Saying that convictions are too often based on inaccurate identification, a report ordered by the New Jersey Supreme Court has called for mandatory pretrial hearings to evaluate the testimony of eyewitnesses in all criminal cases.

The special report, filed Monday, said judges should consider the wealth of scientific studies that explain why the most certain of witnesses may be wrong, and how to ensure better reliability.

If the recommendations are adopted by the state's highest court, New Jersey would have the country's most rigorous system to deal with eyewitness testimony, which is known to carry great weight with juries but has attracted increasing concern.

"It's very significant," said Gary Wells, a psychology professor at Iowa State University who testified before the retired New Jersey judge who examined the issue.

The report was ordered in the appeal of a Camden defendant, Larry R. Henderson, who contended that a lineup in his manslaughter case violated 2001 guidelines on eyewitnesses.

The high court ordered a sweeping inquiry into the issue, with input from the state Attorney General's Office, the state Public Defender, the Association of Criminal Defense Lawyers of New Jersey, and the New York-based Innocence Project.

More than 200 studies, articles, and books were examined, and seven experts testified before retired appellate Judge Geoffrey Gaulkin, who filed the 88-page report with the high court.

Thus far, Gaulkin wrote, "only bits and pieces of the science" that casts doubt on the reliability of eyewitness testimony "have found their way into the New Jersey courtrooms."

Embracing studies that questioned the validity of many eyewitness' reports and holding pretrial hearings to assess eyewitness' reliability, Gaulkin concluded, "would remedy the flaws and inadequacies" in how such evidence is handled and would "improve the assessment of eyewitness reliability by judges and jurors alike."

The unreliability of eyewitness testimony has come to light because of DNA testing. According to the Innocence Project, three-quarters of the 254 defendants exonerated nationally by DNA testing were convicted largely on the basis of eyewitness testimony.

Of the five DNA exonerations in New Jersey, three involved mistaken identifications, the report said.

The problem is not that witnesses lie, but that they are mistaken. Studies attribute the errors to a variety of causes. Line-up or photographic identification practices may be flawed, memory changes with time, and a victim or witness of one race may have difficulty identifying an attacker of another.

The presence of a gun also has a significant effect. It shifts the victim's attention from the perpetrator to the weapon, making identification more difficult.

"The science abundantly demonstrates the many vagaries of memory encoding, storage and retrieval; the malleability of memory; the contaminating effects of extrinsive information; the influence of police interview techniques and identification procedures; and the many other factors that bear on the reliability of eyewitness identifications," Gaulkin wrote.

The state Attorney General's Office would not comment on Monday's report, but prosecutors have expressed concern that too many identifications could be at risk of being suppressed from evidence.

"Under a suppression regime, the people of New Jersey would suffer the injustice of losing evidence, and quite possibly the case, because a witness was threatened and the police tried to reassure him," Special Deputy Attorney General John McNamara Jr. and Deputy Attorney General Deborah Bartolomey wrote in a brief submitted to Gaulkin.

According to the report, witnesses' recollections can be affected by all sorts of variables, including stress, comments by police at a lineup, the passage of time, and "unconscious transference," in which a witness confuses someone near a crime scene with the perpetrator.

To have an identification suppressed under current law, a defendant must show that an identification procedure was influenced by police and that the identification was most likely wrong.

Under recommendations made Monday, the burden would shift to prosecutors, who would have to show during a pretrial hearing that the eyewitness was reliable.

Ezekiel R. Edwards, a lawyer with the Innocence Project who took part in the hearings, said no other state had ordered such an exhaustive study of the issue.

If the high court mandates pretrial hearings in all criminal cases involving eyewitnesses, Edwards said, New Jersey would be at the vanguard nationally.

"New Jersey would be the first place to conduct comprehensive scientifically informed reliability assessments up front," he said.

A problematic witness' testimony might not be suppressed outright, however. Expert witnesses could be called to testify about how memory is affected, or judges might give the jury cautionary instructions.

Wells said that New Jersey already was a front-runner on the issue with the adoption of guidelines in 2001 on how lineups should be conducted to ensure better reliability.

In contrast, he said, Pennsylvania is "not very far along in any of this."

James Doyle, a lawyer who testified before Gaulkin and who wrote True Witness: Cops, Courts, Science, and the Battle Against Misidentification, said the report acknowledges what science and the law have come to understand over the last century.

"It's like finally admitting that the Earth is round," he said.