Science and law have led to the exoneration of hundreds of criminal defendants in the US in recent decades.
But big questions remain: How many other innocent defendants are locked up? How many are wrongly executed?
About one in 25 people imprisoned under a death sentence is likely innocent, according to a new statistical study appearing in the Proceedings of the National Academy of Sciences. And that means it is all but certain that at least several of the 1,320 people executed since 1977 were innocent, the study says.
From 1973 to 2004, 1.6 per cent of those sentenced to death - 138 prisoners - were exonerated and released because of innocence.
But most innocent people sentenced to death are never identified and freed, says professor Samuel Gross of the University of Michigan Law School, the study's lead author.
The difficulty in identifying innocent inmates stems from the fact about 60 per cent of prisoners on death row are resentenced to life imprisonment. Once that happens, their cases no longer receive the exhaustive reviews that the legal system provides for those on death row.
Gross and three other researchers, including a biostatistics expert, looked at the issue using a technique often used in medical studies called survival analysis. Yale University biostatistics expert Theodore Holford, who wasn't part of the study, said the work done by Gross "seems to be a reasonable way to look at these data".
Because of various assumptions, it might be best to use the margin of error in the study and say the innocence rate was probably between 2.8 per cent and 5.2 per cent, said University of South Carolina statistics professor John Grego, who was not part of the study.
The study is the first to use solid and appropriate statistical methods to address questions of exoneration or false convictions, an important subject, Columbia Law School professor Jeffrey Fagan said.
He said the research combined data from three independent sources, a rigorous approach used by few studies on capital punishment.
The research produced an estimate of the percentage of defendants who would be exonerated if they all remained indefinitely on death row, where their cases would be subject to intense scrutiny for innocence.
The study concluded that the number of innocent defendants put to death is "comparatively low. ... Our data and the experience of practitioners in the field both indicate that the criminal justice system goes to far greater lengths to avoid executing innocent defendants than to prevent them from staying in prison indefinitely."
Death sentences represent less than one-tenth of 1 per cent of prison sentences in the US, but they account for 12 per cent of known exonerations of innocent defendants from 1989 to 2012. One big reason is that far more attention and resources are devoted to reviewing and reconsidering death sentences.
"The high rate of exoneration among death-sentenced defendants appears driven by the threat of execution," the study says. "But most death-sentenced defendants are removed from death row and resentenced to life, after which the likelihood of exoneration drops sharply."