"Guilty beyond a reasonable doubt" means, on the one hand, something more than just thinking the defendant probably did it; and on the other hand, something less than being absolutely certain that he did it. It allows for doubts, but only "reasonable" ones.
So you come into the jury room believing that the defendant is guilty and that, whatever doubts you may have about that conclusion, those doubts are not "reasonable" ones. A fellow juror disagrees. He has doubts sufficient to cause him to vote "not guilty". You listen to his argument, but your mind isn't changed. You still vote "guilty". What, exactly, does that vote mean in the context of your assessment of the other juror? They raised doubts, and yet you're voting "guilty beyond a reasonable doubt." At first blush, there seem to be two possibilities:
(1) By voting "guilty beyond a reasonable doubt", you are implicitly saying that their doubts are unreasonable. If they were reasonable, then you'd be obligated to vote "not guilty".To me, the former makes much intuitive sense in terms of the actual concept of "reasonable doubt". Typically in law "reasonableness" is an objective standard. It is meant to transcend individual idiosyncrasies and provide a uniform standard based on the prototypical "reasonable person". And at a very basic level, I'm not sure what it means to say that while someone else's doubts may be reasonable, this person is nonetheless "guilty beyond a reasonable doubt". Once you've made the concession that someone could reasonably have doubts, then by definition the defendant cannot be guilty beyond a reasonable doubt.
(2) By voting "guilty beyond a reasonable doubt", you are making no statement about the reasonableness of anyone else's doubts. In effect, you are stating that it is possible for some people to have "reasonable" doubts and others to have no such doubts.
Yet it is absolutely clear that we do not view reasonable doubt this way. First of all, I suspect that if we did -- if the standard really was "is someone with a contrary view objectively unreasonable?" -- we would almost never convict anyone. But it also is implicit in the structure of appellate review of criminal convictions.
When I was an appellate clerk, we'd of course have a great many cases where defendants challenged their criminal convictions. It was pretty rare that I came across a case where I thought the defendant was "actually innocent", in the sense that I thought it was more likely than not that they didn't commit the accused crime (I can think of one potential candidate -- Paulson v. Newton Correctional Facility, 703 F.3d 416 (8th Cir. 2013)). But there were quite a few where I thought there might be reasonable doubt. I'd read the defendant's account of what happened, and I'd think "that's a live possibility." And of course I don't think I'm unreasonable. So, it seemed to me, reasonable doubt existed. If lack of such doubt is supposedly a prerequisite of a guilty verdict, how do such convictions stand?
The answer is the standard of review, which is quite strict in such cases.
We "reverse only if no reasonable jury could have found [Va] guilty beyond a reasonable doubt." United States v. Morales, 445 F.3d 1081, 1084 (8th Cir.2006) (quoting United States v. Howard, 413 F.3d 861, 864 (8th Cir.2005)). We "view the evidence in the light most favorable to the guilty verdict, granting all reasonable inferences that are supported by that evidence." United States v. Milk, 447 F.3d 593, 598 (8th Cir.2006). "The standard for reviewing a claim of insufficient evidence is strict, and a jury's guilty verdict should not be overturned lightly." United States v. Pizano, 421 F.3d 707, 719 (8th Cir.2005).United States v. Van Nguyen, 602 F. 3d 886, 897 (8th Cir. 2010).
This seems to (and in effect does) shift the burden of proof sharply in favor of upholding convictions. But it suffers from a conceptual problem, since it is effectively smashing two standards of persuasion together. The question for the reviewing court, to paraphrase the Morales decision quoted above, it whether it is unreasonable for a jury to conclude that any doubts raised are unreasonable. Those two "unreasonables" don't play nicely with one another, because again, isn't reasonability supposed to be an objective test? Clearly not, because if so we could just restate the original instructions: are there reasonable doubts or not? The only reason this deferential level of review could possibly be necessary is to uphold verdicts where reasonable doubts do seem to exist. Even granting some allowances for the fact-finders' supposedly superior ability to assess evidence and credibility, the width of the gap between how hard it (supposedly) is to convict ("beyond a reasonable doubt?") and how easy it is to uphold a verdict ("was the jury made up of unreasonable lunatics?") indicates that the test must be subjective.
The end result is that criminal convictions occur -- and probably often occur -- when a reasonable person could have reasonable doubts about the defendant's guilt. That conclusion, to me, seems deeply inconsistent with what we believe as a society tell ourselves about the criminal justice system -- though it seems quite consistent with how it operates in practice. I'm not even necessarily being critical of this result -- the alternative implied by "reasonable doubt" being an objective standard is functional de novo review of jury verdicts, which would mostly obviate the point of having a jury at all. But what it does mean is that "beyond a reasonable doubt" likely is little different from "preponderance of the evidence" -- 12 people (or fewer, depending on the circumstances) subjectively deciding the defendant was probably guilty.
UPDATE: Incidentally, the key case rejecting the claim that a proper "reasonable doubt" instruction must demand that the evidence "exclude every reasonable hypothesis other than that of guilt" is Holland v. United States, 348 U.S. 121, 139-40 (1954).