Finding Truth Is a Slippery Thing But Unlike the Law, Science Has More Time : Science File / An exploration of issues and trends affecting science, medicine and the environment.

Scientific evidence.” The term has such a pure sound, the clear ring of truth.

Yet absolute truth rarely emerges from the laboratory neatly labeled and mature. Instead, it slips out sloppily, develops in fits and starts, and can hang around for a long time unnoticed.

The evolution of scientific truth has a great deal in common with legal truth as it emerges in the messy world of the courtroom--most notably in the O.J. Simpson trial. Both law and science must weigh the reliability of evidence, the trustworthiness of experts and the probability that something is true beyond a reasonable doubt.

Scientific truth, like legal truth, is less a collection of facts than a running argument. And ultimately, scientists and attorneys must present their arguments before a jury of peers.

One big difference between law and science, of course, is in the immediate consequences. If a scientific fact or theory turns out to be false or incomplete, experimentation will eventually set things straight. Nature will show its hand.

But law cannot wait for the luxury of complete information; it has to decide. The Supreme Court recently put it this way: “Scientific truths are subject to perpetual revision. Law, on the other hand, must resolve disputes finally and quickly.”

The luxury of time is one reason that scientists are more comfortable than juries with probable truths. In fact, laws of nature are rarely proved absolutely true. They are only proved untrue.

“The longing for absolute truth has to do with religion, not science,” said historian of science Lorraine Daston of the University of Chicago. Only if scientists perform infinite numbers of experiments can they say for certain that something holds in all possible cases. Theories always hold the door open for revisions based on new revelations.

“Most people think about science as an absolute yes or no, absolute true or false,” said physicist Haim Harari of the Weizmann Institute of Physics. “But there are many things in science at the stage where all you can do is rough approximation.”

The idea that the best science can do is give odds on the truth goes back at least as far as Galileo, according to Harvard University science historian Gerald Holton. “In the historic view, that’s where the notion of absolute truth disappears.”

It was Galileo who abandoned the more cerebral, theory-based truths and decided that the ultimate arbiter should be experiments. Every time the experiment turns out the way theory says it does, the theory gains a degree of validity.

“Experiment makes a hypothesis more and more probable,” Holton said. “But it cannot verify it. That was a significant relaxation of the requirement for useful truth in science. . . . From that time on, truths were statistical.” A recent example of this kind of statistical truth-seeking was the discovery of a subatomic particle called the “top quark” at Fermi National Accelerator Center. Quarks are produced in collisions of subatomic particles at enormous energies. Unseeable, they are identified by the tracks they leave in particle detectors.

But hunting a quark is like looking at a snowy trail packed with hoof prints to track an elk. What seems to have the clear marks of the quark might be some entirely different species with similar prints.

“All we can do is measure the probability that certain attributes were produced by top quarks and a certain probability that those attributes were produced by more prosaic processes [like a stray cosmic ray, or an ordinary particle],” said William Carrithers, head of one of the teams that found the quark. “The whole discovery is an inherently statistical process.”

The same is true of using DNA profiles to “prove” guilt. When short strings of DNA from a spot of blood or semen found at a crime scene are compared to similar strings of DNA from a suspect, the odds that another individual could share that same profile are usually given as less than one in a million. But in a country of 250 million people, that means 250 people could share the same genetic profile.

On the other hand, sometimes odds are truly so long that improbabilities turn into certainties. For example, the odds of an egg breaking when it is dropped are so great that it always happens.

How good, then, do the odds have to be to consider something to be “true”?

In the case of the top quark, physicists made an initial announcement that they had found evidence for the quark in April, 1994, when there was a 1 in 400 chance they were wrong--about the same chance as pulling two aces off the top of a deck of cards. For the bona fide discovery--like the one announced this February--they waited for a certainty more like 1 in 10,000.

“It also depends on whether there’s a very strong prejudice that you should see something [like a quark],” Stanford physicist Stan Wojcicki said. “Then maybe people would accept 1 in 1000.”

In the case of the top quark, there was indeed an extremely strong prejudice. In fact, if it had not been found, the generally accepted model of how matter is constructed would have been all but destroyed.

The odds against a mistake in DNA matching have to be much longer because a human life, not just a scientific discovery, is at stake. The greater number of DNA segments that match, the lower the probability that any particular match is due to random chance.

Early DNA matching made use of very few of these so-called probes, but improved technology has made it possible to compare more and more segments. Now, reliable DNA tests match five or six stretches of the DNA strand; some test eight or more.

“When people were looking at one or two probes, the weight of evidence turned on statistics,” said Harvard biologist Daniel Hartl, one of the most vocal critics of DNA evidence. But the more stretches of DNA compared, the better the accuracy. “Once you have an eightfold match,” Hartl said, “then any [probability of error] you get is going to be very, very small.”

If the DNA samples have been contaminated, all these results are worthless. But even bad DNA evidence, many biologists and legal experts say, is more reliable than other kinds of evidence that juries hear.

Steven Austad, a biologist at the University of Idaho, cites fingerprints as a kind of evidence normally considered foolproof, but in fact, liable to all kinds of identification errors. Even more dubious is evidence from ballistics, barking dogs and especially eyewitness testimony based on all-too-fallible human memory.

“Any psychologist could tell you that kind of identification is prone to error,” Austad said. “Misidentifications are rife. There’s a large psychological literature on this.”

Even Hartl agrees that DNA probes are at least as good as other kinds of evidence: “I figure one [DNA] probe is worth one eyewitness.”

The relative reliability of DNA evidence has been a great tool in freeing people wrongly convicted on the basis of fingerprints and eyewitness testimony--and later proved innocent by DNA matching.

*

The Innocence Project at the Benjamin Cardozo School of Law in New York has overturned nine convictions over the last 18 months using DNA matching. All were people convicted on eyewitness testimony and circumstantial evidence, but exonerated by their DNA.

Unlike proving a genetic match, proving a mismatch--that is, proving that a DNA sample cannot possibly come from a suspect--is a straightforward matter. If two people share the same genetic profile, then either one could be the perpetrator of a crime. But if a DNA sample taken from the crime scene does not match the suspect’s DNA, then it belongs to someone else, and in many cases the suspect goes free.

“With matches, you get a probability,” said Jonathan Oberman of the Innocence Project. “But exclusions are exclusions.”

The same is true in physics: It is possible to say for certain that a particular track cannot be the top quark, but it’s only very, very, very probable that a particular track is the top quark.

“In our society, people don’t realize that a lot of scientific announcements are probabilistic interpretations of data,” said Nobel laureate Burton Richter, director of the Stanford Linear Accelerator Center. “There are some odds that [the scientific results] are wrong.”