Variance, Factual Sufficiency, and the Reach of CCA Authority: CAAF's Decision in United States v. Kershaw

United States v. Kershaw clarifies an important question about the scope of a service Court of Criminal Appeals' factual sufficiency review under the prior version of Article 66, UCMJ. When a discrepancy exists between a fact pleaded by the government and a fact proven at trial — and when that fact is not an integral part of an element of the charged offense — a CCA possesses the authority to conduct a variance analysis and test the discrepancy for prejudice. That authority exists regardless of whether the trier of fact made exceptions and substitutions at trial. If the variance analysis reveals no prejudice, the CCA may affirm despite the discrepancy.

---

CAAF's standard formulation describes variance simply: “A variance between pleadings and proof exists when evidence at trial establishes the commission of a criminal offense by the accused, but the proof does not conform strictly with the offense alleged in the charge.” United States v. Teffeau, 58 M.J. 62, 66 (C.A.A.F. 2003). The doctrine recognizes that real trials rarely match the charge sheet with surgical precision. Witnesses misremember dates. Locations get rounded to the nearest landmark. Names appear in slightly different forms. Variance doctrine asks whether those gaps require reversal or whether the appellant suffered no actual prejudice.

Two kinds of factual gaps require different treatment. A discrepancy that goes to an element of the charged offense — the identity of the larceny victim, for example, where ownership is itself an element — raises a different problem than a discrepancy in surplus factual detail. The latter triggers a prejudice-based variance analysis. The former does not, because the gap reaches the existence of the offense itself.

When variance enters the appellate analysis, courts test for materiality and prejudice. The Court of Military Appeals identified the controlling factors in United States v. Lee, 1 M.J. 15, 17 (C.M.A. 1975), and CAAF refined them in United States v. Teffeau and United States v. Treat, 73 M.J. 331, 336 (C.A.A.F. 2014). The court asks whether the discrepancy:

    • Misled the accused regarding what he needed to defend against at trial;
    • Failed to protect the accused from another prosecution for the same offense;
    • Altered the nature of the offense;
    • Changed the applicable maximum punishment; or
    • Materially infringed the accused's due process rights in any other way.

If the variance fails any of these prongs, it is fatal. If it satisfies all of them, it is non-fatal, and the conviction may stand. Federal appellate practice outside the military uses an analogous framework. See, e.g., United States v. Cochran, 697 F.2d 600 (5th Cir. 1983) (affirming where the indictment alleged “on or about November 1, 1981” and proof showed September 1981 because the defendant suffered no prejudice).

The “On or About” Question Comes First
In date-related cases, the variance inquiry has a threshold step. The phrase “on or about” in a specification covers a range of dates reasonably near the stated date. CAAF held in United States v. Hunt, 37 M.J. 344, 347 (C.M.A. 1993), that the phrase “on or about” “connote[s] a range of days to weeks,” and the Court reaffirmed that reading in United States v. Simmons, 82 M.J. 134, 139 (C.A.A.F. 2022). When the proof falls within that range, no variance exists at all.

CAAF was explicit in Kershaw that this threshold step survives. A CCA must first decide whether the date proven at trial falls within the elastic range that “on or about” allows. Only after concluding that the proven date sits outside that range does the variance question arise.

The Facts of Kershaw
SSgt John D. Kershaw faced two specifications under Article 120b, UCMJ, alleging a lewd act and a sexual act on his niece, FA. The charge sheet placed both offenses “between on or about 1 April 2016 and on or about 30 April 2016.” At trial, the testimony of FA and her mother about the conduct itself was reasonably consistent. The dates were not. Locating the events in time required a complex matrix of past addresses and personal events, and the resulting timeline was muddled.

Trial counsel anticipated the problem and asked the military judge to give a variance instruction. The judge complied. The panel convicted Kershaw of the sexual abuse specification, contrary to his pleas, without making any exceptions or substitutions, and acquitted him of the rape specification.

On appeal, the Air Force Court of Criminal Appeals concluded that the evidence placed the offense “sometime between April 2015 and October 2015,” which the court found to be six months to a year before the charged window. The CCA held that gap was “too substantial” to fit within the “on or about” language. So far, so straightforward.

Then the CCA took an additional step. It declined to conduct a variance analysis at all. Because the military judge had given a variance instruction at trial and the panel had “made no changes” — that is, no exceptions or substitutions — the lower court reasoned that “there is no variance issue” for it to consider on appeal. The court set aside the conviction on factual sufficiency grounds.

The Certified Issue
The acting Judge Advocate General of the Air Force certified the question whether the AFCCA had erred by treating the case as a pure factual sufficiency problem and declining to analyze the discrepancy as a variance. CAAF granted review. United States v. Kershaw, 86 M.J. 133 (C.A.A.F. 2025).

CAAF's Holding
Chief Judge Ohlson, writing for a majority, set aside the AFCCA decision and remanded for a new factual sufficiency review. The Court held that a CCA possesses the authority to conduct a variance analysis on appeal even when the panel made no exceptions or substitutions, provided the discrepant fact is not an integral part of an element of the charged offense. Three steps in the Court's reasoning deserve attention.

Patterson Did Not Forclose Variance Analysis

CAAF first distinguished its recent decision in United States v. Patterson, 86 M.J. 24 (C.A.A.F. 2025), which had affirmed an AFCCA finding of factual insufficiency premised on a date discrepancy. Patterson held only that a CCA may rest a factual insufficiency finding on a pleading-proof gap; it expressly reserved the question whether a CCA may instead run a variance analysis and affirm if no prejudice exists. Kershaw answers that reserved question.

Hunt Permits Appellate Variance Analysis Without Trial-Level Exceptions

The Court drew most directly on Hunt, which had affirmed a lower court that conducted a full variance analysis even without trial-level exceptions and substitutions. Hunt stated, in language CAAF treated as binding rather than dicta: “[A]ssuming a material variance occurred here as a matter of law, we still cannot find reversible error in the judge's denial of the motion for a finding of not guilty. To prevail on that motion appellant must show that he was prejudiced by the variance in this case.” From this passage, CAAF derived a general principle: appellate variance analysis does not depend on what the trier of fact did with exceptions and substitutions.

CAAF connected that conclusion to the broader principle that a CCA conducting factual sufficiency review must independently assess the evidence in the entire record “without regard to the findings reached by the trial court.” On that view, the panel's treatment of exceptions and substitutions cannot serve as a gatekeeper for the appellate variance inquiry.

English Does Not Bar the Approach

The Court then addressed United States v. English, 79 M.J. 116 (C.A.A.F. 2019), which had reversed a CCA that excepted a charged means of force after the proof failed. English held that a CCA cannot use exceptions to create a broader or different offense than the offense charged at trial, and cannot affirm based on a generalized theory of force never submitted to the factfinder. CAAF distinguished those holdings. English speaks to discrepancies that go to elements of the offense; Kershaw speaks to discrepancies that do not. When the discrepancy is in surplus factual detail rather than an element, a CCA does not create a broader or different offense by holding the variance non-fatal.

The Scope of CCA Authority After Kershaw​
First, a CCA retains the option to find a conviction factually insufficient based purely on a pleading-proof gap. Patterson remains good law. The Government's argument that variance analysis is required did not carry the day.

Second, a CCA also possesses an alternative path. Where the discrepant fact is not an integral part of an element, the CCA may treat the gap as a non-fatal variance and affirm. CAAF described the choice between those two paths as discretionary.

Third, the variance authority is not derivative of trial-level fact-finding. The Court was emphatic on this point: “the factfinder below cannot serve as a variance gatekeeper.” A CCA may engage in variance analysis even when the panel returned its verdict without exceptions or substitutions and even when the trial court did not formally rule on a variance motion.

Fourth, the variance authority does not extend to discrepancies that go to elements of the offense. CAAF used a hypothetical assault prosecution to illustrate the boundary. If the specification alleges that the assault occurred in Arlington but the testimony places it in Alexandria, the CCA may except the location word and treat the gap as a non-fatal variance because the location is not an integral part of any element. By contrast, in United States v. Lubasky, 68 M.J. 260 (C.A.A.F. 2010), the discrepancy concerned the identity of the larceny victim — an element of the offense — and no variance fix was available.

Fifth, in date-related cases, the appellate court must work in two steps. The first step asks whether the proof falls within the “on or about” range. Only if it does not does the second step — the variance analysis — begin.

Sixth, the variance authority is permissive, not mandatory. The Court remanded so that the AFCCA could decide whether to apply the variance framework, not because the AFCCA was required to affirm.

Judge Maggs's Dissent
Judge Maggs, joined by Judge Hardy, would have affirmed the AFCCA. He read Hunt as standing primarily for the meaning of “on or about,” and treated its prejudice language as dicta supported only by Lee, a case that involved findings by exceptions and substitutions. The dissent's central concern was a textual one: the prior version of Article 66(d)(1) directs that a CCA “may affirm only such findings of guilty . . . as the Court finds correct in . . . fact.” On the dissent's reading, allowing a CCA to affirm a finding the court itself believes is incorrect in fact runs counter to the statute.

The dissent also worried that the new rule undercuts the limited reach of “on or about,” undermines English's constraints on appellate revision of specifications, and effectively imports a federal-circuit prejudice test that Congress did not authorize. The dissent would have left those policy choices to Congress.

The majority responded that prejudice testing is a traditional appellate function, that nothing in the Article 66 text affirmatively forbids the variance approach, and that placing CCAs in a worse position than other federal appellate courts when applying the longstanding variance framework would conflict with the “awesome, plenary” authority CAAF has long ascribed to those courts.

A Footnote Worth Marking
Tucked into footnote 4 of the majority opinion is a reminder of broader application: when a CCA reviews a conviction resulting from a guilty plea, it applies neither legal nor factual sufficiency review. It instead considers the providence of the plea. See United States v. Faircloth, 45 M.J. 172, 174 (C.A.A.F. 1996). Kershaw's variance holding therefore operates only in cases tried on the merits, not in providence-of-the-plea reviews.