Prof. Larry Laudan
Lecturer in Law and Philosophy
University of Texas at Austin
“‘Not Guilty’: The Misleading Verdict and How It Fails to Serve either Society or the Innocent Defendant”
Most legal systems in the developed world share in common a two-tier verdict system: ‘guilty’ and ‘not guilty’. Typically, the standard for a judgment of guilty is set very high while the standard for a not-guilty verdict (if we can call it that) is quite low. That means any level of apparent guilt less than about 90% confidence that the defendant committed the crime leads to an acquittal (90% being the usual gloss on proof beyond a reasonable doubt, although few legal systems venture a definition of BARD that precise). According to conventional wisdom, the major reason for setting the standard as high as we do is the desire, even the moral necessity, to shield the innocent from false conviction.
There is, however, an egregious drawback to a legal system so structured. To wit, a verdict of ‘not guilty’ tells us nothing whatever about whether it is reasonable to believe that the defendant did not commit the crime. It offers no grounds whatever for inferring that an acquitted defendant probably did not commit the crime. That fact alone should make most of us leery about someone acquitted of a felony. Will a bank happily hire someone recently acquitted of a forgery charge? Are the neighbors going to rest easy when one of them was charged with, and then acquitted of, child molestation?
While the current proof standard provides ample protection to the innocent from being falsely convicted (the false positive rate is ~3%), it does little or nothing to protect the reputation of the truly innocent defendants. If properly understood, it fails to send any message to the general public about how they should regard and treat an acquitted defendant because it fails to tell the public whether it’s likely or unlikely that he committed the crime.
It would not be difficult to remedy this centuries-old mess, both for the public and for the acquitted defendant, by employing a three-verdict system, as the Scots have been doing for some time. Their verdicts are: guilty, guilt not proven and innocent. In a Scottish trial, if guilt is proven beyond a reasonable doubt, the defendant is found guilty; if the jury thinks it more likely than not that defendant committed no crime, his verdict is ‘innocent’; if the jury suspects that defendant did the crime but is not sure beyond all reasonable doubt, the verdict is ‘guilt not proven’. Both the guilt-not-proven verdict and the innocence verdict are officially acquittals in the sense that those receiving it serve no jail time. (This gives a whole new meaning to the well-known phrase ‘going scot-free’.)
The Scottish verdict pattern serves the interests of both the innocent defendant and the general society. The Scots know that if a defendant received an innocent verdict, then the jury believed it likely that he committed no crime and that he should be treated accordingly. That is both important information for the citizenry and a substantial protection for the innocent defendant himself, since the innocent verdict is in effect an exoneration, entailing the likelihood of his innocence.
On the other hand, the Scottish guilt-not-proven verdict sends out the important message to citizens that no other Anglo-Saxon legal system can; to wit, that the acquitted defendant (with a guilt-not-proven verdict) should be treated warily by society since he was probably the culprit, even though he was neither convicted nor punished.
Interestingly, there is ample use of the intermediary verdict. The Scottish government reports in a study of criminal prosecutions in 2005 and 2006 that it turned out that 71% of those defendants tried for homicide and acquitted received a ‘guilt-not-proven’ verdict. That means that about 7-in-10 acquittals for murder in Scotland involved defendants regarded by the jurors as having probably committed the crime. In a more recent analysis, the Scottish government reported that in rape cases some 35% of acquittals resulted in ‘guilt not proven’ verdicts. In murder cases, the probably guilty verdict rate was 27% of all acquittals.
It’s worth adding that Scotland’s intermediary verdict gives us access to information on an error whose frequency no other Western legal system can easily compute: to wit, the frequency of false acquittals. It tells us that, at least in Scotland, the rate of false acquittals hovers between 1-in-4 and 1-in-3. That is crucial information for those of us who believe that a legitimate system of inquiry—whether a legal one or otherwise— must get a handle on its error rates. Without knowing that, we cannot possibly figure out whether the distribution of erroneous verdicts is in line with our beliefs about the respective costs of the two errors.
Scottish criminal law has one other interesting feature worthy of mention in this context: a verdict there requires only a majority vote from the 15 citizens who serve as the jury. By contrast, most American states require a unanimous vote among 12 jurors, contributing to a situation in which mistrials are both expensive and common. They are expensive because they usually lead to re-trials, which are rarely cheap. In some jurisdictions in the US, 20% or more of trials end in a hung jury. Not surprisingly, hung juries in Scottish cases are much less frequent.
 See http://www.scotland.gov.uk/Publications/2006/04/25104019/11.) See also the Scottish Government Statistical Bulletin, Crim/2006/Part 11.
 See Scottish Government, Criminal Proceedings in Scotland, 2013-14, Table 2B.
 A study by Paula Agor et al., (Are Hung Juries a Problem? National Center for State Courts and National Institute of Justice, 2002) found that in Washington, D.C. Superior Courts some 22.4% of jury trials ended in a hung jury; In Los Angeles Superior Courts, the hung jury rate was 19.5%.
- Larry Laudan, “Need Verdicts Come in Pairs?” International Journal of Evidence and Proof, vol. 14 (2010), 1-24.
Previous guest posts:
- Larry Laudan (July 20, 2013): Guest Post: Larry Laudan. Why Presuming Innocence is Not a Bayesian Prior
- Larry Laudan (July 3, 2015): “When the ‘Not-Guilty’ Falsely Pass for Innocent”, the Frequency of False Acquittals (guest post)
Among Laudan’s books:
1977. Progress and its Problems: Towards a Theory of Scientific Growth
1981. Science and Hypothesis
1984. Science and Values
1990. Science and Relativism: Dialogues on the Philosophy of Science
1996. Beyond Positivism and Relativism
2006. Truth, Error and Criminal Law: An Essay in Legal Epistemology
Larry: Thanks so much for your new guest post. Its interesting: I’ve always considered the legal standpoint of not guilty to capture an important aspect of my standpoint about testing in science. It’s not how probable a claim is but how well probed, so the onus is, or should be, similarly on whether a good job has been done in probing the ways in which interpretations of data can be erroneous. Never mind if it’s plausible or believable (on other grounds, perhaps), which is a distinct standpoint. This contrasts with the philosophy of “probabilism” so popular in philosophy of statistics. So now the question is whether this third category of “guilt not proven” is an ever better analogy for the statistical philosophy I favor. I realize that’s not the point of your post. However, I can see real benefits to labeling a lot of reported statistically significant results as “effect not shown” or “effect not statistically proven”–particularly those that don’t replicate. This could be a possibly temporary label until either the credentials are fixed or the reported effect is falsified altogether.
I hope to hear from some legal people.
Thank you for your thoughtful post on the standard of proof in criminal cases.
My sense is that it is misleading to use the term “confidence” in this context, at least when speaking to statisticians, who will tend to understand the term through the lens of Neyman’s frequentist statistical theory and practice. And assuming that anyone could actually quantify his “posterior probability” of guilty, that is probability of guilty given the all the evidence in the case, I am horrified to think that this probability is set at 90% in any civilized, just society. You are certainly correct that American courts refuse to quantify “proof beyond a reasonable doubt,” but many commentators freely attempt to define it, and get it wrong. Here’s my humble attempt of practical certainty for a capital murder case: “Ladies and gentlemen of the jury, it is my privilege to represent the people of this State in this case. I will show that the defendant is guilty of all elements of the crime, beyond a reasonable doubt. Certainly, I have looked very carefully at the evidence before agreeing to bring this prosecution. And if I am ever shown to be wrong, I will trade places with the defendant and take whatever punishment I am asking you or the court to impose upon him.”
Any attempt to quantify the “false conviction” rate must take into account several nuances. First, in the United States, we have a system of plea bargaining in which prosecutors may pressure defendants into compromising their right to trial and accepting a guilty plea. Second, the falsity of the conviction itself has gradations. Someone who may be clearly be guilty of negligent homicide may be false convicted of manslaughter, or of murder. In other words, the falsity may be with respect to the conduct or the mens rea elements of the crime.
The Scots certainly have a more nuanced scheme for jurors to use in evaluating criminal trial evidence, but defendants in Scotland have no constitutional right against “double jeopardy.” I recall that there was a legislative effort to afford defendants such a right, but I lost track of whether the bill was actually enacted. See http://www.scotlawcom.gov.uk/law-reform/law-reform-projects/completed-projects/double-jeopardy/
The implications of a “not guilty” verdict can be seen in the U.S., where the defendant may not be retried for the same crime, but he may be sued in tort for his conduct, as famously happened in the O.J. Simpson case.