Carl Hubbard v. Randee Rewerts

RECOMMENDED FOR PUBLICATION
Pursuant to Sixth Circuit I.O.P. 32.1(b)
File Name: 24a0088p.06

UNITED STATES COURT OF APPEALS
FOR THE SIXTH CIRCUIT

┐
CARL HUBBARD,
│
Petitioner-Appellant, │
> No. 21-2968
│
v. │
│
RANDEE REWERTS, Warden, │
Respondent-Appellee. │
┘

Appeal from the United States District Court for the Eastern District of Michigan at Detroit.
No. 2:13-cv-14540—David M. Lawson, District Judge.

Argued: July 19, 2023

Decided and Filed: April 16, 2024

Before: BATCHELDER, COLE, and NALBANDIAN, Circuit Judges.
_________________

COUNSEL

ARGUED: Alexander Kazam, KING & SPALDING LLP, Washington, D.C., for Appellant.
Marissa A. Wiesen, OFFICE OF THE MICHIGAN ATTORNEY GENERAL, Lansing,
Michigan, for Appellee. ON BRIEF: Alexander Kazam, KING & SPALDING LLP,
Washington, D.C., for Appellant. Marissa A. Wiesen, OFFICE OF THE MICHIGAN
ATTORNEY GENERAL, Lansing, Michigan, for Appellee.
BATCHELDER, J., delivered the opinion of the court in which NALBANDIAN, J.,
joined. COLE, J. (pp. 21–44), delivered a separate dissenting opinion.
_________________

OPINION
_________________

ALICE M. BATCHELDER, Circuit Judge. Petitioner Carl Hubbard was convicted of
first-degree murder in Michigan state court on September 2, 1992. Over two decades later (and
 No. 21-2968 Hubbard v. Rewerts Page 2

long after 28 U.S.C. § 2244’s one-year limitation period had expired), Hubbard filed a petition
for a writ of habeas corpus in the United States District Court for the Eastern District of
Michigan. The district court dismissed the petition as untimely. Hubbard now appeals, arguing
that he is entitled to an equitable exception to the Antiterrorism and Effective Death Penalty Act
of 1996’s (AEDPA) time bar based on a credible showing of actual innocence. See McQuiggin
v. Perkins, 569 U.S. 383, 386 (2013). While Hubbard presents new evidence that impeaches the
State’s case against him, he fails to present evidence affirmatively demonstrating his actual
innocence; he cannot prove that he did not, in fact, commit murder. Accordingly, AEDPA does
not permit him to file an untimely habeas petition. We affirm.

I.

On the night of January 17, 1992, Rodnell Penn was shot in a crime-ridden neighborhood
in Detroit. He was found lying near the curb of Gray Street, just north of the corner of Gray and
Mack (colloquially, the “Gray and Mack” neighborhood), about two hundred yards away from a
party store. Officer Craig Turner arrived at the scene as backup. As EMS loaded Penn’s body
into an ambulance, petitioner Carl Hubbard, who had been walking by the scene, walked up to
Officer Turner and asked what happened. Officer Turner stated that a homicide had occurred.
Hubbard left and came back about ten minutes later, asking if the victim was from the
neighborhood and if he was dead. Officer Turner responded that the victim was not from the
neighborhood and was indeed dead. Hubbard reacted with what Officer Turner later described
as a fake expression of shock. Hubbard then stated, “You think Leonard [] and Charlevois is out
cold, Gray and Mack is.” Officer Turner took this to mean that Hubbard believed that the Gray
and Mack neighborhood had more drugs and violence than nearby locations. Because no one
had indicated this was a drug-related killing and Hubbard’s comment implied that it might be,
Officer Turner asked if Hubbard would have a seat in the patrol car. Hubbard declined and was
allowed to leave.

Four days later, on January 21, Sergeant Joann Kinney took Hubbard into custody and
asked him about Penn’s killing. Hubbard stated that he did not know anything about Penn’s
death until Sergeant Kinney mentioned it. He further stated that he had not seen Penn since the
1980s and was not on Gray Street on January 17. These statements were all false. Hubbard was
 No. 21-2968 Hubbard v. Rewerts Page 3

on Gray, knew about Penn’s death, and was seen discussing drug sales with Penn the day before
his death.

On January 23, neighborhood resident Curtis Collins gave a statement to the police. In
his statement, Collins, signing under the alias “Tony Smith,” claimed that he was in the party
store on Gray and Mack on January 17 when he saw Hubbard and another man arrive at the
store. Collins stated that he left the store about two minutes later and that as he was leaving, he
turned back and noticed Hubbard and the other man leaving the store together and trekking
across Gray and Mack. Collins continued walking down the street, but, he said, when he heard
gunshots he turned and saw Hubbard running through a vacant lot. Collins ran toward the scene
and saw a body lying in the street in front of a house. After obtaining Collins’s statement, the
State charged Hubbard with first-degree murder and possession of a firearm during the
commission of a felony.

The state trial court conducted a preliminary examination on February 4, 1992, before
Hubbard’s actual trial. Collins testified at this preliminary examination, largely repeating his
statement to the police. He recounted that he had left the store and walked about 3–5 feet down
the street when he heard gunshots and turned to see Hubbard running away (inexplicably having
gotten more than 200 yards away from the store in the time Collins took to traverse 5 feet).
During cross-examination, Collins recounted that the time all this occurred was around 8:15 p.m.
(which, on January 17, meant it was dark). He then admitted there were no streetlights or lamps
in the area where Hubbard allegedly shot Penn and that he could not see Hubbard’s face. On
redirect, Collins claimed that even though he could not see Hubbard’s face, he identified him by
the scar on the back of his head. After Collins’s testimony, the trial court asked Hubbard to
stand and turn around and noted that the scar in question was no more than three inches in
length.

—The Trial

The first day of Hubbard’s bench trial commenced on August 31, 1992.1 The
prosecutor’s first witness was Curtis Collins. Right away, Collins recanted and denied he was in

1
Hubbard waived a jury trial.
 No. 21-2968 Hubbard v. Rewerts Page 4

the vicinity of Gray and Mack on January 17. The prosecutor sought to impeach him with his
February 4 testimony. Defense counsel then asked Collins why his story changed between
February and August. Collins replied that when he gave his initial story to the police, he was
mentally unstable because a friend (not Penn) had just died. He also stated he was afraid of
getting into trouble because he had violated parole by taking his ankle monitor off. Lastly, he
claimed that the police threatened to charge him for being “on escape” (presumably referring to
his parole violation, not an actual a jailbreak) and that they would pay him $10,000 to remove
him to Texas as part of a witness protection program if he testified.

The State tried to save its disintegrating case. The prosecutor called another witness,
John Trammel, who placed Hubbard among the spectators at the scene once the police and
medical personnel arrived. The prosecutor next called Leon Penn, the victim Rodnell Penn’s
brother, who testified that Hubbard and Rodnell were together on January 16, the day prior, and
that Hubbard told Rodnell that he would see him the next day. According to Leon, Rodnell Penn
was in the business of selling drugs for Hubbard. The prosecutor’s final witness that day was
Officer Turner, who testified to his seeing Hubbard on the street that night as EMS tended to
Rodnell’s body.

On the second day of trial, the prosecutor called Andrew Smith, another Gray and Mack
neighborhood resident, who said he observed Hubbard on January 17, crossing the street from
the party store to Gray Street with two other men at around 8–9 p.m. About 3–4 minutes later,
while Smith was in the party store, he heard gunshots. He stayed in the store for a few minutes
and then walked to the gunshot area, by which time the police had responded. Smith did not
actually see who shot Penn or anyone fleeing from the scene.

On the third and final day of trial, the prosecutor recalled Curtis Collins. Collins
withdrew his recantation and returned to his former story about identifying Hubbard by the scar
on the back of his head. He stated that he had lied two days earlier because he heard rumors that
his family was threatened. During cross-examination Collins admitted that he had been arrested
and threatened with perjury charges following his initial recantation. After Collins’s testimony,
the prosecutor and defense counsel stipulated to some indisputable facts—facts which the
prosecutor argued established motive: specifically, that Hubbard had a prior murder charge from
 No. 21-2968 Hubbard v. Rewerts Page 5

1988 and that Rodnell Penn was supposed to testify at that murder trial. Although Penn never
testified at that trial, he did testify at the preliminary examination. After this stipulation, the
State rested.

The defense called Raymond Williams, who testified that he was with Curtis Collins in
another part of town from 8–10 p.m. on January 17th. Williams claimed they were gambling.
He remembered the night well because that was when the movie Juice had come out in theaters
and he had gone to see it that night, leaving the house around 10:00 p.m. Roney Fulton testified
to the same fact—Curtis Collins was at his house with Williams.

The defense next called Thomas and Vanessa Spells, friends of Hubbard’s. Thomas
testified that Hubbard was with him at his (Spells’s) house from 6–9 p.m. on January 17 and that
his wife, Vanessa, arrived at their apartment at around 8:15, while he and Hubbard were still
there. Vanessa also testified to this fact. Thomas and Hubbard then left the Spells’s residence to
go to Hubbard’s mother’s house, where she had been taking care of the Spells’s child. The pair
then saw the ambulance and police cruisers on Gray Street and Hubbard walked up to Officer
Turner to ask what was going on.

The defense rested and the court recounted all the evidence presented at trial. The court
found Hubbard not guilty of the firearm charge but guilty of the first-degree murder charge.
While recounting the evidence, the trial court gave some insight into the basis for its verdict,
noting that Collins’s testimony provided little value because, Collins “did look quite nervous”
and was “down right lying to the Court,” and noting that Hubbard had lied to the police.

—Subsequent Proceedings and New Evidence

Hubbard’s attempts to overturn his conviction on both direct appeal and collateral review
proved unsuccessful. People v. Hubbard, 554 N.W.2d 910 (Mich. 1996) (unpublished table
decision); People v. Hubbard, 843 N.W.2d 130 (Mich. 2013) (unpublished table decision).
Furthermore, Hubbard waited over ten years after the conclusion of the direct appeal to file his
state postconviction petition, well beyond the one-year time window during which federal habeas
claims must be brought. See 28 U.S.C. § 2244(d).
 No. 21-2968 Hubbard v. Rewerts Page 6

Notwithstanding AEDPA’s time bar, Hubbard filed this federal habeas petition in 2013
and, after much delay, amended the petition in 2020. The district court dismissed the petition as
untimely but certified Hubbard’s actual-innocence claim for appeal.

On appeal, Hubbard argues that the equitable exception outlined in Schlup v. Delo, 513
U.S. 298 (1995), and applied to AEDPA’s statute of limitations in McQuiggin v. Perkins, 569
U.S. 383 (2013), entitles him to belatedly file his federal habeas petition. Under that exception, a
petitioner may avoid AEDPA’s time bar if he can demonstrate, in light of “new reliable
evidence,” that, more likely than not, “no reasonable juror would have convicted him.” Schlup,
513 U.S. at 324, 327.

Hubbard now presents as “new evidence”:

• A news article describing an investigation by the United States Department of
Justice into the Detroit police department based on a reported practice of coercing
false confessions or testimony. The article named Sergeant Joann Kinney, the
officer who interviewed Hubbard and apparently threatened Collins.
• An affidavit by fellow prison inmate Askia Hill, who claimed that on January 17,
1992, he saw a man named Mark Goings argue with somebody on Gray Street, in
front of “Uncle Peter’s House” (a notorious residence in front of which Penn’s
body was found). He further claimed that he saw the man walk away from Mark
Goings, and that Mark Goings shot him.
• Affidavits from Roy Burford and Emanuel Randall stating that the “word on the
street” was that Mark Goings killed Penn.
• Affidavits from Raad Konja and Samir Konja, the brother co-owners of the party
store, who claimed that they never saw Curtis Collins in their store on January 17,
and that they would have noticed because they had previously banned him from
the store.
• A 2017 affidavit from Curtis Collins, recanting yet again his accusation against
Hubbard. He avers that he was coerced to testify against Hubbard because
Sergeant Kinney, Sergeant Gale, and Assistant Prosecutor Gonzalez threatened to
charge him with the murder of Penn if he did not testify against Hubbard. He
claimed that he had been scared to come forward until 2015, when he learned that
those individuals were no longer working in their government positions.
• A 2018 polygraph report, reporting Collins’s truthfulness when he said that he
was not “present when Carl Hubbard shot that man.”
• A FOIA request to the Detroit Police Department that purportedly demonstrates a
lack of corroborating evidence to Collins’s story at trial. Collins had testified that
 No. 21-2968 Hubbard v. Rewerts Page 7

he took a cab home after seeing the dead body, but when the State subpoenaed the
cab company, it apparently received no helpful records.
• An affidavit from Raymond Williams, who claimed that between August 31,
1992, and September 2, 1992 (the dates of Hubbard’s trial), he heard Collins
crying in a jail cell and moaning about how Sergeant Kinney forced him to lie
about Hubbard.
• An affidavit from Elton Carter who claimed that Collins admitted to lying about
Hubbard’s involvement in Penn’s murder.

II.

Because Hubbard’s federal habeas petition was filed in 2013, AEDPA governs his claim.
White v. Warden, Ross Corr. Inst., 940 F.3d 270, 274 (6th Cir. 2019). We review de novo the
district court’s dismissal of Hubbard’s habeas petition as untimely. Souter v. Jones, 395 F.3d
577, 584 (6th Cir. 2005). We review the district court’s factual findings for clear error. Id.

III.

AEDPA imposes a one-year time bar on federal habeas claims, which, as relevant here.
runs from the latest of, “(A) the date on which the judgment became final by the conclusion of
direct review or the expiration of the time for seeking such review” or, “(D) the date on which
the factual predicate of the claim or claims presented could have been discovered through the
exercise of due diligence.” 28 U.S.C. § 2244(d)(1).2 Despite AEDPA’s clear language barring
untimely petitions, the Supreme Court has held that the statute is subject to an equitable
exception which allows petitioners to ignore the time bar in cases where they can credibly
demonstrate actual innocence. See McQuiggin, 569 U.S. at 386.

This exception requires a petitioner to 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 399 (quoting
Schlup, 513 U.S. at 327). If a petitioner meets this burden, he may belatedly file his underlying
federal habeas claim. See id. at 401. Several facets of this doctrine bear mention.

2
While Hubbard argued before the district court that his petition was timely under § 2244(d)(1)(D), his
argument was rejected, and he was not granted a certificate of appealability on that claim. His only claim before this
court is an equitable-exception claim.
 No. 21-2968 Hubbard v. Rewerts Page 8

First, this equitable-exception doctrine is not a freestanding substantive claim for habeas
relief. The Supreme Court has not decided whether actual innocence is a substantive ground for
relief. Id. at 392 (citing Herrera v. Collins, 506 U.S. 390, 404-05 (1993)). Rather, it allows a
petitioner to overcome a procedural barrier—in this case, AEDPA’s time bar—based on the
“miscarriage of justice” that results from the “incarceration of innocent persons.” Id. (quotation
marks and citation omitted). Actual innocence, in this sense, operates as a “gateway” by which a
petitioner may belatedly file other constitutional and federally cognizable claims. See id. at 393.

Second, the petitioner’s diligence (or lack thereof) in presenting new evidence is not a
threshold barrier to presenting an equitable-exception claim. Id. at 399. Rather, unexplained
delay in presenting new evidence is merely relevant to the credibility of the underlying claim, as
part of a holistic assessment of the record. See id.

Third, an equitable-exception claim requires the presentation of “new reliable evidence.”
Schlup, 513 U.S. at 324. While the Supreme Court has not explicitly defined what evidence
counts as “new,” this court has held that evidence is “new” for the purposes of the actual-
innocence inquiry so long as it was not presented at trial. Souter, 395 F.3d at 595 n.9 (citing
Schlup, 513 U.S. at 324); Freeman v. Trombley, 483 F. App’x 51, 57 (6th Cir. 2012). As for
reliability, the Schlup court illustratively listed “exculpatory scientific evidence, trustworthy
eyewitness accounts, or critical physical evidence” as examples of “reliable evidence.” Schlup,
513 U.S. at 324. Such evidence, the Court stated, is “obviously unavailable in the vast majority
of cases,” because actual-innocence claims are “rarely successful.” Id.

Fourth, while a credible claim of actual innocence requires “new reliable evidence,”
federal courts must not limit their analysis to such evidence. House v. Bell, 547 U.S. 518, 537
(2006). The court must instead look at the entire record, “old and new” evidence, without regard
to its admissibility, before determining whether a petitioner has credibly shown actual innocence
sufficient to overcome a habeas procedural barrier. Id. at 538. Based on the entire record, the
court must then determine whether “no reasonable juror would find [the petitioner] guilty.” Id.
This may require the federal court to make its own credibility determination as to witness
testimony in the record. Id. at 538-39.
 No. 21-2968 Hubbard v. Rewerts Page 9

Fifth, and perhaps most importantly, “‘actual innocence’ means factual innocence, not
mere legal insufficiency.” Bousley v. United States, 523 U.S. 614, 623 (1998) (citing Sawyer v.
Whiteley, 505 U.S. 333, 339 (1992)). And in the “gateway” context, our court has said the same
thing. Gulertekin v. Tinnelman-Cooper, 340 F.3d 415, 427 (6th Cir. 2003). This means that a
petitioner may not pass through the equitable gateway by simply undermining the state’s case.
Rather, he must demonstrate that he factually did not commit the crime. Of course, dismantling
the state’s case is relevant and helpful to the petitioner because it leaves a vacuum to be filled by
an exonerative explanation; but it is not sufficient in and of itself. This distinction between
exonerating evidence and impeachment evidence undergirds both of the Supreme Court’s
landmark equitable-exception cases. Schlup, 513 U.S. at 324; House, 547 U.S. at 552–53.

Start with Schlup. The equitable exception adopted by the Schlup Court rested not on the
notion that all convictions must be supported by constitutionally sufficient evidence, but on the
fact that the writ of habeas corpus, at its core, is an equitable remedy designed to achieve the
“ends of justice.” Schlup, 513 U.S. at 319-20. A miscarriage of justice, in this context, does not
include every legal wrong inflicted on a defendant but is instead confined to the “rare” and
“extraordinary case” in which the petitioner is innocent. Id. at 321. To achieve the ends of
justice and provide an equitable gateway for the innocent, the Court crafted a “standard of proof”
to govern such claims, requiring a petitioner to 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 322, 327.
This probabilistic standard, despite its similarity to the sufficiency standard in Jackson v.
Virginia, 443 U.S. 307 (1979), was not meant to serve as collateral sufficiency review. Schlup,
513 U.S. at 330 (explicitly distinguishing Jackson). Rather, this standard was simply meant to
reflect the degree of proof needed to make a successful actual-innocence claim. Proof of actual
innocence was the end; evaluating the effect of the evidence on a hypothetical factfinder was
merely the means. In this regard, the two inquiries are markedly different. Jackson asks whether
 No. 21-2968 Hubbard v. Rewerts Page 10

sufficient evidence exists such that the government could constitutionally convict. Schlup asks
whether the petitioner actually committed the crime. Id. at 330-31.3

The Court’s analysis in House, 547 U.S. at 518, confirms this reading. In House, a
woman named Carolyn Muncey was murdered and her body found in a ditch on the side of the
road. Id. at 521-22. During the initial search for Muncey after her unexpected disappearance,
her cousin noticed Paul House climbing out of an embankment with a rag, wiping his hands. Id.
at 524. When authorities found Muncey’s body near that ditch, House became the prime suspect
and was charged with her murder. Id. at 524-28. The State presented voluminous evidence, but
primarily relied on the finding of Muncey’s blood on House’s jeans and House’s semen on
Muncey’s person. Id. at 528-29.

On collateral review, House conclusively demonstrated that the blood found on his pants
was due to the negligent spillage of blood from Muncey’s autopsy in the same evidence box
carrying his jeans. Id. at 544. House further proved that the semen found on Muncey was not
his but her husband’s. Id. at 540-41. Despite these new revelations’ effectively dismantling the
State’s entire case, the Court required more. Only after House provided a credible confession by
Muncey’s husband as the true perpetrator did the Court find that he met the actual-innocence test
to permit his federal habeas claim to proceed. Id. at 548-553.

Consider too Cleveland v. Bradshaw, 693 F.3d 626 (6th Cir. 2012), in which this court
found a credible claim of innocence. There, the petitioner produced a reliable recantation by the
state’s star witness. Id. at 629–30. But in addition, the petitioner produced evidence to establish
an alibi—both an alibi witness and contemporaneous flight records. Id. at 637–38. By contrast,
in Davis v. Bradshaw, 900 F.3d 315 (6th Cir. 2018), which concerned a co-defendant involved in
the same crime, we rejected the innocence claim. There, we concluded that Davis had not
“presented similar, reliable alibi evidence.” Id. at 334. Like in House, these cases show that the
relevant new evidence must go to the petitioner’s actual innocence, not merely legal innocence.

3
This is further confirmed by the fact that Schlup allows for the examination of non-admissible evidence,
demonstrating that the object of the Schlup standard is to ascertain truth, as opposed to guaranteeing a criminal
procedural right. Schlup, 513 U.S. at 327-28.
 No. 21-2968 Hubbard v. Rewerts Page 11

The dissent disagrees. It argues that Hubbard is “actually innocent” if he demonstrates
that “any reasonable juror would have reasonable doubt.” Dissent, at 23 (quoting House, 547
U.S. at 538). That is, the dissent claims that this standard defines actual innocence, rather than
simply providing a burden of proof by which courts assess actual innocence.

What’s the difference? The dissent quotes a recent case of ours to explain: “A defendant
who can clearly and convincingly dismantle the government’s case against him could therefore
overcome the procedural bar of § 2244(b)(2)(B)(ii)—by removing that ‘certitude’—even if he
cannot show that he is, in fact, innocent.” Id. (quoting Keith v. Hill, 78 F.4th 307, 315 (6th Cir.
2023) (emphasis added by dissent)). In other words, if a defendant can instill reasonable doubt
as to his guilt by impeaching the State’s case against him, he has proven “actual innocence,”
regardless of whether, in real time and space, it is more likely than not that he actually engaged
in the conduct the state alleges to be criminal.

This argument is wrong. But to fully explain why, we must review the case the dissent
relies on: Keith v. Hill. That case involved a habeas petitioner who, after his conviction for a
drug-related triple homicide, sought to bring a fourth successive habeas petition alleging
violations of Brady v. Maryland, 373 U.S. 83 (1963). Keith, 78 F.4th at 308-09, 312-14.
AEDPA generally bars successive habeas petitions unless, as relevant in Keith,

[T]he facts underlying the claim, if proven and viewed in light of the evidence as
a whole, would be sufficient to establish by clear and convincing evidence that,
but for the constitutional error, no reasonable factfinder would have found the
applicant guilty of the underlying offense.

Id. at 314 (quoting 28 U.S.C. § 2244(b)(2)(B)(ii).

The government in Keith argued that Keith could not meet AEDPA’s successive-petition
standard unless he factually proved his innocence (i.e., that he literally did not commit the
crime). Id. at 315. But the court, accepting the premise that the successive-petition provision
requires a showing of actual innocence, claimed that “actual innocence” was distinct from
“factual innocence.” Id. “Actual innocence,” the court stated, is proved whenever, based on the
evidence as a whole, “no reasonable juror would have convicted [the defendant].” Id. (quoting
 No. 21-2968 Hubbard v. Rewerts Page 12

House, 547 U.S. at 526, and McQuiggin, 569 U.S. at 385, 393-94). “Factual innocence,” the
court distinguished, means what it says—the defendant didn’t do it. Id.

The Keith court’s reading of House and McQuiggin is both inapposite and inaccurate.
Inapposite here because AEDPA’s successive-petition provision—the provision at issue in
Keith—involves a different inquiry than does the equitable-exception standard. The successive-
petition provision considers whether, absent the claimed constitutional error, the new factual
predicates would establish by clear and convincing evidence that no reasonable factfinder would
have found the habeas petitioner guilty. See Baugh v. Nagy, No. 21-1844, 2022 WL 4589117, at
*10 (6th Cir. Sept. 30, 2022) (alternatively arguing—in the successive petition context—that
improper withholding of Brady material did not prejudice defendant because it could only be
used as impeachment evidence, and that the defendant was guilty in any event). But the
equitable-exception standard at issue here requires first a showing that the defendant, based on
the entire record, is actually innocent; only then may the court even consider his assertions of
constitutional error. See McQuiggin, 569 U.S. at 393 (describing equitable-exception claim as a
“gateway” to consideration of constitutional error). Considering the differences between the two
standards, McQuiggin and House were arguably irrelevant to Keith’s analysis.

And Keith reads the Supreme Court’s equitable-exception cases inaccurately. Keith reads
McQuiggin’s and House’s standard of proof—that considering the evidence as a whole, “no
reasonable juror would have convicted [the petitioner]”—as the Supreme Court’s definition of
actual innocence (as the dissent does), rather than its means of assessing actual innocence.
Keith, 78 F.4th at 315 (citing McQuiggin, 569 U.S. at 385, 393-94, and House, 547 U.S. at 526).
In doing so, the Keith court had to claim that “actual innocence” is somehow distinct from
“factual innocence.”4 Id. But the Supreme Court, while adjudicating an equitable-exception
claim, long ago explicitly stated otherwise: “It is important to note in this regard that ‘actual
innocence’ means factual innocence, not mere legal insufficiency.” Bousley, 523 U.S. at 623.

4
Despite the ironic reality that “actual” literally means “existing in fact.” Actual, MERRIAM-WEBSTER,
https://www.merriam-webster.com/dictionary/actual (last accessed Apr. 11, 2024).
 No. 21-2968 Hubbard v. Rewerts Page 13

Keith’s citation to Jackson v. Virginia, 443 U.S. 307, 315 (1979), the seminal sufficiency-review
case, reveals its failure to consider Bousley or Gulertekin. Keith, 78 F.4th at 315.5

The simple premise undergirding our description of actual innocence or factual innocence
(they are equivalent) is that objective, historical events occur, regardless of a third party’s ability
to gather evidence of that occurrence after the fact. Either Hubbard killed Rodnell Penn, or he
did not. The government’s evidence of that event in no way changes what actually occurred.

Of course, the government’s ability to convict Hubbard depends entirely on the proof it
provides. All evidentiary questions must in some way reference a standard of proof or certitude.
That is why a guilty verdict is not a declaration that a defendant objectively committed a crime
(though we often use that shorthand) but rather that the government has provided evidence that
he did so sufficient to dispel all reasonable doubts. Inversely, a “not guilty” verdict is just that—
an acquittal based on the government’s inability to meet its burden of proof. An acquittal is not a
declaration of innocence.

But the error of the dissent (and Keith) is that they equate the two. Dissent, at 23; see
also Keith, 78 F.4th at 315. The dissent conducts its actual-innocence analysis under the
presupposition of a hypothetical new trial, and Keith claims a defendant proves his innocence by
removing the “certitude” of his conviction. But Schlup took pains to distinguish the burden of
proof from the objective reality: “Our reference to Winship is intended merely to demonstrate
that it is quite consistent with our jurisprudence to give content through a burden of proof to the
understanding that a fundamental injustice would result from the erroneous conviction and
execution of an innocent person.” Schlup, 513 U.S. at 326 n.42 (emphasis added).

What role does the burden of proof play in our analysis of an equitable-exception claim?
It establishes the required probability of the objective, historical fact. With the knowledge that
factual innocence is the object of proof, the question of whether “it is more likely than not that
no reasonable juror would have convicted him” simply gives us a metric by which to assess
whether the petitioner has met that burden. Id. at 327. In this regard, we reject the assertion that

5
At any rate, Gulertekin dealt with the gateway innocence context, which is this case, and Keith did not.
Still, the Supreme Court has not indicated that the “innocence” definitions are different in these different contexts—
even though the standards themselves are slightly different.
 No. 21-2968 Hubbard v. Rewerts Page 14

requiring a probabilistic showing of actual innocence necessarily requires “conclusive
exoneration.” Dissent, at 23 (quoting House, 547 U.S. at 553). We simply require that a
petitioner show the probability of his innocence, rather than merely impeach the State’s case.
This means that a defendant must put forth some type of reliable evidence that is exonerative in
nature. That evidence need not conclusively prove his exoneration to make the gateway
showing, but it must, at minimum, go towards his innocence. See Schlup, 513 U.S. at 324
(describing “reliable evidence” of innocence as “exculpatory”); see also Hyman v. Brown, 927
F.3d 639, 665 (2nd Cir. 2019) (recanting eyewitness of shooting did not establish innocence
because failure to inculpate the petitioner does not exonerate the petitioner). In other words, the
new evidence mirrors requiring a showing that the petitioner did not do the crime.

Finally, the Supreme Court has repeatedly said that the actual-innocence remedy is
reserved for only the most extraordinary case. See Calderon v. Thompson, 523 U.S. 538, 558
(1998); House, 547 U.S. at 536–37; Murray v. Carrier, 477 U.S. 478, 496 (1986); Sawyer, 505
U.S. at 339 n.6; McQuiggin v. Perkins, 569 U.S. 383, 393 (2013); Dugger v. Adams, 489 U.S.
401, 410 n.6 (1989); Herrera v. Collins, 506 U.S. 390, 426–27 (1993) (O’Connor, J.,
concurring). It would conflict with that directive to allow a petitioner to overcome procedural
default only by calling the state’s case into question. To be fair, at least one treatise asserts that
“there is no requirement that the petitioner present affirmative proof of innocence. It is enough if
the ‘post-conviction evidence casts doubt on the conviction by undercutting the reliability of the
proof of guilt . . . .’” Fed. Habeas Manual § 9B:80 at 155 (2023) (quoting Sistrunk v. Armenakis,
292 F.3d 669, 673 (9th Cir. 2002) (en banc)). In Sistrunk, the Ninth Circuit rejected the
petitioner’s actual-innocence claim, which was based largely on discrediting the state’s main trial
witness. Sistrunk, 292 F.3d at 675–77. But Sistrunk was decided pre-House. And in a post-
House case, the Ninth Circuit rejected the petitioner’s freestanding actual-innocence claim when
it was “based on recantation testimony alone.” Jones v. Taylor, 763 F.3d 1242, 1248 (9th Cir.
2014). So we are in good company in reading House and Schlup to require petitioners seeking to
meet the equitable exception to point to new evidence that contemplates establishing the
petitioner’s actual innocence—in other words, to make a gateway showing that he didn’t do the
crime (which goes to actual innocence) instead of merely attacking the state’s case (which could
show legal innocence).
 No. 21-2968 Hubbard v. Rewerts Page 15

The genesis of this distinction comes from Sawyer. On “factual” or “actual innocence,”
the Supreme Court said a “prototypical example,” at least in a “colloquial sense” is when the
state “has convicted the wrong person of the crime.”6 Sawyer, 505 U.S. at 340. This would
include those “rare instances” when “it may turn out later . . . that another person has credibly
confessed to the crime, and it is evident that the law has made a mistake.” And in this type of
case, “the concept of ‘actual innocence’ is easy to grasp.” Id. at 340–41.

Importantly, also in Sawyer—the Court notes that the argument that the state has
convicted the wrong person is a common argument made in post-trial motions after conviction.
But these arguments are regularly rejected “because the evidence adduced in support of them
fails to meet the rigorous standards for granting such motions.” Id. at 340. The “rare” instances
that can lead to successful results are the ones, as noted above, whether the petitioner can
establish his innocence. So the genesis of the distinction between arguments about legal
sufficiency, on the one hand, and actual/factual innocence is in Sawyer.

What does all this mean? The bottom line is that Hubbard cannot satisfy the actual-
innocence standard only by undermining the state’s case alone. First, that scenario is not the
“prototypical” scenario, as explained above, that Sawyer outlines. And that prototypical scenario
is contrasted with post-trial sufficiency arguments. Second, the requirement that “new reliable
evidence” be produced, “whether it be exculpatory scientific evidence, trustworthy eyewitness
accounts, or critical physical evidence,” Schlup, 513 U.S. at 324, is inconsistent with simply
undermining the state’s case. These types of new evidence (though admittedly not exclusive) all
would appear to contemplate establishing the factual innocence of the petitioner. The Court was
contemplating that the mine-run case would involve factual innocence. In other words, the
character of the new evidence that the Court requires mirrors requiring a showing that the

6
Under habeas law, courts have drawn distinctions drawn between the § 2244(b)(2)(B)(ii) context—a direct
descendent of Sawyer; the “gateway” situation, which comes from Murray v. Carrier, 477 U.S. 478 (1986); and so-
called freestanding actual-innocence claims, see Carriger v. Stewart, 132 F.3d 463, 476 (9th Cir. 1997). At the
Supreme Court, however, the constitutional implications of a freestanding actual-innocence claim remain “an open
question.” Dist. Att’y’s Off. v. Osborne, 557 U.S. 52, 71–72 (2009) (assuming without deciding a “federal
constitutional right to be released upon” a freestanding actual-innocence claim). But the distinction in those cases is
the burden of proof that the petitioner must meet, i.e., clear and convincing evidence, § 2244(b)(2)(B)(ii), versus
more likely than not, Bousley, 52 U.S. at 623, versus requiring an “extraordinarily high” threshold showing, see
Herrera v. Collins, 506 U.S. 390, 417 (1993). It is not about what “innocence” means.
 No. 21-2968 Hubbard v. Rewerts Page 16

petitioner didn’t do the crime. To show he is entitled to equitable tolling, Hubbard must present
new evidence that contemplates establishing his actual innocence—a lower bar than conclusively
establishing that he did not do the crime, but one that requires him to do more than only
undermine the state’s case.

With the law established, we proceed to analyze Hubbard’s newly presented evidence.

IV.

Hubbard presents new evidence that purportedly demonstrates that Curtis Collins’s
incriminating testimony is false and that a different man killed Rodnell Penn. We consider its
reliability, and whether it establishes his innocence, in turn.7

—Curtis Collins’s Testimony

Hubbard presents numerous affidavits purportedly demonstrating that Curtis Collins was
likely not at the party store that night, and therefore could not have seen Hubbard shoot Penn.
We first consider whether this evidence is reliable.

A vacillating witness who changes his story multiple times is often presumed to be
unreliable. See Davis v. Bradshaw, 900 F.3d 315, 330 (6th Cir. 2018) (a change in story could
“make it more difficult for reasonable juror to find it reliable”). In such circumstances, a
reasonable juror would likely need “corroborating evidence” to discern which version of the
witness’s story is true. Id. Unexplained delay between long past events and a witness’s decision
to testify to those events also casts a pall of unreliability over the testimony. See McQuiggin,
569 U.S. at 399.

Collins’s recantation affidavit marks the third time he has changed his story. This alone
makes it difficult to credit his account unless reliable corroborating evidence indicates his
truthfulness. The corroborating evidence that Hubbard now presents, however, is weak. All of

7
Contra the dissent, we are not isolating our consideration of the evidence. Dissent, at 41. We do evaluate
how the evidence interacts and aggregately contributes or detracts from a showing of innocence. Infra, at 17-19.
We simply have chosen to organize our analysis differently than the dissent would.
 No. 21-2968 Hubbard v. Rewerts Page 17

the corroborating affidavits are decades after the fact, and many do not explain the reason for the
delay.

While we could laboriously examine the motive of each affiant or criticize the reliability
of the polygraph examination or question the relevance of the DOJ study, we need not do so here
because Collins’s recantation is not reliable. Even if it were, it does not go towards Hubbard’s
innocence—it goes only toward undermining the state’s case. Even if we discredit Collins’s
testimony from Hubbard’s trial, Hubbard is left with only the lack of an incriminating witness,
not the presence of an exonerating one. See Hyman, 927 F.3d at 665. And, as stated previously,
Hubbard must present some reliable exonerative evidence to succeed in his equitable-exception
claim. So even though a reliable recantation could be one part of the actual-innocence analysis,
Collins’s unreliable recantation does not warrant consideration.

—An Alternative Suspect

Hubbard’s evidence pointing to Mark Goings as an alternative suspect does not fit that
bill. The first affidavit, from fellow inmate Askia Hill, claims that Hill saw Mark Goings arguing
with someone on Gray Street on January 17, 1992. When the unknown man turned to walk
away, Mark Goings shot him. The other two affidavits, from Roy Burford and Emmanual
Randall, claim that the “word on the street” was that Goings shot Penn.

But these affidavits are patently unreliable. Consider Hill’s affidavit. Although Hill
claims that he does not know Hubbard, he was incarcerated with him at the time of the affidavit,
rendering both that statement and his motives suspect. Herrera, 506 U.S. at 423 (O’Connor, J.,
concurring) (“It seems that, when a prisoner’s life is at stake, he often can find someone new to
vouch for him.”). But even disregarding that, Hill’s affidavit bears no indicia of reliability. Hill
waited almost twenty years to come forward and tell anyone that the State had supposedly
convicted the wrong man. His only explanation for this delay is that he was “afraid for [his] life”
and “didn’t want any trouble with anybody in the neighborhood.” This hardly clears the cloud of
skepticism over the twenty-year silence. Nor does Hill provide any corroborating evidence to
bolster his account. The affidavit is barebones, claiming only that Hill saw Mark Goings kill
someone (presumably Penn) in the Gray and Mack neighborhood that night. The affidavit lacks
 No. 21-2968 Hubbard v. Rewerts Page 18

even a shred of information corroborating that Hill was in the vicinity of the Gray and Mack
neighborhood on January 17. Without more, Hubbard’s actual-innocence-gateway claim cannot
prevail based on this affidavit alone.

The Burford and Randall affidavits fare no better. Both are unreliable hearsay. While we
can consider hearsay evidence when adjudicating an actual-innocence gateway claim, we do not
have to give this “word on the street” hearsay a level of credibility that it does not deserve.

Even if we credited all this evidence as true, it does not actually establish what Hubbard
claims. Believing every word of all three affidavits establishes only that 1) Mark Goings shot
someone on Gray and Mack, and 2) the rumor in the neighborhood was that Mark Goings shot
Rodnell Penn. This evidence does not actually establish that anyone, by first-hand account, saw
Mark Goings shoot Rodnell Penn.

—The Alibi Witnesses

Although Hubbard has not presented any new evidence regarding his potential alibi, we
review the whole record, “old and new,” when faced with an actual innocence claim. House, 547
U.S. at 538. As explained previously, Collins’s recantation is unreliable. However, to ensure we
examine the whole record rather than isolated segments, we will reconsider the weight of
Hubbard’s alibi witnesses in light of Collins’s recantation. Thomas and Vanessa Spells testified
that Hubbard was with them at their apartment from approximately 6–9 p.m. on January 17, the
time of Penn’s shooting. Previously, this testimony, by people with a friendly relationship to
Hubbard and therefore perhaps a motive to lie, was set against two witnesses (Collins and
Andrew Smith) who testified that Hubbard was near the party store during that time. Now,
however, assuming Collins’s recantation is believable, the Spells’s story seems somewhat more
likely.

Unfortunately for Hubbard, however, their story falls short of the probabilistic burden
established in Schlup. Consider first Smith’s incriminatory testimony. While he did not see
Hubbard fleeing from the vicinity of the shooting, he claims he did see Hubbard near the party
store at around 8 p.m. with two other men, and a few minutes later he came out of the party store
and saw a dead body down the street. This directly contradicts the Spells’ story that Hubbard
 No. 21-2968 Hubbard v. Rewerts Page 19

was at their apartment. Furthermore, the prosecutor established a plausible motive for Hubbard
to kill Penn: Penn had previously agreed to testify against Hubbard at a prior murder trial. On
top of that, Hubbard and Penn were dealing drugs together around the time of Penn’s death, and
Hubbard planned to see Penn the day of Penn’s death. In light of all this evidence, even
assuming Collins’s recantation is reliable, a reasonable trier of fact could still credit Smith’s
testimony as more reliable than the Spells’s.

V.

Hubbard argues that even if we find he cannot make out an equitable-exception claim on
this record, we should remand for an evidentiary hearing. Typically, “[i]n deciding whether to
grant an evidentiary hearing, a federal court must consider whether such a hearing could enable
an applicant to prove the petition’s factual allegations, which, if true, would entitle the applicant
to federal habeas relief.” Schriro v. Landrigan, 550 U.S. 465, 474 (2007). And we must consider
the actual-innocence standard in determining whether an evidentiary hearing is appropriate. See
Turner v. Romanowski, 409 F. App’x 922, 930 (6th Cir. 2011).

An evidentiary hearing wouldn’t do much to bolster the inherent unreliability of
Hubbard’s evidence. Even if we credited all his evidence as true, for the reasons explained
above, it is not more likely than not that no reasonable juror would have convicted Hubbard. In
other words, “[n]either the existing record nor the additional affidavits submitted by [Hubbard]
are sufficient to cast any significant doubt on the guilty verdict rendered by the state court, and
the district court was not required to hold an evidentiary hearing on such insubstantial factual
allegations.” Id.

VI.

Hubbard waited over twenty years after his conviction to file his federal habeas petition.
He now asks us to ignore this delay and allow his habeas petition to be considered on the merits,
all in the hopes of overturning a conviction more than thirty years old. But the State of Michigan
has an interest in the finality of Hubbard’s conviction. Congress recognized as much by passing
AEDPA and imposing a one-year time bar on federal habeas claims. That time bar can be
ignored only when a petitioner shows that the State has imprisoned an innocent person.
 No. 21-2968 Hubbard v. Rewerts Page 20

Hubbard’s new evidence does not meet that burden. He must therefore comply with the same
law with which all other habeas petitioners must comply. Accordingly, we AFFIRM the district
court’s judgment.
 No. 21-2968 Hubbard v. Rewerts Page 21

_________________

DISSENT
_________________

COLE, Circuit Judge, dissenting. Petitioner Carl Hubbard appeals the district court’s
dismissal of his petition for a writ of habeas corpus. Hubbard, currently incarcerated in a Carson
City, Michigan correctional facility, was convicted of the murder of Rodnell Penn in 1992. His
28 U.S.C. § 2244 petition was dismissed by the district court as untimely and ineligible for
equitable tolling of AEDPA’s one-year statute of limitations. But the district court granted a
certificate of appealability on the latter ground, which is now before this court. Because I find
that Hubbard has demonstrated a credible claim of actual innocence, I would hold he is entitled
to equitable tolling. I would therefore reverse the district court’s dismissal of his habeas petition
and allow Hubbard to pursue the merits of his underlying claims, and would, at a minimum,
remand for an evidentiary hearing on his new reliable evidence. For these reasons, I respectfully
dissent.

I.

Before discussing the specifics of Hubbard’s case, it is important to clarify a few of the
governing legal standards, as I take issue with the majority’s interpretation of the standard of
review applied to the underlying facts and, more importantly, the actual innocence requirement
as articulated in our court’s and the Supreme Court’s binding precedent.

We review a district court’s dismissal of a writ of habeas corpus as barred by the statute
of limitations de novo. Souter v. Jones, 395 F.3d 577, 584 (6th Cir. 2005). We also review a
district court’s refusal to apply equitable tolling based on actual innocence de novo, reflecting
that a claim of actual innocence is “primarily a question of law” on which this court “do[es] not
defer to the district court’s judgment.” McSwain v. Davis, 287 F. App’x 450, 459 (6th Cir. 2008)
(citing House v. Bell, 547 U.S. 518, 539–40 (2006)).

Importantly, while a district court’s factual findings are typically reviewed for clear error,
Souter, 395 F.3d at 584, factual findings made without an evidentiary hearing—as is the case
 No. 21-2968 Hubbard v. Rewerts Page 22

here—are reviewed de novo, Burton v. Renico, 391 F.3d 764, 770 (6th Cir. 2004) (citing Bugh v.
Mitchell, 329 F.3d 496, 500 (6th Cir. 2003)).

The Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA) established a one-
year limitations period during which a state prisoner can bring a federal habeas corpus petition.
28 U.S.C. § 2244(d)(1). As Hubbard does not challenge the district court’s conclusion that his
petition is untimely, this issue is waived. Gregory v. City of Louisville, 444 F.3d 725, 737 (6th
Cir. 2006). So I, like the majority, focus on his remaining avenue for relief: equitable tolling.

AEDPA’s limitations period is not jurisdictional, and does not require courts to dismiss a
claim as soon as the “clock has run.” Day v. McDonough, 547 U.S. 198, 208 (2006). “[A]
petitioner who misses the deadline may still maintain a viable habeas action if the court decides
that equitable tolling is appropriate.” Allen v. Yukins, 366 F.3d 396, 401 (6th Cir. 2004).

Since Souter, this court has embraced equitable tolling of AEDPA’s one-year period
where a petitioner presents a credible claim of actual innocence. Cleveland v. Bradshaw, 693
F.3d 626, 631 (6th Cir. 2012) (citing Souter, 395 F.3d 577). A credible claim of actual
innocence operates as a “gateway” through which a petitioner may pass and argue the merits of
his underlying constitutional claims “despite a procedural bar that would ordinarily preclude
such review.” Id. at 632 (citing Schlup v. Delo, 513 U.S. 298, 315 (1995)). A petitioner is
“actually innocent” when he (1) presents “new reliable evidence” of his innocence that, (2) when
considered with all the old evidence in the record, makes it “more likely than not that no
reasonable juror would have convicted him.” Schlup, 513 U.S. at 324, 327.

The gap between my view of the case and the majority’s is at its widest with respect to
this second requirement, so I begin there. If it is established that a petitioner has presented “new
reliable evidence,” we then evaluate the second prong of the inquiry: whether “it is more likely
than not that no reasonable juror would have found petitioner guilty beyond a reasonable doubt.”
Schlup, 513 U.S. at 327–28. For this inquiry, we consider the “evidence in its entirety”: “all the
evidence, old and new, incriminating and exculpatory, without regard to whether it would
necessarily be admitted under rules of admissibility that would govern at trial.” House, 547 U.S.
at 537 (cleaned up) (citing Schlup, 513 U.S. at 327–28).
 No. 21-2968 Hubbard v. Rewerts Page 23

The individual pieces of evidence need not, on their own, support a finding of actual
innocence; “a reasonable jury might well disregard” a piece of evidence “[i]f considered in
isolation.” Id. at 552. Some pieces of evidence may “reinforce other doubts” as to the
petitioner’s guilt, while some pieces may support an inference of guilt. See id. at 553–54. This
probability inquiry looks at the aggregate impact of all of the evidence, not the myopic impact of
each piece itself. Id. at 537–38; contra, e.g., Maj. Op. at 18 (that “the Spells’s story . . . falls
short of the probabilistic burden”). This approach mirrors how the government often secures
convictions: In lieu of a smoking gun, a verified eyewitness, or other airtight evidence, the
government presents a compilation of circumstantial evidence in a way that, together, may allow
a judge or jury to draw the conclusion that the accused did, in fact, do the thing the government
says they did. Indeed, this is precisely the foundation on which Hubbard was convicted.

So, to be eligible for equitable tolling, Hubbard must make “a credible claim of actual
innocence.” McQuiggin v. Perkins, 569 U.S. 383, 391 (2013). This “actual innocence” inquiry
does not require him to conclusively prove his innocence. Instead, Hubbard meets this burden—
and is therefore “actually innocent”—if he demonstrates that “no reasonable juror would find
him guilty beyond a reasonable doubt—or, to remove the double negative, that . . . any
reasonable juror would have reasonable doubt.” House, 547 U.S. at 538.

To be sure, a petitioner can be actually innocent even without “conclusive exoneration.”
Id. at 553. We have recently emphasized as much:

Proof beyond a reasonable doubt requires jurors to “reach a subjective state of
near certitude of the guilt of the accused”; it is proof “so convincing that you
would not hesitate to rely and act on it in making the most important decisions in
your own [life].” A defendant who can clearly and convincingly dismantle the
government’s case against him could therefore overcome the procedural bar of
§ 2244(b)(2)(B)(ii)—by removing that “certitude”—even if he cannot show that
he is, in fact, innocent.

Keith v. Hill, 78 F.4th 307, 315 (6th Cir. 2023) (emphasis added) (citations omitted).

The majority nonetheless concludes that Hubbard fails to meet his burden after requiring
he “demonstrate that he factually did not commit the crime.” Maj. Op. at 9. But I am unaware
of, and am not directed to, a case demanding as much. More importantly, our court just
 No. 21-2968 Hubbard v. Rewerts Page 24

explicitly denounced such a requirement in a published opinion. See Keith, 78 F.4th at 315. In
so doing, we rejected the use of the “‘actual innocence’ standard as interchangeable with ‘factual
innocence,’” clarifying that a petitioner need not show that “he did not in fact commit the subject
crime.” Id.

As we have acknowledged, “the teachings of precedent are not always as clear as we
might wish[,] [e]specially in a complicated area like habeas.” Wright v. Spaulding, 939 F.3d
695, 699–700 (6th Cir. 2019). But one thing is crystal clear: “[T]he holding of a published
panel opinion binds all later panels unless overruled or abrogated en banc or by the Supreme
Court.” Id. at 700. Here, that holding is Keith, a holding that is also clear: Under Supreme
Court and Sixth Circuit precedent, a defendant can overcome AEDPA’s procedural bar “even if
he cannot show that he is, in fact, innocent.” Keith, 78 F.4th at 315.

Even before Keith, though, the Supreme Court, and the statute itself, made the actual
innocence inquiry clear: whether “no reasonable factfinder would have found the [petitioner]
guilty of the underlying offense.” § 2244(b)(2)(B)(ii); Calderon v. Thompson, 523 U.S. 538, 559
(1998); House, 547 U.S. at 537; McQuiggin, 569 U.S. at 395. So, in line with Schlup and its
progeny, our analysis turns on whether Hubbard has presented new reliable evidence collected
since trial that, when considered with the record as a whole, “raises sufficient doubt about his
guilt and undermines confidence in the result of his trial,” Souter, 395 F.3d at 590—not whether
Hubbard has affirmatively demonstrated that he did not kill Rodnell Penn. The majority’s focus
on the idea of “factual innocence” is therefore misguided under both Supreme Court and our
circuit’s precedent.

Having articulated the underlying standards, I start from the top.

II.

Rodnell Penn was shot and killed down the street from a party store in Detroit, Michigan
on January 17, 1992, around 9:30 p.m. As Hubbard walked by after Penn’s body had been
loaded into an ambulance, he asked Officer Craig Turner what happened and learned there had
been a homicide. Hubbard was taken into custody four days later and interrogated by Sergeant
Joann Kinney.
 No. 21-2968 Hubbard v. Rewerts Page 25

Curtis Collins gave a statement to police under the alias “Tony Smith,” claiming he saw
Hubbard at the party store the night of the shooting. The next day, Hubbard was charged with
first-degree murder.

At the preliminary examination, the State called three witnesses: Collins, Officer Turner,
and Sergeant Kinney. Here, Collins testified that he was in the party store for “five or ten
minutes” that night, encountered Hubbard with Penn, and left the store before Hubbard.
According to Collins, when he was “three” or “five” feet away from the store, he heard gun shots
and saw Hubbard running through a field. He claimed to have recognized Hubbard as he ran
based on “the scar on the back of his head” but noted that there were no streetlights, lamps, or
house lights nearby and that it was dark. Collins testified that he “jumped in a cab and went
home” after he saw the victim’s body.

A.

Hubbard’s single-judge trial began on August 31, 1992, and lasted three days. At trial,
the government presented no eyewitnesses or physical evidence connecting Hubbard to the
murder scene. The prosecution’s key witness, and the only evidence tying Hubbard to the scene
of Penn’s murder at the time of the shooting, was Curtis Collins.

On the first day of trial, the government reiterated Collins’s testimony from the
preliminary examination in its opening statement, but Collins recanted this testimony once he
took the stand that same day. He instead asserted he was not at the party store at any point that
night and therefore did not see Hubbard at the store or running across the field near Penn’s
murder. To explain this change, Collins explained his fear of getting in trouble with the police as
his reason for initially wrongly implicating Hubbard—specifically, threats that he would be
charged with Penn’s murder if he did not place the blame on Hubbard. This fear of the police
was real, as he was subsequently arrested for perjury as a result of the change in his testimony.

The government’s next witness, John Trammel, testified that he saw Hubbard wearing a
black jacket and standing “among the spectators” near the scene of the crime after the ambulance
arrived.
 No. 21-2968 Hubbard v. Rewerts Page 26

The second day of trial commenced. The police department’s evidence technician,
Randy Richardson, testified that the scene was “fairly dark,” confirmed that there was no street
lighting nearby, and estimated the distance between Penn’s body and the party store was about
375 feet.

Another witness, Andrew Smith, testified to seeing Hubbard with “two other guys” while
Smith was on his way to the party store that night, but could not remember what time that was or
who he allegedly saw Hubbard with. He testified that he did not see Collins, whom he knew,
anywhere in the area that night and that he stayed in the party store until the police arrived.

Lucinka Gross arrived at the party store after encountering Penn’s dead body on the street
on her way there and asked the store employees to call the police. She testified that she did not
see either Smith or Collins that night.

On the third day of trial, after being charged with perjury, Collins took the stand again
and withdrew his recantation—meaning, he changed his story to be closer to that of the
preliminary examination, implicating Hubbard yet again. Collins said unknown individuals
threatened him on the street after his testimony on the first day of trial. Collins’s new testimony
did not clarify how he had left the store three or four minutes before Hubbard but had made it
only, at most, 25 to 30 feet, yet Hubbard somehow traveled a few hundred feet away—from the
store to where the shooting occurred—in the same time period. Collins again said he recognized
Hubbard by the scar on the back of his head as he never saw his face. Collins’s perjury charges
were subsequently dropped, and he did not face perjury allegations for the change to his
testimony between the first and third days of trial.

Defense counsel called four witnesses. Collins’s “best friend[],” Raymond Williams,
testified that he and Collins were at a friend’s house gambling the night of Penn’s murder. That
friend, Roney Fulton, agreed, stating that Collins spent “all day” and evening at Fulton’s house
on January 17, 1992. Williams said he was with Collins until at least 10:00 p.m., when Williams
left to see a movie. Neither friend knew Hubbard well.

Hubbard’s friends, Thomas and Vanessa Spells, independently testified as to an alibi for
Hubbard: that Hubbard spent the relevant part of the evening of January 17, 1992, at their house
 No. 21-2968 Hubbard v. Rewerts Page 27

and otherwise with them. Thomas and Vanessa both testified that Thomas and Hubbard left their
house around 10:00 p.m. On their way to pick up the Spells’s baby from Hubbard’s mother,
Thomas and Hubbard saw an ambulance and Hubbard spoke to a detective. Thomas and
Hubbard arrived back at the Spells’s with the baby around 10:30 p.m.

After agreeing with defense counsel that “Collins’s testimony at times was very
conflicting and downright lying,” the court convicted Hubbard of first-degree murder. He was
subsequently sentenced to life without parole.

B.

Hubbard’s conviction was affirmed on direct appeal, where the Michigan Court of
Appeals acknowledged that “[t]he evidence upon which [Hubbard] was convicted was entirely
circumstantial.” (Mich. Ct. of Appeals, R.56-8, PageID 2560.) Hubbard then sought state post-
conviction relief, which was denied at multiple points. See People v. Hubbard, No. 92-001856
(Wayne Cnty. Cir. Ct. Mar. 15, 2012); reconsideration denied No. 92-001856 (Wayne Cnty. Cir.
Ct. May 31, 2012); People v. Hubbard, No. 311427 (Mich. Ct. App. May 7, 2013); leave denied
843 N.W. 2d 130 (Mich. 2013) (table).

Hubbard’s habeas petition was signed and dated on October 22, 2013, which operates as
the filing date for purposes of AEDPA’s statute of limitations. See Cretacci v. Call, 988 F.3d
860, 865 (6th Cir. 2021) (citing Houston v. Lack, 487 U.S. 266, 270 (1988)). After procedural
back and forth between the state and federal courts, this court reopened Hubbard’s federal habeas
case and allowed him to amend his petition on July 15, 2020. His petition sought relief on
several constitutional grounds, including due process claims arising from the prosecutor’s
coercion of Collins and withholding of evidence, and claims of ineffective assistance of trial and
appellate counsel. Hubbard argued that AEDPA’s statute of limitations was tolled because he
had a colorable claim of actual innocence, and requested an evidentiary hearing.

The district court did not address Hubbard’s request for an evidentiary hearing and
dismissed his petition with prejudice. In dismissing, the court found that the statute of
limitations under § 2244 had expired and that Hubbard was not entitled to equitable tolling of the
limitations period. As to this latter ground, however, the court found that “reasonable jurists
 No. 21-2968 Hubbard v. Rewerts Page 28

could debate whether evidence obtained by [Hubbard] after trial which suggests that he did not
commit the murder for which he was convicted could justify the application of equitable tolling
to excuse the untimely filing of the petition,” and granted a limited certificate of appealability.

Hubbard timely appealed, which brings us to the present dispute. On appeal, Hubbard
again asserts a claim of actual innocence premised on new reliable evidence that he submits
justifies equitable tolling of his habeas petition so that his underlying constitutional claims can be
heard on the merits.

C.

It is important to note the specifics of the evidence submitted by Hubbard at various
junctures in his state and federal post-conviction proceedings. The following summaries are
taken verbatim from the preceding district court opinion, organized chronologically.1

• 1/2/2008 affidavit of Elton Carter:
Carter asserts that Collins told him that his September 2, 1992 trial testimony was
coerced and that the police threatened to charge him if he did not agree to so
testify. Carter avers that Collins told him that he wasn’t at the crime scene.
Collins revealed this information after the petitioner was found guilty. There are
two dates on this affidavit: 1/28/04 (the date of the affiant’s subscription) and
1/2/08 (the date of the notary’s certification).

• 6/25/2009 affidavit of Emanuel Randall:
Randall swears that Collins was not near the crime scene on January 17, 1992.
Randall avers that Collins was with him and Raymond Williams playing a dice
game when the men received a call that someone had been killed. Randall also
states that the “word on the street” was that Mark Goings killed the victim,
Rodnell Penn, because Goings believed that Penn had killed Goings’[s] brother a
few weeks earlier. Randall stated that no one understood why Collins would
falsely accuse the petitioner of shooting the victim.

1
Op. & Order, R. 66, PageID 4179–82 (discussing Carter Aff., R. 1, PageID 72–73; R. 26, PageID 1474–
75; Randall Aff., R. 1, PageID 64–65; R. 26, PageID 1459–60; Hill Aff., R. 1, PageID 57–59; R. 26, PageID 1451–
53; Williams First Aff., R. 1, PageID 67–68; R. 26, PageID 1462–63; Burford Aff., R. 1, PageID 61–62; R. 26,
PageID 1471–72; Williams Second Aff., R. 1, PageID 70; R. 26, PageID 1465; Hubbard Decl., R. 1, PageID 75; R.
26, PageID 1474–75; Samir and Raad Konja Affs., R. 26, PageID 1477, 1479; Checker Subpoena, R. 51, PageID
2309–10, 2312–17; Collins Aff., R. 51, PageID 2253–54; Collins Polygraph, R. 51, PageID 2306–07).
No. 21-2968 Hubbard v. Rewerts Page 29

• 2/1/2011 affidavit of Askia Hill:
Hill avers that he saw Mark Goings shoot the victim on January 17, 1992, but that
he never told anyone because he feared for his life. Hill did not know the
petitioner but had seen him in the neighborhood. Hill indicates that the petitioner
was not the shooter.

• 5/23/2011 affidavit of Raymond Williams:
Williams swears that while being detained at the Detroit Police Homicide Section
between August 31 and September 2, 1992, he overheard Collins crying in his
nearby cell. When Williams asked Collins what was wrong, Collins indicated that
two police officers made him testify falsely against the petitioner at his trial on
September 2, 1992. Williams advised Collins not to lie because the men weren’t
near the crime scene that night. Collins informed Williams that if he did not
implicate the petitioner, the police would charge him with the murder. Williams
never told anyone about what Collins had told him while in lockup until he
contacted the petitioner in late 2010 and 2011.

• 9/8/2011 affidavit of Roy Burford:
Burford avers that he was at the Special K party store on January 17, 1992, where
the shooting took place, from 6:00 p.m. until closing and that at one point the
store owner called police. Burford states he saw neither Curtis Collins nor the
petitioner that night in the store or near it, that Collins informed Burford that he
“lied on” the petitioner because the petitioner robbed him in 1986, and that after
the petitioner was convicted, people were saying that Mark Goings was the actual
killer.

• 1/9/2012 second affidavit of Raymond Williams:
Williams avers in his second affidavit that on October 2, 2011, he discussed the
case with the Special K party store owner, Steve Konja, who informed Williams
that never saw Collins in the store on January 17, 1992.

• 10/22/2013 unsworn declaration of Hubbard:

[Hubbard] asserts that:
• He was unaware of the contents of Hill’s affidavit until January 2011,
when they had a random and unplanned encounter while incarcerated.
• He was unaware of the contents of Burford’s affidavit until August 2011,
when they had a chance encounter while incarcerated, and in October
2011, when Burford talked to Steve Konja.
• He was unaware of the contents of Randall’s affidavit until June 2009,
when they had a chance encounter while incarcerated.
No. 21-2968 Hubbard v. Rewerts Page 30

• He was unaware of the contents of Carter’s affidavit until January 2004,
when Carter wrote to him.
• It was only through further conversation with Raymond Williams that the
petitioner was able to get Williams’ second affidavit.

• 7/28/2014 affidavits of Raad and Samir Konja:
[Samir] Konja indicates that he was a co-owner of the Special K Party Store on
January 17, 2014. Konja states that neither he nor his brothers permitted Collins
in the store because of problems they had with him. Konja indicates that he was
never spoken to by the police or the prosecutor[.] Raad Konja’s affidavit mirrors
that of his brother except he also indicates that Collins was not in the store on the
night of the murder.

• 1/14/2016 request to subpoena records from the Checker Cab Company:
The petitioner has evidence that the assistant prosecutor in his case requested
Sergeant Kinney to subpoena the records from the Checker Cab Company to
attempt to corroborate whether Collins, in fact, took a taxicab from the location of
the shooting. The petitioner only discovered the existence of the subpoena after
filing a Freedom of Information Act request and receiving the information on
January 14, 2016.

• 10/31/2017 affidavit of Curtis Collins:
Curtis Collins signed an affidavit averring that he was not present anywhere near
the Special K Party Store on January 17, 1992. Collins avers that he did not see
the petitioner fleeing from where the victim was found shot to death. Collins
claims that Sergeant Kinney forced him to testify falsely at the petitioner’s
preliminary examination that he saw the petitioner fleeing the murder scene.
Collins states he testified truthfully on the first day of trial. He says that he spent
two days at the 1300 Precinct [Detroit Police Department Headquarters] where he
was threatened by Sergeant Kinney and Sergeant Gale that he would be charged
with the victim’s murder if he didn’t implicate the petitioner. Collins also states
that as a result of this coercion, he returned on the third day of the petitioner’s
trial and falsely implicated him in the victim’s murder. Collins avers that he
contacted Raymond Williams in 2014 and told him that he had perjured himself at
the petitioner’s trial but only did so because the assistant prosecutor and the police
had threatened him and he did not want to go to jail for perjury. Collins told
Williams he would sign an affidavit to that effect. Collins went to prison in 2014
and 2015 and during the nine months there, realized how difficult prison was.
Upon his release from prison, Collins learned that the assistant prosecutor in the
petitioner’s case was no longer working and the police on his case were now
retired. Collins said he no longer had to worry about threats from these
 No. 21-2968 Hubbard v. Rewerts Page 31

individuals to prosecute him and was “tired from running from the fact that he had
put an innocent man, Carl Hubbard, in prison.”

• 2/21/2018 polygraph examination of Curtis Collins:
Petitioner attached as an exhibit to his second amended petition the results of a
polygraph examination performed by Michael Anthony on Curtis Collins on
February 21, 2018. Anthony asked the following questions of Collins:

1. Did you see Carl Hubbard shoot that man?

Answer: No

2. Did you see Carl Hubbard shoot anyone at Gray and Mack in January of 1992?

Answer: No

3. Were you present when Carl Hubbard shot that man?

Answer: No

Anthony opined that Collins was being truthful regarding his answers to these questions.

III.

With the proper context in mind, I now turn back to what has come to be known as the
Schlup standard: whether Hubbard has presented new reliable evidence that makes it “more
likely than not that no reasonable juror would have convicted [him]” based on all the evidence,
old and new. Schlup, 513 U.S. at 327–28.

A.

Because Schlup instructs that additional evidence of actual innocence must be new and
reliable before it can be considered, 513 U.S. at 324, I examine both characteristics before
considering the probabilistic question.

1.

In this circuit, evidence is new so long as it was never presented at trial. Souter, 395 F.3d
at 595 n.9 (quoting Schlup, 513 U.S. at 324). Some aspects of newness—notably, age and timing
of the evidence’s submission—factor into reliability. Freeman v. Trombley, 483 F. App’x 51, 57
 No. 21-2968 Hubbard v. Rewerts Page 32

(6th Cir. 2012) (citing House, 547 U.S. at 537). Whether the evidence is “new” is not at issue
here, as both the district court and the government focus their analyses on the reliability of the
new evidence.

True, the government refers to the evidence as “purportedly new,” which is hardly a
concession. But the government provides no rationale as to why any of the additional evidence
is not new, instead focusing on the evidence’s reliability. The district court analyzed the
additional evidence in more detail in its timeliness inquiry, as Hubbard argued that “[AEDPA’s
one-year] limitations period should not start when his conviction became final, because he has
newly discovered evidence [challenging his conviction].” (Op. & Order, R. 66, PageID 4178.)
This analysis has no bearing on the question of newness, as timeliness turns on when Hubbard
discovered the “factual predicate” for his underlying claims, not when he discovered the actual
pieces of evidence, as is relevant for equitable tolling. Compare § 2244(d)(1)(D) (referring to
“the date on which the factual predicate of the claim or claims presented could have been
discovered through the exercise of due diligence” (emphasis added)), with Cleveland, 693 F.3d at
636 (“Under the district court’s reasoning, Green’s affidavit was ‘available’ to Lloyd at the time
of his trial. However, this had no effect on the Court’s determination that the information
contained in Green’s affidavit constituted new evidence.” (emphasis added)).

Accepting the evidence as new, I move on to its reliability.

2.

Our circuit does not appear to have one succinct definition of reliability, but a few
characteristics warrant discussion based on the evidence at play in this case.

While Schlup provided three examples of presumably reliable evidence when announcing
the rule—“exculpatory scientific evidence, trustworthy eyewitness accounts, or critical physical
evidence,” 513 U.S. at 324—we have since emphasized that “the examples following the words
‘new reliable evidence’ [in Schlup] were not meant to be an exhaustive list of everything upon
which an actual innocence claim may be based,” Souter, 395 F.3d at 593 n.8. Hubbard correctly
asserts as much, quoting an out-of-circuit case to articulate that “there are no categorical limits
 No. 21-2968 Hubbard v. Rewerts Page 33

on the type of evidence that can be offered under Schlup.” (Appellant Br. 28 (quotations
omitted) (quoting Howell v. Superintendent Albion SCI, 978 F.3d 54, 60 (3d Cir. 2020)).)

Generally speaking, we consider whether the source has an “evident motive to lie,” such
as if the new evidence is proffered without a reward or pressure, and whether the source is
willing to testify to their statements. See House, 547 U.S. at 552; Cleveland, 693 F.3d at 640;
see also Jimenez v. Lilley, No. 16CIV8545, 2017 WL 4535946, at *10 (S.D.N.Y. Oct. 10, 2017),
report and recommendation adopted, 2018 WL 2768644 (S.D.N.Y. June 7, 2018)
(acknowledging that an affiant’s “willingness to testify” live may also support his reliability).
While some courts subject statements from “inmates, suspects, or friends or relations of the
accused” to additional scrutiny, evidence from these sources is not per se unreliable. House, 547
U.S. at 552. These statements are instead considered with an eye toward the same reliability
concerns as any other source—concerns which often serve as the basis for the rejection of
statements from “suspicious” sources, inasmuch as the statements are not rejected because of the
source itself. Compare Cleveland, 693 F.3d at 641 (deeming evidence from a petitioner’s
neighbor reliable due to lack of bias), with Milton v. Sec’y, Dep’t of Corr., 347 F. App’x 528,
531 (11th Cir. 2009) (discounting evidence due to relationship and unexplained delay).

An individual’s status as a recanting witness is relevant to—and can detract from—their
reliability. United States v. Chambers, 944 F.2d 1253, 1264 (6th Cir. 1991); see also Byrd v.
Collins, 209 F.3d 486, 508 n.16 (6th Cir. 2000). But recantation status alone is insufficient to
render otherwise convincing or helpful evidence unreliable. Cleveland, 693 F.3d at 640 (finding
a recantation to be credible); see also Fairman v. Anderson, 188 F.3d 635, 646–47 (5th Cir.
1999) (“While Prewitt’s status as a recanting witness detracts from the credibility of his new
testimony . . . it is not a bar to the acceptance of such testimony. Indeed, . . . Prewitt proffered a
convincing reason for his recanting affidavit: the prosecution coerced him to lie at Fairman’s trial
by threatening to charge him with murdering Jones.” (internal citation and parenthetical
omitted)). In fact, the circumstances surrounding a recantation can make it more credible than
one’s prior inconsistent statements, such as if a petitioner does not subsequently withdraw the
currently asserted testimony. See Cleveland, 693 F.3d at 640.
 No. 21-2968 Hubbard v. Rewerts Page 34

And the same is true for the passage of time, which we have explicitly rejected as a
standalone basis on which to deem an affidavit unreliable. Id. at 641. “[A] federal habeas court,
faced with an actual-innocence gateway claim, should count unjustifiable delay on a habeas
petitioner’s part, not as an absolute barrier to relief, but as a factor in determining whether actual
innocence has been reliably shown.” McQuiggin, 569 U.S. at 387 (emphasis added). So, we
look to see if the affiant provided a reasonable justification for failing to come forward earlier.
Cleveland, 693 F.3d at 640–41. “Reasonable” is not an insurmountable bar: We have deemed
lack of appreciation for the importance of one’s statements to a case and “personal issues during
[the relevant] period” as adequate justifications for waiting to come forward, particularly where
that person has no reason to lie. Id.; see also House, 547 U.S. at 552. On the other hand, we
view “tactical maneuvers,” like “waiting to provide the statement until” the declarant “was no
longer subject to further punishment” for their actions in the charged crime, as “highly
suspect.” In re Byrd, 269 F.3d 561, 574 (6th Cir. 2001).

All of this notwithstanding, where neither the government nor the district court attacks a
piece of evidence’s reliability, the appellate court can proceed on the petitioner’s assertions of
reliability. See Cleveland, 693 F.3d at 638. As neither the district court nor the government
makes any argument attacking the reliability of the affidavits from the Konjas, the party store
owners, I presume these are reliable. This is important, as the Konjas provide yet another first-
hand account corroborating Collins’s recantation, explaining that Collins both was not and could
not have been in the party store the night of Penn’s murder: Not only do they own the store,
were working that night, and did not see Collins near the store that night, but they had banned
Collins from the store, so he was not allowed in.

With these principles laid out, I analyze each piece of evidence’s reliability. Schlup, 513
U.S. at 327–28.

Collins. I review Collins’s affidavit with the importance of his trial testimony to
Hubbard’s conviction in mind.

The existence of a delay in coming forward on its own does not render an affidavit
unreliable. Cleveland, 93 F.3d at 641. Instead, the inquiry takes into account all of the case-by-
 No. 21-2968 Hubbard v. Rewerts Page 35

case situations in which long-belated affidavits factor into a petitioner’s actual innocence claims.
See, e.g., Cleveland, 693 F.3d at 639–40 (crediting a recanting affidavit offered 15 years after
trial); Howell, 978 F.3d at 60–61 (granting evidentiary hearing based on recantations three
decades later); Arnold v. Dittmann, 901 F.3d 830, 838–39 (7th Cir. 2018) (remanding for an
evidentiary hearing to determine a recantation’s reliability); Teleguz v. Pearson, 689 F.3d 322,
331–32 (4th Cir. 2012) (remanding for consideration of an actual-innocence claim based on new
evidence, including recantation affidavits).

Crucially, Collins’s delay was not unexplained. Collins’s fear of the involved
prosecution and police—who, as Collins asserts, threatened him into incriminating Hubbard—
abated only after he learned both were no longer working on Hubbard’s case. (Collins Aff., R.
51, PageID 2253–54.) This was not until after his release from prison in 2015, during which his
guilt over “put[ting] an innocent man, Carl Hubbard, in prison” grew. (Id.) It is unfair to
characterize Collins’s delay as 25 years without considering the fact-specific context of his
delay, including well-documented and corroborated fear of the prosecution for all of those years.
Other circuits have explicitly found factually similar coercion—threats to charge the witness
with the murder—to be an adequate justification for delay. See, e.g., Fairman, 188 F.3d at 646–
47 (“While Prewitt’s status as a recanting witness detracts from the credibility of his new
testimony . . . it is not a bar to the acceptance of such testimony. Indeed, . . . Prewitt proffered a
convincing reason for his recanting affidavit: the prosecution coerced him to lie at Fairman’s trial
by threatening to charge him with murdering Jones.”). The same is true here.

Taking all of the circumstances together, Collins is not closely aligned with the
petitioner, has an adequate explanation for his delay in coming forward, and received no benefit
from testifying in a way that favors the petitioner. In fact, regarding this latter point, the opposite
is true: Collins faced punishment for testifying about his recollection of the evening that favored
Hubbard in the form of a perjury charge—a charge that was dropped only after he was recalled
and incriminated the petitioner—and Collins was given a deal by which he would remain on
parole or probation if he testified against the petitioner. To the extent the government refers to
“Collins’s most recent recantation” as “yet another change in his story,” the same would be true
of Collins’s day three testimony, which recanted his testimony from the first day of trial. Yet the
 No. 21-2968 Hubbard v. Rewerts Page 36

prosecution had no issues accepting Collins’s “flipped” testimony when it supported their goal:
Hubbard’s conviction. I am otherwise unaware of a bright line rule that renders his third story
reliable but his fourth unreliable. Contra Maj. Op. at 18.

Our review is de novo, but the district court’s analysis of Collins’s affidavits bears
mentioning, as the district court discounted Collins’s affidavit for a different reason than the
majority: the length of time between trial and submission of his affidavit. In support of its
finding that this delay bars Collins’s most recent affidavit from holding weight, the district court
relied on two cases. But as the affidavits in those cases could not be relied on due to other
indicia of unreliability, those cases do not seal the fate of Collins’s affidavit.

First, in Lewis v. Smith, we accepted the district court’s conclusion that the petitioner’s
girlfriend’s decision to come forward with an undated recanting affidavit two years post-trial was
not “cause” to excuse the party’s failure to raise the issue in state court. 100 F. App’x. 351, 355
(6th Cir. 2004). In so finding, we relied on dicta explaining this court’s “suspicion of
exculpatory affidavits submitted by someone closely aligned with a defendant.” Id. (emphasis
added) (citing United States v. Willis, 257 F.3d 636, 647 (6th Cir. 2001)). It was not the delay,
on its own, that led to the affidavit’s demise; rather, the relationship between the petitioner and
affiant set it over the edge.

Then, in Strayhorn v. Booker, we rejected an affidavit that “d[id] not explain why the
witnesses waited nearly two years after petitioner’s trial to come forward” and that came from a
source who “received the supposed benefit from testifying against petitioner—leniency with
respect to his own crimes—and had nothing to lose by recanting.” 718 F. Supp. 2d 846, 874
(E.D. Mich. 2010). So, that affiant’s statements were unreliable due to a combination of three
things: the delay, the lack of explanation for the delay, and the benefit received from testifying
in the way they seek to now.

Because Collins’s affidavit is distinguishable from both of the above cases, and for the
reasons stated above, I find Collins’s affidavit sufficiently reliable.

Hill, Burford, and Randall. The majority then writes off the Hill, Burford, and Randall
affidavits as “patently unreliable.” Maj. Op. at 17. I disagree.
 No. 21-2968 Hubbard v. Rewerts Page 37

The district court characterized Hill’s affidavit as Hubbard’s main proof of innocence,
because Hill testifies that he saw Mark Goings shoot Penn. The district court says Hill’s
incarceration with Hubbard and Hill’s delay in coming forward “cast a pall of suspicion” on his
affidavit. (Op. & Order, R. 66, PageID 4188.)

As to Hill’s delay, Hill explained that he “never told anybody” what he witnessed
because he lived in the same neighborhood where the shooting took place and so “was afraid for
[his] life.” (Hill Aff., R. 51, PageID 2238.) His delay, then, is not unjustified. See McQuiggin,
569 U.S. at 387. This delay is just as rational as not knowing the importance of one’s testimony
to a case or a “personal issue[],” see Cleveland, 693 F.3d at 641, and is much more than a desire
to not “get involved,” which has been labeled an inadequate justification for delay, Milton, 347
F. App’x at 531 (citing Schlup, 513 U.S. at 332). And while Hill was incarcerated with Hubbard,
there is no evidence that he has a motive to lie: He testified that he does not know Hubbard
“personally.” So his affidavit does not have “the same risk of bias as an affidavit made by close
friends or relations of [Hubbard].” Cleveland, 693 F.3d at 641; see, e.g., Freeman, 483 F. App’x
at 60 (discrediting an exculpatory affidavit where the affiant—the petitioner’s girlfriend—was
pregnant with the petitioner’s child at the time of the murder and continued to co-parent with him
at the time of her affidavit).

What is more, with Collins’s damaging testimony effectively rendered null and void, Hill
is not simply a “possible eyewitness . . . among purported eyewitnesses.” (Appellee Br. 37.) He
is, at this point, the only alleged eyewitness. The government attempts to discredit Hill’s
eyewitness account by pointing to an alleged lack of cross-corroboration of Hill’s “presence near
the scene of the crime back in 1992. For example, none of the other witnesses, new or old, have
testified that they saw Hill or Goings near or at the scene of the crime.” (Appellee Br. 37.) But
the government cannot have its cake and eat it too when it comes to the reliability of evidence:
Collins is willing to testify that he was not at the scene of the crime, removing the prosecution’s
only key witness. Yet the government clings to the reliability of Collins’s testimony to the
contrary to counter Hill’s eyewitness testimony. To the extent that the government secured
Hubbard’s conviction based on evidence from one witness’s account when that witness’s
presence at the scene was not corroborated, it makes sense to consider another witness’s account
 No. 21-2968 Hubbard v. Rewerts Page 38

who claims to have actually witnessed the murder even if his presence is not independently
corroborated. Given the circumstances surrounding Hill’s affidavit, I find it to be reliable.

As to Burford and Randall, substantively, their affidavits are not entirely “word on the
street.” For example, Randall provides corroborated personal knowledge that Collins was not on
Gray Street the night of the murder because he was with him that night playing dice. Randall is
one of many people to negate the idea that Collins was at the murder scene, including a
consistent corroborating account that Randall, Williams, and Collins were together. This cuts in
favor of the reliability of the notion that Collins was in fact not at the scene of the crime, thereby
weakening the prosecution’s case.

Some of the “word on the street,” including Randall’s assertion that he heard that “Mark
Going [sic] kill[ed]” Penn, cannot be discredited due to its hearsay nature alone. (Randall Aff.,
R. 51, PageID 2321–22.); House, 547 U.S. at 537–38 (instructing habeas courts to consider all
evidence without regard for the rules of admissibility). Multiple individuals assert that Mark
Goings is responsible for Penn’s death for the same reason, so those reports cross-corroborate
each other, thereby increasing each statement’s reliability. See Goodwin, 552 F. App’x at 547.
And having a cross-corroborated suspect becomes even more important where the prosecution’s
only key witness connecting Hubbard to Penn’s death has been increasingly called into question.

In rejecting Burford’s affidavit as unreliable hearsay, the district focused on only one
aspect of his testimony—that community members believe Mark Goings to be the actual killer.
But this overlooks—and does not call into question—the more significant, novel aspect of
Burford’s affidavit: his personal knowledge that he was at the party store the night of Penn’s
murder and that neither Hubbard nor Collins were there at any point.

While neither affidavit provides a reason for their delay, the record lacks information
about any bias, reward, or reason to lie, beyond that Randall and Collins knew each other, which
alone is insufficient to undermine his otherwise reliable statements. And Randall’s accounts are
cross-corroborated, which weighs in favor of the statement’s reliability. The same is true of
Burford’s assertion that Collins “lied on” Hubbard, as it is cross-corroborated and is relevant to
the extent that this is further proof that Collins did, in fact, lie when implicating Hubbard. Given
 No. 21-2968 Hubbard v. Rewerts Page 39

the corroborated nature of their statements, I find Burford and Randall’s affidavits to be reliable
and important.

Other Evidence. The majority discards much of the other new evidence—including, at
least, Williams’s and Carter’s affidavits claiming personal admissions from Collins that his
testimony was untruthful, Collins’s polygraph, the FOIA request for Collins’s alleged cab ride
receipts, and the DOJ investigation implicating Sergeant Kinney for coercing false testimony—
as “weak,” based on the delay of some and lack of explanation for these delays for some, and as
not being independently “exonerating.” Maj. Op. at 17–19.

But these pieces of evidence are also reliable. First, Williams’s delay is not unexplained.
He did not understand the importance of his testimony until Hubbard contacted him, after which
there was hardly a delay. Second, Sergeant Kinney’s rampant coercion became more important
once Hubbard had more proof that Collins was, in fact, coerced, which occurred once they were
imprisoned together years after the fact. Such a delay is justified. The same goes for the cab
records, which Hubbard did not have access to until early 2016.

Regarding the polygraph, the district court made a logical jump that because polygraph
examinations are not considered admissible or scientifically valid in Michigan state courts, they
are not “exculpatory evidence” and are therefore not “new reliable evidence.” (Op. and Order,
R. 6,, PageID 4191–92.) This is flawed for more than one reason. First, the Supreme Court
instructs lower courts to consider all evidence without regard to whether it would necessarily be
admitted under “rules of admissibility that would govern at trial.” House, 547 U.S. at 537–38.
Second, evidence does not need to fall under any of Schlup’s three example categories to be
considered new reliable evidence, as that list was “not meant to be [] exhaustive.” Souter, 395
F.3d at 593 n.8.

More generally, evidence need not, on its own, exonerate a petitioner in order to be
reliable, so long as it generally supports a claim of innocence. Cleveland, 693 F.3d at 633.
Corroborated evidence that the State’s only witness connecting Hubbard to Penn’s murder lied
about multiple aspects of his testimony and was not anywhere near the scene goes towards
Hubbard’s innocence. Imagine a Jenga tower. The tower might stand strong as one block is
 No. 21-2968 Hubbard v. Rewerts Page 40

removed. And another. But when a crucial block is removed, the tower tumbles—and the player
loses. Collins’s testimony is this crucial block to the State’s case. Without it, the State loses.
The further evidence that places Collins elsewhere represents each tap to this crucial block. It is
not each tap on its own that knocks this block out of place, but rather the aggregate impact of
all—as is the case with this evidence.

Underlying the majority’s brushing aside of this evidence appears to be the sentiment that
much of this evidence is provided only to undercut the validity of Collins’s most recent affidavit
and so the evidence is duplicative or cumulative in some way—as the district court touched on
and the government argued. (See Appellee Br. 42, 31 n.3; Op. & Order, R. 66, PageID 4186.)
But whether evidence is cumulative or duplicative is more relevant to a statute of limitations
“factual predicate” argument, which Hubbard waived. See § 2244(b)(2)(B)(i). Evidence is
“redundant” for purposes of Hubbard’s waived timeliness argument if it is “additional
information about an issue that Hubbard was aware of several years earlier” even if that specific
piece of evidence had not been presented before. (Op. & Order, R. 66, PageID 4184 (quoting
Jefferson v. United States, 730 F.3d 537, 547 (6th Cir. 2013)).) But as to equitable tolling,
additional information about a known issue functions as cross-corroboration, supporting a
source’s credibility and the evidence’s reliability, particularly where multiple parties present the
same information. United States v. Goodwin, 552 F. App’x 541, 547 (6th Cir. 2014); see also
United States v. Leppert, 408 F.3d 1039, 1041 (8th Cir. 2005) (finding that “the cross-
corroboration of some details of the statements of [the informant] and the [confidential
informant] supports the reliability of [the informant’s] statements as a whole”). “By telling
consistent yet independent stories, informants provide cross-corroboration, and enhance the
reliability” of the information. Goodwin, 552 F. App’x at 547 (cleaned up) (quoting United
States v. Lancaster, 145 F. App’x 508, 510 (6th Cir. 2005)).

Beyond cross-corroboration, even where evidence makes similar points, new evidence is
not to be brushed aside as “cumulative” where the evidence “do[es] not merely add to the
defense, but also deduct[s] from the prosecution.” Souter, 395 F.3d at 593. The “cumulative”
nature of the evidence is relevant, then, only to the extent that it piles on to make one
interpretation of the facts more likely than a lesser supported interpretation. Hubbard’s case
 No. 21-2968 Hubbard v. Rewerts Page 41

tracks Souter in this regard: In both cases, the petitioner’s new evidence—evidence
underpinning a pre-existing factual predicate—called the prosecution’s key evidence into
question (the bottle in Souter and Collins’s testimony here). Id. Evidence supporting the fact
that Collins was indeed not near the party store the night of the murder—a pre-existing factual
predicate—takes on a new significance where Collins’s own account of his proximity to the
murder has changed. See id. The stronger the case against the government’s portrayal of
Collins, the more the evidence “deduct[s] from the prosecution.” In this way, whether evidence,
like Williams’s and Carter’s affidavits, is duplicative is irrelevant to its reliability. Here, this
evidence is reliable, which ends our inquiry (for now).

B.

Having established that Hubbard submitted new reliable evidence in support of his claim
of actual innocence, I turn to the probabilistic inquiry: whether “any reasonable juror would
have reasonable doubt” about Hubbard’s conviction based on “all the evidence, old and new,
incriminating and exculpatory[.]” House, 547 U.S. at 538 (cleaned up) (citing Schlup, 513 U.S.
at 327–28); see supra pp. 1–5. While “the Schlup standard is demanding,” it “does not require
absolute certainty about the petitioner’s guilt or innocence.” House, 547 U.S. at 538; see Keith,
2023 78 F.4th at 315; contra Maj. Op. at 2 (“[H]e fails to present evidence affirmatively
demonstrating his actual innocence; he cannot prove that he did not, in fact, commit murder.”),
Maj. Op. at 9 (“Rather, he must demonstrate that he factually did not commit the crime.”).

Not only does the majority err in requiring factual innocence, but it also errs in its
isolated consideration of the evidence. The majority’s discussion of the evidence’s reliability
appears to answer the Schlup probability question based on each piece of evidence individually.
For example, the opinion concludes that Hubbard’s alibi (from Thomas and Vanessa Spells)
“falls short” of Schlup’s probabilistic burden. Maj. Op. at 18. But we have been instructed to
evaluate the evidence in totality. House, 547 U.S. at 538; Keith, 78 F.4th at 317.

To put this standard into perspective, the Souter court granted relief in a somewhat
analogous situation: where only one piece of evidence—there, a bottle; here, Collins’s day three
testimony—directly tied the petitioner to the victim’s death. 395 F.3d at 596–97. Once the
 No. 21-2968 Hubbard v. Rewerts Page 42

bottle was called into question by new reliable evidence, other evidence—new and old—became
more relevant and more important, undermining the prosecution’s original case and conviction.
The same was true in House, where the “central forensic proof connecting [the petitioner] to the
crime” was “called into question” based on the “substantial evidence” to the contrary. 547 U.S.
at 554.

And the same is true here. Multiple aspects of Collins’s day three testimony are less than
plausible—or even implausible—based on the record as a whole at this point. This is even more
pronounced when considering the new reliable evidence, which bolsters support for the veracity
of Collins’s day one testimony that he reasserts as the truth and to which he is willing to testify
and be cross-examined. As discussed in the reliability analysis, the district court’s main
contentions with Hubbard’s evidence are that it is cumulative, hearsay, and from someone who
was incarcerated at some point.

But none of these challenges negates the aggregate impact of Hubbard’s new reliable
evidence, as we are required to evaluate. See House, 547 U.S. at 538; Keith, 78 F.4th at 317. If
Hubbard’s trial had included all of the evidence in the record before us now, it is likely that
Collins would repeat his day one testimony, so Hubbard is not connected to the scene at the time
of the murder, and Hill would testify that he witnessed Mark Goings shoot Penn. When you
remove Collins from the scene of Penn’s murder, as the evidence would indicate, and add in a
more-corroborated suspect, the case against Hubbard seemingly falls apart.

This is true even if the evidence does not establish that Goings shot Penn, because the
question at this stage is not whether Hubbard can construct a fail-proof case against an
exonerating suspect. Contra Maj. Op. at 17–20. The question we are required to answer is
whether “any reasonable juror would have reasonable doubt.” House, 547 U.S. at 538. This
inquiry remains the same regardless of whether we label “McQuiggin’s and House’s standard of
proof—that considering the evidence as a whole, ‘no reasonable juror would have convicted [the
petitioner]’—as the Supreme Court’s definition of actual innocence” or “its means of assessing
actual innocence.” Maj. Op. at 12; see Hilliard v. United States, 157 F.3d 444, 450–51
(interpreting Bousley v. United States, 523 U.S. 614, 623 (1998)).
 No. 21-2968 Hubbard v. Rewerts Page 43

Applying Bousley here, “the focus of our inquiry is limited to whether no reasonable
juror would have otherwise concluded that” Hubbard shot Penn. Hilliard, 157 F.3d at 450–51.
Hubbard’s new evidence, taken with the old, unequivocally “chips away at the rather slim
circumstantial evidence upon which [Hubbard] was convicted,” Souter, 395 F.3d at 592 (cleaned
up), thereby instilling “reasonable doubt” in “any reasonable juror,” House, 547 U.S. at 538.
This doubt may not be the result of an “exonerating” witness in Hubbard’s case. But as is the
case at trial, reasonable doubt is doubt “based on reason which arises from the evidence or lack
of evidence.” Jackson v. Virginia, 443 U.S. 306, 317 n.9 (1979) (emphasis added). In the same
way an exonerating witness could, surely, be the source of a reasonable juror’s reasonable doubt,
so, too, can the lack of incriminating witness when taken with all of the other evidence. And
here, any reasonable juror would doubt a conviction based on such a shaky foundation that all
has since collapsed: where the new evidence impeaches the key parts of the old evidence, the
State’s key witness has been discredited, and an unquestioned suspect is at play. Such doubt
demands relief.

I need not decide whether I would grant Hubbard relief under the majority’s flawed
recitation of the standard—whether Hubbard “affirmatively demonstrat[ed]” or “prove[d] that he
did not, in fact, commit murder.” Maj. Op. at 2. Certainly, if a petitioner were to prove as much,
they, too, would be “actually innocent.” Nonetheless, under the Schlup standard, including its
interpretation by Keith, Hubbard is “actually innocent”: He has presented new reliable evidence
that, when considered with the record as a whole, undermines confidence in his conviction such
that no reasonable jury would be able to convict him beyond a reasonable doubt. Because he has
therefore made a credible showing of actual innocence, Hubbard is entitled to pass through
AEDPA’s gateway in order to have his constitutional claims heard on the merits. See Cleveland,
693 F.3d at 632.

I believe Keith is rightly and soundly decided, but I can appreciate the frustration that
accompanies being bound to precedent with which you disagree. See, e.g., Nayed v. Garland,
No. 22-4002, 2023 WL 5237458, at *5 (6th Cir. Aug 15, 2023) (Cole, J., concurring).
Nonetheless, “[t]he prior decision remains controlling authority[.]” Salmi v. Sec. of Health and
Human Servs., 774 F.2d 685, 689 (6th Cir. 1985). And under Keith, Hubbard deserves relief.
 No. 21-2968 Hubbard v. Rewerts Page 44

C.

Hubbard alternatively requested an evidentiary hearing in the case that we do not reverse
the district court’s dismissal of his habeas petition. As I find Hubbard to be “actually innocent”
based on the record before us today, an evidentiary hearing is unnecessary. But I understand that
the additional reliable evidence may bring new questions to light. Should one require answers to
these questions before resolving his actual innocence inquiry, there is a remedy for that—a
remedy Hubbard has repeatedly requested: an evidentiary hearing.

“[I]t may frequently be appropriate to require the district court to hold an evidentiary
hearing to enable a procedurally-barred habeas petitioner to develop the factual record necessary
to support equitable tolling under the actual innocence standard.” McSwain, 287 F. App’x at
461–62. “[A] petitioner is due some form of hearing suited to the circumstances, [u]nless the
motion and the files and records of the case conclusively show that the prisoner is entitled to no
relief.” Christopher v. United States, 605 F. App’x 533, 537 (6th Cir. 2015) (emphasis added).

Based on the new reliable evidence, the record cannot be said to “conclusively show that
[Hubbard] is entitled to no relief.” Id. For the reasons explained above, particularly given the
hypocritical treatment of affiants (for example, using Collins’s testimony for conviction but
rejecting Hill’s for equitable tolling), these individuals deserve a chance to testify in court to
have their reliability ascertained once and for all. Many of our sister circuits have granted
evidentiary hearings based on long-delayed affidavits, and I see no reason to not follow suit here.
See, e.g., Howell, 978 F.3d at 60–62 (granting evidentiary hearing based on recantations three
decades later); Dittmann, 901 F.3d at 838–39, 842 (remanding for an evidentiary hearing to
determine a recantation’s reliability); Teleguz, 689 F.3d at 331–32 (remanding for consideration
of an actual-innocence claim based on new evidence, including recantation affidavits).

IV.

For these reasons, I would reverse the district court’s denial of Hubbard’s habeas petition
or, at a minimum, remand for an evidentiary hearing. Therefore, I respectfully dissent.