Keck v. Bonn: Sixth Circuit Reaffirms AEDPA–Strickland “Double Deference,” Making Expert-Selection Strategy Virtually Unassailable on Federal Habeas

Introduction

In Allen W. Keck v. Dale Bonn, the U.S. Court of Appeals for the Sixth Circuit (opinion by Judge John K. Bush, joined by Judges Stranch and Davis) affirmed the denial of a federal habeas petition challenging a 2018 Michigan conviction for first-degree child abuse. Although designated “not recommended for publication,” the opinion offers a rigorous application of the Antiterrorism and Effective Death Penalty Act (AEDPA) and the Sixth Amendment ineffective-assistance framework under Strickland v. Washington. It also elaborates—drawing on the Sixth Circuit’s recent en banc articulation in Hodge v. Plappert—how high the bar remains for habeas petitioners when state courts have adjudicated ineffective-assistance claims on the merits.

Keck’s petition raised three ineffective-assistance claims: (1) defense counsel’s decision to rely on one expert, child-abuse pediatrician Dr. Marcus DeGraw, rather than a suite of additional experts; (2) counsel’s failure to provide retinal imaging to Dr. DeGraw pretrial; and (3) counsel’s failure to adequately impeach the State’s pediatric ophthalmology witness, Dr. Elana Gianfermi, with a prior inconsistent statement about timing of injuries. The Sixth Circuit concluded that none of these claims could overcome AEDPA’s deference to state-court adjudications—especially given Strickland’s own deference to strategic decisions—what the Court aptly termed a “doubly deferential” standard.

The case arises from injuries to Keck’s three-month-old daughter, CK, whom he was watching with another toddler in the home. Medical professionals diagnosed injuries consistent with nonaccidental trauma. The prosecution also presented evidence of Keck’s prior conviction for the second-degree murder of another infant child decades earlier. A Michigan jury convicted Keck of first-degree child abuse, and the Michigan Court of Appeals affirmed. After the state supreme court denied review, a federal district court denied habeas relief but issued a certificate of appealability. The Sixth Circuit now affirms.

Summary of the Opinion

The Sixth Circuit:

Applied AEDPA deference to the last reasoned state-court decision (the Michigan Court of Appeals), and adopted that court’s factual findings under 28 U.S.C. § 2254(e)(1) because Keck did not rebut them by clear and convincing evidence.
Emphasized the “formidable barrier” to habeas relief and, for ineffective-assistance claims, the “double deference” that results from layering AEDPA deference over Strickland deference to counsel’s decisions.
Held that counsel’s decision to rely on Dr. DeGraw as the defense’s sole medical expert was a quintessential strategic choice that was reasonable on this record and therefore not deficient under Strickland.
Rejected the retinal-imaging claim because Keck failed to show that earlier access would have altered Dr. DeGraw’s testimony or the outcome; the record showed DeGraw would defer to the treating ophthalmologist’s in-person observations regardless.
Agreed that impeachment of Dr. Gianfermi was suboptimal, but found no Strickland prejudice given the weight of the other medical evidence (including separate timing testimony on head injuries) and evidence of Keck’s prior conviction; thus, the state court’s no-prejudice ruling was not an unreasonable application of Strickland.

Detailed Analysis

Standards Applied and Framework

The Court began by invoking AEDPA’s key provisions:

Section 2254(d) limits federal habeas relief to cases where the state court’s adjudication was either contrary to, or involved an unreasonable application of, clearly established Supreme Court law, or rested on an unreasonable determination of the facts.
Section 2254(e)(1) imposes a presumption of correctness on state-court factual findings, rebuttable only by clear and convincing evidence. Keck did not meaningfully contest those findings.

The Court then applied Strickland’s two-part test—deficient performance and prejudice—underscoring that both prongs are highly deferential to counsel’s decisions at trial, and that under AEDPA the question is whether any fair-minded jurist could agree with the state court’s Strickland analysis.

Precedents Cited and Their Role

Strickland v. Washington, 466 U.S. 668 (1984): The foundational two-prong test for ineffective assistance—objective unreasonableness and prejudice (a reasonable probability of a different outcome sufficient to undermine confidence in the verdict).
Harrington v. Richter, 562 U.S. 86 (2011): Clarifies that the AEDPA–Strickland overlay is “doubly deferential,” and relief lies only if there is no reasonable argument that counsel satisfied Strickland’s standards. Also defines the prejudice standard as requiring a “substantial, not just conceivable” likelihood of a different result.
Cullen v. Pinholster, 563 U.S. 170 (2011): Reiterates the “substantial, not just conceivable” prejudice threshold and emphasizes deference to state-court adjudications.
Dunn v. Reeves, 594 U.S. 731 (2021): Reaffirms the strong presumption of reasonableness for counsel’s choices.
Hinton v. Alabama, 571 U.S. 263 (2014): Recognizes that selection of an expert is a paradigmatic strategic choice; performance may be deficient where the choice is driven by a mistaken belief about funding rather than strategy. The Sixth Circuit distinguished Hinton because Keck’s counsel made a strategic choice to retain a credible child-abuse pediatrician; there was no funding mistake.
Rompilla v. Beard, 545 U.S. 374 (2005): Duty to investigate is reasonable, not limitless; counsel need not “scour the globe.” The Court used Rompilla to reject arguments that counsel should have done substantially more expert shopping.
Wiggins v. Smith, 539 U.S. 510 (2003): Found ineffective assistance in a capital sentencing context for failure to present mitigation; cited here only to distinguish the very different posture.
Burt v. Titlow, 571 U.S. 12 (2013), and McGowan v. Burt, 788 F.3d 510 (6th Cir. 2015): Emphasize AEDPA’s “formidable barrier” to federal habeas relief, especially when state courts have adjudicated the claim on the merits.
Hodge v. Plappert, 136 F.4th 648 (6th Cir. 2025) (en banc): Clarifies the “contrary to” and “unreasonable application” clauses and underscores that a state decision must be so wrong that it is “completely impossible for a fair-minded jurist” to find it consistent with Supreme Court precedent—an articulation the panel quotes and applies here.
Virginia v. LeBlanc, 582 U.S. 91 (2017) (per curiam), and Shinn v. Kayer, 592 U.S. 111 (2020) (per curiam): Further emphasize the extraordinarily high “objective unreasonableness” threshold under § 2254(d).
Michigan authorities: People v. Ginther, 212 N.W.2d 922 (Mich. 1973) (establishing evidentiary hearings to build an extra-record basis for IAC claims) and People v. Ackley, 870 N.W.2d 858 (Mich. 2015) (finding IAC where counsel retained no helpful expert and failed to consult recommended, sympathetic experts). The panel distinguishes Ackley because Keck’s counsel retained a supportive expert (indeed, the only child-abuse pediatrician to testify) whose testimony actually undergirded the defense theory.

Legal Reasoning and Application

1) Counsel’s reliance on a single defense expert (Dr. Marcus DeGraw)

Keck argued that counsel was ineffective for calling only one expert—a child-abuse pediatrician—rather than multiple specialists (e.g., pediatric neurology, biomechanics, neurosurgery). The Court rejected this argument under Strickland’s deference to strategic choices and AEDPA’s layered deference to state-court adjudications:

Expert selection is a “paradigmatic” strategic call and “virtually unchallengeable” when made after reasonable investigation (Hinton; Strickland).
Counsel reasonably chose Dr. DeGraw, a well-credentialed expert who had testified for both prosecution and defense, enhancing perceived neutrality. Keck himself wanted DeGraw called. DeGraw’s testimony directly supported the defense theory: he questioned rib and femur fractures, downplayed the retinal findings, and opined that the head injuries could be consistent with another child sitting on CK sometime earlier.
Post-hoc expressions of regret by trial counsel (that more experts might have helped) are irrelevant; Strickland uses an objective standard, not subjective hindsight. The question is whether any reasonable lawyer could have made the same call. On this record, yes.
Calling the additional Ginther-hearing experts could have backfired: one held minority “fringe” views (and had limited explanatory scope), another was a biomechanical engineer without a medical explanation for noncranial injuries (risking jury confusion), and a third had been widely criticized and censured for courtroom theories. Avoiding credibility landmines and a “battle of the experts” was a reasonable, prudent strategy (see Richter’s caution about esoteric scientific detours).

Under AEDPA-augmented Strickland, the dispositive inquiry is whether there is “any reasonable argument” that counsel acted reasonably. The panel concluded there plainly was.

2) Failure to provide retinal imaging to Dr. DeGraw before trial

Keck next contended that counsel should have supplied the retinal images to DeGraw pretrial to prepare for cross-examination of the State’s ophthalmology witness. The Court held Keck could not show either deficient performance or prejudice:

DeGraw had the medical records and treating ophthalmologist’s findings pretrial; Keck did not explain how access to the images would have changed DeGraw’s testimony.
More importantly, DeGraw testified that he would defer to the treating ophthalmologist’s in-person observations regardless, because direct visualization of the retina by a treating specialist is paramount. That deference therefore did not flow from a lack of images.
DeGraw’s retinal testimony was, on balance, favorable to the defense (describing the hemorrhages as not “terribly significant” or severe and disputing the State expert’s conclusions). Even assuming arguendo a lapse, Keck could not establish Strickland prejudice, much less show the state court’s contrary finding was objectively unreasonable under § 2254(d).

3) Inadequate impeachment of the State’s pediatric ophthalmologist (Dr. Elana Gianfermi)

The Michigan Court of Appeals found counsel’s cross-examination deficient for failing to confront Dr. Gianfermi with a prior inconsistent statement about the timing of retinal hemorrhages (less than a week vs. possibly up to two weeks). It nevertheless held there was no prejudice. The Sixth Circuit agreed that the no-prejudice determination was not an unreasonable application of Strickland:

Timing of retinal hemorrhages was not the only timeline evidence. A pediatric neuroradiologist opined that CK’s head injuries (skull fractures and subdural hematoma) were less than a week old, independently tightening the window.
The overall case against Keck was “overwhelming”: multiple treating physicians concluded the injuries reflected nonaccidental trauma and rejected the “rough play by siblings” theory; the jury also heard extensive evidence of Keck’s prior conviction for the murder of another three-month-old under similar circumstances.
Strickland requires a “substantial,” not merely conceivable, likelihood of a different outcome, and AEDPA relief requires a showing that no fair-minded jurist could agree with the state court’s no-prejudice determination. Keck could not meet that “doubly deferential” threshold.

Key Doctrinal Reinforcements

“Double deference” governs ineffective-assistance claims on federal habeas: Strickland’s high tolerance for reasonable strategy choices is filtered through AEDPA’s even higher bar for overturning a state-court merits decision. The operative question is whether any reasonable jurist could uphold counsel’s performance.
AEDPA’s “unreasonable application” standard—as articulated in Hodge v. Plappert—requires error so clear that it is “completely impossible for a fair-minded jurist” to reconcile the state court’s decision with Supreme Court precedent. That is an exacting test.
Expert-selection decisions are virtually unassailable when grounded in reasonable investigation and strategy, particularly where proposed alternative experts carry credibility risks, fringe views, or limited explanatory value.
Hindsight critiques, including counsel’s own later misgivings, do not establish deficiency; Strickland is an objective, not subjective, standard.
Prejudice under Strickland is demanding; where the State’s evidence is robust and multifaceted, isolated shortcomings (e.g., a missed impeachment point) rarely suffice.

Impact and Implications

The opinion, though unpublished, has practical and doctrinal implications for habeas litigation and trial practice:

For habeas petitioners: Overcoming AEDPA in ineffective-assistance claims that second-guess expert selection is extraordinarily difficult. Petitioners must demonstrate that no reasonable jurist could accept the state court’s approval of counsel’s strategy. Simply proposing “additional experts” or postulating marginal improvements in testimony is not enough—especially if those experts present credibility vulnerabilities or do not meaningfully address the full spectrum of injuries.
For defense counsel in child-abuse prosecutions: Selecting a credible, mainstream expert who can directly address the State’s medical theories is sound strategy. Courts will not fault decisions to avoid fringe or frequently censured experts, nor will they require counsel to create a “battle of the experts” when doing so risks confusing jurors or diluting a clear defense narrative.
Record-building in state court: The Court’s reliance on a Ginther hearing record underscores the importance of developing expert-related IAC claims in state postconviction proceedings. Under Pinholster, federal review of § 2254(d)(1) claims is tethered to the state-court record; robust state evidentiary development can be critical—though still insufficient unless the state court’s resolution is unreasonably wrong.
Impeachment claims and prejudice: Even where a state court acknowledges deficient performance (as with the missed impeachment), habeas relief often founders on prejudice if other evidence independently supports the verdict. Here, additional medical testimony and the admission of a prior conviction significantly blunted any effect better impeachment might have had.
Use of prior bad acts evidence in prejudice analysis: While the admissibility of prior-act evidence was not before the Sixth Circuit, its presence weighed heavily against finding Strickland prejudice. Practitioners should recognize that the totality-of-the-evidence frame under Strickland can make relief elusive when Rule 404(b)-type or analogous state-law evidence is properly in the record and strong.

Complex Concepts Simplified

AEDPA deference (28 U.S.C. § 2254(d)): Federal courts may not grant habeas relief simply because they would have decided a case differently. Relief is available only if the state court’s decision was not just wrong, but unreasonably wrong, in a way no fair-minded jurist could defend in light of U.S. Supreme Court precedent.
Presumption of correctness (28 U.S.C. § 2254(e)(1)): State-court factual findings are presumed correct. A petitioner must rebut them by “clear and convincing evidence,” a very high standard.
Strickland’s “double deference” on habeas: First, courts defer to counsel’s strategic judgments; second, federal courts defer to the state court’s assessment of those judgments. The question becomes extremely narrow: Is there any reasonable argument that counsel met Strickland’s standard?
“Unreasonable application” of clearly established law: A state court unreasonably applies Supreme Court law when it applies the correct rule in a way that is objectively unreasonable—not just incorrect. Per Hodge v. Plappert, the bar is crossed only when it is “completely impossible” for a fair-minded jurist to agree with the state court.
Ginther hearing (Michigan): An evidentiary hearing to develop facts outside the trial record, often used for ineffective-assistance claims that require proof of what counsel did or did not do and why.

Conclusion

Keck v. Bonn is a textbook application of AEDPA–Strickland “double deference.” It underscores that challenging counsel’s expert-selection strategy on habeas is a steep uphill climb, particularly where counsel selected a credible, supportive expert and avoided witnesses with evident credibility or methodological problems. The panel’s analysis—rooted in Supreme Court precedents and the Sixth Circuit’s en banc articulation in Hodge—confirms that habeas relief will be denied unless the petitioner demonstrates that the state court’s rejection of the Strickland claim was not only wrong but so unreasonable that no fair-minded jurist could defend it.

For practitioners, the opinion reinforces the importance of strategic coherence in expert selection, thorough but targeted investigation, and building a full record in state court. For habeas litigants, it is a cautionary tale: without a truly egregious breakdown of representation or a clear, outcome-determinative error, ineffective-assistance claims—especially those second-guessing expert choices—will almost invariably fail under AEDPA’s formidable standards.