United States v. Phillip Barnes

RECOMMENDED FOR PUBLICATION

Pursuant to Sixth Circuit I.O.P. 32.1(b) File Name: 25a0226p.06

UNITED STATES COURT OF APPEALS

FOR THE SIXTH CIRCUIT

UNITED STATES OF AMERICA,

Plaintiff-Appellee, v.

NATHANIEL DURHAM (23-5162); PHILLIP BARNES (23-

5173),

Defendants-Appellants. ┐

│

│

│

│

│

│

│

│

┘

Nos. 23-5162/5173 Appeal from the United States District Court for the Western District of Kentucky at Louisville. No. 3:21-cr-00012-3—Benjamin J. Beaton, District Judge.

Decided and Filed: August 14, 2025

Before: MOORE, BUSH, and NALBANDIAN, Circuit Judges.

_________________

COUNSEL

ON BRIEF: Larry D. Simon, Louisville, Kentucky, for Nathaniel Durham. Britt M. Cobb, WILLEY & CHAMBERLAIN LLP, Grand Rapids, Michigan, for Phillip Barnes. Amanda E. Gregory, UNITED STATES ATTORNEY'S OFFICE, Louisville, Kentucky for Appellee. The court issued a PER CURIAM opinion and judgment. BUSH, J. (pg. 12), delivered a separate concurring opinion. MOORE, J. (pp. 13-26), delivered a separate opinion concurring in the judgment in part and dissenting from the judgment in part.

_________________

OPINION

_________________ PER CURIAM. Following then-existing Sixth Circuit precedent, the district court determined Nathaniel Durham and Phillip Barnes were subject to the Armed Career Criminal >

1

Act's mandatory minimum term of imprisonment because each had at least three prior qualifying convictions for offenses that occurred on different occasions. The Supreme Court's intervening decision in Erlinger v. United States, however, made clear that a jury, not a judge, must determine whether crimes occurred on different occasions. 602 U.S. 821 (2024). In this appeal, we must determine whether the district court's failure to submit the different-occasions question to a jury was harmless. Our answer differs between Defendants. For the reasons that follow, we

AFFIRM the district court's judgment as to Durham, VACATE Barnes's sentence, and

REMAND his case for further proceedings.

I.

In 2020, law enforcement arrested Durham and Barnes outside an apartment in Louisville, Kentucky. Both men possessed firearms. Because each had been previously convicted of a crime punishable by more than a year imprisonment, a federal grand jury indicted Durham and Barnes for violating 18 U.S.C. § 922(g)(1). Barnes pleaded guilty, and a jury convicted Durham for the same offense.

A.

This appeal focuses on the sentencing phase of each Defendant's case. At the time of their offenses, federal law generally provided for a punishment of up to ten years' imprisonment.

2

prior convictions qualify as violent felonies or serious drug offenses. See United States v. Taylor, 495 U.S. 575, 600-02 (1990). Second, there must be a finding that at least three of those qualifying offenses were "committed on occasions different from one another." 18 U.S.C § 924(e)(1). Who makes that second finding? The answer has changed over time. For most of ACCA's existence, our precedent required district court judges to decide whether prior offenses were committed on different occasions. See United States v. Burgin, 388 F.3d 177, 182-87 (6th Cir. 2004). But just last year, the Supreme Court held that the Fifth and Sixth Amendments require a jury to make that determination. Erlinger, 602 U.S. at 833-35.

B.

Back to Durham and Barnes. At sentencing, the United States sought to apply ACCA's increased penalty provision to both Defendants. It contended, and the parties do not now dispute, that each Defendant had three or more prior convictions for qualifying offenses. But neither Defendant conceded that their offenses were committed on different occasions. Defendants' respective sentencing hearings occurred before Erlinger. So, following existing Sixth Circuit precedent, the district court concluded that it must decide whether Durham and Barnes committed three offenses on different occasions. And examining the facts of the prior convictions, the court held that each Defendant fell within § 924(e)(1)'s penalty provision. Consistent with the statute's mandatory minimum, the district court ultimately sentenced both Durham and Barnes to fifteen years' imprisonment.

II.

Durham and Barnes timely appealed their sentences. Relying on the intervening decision in Erlinger, both contend that their sentences must be vacated because the district court, not a jury, found that their prior offenses occurred on different occasions. Other subsequent developments, however, complicate matters. Our recent decision in United States v. Campbell held that a district court's pre-Erlinger failure to submit the different-occasions question to a jury is subject to harmless error analysis. 122 F.4th 624, 630-31 (6th Cir. 2024). That means an appellate court may affirm a district court's different-occasions finding if the court can conclude

3

beyond a reasonable doubt that the jury would have found the defendant's offenses occurred on different occasions. See, e.g., United States v. Robinson, 133 F.4th 712, 723-25 (6th Cir. 2025). In this appeal, we therefore must determine whether the district court's failure to submit the different-occasions question to a jury was harmless. We begin by discussing the different- occasions requirement and our precedent applying Campbell's harmless error holding. We then apply those standards to Defendants' prior offenses. We take Durham first, then Barnes.

A.

1.

As we've explained, to trigger § 924(e)(1), a defendant must have at least three previous convictions for offenses that were "committed on occasions different from one another." For much of ACCA's existence, we interpreted that provision to focus on whether offenses occurred simultaneously or were part of the same criminal episode. See, e.g., United States v. Graves, 60 F.3d 1183, 1186-87 (6th Cir. 1995). Our cases typically defined the scope of a criminal episode narrowly. So, for example, we held that two armed robberies of different victims, at different locations, committed thirty minutes apart, were committed on different occasions. United States

v. Brady, 988 F.2d 664, 668-69 (6th Cir. 1993) (en banc). Over time, the federal courts of appeals splintered over the meaning of ACCA's occasions requirement. Some courts focused narrowly on the timing of the offense—whether the offenses occurred "sequentially rather than simultaneously"—while others considered a broader set of facts and circumstances. See Wooden v. United States, 595 U.S. 360, 365 (2022). Enter the case of William Wooden. One evening, he unlawfully entered a one-story storage facility. Id. at 363. Once inside, Wooden and his accomplices broke into ten different storage units, proceeding from unit to unit by crushing the drywall between them. Id. He was convicted of ten burglary offenses for his conduct (one for each unit). Id. In a later ACCA prosecution, we held Wooden's ten burglary offenses occurred on different occasions. United States v. Wooden, 945 F.3d 498, 504-06 (6th Cir. 2019).

4

The Supreme Court granted certiorari in Wooden's case and reversed. See 595 U.S. at 366-76. In doing so, the Court clarified that by "occasion[]," ACCA seeks to identify crimes that are not part of a single "event, occurrence, happening, [] episode," or "uninterrupted course of conduct." Id. at 367, 369. And it adopted a wholistic, multi-factor inquiry to determine whether offenses occurred on different occasions. Id. at 366-69. As we later explained, "[t]hose factors include the timing, the locations, the character, and the relationship of the offenses."

Campbell, 122 F.4th at 629. Wooden itself explained that the first two factors are often dispositive. See 595 U.S. at 369-70 ("In many cases, a single factor—especially of time or place—can decisively differentiate occasions. Courts, for instance, have nearly always treated offenses as occurring on separate occasions if a person committed them a day or more apart, or at a 'significant distance.'" (citation omitted)); see also Campbell, 122 F.4th at 629, 632. But given the right set of facts, other factors can inform the analysis. See, e.g., Wooden, 595 U.S. at 369 ("[T]he character and relationship of the offenses may make a difference: The more similar or intertwined the conduct giving rise to the offenses—the more, for example, they share a common scheme or purpose—the more apt they are to compose one occasion.").²

After Wooden, this court had little difficulty applying its framework to ACCA cases involving retail robberies and residential burglaries. See, e.g., United States v. Williams, 39 F.4th 342, 350-51 (6th Cir. 2022) (concluding robberies that occurred at three or four different retail locations over a month and a half period occurred on different occasions); United States v. Lovell, No. 20-6287, 2023 WL 1879530, at *1-3 (6th Cir. Feb. 10, 2023) (holding the district court did not plainly err in concluding that burglaries targeting homes in numerous locations over more than a month occurred on different occasions).

2 The partial dissent objects to our framing of the remaining Wooden factors. But we fail to see how certain factors, like time and location, can often be "decisive[]," Wooden, 595 U.S. at 370, and yet how other factors will always meaningfully impact the different-occasions analysis, as the partial dissent suggests. After all, Wooden explained that these remaining factors only "may make a difference." Id. at 369 (emphasis added). And if there were any doubt, Campbell confirms our approach when it explained, in the context of harmless error review, that although these other factors can be relevant, "[i]n many instances, however, the analysis can be limited to inquiries like whether the offenses were committed a 'day or more apart' or at 'significant' distances, factors that can singularly decide the question." Campbell, 122 F.4th at 629 (quoting Wooden, 595 U.S. at 370).

5

2.

Erlinger, however, presents a new challenge to cases on direct appeal where the district court made the different-occasions finding. No longer may we ask whether, in our view, a defendant's offenses were committed on different occasions. Instead, we must now ask whether, considering the record in its totality, the United States has proved beyond a reasonable doubt that any reasonable juror would have found that the defendant's offenses occurred on different occasions. United States v. Kimbrough, 138 F.4th 473, 477 (6th Cir. 2025); United States v. Cogdill, 130 F.4th 523, 528 (6th Cir. 2025).³Since Erlinger, we have addressed this question on five separate occasions.

Start with the cases in which we found the Erlinger error harmless. In Robinson, we held that the error was harmless where the relevant offenses occurred years apart, involved different crimes, and were separated by intervening criminal sentences. 133 F.4th at 723-25. Likewise, in Campbell, we affirmed an ACCA sentence despite the Erlinger error because the relevant drug trafficking offenses were separated by two (or three) months and occurred at

"geographically distinct locations." 122 F.4th at 632-33. And in United States v. Thomas, we found the Erlinger error harmless where the relevant offenses occurred years apart, took place in different locations, involved different crimes, and targeted different victims. 142 F.4th 412, 418 (6th Cir. 2025).

In two cases, however, we held the Erlinger error was not harmless. In Cogdill, we vacated an ACCA sentence based on drug offenses that occurred three months apart. 130 F.4th at 529-31. Although the offenses were separated by a similar period of time to those at issue in Campbell, the panel held that "on its own, a gap in time between offenses . . . does not prove beyond a reasonable doubt" that offenses occurred on different occasions. Id. at 530. Essential to the panel's conclusion, however, was the absence of record evidence to support the

3 We are somewhat puzzled by the partial dissent's criticism of our phrasing of the harmless error standard, which we pull directly from the same precedent on which the partial dissent itself quotes. Compare Cogdill, 130 F.4th at 528 ("In this context, harmlessness turns on whether the government proved beyond a reasonable doubt— through 'relevant and reliable information' in the record—that, absent the error, any reasonable jury would have found that [the defendant] committed the prior offenses on different occasions." (citation omitted)), with Opinion of Moore, J., at 2 (quoting same passage).

6

government's evidentiary burden. See Thomas, 142 F.4th at 419 n.1 (noting Cogdill turned, in significant part, on "ambiguities in the factual record" as to "key factual details"). Specifically, the record was silent on the precise location of the offenses and "how the[] crimes took place."

Cogdill, 130 F.4th at 529. This trim record raised a reasonable doubt over whether the crimes occurred on different occasions. Id. at 529-31. Similarly, in Kimbrough, over the dissent of Judge Readler, we vacated an ACCA sentence where the relevant burglary offenses were separated by (at most) nine days, occurred at locations separated by only two miles, involved the same modus operandi, and were not separated by any intervening arrests. 138 F.4th at 477-79. The panel similarly noted that the "record [was] unclear" as to key facts that might cause the offense to have occurred on different occasions. See id. at 478 (noting inconsistencies about the dates of the prior offenses); see also Thomas, 142 F.4th at 419 n.1 (reiterating the centrality of record gaps to Kimbrough's holding).

B.

Against this legal backdrop, we move to Durham's prior offenses and ask whether the Erlinger error was harmless under our precedent. The district court determined he was subject to § 924(e)(1) based on ten burglary offenses that occurred on ten different dates in eleven different locations. For clarity's sake, we divide these offenses into four buckets: 1) four burglaries in October 2009; 2) the burglary of two different homes on October 19, 2010; 3) a burglary on December 15, 2010; and 4) four burglaries between January 4, and January 28, 2011. Durham does not seriously contest that his October 2009 offenses occurred on different occasions than his 2010 and 2011 offenses. So even assuming the four October 2009 burglaries occurred on one "occasion," the question becomes whether at least two of his 2010 and 2011 offenses occurred on different occasions. We conclude it is beyond reasonable doubt that a jury would have found that at least two of the 2010 and 2011 offenses occurred on different occasions.

Take, for example, the two October 2010 burglaries and the January 27, 2011, burglary. The offenses are separated by a gap in time and space that clearly establishes the offenses occurred on different occasions. See Thomas, 142 F.4th at 418; Campbell, 122 F.4th at 629; cf.

7

Wooden, 595 U.S. at 369-70 ("In many cases, a single factor—especially of time or place—can decisively differentiate occasions. Courts . . . have nearly always treated offenses as occurring on separate occasions if a person committed them a day or more apart, or at a 'significant distance.'" (citation omitted)). As to time, more than three months separated the offenses. As to distance, the January 27 burglary occurred more than 15 miles from the houses burglarized in October 2010. What's more, the offenses involved victims with no apparent connection to each other. See, e.g., Thomas, 142 F.4th at 418; United Staes v. Stowell, 82 F.4th 607, 609-11 (8th Cir. 2023) (en banc) (holding Erlinger error harmless where a multi-day gap separated the offenses and the offenses involved different victims).

That makes Durham's offenses more like those at issue in Campbell, where we found the Erlinger error harmless given the months-long gap and significant distance between the relevant offenses. See 122 F.th at 632; see also Williams, 39 F.4th at 350.⁴And it distinguishes Durham's case from Wooden—where the relevant offenses occurred in a single evening at the same location, 595 U.S. at 363—and Kimbrough—where the relevant offenses were separated by (at most) nine days and two miles, 138 F.4th at 478.

Cogdill, on which the partial dissent also relies, is likewise distinguishable. As Thomas explained, Cogdill's harmless error holding turned, in significant part, on "ambiguities in the factual record" as to "key factual details," Thomas, 142 F.4th at 419 n.1, none of which are present here. Unlike Cogdill, the record here demonstrates the precise location of Durham's offenses, a factor that, as we've explained, significantly impacts the different-occasions analysis. The partial dissent also suggests sufficient doubt can be created because the offenses were part of a general scheme to burglarize residences, presumably for a common purpose (financial gain). But we have previously rejected framing these Wooden factors at such a high level of generality, as doing so "would fail to honor [ACCA's] text." See Campbell, 122 F.4th at 632.⁵

4 Although Williams was decided pre-Erlinger, we have relied on its different-occasions analysis in determining whether an Erlinger error is harmless. See, e.g., Campbell, 122 F.4th at 632.

5 The partial dissent's alternative approach results from its failure to keep the focus on the actual statutory language—"committed on occasions different from one another." § 924(e)(1). It instead zeroes in on certain minor definitional factors or semantic features, like "scheme" and "purpose," frames those features at a high level of generality, and then applies them in a manner inconsistent with the objective meaning of the phrase that they help

8

Put simply, no reasonable juror (or speaker of the English language) could conclude Durham's October 19, 2010, and January 27, 2011, offenses occurred on the same occasion. The failure to submit this question to the jury, therefore, was harmless, and we affirm Durham's sentence under ACCA.⁶

C.

We move next to Barnes. The district court determined he was subject to § 924(e)(1) based on four first-degree robbery offenses that occurred on four different dates: August 6, 2008,⁷August 18, 2008, August 31, 2008, and June 11, 2009. In each instance, Barnes robbed a Family Dollar store in Jefferson County, Kentucky. The first two robberies occurred at the same store, while the latter two occurred at different locations.

Barnes does not seriously contest that his August 2008 offenses occurred on a different occasion than his June 2009 offense. The question then becomes whether at least two of his August 2008 offenses occurred on different occasions. Or, more precisely, the question is define. The partial dissent's approach is like saying a married adult male could be reasonably called a "bachelor"

because two definitional elements of "bachelor" are that one is an adult and a male. Or take a more concrete example. Imagine a baseball fan who holds season tickets to see his beloved Boston Red Sox's home games at Fenway Park and religiously attends each game. No reasonable juror could conclude that the game the fan attended on Opening Day occurred on the same occasion as a September series finale against the team's archnemesis, the New York Yankees. That's despite both games being part of the same overall scheme (a baseball season), their occurrence at the same location, the fan's attendance at the games for the same general purpose (enjoying one of America's great pastimes) and the "near-consecutive nature," Opinion of Moore, J., at 14, of the games that the fan also attended in the months between. As we explained in Campbell, in such a setting, relying on such a general connection between the two distinct events "would fail to honor [ACCA's] text."

122 F.4th at 632.

6 In Durham's supplemental brief, he suggests there is a double-jeopardy problem by sentencing him to an enhanced penalty under ACCA. This argument, raised for the first time in supplemental briefing, is not properly before us because "[t]here appears to be no reason why [Durham] could not have made this argument earlier."

United States v. Sain, No. 22-6131, 2025 WL 801366, at *2 (6th Cir. Mar. 13, 2025). Still, the crux of his argument is that after United States v. Bell, the government can't "marshal its evidence against Durham a second time." 37 F.4th 1190 (6th Cir. 2022). But Thomas rejected this exact argument. See 142 F.4th at 420-23. So either way, double jeopardy doesn't preclude our finding that Durham can be subject to an enhanced ACCA penalty.

7 There is some discrepancy in the record as to whether this offense occurred on August 6 or 8. Barnes's supplemental brief represents that the offense occurred on August 6, so we proceed on that assumption.

9

whether we are convinced beyond a reasonable doubt that a jury would conclude at least two of these offenses occurred on different occasions.⁸

The temporal and physical proximity of Barnes's August 2008 offenses make his case closer than Durham's. The offenses are separated by less than a month, and the first two offenses occurred at the same location. That said, the August 6 and 31 offenses were separated by several weeks and more than ten miles.

As an original matter, a court might conclude that no ordinary speaker would say events occurring some twenty-five days apart at locations separated by more than ten miles occurred on the same occasion. See Sent'g Op., R. 155, PageID 814 ("What is not plausible, given ordinary English usage and the interpretation set forth in Wooden, is saying that Barnes robbed three Family Dollars on a single 'occasion' that stretched across 25 days in August 2008."); cf. Kimbrough, 138 F.4th at 480-86 (Readler, J., concurring in part & dissenting in part). But we do not write on a blank slate. And the relationship between Barnes's offenses share some meaningful similarities with the relationship between the offenses at issue in Kimbrough and Cogdill.

Ultimately, we hold that these similarities preclude us from finding that the Erlinger error was harmless. Kimbrough does much of the work. Like the burglaries at issue there, Barnes's robberies involved largely the same modus operandi (stealing cash at gunpoint), 138 F.4th at 478, and though two were separated by some distance, they involved a shared victim, Family Dollar. Barnes's robberies also were not separated by intervening arrests, and the offenses were charged in the same document. The similarities with Kimbrough go on. Like the offenses there,

8 In its supplemental brief, the United States changed its earlier position and argued that a different standard should apply to Barnes's sentence because his conviction resulted from a guilty plea, rather than a trial. Relying on a recent Fourth Circuit decision, the United States now maintains that our harmless error analysis should ask whether the government has "show[n] beyond a reasonable doubt that if [the defendant's] indictment had alleged the 'different occasions' element of ACCA and if [the defendant] had been correctly advised at his plea hearing that he 'was entitled to have a jury resolve'" the different occasions issue, the defendant "would have nonetheless waived that right and admitted as part of his guilty plea that his prior offenses were committed on different occasions."

United States v. Brown, 136 F.4th 87, 97 (4th Cir. 2025) (citations omitted). We need not decide whether Brown's alternative standard should apply in the guilty plea context because we conclude the United States fails to meet its burden under either standard. At his change of plea hearing, Barnes objected to ACCA's application without submitting the different-occasions question to a jury and declined to agree with the United States's characterization of his prior offenses.

10

Barnes pleaded guilty to his offenses on the same day and received concurrent sentences. Id. Although there are some differences,⁹we struggle to see how those distinctions are material here in light of the United States's high burden of proof. And our duty to apply the holdings of prior published opinions also requires us to fairly characterize and faithfully apply the rationale employed to reach those holdings.

III.

For these reasons, we affirm the district court's judgment as to Durham, vacate Barnes's sentence, and remand his case for further proceedings consistent with this opinion.

9 As to Kimbrough, Barnes's offenses were separated by a greater distance (ten miles instead of two miles) and period of time (twenty-five days instead of (at most) nine days). Cogdill differs in other respects. There, the record was silent on where the crimes occurred, so we assumed that the offenses could have occurred at the same location. 130 F.4th at 529. Not so here. And unlike the drug offenses in Cogdill, there is no evidence that Barnes's burglaries were part of an ongoing criminal act or scheme. See Kimbrough, 138 F.4th at 483-84 (Readler, J., concurring in part & dissenting in part) (distinguishing discrete burglaries of separate locations from ongoing drug distribution offenses); Sent'g Op., R. 155, PageID 814 ("[N]othing in the record before us indicates anything connected the burglaries other than the store brand and city.").

11

_________________

CONCURRENCE

_________________ JOHN K. BUSH, Circuit Judge, concurring. Absent circuit precedent, this is an easy case, even as it pertains to Phillip Barnes. No reasonable speaker of the English language would say discrete criminal offenses occurring no less than twenty-five days apart at locations separated by no less than ten miles occurred on the same occasion. The offenses at issue in this case, accordingly, fit squarely within the category of cases in which Wooden v. United States instructed that differences in time and place "decisively differentiate occasions." 595 U.S. 360,

370 (2022).

Although this court's first foray into the Erlinger harmless error thicket, United States v. Campbell, 122 F.4th 624 (6th Cir. 2024), supports this straightforward resolution, two of our subsequent published opinions take it off the table. See United States v. Cogdill, 130 F.4th 523 (6th Cir. 2025); United States v. Kimbrough, 138 F.4th 473 (6th Cir. 2025). Like Judge Readler, I find Cogdill and Kimbrough inconsistent with Wooden and impossible to square with how any reasonable speaker of the English language would understand the concept of different occasions. See Kimbrough, 138 F.4th at 480-86 (Readler, J., concurring in part & dissenting in part). It is even possible that, under one reading of Cogdill and Kimbrough, Campbell arguably should have reached the opposite result.

But we do not approach this case on a clean slate, and I do not see Cogdill and Kimbrough as fitting within the narrow category of cases in which we may look past certain published opinions based on their inconsistency with a prior published opinion. See Counts v. General Motors LLC, 139 F.4th 576, 582 (6th Cir. 2025) ("[I]t takes more than spirited disagreement about a rule's application to particular facts, for one panel lawfully to disregard a published decision of another."). As I see it, the per curiam opinion does its best to fairly characterize, reconcile, and apply all of our prior published opinions in this area. For that reason, I concur in the opinion in full.

12

_____________________________

CONCURRENCE / DISSENT

_____________________________ KAREN NELSON MOORE, Circuit Judge, concurring in the judgment in part and dissenting in part. I do not agree that "no reasonable juror (or speaker of the English language) could conclude" that Durham's prior offenses occurred on the same occasion. Maj. Op. at 9. In any case, the question should be whether a reasonable juror could find reasonable doubt about whether the offenses occurred on different occasions. Nor do I agree with Judge Bush that United States v. Cogdill and United States v. Kimbrough are "inconsistent with Wooden and impossible to square with how any reasonable speaker of the English language would understand the concept of different occasions." Bush Op. at 12. In my view, this assertion represents an attempt to apply a renewed version of the test this circuit used before the Supreme Court struck it down in Wooden v. United States, 595 U.S. 360 (2022), and substitutes the majority's judgment for that of a jury. This court's post-Erlinger harmless-error precedents are completely consistent, are readily distinguishable from one another based on their facts, and support remand in both Durham's and Barnes's cases. I therefore respectfully dissent from the majority's affirmance of Durham's sentence and concur in its judgment as to Barnes's sentence.

I. STANDARD OF REVIEW

Everyone agrees on a few key principles. First, the Armed Career Criminal Act (ACCA) imposes a lengthy mandatory-minimum sentence on defendants who (1) are convicted of being a felon in possession of a firearm under 18 U.S.C. § 922(g), and (2) "ha[ve] three previous convictions . . . for a violent felony or a serious drug offense, or both, committed on occasions different from one another." 18 U.S.C. § 924(e)(1).

Second, pursuant to Erlinger v. United States, in order for such a sentence enhancement under the ACCA to comply with the Fifth and Sixth Amendments, the issue of whether a defendant's prior offenses occurred on at least three different occasions must be submitted to a jury and proven beyond a reasonable doubt (or admitted on the record by the defendant). 602 U.S. 821, 834-35, 840 (2024).

13

Third, under Wooden v. United States, the different-occasions inquiry is "multi-factored."

595 U.S. at 369. "Offenses committed close in time, in an uninterrupted course of conduct, will often count as part of one occasion; not so offenses separated by substantial gaps in time or significant intervening events." Id. "Proximity of location is also important; the further away crimes take place, the less likely they are components of the same criminal event." Id. "And the character and relationship of the offenses may make a difference: The more similar or intertwined the conduct giving rise to the offenses—the more, for example, they share a common scheme or purpose—the more apt they are to compose one occasion." Id. Fourth, based on the foregoing two principles, it was error for the district court to undertake the different-occasions analysis for itself. Fifth, under binding Sixth Circuit precedent, we should now review for harmless error. United States v. Campbell, 122 F.4th 624, 630-31 (6th Cir. 2024).

An error is harmless if the beneficiary of the error, (here, the government) "prove[s] beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained." Chapman v. California, 386 U.S. 18, 24 (1967). Put another way, "[a]n error is harmful if we lack fair assurance 'that the judgment was not substantially swayed by the error.'"

United States v. Grogan, 133 F.4th 553, 563 (6th Cir. 2025) (quoting United States v. Craig, 953 F.3d 898, 906 (6th Cir. 2020)). "In this context, harmlessness turns on whether the government proved beyond a reasonable doubt—through 'relevant and reliable information' in the record— that, absent the error, any reasonable jury would have found that [the defendants] committed the prior offenses on different occasions." United States v. Cogdill, 130 F.4th 523, 528 (6th Cir. 2025) (citing Campbell, 122 F.4th at 630-33). The error is not harmless if there is reasonable doubt about whether, had the different-occasions inquiry been put to the jury, the result would have been the same.

The majority's wording of the harmless-error standard undersells the issue, the government's burden, and the ultimate issue for the jury. See, e.g. Maj. Op. at 9-10 (asserting that "more precisely, the question is whether we are convinced beyond a reasonable doubt that a jury would conclude at least two of these offenses occurred on different occasions"). It would be most precise, however, to posit the question as whether any reasonable jury could find

14

reasonable doubt as to whether the offenses occurred on different occasions.¹The majority ultimately concludes beyond a reasonable doubt that a reasonable jury would necessarily find that Durham's October 19, 2010 and January 27, 2011 offenses were committed on different occasions. Id. at 9. But given Judge Bush's concerns about consistency in our case law, this nuance is important.

Some of our cases, including those on which Campbell relied, have stated the standard this way: "Errors are deemed harmless when 'the reviewing court may confidently say, on the whole record, that the constitutional error was harmless beyond a reasonable doubt.'" United States v. Henderson, 626 F.3d 326, 333 (6th Cir. 2010) (quoting United States v. McGee, 529 F.3d 691, 697 (6th Cir. 2008)); see Campbell, 122 F.4th at 631. This is not incorrect, but it is not particularly helpful. Henderson's framing is akin to saying that an error is harmless only if we are really sure it is harmless. And that statement is somewhat tautological in that it does not explain what it actually means for an error to be harmless. Instead, it speaks only to the degree of certainty we must have in order to affirm on the grounds that an error was harmless. Moreover, under Erlinger, a unanimous jury must itself find, beyond a reasonable doubt, that the offenses were committed on at least three different occasions. Erlinger, 602 U.S. at 840.²This is the nuance the majority elides—we should not ask whether a reasonable jury would find that the offenses occurred on different occasions, but whether a reasonable jury could have a reasonable doubt about whether they did. If just one reasonable juror could find reasonable doubt about whether the offenses were committed on different occasions, there is reasonable doubt about whether the error was harmless and we must reverse. Cogdill, 130 F.4th

1 The majority does not appreciate this nuance. Maj. Op. at 6 n.3. The government must prove to us beyond a reasonable doubt that, had the different-occasions inquiry been put to a jury, the government would have carried its burden to convince every reasonable juror beyond a reasonable doubt. The exacting nature of this standard is a feature, not a bug. See infra pp. 18-20, 23 n.10. Meanwhile, the majority's language—which is specifically quoted above—asks whether a hypothetical reasonable jury would agree with the majority's determination. But just because one reasonable jury might agree with the majority that offenses were committed on different occasions does not prevent another reasonable jury from concluding the opposite on the same facts. And, if those facts would allow the second jury to so conclude, the error is not harmless.

2 Really, because Erlinger requires that the different-occasions inquiry be proven to a unanimous jury beyond a reasonable doubt, 602 U.S. at 840, an error is not harmless if just one reasonable juror on one reasonable jury could find reasonable doubt.

15

at 529 ("To prove that the Erlinger error here was harmless, the government must show, beyond a reasonable doubt, that any rational jury would have found that all three of Cogdill's predicate offenses were 'committed on occasions different from one another.'" (emphasis added) (quoting

18 U.S.C. § 924(e)(1))). I have reasonable doubt about whether a reasonable jury would be compelled to find that either defendant's prior offenses were, beyond a reasonable doubt, committed on at least three different occasions. My conclusion is supported by Wooden and our post-Erlinger case law, which, unlike Judge Bush, I find readily reconcilable based on the different facts at play in each case.

II. POST-ERLINGER HARMLESS-ERROR PRECEDENT

In Campbell and Cogdill, we distilled from Wooden three relevant factors that a jury would consider in determining whether prior offenses were committed on different occasions:

(1) timing, (2) proximity of location, and (3) character and relationship of the offenses. Cogdill,

130 F.4th at 528; accord Campbell, 122 F.4th at 629. In Campbell, we had little trouble determining beyond a reasonable doubt that any reasonable jury necessarily would find that Campbell's prior offenses had occurred on different occasions: "The offenses at issue [were] a Chattanooga drug trafficking offense in August 1992, a Norfolk drug trafficking offense in December 1992, and another Norfolk drug trafficking offense from March 1993." Campbell, 122 F.4th at 632. We noted that months separated each offense and that the first two offenses were "remote as a matter of proximity: eight hundred miles as the bird flies separated the Tennessee offense from those in Virginia." Id. Lastly,

"Campbell's predicates were for wholly distinct crimes." Id. In Robinson, another published post-Erlinger case where we found harmless error, the ACCA predicate offenses were even more distinct from one another: they consisted of "(1) a conviction for a murder in 1991; (2) a conviction for [] possession with the intent to sell hydrocodone in 2013; and (3) a conviction for [] possession with the intent to sell heroin in 2017." United States v. Robinson, 133 F.4th 712, 723 (6th Cir. 2025) (citations omitted). In Robinson "the facts . . . ma[de] it easier for us to find the Erlinger error harmless than the facts in

16

Campbell." Id. at 724. "The latter two drug offenses . . . involved different drugs and took place years apart. A court had also punished and sentenced Robinson for one of the drug offenses before he committed the other one." Id.

By contrast, like the facts in this case, the facts in United States v. Cogdill and United States v. Kimbrough made it much harder to predict whether a reasonable jury would find beyond a reasonable doubt that the predicate offenses there occurred on different occasions. In Cogdill, the harmless-error analysis boiled down to a dispute over whether a reasonable jury would necessarily find beyond a reasonable doubt that two of Cogdill's prior offenses—selling or delivering methamphetamine on June 12, 2014 in Bradley County, Tennessee, and possession with intent to sell or deliver methamphetamine on September 15, 2024 in Bradley County, Tennessee—were committed on different occasions. 130 F.4th at 529-30. True, Cogdill's offenses were separated by three months, id., not so different from the offenses in Campbell, which were committed three and five months apart, respectively, Campbell, 122 F.4th at 632. But the similarities ended there.

Cogdill was convicted of twice trafficking the same drug in the same county in the span of three months. Cogdill, 130 F.4th at 529. We found it relevant that "[n]o one arrested Cogdill between the June offense and the September offense, and he ultimately was convicted of both offenses on the same date." Id. "For all the record show[ed], . . . Cogdill might have obtained a large quantity of methamphetamine, sold some of that stash in June, and then 'was busted' for possessing with intent to sell the rest in September." Id. Given the record at issue in Cogdill, a reasonable jury could conclude that the crimes were "'similar or intertwined' and 'share[d] a common scheme or purpose,'" and were committed on the same occasion despite the gap in time. Id. (alteration in Cogdill) (quoting Wooden, 595 U.S. at 369). Certainly, a reasonable jury could also conclude the opposite, but we could not be sure beyond a reasonable doubt that any reasonable jury would consider the occasions to be different beyond a reasonable doubt. In United States v. Kimbrough, the dispute was centered on two 2016 burglaries in Tennessee. 138 F.4th 473, 475 (6th Cir. 2025). The two burglaries occurred less than three miles apart, "increas[ing] the likelihood that the burglaries were 'components of the same criminal event.'" Id. at 478 (quoting Wooden, 595 U.S. at 369). "[T]he record [was] unclear as

17

to exactly when in 2016 the two burglaries occurred," but indicated that they were both either reported or committed in March 2016. Id. Kimbrough was arrested for both offenses on the same date and "[t]here were no intervening arrests or other 'significant intervening events.'" Id. (quoting Wooden, 595 U.S. at 369). So, too, did Kimbrough plead guilty to both offenses on the same date, and he served concurrent sentences for them. Id. Although with this evidence in mind a reasonable jury could find that the offenses were committed on different occasions, it could also, we held, reasonably find that the offenses were committed on the same occasion, as that inquiry was defined in Wooden. Id. at 479-80. It was the latter conclusion, the former notwithstanding, that compelled reversal in Kimbrough, just as it should here. Contrary to the majority's implication (and Judge Bush's opinion), there is no conflict between Campbell and other cases finding harmless error on the one hand, and Cogdill and Kimbrough on the other. And they are all supported by Wooden. As should be obvious from the above recitation, each panel came to its own conclusion by applying the exact same three factors—the timing, proximity of location, and character and relationship of the offenses—to four different sets of facts. There is nothing unusual or untoward about this. It should go without saying that a factual, context-specific inquiry will come out differently when applied to distinguishable facts; this is a feature, not a bug. And I do not read the majority's or Judge Bush's opinions to contend that the circumstances in Campbell and Robinson, for example, were not in fact distinguishable from those in Cogdill and Kimbrough. In any case, reasonable juries can come out differently when applying such an inquiry to the same facts. See Coopersville Co- op. Creamery Co. v. Lemon, 163 F. 145, 152 (6th Cir. 1908) ("That juries might disagree with one another as to a [factual issue], and so some would be compelled to pay and others escape, may be conceded. But is not this so with respect to many questions which for centuries have gone to the jury?").

When the Supreme Court reversed this court in Wooden, it made exceedingly clear that the different-occasions inquiry was factual and context-specific. See Wooden, 595 U.S. at 369 (relying on the many ordinary meanings of the term "occasion" to hold that the different- occasions inquiry is "multi-factored" and that "a range of circumstances may be relevant to identifying episodes of criminal activity"). And no one can deny that the inquiry's factual nature

18

is exactly why the Court held in Erlinger that the question must be submitted to a jury. Erlinger,

602 U.S. at 834 (holding that there is "no doubt" that the Constitution requires the "fact-laden task" of delineating different occasions to be "resolved by a unanimous jury beyond a reasonable doubt"); see also id. at 828 (acknowledging the "intensely factual nature of th[e] [ACCA occasions] inquiry," which "can require an examination of a 'range' of facts, including whether the defendant's past offenses were 'committed close in time,' whether they were committed near to or far from one another, and whether the offenses were 'similar or intertwined' in purpose and character" (quoting Wooden, 595 U.S. at 369)).

This makes sense; juries, not judges, specialize in factual, context-specific inquiries, and, based on our long historical tradition, they do so nowhere more than in criminal cases. See Duncan v. Louisiana, 391 U.S. 145, 157-58 (1968). "[W]hen juries differ with the result at which the judge would have arrived, it is usually because they are serving some of the very purposes for which they were created and for which they are now employed"—to "'confirm[]'"

"'the truth of every accusation'" by applying "the common-sense judgment of a jury" composed of the defendant's peers, rather than "the more tutored but perhaps less sympathetic reaction of the single judge." Id. at 151-52, 156-57 (quoting 4 W. Blackstone, Commentaries on the Laws of England 349-50 (Cooley ed. 1899)).

Properly framed, the harmless-error inquiry underscores that the government carries a heavy burden to show that the violation of Barnes's and Durham's Fifth and Sixth Amendment rights was harmless. See Eberhart v. Bordenkircher, 605 F.2d 275, 279 (6th Cir. 1979) (holding the government to its "heavy burden" to show that a constitutional error was harmless); Campbell, 122 F.4th at 636 (Davis, J., concurring) (contrasting plain-error review, "where the burden is on the defendant" and which "spring[s] from . . . defense-related negligence," with harmless-error review, where "the government bears the burden of showing that the Erlinger error did not contribute to the challenged outcome because such errors implicate a defendant's trial right guarantee"). As applied, this burden is heavy in part because the government also bears the burden to prove its underlying case beyond a reasonable doubt. Erlinger, 602 U.S. at 834-35, 840. Effectively, the government must prove to us beyond a reasonable doubt that it will prove its case to a jury beyond a reasonable doubt. See id.

19

Accordingly, our application of the standard must give due respect to the jury's function and avoid substituting our judgment for the jury's unless we can be sure beyond a reasonable doubt that every reasonable jury would agree with us. Id. at 842 ("In a free society respectful of the individual, a criminal defendant enjoys the right to hold the government to the burden of proving its case beyond a reasonable doubt to a unanimous jury of his peers 'regardless of how overwhelmin[g]' the evidence may seem to a judge." (quoting Rose v. Clark, 478 U.S. 570, 578 (1986))). For the reasons that follow, I agree with the majority that a reasonable jury could find reasonable doubt about whether Barnes committed his offenses on different occasions. I am not convinced, however, that every reasonable jury would agree, beyond a reasonable doubt, with the majority's conclusions about Durham's offenses. Thus I would vacate both Durham's and Barnes's sentences and remand both cases.

III. BARNES

For Barnes's ACCA enhancement, the government relies on the following four predicate offenses:³two first-degree robberies at a single Family Dollar store on August 8 and 18, 2008, a first-degree robbery at a different Family Dollar location on August 31, 2008, and another first- degree robbery at a third Family Dollar location on June 11, 2009. Appellee Br. at 2. Each Family Dollar was located in the Louisville area. Id.

As the majority wisely concludes, our case law establishes that a reasonable jury, considering the factors identified in Wooden, could find that some or all of Barnes's prior offenses occurred on the same occasion for ACCA purposes, or could have at least reasonable doubt about whether they occurred on different occasions. Kimbrough, 138 F.4th at 478-79. Less than a month went by between the first three offenses. R. 133 (Barnes PSR at 16) (Page ID #674). The offenses were strikingly similar in that, each time, Barnes used a handgun to rob a Family Dollar store, and he was not charged with targeting businesses of any other type. Id. And Barnes was arrested for all four offenses on the same date, pleaded guilty to all four on the

3 Like the parties and the majority, I rely largely on the presentence reports in these cases for their recitations of each defendant's prior offenses, and also on the government's representations about which offenses it believes to be ACCA predicates.

20

same date, and was sentenced to concurrent terms of imprisonment for all four on the same date. Id. at 15-16 (Page ID #673-74).

A reasonable jury, faced with this evidence, could conclude that the crimes were

"'similar or intertwined' and 'share[d] a common scheme or purpose.'" Cogdill, 130 F.4th at 529 (alteration in Cogdill) (quoting Wooden, 595 U.S. at 369).⁴Nor, based on our precedent, do the timing and physical-proximity factors require a different outcome. The record indicates that Barnes committed his offenses closer in time and location to one another than Cogdill did. Compare R. 133 (Barnes PSR at 16) (Page ID #674) (describing offenses committed within no more than two weeks and within a ten mile radius of one another) with Cogdill, 130 F.4th at 529-30 (describing offenses committed three months apart in the same 300-plus-square-mile county). Under Wooden, Erlinger, and our precedents, these facts could create in the jury reasonable doubt about whether the offenses occurred on different occasions. Cogdill, 130 F.4th at 529-30. And, because an error is not harmless if just one reasonable juror could find reasonable doubt, reversal is required. See id. at 529; Erlinger, 602 U.S. at 834-35, 840.

IV. DURHAM

Durham's prior convictions present a closer question for the ACCA occasions inquiry, and that closeness itself demonstrates the importance of the jury's role. Unlike the majority, I believe that binding, consistent precedent requires such a close question to be resolved by a jury of Durham's peers, not by judges on the Court of Appeals.

The majority relies on four "buckets" of offenses as supporting the ACCA enhancement to Durham's sentence: "1) four burglaries in October 2009; 2) the burglary of two different homes on October 19, 2010; 3) a burglary on December 15, 2010; and 4) four burglaries between

4 Any "differences" the majority notes between Barnes's offenses and those in Cogdill and Kimbrough, Maj. Op. at 11 & n.9, do not change this conclusion. Certainly, the distance between Barnes's offenses and the purportedly distinct nature of burglaries as compared to drug-trafficking offenses are pieces of evidence that could support a jury verdict that the offenses occurred on different occasions. But this does not matter if the record also creates reasonable doubt about whether the evidence did or could support a jury finding to the contrary. See Grogan, 133 F.4th at 564 ("[S]ince the government bears the burden of showing an error was harmless, if we are uncertain, 'the tie goes to the criminal defendant.'" (quoting A.K. ex rel. Kocher v. Durham Sch. Servs., L.P., 969 F.3d 625, 629 (6th Cir. 2020))). We must therefore reverse and let the jury conduct that inquiry for itself.

21

January 4, and January 28, 2011." Maj. Op. at 7. The majority then plucks dates from the temporal extremes and declares that no reasonable juror or speaker of the English language could conclude that they occurred on the same occasion. Id. at 7-8.⁵The majority's analysis seems to hinge on "a gap in time and space that clearly establishes the offenses occurred on different occasions." Id. ("[M]ore than three months separated the offenses [and] . . . the January 27 burglary occurred more than 15 miles from the houses burglarized in October 2010."). But in Cogdill, the offenses were also separated by about three months and the record reflected only that they occurred in the same county, potentially hundreds of miles apart, yet we held the error harmful. Cogdill, 130 F.4th at 529-30.

Although in Campbell we found harmless error where the predicate offenses occurred several months apart, the Supreme Court has made abundantly clear that timing is not the only factor a jury may consider in its "multi-factored" analysis. Wooden, 595 U.S. at 369. Campbell is distinguishable on the other two factors—proximity of location and character and relationship of the offenses. At least two of the three offenses in Campbell were separated by "eight hundred miles as the bird flies," much less proximate than the offenses here. Campbell, 122 F.4th at 632.

"And, by and large, Campbell's predicates were for wholly distinct crimes," unlike here where every single offense was a second-degree burglary. Id.

Even if there were no instructive circuit precedent on point, the majority's approach would ignore the obvious near-consecutive nature of the 2010 and 2011 offenses. It is one thing to ask if an October burglary occurred on a different occasion than a burglary the following January. It is quite another to ask whether a string of burglaries occurring at practically monthly intervals from October 2010 to February 2011 occurred on different occasions. The Supreme Court's precedent, as well as our own, has established that a jury may find reasonable doubt that a defendant's prior offenses were committed on different occasions based on the fact that the offenses are "similar or intertwined" or "share a common scheme or purpose."⁶Wooden, 595

5 This, despite the fact that, as explained above, the question should be whether a reasonable juror could have reasonable doubt about whether the offenses occurred on different occasions. See supra Part I.

6 The majority suggests that this approach is akin to "saying a married adult male could be reasonably called a 'bachelor' because two definitional elements of 'bachelor' are that one is an adult and a male." Maj. Op. at 8 n.5. This misses the point. To indulge in the majority's hypothetical, my approach is more accurately

22

U.S. at 369; accord Cogdill, 130 F.4th at 529; Kimbrough, 138 F.4th at 478-79; see also Campbell, 122 F.4th at 629 (acknowledging that "offenses with a 'common scheme' are more likely to be the same occasion" (quoting Wooden, 595 U.S. at 369)). Moreover, taken together, the regular frequency of Durham's offenses, along with their relative proximity to one another,⁷

and their shared character—all second-degree burglaries—could provide a reasonable jury with reasonable doubt that Durham's offenses occurred on different occasions.⁸That is enough to require remand here.

Would a jury probably find that the different-occasions test was satisfied? Maybe.⁹Can we be sure beyond a reasonable doubt that any reasonable jury would itself be sure beyond a reasonable doubt that the offenses occurred on different occasions? No.¹⁰

characterized this way: if there is conflicting evidence about whether the man is married—for example conflicting evidence about whether his divorce was ever finalized—that evidence must be put to a jury to determine whether the events required to finalize the divorce ever occurred.

7 After all, a car can travel the relevant distance of 15 miles in less than 15 minutes when the conditions are right. My disagreement with the majority as to what constitutes a "significant distance," see Maj. Op. at 8, illustrates that the line-drawing exercise in which the majority engages is not so clear cut. A jury might not agree that, in context, 15 miles is a significant distance, as the term used in Wooden and the majority opinion.

8 The majority asserts that Campbell forecloses considering the character and relationship of offenses at this level of generality. But Campbell merely states that the defendant's argument there, "that all offenses with the purpose of 'trafficking' drugs occur on the same occasion . . . would fail to honor the statute's text." 122 F.4th at

632. It could hardly be otherwise, given that the ACCA explicitly contemplates that drug offenses, specifically, can be predicates distinct from one another. 18 U.S.C. § 924(e)(1). In any case, I do not suggest, as the defendant in Campbell did, that a shared purpose alone is dispositive. See supra p. 23 (considering the frequency, proximity, and character of the offenses "taken together"). Finally, the shared characteristic in Campbell—a purpose of trafficking—was more general than the characteristics shared by Durham's offenses. For example, although both of Campbell's prior offenses could be characterized as having to do with drug trafficking, they involved two different drugs. 122 F.4th at 632.

9 I do not claim that Durham's offenses did in fact occur on the same occasion—that is for a jury to decide. The facts need only be capable of creating reasonable doubt about the conclusion that the offenses occurred on different occasions. This is not a high bar on these facts. See supra pp. 18-20; infra p. 23 n.10.

10 If it seems like the correct framing of the relevant question will result in remanding for jury consideration in a great many post-Erlinger harmless-error cases, this is by design. The burden is on the government to convince us to affirm despite the acknowledged error, Grogan, 133 F.4th at 563-64, and under binding Supreme Court precedent, the different-occasions inquiry is presumptively a question for the jury, Erlinger, 602 U.S. at 834-35. We need not worry about opening the so-called flood gates to Erlinger-error litigation under the harmless-error standard. As time proceeds and district courts, adapting to Erlinger, submit the different-occasions issue to juries, the issue will largely die out.

23

Despite Wooden's clear, multi-factored approach, some members of this court seem to have trouble letting go of their preferred, timing-and-location-based bright-line rule. See Kimbrough, 138 F.4th at 480 (Readler, J., concurring in part and dissenting in part) ("Can crimes committed nine days apart ever occur on the same 'occasion'? Consult a dictionary definition of the term, and the answer is surely no." (emphasis added)); Campbell, 122 F.4th at 629 ("In many instances, however, the analysis can be limited to inquiries like whether the offenses were committed a 'day or more apart' or at 'significant' distances, factors that can singularly decide the question." (quoting Wooden, 595 U.S. at 370)). True, the Supreme Court did write that "[i]n many cases, a single factor—especially of time or place—can decisively differentiate occasions."

Wooden, 595 U.S. at 369-70. But the Supreme Court did not hold, as Campbell implies, that

"whether the offenses were committed a 'day or more apart' . . . can singularly decide the [occasions] question." Campbell, 122 F.4th at 629 (quoting Wooden, 595 U.S. at 370) (emphasis and alteration added).¹¹ Rather, the Supreme Court was describing how courts have historically interpreted the word "occasion" for ACCA purposes. Wooden, 595 U.S. at 370. And, here, the majority acknowledges that Wooden represented a break from our past practice. Maj. Op. at 4-5. How could it not given that, in Wooden, the Supreme Court reversed a decision of this very court?

Still, the majority clings to timing, and to a lesser degree, location, as "often dispositive"

and concedes only that "other factors can inform the analysis" "given the right set of facts." Id. at 5 (emphasis added). But who decides when the set of facts is right for the nature and relationship of the offenses to factor into the occasions analysis? Under Erlinger, the jury decides. To the extent my colleagues seek to limit the different-occasions analysis in most cases to include only the time and proximity factors, this not only disregards the Supreme Court's analysis, but also reflects a value judgment about which factors should carry the most weight in any particular case. Contrary to some judges' attempts to reestablish a bright-line rule when offenses are committed a day or more apart, the Supreme Court has been clear that "no particular lapse of time or distance between offenses automatically separates a single occasion from

11 Campbell did also apply the nature-and-relationship-of-the-offenses factor and found that it pointed in the same direction as the timing and location factors. 122 F.4th at 632.

24

distinct ones." Erlinger, 602 U.S. at 841. "[Q]ualitative assessment" may be required, id., yet the majority here resolves Durham's case quantitatively and to the exclusion of qualitative factors. A jury may decide that is the proper way to consider the factors in a specific case, but, at this stage in this case, we may not. See Erlinger. 602 U.S. at 842 ("We do not question [the] assessment that in many cases the occasions inquiry will be 'straightforward.' . . . Often, a defendant's past offenses will be different enough and separated by enough time and space that there is little question he committed them on separate occasions. But none of that means a judge rather than a jury should make the call." (citation omitted)).

It is for the jury to determine which factors should carry the most weight as applied to specific facts. See Wooden, 595 U.S. at 384-85 (Gorsuch, J., concurring) (describing the occasions inquiry as a "'multi-factored' balancing test" (quoting Wooden, 595 U.S. at 369)). To hold, as the majority effectively does today, that Durham's offenses occurred so far apart in space and time that no reasonable jury could find the nature and relationship of the offenses to be worth considering at all is to presuppose the weight the jury would attach to each factor given any particular set of facts. When, on harmless-error review, the factors point in conflicting directions, as they arguably do here, it will ordinarily be for the jury to determine which factor outweighs the other,¹²just as it is up to a jury to decide whether to believe one witness over another.

This is not to say that, for harmless-error purposes, time or proximity can never be decisive. The Supreme Court acknowledged, albeit not under the harmless-error standard, that they could be. Id. at 369-70. There may indeed be cases where the offenses occurred so far apart in time and space that no reasonable jury could be swayed by the character and relationship of the offenses, allowing us to find harmless error.¹³This is not that case. For the reasons stated

12 Note that in each case where this court has found harmless Erlinger error, all three factors pointed toward the conclusion that the offenses were committed on different occasions. Campbell, 122 F.4th at 632; Robinson, 133

25

above, it is not beyond doubt that a reasonable jury could find reasonable doubt as to whether Durham's offenses occurred on different occasions.

I respectfully dissent from the majority's affirmance of Durham's sentence. And, for the reasons expressed above, I concur simply in the judgment regarding Barnes. Accordingly, I would vacate both Durham's and Barnes's sentences and remand in both cases.

26