In Washington v. Recuenco, a lopsided 7-2 majority held that the failure to have the jury rather the judge decide a fact that increases the maximum sentence is subject to harmless error analysis. The case is a perfect illustration of harmless error. The jury convicted the defendant of assault. They found he used a deadly weapon. The only weapon involved was a gun. The judge added a 3-year sentence enhancment for use of a gun, but after the trial the Supreme Court held in another case, Blakely v. Washington, that these kinds of findings have to be made by a jury. The "error" in this case is harmless, because the jury had actually found all the facts in genuine dispute.

In United States v. Gonzalez-Lopez, the defendant was denied the lawyer of his choice, because the trial judge erroneously refused to admit an out-of-state lawyer pro hac vice. By a vote of 5-4, the Court held that this was grounds for reversal regardless of whether the "second choice" lawyer did an equal or better job. Justice Scalia wrote the opinion, and Justice Alito wrote the dissent.

In the 1993 case of Sullivan v. Louisiana, a unanimous opinion by Justice Scalia, the Court held that an erroneous jury instruction on the burden of proof can never be harmless error. The opinion went further than it needed to and declared that because of this error there was no real jury verdict, and "no object, so to speak, upon which the harmless-error scrutiny can operate."

Six years later, in Neder v. United States, a defendant cited Sullivan in support of a claim that a failure to have the jury find one element of an offense, materiality of false statements in a fraud case, could never be harmless. It was a multimillion dollar fraud, and materiality was never in doubt. A holding that this error could not be harmless would have been contrary to a considerable body of precedent, and the Court in Neder disclaimed the expansive language of Sullivan. Justice Scalia wrote a strident dissent.

Yet the Neder opinion went too far itself. Chief Justice Rehnquist characterized the Court's prior cases of errors that can never be harmless ("structural errors") as errors that necessarily render a trial fundamentally unfair. That was manifestly not correct. In 1984, McKaskle v. Wiggins classified as structural error the erroneous failure to let a defendant represent himself. Because defendants who exercise their constitutional right to be a fool generally botch their cases, this error typically makes the trial more fair rather than less. In 1986, Vasquez v. Hillery held that it was not harmless error that the defendant was indicted by an improperly selected grand jury, even though he was subsequently convicted by a properly selected trial jury. The proper trial verdict of guilty beyond a reasonable doubt eliminated any possibility that the defendant was indicted without probable cause, yet reversal was still required.

Despite its expansive language, the Court has shown no inclination to overrule the holding of Neder on its facts, and that holding compelled the conclusion in Recuenco. The whole point of the Apprendi-Blakely line of cases is to treat sentencing factors like elements of offenses when they raise the ceiling for the allowable sentence. If failure of the jury to find an element can be harmless, so can failure to find a sentence enhancement factor. Justice Thomas's straightforward opinion for the Court has no difficulty reaching this conclusion, and Justice Scalia concurs in this application of Neder's core holding.

In Gonzalez-Lopez, though, a majority of the Court is willing to put aside Neder's unnecessary and expansive language. Justice Alito in dissent relies on the language of Neder that only errors that necessarily render a trial unfair are structural. The majority simply ignores that part of Neder and relies on earlier cases that a "structural" error is one where the effect of the error cannot be quantified.

The holdings of all these cases can be reconciled, but their broad language cannot. The Supreme Court has said several times, "It is to the holdings of our cases, rather than their dicta, that we must attend...." It would be better for the Court as an institution if they just refrained from the dicta (statements unnecessary to the opinion) in the first place. For the Court to contradict or ignore statements in its own precedents can only weaken public respect for the Court as an institution. The judicial power is to decide cases and controversies. The better way to approach that task is to say as much as needs to be said to decide the case and explain the decision. Then stop.