The most vexing issue in an appeal is often not whether there was error, but its significance. CrimR 52 says that errors which don't affect the "substantial rights" of the accused are harmless and can be disregarded; the Supreme Court has held that even errors of constitutional dimension can be ignored if they are "harmless beyond a reasonable doubt."
So let's say you have a case where the defendant is tied to a shooting. During the trial, the judge allows the police officer to testify that someone told him that the defendant had previously threatened the victim. It shouldn't have come in; it's hearsay, and it's testimonial under Crawford v. Washington. Still, the state's got a pretty strong case: three eyewitnesses, a positive gunshot residue test. What's more, besides the tainted testimony of the cop, there's other evidence that the defendant harbored a grudge toward the victim, and two other people testify that he admitted the shooting to them. You could make a pretty good claim that the remaining evidence was more than sufficient to convict. That's how a lot of appellate courts handle the issue: you'll find innumerable decisions shrugging off claims of error on the basis of the "overwhelming evidence" against the defendant.
But what if I told you that, after eight hours of deliberation, the jury asked whether they could have the officer's testimony about the threat -- the inadmissible testimony -- read back to them, and that after the judge did so, they returned shortly afterward with a guilty verdict. Or what if I told you that this happened after eight days of deliberation? Still think the error was harmless?
These scenarios reflect the differences in how harmless error is evaluated. In the first, the court looks only at the evidence remaining after the error is discounted and decide whether it's sufficient for conviction. In the second, the court looks at how the error affected the jury, and attempts to determine whether it contributed to the conviction. The second approach comes in two flavors: looking at how it might have affected a "reasonable" jury, and looking at how it affected this particular jury.
The Supreme Court took a look at that a couple of weeks back in the oral argument in Vazquez v. United States. Oddly enough, the Court has never decided on one particular standard of how harmless error is to be evaluated; it's vacillated between holding that the appropriate test is "whether there is a reasonable possibility that the evidence complained of might have contributed to the conviction, others that say the the appellate court should "simply review the remainder of the evidence against the defendant to determine whether" the error was harmless. The Ohio Supreme Court has been similarly inconsistent; in State v. Hale the court found the error harmless because of "overwhelming" evidence of guilt, but just two years earlier had held in State v. Conwaythat the appellate court must "decide the probable impact of the error on the minds of the average juror."
Although the two tests aren't mutually exclusive -- it may well be that the evidence is so overwhelming no error could have had any impact on the verdict -- this isn't just splitting hairs. As Kagan pointed out during the argument in Vazquez, studies show that the standard used is often outcome-determinative: courts which look only at the remaining evidence tend to find reversible error less frequently than courts which look at how the error might have impacted the jury's decision. That's obvious from the example I gave above. If you just look at the evidence against the defendant, you could easily come to the conclusion that the error didn't matter. But if you look at what the jury did, it's obvious that the inadmissible error played a significant role in their deliberations, and in the outcome of the case.
So what's the Court going to do in Vazquez? Nothing; the other day it dismissed the petition for certiorari as having been improvidently granted. The reason for that became apparent during the first few minutes of oral argument. Instead of arguing whether review looked at whether the evidence was overwhelming or how it might have affected the jury, both parties agreed that the latter standard was appropriate: the only point of dispute was whether review focused on the effect of the error on the average jury or on the particular jury which heard the case. Muddying the waters further was that it wasn't clear which standard the lower court had applied, or whether the result would have been any different if the lower court had used Vazquez's standard.
That's a shame, because it was a missed opportunity to clarify what the standard should be. It's not exactly rare for the issue of harmless error to arise; in fact, any time a court of appeals finds error, the next step is to determine whether it's harmless.
And there's another issue here: who gets to decide guilt? Under our system, it's the jury, a conclusion reinforced by the decisions in the Apprendi/Blakely line. When you have a court of appeals deciding whether the evidence is "overwhelming," with no consideration of whether the jury found it so, you're essentially allowing the appellate courts to determine whether a defendant is guilty.
So the Court in Alvarez missed an opportunity to clarify how to determine harmless error, but that doesn't mean you have to. Appellate tip for the day: if you're doing this sort of work, you're arguing error, and if you're arguing error, you've got to point the court in the right direction. That means acclimating them to the idea of looking at it from the jury's perspective, not their own.