(Slip Opinion)              OCTOBER TERM, 2015                                       1

                                       Syllabus

         NOTE: Where it is feasible, a syllabus (headnote) will be released, as is
       being done in connection with this case, at the time the opinion is issued.
       The syllabus constitutes no part of the opinion of the Court but has been
       prepared by the Reporter of Decisions for the convenience of the reader.
       See United States v. Detroit Timber & Lumber Co., 

, 337.


SUPREME COURT OF THE UNITED STATES

                                       Syllabus

                     WELCH v. UNITED STATES

CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR
                THE ELEVENTH CIRCUIT

     No. 15–6418. Argued March 30, 2016—Decided April 18, 2016
Federal law makes the possession of a firearm by a felon a crime pun-
  ishable by a prison term of up to 10 years, 

. §§922(g),
  924(a)(2), but the Armed Career Criminal Act of 1984 increases that
  sentence to a mandatory 15 years to life if the offender has three or
  more prior convictions for a “serious drug offense” or a “violent felo-
  ny,” §924(e)(1). The definition of “violent felony” includes the so-
  called residual clause, covering any felony that “otherwise involves
  conduct that presents a serious potential risk of physical injury to
  another.” §924(e)(2)(B)(ii). In Johnson v. United States, 576 U. S.
  ___, this Court held that clause unconstitutional under the void-for-
  vagueness doctrine.
     Petitioner Welch was sentenced under the Armed Career Criminal
  Act before Johnson was decided. On direct review, the Eleventh Cir-
  cuit affirmed his sentence, holding that Welch’s prior Florida convic-
  tion for robbery qualified as a “violent felony” under the residual
  clause. After his conviction became final, Welch sought collateral re-
  lief under 

. §2255, which the District Court denied. The
  Eleventh Circuit then denied Welch a certificate of appealability.
  Three weeks later, this Court decided Johnson. Welch now seeks the
  retroactive application of Johnson to his case.
Held: Johnson announced a new substantive rule that has retroactive
 effect in cases on collateral review. Pp. 6–15.
      (a) An applicant seeking a certificate of appealability in a §2255
 proceeding must make “a substantial showing of the denial of a con-
 stitutional right.” §2253(c)(2). That standard is met when “reasona-
 ble jurists could debate whether . . . the petition should have been re-
 solved in a different manner.” Slack v. McDaniel, 

, 484.
 The question whether Welch met that standard implicates a broader
2                      WELCH v. UNITED STATES

                                  Syllabus

    legal issue: whether Johnson is a substantive decision with retroac-
    tive effect in cases on collateral review. If so, then on the present
    record reasonable jurists could at least debate whether Welch should
    obtain relief in his collateral challenge to his sentence. Pp. 6–7.
         (b) New constitutional rules of criminal procedure generally do
    not apply retroactively to cases on collateral review, but new sub-
    stantive rules do apply retroactively. Teague v. Lane, 

,
    310; Schriro v. Summerlin, 

, 351. Substantive rules al-
    ter “the range of conduct or the class of persons that the law punish-
    es,” 

Procedural rules, by contrast, “regulate only the
    manner of determining the defendant’s culpability.” 

Johnson is substantive. Before Johnson, the residual
    clause could cause an offender to face a prison sentence of at least 15
    years instead of at most 10. Since Johnson made the clause invalid,
    it can no longer mandate or authorize any sentence. By the same log-
    ic, Johnson is not procedural, since it had nothing to do with the
    range of permissible methods a court might use to determine whether
    a defendant should be sentenced under the Act, see 

. Pp. 7–9.
         (c) The counterarguments made by Court-appointed amicus are
    unpersuasive. She contends that Johnson is a procedural decision
    because the void-for-vagueness doctrine is based on procedural due
    process. But the Teague framework turns on whether the function of
    the rule is substantive or procedural, not on the rule’s underlying
    constitutional source. Amicus’ approach would lead to results that
    cannot be squared with prior precedent. Precedent also does not
    support amicus’ claim that a rule must limit Congress’ power to be
    substantive, see, e.g., Bousley v. United States, 

, or her
    claim that statutory construction cases are an ad hoc exception to
    that principle and are substantive only because they implement the
    intent of Congress. The separation-of-powers argument raised by
    amicus is also misplaced, for regardless of whether a decision in-
    volves statutory interpretation or statutory invalidation, a court
    lacks the power to exact a penalty that has not been authorized by
    any valid criminal statute. Pp. 10–15.
Vacated and remanded.

   KENNEDY, J., delivered the opinion of the Court, in which ROBERTS,
C. J., and GINSBURG, BREYER, ALITO, SOTOMAYOR, and KAGAN, JJ.,
joined. THOMAS, J., filed a dissenting opinion.
                        Cite as: 578 U. S. ____ (2016)                              1

                             Opinion of the Court

     NOTICE: This opinion is subject to formal revision before publication in the
     preliminary print of the United States Reports. Readers are requested to
     notify the Reporter of Decisions, Supreme Court of the United States, Wash-
     ington, D. C. 20543, of any typographical or other formal errors, in order
     that corrections may be made before the preliminary print goes to press.


SUPREME COURT OF THE UNITED STATES
                                   _________________

                                   No. 15–6418
                                   _________________


GREGORY WELCH, PETITIONER v. UNITED STATES
 ON WRIT OF CERTIORARI TO THE UNITED STATES COURT OF
          APPEALS FOR THE ELEVENTH CIRCUIT
                                 [April 18, 2016]

  JUSTICE KENNEDY delivered the opinion of the Court.
  Last Term, this Court decided Johnson v. United States,
576 U. S. ___ (2015). Johnson considered the residual
clause of the Armed Career Criminal Act of 1984, 

. §924(e)(2)(B)(ii). The Court held that provision
void for vagueness. The present case asks whether John-
son is a substantive decision that is retroactive in cases on
collateral review.
                             I
   Federal law prohibits any felon—meaning a person who
has been convicted of a crime punishable by more than a
year in prison—from possessing a firearm. 

.
§922(g). A person who violates that restriction can be
sentenced to prison for up to 10 years. §924(a)(2). For
some felons, however, the Armed Career Criminal Act
imposes a much more severe penalty. Under the Act, a
person who possesses a firearm after three or more convic-
tions for a “serious drug offense” or a “violent felony” is
subject to a minimum sentence of 15 years and a maxi-
mum sentence of life in prison. §924(e)(1). Because the
ordinary maximum sentence for a felon in possession of a
firearm is 10 years, while the minimum sentence under
2                 WELCH v. UNITED STATES

                      Opinion of the Court

the Armed Career Criminal Act is 15 years, a person
sentenced under the Act will receive a prison term at least
five years longer than the law otherwise would allow.
   The Act defines “violent felony” as
    “any crime punishable by imprisonment for a term ex-
    ceeding one year . . . that—
    “(i) has as an element the use, attempted use, or
    threatened use of physical force against the person of
    another; or
    “(ii) is burglary, arson, or extortion, involves use of
    explosives, or otherwise involves conduct that pre-
    sents a serious potential risk of physical injury to an-
    other.” §924(e)(2)(B).
   Subsection (i) of this definition is known as the elements
clause. The end of subsection (ii)—“or otherwise involves
conduct that presents a serious potential risk of physical
injury to another”—is known as the residual clause. See

, at ___ (slip op., at 2). It is the residual
clause that Johnson held to be vague and invalid.
   The text of the residual clause provides little guidance
on how to determine whether a given offense “involves
conduct that presents a serious potential risk of physical
injury.” This Court sought for a number of years to de-
velop the boundaries of the residual clause in a more pre-
cise fashion by applying the statute to particular cases. See
James v. United States, 

(2007) (residual
clause covers Florida offense of attempted burglary);
Begay v. United States, 

(2008) (residual
clause does not cover New Mexico offense of driving under
the influence of alcohol); Chambers v. United States, 

(2009) (residual clause does not cover Illinois
offense of failure to report to a penal institution); Sykes v.
United States, 

(2011) (residual clause covers
Indiana offense of vehicular flight from a law-enforcement
                 Cite as: 578 U. S. ____ (2016)            3

                     Opinion of the Court

officer). In Johnson, a majority of this Court concluded
that those decisions did not bring sufficient clarity to the
scope of the residual clause, noting that the federal courts
remained mired in “pervasive disagreement” over how the
clause should be interpreted. Johnson, 576 U. S., at ___
(slip op., at 9).
   The Johnson Court held the residual clause unconstitu-
tional under the void-for-vagueness doctrine, a doctrine
that is mandated by the Due Process Clauses of the Fifth
Amendment (with respect to the Federal Government) and
the Fourteenth Amendment (with respect to the States).
The void-for-vagueness doctrine prohibits the government
from imposing sanctions “under a criminal law so vague
that it fails to give ordinary people fair notice of the con-
duct it punishes, or so standardless that it invites arbi-
trary enforcement.” Id., at ___ (slip op., at 3). Johnson
determined that the residual clause could not be recon-
ciled with that prohibition.
   The vagueness of the residual clause rests in large part
on its operation under the categorical approach. The
categorical approach is the framework the Court has
applied in deciding whether an offense qualifies as a vio-
lent felony under the Armed Career Criminal Act. See id.,
at ___ (slip op., at 4). Under the categorical approach, “a
court assesses whether a crime qualifies as a violent fel-
ony ‘in terms of how the law defines the offense and not in
terms of how an individual offender might have committed
it on a particular occasion.’ ” 

at 141). For purposes of the residual clause, then, courts
were to determine whether a crime involved a “serious
potential risk of physical injury” by considering not the
defendant’s actual conduct but an “idealized ordinary case
of the crime.” 576 U. S., at ___ (slip op., at 12).
   The Court’s analysis in Johnson thus cast no doubt on
the many laws that “require gauging the riskiness of
conduct in which an individual defendant engages on a
4                 WELCH v. UNITED STATES

                      Opinion of the Court

particular occasion.” 

failed not
because it adopted a “serious potential risk” standard but
because applying that standard under the categorical
approach required courts to assess the hypothetical risk
posed by an abstract generic version of the offense. In the
Johnson Court’s view, the “indeterminacy of the wide-
ranging inquiry” made the residual clause more unpre-
dictable and arbitrary in its application than the Consti-
tution allows. Id., at ___ (slip op., at 5). “Invoking so
shapeless a provision to condemn someone to prison for 15
years to life,” the Court held, “does not comport with the
Constitution’s guarantee of due process.” Id., at ___ (slip
op., at 10).
                               II
  Petitioner Gregory Welch is one of the many offenders
sentenced under the Armed Career Criminal Act before
Johnson was decided. Welch pleaded guilty in 2010 to one
count of being a felon in possession of a firearm. The
Probation Office prepared a presentence report finding
that Welch had three prior violent felony convictions,
including a Florida conviction for a February 1996 “strong-
arm robbery.” The relevant Florida statute prohibits
taking property from the person or custody of another with
“the use of force, violence, assault, or putting in fear.” Fla.
Stat. §812.13(1) (1994). The charging document from the
1996 Florida case tracked that statutory language. App.
187a. The 2010 federal presentence report provides more
detail. It states that, according to the robbery victim,
Welch punched the victim in the mouth and grabbed a
gold bracelet from his wrist while another attacker
grabbed a gold chain from his neck.
  Welch objected to the presentence report, arguing (as
relevant here) that this conviction was not a violent felony
conviction under the Armed Career Criminal Act. The
District Court overruled the objection. It concluded that
                 Cite as: 578 U. S. ____ (2016)            5

                     Opinion of the Court

the Florida offense of strong-arm robbery qualified as a
violent felony both under the elements clause, 

.
§924(e)(2)(B)(i), and the residual clause, §924(e)(2)(B)(ii).
The District Court proceeded to sentence Welch to the
Act’s mandatory minimum sentence of 15 years in prison.
  The Court of Appeals for the Eleventh Circuit affirmed.
That court did not decide whether the conviction at issue
could qualify as a violent felony under the elements
clause. Instead, it held only that the conviction qualified
under the residual clause. This Court denied certiorari,
see Welch v. United States, 568 U. S. ___ (2013), and
Welch’s conviction became final.
  In December 2013, Welch appeared pro se before the
District Court and filed a collateral challenge to his con-
viction and sentence through a motion under 

.
§2255. He argued, among other points, that his strong-
arm robbery conviction itself was “vague” and that his
counsel was ineffective for allowing him to be sentenced as
an armed career criminal. The District Court denied the
motion and denied a certificate of appealability.
  Still proceeding pro se, Welch applied to the Court of
Appeals for a certificate of appealability. His application
noted that Johnson was pending before this Court. Welch
argued, in part, that his “armed career offender status is
unconstitutional and violate[s] [his] Fifth Amendment
right to notice of the state priors.” App. 20a. Two months
later, Welch filed a motion asking the Court of Appeals to
hold his case in abeyance until Johnson could be decided,
“based on the fact he was sentenced under the [residual
clause].” App. 15a.
  In June 2015, the Court of Appeals entered a brief
single-judge order denying the motion for a certificate of
appealability. Less than three weeks later, this Court
issued its decision in Johnson holding, as already noted,
that the residual clause is void for vagueness. Welch filed
a motion asking the Court of Appeals for additional time
6                 WELCH v. UNITED STATES

                      Opinion of the Court

to seek reconsideration of its decision in light of Johnson,
but the court returned that motion unfiled because
Welch’s time to seek reconsideration already had expired.
   Welch then filed a pro se petition for certiorari. His
petition presented two questions: whether the District
Court erred in denying his §2255 motion because his
Florida robbery conviction does not qualify as a violent
felony conviction under the Armed Career Criminal Act;
and whether Johnson announced a substantive rule that
has retroactive effect in cases on collateral review. Pet. for
Cert. i. This Court granted the petition. 577 U. S. ___
(2016). Because the United States, as respondent, agrees
with Welch that Johnson is retroactive, the Court ap-
pointed Helgi C. Walker as amicus curiae in support of the
judgment of the Court of Appeals. She has ably dis-
charged her responsibilities.
                              III

                               A

   This case comes to the Court in a somewhat unusual
procedural posture. Under the Antiterrorism and Effec-
tive Death Penalty Act of 1996, there can be no appeal
from a final order in a §2255 proceeding unless a circuit
justice or judge issues a certificate of appealability. 

. §2253(c)(1). A certificate of appealability may
issue “only if the applicant has made a substantial show-
ing of the denial of a constitutional right.” §2253(c)(2).
That standard is met when “reasonable jurists could de-
bate whether (or, for that matter, agree that) the petition
should have been resolved in a different manner.” Slack v.
McDaniel, 

, 484 (2000). Obtaining a certifi-
cate of appealability “does not require a showing that the
appeal will succeed,” and “a court of appeals should not
decline the application . . . merely because it believes the
applicant will not demonstrate an entitlement to relief.”
Miller-El v. Cockrell, 

, 337 (2003).
                 Cite as: 578 U. S. ____ (2016)            7

                     Opinion of the Court

  The decision under review here is the single-judge order
in which the Court of Appeals denied Welch a certificate of
appealability. Under the standard described above, that
order determined not only that Welch had failed to show
any entitlement to relief but also that reasonable jurists
would consider that conclusion to be beyond all debate.
See 

. The narrow question here is
whether the Court of Appeals erred in making that deter-
mination. That narrow question, however, implicates a
broader legal issue: whether Johnson is a substantive
decision with retroactive effect in cases (like Welch’s) on
collateral review. If so, then on the present record reason-
able jurists could at least debate whether Welch should
obtain relief in his collateral challenge to his sentence. On
these premises, the Court now proceeds to decide whether
Johnson is retroactive.
                             B
   The normal framework for determining whether a new
rule applies to cases on collateral review stems from the
plurality opinion in Teague v. Lane, 

(1989).
That opinion in turn drew on the approach outlined by the
second Justice Harlan in his separate opinions in Mackey
v. United States, 

(1971), and Desist v. United
States, 

(1969). The parties here assume that
the Teague framework applies in a federal collateral chal-
lenge to a federal conviction as it does in a federal collat-
eral challenge to a state conviction, and we proceed on
that assumption. See Chaidez v. United States, 568 U. S.
___, ___, n. 16 (2013); Danforth v. Minnesota, 

, 269, n. 4 (2008).
   Under Teague, as a general matter, “new constitutional
rules of criminal procedure will not be applicable to those
cases which have become final before the new rules are

. Teague and its progeny
recognize two categories of decisions that fall outside this
8                WELCH v. UNITED STATES

                     Opinion of the Court

general bar on retroactivity for procedural rules. First,
“[n]ew substantive rules generally apply retroactively.”
Schriro v. Summerlin, 

, 351 (2004); see
Montgomery v. Louisiana, 577 U. S. ___, ___ (2016) (slip
op., at 6); 

, 311. Second, new “ ‘wa-
tershed rules of criminal procedure,’ ” which are proce-
dural rules “implicating the fundamental fairness and accu-
racy of the criminal proceeding,” will also have retroactive
effect. Saffle v. Parks, 

, 495 (1990); see

–313.
   It is undisputed that Johnson announced a new rule.
See 

(“[A] case announces a new rule
if the result was not dictated by precedent existing at the
time the defendant’s conviction became final”). The ques-
tion here is whether that new rule falls within one of the
two categories that have retroactive effect under Teague.
The parties agree that Johnson does not fall into the
limited second category for watershed procedural rules.
Welch and the United States contend instead that John-
son falls into the first category because it announced a
substantive rule.
   “A rule is substantive rather than procedural if it alters
the range of conduct or the class of persons that the law
punishes.” 

. “This includes
decisions that narrow the scope of a criminal statute by
interpreting its terms, as well as constitutional determina-
tions that place particular conduct or persons covered by
the statute beyond the State’s power to punish.” 

(citation omitted); see 

, at ___
(slip op., at 6). Procedural rules, by contrast, “regulate
only the manner of determining the defendant’s culpabil-
ity.” 

. Such rules alter “the
range of permissible methods for determining whether a
defendant’s conduct is punishable.” 

produce a class of persons convicted of conduct the law
does not make criminal, but merely raise the possibility
                 Cite as: 578 U. S. ____ (2016)           9

                     Opinion of the Court

that someone convicted with use of the invalidated proce-
dure might have been acquitted otherwise.” 

  Under this framework, the rule announced in Johnson
is substantive. By striking down the residual clause as
void for vagueness, Johnson changed the substantive
reach of the Armed Career Criminal Act, altering “the
range of conduct or the class of persons that the [Act]
punishes.” 

. Before Johnson, the
Act applied to any person who possessed a firearm after
three violent felony convictions, even if one or more of
those convictions fell under only the residual clause. An
offender in that situation faced 15 years to life in prison.
After Johnson, the same person engaging in the same
conduct is no longer subject to the Act and faces at most
10 years in prison. The residual clause is invalid under
Johnson, so it can no longer mandate or authorize any
sentence. Johnson establishes, in other words, that “even
the use of impeccable factfinding procedures could not
legitimate” a sentence based on that clause. United States
v. United States Coin & Currency, 

, 724
(1971). It follows that Johnson is a substantive decision.
  By the same logic, Johnson is not a procedural decision.
Johnson had nothing to do with the range of permissible
methods a court might use to determine whether a de-
fendant should be sentenced under the Armed Career
Criminal Act. See 

. It did not,
for example, “allocate decisionmaking authority” between
judge and jury, ibid., or regulate the evidence that the
court could consider in making its decision, see Whorton v.
Bockting, 

, 413–414, 417 (2007); 

–701 (opinion of Harlan, J.). Unlike those
cases, Johnson affected the reach of the underlying statute
rather than the judicial procedures by which the statute is
applied. Johnson is thus a substantive decision and so has
retroactive effect under Teague in cases on collateral
review.
10               WELCH v. UNITED STATES

                     Opinion of the Court 


                             C

   Amicus urges the Court to adopt a different understand-
ing of the Teague framework. She contends courts should
apply that framework by asking whether the constitu-
tional right underlying the new rule is substantive or
procedural. Under that approach, amicus concludes that
Johnson is a procedural decision because the void-for-
vagueness doctrine that Johnson applied is based, she
asserts, on procedural due process.
   Neither Teague nor its progeny support that approach.
As described above, this Court has determined whether a
new rule is substantive or procedural by considering the
function of the rule, not its underlying constitutional
source. See, e.g., 

–353. That is for
good reason. The Teague framework creates a balance
between, first, the need for finality in criminal cases, and
second, the countervailing imperative to ensure that crim-
inal punishment is imposed only when authorized by law.
That balance turns on the function of the rule at issue, not
the constitutional guarantee from which the rule derives.
If a new rule regulates only the procedures for determin-
ing culpability, the Teague balance generally tips in favor
of finality. The chance of a more accurate outcome under
the new procedure normally does not justify the cost of
vacating a conviction whose only flaw is that its proce-
dures “conformed to then-existing constitutional stand-
ards.” 

. On the other hand, if a new
rule changes the scope of the underlying criminal proscrip-
tion, the balance is different. A change of that character
will “necessarily carry a significant risk that a defendant
stands convicted of ‘an act that the law does not make
criminal.’ ” Bousley v. United States, 

, 620
(1998) (quoting Davis v. United States, 

, 346
(1974)). By extension, where the conviction or sentence in
fact is not authorized by substantive law, then finality
interests are at their weakest. As Justice Harlan wrote,
                  Cite as: 578 U. S. ____ (2016)            11

                      Opinion of the Court

“[t]here is little societal interest in permitting the criminal
process to rest at a point where it ought properly never to
repose.” 

(opinion of Harlan, J.).
   The Teague balance thus does not depend on whether
the underlying constitutional guarantee is characterized
as procedural or substantive. It depends instead on
whether the new rule itself has a procedural function or a
substantive function—that is, whether it alters only the
procedures used to obtain the conviction, or alters instead
the range of conduct or class of persons that the law pun-
ishes. See 

; Montgomery, 577 U. S.,
at ___ (slip op., at 14). The emphasis by amicus on the
constitutional guarantee behind the new rule, then, would
untether the Teague framework from its basic purpose.
   The approach amicus suggests also would lead to results
that cannot be squared with prior precedent. Decisions
from this Court show that a rule that is procedural for
Teague purposes still can be grounded in a substantive
constitutional guarantee. For instance, the Court has
adopted certain rules that regulate capital sentencing
procedures in order to enforce the substantive guarantees
of the Eighth Amendment. The consistent position has
been that those rules are procedural, even though their
ultimate source is substantive. See, e.g., Beard v. Banks,

, 408, 416–417 (2004); Sawyer v. Smith, 

, 233, 241–242 (1990). From the converse per-
spective, there also can be substantive rules based on
constitutional protections that, on the theory amicus
advances, likely would be described as procedural. For
instance, a decision that invalidates as void for vagueness
a statute prohibiting “conduct annoying to persons passing
by,” cf. Coates v. Cincinnati, 

, 612, 614
(1971), would doubtless alter the range of conduct that the
law prohibits. That would make it a substantive decision
under our precedent, see 

, even if
the reasons for holding that statute invalid could be char-
12                WELCH v. UNITED STATES

                      Opinion of the Court

acterized as procedural.
   Amicus next relies on language from this Court’s cases
describing substantive decisions as those that “place par-
ticular conduct or persons . . . beyond the State’s power to
punish,” 

or that “prohibi[t] a certain category of
punishment for a class of defendants because of their
status or offense,” 

(internal quota-
tion marks omitted). Cases such as these, in which the
Constitution deprives the Government of the power to
impose the challenged punishment, “represen[t] the clear-
est instance” of substantive rules for which retroactive
application is appropriate. 

(opinion
of Harlan, J.). Drawing on those decisions, amicus argues
that Johnson is not substantive because it does not limit
Congress’ power: Congress is free to enact a new version of
the residual clause that imposes the same punishment on
the same persons for the same conduct, provided the new
statute is precise enough to satisfy due process.
   Although this Court has put great emphasis on substan-
tive decisions that place certain conduct, classes of per-
sons, or punishments beyond the legislative power of
Congress, the Court has also recognized that some sub-
stantive decisions do not impose such restrictions. The
clearest example comes from 

. In Bousley,
the Court was asked to determine what retroactive effect
should be given to its decision in Bailey v. United States,

(1995). Bailey considered the “use” prong of

. §924(c)(1), which imposes increased penalties
on the use of a firearm in relation to certain crimes. The
Court held as a matter of statutory interpretation that the
“use” prong punishes only “active employment of the
firearm” and not mere 

. The
Court in Bousley had no difficulty concluding that Bailey
was substantive, as it was a decision “holding that a sub-
stantive federal criminal statute does not reach certain
conduct.” 

, at 620; see 

, at
                 Cite as: 578 U. S. ____ (2016)          13

                     Opinion of the Court

354 (“A decision that modifies the elements of an offense is
normally substantive rather than procedural”). The Court
reached that conclusion even though Congress could (and
later did) reverse Bailey by amending the statute to cover
possession as well as use. See United States v. O’Brien,

, 232–233 (2010) (discussing statutory
amendment known as the “Bailey fix”). Bousley thus
contradicts the contention that the Teague inquiry turns
only on whether the decision at issue holds that Congress
lacks some substantive power.
   Amicus recognizes that Bousley does not fit the theory
that, in her view, should control this case. She instead
proposes an ad hoc exception, contending that Bousley
“recognized a separate subcategory of substantive rules”
for decisions that interpret statutes (but not those, like
Johnson, that invalidate statutes).        Brief for Court-
Appointed Amicus Curiae in Support of Judgment Below
40. For support, amicus looks to the separation-of-powers
doctrine. Her argument is that statutory construction
cases are substantive because they define what Congress
always intended the law to mean—unlike Johnson, which
struck down the residual clause regardless of Congress’
intent.
   That argument is not persuasive. Neither Bousley nor
any other case from this Court treats statutory interpreta-
tion cases as a special class of decisions that are substan-
tive because they implement the intent of Congress.
Instead, decisions that interpret a statute are substantive
if and when they meet the normal criteria for a substan-
tive rule: when they “alte[r] the range of conduct or the
class of persons that the law punishes.” 

.
   The separation-of-powers argument that amicus raises
is also misplaced. Bousley noted that the separation of
powers prohibits a court from imposing criminal punish-
ment beyond what Congress meant to 

              WELCH v. UNITED STATES

                     Opinion of the Court

620–621 (“[I]t is only Congress, and not the courts, which
can make conduct criminal”). But a court likewise is
prohibited from imposing criminal punishment beyond
what Congress in fact has enacted by a valid law. In
either case a court lacks the power to exact a penalty that
has not been authorized by any valid criminal statute.
   Treating decisions as substantive if they involve statu-
tory interpretation, but not if they involve statutory inval-
idation, would produce unusual outcomes. “It has long
been our practice . . . before striking a federal statute as
impermissibly vague, to consider whether the prescription
is amenable to a limiting construction.” Skilling v. United
States, 

, 405–406 (2010). Amicus acknowl-
edges that a decision that saves a vague statute by adopt-
ing a limiting construction is substantive, so anyone who
falls outside the limiting construction can use that deci-
sion to seek relief on collateral review. But amicus also
contends that, if a court takes the further step of striking
down the whole statute as vague, that decision is proce-
dural, so no one can use it to seek relief on collateral re-
view. That arbitrary distinction has no place in the
Teague framework.
   It should be noted, of course, that not every decision
striking down a statute is ipso facto a substantive deci-
sion. A decision that strikes down a procedural statute—
for example, a statute regulating the types of evidence
that can be presented at trial—would itself be a proce-
dural decision. It would affect only the “manner of determin-
ing the defendant’s culpability,” not the conduct or persons
to be punished. 

(emphasis
deleted). A decision of this kind would have no retroactive
effect under Teague unless it could be considered a “water-
shed” procedural rule. See 

–313.
Johnson, however, struck down part of a criminal statute
that regulates conduct and prescribes punishment. It
thereby altered “the range of conduct or the class of per-
                 Cite as: 578 U. S. ____ (2016)           15

                     Opinion of the Court

sons that the law punishes.” 

. It
follows that Johnson announced a substantive rule that
has retroactive effect in cases on collateral review.
                        *    *     *
  It may well be that the Court of Appeals on remand will
determine on other grounds that the District Court was
correct to deny Welch’s motion to amend his sentence. For
instance, the parties continue to dispute whether Welch’s
strong-arm robbery conviction qualifies as a violent felony
under the elements clause of the Act, which would make
Welch eligible for a 15-year sentence regardless of John-
son. On the present record, however, and in light of to-
day’s holding that Johnson is retroactive in cases on col-
lateral review, reasonable jurists at least could debate
whether Welch is entitled to relief. For these reasons, the
judgment of the Court of Appeals is vacated, and the case
is remanded for further proceedings consistent with this
opinion.
                                            It is so ordered.
                 Cite as: 578 U. S. ____ (2016)            1

                    THOMAS, J., dissenting

SUPREME COURT OF THE UNITED STATES
                         _________________

                         No. 15–6418
                         _________________


GREGORY WELCH, PETITIONER v. UNITED STATES
 ON WRIT OF CERTIORARI TO THE UNITED STATES COURT OF
          APPEALS FOR THE ELEVENTH CIRCUIT
                        [April 18, 2016]

   JUSTICE THOMAS, dissenting.
   Last Term the Court held in Johnson v. United States,
576 U. S. ___ (2015), that because the residual clause of
the Armed Career Criminal Act of 1984 (ACCA), 

. §924(e)(2)(B)(ii), “combin[es] indeterminacy about
how to measure the risk posed by a crime with indetermi-
nacy about how much risk it takes for the crime to qualify
as a violent felony,” it is unconstitutionally vague. 576
U. S., at ___ (slip op., at 6). Federal prisoners then sought
to invoke Johnson as a basis for vacating their sentences
in federal collateral review proceedings. See 

.
§2255(a).
   Today the Court holds that Johnson applies retroactively
to already final sentences of federal prisoners. That
holding comes at a steep price. The majority ignores an
insuperable procedural obstacle: when, as here, a court
fails to rule on a claim not presented in a prisoner’s §2255
motion, there is no error for us to reverse. The majority
also misconstrues the retroactivity framework developed
in Teague v. Lane, 

(1989), and its progeny,
thereby undermining any principled limitation on the
finality of federal convictions. I respectfully dissent.
                            I
 As the majority observes with considerable understate-
ment, “[t]his case comes to the Court in a somewhat un-
2                WELCH v. UNITED STATES

                    THOMAS, J., dissenting

usual procedural posture.” Ante, at 6. This case arises from
petitioner Gregory Welch’s challenge to the Eleventh
Circuit’s denial of a certificate of appealability.
§2253(c)(1). In other words, Welch asks the Court to
review the Eleventh Circuit’s refusal to allow him to ap-
peal the claims he raised in a motion to vacate his sen-
tence and lost in the District Court. But Welch never
claimed that the residual clause was unconstitutionally
vague in his §2255 motion, let alone that Johnson applies
retroactively. Accordingly, courts below addressed neither
issue. Indeed, Johnson was not even decided when the
courts below issued their rulings. Those deficiencies
should preclude us from deciding in this case whether
Johnson is retroactive.
  Our role in reviewing the denial of a certificate of ap-
pealability is far more circumscribed than normal appel-
late review. The text of 

. §2253 confirms this.
Defendants can appeal their convictions and sentences as
a matter of right on direct review, but §2253 deprives
courts of appeals of jurisdiction to review the denial of a
petitioner’s motion for federal postconviction relief unless
he obtains a “certificate of appealability.” §2253(c)(1).
And he can obtain that certificate only if he makes “a
substantial showing of the denial of a constitutional right.”
§2253(c)(2); see Miller-El v. Cockrell, 

, 335–
336 (2003).
  Accordingly, this Court has instructed that review of the
denial of a certificate of appealability is a retrospective
inquiry into whether the movant’s claims, as litigated in
the district court, warrant further proceedings—not
whether there is any conceivable basis upon which the
movant could prevail. Courts must ask whether “reason-
able jurists would find the district court’s assessment of
the constitutional claims debatable or wrong.” Slack v.
McDaniel, 

, 484 (2000) (emphasis added).
They are to “look to the District Court’s application of [the
                 Cite as: 578 U. S. ____ (2016)            3

                    THOMAS, J., dissenting

Antiterrorism and Effective Death Penalty Act of 1996
(AEDPA)] to petitioner’s constitutional claims and ask
whether that resolution was debatable.” 

(emphasis added).
   Until today, we did not require courts of appeals to
consider all possible constitutional issues that might
warrant relief as part of this inquiry. Those courts instead
looked to how the movant framed his case in his motion to
vacate. Even if, for example, a district court denies ha-
beas relief based on procedural default and never reached
the merits, the movant must establish not only that the
procedural ruling is “debatable” but also that his motion
“state[d] a valid claim of the denial of a constitutional
right.” 

.
   Requiring a court of appeals to consider arguments not
raised in a §2255 motion is also at odds with how the
Court has described the certificate-of-appealability in-
quiry. The Court has called the decision whether to grant
a certificate of appealability a “threshold” inquiry that
“forbids” reviewing courts to engage in “full consideration
of the factual or legal bases adduced in support of the
claims.” 

. That is because, in the
Court’s view, the point of “[t]he [certificate of appealabil-
ity] process [is to] scree[n] out issues unworthy of judicial
time and attention and ensur[e] that frivolous claims are
not assigned to merits panels.” Gonzalez v. Thaler, 

, 145 (2012). There would be no surer way to
transform this determination into a full-blown merits
appeal than to require courts of appeals to consider all
conceivable bases for relief that the movant failed to raise.
   Welch’s failure to raise any Johnson-related claim in the
District Court should, therefore, bar the Eleventh Circuit
and this Court from addressing whether Johnson applies
retroactively. Welch’s §2255 motion omitted any claim
4                   WELCH v. UNITED STATES

                        THOMAS, J., dissenting

that his sentence was invalid because the ACCA’s residual
clause is unconstitutionally vague.* Unsurprisingly, the
District Court did not address a vagueness claim that
Welch had not raised. Nor did (or could) the District
Court assess whether Johnson applies retroactively, for
we decided Johnson after the District Court ruling. In
sum, when Welch raised the vagueness of the residual
clause for the first time in his Eleventh Circuit application
for a certificate of appealability, it was too late.
  The Government responds to this issue by attempting,
in its reply brief in this Court, to “expressly waiv[e] any
procedural default defense against petitioner on his John-
son claim.” Reply Brief for United States 22. But this
case has not been framed as one involving a “procedural
default,” which ordinarily refers to the affirmative defense
that a petitioner defaulted his claim in some earlier pro-
ceeding. See McCleskey v. Zant, 

, 490 (1991)
(describing, in the context of a federal habeas petition
brought by a state prisoner, “procedural default” as the
“presumption against [federal] habeas adjudication . . . of
claims defaulted in state court”); see also, e.g., Jenkins v.
Anderson, 

, 234, n. 1 (1980) (noting that
procedural default is an affirmative defense that must be
raised); 

. §2254 Rule 5 (requiring the Govern-
ment to “state whether any claim in the petition is barred
by a . . . procedural bar” in the answer to the motion).
  Welch instead failed to raise that claim in this proceed-
ing by failing to present it in his motion to vacate his
sentence. And the Court of Appeals, when deciding

——————
   * Welch’s §2255 motion did assert that his “robbery under Florida
[statutes] is ambiguous, vague, and was without any violence and or
physical force,” App. 96a, and that Florida robbery “has multi-
meanings.” 

But challenging the vagueness of Florida law is
quite different from the argument Welch needed to assert a Johnson
claim: that the residual clause is itself unconstitutionally vague.
                 Cite as: 578 U. S. ____ (2016)            5

                    THOMAS, J., dissenting

whether to grant a certificate of appealability, cannot be
expected to look beyond the claims presented in that
motion in conducting its threshold inquiry about whether
“reasonable jurists would find the district court’s assess-
ment of the constitutional claims debatable or wrong.”

(emphasis added). Although the
Government purports to waive any forfeiture defense now,
it cannot alter what was before the Court of Appeals.
After Welch failed to raise a Johnson claim in his §2255
motion and the Eleventh Circuit denied a certificate of
appealability, the Government could not inject the claim
into the case.
   Rather than grappling with these issues, the majority
distorts the standard for reviewing certificates of appeal-
ability by asking whether reasonable jurists would debate
the “conclusion” that Welch “failed to show any entitle-
ment to relief.” Ante, at 7. The majority thereby trans-
forms what should be a quick “overview of the claims in
the habeas petition,” 

(emphasis
added), to a searching review of the “conclusion” that a
prisoner is not entitled “any” collateral relief. Ante, at 7
(emphasis added). That is preposterous. The Eleventh
Circuit, according to the majority, erred by denying Welch
the opportunity to “appeal” a claim that he failed to raise,
in part because a decision that did not yet exist when the
Eleventh Circuit ruled may entitle him to relief. The
majority’s view of AEDPA demands judicial clairvoyance;
Courts of Appeals can avoid reversal only by inventing
arguments on the movant’s behalf.
                             II
  After bypassing what should have been an insurmount-
able procedural hurdle, the majority then gets the merits
wrong. The retroactivity rules the Court adopted in
Teague v. Lane, 

, generally foreclose prison-
ers from collaterally challenging their convictions based on
6                WELCH v. UNITED STATES

                    THOMAS, J., dissenting

new decisions that postdate their convictions and sentences.
The only exceptions to that bar are for decisions that
announce a new substantive rule or a new “watershed”
procedural rule. See ante, at 7–8. All agree that Johnson
announced a new rule and that it is not a “watershed”
procedural rule. See ante, at 8. But the rule in Johnson
also does not satisfy our criteria for substantive rules. The
majority concludes otherwise, ante, at 8–9, but its ap-
proach fails under Teague’s own terms and erodes any
meaningful limits on what a “substantive” rule is.
                            A
  The Court has identified two types of substantive rules,
and Johnson’s rule of decision fits neither description. It
is not a new substantive constitutional rule, nor does it
narrow the scope of a criminal statute through statutory
construction.
                             1
   Time and again, the Court has articulated the test for
defining a substantive constitutional rule as follows: The
rule must “place particular conduct or persons covered by
the statute beyond the State’s power to punish.” Schriro
v. Summerlin, 

, 352 (2004); see also Beard v.
Banks, 

, 416 (2004) (similar); Penry v.
Lynaugh, 

, 330 (1989) (rule is substantive if
“the Constitution itself deprives the State of the power to
impose a certain penalty”). This is also the test the Court
has purported to apply in case after case. See, e.g., Sawyer
v. Smith, 

, 233, 241 (1990) (prohibiting pros-
ecutors from misleading the jury to believe that it was not
responsible for a death sentence was a nonsubstantive
rule that did not “place an entire category of primary
conduct beyond the reach of the criminal law” or “prohibit
imposition of a certain type of punishment for a class of
defendants because of their status or offense”). Our prec-
                  Cite as: 578 U. S. ____ (2016)             7

                     THOMAS, J., dissenting

edents thus make clear the distinction between substan-
tive and nonsubstantive constitutional rules. A rule that
“because [a State] has made a certain fact essential to the
death penalty, that fact must be found by a jury,” is not
substantive; it had no effect on the “range of conduct a
State may criminalize.” 

–354.
But a rule in which this Court “ma[de] a certain fact es-
sential to the death penalty . . . would be substantive”; it
would change the range of conduct warranting a death
sentence. 

   Under these principles, Johnson announced a new
constitutional rule, but it is not substantive. Johnson’s
new constitutional rule is that a law is unconstitutionally
vague if it “requires a court to picture the kind of conduct
that the crime involves in ‘the ordinary case,’ and to judge
whether that abstraction presents a serious potential risk”
of some result. 576 U. S., at ___ (slip op., at 4). Such laws
are vague because they simultaneously create “indetermi-
nacy about how to measure the risk posed by a crime with
indeterminacy about how much risk it takes for the crime
to qualify” as the described offense. Id., at ___ (slip op., at
6). Together, those two indeterminacies “produc[e] more
unpredictability and arbitrariness than the Due Process
Clause tolerates.” 

is not substantive under our precedents.
It does not preclude the Government from prohibiting
particular conduct or deem any conduct constitutionally
protected. The Government remains as free to enhance
sentences for federal crimes based on the commission of
previous violent felonies after Johnson as it was before.
Cf. Butler v. McKellar, 

, 415 (1990) (deciding
that a new rule was not substantive because “[t]he pro-
scribed conduct” was “capital murder, the prosecution of
which is, to put it mildly, not prohibited by the rule”). Nor
does Johnson’s vagueness rule place any class of persons
or punishment off limits. There is no category of offender
8                WELCH v. UNITED STATES

                    THOMAS, J., dissenting

that Congress cannot subject to an enhanced sentence
after Johnson. See James v. United States, 

,
230 (2007) (Scalia, J., dissenting) (Congress “very easily”
could “subjec[t] all repeat offenders to a 15-year manda-
tory minimum prison term” in differently worded statute
(emphasis deleted)). The only constraint Johnson imposes
is on the manner in which the Government can punish
offenders. To avoid “fail[ing] to give ordinary people fair
notice” or “invit[ing] arbitrary enforcement,” 576 U. S., at
___ (slip op., at 3), Congress must be clearer in describ-
ing what conduct “otherwise . . . presents a serious poten-
tial risk of physical injury to another.” 

.
§924(e)(2)(B)(ii).
                              2
   Johnson also does not fit within the second type of
substantive rule this Court has recognized, which consists
of “decisions that narrow the scope of a criminal statute
by interpreting its terms.” 

;
see 

(contrasting these rules with “constitu-
tional determinations” that rule out punishing conduct or
persons).
   The Court has invoked this subset of new rules just
once, in Bousley v. United States, 

(1998).
Bousley held that Bailey v. United States, 

(1995), which interpreted a federal firearms sentencing
enhancement to require proof of “ ‘active employment of
the firearm’ ” as an element, applied 

–617. The Court explained that Teague’s bar
on retroactively applying “procedural rules” is “inapplic-
able to the situation in which this Court decides the mean-
ing of a criminal statute enacted by 

. Moreover, the Court reasoned, “decisions of this
Court holding that a substantive federal criminal statute
does not reach certain conduct” share a key commonality
with “decisions placing conduct beyond the power of the
                 Cite as: 578 U. S. ____ (2016)           9

                    THOMAS, J., dissenting

criminal law-making authority to proscribe”: both “neces-
sarily carry a significant risk that a defendant stands
convicted of an act that the law does not make criminal.”

omitted). The Court thus
classified decisions “holding that a substantive federal
criminal statute does not reach certain conduct” as sub-
stantive. 

so readily assume that Bousley applies here.
Until today, Bousley applied only to new rules reinterpret-
ing the text of federal criminal statutes in a way that
narrows their reach. Johnson announced no such rule. It
announced only that there is no way in which to narrow
the reach of the residual clause without running afoul
of the Due Process Clause. 576 U. S., at ___–___ (slip op.,
at 10–12).
   The majority protests that applying different retroactiv-
ity principles to constitutional and statutory rules pro-
duces “unusual outcomes” because a decision interpreting a
statute’s text to narrow its scope may be retroactive, while
a decision declaring the provision unconstitutional might
not be. Ante, at 14. But such outcomes are an inevitable
byproduct of the Court’s retroactivity jurisprudence, not a
unique consequence of this case. Take a statute allowing
the Federal Government to prosecute defendants for “seri-
ous crimes involving interstate commerce” of which they
were acquitted in state court. See Bartkus v. Illinois, 

(1959) (no double jeopardy bar to such prosecu-
tions). Suppose the Court, concerned that there might be
a double jeopardy problem after all, narrowed the meaning
of “serious crimes involving interstate commerce” to en-
compass only felonies that would be subject to a statutory
maximum sentence of life imprisonment. Anyone acquit-
ted of a minor felony by the State but convicted by the
Federal Government before this Court’s narrowing con-
struction could, in the majority’s view, seek relief on col-
lateral review under Bousley. See ante, at 14. But if the
10               WELCH v. UNITED STATES

                    THOMAS, J., dissenting

Court ruled that the Double Jeopardy Clause bars all
federal reprosecutions, I doubt that rule would be retroac-
tive. That rule dictates when a defendant may be tried
and convicted of certain conduct—not the substance of the
crime for which the defendant is tried, or the punishment
imposed.
   The Court’s historical justifications for retroactivity
underscore the reasons for treating statutory and constitu-
tional rules differently. The Court in the 1950’s “ex-
tend[ed] the scope of habeas to all alleged constitutional
errors” to “forc[e] trial and appellate courts in both the
federal and state system to toe the constitutional mark” in
the face of perceived systemic violations. Mackey v. United
States, 

, 687 (1971) (Harlan, J., concurring
in judgments in part and dissenting in part). That devel-
opment led to the Teague framework allowing retroactivity
for certain types of constitutional rules. See Montgomery
v. Louisiana, 577 U. S. ___, ___–___ (2016) (Scalia, J.,
dissenting) (slip op., at 2–3) (recounting history). But this
Court has never suggested that lower courts had similar
difficulties in interpreting the reach of criminal statutes,
such that the retroactivity rules should be the same.
Rather, the history suggests that the failure to apply a
narrowing construction of a criminal statute is a qualita-
tively different type of error.
                              B
  The majority instead determines whether a rule is
substantive by looking to the “function of the rule,” ante,
at 10, and asking “whether the new rule itself has a proce-
dural function or a substantive function,” ante, at 11. This
apparently means that courts should divine the effect of a
new rule and decide whether that effect alters the sub-
stantive elements of a crime or sentence. All that matters,
the majority says, is that the vagueness rule announced in
Johnson had the effect of invalidating the residual clause
                   Cite as: 578 U. S. ____ (2016)              11

                      THOMAS, J., dissenting

and, as a result of its invalidation, the residual clause “can
no longer mandate or authorize any sentence.” Ante, at 9
(“striking down the residual clause” is what “changed the
substantive reach of [ACCA]”).
   That approach is untenable. It brushes aside the rule of
decision, which is where all of our prior precedents begin
and end for purposes of applying Teague. When deciding
whether rules are substantive, our cases have homed in on
the rule that would apply not just to the specific statute at
hand, but in similar, future circumstances. Thus, just this
Term, the Court defined the rule announced in Miller v.
Alabama, 567 U. S. ___ (2012), as: The Eighth Amend-
ment “prohibit[s] . . . mandatory life without parole for
juvenile offenders”—not that Alabama’s juvenile-
sentencing statute flouts the Eighth Amendment. Mont-

, at ___ (slip op., at 14). Likewise, the rule
announced in Ring v. Arizona, 

(2002), was
that “a sentencing judge, sitting without a jury, [may not]
find an aggravating circumstance necessary for imposition
of the death penalty”—not that provisions of Arizona’s
death penalty statute violate the Sixth Amendment.

(internal quotation marks omit-
ted; alteration in original). By jettisoning that approach
and focusing solely on Johnson’s effect (the invalidation
of the residual clause), the majority departs from our
precedents.
   The majority’s focus on the effect of a decision breaks
down all meaningful distinctions between “new” and “old”
rules, or “substantive” and “procedural” ones. The first
step of the Teague inquiry assesses whether the rule
is “new” by looking to whether prior precedents dictated
the rule of decision—not its effects. See, e.g., Chaidez v.
United States, 568 U. S. ___, ___ (2013) (slip op., at 4)
(internal quotation marks omitted). But if, as the majority
contends, the “function of the rule,” ante, at 10, is the relevant
baseline, then every case invalidating a statute or a sen-
12               WELCH v. UNITED STATES

                    THOMAS, J., dissenting

tence establishes a “new” rule. The law’s invalidation
would be a departure from any prior decision that inter-
preted the law as if it were operative. Likewise, if any
decision has the effect of invalidating substantive provi-
sions of a criminal statute, it is a substantive rule under
the majority’s approach no matter what the reason for the
statute’s invalidation.
   The majority denies that “every decision striking down a
statute is ipso facto a substantive decision,” saying that
only when a decision invalidates a provision that “regu-
lates conduct and prescribes punishment” is it retroactive.
Ante, at 14. But that still transforms innumerable proce-
dural rules into substantive ones. Take a state law that
defines the crime of robbery and specifies that only 10 of
the 12 jurors need to vote to convict someone of that crime.
If this Court were to reverse Apodaca v. Oregon, 

(1972), and hold that the Sixth Amendment requires
unanimous jury verdicts, the portion of the statute allow-
ing nonunanimity would be invalid. But assume that the
state statute was nonseverable: the Court’s jury unanim-
ity rule, undoubtedly “procedural,” would have the effect of
invalidating not only the portion of the state statute re-
garding unanimity but also the provision defining the
crime of robbery, a provision that “regulates conduct.”
Ante, at 14. To the majority, these effects would make the
rule substantive. That approach is mistaken, and would
also produce arbitrary results. Suppose most States had
similar statutes, but that some had robust severability
provisions and others did not. In some States, the decision
would be procedural; elsewhere, it would be substantive,
producing a patchwork of statute-specific outcomes.
   Finally, the majority flips Teague on its head with its
alternative contention that Johnson must have announced
a substantive rule because it is “not a procedural decision.”
Ante, at 9. Teague is a general rule against retroactivity,
see ante, at 7–8, and all new rules are barred unless they
                  Cite as: 578 U. S. ____ (2016)           13

                     THOMAS, J., dissenting

fit within the exceptions for substantive or “watershed”
procedural 

–311 (plurality opin-
ion). To say that a rule is “not . . . procedural” is not
enough. And even if it were, the rule in Johnson fits
better within the Court’s descriptions of procedural rules
than substantive ones. “Procedural rules . . . are designed
to enhance the accuracy of a conviction or sentence by
regulating ‘the manner of determining the defendant’s
culpability.’ ” Montgomery, 577 U. S., at ___ (slip op., at 9)
(quoting 

). And the rule in Johnson
regulates only the manner in which Congress defined a
sentencing enhancement, not the conduct that triggers the
punishment. E.g., Smith v. Goguen, 

, 572–
573 (1994) (vagueness doctrine “requires legislatures to
set reasonably clear guidelines” to give “fair notice or
warning” and “prevent arbitrary and discriminatory en-
forcement” (internal quotation marks omitted)).
                              III
   Today’s opinion underscores a larger problem with our
retroactivity doctrine: The Court’s retroactivity rules have
become unmoored from the limiting principles that the
Court invoked to justify the doctrine’s existence. Under
Teague itself, the question whether Johnson applies retro-
actively would be a straightforward “No.” If this question
is close now, that is only because the Court keeps moving
the goalposts.
   As the majority observes, the foundations of our ap-
proach to retroactivity in collateral review come from
Justice Harlan’s separate opinions in Desist v. United
States, 

(1969), and Mackey v. United States,

. Ante, at 7. There, Justice Harlan con-
fronted a now-familiar problem: how to address the con-
sequences of an ever-evolving Constitution. He responded
with an approach to retroactivity that placed at the fore-
front the need for finality in the criminal process. See, e.g,
14                WELCH v. UNITED STATES

                     THOMAS, J., 

–683. In his view, very few rules that
emerged after a prisoner’s conviction—including constitu-
tional rules—warranted disturbing that conviction. See

Justice Harlan saw only “two exceptions”:
“bedrock procedural” rules, 

and “[n]ew
‘substantive due process’ rules” removing “certain kinds of
primary, private individual conduct beyond the power of
the criminal law-making authority to proscribe,” 

As examples of the latter category, he cited such
rules as that the First Amendment forbids criminalizing
flag burning, that the right to privacy precludes the Gov-
ernment from prosecuting distributors of contraception,
and that the “freedom to marry” and equal protection
principles immunize couples from being punished for
entering into interracial marriages. 

n. 7.
These “ ‘substantive due process’ rules,” Justice Harlan
explained, were “on a different footing” because “the writ
has historically been available for attacking convictions on
such grounds.” 

Moreover, society has an
“obvious interest in freeing individuals for punishment for
conduct that is constitutionally protected.” 

And granting relief for such claims would not require
retrials. 

Justice Harlan’s approach, 

(plurality opinion), it agreed that to
preserve “the principle of finality which is essential to the
operation of our criminal justice system,” 

“new
rules generally should not be applied retroactively to cases
on collateral review,” 

Teague thus adopted
Justice Harlan’s two exceptions for “watershed rules of
criminal procedure” and rules that “accord constitutional
protection to . . . primary activity.” 

see 

   The Court then swiftly discarded the limitations that
Teague adopted. Penry proclaimed the retroactivity of
rules barring certain punishments, even though the
                  Cite as: 578 U. S. ____ (2016)           15

                     THOMAS, J., dissenting

Court’s constant revision of the Eighth Amendment pro-
duces an “ever-moving target of impermissible punish-
ments.” Montgomery, 577 U. S., at ___ (slip op., at 11)
(Scalia, J., dissenting) (emphasis deleted); see id., at ___–
___ (slip op., at 10–11). Bousley extended retroactive relief
for federal prisoners raising statutory claims, not just
constitutional ones. 

–617, 620–621.
Montgomery extended Teague to state postconviction
proceedings, enshrined Teague as a constitutional com-
mand, and redefined substantive rules to include rules
that require sentencers to follow certain procedures in
punishing juveniles. Now the majority collapses Teague’s
substantive-procedural distinction further, allowing any
rule that has the incidental effect of invalidating substan-
tive provisions of a criminal statute to become a substan-
tive rule.
   Today’s decision, like those that preceded it, professes to
venerate Justice Harlan’s theory of retroactivity. See
ante, at 7; 

, at ___ (slip op., at 8–10).
This rings hollow; these decisions spell its ruin. The Court
adopted Justice Harlan’s approach to retroactivity because
it shared his conviction that “there [must] be a visible end
to the litigable aspect of the criminal process.” 

; see 

(plurality opinion)
(similar). With the Court’s unprincipled expansion of
Teague, every end is instead a new beginning.
                        *    *     * 

  For these reasons, I respectfully dissent.