Government “Pivot” Advocacy and Plain-Error Limits on Plea-Agreement Breach Claims at Sentencing

1. Introduction

In United States v. William Manderville (3d Cir. Aug. 13, 2026) (not precedential), the Third Circuit affirmed William Manderville’s conviction and 108-month sentence following guilty pleas to methamphetamine- and heroin-trafficking offenses under 21 U.S.C. §§ 846 and 841(a)(1), (b)(1)(B).

The plea agreement included a stipulated “reasonable” sentencing range of 60 to 84 months and an agreement not to appeal a sentence within that range. The District Court, however, calculated an advisory Guidelines range of 151 to 188 months and imposed 108 months—well above the stipulated range but substantially below the Guidelines.

On appeal, Manderville argued (i) the Government breached the plea agreement at sentencing by “poison[ing] the well” with aggravating rhetoric and by “pitch[ing]” for a “significant” sentence, and (ii) his plea was not knowing and voluntary because he believed he was guaranteed 5 to 7 years.

2. Summary of the Opinion

The Third Circuit rejected both challenges under plain-error review. On the alleged breach, the Court held that the Government’s remarks were not clearly inconsistent with the plea agreement—especially given that the Government expressly requested a downward variance to the top of the stipulated range and “pivoted” to mitigating facts. Even assuming error, any failure by the District Court to find a breach was not “clear or obvious.”

On the voluntariness challenge, the Court held the record foreclosed Manderville’s claim: the plea agreement and colloquy repeatedly stated sentencing was “within the sole discretion of the sentencing judge” and the stipulated range did not bind the Court. Manderville did not show a reasonable probability he would have gone to trial absent any alleged error.

3. Analysis

A. Precedents Cited

  • Anders v. California, 386 U.S. 738 (1967)
    The decision’s procedural backdrop involved Anders practice. The panel noted prior counsel’s Anders submission was found inadequate in an earlier appeal-related disposition (United States v. Manderville, 2025 WL 1604505), and that the case “moved beyond the Anders construct” once the Court ordered merits briefing. This frames the Court’s willingness to engage substantive issues rather than dismiss them summarily under Anders.
  • United States v. Olano, 507 U.S. 725 (1993)
    Olano supplied the plain-error framework. Because Manderville raised the alleged breach and plea-voluntariness issues for the first time on appeal, the Court emphasized that relief requires error that is “clear or obvious,” among other requirements. The panel resolved the breach claim at what it described as “Step Two of Olano”—lack of clarity/obviousness.
  • United States v. Williams, 510 F.3d 416 (3d Cir. 2007) (quoting United States v. Hodge, 412 F.3d 479 (3d Cir. 2005))
    These cases provided the governing contract-like lens for plea agreements: courts evaluate “the parties’ reasonable understanding of the agreement” and whether challenged conduct is consistent with it. This framing set the baseline for assessing whether prosecutorial sentencing advocacy crossed the line.
  • United States v. Davis, 105 F. 4th 541 (3d Cir. 2024)
    Manderville relied on Davis for the proposition that the Government can breach a plea agreement by overemphasizing aggravating facts in a manner that undermines its promised recommendation. The panel distinguished Davis by stressing that in Manderville the Government “made such a pivot” to mitigation and advocated for a sentence “substantially below the Guidelines range.” The panel also invoked Davis’s own suggested cure: after describing aggravation, a prosecutor may avoid breach by pivoting to why a lower sentence is still warranted.
  • United States v. Nolan-Cooper, 155 F.3d 221 (3d Cir. 1998)
    Like Davis, Nolan-Cooper was used as a contrast case where the prosecution’s approach was found inconsistent with the plea deal. The panel distinguished it on the plea agreement’s text here, which permitted the Government to provide “the full nature and extent of Manderville’s activities and relevant conduct.”
  • United States v. Thornton, 327 F.3d 268 (3d Cir. 2003)
    Cited to explain that ineffective-assistance-type complaints about trial counsel’s performance are typically left for collateral review, not resolved on direct appeal on the existing record.
  • United States v. Agarwal, 24 F.4th 886 (3d Cir. 2022)
    Reinforced that unpreserved challenges to the plea proceedings are reviewed only for plain error.
  • United States v. Dominguez Benitez, 542 U.S. 74 (2004)
    Supplied the prejudice standard for unpreserved Rule 11 errors: a defendant must show a reasonable probability that, but for the error, he would not have pleaded guilty. The panel held Manderville did not meet that burden.

B. Legal Reasoning

1. Alleged breach of the plea agreement at sentencing

The Court treated the breach issue as an unpreserved claim and applied plain-error review. It then assessed the Government’s sentencing presentation through the “reasonable understanding” standard from Williams/Hodge.

Manderville’s “poison[ing] the well” theory had two principal components: (i) the Government’s opening comments about the scourge of drug trafficking and its harms, and its emphasis on Manderville’s criminal history and recidivism; and (ii) the Government’s closing “pitch” for a “significant” sentence allegedly ambiguous between the stipulated range and the Guidelines range.

The panel rejected ambiguity as a factual matter: the Government stated at the outset it was “requesting a downward variance to the top of the stipulated range that the parties have agreed to in the plea agreement,” and the Court expressly noted it understood the parties sought 5 to 7 years.

The panel also distinguished Davis and Nolan-Cooper by focusing on (a) what the plea agreement allowed (the Government could describe “the full nature and extent” of conduct), and (b) the structure of the Government’s advocacy. Importantly, the Government did not merely dwell on aggravation; it “pivoted” to mitigation (excellent pretrial conduct, lower recidivism risk) and advocated a sentence well below the Guidelines. That “pivot” mirrored Davis’s suggested way a prosecutor might cure the risk of breach after describing aggravating facts.

Finally, the panel treated the “significant” language as non-transgressive in context: both sides used similar language (defense counsel called 5 years “very significant”), and the Government’s “significant” request was tied to the top of the stipulated range rather than the Guidelines.

Even assuming a breach could be argued, the Court held any error was not “clear or obvious” under Olano. It also underscored that the District Court considered the agreement and mitigating points and still imposed a large downward variance—though not as low as the parties’ stipulated range.

2. Knowing and voluntary plea

Manderville’s pro se argument hinged on a claimed belief that his plea guaranteed a 5-to-7-year sentence. The panel resolved this primarily by reference to the written plea agreement and plea colloquy: the agreement stated the sentence was “within the sole discretion of the sentencing judge,” the stipulations did not bind the Court, and Manderville acknowledged understanding those terms.

Applying Dominguez Benitez, the panel concluded Manderville failed to show a reasonable probability that any supposed Rule 11 error affected his decision to plead guilty, and thus failed to establish plain-error prejudice.

C. Impact

Although designated “NOT PRECEDENTIAL,” the opinion is practically significant in three ways:

  1. Clarifies how “breach-by-rhetoric” claims fare under plain-error review. When a defendant fails to object at sentencing, the appellate hurdle is high; even potentially aggressive prosecutorial framing may not be “clear or obvious” breach when the plea agreement authorizes discussion of “the full nature and extent” of conduct and the prosecutor explicitly requests the bargained-for range.
  2. Signals the importance of the “pivot” emphasized in United States v. Davis. The panel operationalizes Davis’s dictum: a prosecutor may discuss aggravation but should then affirmatively connect mitigating factors to the promised recommendation to avoid undermining the bargain.
  3. Reinforces the primacy of plea text and colloquy in voluntariness challenges. Clear “court not bound” language, coupled with defendant acknowledgments, makes it difficult to later claim a guaranteed sentencing cap—especially on plain-error review.

4. Complex Concepts Simplified

  • Plea agreement “breach”: A plea deal is treated like a contract; the question is whether the Government’s conduct matches what the defendant reasonably understood the Government promised.
  • “Poisoning the well”: A rhetorical claim that the prosecutor, while ostensibly recommending a lower sentence, front-loads the presentation with aggravating facts so strongly that it effectively defeats the promised recommendation.
  • Guidelines range vs. stipulated range: The Guidelines range is the advisory range calculated under the Sentencing Guidelines; a stipulated range is a negotiated recommendation between parties, which (unless binding under a different plea mechanism) does not constrain the judge.
  • Downward variance: A sentence below the advisory Guidelines range based on statutory sentencing factors (18 U.S.C. § 3553(a)).
  • Plain error (Rule 52(b)): If an issue was not raised below, an appellate court corrects it only if the error is clear/obvious and affected substantial rights (and, typically, seriously affects the fairness or integrity of judicial proceedings).
  • Anders brief: A procedure allowing defense counsel to seek withdrawal when counsel believes the appeal lacks nonfrivolous issues; the court must still independently review. Here, the Third Circuit required merits briefing, reflecting its gatekeeping role.

5. Conclusion

United States v. William Manderville affirms that not every forceful description of offense harm or recidivism constitutes a plea-agreement breach—particularly where the agreement permits full discussion of relevant conduct, the Government clearly asks for the stipulated range, and it “pivots” to mitigation consistent with its recommendation. The decision also underscores that “court not bound” plea terms, confirmed on the record, defeat later claims of a guaranteed sentencing range, especially under plain-error review.