Failure to Investigate a Known Exculpatory Eyewitness Can Warrant Habeas Relief Despite AEDPA Deference

Case: Maleek Jones v. Sharonda Carlos, et al., No. 23-7097
Court: United States Court of Appeals for the Second Circuit
Decision: November 10, 2026
Opinion: Judge Parker, joined by Judge Lee; Judge Sullivan dissenting.

1. Introduction

This divided decision addresses when defense counsel’s failure to investigate and present a known, potentially exculpatory eyewitness is sufficiently serious to justify federal habeas relief from a state conviction. Its importance lies in the demanding procedural setting: the federal court was required not merely to find ineffective assistance, but to conclude that the state court’s rejection of the claim was objectively unreasonable under the Antiterrorism and Effective Death Penalty Act of 1996, commonly called AEDPA.

Maleek Jones was convicted in 1995 of murdering Eddie Harp, conspiracy to commit murder, and carrying a pistol without a permit. He received a 65-year sentence. The prosecution principally relied on Tyrone Spears, an admitted participant in the shooting who implicated Jones after obtaining a substantially reduced charging exposure through cooperation.

Defense counsel Leo Ahern knew that Sheila McCray, a hospital nurse and the only disinterested eyewitness identified by the majority, had given police an account inconsistent with the prosecution’s three-shooter theory. She described two men whose appearances did not match Jones and expressly rejected Jones’s photograph as depicting either man she had observed. Nevertheless, Ahern never contacted her.

The Second Circuit affirmed the federal district court’s grant of habeas relief. The majority held that, on this record, the failure to investigate McCray and present her testimony was constitutionally deficient and prejudicial, and that the contrary state-court conclusion could not survive even AEDPA’s exceptionally deferential review. The decision does not impose a categorical obligation to interview or call every favorable witness. It instead identifies circumstances in which an unexplained investigative omission cannot reasonably be defended as trial strategy.

2. Background and Procedural History

The shooting and competing accounts

Harp was killed outside Saint Raphael’s Hospital in New Haven, Connecticut, during the early morning of October 14, 1992. The prosecution’s theory connected the killing to hostility between Jones’s street gang, the Red Line, and a rival group associated with Harp’s friend Steven Gary.

Spears testified that he, Gene “Pepper” John, and Jones fired at Harp’s vehicle. Pepper died before trial. Spears, who otherwise faced potential murder liability and a much longer sentence, became the prosecution’s principal witness under a cooperation agreement.

McCray supplied a materially different account shortly after the shooting. According to the police materials discussed in the opinion, she described two armed men, both slim and weighing approximately 130 to 140 pounds. Jones weighed approximately 200 pounds and had a lighter complexion than the men she described. When police directed her attention to Jones’s photograph in an array, she said he was not one of the men she had seen.

The defense pursued a two-shooter theory: Spears and Pepper committed the killing without Jones. Several witnesses recounted statements by Spears or Pepper that referred to their own involvement without mentioning Jones. However, the trial court excluded Lee Bember’s proposed testimony that Pepper had confessed to committing the killing with Spears. No physical evidence linked Jones to the crime, and the majority also identified tension between Spears’s account and the vehicle’s bullet damage.

The investigation that did not occur

Ahern had access to the prosecution’s investigation file and knew about McCray’s statements well before trial. He nevertheless did not telephone, visit, interview, or subpoena her. Instead, he advised Jones’s family to retain an investigator. Their investigator stopped after approximately 16.5 hours because the family could not pay, and never interviewed McCray.

The majority emphasized that investigative funding was available through the Office of the Public Defender, notwithstanding Ahern’s appointment as an outside Special Public Defender. It also stressed that McCray was readily traceable through her local residence and hospital employment. Rather than presenting her, Ahern unsuccessfully requested a missing-witness instruction permitting the jury to infer that her testimony would have been unfavorable to the prosecution.

State and federal proceedings

  • Direct appeal: Connecticut’s appellate court affirmed the conviction in 1997. It rejected the evidentiary challenge and directed that the inadequate-investigation issue be pursued through habeas proceedings.
  • First state habeas proceeding: A petition filed in 1998 was tried in 2009. It raised broad ineffective-assistance and actual-innocence claims but did not develop the McCray issue. Spears recanted his trial testimony, but the court found the recantation unpersuasive and denied relief.
  • Second state habeas proceeding: Filed in 2012, this petition challenged first habeas counsel Bruce McIntyre’s failure to investigate McCray. She testified in 2015 for the first time in any proceeding concerning the shooting. Relief was denied in 2016, and subsequent state review was unsuccessful.
  • Federal proceedings: An earlier federal petition was dismissed for failure to exhaust state remedies. Jones commenced the proceeding at issue in 2019. In 2023, the district court granted relief on ineffective-assistance and exclusion-of-evidence grounds. Jones was released that year.

The Commissioner appealed. The Second Circuit affirmed on ineffective assistance alone, leaving the separate evidentiary ground undecided.

3. Summary of the Opinion

The majority reached three principal conclusions:

  1. Deficient investigation: Ahern lacked a reasonable professional basis for failing to contact a known, accessible, disinterested witness whose account directly supported his chosen defense.
  2. Deficient presentation: Because no reasonable professional judgment supported the investigative omission, the resulting decision not to call McCray could not be insulated as an informed strategic choice.
  3. Prejudice sufficient to satisfy AEDPA: McCray’s contemporaneous account could have materially undermined a prosecution case dependent on one interested eyewitness. The state court’s contrary conclusion was an unreasonable application of Strickland v. Washington.

The holding concerns the constitutional reliability of Jones’s trial, not a judicial declaration of factual innocence. Nor did the Second Circuit hold that the exclusion of Pepper’s confession independently violated the Constitution.

4. Analysis: The Governing Legal Framework

Two requirements under Strickland

Under Strickland v. Washington, 466 U.S. 668 (1984), a defendant must establish both objectively unreasonable representation and resulting prejudice. Courts evaluate performance from counsel’s perspective at the time, avoiding hindsight and presuming reasonable professional judgment.

That presumption protects informed tactical choices. It does not eliminate counsel’s obligation to investigate. Counsel must either undertake reasonable investigation or reasonably determine that particular investigation is unnecessary.

Prejudice requires a reasonable probability of a different result sufficient to undermine confidence in the outcome. It does not require proof that acquittal was certain, that acquittal was more likely than not, or that the defendant was actually innocent. At the same time, a merely conceivable effect on the verdict is insufficient.

AEDPA’s additional restriction

Under 28 U.S.C. § 2254(d), a federal court generally cannot disturb a state merits adjudication simply because it believes the state court was wrong. Relief requires, among other possible statutory grounds, a decision contrary to or unreasonably applying clearly established Supreme Court law.

For ineffective-assistance claims, this produces “doubly deferential” review: deference to counsel’s professional judgment and deference to the state court’s evaluation of that judgment. The governing question is whether any fairminded jurist could sustain the state court’s application of Strickland v. Washington.

The Second Circuit reviewed the district court’s habeas ruling de novo, but that appellate standard did not remove AEDPA’s restrictions on reviewing the underlying state decision. The majority expressly concluded that the state decision failed the more demanding federal standard.

5. Precedents Cited and Their Influence

A. Supreme Court decisions governing ineffective assistance

Strickland v. Washington, 466 U.S. 668 (1984)

This was the controlling constitutional authority. The majority relied on its investigation requirement, its distinction between informed strategy and inadequately investigated choices, and its contextual approach to prejudice. Particularly important was its recognition that errors are more likely to affect a verdict weakly supported by the evidence.

The dissent invoked the same decision’s protective principles: strong presumptions of competence, substantial deference to investigative judgments, and the possibility that already-known facts can reduce the need for further inquiry. The disagreement was therefore principally about application, not the governing test.

Wiggins v. Smith, 539 U.S. 510 (2003)

Through its discussion of Rivas v. Fischer, the majority used this decision to frame the investigative question: would the evidence already known to counsel lead a reasonable attorney to investigate further? McCray’s police statements were not peripheral information. They contradicted the prosecution’s central account and supported counsel’s own defense theory.

Rompilla v. Beard, 545 U.S. 374 (2005)

This decision supported consulting the ABA Standards as guides to professionally reasonable representation. The majority examined standards in force at the time of Jones’s trial, which called for prompt investigation and exploration of relevant factual avenues. Those standards informed the constitutional analysis; they were not treated as an independently enforceable constitutional code.

Williams v. Taylor, 529 U.S. 362 (2000)

The majority cited this decision to show that potentially unfavorable consequences do not automatically excuse failing to investigate and present favorable evidence. Although its cited discussion concerned mitigating evidence, its relevance here was the need to assess concrete benefits and risks rather than rely on generalized apprehension.

Hinton v. Alabama, 571 U.S. 263 (2014)

This provided the principal analogy concerning defense funding. There, counsel’s unreasonable failure to understand available public resources caused the use of an expert counsel considered inadequate. The majority treated Ahern’s failure to investigate available funding as similarly outside protected strategy.

The dissent disputed the analogy on two grounds: counsel had been informed before trial that public funding was available, and he could have interviewed McCray personally without retaining an investigator. The funding issue therefore rested on a contested reading of the record. The majority’s broader holding also depended on the unexplained failure to contact McCray at all.

B. Supreme Court decisions defining AEDPA deference

Harrington v. Richter, 562 U.S. 86 (2011)

This was the central authority on the distinction between an incorrect state decision and an unreasonable one. The State and dissent emphasized that a federal court must consider arguments that supported, or could have supported, the state outcome. They also relied on its requirement that a different result be substantially likely rather than merely conceivable.

The majority accepted these principles but concluded that the record supplied no reasonable justification for ignoring McCray and no reasonable basis for finding the omission harmless under Strickland v. Washington.

Burt v. Titlow, 571 U.S. 12 (2013)

The majority cited this decision for the “doubly deferential” standard. The dissent additionally invoked its presumption that counsel acted within the broad range of reasonable professional assistance. Its importance was to place the burden on Jones to establish unreasonableness, rather than requiring counsel to prove competence.

Cullen v. Pinholster, 563 U.S. 170 (2011)

The majority relied on the requirement that Jones show it was necessarily unreasonable for the state court to conclude that he had not overcome the presumption of competence. The dissent stressed the further obligation to entertain possible explanations for counsel’s conduct, even when counsel later cannot articulate them. The majority found the suggested explanations inconsistent with the actual record.

Klein v. Martin

Both opinions cited this 2026 Supreme Court decision for AEDPA’s strict limitations. The majority acknowledged it as a warning against expansive federal review. The dissent used it to emphasize that federal courts sometimes must deny relief even when they would have decided the constitutional issue differently in the first instance.

Knowles v. Mirzayance, 556 U.S. 111 (2009)

Cited by the dissent, this decision reinforced that an AEDPA petitioner must clear a substantially higher threshold than showing that the state court incorrectly applied the ineffective-assistance standard.

Mays v. Hines, 592 U.S. 385 (2021)

The dissent invoked this decision’s formulation that the state court must have erred so seriously that every fairminded jurist would disagree. It sharpened Judge Sullivan’s criticism that the majority was substituting its own evaluation for a permissible state judgment.

McCarthy v. Hernandez, 608 U.S. 809 (2026)

Also cited only by the dissent, this decision reinforced the limits Congress placed on federal habeas authority. Its role was institutional: dissatisfaction with those limits does not authorize a court to disregard them.

C. Second Circuit decisions applying these principles

Carmichael v. Chappius, 848 F.3d 536 (2d Cir. 2017)

This supplied the de novo standard for appellate review of the district court’s grant of habeas relief. It concerned the appellate court’s relationship to the federal district court, not a relaxation of deference to the state adjudication.

Cox v. Donnelly, 387 F.3d 193 (2d Cir. 2004)

The majority cited this decision as circuit authority restating the two components of an ineffective-assistance claim: deficient performance and a reasonable probability of a different outcome.

Greiner v. Wells, 417 F.3d 305 (2d Cir. 2005)

This was the most contested circuit precedent. It recognizes both the substantial protection afforded witness-selection decisions and the fact-specific nature of the investigative duty. It also distinguishes circumstances in which a complaint about investigation is effectively a complaint about not presenting evidence.

The State argued that Ahern already knew McCray’s account, so the real issue was his strategic decision not to call her. The majority distinguished the precedent because the potential witnesses there were reported to be unwilling or afraid to speak, whereas nothing indicated that McCray would have refused an interview. Knowing the contents of police reports was not, on these facts, equivalent to investigating her potential testimony.

The dissent read the same precedent as supporting deference where counsel already possessed substantial information about a witness and might reasonably decline further inquiry.

Sacco v. Cooksey, 214 F.3d 270 (2d Cir. 2000)

Quoted through Greiner v. Wells, this decision supported the proposition that counsel ordinarily should not be faulted for avoiding a potentially fruitful course carrying a significant downside. The majority found no adequately grounded, witness-specific downside here.

Rivas v. Fischer, 780 F.3d 529 (2d Cir. 2015)

This was the majority’s closest circuit analogy. It involved an unreasonable failure to investigate a central prosecution proposition and the resulting absence of important contradictory evidence. The majority considered Ahern’s omission even less defensible because he could not identify a meaningful strategy supporting it.

The case also informed prejudice analysis: when the prosecution’s decisive witness goes effectively uncontradicted because counsel failed to investigate credible opposing evidence, confidence in the verdict can be undermined even under AEDPA.

Lindstadt v. Keane, 239 F.3d 191 (2d Cir. 2001)

This supported the majority’s rejection of a generalized concern about adverse evidence as an adequate substitute for investigation. Counsel must evaluate a potential witness rather than assume that uncertainty itself justifies inaction.

Waiters v. Lee, 857 F.3d 466 (2d Cir. 2017)

This supplied the AEDPA formulation for reviewing a state finding of no prejudice: the petitioner must establish that the finding was contrary to, or an unreasonable application of, the governing Supreme Court standard.

Pavel v. Hollins, 261 F.3d 210 (2d Cir. 2001)

The majority relied on this decision for the contextual importance of the prosecution’s evidentiary strength. A comparatively limited omission may undermine confidence in a thin case even when it would not materially affect an overwhelmingly supported conviction. This is not a different legal threshold; it is an assessment of prejudice against the actual trial record.

These circuit decisions helped explain and apply the constitutional principles. They did not themselves supply the “clearly established Federal law” required by § 2254(d)(1), which must come from the Supreme Court.

D. Connecticut decisions forming the procedural record

State v. Jones, 700 A.2d 710 (Conn. App. Ct. 1997)

This direct-appeal decision upheld the exclusion of Bember’s proposed testimony and treated the inadequate-investigation argument as an ineffective-assistance claim requiring habeas development. Its significance was procedural and evidentiary, rather than supplying the federal constitutional standard.

Jones v. Warden, 2009 WL 2961443 (Conn. Super. Ct. Aug. 13, 2009)

The first habeas court found insufficient proof of prejudice and rejected Spears’s recantation. It considered his detailed original testimony more credible than his later withdrawal of the accusation. The federal majority emphasized that McCray’s potential contribution was not meaningfully developed in that proceeding.

Jones v. Warden, 2016 WL 921751 (Conn. Super. Ct. Feb. 11, 2016)

This second habeas decision was central to the dispute. It considered whether first habeas counsel was ineffective for failing to investigate McCray, questioned her recollection almost a quarter-century after the shooting, and concluded that her testimony did not undermine confidence in the criminal verdict. The majority found its treatment of the underlying trial issue objectively unreasonable; the dissent considered it a permissible assessment.

Jones v. Comm'r of Corr.

The decisions reported at 40 A.3d 344 (Conn. App. Ct. 2012), 47 A.3d 884 (Conn. 2012), 194 A.3d 905 (Conn. App. Ct. 2018), and 204 A.3d 1159 (Conn. 2019) recorded the unsuccessful appeals and certification requests following the two state habeas proceedings. Their principal significance was the completion of state review.

The dissent also referred to the 2018 affirmance as Jones v. Comm'r of Correction, 185 Conn. App. 906, 194 A.3d 905 (2018). That reference identifies the affirmance of the second habeas judgment rather than a separate substantive federal rule.

6. Analysis: Why the Majority Found a Constitutional Violation

A. The lead was central, specific, and accessible

The majority’s conclusion depended on a convergence of circumstances. McCray was not a speculative witness whose relevance emerged only after trial. Counsel knew that she had witnessed relevant events, that her description excluded Jones, and that her account supported the defense’s central theory. There was also no apparent obstacle to locating her.

Her independence mattered. Other defense witnesses were connected to the alleged participants or the Red Line. A hospital nurse without those associations potentially offered a materially different source of credibility. Independence did not guarantee accuracy, but it made investigation especially important.

B. Uncertainty called for inquiry, not avoidance

Ahern’s explanation—that witnesses sometimes say unexpected, harmful things—did not identify a concrete danger associated with McCray. In the majority’s view, that uncertainty demonstrated why he should have interviewed her before deciding whether to present her.

His later acknowledgment that failing to seek her out may have been a mistake was relevant but not independently dispositive. The constitutional inquiry remained objective: whether the circumstances supplied a reasonable basis for his conduct. The majority found none.

C. Reviewing reports did not resolve the investigative duty

The investigator’s report presented an important point of disagreement. The dissent emphasized its substantial discussion of McCray and references to conflicting statements. The majority stressed that it remained a third-hand account, omitted important physical-description evidence, and did not reflect any contact with her.

The broader principle is not that an attorney must personally interview every witness mentioned in discovery. It is that possession of reports does not necessarily discharge the investigative duty where a readily available witness could provide decisive support for the defense and counsel has not reasonably assessed her testimony.

D. The omission affected an unusually vulnerable prosecution case

Spears was both an admitted participant and a beneficiary of cooperation. No physical evidence linked Jones to the shooting. Although the defense challenged Spears and presented other accounts, the jury never heard an independent eyewitness contradict his three-person narrative.

McCray’s account therefore offered more than cumulative impeachment. Her description of two men, her physical descriptions, and her rejection of Jones’s photograph could have supported reasonable doubt about both the number of participants and Jones’s identity as a participant.

The majority did not require her testimony to make a third shooter logically impossible. That would confuse the defendant’s burden to establish prejudice with a burden to prove innocence. The relevant question was whether her evidence created a reasonable probability that the jury would doubt guilt.

E. The appropriate counterfactual concerned the 1995 trial

The majority’s most distinctive prejudice analysis concerned timing. The second state habeas court questioned McCray’s recollection in 2015, including discrepancies with the hospital security guard’s account. But the alleged constitutional injury occurred when counsel failed to investigate and present her in 1995.

Her contemporaneous police statements therefore mattered greatly. Memory problems nearly twenty-five years after the event did not necessarily establish how she would have testified much closer to it. The majority also viewed discrepancies about encounters with security personnel and police arrival as peripheral to the decisive issues of shooter numbers and identification.

This reasoning should not be read as generally eliminating deference to state credibility findings. The majority instead concluded that the state court’s credibility assessment, in its timing and subject matter, did not reasonably answer the trial-focused prejudice question.

7. The Dissent: Strategy, Causation, and Federalism

Judge Sullivan’s dissent challenged both components of the claim and, more fundamentally, the majority’s application of AEDPA.

  • Existing information could justify stopping: Counsel already possessed police materials and an investigator’s report detailing McCray’s account. A fairminded court could view another inquiry as unnecessary.
  • Funding was not necessarily misunderstood: The dissent read the record as showing that counsel knew before trial that public funds were available.
  • Possible strategic explanations deserved weight: Concerns about inconsistent statements, unforeseen testimony, and McCray’s limited observation could support declining to call her.
  • Her account did not conclusively exclude a third participant: The dissent emphasized her later acknowledgment that she did not actually observe anyone firing and noted features of her account that could corroborate Spears.
  • The prosecution might have rebutted her: Other potential identification witnesses could have been called in response, although their testimony was not presented at the criminal trial.
  • Failure to investigate required causal prejudice: If further inquiry would reveal nothing materially new, the dissent questioned whether it would have changed counsel’s presentation or the verdict.

The majority answered that these explanations did not reasonably fit counsel’s actual conduct and that omitting McCray’s evidence left the prosecution’s only testifying eyewitness unopposed by another eyewitness.

The division exposes a recurring habeas tension: courts must reconstruct reasonable explanations for counsel’s choices without converting every omission into hypothetical strategy. Conversely, they must assess constitutional reliability without using habeas review simply to replace a state court’s permissible judgment with their own.

8. Potential Impact and Limits

Consequences for defense practice

The opinion strongly supports investigating an identified, available witness whose account directly advances the defense’s core theory. It also cautions appointed counsel against treating an indigent family’s inability to pay as the end of the inquiry when public investigative resources exist.

Its practical lesson is to distinguish three separate judgments: whether to investigate, what investigation is reasonably necessary, and whether to present the witness. A defensible answer to the last question ordinarily requires an adequate basis for the earlier ones.

Consequences for postconviction litigation

The decision highlights the importance of developing what a missing witness could actually have contributed. Jones’s first habeas proceeding did not adequately develop McCray’s account; the second finally produced her testimony and a fuller record.

It also encourages careful identification of the relevant counterfactual. When trial counsel’s omission is challenged, later testimony must be evaluated for what it reveals about the original trial, rather than allowing later memory deterioration automatically to determine the outcome.

Limits on the precedent

The ruling remains highly fact-dependent. It does not establish that:

  • Every uncontacted favorable witness proves ineffective assistance;
  • Seeing fewer participants necessarily disproves a prosecution theory involving more participants;
  • A weak prosecution case automatically satisfies AEDPA;
  • A lawyer’s inability to remember a strategy automatically defeats the presumption of competence; or
  • A disputed state evidentiary ruling independently warrants federal habeas relief.

The court neither accepted Spears’s recantation as proof of innocence nor resolved Jones’s remaining grounds for relief. Its binding significance rests on the combined investigative failure, the importance of the omitted evidence, and the unreasonableness of the state court’s application of the governing constitutional standard.

9. Complex Concepts Simplified

Habeas corpus
A proceeding through which a person challenges the legal basis of custody. Federal habeas review of a state conviction is not an unrestricted new appeal.
AEDPA deference
A statutory restriction requiring federal courts to respect reasonable state-court decisions, even when the federal court would have decided the issue differently.
Doubly deferential review
The combined respect owed to defense counsel’s reasonable professional choices and to the state court’s reasonable assessment of those choices.
Deficient performance
Representation falling below an objective standard of professional reasonableness—not simply unsuccessful advocacy.
Prejudice
A sufficiently serious effect on the reliability of the result. Here, it meant a reasonable probability that adequate representation would have caused the jury to doubt Jones’s guilt.
Exculpatory evidence
Evidence tending to show that the accused did not commit the offense. It need not conclusively establish innocence.
Missing-witness instruction
An instruction that can permit an unfavorable inference from a party’s failure to present a witness. Ahern requested one concerning McCray, but the trial court refused it.
Exhaustion
The requirement that state courts first receive an appropriate opportunity to address a prisoner’s federal claims before federal habeas relief is sought.
De novo appellate review
Independent appellate review of the district court’s legal ruling. It does not cancel AEDPA’s separate protection of the state judgment.

10. Conclusion

Maleek Jones v. Sharonda Carlos demonstrates that AEDPA’s demanding standard is not an absolute bar to relief. Where counsel inexplicably fails to investigate an accessible, independent eyewitness whose contemporaneous account directly supports the defense, and the prosecution depends on a single interested eyewitness, the resulting constitutional failure can render a contrary state decision objectively unreasonable.

The decision’s central lesson is precise: deference protects reasonable professional judgment, not an unsupported omission relabeled as strategy. Its divided outcome also underscores how exceptional—and contested—federal habeas relief remains.