also Schlup, 513 U.S. at 326-27, 115 S.Ct. 851.
. Schlup incorporates the standard of proof of Murray v. Carrier, 477 U.S. 478, 106 S.Ct. 2639, 91 L.Ed.2d 397 (1986), and therefore involves a different inquiry than mere sufficiency of the evidence. See Schlup, 513 U.S. at 324, 330, 115 S.Ct. 851. As the Supreme Court explained in Schlup, the Carrier standard differs from the standard of Jackson v. Virginia, 443 U.S. 307, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979), applicable to claims of insufficient evidence, in at least two respects. Schlup, 513 U.S. at 330, 115 S.Ct. 851. First, while a Jackson analysis generally precludes assessment of witness credibility, a habeas court applying Carrier "may have to make some credibility assessments." Id.; House, 547 U.S. at 538-39, 126 S.Ct. 2064. Second, Jackson asks whether a trier of fact "could," i.e. has the power to, reach its conclusion, while Carrier asks whether a trier of fact "would," i.e. is likely to, reach a particular conclusion. Schlup, 513 U.S. at 330, 115 S.Ct. 851; House, 547 U.S. at 538, 126 S.Ct. 2064 ("[T]he [Schlup] inquiry requires the federal court to assess how reasonable jurors would react to the overall, newly supplemented record.”). Hence, "though under Jackson the mere existence of sufficient evidence to convict would be determinative of [a] petitioner's claim, that is not true under Carrier." Schlup, 513 U.S. at 330, 115 S.Ct. 851.
. To maintain consistency with trial testimony, our recitation of the facts often will refer to Richard Lee and others by their first names.
. According to Fowler, Matthew’s father encouraged Matthew to be truthful, but otherwise did not participate in the interview.
. Bruck doubts that Cheryll could remember her conversation with Matthew and notes that Cheryll’s testimony changed when police confronted her. But the jury also heard testimony on Matthew's disclosure from Matthew and Detective Fowler, which corroborated Cheryll's version of events.