Cavazos v Smith I

Cavazos is a habeas petition involving a grandmother convicted of killing her grandson by violently shaking him. The case implicates two important issues involving courtroom science. In this post, I’ll look at what it suggests about the role of dueling experts. In my next post, I’ll look at how the dissent—which dives head-first into the medical literature on shaken baby syndrome—uses scientific evidence.

The question at the heart of Cavazos is whether a 7-week old boy, Etzel Glass, died one night from SIDS or due to shaken baby syndrome (SBS) at the hands of his grandmother, Shirley Smith. At trial, the jurors heard from dueling pathologists about the likely cause of death, and they ultimately accepted the story of the prosecution’s doctors and convicted Smith.

The evidence in the case was ambiguous. Many (though not all) of the “tell-tale” signs of SBS were missing, and the state’s experts concluded that death resulted from an undetected—and perhaps undetectable—shearing of the brainstem. Smith’s pathologists noted that Etzel was a low birth-weight child born with jaundice and a heart murmur, all of which are risk factors for SIDS. A classic dueling-experts situation.

While all the state appellate courts and the federal district court upheld Smith’s conviction, the Ninth Circuit tossed it on the grounds that the conviction was based on an unreasonable determination of the facts, since there was “no evidence to permit an expert conclusion one way or the other.” Though a rather narrow claim, I think it actually points to a potentially far more radical take on dueling experts that I want to consider^* here.

The California Court of Appeal, in upholding the conviction, and the Ninth Circuit, in reversing it, lay out potentially different conceptions of the jury’s role in the presence of dueling experts. The CCA adopts the conventional view on jury fact-finding in complex scientific cases: “The expert opinion evidence … was conflicting. It was for the jury to resolve the conflicts. The evidence was substantial and sufficient to support the jury’s conclusions….” The Ninth Circuit, on the other hand, argues that the scientific uncertainty here bars a beyond-a-reasonable-doubt conclusion about guilt as a matter of law.

In a narrow sense, the Ninth Circuit’s claim, if true,^** is perhaps not all that daring. It seems to be arguing that the state’s evidence, in isolation, is insufficient. But the court also points to other exculpating empirical evidence (or at least empirical claims^***) about Etzel’s social environment to bolster its insufficiency claim. So it isn’t entirely clear that the panel is reversing because the state provided no evidence, but because it provided insufficient evidence given the evidence brought to bear by the defense.

One reading of this would simply be “judicial activism” of a sort: the Ninth Circuit replaced the jury’s weighing of expert testimony with its own. And perhaps that is what actually happened; and if so, the Supreme Court’s majority is surely right that that is precluded by AEDPA and Jackson v Virginia.

But there is a different, broader way to read the Circuit court opinion that in some ways avoids the AEDPA problem (but introduces a host of other problems): what if we have a blanket rule that when both the state and the defense introduce plausible, reliable and contradictory scientific evidence  about an essential element of a crime, the state automatically loses? (I will note here that accurately making the prima facie assessment of reliability and credibility is a tough issue I’ll return to again and again this year but put aside for now.)

Think of it this way: when both sides introduce credible expert testimony on an element of a crime, perhaps there is no uncertain fact for a jury to “find.” What we have is a “known unknown”: we know that we do not know the answer.^† And not knowing seems to be reasonable doubt; or, phrased more carefully, in the presence of a “known unknown” it is unreasonable to say that you know one way or the other. And at least in criminal cases, that should tip the balance automatically in favor of the defense.

Of course, juries are already free to find a “known unknown” if they wish. I want to push the issue further: as a matter of law, should we compel such a non-finding? Even before getting to questions of jury competence, is there some sort of particularly important meta-evidence we get from credible dueling experts? Does the very inability to agree suggest, at least in criminal cases, that as a matter of law there is no fact for the jury to find? We know that we don’t know, and it is unreasonable to argue that you do know in such a setting.

Even if you’re not convinced by this argument—and I’m still trying to decide if I am—the jurors’ lack of epistemic competence may still play a role: while perhaps in theory there is some fact to be “found” here, the people we are asking to find it are effectively blind. Jury pools are not well-educated: at best about half are college graduates (and this from a study set in Connecticut, the sixth-best educated state in the country), and few college graduates have real math/science training. So if dueling experts do not theoretically demand a non-finding, does the clear inability of the jury pool to make the necessary finding along rational lines do so pragmatically?

This is a radical claim, I admit, and I’m generally wary of let’s-start-from-scratch proposals; if nothing else, Tal Golan’s history of the centuries-long failure to meaningfully reform courtroom science proves that even more-modest proposal generally seem doomed. On the other hand, Mirjan Damaška claims that the current “scientization” of the law is the biggest challenge it’s faced since proof-by-ordeal was replaced by proof-by-jury, so maybe this is the time to Think Big.

That said, it may not be that the solution requires a radical reformation of the law, just substantially greater honesty. Smith was convicted under Cal. Penal Code § 273ab, which punishes:

any person, having the care … of a child …, who assaults that child … resulting in the child’s death ….

Under my suggestion here, Smith could not be convicted since we cannot prove that she assaulted the child. So what if we rewrote the statute to punish:

any person, having the care … of a child …, who plausibly assaults that child … resulting in the child’s death ….

In effect, this is the statute under which Smith was convicted, given the ambiguous evidence and the limited epistemological competence of the jury; maybe we even need to replace “plausibly” with “potentially.” Under this statute, the state’s experts can advance a causal story and the defense can try to show that it is preposterous. Perhaps jurors are far more competent at deciding whether the state is telling a viable story more than deciding whose story is more convincing.

Forcing us to rewrite statutes along these lines lays bare the implications of the legal fiction of “jury factfinding” in situations of scientific complexity and "known unknowns." They are not finding that x caused y; at best they are finding that x perhaps caused y, or that it is not completely impossible that x caused y, or something along those lines. The revised statute may be unpopular—how can we convict someone for possibly causing a death?—but again, stripping away the legal fiction of “jury factfinding” here, that is what we are doing.

(It is worth point out, briefly, that administrative law is honest in just this way: agencies are allowed to issue regulations when the causal relationship is unknown.)

Eliminating the fiction surely must advance justice. Perhaps we want a lesser “plausible” offense, so we punish people less in the presence of scientific complexity. Perhaps we don’t want the “plausible” statute at all, and so eliminating the legal fiction makes us appreciate the extent to which we have been wrongly punishing people. These are all normative questions, but it should be clear now that the legal fiction that concerns me has saved us from having to ask them. And they are not easy questions to ask.

And, it is worth pointing out, these questions are actually easier to answer in the criminal context, where we are more comfortable stacking the deck against one party. How to handle known unknowns is even tougher in the tort context, which is an issue I will also consider in future posts.

 

^* I almost used “unpack” rather than “consider,” but I am really working hard to purge law-professor-slang from my vocabulary.

^** As the Supreme Court points out, the Ninth Circuit’s statement of the facts is not entirely correct. While the state’s experts admit that they could not find evidence of the brainstem shear, and that in fact such evidence may have been impossible to find, the expert also said he could rule out SIDS due to other symptoms. Thus the state’s expert was not just guessing, but instead was saying “there are two possibilities, I believe I can eliminate one, and the evidence we have is consistent with the second.”

^*** The magistrate judge states that grandmothers rarely are charged with SBS, that there was no evidence of social stress that could have caused the defendant to snap, etc. It is unclear if the magistrate cited (or at least encountered) studies to validate these claims, although they are in general facially plausible.

^† It is unfortunate that Donald Rumsfeld has been mocked for using this term, and it is unfortunate that the term now often elicits a snide laugh because of its association with Rumsfeld. The concept is a profoundly important one, and whatever your impression of Rumsfeld he used it completely correctly in his speech.