Quite a lot of legal systems have used torture to try to force defendants and/or witnesses to tell the truth. Moral problems aside, the obvious problem is that an innocent defendant may confess, or a witness give the testimony he thinks the interrogator wants, to avoid the torture.
Obvious not only to us but to them. The legal literature of Imperial China, where torture of both witnesses and defendants was legal, contains accounts of cases where a clever official figured out that a confession under torture was bogus and successfully identified the real criminal.
Much of what we know about the Athenian legal system comes from orations written by professional orators to be spoken by plaintiffs or defendants. They include two orations, written by the same orator for different cases, one of which argues that testimony under torture (of slaves) is entirely reliable, one of which gives the obvious argument for why it is not—in each case, of course, supporting his client's position.
I recently came across an ingenious, if imperfect, solution to the problem in what is apparently the oldest surviving Germanic law code—the Visigothic Forum Judicum
. Alaric accuses Beremud of murder. There is not enough evidence to convict Beremud, so he is tortured to make him confess. Under the code:
“The judge shall take the precaution to compel the accuser to specifically describe the alleged offence, in writing; and after he has done so, and presented it privately to the judge, the torture shall proceed; and if the confession of him who is subjected to the torture should correspond with the terms of the accusation, his guilt shall be considered to be established. But if the accusation should allege one thing, and the confession of the person tortured the opposite, the accuser must undergo the penalty hereinbefore provided; because persons often accuse themselves of crime while being tortured.
If Beremud is innocent, he won't know the details of the offense—where and when the victim was killed, with what weapon, what wounds were inflicted. However much he wants to stop the pain, he can't confess what he does not know. This assumes, of course, that an innocent defendant will not know the details of the crime. The code dealt with that problem:
But if the accuser, before he has secretly given the written accusation to the judge as aforesaid, should, either in his own proper person, or by anyone else, inform the party of what he is accused, then it shall not be lawful for the judge to subject the latter to torture, because the alleged offence has become publicly known.”
There is at least one problem with this solution to the problem of false confessions. A guilty defendant, knowing the law, may confess under torture and deliberately get the details wrong. My conjecture is that that problem was dealt with by having the interrogator keep up the torture until either the defendant confessed with the right details or the interrogator was convinced that he could not. I suspect that most of us, having had the good fortune not to have been subjected to torture, overestimate how easy is to hold out when saying the right words will end the present pain.
It may occur to some readers that a variant of the Visigothic solution survives in modern practice. Police officers are not supposed to torture defendants, but they are allowed to interrogate them, and interrogation sometimes leads to confessions—which may make one suspect that it comes closer to torture than it is supposed to. One way of checking whether the confession is for real is to see whether it contains details that an innocent defendant would not know.
Of course, this depends on the honesty of the officers doing the interrogation—they can, and for all I know sometimes do, subvert the process by feeding the defendant information that they have and he, if innocent, does not.
P.S. A talk I just heard on the problem of jailhouse snitches points out a different modern version of the problem—the "confession" by a defendant that may have been invented by the person it was supposedly confessed to, a fellow criminal eager to testify in exchange for lenient treatment in his own case. Here again, one test of truth is whether the confession contains details that the real criminal would know and an innocent defendant, or the fellow criminal who claims to have heard the confession, would not. Here again, the problem is that the snitch may have been fed that information by the police or prosecution, or obtained it himself in some other way.
One further problem, in that case, is that the snitch may be, in some cases turned out to have been, the real criminal, eager to convict someone else for his crime. And well informed about the details.