When I see a police dog inside a train station or at a public gathering, I feel safer. I figure it is there to protect us from explosives, and if it sniffs out drugs along the way, well, that's against the law, too.

But what if it turns out that the dogs aren't all that good at the job the police are giving them? If that's the case, should we think differently about when the police should use dogs to sniff us and our belongings, especially in the privacy of the home?

That's the question in two cases now before the Supreme Court. In the first case, Florida vs. Jardines, the Miami police used a trained detection dog named Franky to check for drugs at the home of Joelis Jardines after getting a Crime Stopper tip. Led onto the porch, Franky sat down by the front door, his sign for alerting his handler that something smelled funny. The police got a search warrant and found a pot-growing business inside the house.

In the second case, Florida vs. Harris, a sheriff's deputy pulled over Clayton Harris because the truck he was driving had an expired license plate. Harris was shaking and breathing fast, so the deputy asked for permission to search his truck. When Harris said no, the deputy brought his trained dog, Aldo, over to the truck, and Aldo alerted him to a smell on one of the door handles. With that as his basis for a search, the deputy looked inside the truck and found ingredients for making methamphetamine.

Both Jardines and Harris challenged the searches as violations of their Fourth Amendment rights. They argue that the police should have gotten a warrant before they deployed Franky and Aldo.

Back in 1983, however, the Supreme Court said that a dog sniff isn't a search under the Fourth Amendment, which means the police don't need a warrant. The idea was that dogs are sui generis as a detection method because the only information they can provide is the presence or absence of drugs. Since nothing else is revealed, there's no privacy at stake. Key to the court's reasoning was "the assumption that sniffing dogs do not err."

That's how Justice David Souter put it in 2005, in dissenting from another pro-dog ruling. Souter said, in response to the argument that dogs can only detect the presence or absence of drugs, that "the infallible dog, however, is a creature of legal fiction." Souter's best evidence was an Illinois study showing that trained dogs tasked with sniffing for drugs came up with false positives between 12 percent and 60 percent of the time. In fact, there'd been hints of the false positive problem earlier: In Indiana in 1979, 26 trained dogs sniffed 2,780 junior high and high school students, running their noses along the kids' legs. They identified 50 students as drug carriers. The kids were strip-searched or told to empty their pockets. Thirty-five of the 50 students were clean.

Another problem for gauging the reliability of canines is the bias of their handlers. In a 2011 study published in Animal Cognition, the sniffing accuracy of 18 trained dogs was tested over two days. The researchers concluded that handlers cue their dogs, deliberately or not, and this affects the animals' accuracy. They hark back to an early-20th-century horse named Clever Hans, who appeared to be capable of counting, but whose real cleverness lay in picking up the nods and bends and smiles of his handler. When Clever Hans couldn't see the person giving him directions, he couldn't count.

Does this matter? It might if you're the Supreme Court deciding whether a dog can sniff the front door of a house without a warrant. In its own brief, the National Police Canine Association reports that in two and a half years of service, Franky sniffed out narcotics in nearly 400 of 656 attempts. We're not told the dog's rate of false positives. Instead the canine association reassures us that when Franky didn't alert his handler to a scent, property owners were none the wiser - their privacy wasn't violated because they never knew their things had been sniffed.

But that, too, is questionable. Dogs cue to chemicals that are found in illegal drugs and also in all kinds of household products. Pickles and glue share an ingredient with heroin. Cocaine shares one with insecticides, perfume and food flavoring. In previous cases, the Supreme Court has ruled that the police must have a warrant to uncover intimate details at home - the temperature of the rooms in a house, for example. Maybe we don't want dogs to come in without a warrant and tell the police what kind of perfume someone wears either.

Emily Bazelon is a senior editor at Slate.