Wednesday, November 04, 2009

Phony dog evidence

The NY Times reports:
HOUSTON — A dog’s sniff helped put Curvis Bickham in jail for eight months. Now that the case against him has been dropped, he wants to tell the world that the investigative technique that justified his arrest smells to high heaven.

The police told Mr. Bickham they had tied him to a triple homicide through a dog-scent lineup, in which dogs choose a suspect’s smell out of a group. The dogs are exposed to the scent from items found at crime scene, and are then walked by a series of containers with samples swabbed from a suspect and from others not involved in the crime. If the dog finds a can with a matching scent, it signals — stiffening, barking or giving some other alert its handler recognizes. ...

Mr. Myers, the animal behavior expert, suggested that handlers like Deputy Pikett might believe in the dogs and the methods, but might allow samples to become contaminated or inadvertently allow the dogs to pick up on subtle, even unconscious signals from handlers or detectives.

“They just don’t realize they’re doing it wrong,” he said.
No, it should even obvious to a dumb judge that this is unscientific.

No test is 100% foolproof. For this test to be scientific, someone has to do an objective measure of the reliability of the test. The judge just has to ask, "What is the error rate? Show me how you measured the error rate?"

This dog sniff expert has probably never measured the error rate. He just has a gut feeling that it works. Then it is not scientific, and should not be admissable in court.

No comments: