In 1967, the Supreme Court held in Chapman v. California, 386 U.S. 18, that California could not apply its long-established harmless error standard to federal constitutional errors. The long-established federal standard of Kotteakos v. United States, 328 U.S. 750 (1946), whether "the error had substantial and injurious effect or influence in determining the jury's verdict," was not good enough, either. Chapman required "the beneficiary of a constitutional error to prove beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained."

Drawing such a line between constitutional and nonconstitutional errors would have made some sense in the days when only fundamental errors going to the heart of the fairness of the proceeding were considered constitutional, but that day was already past when Chapman was decided. The Fourth Amendment exclusionary rule was already excluding evidence for reasons having nothing whatever to do with the fairness of the trial, and the Miranda rule was tossing out confessions regardless of how clearly voluntary they were. Even so, Chapman is precedent, and the Court has shown no inclination to overrule it.

In 1993, the Court asked in Brecht if it was bound to apply Chapman on habeas corpus as well as on direct review of state judgments. The Court held it was not and decided that the Kotteakos standard was better suited to the "secondary and limited" proceeding of habeas corpus. That proceeding is not for the purpose of providing a second appeal, but rather to fix grievous wrongs. A technical error that was unlikely to have made any difference does not qualify.

In the course of the discussion, the Brecht Court noted that the state courts had previously considered the harmlessness of the error under the Chapman standard. The Eighth Circuit alone has held that Brecht is limited to that situation, and Chapman still applies when the state court, for whatever reason, did not do a Chapman harmless error analysis. All the other circuits to squarely address the question have gone the other way.

That issue should present the Court with little difficulty. The language of Brecht itself is clear enough that its rule applies to all federal habeas cases. The rule it created was not a rule of deference to the state court decision. Congress would enact that rule three years later. The primary rationale of Brecht was limitation of the drastic remedy of collateral attack on a final judgment to cases that really deserve it.

The Fry case contains an interesting wrinkle, though, in that the underlying rule is that of Chambers v. Mississippi, 410 U.S. 284 (1973). Can a Chambers violation ever be harmless?

For most rules, a finding of error must be followed by an inquiry as to whether the error was harmless. There are two classes where this is not necessary, which I call the "error is never harmless" group and the "harmless is never error" group. The first group is structural errors such as a complete denial of counsel. The second group consists of rules where harmfulness is built into the definition of the rule. For example, there is no such thing as a harmless Brady error — failure to disclose material exculpatory evidence — because if the failure to disclose was harmless then the evidence wasn't material, and there was no constitutional duty to disclose it in the first place. See Kyles v. Whitley, 514 U.S. 419, 436 (1995).

In United States v. Scheffer, 523 U.S. 303, 315-317 (1998) the Court reminded us that the Chambers line does not steamroller the entire rules of evidence and give criminal defendants carte blanche to introduce anything and everything. Only where the exclusion "undermined the fundamental elements of the defendant's defense" is there a constitutional error. If that is the threshold to be met to show an error, how could the error then be harmless under Brecht/Kotteakos?

Counsel for Fry protests that the Ninth Circuit's finding of Chambers error is inconsistent with its finding that the error was harmless. He has a point.  But which of those two decisions was wrong? The answer is fact-intense, and the facts of this case are complex.