I have always had a gut reaction to dog searches. On the one hand, their senses are fantastic. Nothing better than a good bird dog retrieving game you would never find from the briar patch. Don't mind the bomb-sniffers so much either. And there's no question that dogs are highly effective at finding drugs in cars, suitcases, etc. The skill set is there, the problem is in its unscientific application. I guess what I'm saying is, unless there's some standard of reliability, I don't want anyone's car being torn apart or home invaded based on just a positive drug dog alert. Particularly, as shown in a recently published study in the Journal of Animal Cognition (January 2011), when the dogs may be induced into making false positives by picking up subtle cues from the handler. Turns out the dogs aren't the problem, it's us.
The study by the University of California, Davis, concluded that the rate of false positives (where the dog alerts but there was no scent present) was influenced by the handler's own belief that drugs (the scent) were present. The study used 18 handler-detection dog teams, all from law enforcement. Each team was certified by an agency for detection of drug, bomb, or both. The dogs had between two and seven years experience, while the handlers had as much as 18 years experience. Although the experiment never exposed the dogs to any of the actual target scents they were looking for, there were over 200 false positives, and included every team. The situation that led to the most false positives was when the handler was given cues that the scent would be present, even more so than when decoy scents like sausage were used to fool the dogs.
The study is very readable, and somewhat ingenious. It involved rooms with sausage and tennis balls, so if that's not enough to get you to read it, I don't know what is.