Roanoke Times
Copyright (c) 1995, Landmark Communications, Inc.
DATE: THURSDAY, October 5, 1995 TAG: 9510050078
SECTION: EDITORIAL PAGE: A-13 EDITION: METRO
SOURCE: BRUCE S. WEIR
DATELINE: LENGTH: Medium
Was it too much to expect a jury of 12 nonstatisticians to have understood the overwhelming nature of the DNA matches between 45 blood stains collected at the Bundy crime scene, in Simpson's Bronco and at his Rockingham estate?
Did they reject figures like 1 in 57 billion, the statistical frequency mentioned by prosecutor Marcia Clark in her closing argument to show it was Simpson's blood on the Bundy rear gate - because of the defense's arguments about police conspiracy, or because the numbers were based on complex statistical arguments that left them bewildered and intimidated?
My interest in this is personal as well as professional: I was the expert witness called by the prosecution to tell the jury that these astronomical numbers are based on good scientific arguments. I would like to think the jury accepted my statements. But it would have been a lot easier if prosecutors could have simply said that forensics experts had found DNA matches and that this pointed to Simpson's involvement with the crime.
A radical concept? Not really. After all, fingerprint experts can testify that a defendant's prints were found at a crime scene, and no numbers or scientific arguments are required to support their conclusion. Why then should a jury be required to understand complex genetic statistics in order to decide a case involving DNA evidence? Does anyone really understand what ``one in 57 billion'' means?
These are not theoretical questions. Public perception about the meaning of DNA matches will affect court cases for years. The use of DNA forensics is on trial as well as Simpson. If confidence in this new technology is lost, we will all be the poorer. Conviction of the guilty is only one use of DNA forensics. It is also used to clear the innocent, to identify fathers in paternity disputes, to identify war remains and to protect proprietary crop varieties.
So what should be done?
For the present, I believe forensic scientists must continue to give mind-blowing statistical frequencies to support our conclusions about the strength of DNA matches, providing we are clear that the numbers presented are meant only to indicate the relative rarity of the kinds of DNA profiles being analyzed and not as a probability of guilt or innocence.
For instance, blood collected from the rear gate at the Bundy crime scene had 14 matching DNA bands, and each of these occurs in about 10 percent of the population. The odds of someone having all 14 bands in their DNA profile, as Simpson does, is one in 57 billion. This does not mean there is only a one in 57 billion chance that someone other than Simpson committed the murders. That is a different issue.
Ironically, the extreme power of detailed DNA profile matches brings the danger of disbelief in or misinterpretation of the numbers.
Nonetheless, I believe that the forensics use of DNA will eventually become as readily accepted as fingerprints. The amount of detail in the profiles is becoming so great that the chance of finding two people with the same type is already essentially nil.
Here's an analogy: Suppose a slot machine had 14 wheels, each with all 94 symbols found on a standard computer keyboard. Suppose you pull the lever and the wheels spell out ``Nicole Simpson.'' This is analogous to the strength of the match of Nicole Brown Simpson's blood on the socks found in Simpson's bedroom. As a statistician I could give the odds on this occurring. But anyone who has tried their luck on a slot machine with just three wheels can understand that coincidence is the least likely explanation. In this case, statistical frequencies are not necessary.
The DNA in all 45 blood stains, some with as many as 42 matching components, points to Simpson, Nicole Brown Simpson and Ron Goldman as being the contributors to these stains. In an attempt to convey how strong these matches are, the prosecution presented frequency estimates for the profiles. I look forward to the day when prosecution and defense agree that matches are very unlikely to be coincidences and dispense with numbers that are beyond our comprehension.
Dr. Bruce Weir, professor of statistics and genetics at North Carolina State University, has published nearly 100 scholarly articles and books on the statistics of genetics. He wrote this column for The Charlotte Observer.
- Knight-Ridder Newspapers
by CNB