USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 1 of 14
PUBLISHED
UNITED STATES COURT OF APPEALS FOR THE FOURTH CIRCUIT
No. 21-4250
UNITED STATES OF AMERICA,
Plaintiff – Appellee,
v.
KEITH EUGENE NELSON,
Defendant – Appellant.
Appeal from the United States District Court for the Southern District of West Virginia, at Bluefield. David A. Faber, Senior District Judge. (1:09-cr-00146-1)
Argued: May 5, 2022 Decided: June 17, 2022
Before WYNN, HARRIS, and RUSHING, Circuit Judges.
Vacated and remanded by published opinion. Judge Harris wrote the opinion, in which Judge Wynn and Judge Rushing joined.
ARGUED: Stephen J. van Stempvoort, MILLER JOHNSON, Grand Rapids, Michigan, for Appellant. Andrew Dayne Isabell, OFFICE OF THE UNITED STATES ATTORNEY, Beckley, West Virginia, for Appellee. ON BRIEF: Lisa G. Johnston, Acting United States Attorney, OFFICE OF THE UNITED STATES ATTORNEY, Charleston, West Virginia, for Appellee. USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 2 of 14
PAMELA HARRIS, Circuit Judge:
Upon revocation of a term of supervised release, Keith Eugene Nelson was
sentenced to what the district court believed was a mandatory minimum term of five
additional years of supervised release. On appeal, he argues for the first time that the
district court was mistaken, and that the statute governing his revocation sentence, 18
U.S.C. § 3583(h), does not impose any mandatory term of supervised release. And the
district court likewise erred, he contends, in determining that the applicable Sentencing
Guidelines provision recommended a five-year term of supervised release.
We agree with Nelson. The text of § 3583(h) is plain, and it provides for a
maximum but not a minimum term of supervised release. The relevant Guidelines
provision tracks the same clear text. Because the district court, in concluding otherwise,
committed a plain error that requires correction, we vacate Nelson’s sentence and remand
for resentencing.
I.
The events that led to this appeal began nearly two decades ago, in 2002, when
Nelson was convicted of two sex offenses in Illinois state court. As a result, Nelson was
required to register as a sex offender, 34 U.S.C. § 20913(a), which he did. Years later,
however, he moved from Illinois to West Virginia, where he failed to register. And when
pulled over for a traffic violation in April 2009, he admitted this lapse to a police officer.
Soon after, Nelson was indicted for, and pleaded guilty to, knowingly failing to
update his sex-offender registration. See 18 U.S.C. § 2250(a). When sentenced for that
2 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 3 of 14
offense – at what we will call his “original sentencing” – Nelson faced a maximum term of
ten years’ imprisonment. Id. Under 18 U.S.C. § 3583, which governs supervised release
after imprisonment, his § 2250(a) conviction also subjected him to a mandatory supervised
release term of no less than five years, with a maximum of life. See id. § 3583(k).
Consistent with those statutory dictates, the court sentenced Nelson to 41 months’
imprisonment and 25 years’ supervised release.
In the years after Nelson’s release from that original prison term, he often violated
the conditions of his supervised release, leading the district court repeatedly to revoke his
release and impose additional terms of imprisonment and supervised release. See id.
§ 3583(e)(3), (h) (authorizing imposition of imprisonment and supervised release after
revocation). This appeal concerns Nelson’s fifth revocation proceeding. In April 2021, a
probation officer notified the court that Nelson had violated several conditions of his
release and recommended that the court revoke Nelson’s release and send him back to
prison. The probation officer also calculated the applicable sentencing range, concluding,
as relevant here, that “the Court [was] authorized to impose a supervised release term of at
least five years to life upon any revocation.” J.A. 112 (emphasis added); see J.A. 114
(“Violation Worksheet” setting forth supervised release range from “[a] minimum of five
years to life, less any imprisonment imposed”).
At a revocation hearing, Nelson admitted to each charged violation. The court then
explained its view of the applicable sentencing regime, concluding that the “statutory
available sentence” was “24 months in custody and supervised release of five years to life,”
and that the Sentencing Guidelines also called for “a new term of supervised release of five
3 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 4 of 14
years.” J.A. 69; see 18 U.S.C. § 3583(e)(3) (capping revocation prison sentences at 24
months when the original offense, as here, is a Class C felony). No party objected, and the
court imposed a sentence of 24 months’ imprisonment and five years’ supervised release.
Nelson timely appealed.
II.
On appeal, Nelson argues that the district court committed procedural error in
calculating his sentencing range. According to Nelson, it is not the case, as the district
court determined, that he is subject to a five-year mandatory minimum term of supervised
release upon revocation. Instead, he contends, there is no minimum supervised release
term at all, because the statute governing supervised release after revocation, 18 U.S.C.
§ 3583(h), speaks only of a maximum term, not a minimum, and the relevant Guidelines
policy statement mirrors the statute’s terms.
Because Nelson concededly did not raise this objection to the district court, our
review is for plain error only. See Fed. R. Crim. P. 52(b) (“A plain error that affects
substantial rights may be considered even though it was not brought to the court’s
attention.”). To establish plain error, Nelson must show that (1) an error was made, (2) the
error was plain, and (3) the error affected his substantial rights. See United States v. Green,
996 F.3d 176, 185 (4th Cir. 2021) (citing United States v. Olano, 507 U.S. 725, 732 (1993)).
Even then, we will exercise our discretion to correct such an error only if declining to do
so “would result in a miscarriage of justice or would otherwise seriously affect the fairness,
4 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 5 of 14
integrity or public reputation of judicial proceedings.” Id. (internal quotation marks
omitted).
Although this standard is strict, Nelson has met it here.
A.
We first address whether the district court committed an error. The district court, at
the recommendation of Nelson’s probation officer, determined that the minimum term of
supervised release it could impose under § 3583(h) was five years, and that the Sentencing
Guidelines also called for a five-year term of supervised release. See J.A. 69 (holding that
the “statutory available sentence” included “supervised release of five years to life,” and
that the Guidelines recommended “a new term of supervised release of five years”). That
is incorrect. As we explain below, neither § 3583 nor the Guidelines provides any lower
limit – mandatory or advisory – on supervised release terms imposed after revocation.
All agree that at Nelson’s original sentencing for his § 2250(a) sex-offender
registration offense, the supervised release portion of his sentence was governed by
§ 3583(k). And under § 3583(k), it is clear and undisputed that Nelson was subject to a
mandatory minimum supervised release term of five years and a maximum term of life.
See 18 U.S.C. § 3583(k) (“[T]he authorized term of supervised release . . . for any offense
under section . . . 2250 . . . is any term of years not less than 5, or life.”).
This appeal, however, does not concern Nelson’s original sentencing. Instead, we
are dealing with his sentencing at a subsequent revocation proceeding, which is governed
by a separate provision of § 3583. Section 3583(h), titled “Supervised Release Following
Revocation,” provides as follows:
5 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 6 of 14
When a term of supervised release is revoked and the defendant is required to serve a term of imprisonment, the court may include a requirement that the defendant be placed on a term of supervised release after imprisonment. The length of such a term of supervised release shall not exceed the term of supervised release authorized by statute for the offense that resulted in the original term of supervised release, less any term of imprisonment that was imposed upon revocation of supervised release.
We think the meaning of that provision is clear. First, § 3583(h) allows, but does not
require, a court to impose a new term of supervised release upon revocation. Id. (“[T]he
court may include a [supervised release] requirement[.]” (emphasis added)). In other
words, there is no mandatory supervised release at all under § 3583(h). See United States
v. Teague, 8 F.4th 611, 616 (7th Cir. 2021) (“The use of the term ‘may’ . . . is unambiguous:
it is up to the court whether to add additional supervised release after the defendant serves
his additional time in prison.”). And second, if a court chooses to impose a term of
supervised release, § 3583(h) establishes only an upper limit on its length: The term “shall
not exceed” the supervised release term authorized at the defendant’s original sentencing.
Section 3583(h), that is, “prescribes a ceiling for additional supervised release, but no
floor.” Id. (holding that § 3583(h) does not incorporate mandatory minimum supervised
release term from original sentencing on § 841(b)(1) drug offense); see United States v.
Campos, 922 F.3d 686, 687–88 (5th Cir. 2019) (same); United States v. Roebuck, 761 F.
App’x 98, 103–04 (3d Cir. 2019) (unpublished) (same).
The result in Nelson’s case is equally clear. Section 3583(k) provided a mandatory
five-year term of supervised release at his original sentencing on his § 2250(a) offense.
But that mandatory five-year term did not apply at his subsequent revocation sentencing,
governed by § 3583(h). Instead, § 3583(h) made supervised release discretionary, and
6 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 7 of 14
imposed only one condition on any term of supervised release: that it “not exceed” the
lifetime term authorized by § 3583(k) at Nelson’s original sentencing. The district court’s
contrary conclusion – that § 3583(h) imported not only § 3583(k)’s maximum term of
supervised release but also its five-year mandatory minimum – simply cannot be reconciled
with the statutory text.
The same is true of the Guidelines policy statement on “Revocation of Probation or
Supervised Release.” See U.S.S.G. § 7B1.3. The relevant portion exactly tracks the
language of § 3583(h), again providing that upon revocation “the court may include” a
supervised release requirement and that the length of any such term of supervised release
“shall not exceed” the term of supervised release originally authorized by statute, less any
term of imprisonment imposed upon revocation. Id. § 7B1.3(g)(2) (emphasis added). So
that provision, too, as we already have recognized, addresses only an upper limit on terms
of supervised release, and “does not suggest a lower end to the range.” United States v.
Forrester, 740 F. App’x 269, 272–73 (4th Cir. 2018) (unpublished) (per curiam) (“[T]he
district court’s calculation of only the maximum term of supervised release was proper.”).
Again, the district court’s contrary conclusion – that the Sentencing Guidelines called for
a “new term of supervised release of five years” in Nelson’s case, see J.A. 69 – is in error.
The government offers little by way of response to this straightforward reading of
the relevant texts. This is perhaps not surprising; in two other federal circuit courts, the
government has conceded that the only plausible interpretation of § 3583(h) is the one we
adopt today. See Teague, 8 F.4th at 612, 615; Campos, 922 F.3d at 688. Here, however,
the government advances a different theory: Under § 3583(h), the government suggests, a
7 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 8 of 14
district court may choose not to impose any supervised release at all, but if it does impose
a term of supervised release, then § 3583(h) incorporates § 3583(k)’s original sentencing
scheme in its entirety, minimum and maximum alike. So for Nelson, the government
would have it, the available statutory range was a term of supervised release of either zero
or, as at his original sentencing under § 3583(k), at least five years – but nothing in
between.
We may put to one side the oddness of that result. The fundamental problem is that
the government’s novel reading is not faithful to § 3583(h)’s text. The government’s
“zero” option, to be sure, does account for the permissive “may” in § 3583(h), making the
imposition of supervised release discretionary. But the government cannot account for the
critical phrase “shall not exceed,” which makes clear that § 3583(h) imposes only a
maximum term of supervised release and not a minimum. See, e.g., Campos, 922 F.3d
at 687–88. For the government’s theory to work, “shall not exceed” would have to become
“shall not exceed or fall below”: “The length of such term of supervised release shall not
exceed [or fall below] the term of supervised release authorized by statute” at the original
sentencing. But those, of course, are not the words that Congress chose. 1
1 In support of its theory, the government relies chiefly on our unpublished decision in United States v. Thompson, 794 F. App’x 350 (4th Cir. 2020) (per curiam). With respect, we fail to see its relevance. Thompson addresses the maximum term of supervised release allowed by § 3583(h) and, specifically, § 3583(h)’s “subtraction rule,” incorporating the maximum term authorized at the original sentencing “less any term of imprisonment that was imposed upon revocation of supervised release.” Id. at 351 (quoting 18 U.S.C. § 3583(h)). But there is no dispute here about the maximum supervised release term to which Nelson is subject. And Thompson has nothing to say on the question that is in (Continued) 8 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 9 of 14
B.
Next, we consider whether this error is plain – that is, whether it is “clear or obvious,
rather than subject to reasonable dispute.” Puckett v. United States, 556 U.S. 129, 135
(2009). Again, we answer in the affirmative.
The district court’s error in calculating Nelson’s supervised release sentencing range
is “plain” for purposes of our Rule 52(b) review because the statutory text – as set out in
§ 3583(h) and replicated in section 7B1.3(g)(2) of the Sentencing Guidelines – provides a
“clear” and “obvious” answer to the question before us. Id. Often, in undertaking this
inquiry, we look to controlling decisions from our circuit or the Supreme Court, or a
consensus among other circuits. See, e.g., Green, 996 F.3d at 185. But “an error is [also]
plain if . . . the explicit language of a statute or rule resolves the question” before us.
United States v. Davis, 855 F.3d 587, 595–96 (4th Cir. 2017). It is only “[w]here the
explicit language of a statute or rule does not specifically resolve an issue” that we must
resort to judicial authority to decide whether an error is plain. United States v. Beasley,
495 F.3d 142, 149 (4th Cir. 2007) (quoting United States v. Lejarde-Rada, 319 F.3d 1288,
1291 (11th Cir. 2003)) (consulting case law only after finding the relevant statute to be
“ambiguous”). 2
dispute: whether a defendant is subject, under § 3583(h), to a minimum supervised release term imported from his original sentencing. 2 Indeed, our sister circuits routinely rely solely on plain text to find plain error. See, e.g., United States v. Sholley-Gonzalez, 996 F.3d 887, 899 (8th Cir. 2021); United States v. Wang, 944 F.3d 1081, 1089 (9th Cir. 2019); United States v. Murphy, 942 F.3d 73, 86 (2d Cir. 2019); United States v. Courtney, 816 F.3d 681, 685–86 (10th Cir. 2016); United (Continued) 9 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 10 of 14
Here, we agree with Nelson that the “explicit language” of § 3583(h) and U.S.S.G.
§ 7B1.3(g)(2) “specifically resolve[s]” the issue we confront. See id. (quoting Lejarde-
Rada, 319 F.3d at 1291). As explained above, the terms “may” and “shall not exceed”
cannot reasonably be read to impose or suggest a mandatory or minimum term of
supervised release. See Roebuck, 761 F. App’x at 104 (finding same error to be “plain”
under Rule 52(b) because “it is clear that the Guidelines and § 3583(h) each contain ‘cannot
exceed’ language and do not require a subsequent term of supervised release”). Although
this case marks our circuit’s first encounter with this question – likely because of the clarity
of the governing statute – it does not take controlling authority to confirm that “exceed”
means “exceed.”
Although our analysis does not depend on it, we note that other federal circuit courts
have had occasion to address this question, and uniformly have concluded, as we do today,
that § 3583(h) does not impose any minimum term of supervised release. See Green, 996
F.3d at 185 (relying in part on consensus in other circuits to establish plain error). As
described above, three circuits, construing the same language before us today, have held
that § 3583(h) “prescribes a ceiling for additional supervised release, but no floor” – and
also that the statutory text, by itself, is sufficiently clear that it is plain error to conclude
otherwise. See Teague, 8 F.4th at 616; Campos, 922 F.3d at 687–88; Roebuck, 761 F.
States v. Husmann, 765 F.3d 169, 177 (3d Cir. 2014); In re Sealed Case, 573 F.3d 844, 851–52 (D.C. Cir. 2009); United States v. Maturin, 488 F.3d 657, 663 (5th Cir. 2007).
10 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 11 of 14
App’x at 103–04. 3 We are aware of no authority to the contrary. So in addition to the clear
text of the statute, we have here as well a three-circuit consensus confirming the plainness
of the error in question. See United States v. Maxwell, 285 F.3d 336, 342 (4th Cir. 2002)
(finding plain error, in the absence of controlling authority, in light of clear statutory text
and three-circuit consensus).
C.
That brings us to the third plain-error factor: whether Nelson has shown that the
error in calculating his supervised release sentencing range affected his substantial rights.
To make this showing, a defendant must establish a “‘reasonable probability’ of a different
outcome – in this context, a lower sentence – absent the error.” Green, 996 F.3d at 186
(quoting Molina-Martinez v. United States, 136 S. Ct. 1338, 1345 (2016)). Again, Nelson
meets this standard.
In the context of a Sentencing Guidelines error, the Supreme Court has instructed
that when a defendant is sentenced under an incorrect advisory range, “the error itself can,
and most often will, be sufficient to show a reasonable probability” that his sentence would
have been different had the district court used the correct “framework.” Molina-Martinez,
136 S. Ct. at 1345–46. In the “ordinary case,” in other words, a miscalculation of a
3 In all three of those cases, the specific question was whether § 3583(h) incorporates the mandatory minimum supervised release term that governs in an original sentencing for a § 841(b)(1) drug offense; here, the question is whether § 3583(h) incorporates the mandatory minimum supervised release term that governed, under § 3583(k), in Nelson’s original sentencing for his sex-offender registration offense. Either way, the answer turns entirely on the interpretation of § 3583(h), and so these cases are directly on point.
11 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 12 of 14
Guidelines range – which, though advisory, is the “lodestar” for sentencing – will be
enough to establish the necessary effect on substantial rights. Green, 996 F.3d at 186
(quoting Molina-Martinez, 136 S. Ct. at 1346). And the same, of course, must be true when
a district court, as here, also incorrectly calculates a statutory sentencing range – which is
not advisory but mandatory, so that the court has no discretion to go below what it
mistakenly believes to be a statutory floor. See Campos, 922 F.3d at 689 (“[T]he logic of
Molina-Martinez applies with even more force to this [§ 3583(h)] mistake.”).
The government offers no ground for departing from the Molina-Martinez analysis,
and this seems to be exactly the “ordinary case” that Molina-Martinez had in mind. 4 It is
true, as the government notes, that the district court gave reasons for the sentence it
imposed, and that Nelson’s five-year supervised release term falls well below the statutory
maximum. But the government has pointed us to no record evidence – and none is apparent
– that the district court’s five-year sentence “was imposed independent of its error” in
calculating a five-year minimum term of supervised release under § 3583(h) and a
comparable recommendation under the Guidelines. Green, 996 F.3d at 186–87; see
Molina-Martinez, 136 S. Ct. at 1346–47. Instead, as in Molina-Martinez, the district court
“said nothing to suggest that it would have imposed” the same sentence regardless of the
4 Molina-Martinez, we recognize, addressed a carceral sentence, not supervised release. The government does not suggest that this distinction is relevant here, and for good reason: We have held already that because the “terms and conditions of supervised release are a substantial imposition on a person’s liberty,” a plainly erroneous extension of a supervised release term, like a prison term, affects “substantial rights” and is correctable under Rule 52(b). See Maxwell, 285 F.3d at 342.
12 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 13 of 14
statutory and Guidelines sentencing ranges. 136 S. Ct. at 1348. And as in Molina-
Martinez, the five-year sentence the court “selected is conspicuous for its position as the
lowest sentence within what the District Court believed to be the applicable range.” Id.
at 1347. Under these circumstances, Nelson has shown that the plain error in calculating
his sentencing range affected his substantial rights for purposes of Rule 52(b) review.
D.
Last, “where a plain error in calculating the Guidelines sentencing range affects
substantial rights, the fourth prong of plain error – whether an error affects the ‘fairness,
integrity, or public reputation of judicial proceedings’ – ordinarily will be satisfied, as
well.” Green, 996 F.3d at 187 (quoting Rosales-Mireles v. United States, 138 S. Ct. 1897,
1908 (2018)). “Although there may be cases in which ‘countervailing factors’ make
correction of a plain error inappropriate, the government has presented none here, nor made
any other argument regarding the fourth factor.” Id. (quoting Rosales-Mireles, 138 S. Ct.
at 1909). We likewise see no reason why this plain error should go uncorrected, and
accordingly vacate Nelson’s revocation sentence and remand.
Our vacatur of Nelson’s sentence includes both its carceral and supervised release
aspects, and our remand is for a full resentencing at which the district court can determine,
in light of the correct supervised release range, the appropriate overall sentence in this case.
That is the relief Nelson requests, and the government has not argued for a different result.
It also accords with our usual practice of giving the district court the chance to reconsider
13 USCA4 Appeal: 21-4250 Doc: 41 Filed: 06/17/2022 Pg: 14 of 14
its sentencing decision as a whole when we disturb one portion thereof. 5 See, e.g., United
States v. Singletary, 984 F.3d 341, 346 & n.4 (4th Cir. 2021) (vacating “entire sentence”
after finding supervised release conditions improper, “given that custodial and supervised
release terms are components of one unified sentence” (cleaned up)); cf. United States v.
Pratt, 915 F.3d 266, 275 (4th Cir. 2019) (explaining that “a full resentencing is typically
appropriate” upon vacatur of some but not all of a defendant’s convictions).
III.
For the reasons given above, we vacate Nelson’s sentence and remand for
proceedings consistent with this opinion.
VACATED AND REMANDED
5 This custom is not an inflexible rule. But it arises from the sound recognition that a “criminal sentence is a package of sanctions,” and that a district court’s sentencing intent “may be undermined by altering one portion of the calculus” while freezing others in place. Pepper v. United States, 562 U.S. 476, 507 (2011) (internal quotation marks omitted). In this case, where neither party has argued for a different outcome, we are satisfied that allowing the district court holistically to revisit its sentencing decision is an appropriate result.