Raise Your Right Paw ...

The use of canine identification evidence in courts is in the spotlight again, this time in a New Jersey trial:

The presiding judge at the Newark schoolyard slayings trial today has allowed testimony from a dog handler for the Essex County Sheriff's Office whose bloodhound tracked a scent from a cell phone found near the scene to the defendant's home several blocks away. The dog handler, Sheriff's Officer William Caldwell, is expected to take the stand this morning.

The ruling by Superior Court Judge Michael L. Ravin came over the objections of the defense attorney for Rodolfo Godinez, who is the first of six defendants to stand trial for the Aug. 4, 2007, triple killing behind Mount Vernon School.

Is canine identification evidence "junk science," as the defense attorney claims? I'm no forensic scientist, but in studying the history of the German Shepherd Dog for a lecture, I came across a controversy from 1920s Germany on the reliability of the breed's evidence as a scent tracker. Of course, if the dog leads investigators to independent incriminating material like actual stolen property in a suspect's house, the question may be moot. But if other evidence is weak? In an issue of the journal of the original German Shepherd Dog society in the Cleveland Public Library, one of only a few library copies in North America, I discovered that even well-trained dogs were found to be so eager to please their handlers, so sorry to disappoint them, that they faked positive identification.

If (as the bestseller title suggests) dogs never lie about love, the downside is that without any intentional encouragement from their masters, they may do so for love. That's no fault of the dog; dogs aren't scientific instruments any more than human beings are. But it must be a challenge to training, which may reward only "success."

It's worth remembering that a canine false positive was involved in the anthrax investigation, as reported on the journal Nature's website, pointing to an apparent leak of bloodhound identification pointing to Steven Hatfill. As David Freed writes in the current print Atlantic:

Agents conducted a second search five weeks later amid a repeated media circus. This time they came equipped with a warrant and bloodhounds. The dogs, Hatfill would later learn, had been responsible for false arrests in other cases. Hatfill says he innocently petted one of hounds, named Tinkerbell. The dog seemed to like him. "He's identified you from the anthrax letters!" Tinkerbell's handler exclaimed.

"It took every ounce of restraint to stop from laughing," Hatfill recalls. "They said, 'We know you did it. We know you didn't mean to kill anyone.' I said, 'Am I under arrest?' They said no. I walked out, rented a car, and went to see an attorney about suing the hell out of these people."

As the Nature site observes:

Research has produced conflicting claims about bloodhounds' reliability. The most encouraging results suggest that they can correctly identify a suspect out of a line-up around 85% of the time. But one study found that when the person whose scent was taken is omitted from a line-up, dogs choose someone at random almost half the time anyway. . .

[Jerry] Nichols [president of the Law Enforcement Bloodhound Association] and others say that dogs can also respond to unintentional signals from their handlers, who should therefore not be told about where the dogs are expected to react. But as the anthrax investigation drags on, the chances of a thorough double-blind study being done on this or any other technical issues raised seem to be remote.

Unfortunately the adversarial system does not always assure the reliability of expert testimony, defense as well as prosecution, human as well as canine.