LOS ANGELES (AP) _ California trial judges generally have accepted DNA tests in criminal cases, but appeals courts have split on whether a sophisticated form of the tests should be allowed.

At least two appeals courts have ruled that Restriction Fragment Length Polymorphism (RFLP) tests should have been disallowed in criminal trials because scientists couldn't agree on how to interpret the results.

The tests are based on comparisons of DNA fragments from hair or tissue samples at a crime scene, with samples taken from suspects and victims.

Prosecutors and defense attorneys agree that obvious differences in segments of DNA can easily eliminate a suspect. The disagreements arise when the tests are proposed to positively identify a suspect.

Because there are no known eyewitnesses, prosecutors in the O.J. Simpson case want to use DNA tests to help prove that he killed his ex-wife and her friend.

California requires that scientific methods be generally accepted in the scientific community before they are admissible in court. Appeals court judges have found no problems with the RFLP test process but have found disagreement among experts on how to read the results.

None of the appellate rulings reversed verdicts.

Judges in Washington, Arizona, New Hampshire and Massachusetts have ruled that disputes about statistical analysis render the RFLP evidence inadmissible. Twenty-six states allow the tests as evidence. The remaining states have no formal policy.

George Clark, a deputy district attorney in San Diego, said there had been no appellate rulings yet in California on the admissibility of results from a simpler form of DNA testing, called Polymerase Chain Reaction (PCR) tests.

He said he knew of only two trials where PCR test results were disallowed.

In one of those, a 1991 murder case, seven of eight expert witnesses testified to general acceptance of the technique. But one said it was not ready for the courtroom and the trial judge in Sacramento agreed.

The U.S. Supreme Court loosened standards last year for the admission of expert testimony in federal cases, making general scientific consensus only a consideration, not a requirement, for courtroom use.