There is another limitation to the RFLP method. "Each additional probing adds four or five days. With each additional probing, you have to wash the gel," Ferrara says. "After the third or fourth probing, you start to lose too much DNA to maintain accuracy."

The catch-22 of the testing is that no one knows how many markers need to be tested. "No one has defined what that threshold might be," said Sensabaugh. "An ad hoc number like 8 to 10 provides the presumption of rebuttable proof of identification."

But Mark Stolorow, director of operations for Cellmark Diagnostics of Germantown, Md., says his firm generally uses five radioactive probes. "We can generally get at least 99.99 percent exclusion based on five probes," Stolorow says.

PCR analysis, using a technique whose inventors won the 1993 Nobel Prize, is a sort of molecular copying process that zeros in on and amplifies a particular region of DNA. The differences at that site can then be distinguished through specialized probes that tiger a color change or through gel electrophoresis followed by a silver stain.

"PCR is a quick process that requires very small samples but is not as specific, and contamination is more of a problem," Ferrara says. But, even "the saliva on the back of a postage stamp or on the butt of a cigarette has enough DNA to analyze."

However, the PCR method doesn't always provide numbers as small as RFLP.

The two methods can be used in concert to get even better probabilities, Ferrara says. "Let's say that you do the PCR tests and get numbers of 1 in 20 frequency. The RFLP then gives you, let's say, 1 in 100 thousand. Put them together, and you get 1 in 2 million."

The Controversy

Lewontin and Hartl have fueled controversy over the statistics, not the techniques, of DNA profiling. "The foundation of the molecular biology is well laid," says James F. Crow, professor emeritus of genetics at the University of Wisconsin. "The difficulties are matters of the interpretation" of the profiles.

The original statistical calculation used strict multiplication of probabilities.

A given part of a profile is compared to a representatives database of other profiles to determine the number of times that part would reoccur in the database.

Under the multiplication rule, if one probe gave odds of one in 20 and a second gave odds of one in five, the total odds reported were one in 100.

"Some population geneticists felt that strict multiplication was not a sufficiently valid and conservative method because the size of the databases was considered too small to justify such large numbers (e.g., a database of 100 profiles leading to a calculation of a frequency of 1 in 100 million)," Ferrara says.

The National Research Council (NRC), an arm of the National Academy of Sciences, published a report in 1992 on "DNA Technology in Forensic Science." The report developed a technique of interpretation called the "ceiling principle."

"The ceiling compromise puts a limiter on how we employ multiplication so the numbers aren't so big," says Ferrara, who does not "personally ascribe" to the principle.