hound evidence produced by Slavin and Hamm, said the critics are misinformed. She said the California handlers are very careful to test their techniques before trying them on real cases.

"They've been working at the forefront of this kind of evidence," said Do. She was skeptical at first but has been won over: "You really have to see it to believe it."

Traditionally, a bloodhound handler uses a "scent article" to start his dog looking for a trail left by the person who had contact with the article. For example, a dog might be given an item of a missing child's clothing.

Then, on command, the dog sniffs around an area for a matching scent. If the dog picks up the child's scent, it trails with nose to the ground until it finds the child or the point where, for instance, the child was pulled into a car.

Instead of using the original scent article, handlers often put a small gauze pad on the item - clothing, facial tissue, a steering wheel - and allow the gauze to absorb the scent before preserving the "scent pad" in a plastic bag.

Pushing the limits

But the Californians have pushed the limits of the bloodhound art, taking scent off shell casings from firearms used in crimes, fragments of exploded bombs - and the decontaminated anthrax letters.

The Californians often use a $895 machine called a Scent Transfer Unit, resembling a small vacuum cleaner, that is designed to draw the scent off the article and deposit it on the pad. One of the machine's inventors is Larry R. Harris, a veteran bloodhound handler who trains with Slavin, Kift and Hamm.

Neither of the two police bloodhound associations has endorsed the Scent Transfer Unit. Officers of the two groups say it offers little advantage over using a gauze pad alone and in fact might confound matters. They contend that an older scent might linger in the machine when it is used on a new case - a charge its users deny.

In addition, bloodhound experts say, the Californians have been quite aggressive in using the dogs not only to follow fugitives or missing persons, but also to identify potential suspects - such as Hatfill - out of a number of people who might have committed the crime.

A false positive

That raises the possibility of a false positive - the identification of an innocent person as the perpetrator, with dire consequences. Some handlers say that there is always a chance that an eager-to-please dog will identify someone even if there is no scent match.

I. Lehr Brisbin Jr., a University of Georgia biologist who has a doctorate in animal behavior, is one of the few scientists who has actually tried to test dog handlers' claims that a bloodhound can accurately pick out a perpetrator from several suspects. He has conducted numerous experiments in which the bloodhound tried to pick from a half-dozen people the one whose scent was on a baseball cap, he said.

No dog was able to do it consistently.

"As a scientist, what they're supposed to have done [in the anthrax case] sounds like a miracle," said Brisbin, a bloodhound handler himself. "Every time I ask a dog to identify a suspect under controlled conditions, the dog can't do it."

Indeed, a federal jury awarded $1.7 million last year to a man wrongly accused of rape after police identified him in part based on the use of Slavin's bloodhound, TinkerBelle. DNA evidence later proved the man, Jeffrey Allen Grant, had not committed the rape.

Hatfill's attorney, Victor M. Glasberg, has suggested that such a mistake might have occurred with the bloodhounds used in the anthrax case, which he has ridiculed as "bionic dogs."

While few details of what the California handlers did in the anthrax case are known, Hatfill said through a spokesman that a bloodhound entered a room where he was sitting and approached him, prompting one agent to call out that the dog was identifying him.

Similarly, William C. Patrick III, another bioterrorism expert, said he and his wife were asked to stand on their lawn in Frederick, and two bloodhounds were led near them.

"They released the dogs, maybe 10 feet away," Patrick said. "My wife and I are dog lovers and we called them, and they walked up and we patted them." Patrick said he was told the dogs had not identified him as the perpetrator.