A comprehensive study of 328 criminal cases over the last 15 years in which the convicted person was exonerated suggests that there are thousands of innocent people in prison today.
Almost all the exonerations were in murder and rape cases, and that implies, according to the study, that many innocent people have been convicted of less serious crimes. But the study says they benefited neither from the intense scrutiny that murder cases tend to receive nor from the DNA evidence that can categorically establish the innocence of people convicted of rape.
Prosecutors, however, have questioned some of the methodology used in the study, which was prepared at the University of Michigan and supervised by a law professor there, Samuel R. Gross. They say that the number of exonerations is quite small when compared with the number of convictions during the 15-year period. About 2 million people are in American prisons and jails.
The study identified 199 murder exonerations, 73 of them in capital cases. It also found 120 rape exonerations. Only nine cases involved other crimes. In more than half of the cases, the defendants had been in prison for more than 10 years.
The study's authors said they picked 1989 as a starting point because that was the year of the first DNA exoneration. Of the 328 exonerations they found in the intervening years, 145 involved DNA evidence.
In 88 percent of the rape cases in the study, DNA evidence helped free the inmate. But biological evidence is far less likely to be available or provide definitive proof in other kinds of cases. Only 20 percent of the murder exonerations involved DNA evidence, and almost all of those were rape-murders.
The study, which will be presented Friday at a conference of defense lawyers in Austin, Tex., also found that very different factors contributed to wrongful convictions in rape and murder cases.
Some 90 percent of false convictions in the rape cases involved misidentification by witnesses, very often across races. In particular, the study said black men made up a disproportionate number of exonerated rape defendants.
The racial mix of those exonerated, in general, mirrored that of the prison population, and the mix of those exonerated of murder mirrored the mix of those convicted of murder. But while 29 percent of those in prison for rape are black, 65 percent of those exonerated of the crime are.
Interracial rapes are, moreover, uncommon. Rapes of white women by black men, for instance, represent less than 10 percent of all rapes, according to the Justice Department. But in half of the rape exonerations where racial data was available, black men were falsely convicted of raping white women.
"The most obvious explanation for this racial disparity is probably also the most powerful," the study says. "White Americans are much more likely to mistake one black person for another than to do the same for members of their own race."
On the other hand, the study found that the leading causes of wrongful convictions for murder were false confessions and perjury by co- defendants, informants, police officers or forensic scientists.
A separate study considering 125 cases involving false confessions was published in the North Carolina Law Review last month and found that such confessions were most common among groups vulnerable to suggestion and intimidation.
"There are three groups of people most likely to confess," said Steven A. Drizin, a law professor at Northwestern, who conducted the study with Richard A. Leo, a professor of criminology at the University of California, Irvine. "They are the mentally retarded, the mentally ill and juveniles."
Professor Drizin, too, said that false confessions were most common in murder cases.
"Those are the cases where there is the greatest pressure to obtain confessions," he said, "and confessions are often the only way to solve those crimes."
Professor Drizin said that videotaping of police interrogations would cut down on false confessions.
The authors of the Michigan study offered dueling rationales for the murder exonerations, and both reasons, they said, were disturbing.
There may be more murder exonerations, they said, because the cases attract more attention, especially when a death sentence is imposed. Death row inmates represent a quarter of 1 percent of the prison population but 22 percent of the exonerated.
That suggests that innocent people are often convicted in run-of-the-mill cases. Indeed, the study says, "if we reviewed prison sentences with the same level of care that we devote to death sentences, there would have been over 28,500 non-death-row exonerations in the past 15 years rather than the 255 that have in fact occurred."
The study offered a competing theory, as well. Mistakes, it said, may be more likely in murder cases and far more likely in capital cases.
"The truth," the study concludes, "is clearly a combination of these two appalling possibilities."
Critics of the Michigan study questioned its methodology, saying it overstated the number of authentically innocent people. The study calls every nullification of a conviction by a governor, court or prosecutor declaring a person not guilty of a crime an exoneration.
In Astoria, Ore., Joshua Marquis, the district attorney for Clatsop County, said that many of the people exonerated under the study's definition may nonetheless have committed the crimes in question, though the evidence may have become too weak to prove that beyond a reasonably doubt.
"The real number of people on death row exonerated in the sense of being actually innocent in the modern era of the death penalty is about 25 to 30," Mr. Marquis said. The Michigan study put the number at 73.
He added that even the error rate suggested by the study was tolerable given the American prison population.
"We all agree that it is better for 10 guilty men to go free than for one innocent man to be convicted," Mr. Marquis said. "Is it better for 100,000 guilty men to walk free rather than have one innocent man convicted? The cost-benefit policy answer is no."
At the University of Michigan, Professor Gross said that was the wrong calculus.
"No rate of preventable errors that destroy people's lives and destroy the lives of those close to them is acceptable," he said.
Barry Scheck, a founder of the Innocence Project, said Mr. Marquis's analysis ignored another point.
"Every time an innocent person is convicted," Mr. Scheck said, "it means there are more guilty people out there who are still committing crimes."
Study Suspects Thousands of False Convictions