A new study takes a crack at quantifying it.

Criminal justice scholars often say that the true number of innocent people convicted of crimes is unknown--in fact, unknowable. A new University of Michigan study challenges that belief in one important context.

Among defendants sentenced to death in the United States since 1973, at least 2.3 percent--and possibly more--were falsely convicted, said U-M law professor Samuel Gross in a study co-authored by Barbara O'Brien, a professor at Michigan State University College of Law.

If defendants who were sentenced to prison had been freed because of innocence at the same rate as those who were sentenced to death, there would have been nearly 87,000 non-death row exonerations in the United States from 1989 through 2003, rather than the 266 that were reported, the study said.

"The main thing we can safely conclude from exonerations of falsely convicted defendants is that there are many other false convictions that we have not discovered," said Gross, whose research has focused on the death penalty, false convictions and eyewitness identification.

Since 1989, nearly all exonerations in the United States fall into three categories: rape convictions, because of post-conviction DNA testing; murder convictions, and especially death sentences, which are subjected to much more detailed post-conviction reinvestigation than other convictions; and a few groups of false drug and gun possession convictions that were produced by concerted programs of police perjury that later unraveled.

As result, researchers know little about false convictions among crimes of violence other than murder or rape, even though false convictions for robbery could greatly outnumber those for rape and murder. And researchers know next to nothing about false convictions for other types of crimes, such as property crimes, misdemeanors and white collar crimes.