SHEPARD V. UNITED STATES (03-9168) 544 U.S. 13 (2005)
348 F.3d 308, reversed and remanded.
[ Souter ]
[ Opinion of Thomas ]
[ O’Connor ]
HTML version
PDF version
HTML version
PDF version
HTML version
PDF version
HTML version
PDF version

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, Washington, D. C. 20543, of any typographical or other formal errors, in order that corrections may be made before the preliminary print goes to press.


No. 03—9168



[March 7, 2005]

Justice Souter delivered the opinion of the Court, except as to Part III.

Title 18 U.S.C. § 924(e) (2000 ed. and Supp. II), popularly known as the Armed Career Criminal Act (ACCA), mandates a minimum 15-year prison sentence for anyone possessing a firearm after three prior convictions for serious drug offenses or violent felonies. The Act makes burglary a violent felony only if committed in a building or enclosed space (“generic burglary”), not in a boat or motor vehicle. In Taylor v. United States, 495 U.S. 575 (1990), we held that a court sentencing under the ACCA could look to statutory elements, charging documents, and jury instructions to determine whether an earlier conviction after trial was for generic burglary. The question here is whether a sentencing court can look to police reports or complaint applications to determine whether an earlier guilty plea necessarily admitted, and supported a conviction for, generic burglary. We hold that it may not, and that a later court determining the character of an admit-
ted burglary is generally limited to examining the statutory definition, charging document, written plea agreement, transcript of plea colloquy, and any explicit factual finding by the trial judge to which the defendant assented.


Petitioner Reginald Shepard was indicted under 18 U.S.C. § 922(g)(1), barring felons from possessing a firearm, and pleaded guilty. At sentencing the Government claimed that Shepard’s prior convictions raised his sentencing range from between 30 and 37 months (under the United States Sentencing Guidelines) to the 15-year minimum required by §924(e), pointing to four prior convictions entered upon Shepard’s pleas of guilty under one of Massachusetts’s two burglary statutes.1 Whereas the Government said that each conviction represented a predicate ACCA offense of generic burglary, the District Court ruled that Taylor barred counting any of the prior convictions as predicates for the mandatory minimum. 125 F. Supp. 2d 562, 569 (Mass. 2000).

In Taylor we read the listing of “burglary” as a predicate “violent felony” (in the ACCA) to refer to what we called “generic burglary,” an “unlawful or unprivileged entry into, or remaining in, a building or structure, with intent to commit a crime.” 495 U.S., at 599. Because statutes in some States (like Massachusetts) define burglary more broadly, as by extending it to entries into boats and cars, we had to consider how a later court sentencing under the ACCA might tell whether a prior burglary conviction was for the generic offense.2 We held that the ACCA generally prohibits the later court from delving into particular facts disclosed by the record of conviction, thus leaving the court normally to “look only to the fact of conviction and the statutory definition of the prior offense.” Id., at 602. We recognized an exception to this “categorical approach” only for “a narrow range of cases where a jury [in a State with a broader definition of burglary] was actually required to find all the elements of” the generic offense. Ibid. We held the exception applicable “if the indictment or information and jury instructions show that the defendant was charged only with a burglary of a building, and that the jury necessarily had to find an entry of a building to convict … .” Ibid. Only then might a conviction
under a “nongeneric” burglary statute qualify as an ACCA predicate.

In this case, the offenses charged in state complaints were broader than generic burglary, and there were of course no jury instructions that might have narrowed the charges to the generic limit. The Government nonetheless urged the District Court to examine reports submitted by the police with applications for issuance of the complaints, as a way of telling whether Shepard’s guilty pleas went to generic burglaries notwithstanding the broader descriptions of the offenses in the complaints, descriptions that tracked the more expansive definition in Massachusetts law. The court concluded that Taylor forbade this, and that investigation within the Taylor limits failed to show that Shepard had three generic burglary convictions. The court accordingly refused to consider the 15-year mandatory minimum, though it did sentence Shepard somewhat above the standard level under the Sentencing Guidelines, on the ground that his criminal history category under the Guidelines did not do justice to his ample criminal record.

On appeal the First Circuit, following its earlier decision in United States v. Harris, 964 F.2d 1234 (1992), vacated the sentence and ruled that complaint applications and police reports may count as “sufficiently reliable evidence for determining whether a defendant’s plea of guilty constitutes an admission to a generically violent crime,” 231 F.3d 56, 67 (2000). As to each of Shepard’s prior convictions, the court remanded the case for the District Court to determine whether there was “sufficiently reliable evidence that the government and the defendant shared the belief that the defendant was pleading guilty to a generically violent crime.” Id., at 70.

The District Court again declined to impose the 15-year mandatory minimum, even though the Government supplemented its earlier submission with police reports or complaint applications on two additional burglary convictions. The District Judge noted that the only account of what occurred at each of the prior plea hearings came from an affidavit submitted by Shepard, who stated “that none of the details in th[e police] reports w[as] ever mentioned at his pleas,” that “the reports themselves were never read by the judge to him during the plea colloquy,” and that at no time “was he ever asked if the information contained in the … [r]eports w[as] true.” 181 F. Supp. 2d 14, 19 (Mass. 2002). Shepard further swore that “with respect to each report: [he] did not admit the truth of the information contained in the … [r]eport as part of [his] plea and [had] never admitted in court the facts alleged in the report … .” Id., at 19—20 (internal quotation marks omitted). Based on this, the District Court found that the Government had failed to carry its burden to demonstrate that Shepard had pleaded to three generic burglaries.

The Court of Appeals again vacated the sentence. After observing that Shepard had never “seriously disputed” that he did in fact break into the buildings described in the police reports or complaint applications, 348 F.3d 308, 311 (2003), the court rejected the District Court’s conclusion that the Government had not shown the requisite predicate offenses for the 15-year minimum sentence, id., at 314. The case was remanded with instructions to impose that sentence.

We granted certiorari, 542 U.S. ___ (2004), to address divergent decisions in the Courts of Appeals applying Taylor when prior convictions stem from guilty pleas, not jury verdicts. We now reverse.


We agree with the First Circuit (and every other Court of Appeals to speak on the matter) that guilty pleas may establish ACCA predicate offenses and that Taylor’s reasoning controls the identification of generic convictions following pleas, as well as convictions on verdicts, in States with nongeneric offenses. See 348 F.3d, at 312, n. 4 (citing cases). Shepard wisely refrains from challenging this position, for the ACCA nowhere provides that convictions in tried and pleaded cases are to be regarded differently. It drops no hint that Congress contemplated different standards for establishing the fact of prior convictions, turning on the basis of trial or plea. Nothing to that effect is suggested, after all, by the language imposing the categorical approach, which refers to predicate offenses in terms not of prior conduct but of prior “convictions” and the “element[s]” of crimes. Taylor, 495 U.S., at 600—601 (citing 18 U.S.C. § 924(e)). Nor does the Act’s legislative history reveal a lesser congressional preference for a categorical, as distinct from fact-specific, approach to recognizing ACCA predicates in cases resolved by plea. Taylor, 495 U.S., at 601. And certainly, “the practical difficulties and potential unfairness of a factual approach are daunting,” ibid., no less in pleaded than in litigated cases. Finally, nothing in Taylor’s rationale limits it to prior jury convictions; our discussion of the practical difficulties inherent in looking into underlying circumstances spoke specifically of “cases where the defendant pleaded guilty, [in which] there often is no record of the underlying facts.” Ibid. Our job, then, is to find the right analogs for applying the Taylor rule to pleaded cases.

The Taylor Court drew a pragmatic conclusion about the best way to identify generic convictions in jury cases, while respecting Congress’s adoption of a categorical criterion that avoids subsequent evidentiary enquiries into the factual basis for the earlier conviction. The Court held that generic burglary could be identified only by referring to charging documents filed in the court of conviction, or to recorded judicial acts of that court limiting convictions to the generic category, as in giving instruction to the jury.

The Court did not, however, purport to limit adequate judicial record evidence strictly to charges and instructions, id., at 602 (discussing the use of these documents as an “example”), since a conviction might follow trial to a judge alone or a plea of guilty. In cases tried without a jury, the closest analogs to jury instructions would be a bench-trial judge’s formal rulings of law and findings of fact, and in pleaded cases they would be the statement of factual basis for the charge, Fed. Rule Crim. Proc. 11(a)(3), shown by a transcript of plea colloquy or by written plea agreement presented to the court, or by a record of comparable findings of fact adopted by the defendant upon entering the plea.3 With such material in a pleaded case, a later court could generally tell whether the plea had “necessarily” rested on the fact identifying the burglary as generic, Taylor, supra, at 602, just as the details of instructions could support that conclusion in the jury case, or the details of a generically limited charging document would do in any sort of case.

The Government argues for a wider evidentiary cast, however, going beyond conclusive records made or used in adjudicating guilt and looking to documents submitted to lower courts even prior to charges. It argues for considering a police report submitted to a local court as grounds for issuing a complaint under a nongeneric statute; if that report alleges facts that would satisfy the elements of a generic statute, the report should suffice to show that a later plea and conviction were for a predicate offense under the ACCA. There would be no reason for concern about unavailable witnesses or stale memories, the Government points out, and such limited enquiry would be consistent with Taylor because “[t]he underlying purpose [would be] the same as in examining the charging paper and jury instructions (which the Court endorsed in Taylor): to determine the nature of the offense of which petitioner was convicted, rather than to determine what he actually did.” Brief for United States 22—23. The Government stresses three points.

First, it says that the more accommodating view of evidence competent to prove that the plea was to a generic offense will yield reliable conclusions. Although the records of Shepard’s pleas with their notations that he “[a]dmit[ted] suff[icient] facts” do not necessarily show that he admitted entering buildings or structures, as would be true under a generic burglary statute or charge, the police reports suffice to show that the record of admitting sufficient facts “can only have plausibly rested on petitioner’s entry of a building.” Id., at 25.

Second, the Government pulls a little closer to Taylor’s demand for certainty when identifying a generic offense by emphasizing that the records of the prior convictions used in this case are in each instance free from any inconsistent, competing evidence on the pivotal issue of fact separating generic from nongeneric burglary. “[T]here is nothing in the record to indicate that petitioner had pleaded guilty based on entering a ship or vehicle on any of the occasions at issue.” Brief for United States 16.

Finally, the Government supports its call for a more inclusive standard of competent evidence by invoking the virtue of a nationwide application of a federal statute unaffected by idiosyncrasies of record keeping in any particular State. A bar on review of documents like police reports and complaint applications would often make the ACCA sentencing enhancement “hinge on the happenstance of state court record-keeping practices and the vagaries of state prosecutors’ charging practices.” Brief in Opposition 13 (internal quotation marks omitted).

On each point, however, the Government’s position raises an uncomfortable implication: every one of its arguments could have been pressed in favor of an enquiry beyond what Taylor allows when a jury conviction follows nongeneric instructions, and each is therefore as much a menace to Taylor as a justification for an expansive approach to showing whether a guilty plea admitted the generic crime. If the transcript of a jury trial showed testimony about a building break, one could say that the jury’s verdict rested on a finding to that effect. If the trial record showed no evidence of felonious entrance to anything but a building or structure, the odds that the offense actually committed was generic burglary would be a turf accountant’s dream. And, again, if it were significant that vagaries of abbreviated plea records could limit the application of the ACCA, the significance would be no less when the disputed, predicate conviction followed a jury trial and the stenographic notes of the charge had been thrown away.

The Government’s position thus amounts to a call to ease away from the Taylor conclusion, that respect for congressional intent and avoidance of collateral trials require that evidence of generic conviction be confined to records of the convicting court approaching the certainty of the record of conviction in a generic crime State. But that limitation was the heart of the decision, and we cannot have Taylor and the Government’s position both.

There is not, however, any sufficient justification for upsetting precedent here. We are, after all, dealing with an issue of statutory interpretation, see, e.g., Taylor, 495 U.S., at 602, and the claim to adhere to case law is generally powerful once a decision has settled statutory meaning, see Patterson v. McLean Credit Union, 491 U.S. 164, 172—173 (1989) (“Considerations of stare decisis have special force in the area of statutory interpretation, for here, unlike in the context of constitutional interpretation, the legislative power is implicated, and Congress remains free to alter what we have done”). In this instance, time has enhanced even the usual precedential force, nearly 15 years having passed since Taylor came down, without any action by Congress to modify the statute as subject to our understanding that it allowed only a restricted look beyond the record of conviction under a nongeneric statute.4


Developments in the law since Taylor, and since the First Circuit’s decision in Harris, provide a further reason to adhere to the demanding requirement that any sentence under the ACCA rest on a showing that a prior conviction “necessarily” involved (and a prior plea necessarily admitted) facts equating to generic burglary. The Taylor Court, indeed, was prescient in its discussion of problems that would follow from allowing a broader evidentiary enquiry. “If the sentencing court were to conclude, from its own review of the record, that the defendant [who was convicted under a nongeneric burglary statute] actually committed a generic burglary, could the defendant challenge this conclusion as abridging his right to a jury trial?” Taylor, supra, at 601. The Court thus anticipated the very rule later imposed for the sake of preserving the Sixth Amendment right, that any fact other than a prior conviction sufficient to raise the limit of the possible federal sentence must be found by a jury, in the absence of any waiver of rights by the defendant. Jones v. United States, 526 U.S. 227, 243, n. 6 (1999); see also Apprendi v. New Jersey, 530 U.S. 466, 490 (2000).

The Government dismisses the relevance of the Jones-Apprendi implementation of the jury right here by describing the determination necessary to apply the ACCA as “involv[ing] only an assessment of what the state court itself already has been ‘required to find’ in order to find

the defendant guilty.” Brief for United States 38 (quoting Taylor, supra, at 602). But it is not that simple. The problem is that “what the state court … has been ‘required to find’ ” is debatable. In a nongeneric State, the fact necessary to show a generic crime is not established by the record of conviction as it would be in a generic State when a judicial finding of a disputed prior conviction is made on the authority of Almendarez-Torres v. United States, 523 U.S. 224 (1998). The state statute requires no finding of generic burglary, and without a charging document that narrows the charge to generic limits, the only certainty of a generic finding lies in jury instructions, or bench trial findings and rulings, or (in a pleaded case) in the defendant’s own admissions or accepted findings of fact confirming the factual basis for a valid plea. In this particular pleaded case, the record is silent on the generic element, there being no plea agreement or recorded colloquy in which Shepard admitted the generic fact.

Instead, the sentencing judge considering the ACCA enhancement would (on the Government’s view) make a disputed finding of fact about what the defendant and state judge must have understood as the factual basis of the prior plea, and the dispute raises the concern underlying Jones and Apprendi: the Sixth and Fourteenth Amendments guarantee a jury standing between a defendant and the power of the state, and they guarantee a jury’s finding of any disputed fact essential to increase the ceiling of a potential sentence. While the disputed fact here can be described as a fact about a prior conviction, it is too far removed from the conclusive significance of a prior judicial record, and too much like the findings subject to Jones and Apprendi, to say that Almendarez-Torres clearly authorizes a judge to resolve the dispute. The rule of reading statutes to avoid serious risks of unconstitutionality, see Jones, supra, at 239, therefore counsels us to limit the scope of judicial factfinding on the disputed generic character of a prior plea, just as Taylor constrained judicial findings about the generic implication of a jury’s verdict.5


We hold that enquiry under the ACCA to determine whether a plea of guilty to burglary defined by a nongeneric statute necessarily admitted elements of the generic offense is limited to the terms of the charging document, the terms of a plea agreement or transcript of colloquy between judge and defendant in which the factual basis for the plea was confirmed by the defendant, or to some comparable judicial record of this information.

The judgment of the Court of Appeals is reversed, and the case is remanded for further proceedings.

It is so ordered.

The Chief Justice took no part in the decision of this case.


1. The Government initially cited a fifth prior burglary conviction, but after failing to obtain adequate documentation about this conviction the Government focused on the other four.

2. Although Taylor involved prior burglaries, as this case does, our holding in Taylor covered other predicate ACCA offenses. 495 U.S., at 600.

3. Several Courts of Appeals have taken a similar view, approving the use of some or all of these documents. United States v. Bonat, 106 F.3d 1472, 1476—1477 (CA9 1997); United States v. Maness, 23 F.3d 1006, 1009—1010 (CA6 1994); United States v. Smith, 10 F.3d 724, 733—734 (CA10 1993) (per curiam) (construing United States Sentencing Commission, Guidelines Manual §4B1.2 (Nov. 1990)).

4. Like the Government, the dissent would allow district courts to examine a wider range of documents than we approve today, and its proposal is no more consistent with Taylor than the Government’s. Taylor is clear that any enquiry beyond statute and charging document must be narrowly restricted to implement the object of the statute and avoid evidentiary disputes. In the case before it, the Court drew the line after allowing courts to review documents showing “that the jury necessarily had to find an entry of a building to convict.” 495 U.S., at 602; see also ibid. (permitting a sentencing court to look beyond the state statute “in a narrow range of cases where a jury was actually required to find all the elements of generic burglary”). As we say in the text, there are certainly jury trials with record documents like those at issue here, never introduced at trial but “uncontradicted,” post, at 3 (opinion of O’Connor, J.), and “internally consistent,” ibid., with the evidence that came in. The dissent would presumably permit examination of such documents, but Taylor assuredly does not. The only way to reconcile the dissent’s approach with Taylor is to say that in Taylor the prior convictions followed jury verdicts while in this case each prior conviction grew out of a guilty plea. See post, at 9 (“Taylor itself set no rule for guilty pleas”). But Taylor has no suggestion that its reasoning would not apply in plea cases, and its discussion of the practical difficulties specifically referred to prior guilty pleas. 495 U.S., at 601. Moreover, as we have noted, see supra, at 5, and as the dissent nowhere disputes, the ACCA provides no support for such a distinction. We decline to create a distinction that Congress evidently had no desire to draw, that Taylor did not envision, and that we would be hard pressed to explain.

5. The dissent charges that our decision may portend the extension of Apprendi v. New Jersey, 530 U.S. 466 (2000), to proof of prior convictions, a move which (if it should occur) “surely will do no favors for future defendants in Shepard’s shoes.” Post, at 11. According to the dissent, the Government, bearing the burden of proving the defendant’s prior burglaries to the jury, would then have the right to introduce evidence of those burglaries at trial, and so threaten severe prejudice to the defendant. It is up to the future to show whether the dissent is good prophesy, but the dissent’s apprehensiveness can be resolved right now, for if the dissent turns out to be right that Apprendi will reach further, any defendant who feels that the risk of prejudice is too high can waive the right to have a jury decide questions about his prior convictions.