Does the nose know? Not really.
While law enforcement routinely rely on the alerts of drug dogs as a pretext to search the inside of a driver’s vehicle for illicit drugs, a new study reveals that these animals are alarmingly prone to error.
A team of researchers from the United States and Poland assessed the ability of trained dogs to accurately detect the presence various controlled substances, including marijuana, hashish, amphetamines, cocaine and heroin. Although drug dogs were relatively likely to reliably detect the presence of contraband in controlled, indoor environments (e.g., a room that was already familiar to the dog), their accuracy diminished significantly in situations that would mimic those of a real-world traffic stop.
In scenarios where dogs accessed the perimeter of a motor vehicle, the animals alerted accurately to contraband only 64 percent of the time. Fifteen percent of the time, dogs failed to recognize the presence of illicit drugs. More alarmingly, some 22 percent of the time, the dogs indicated that drugs were present when, in fact, they were not.
Drug dogs’ failure rates were even more pronounced in situations where the animals had access to the inside of a vehicle. In this scenario, dogs correctly responded to the presence of contraband only 58 percent of time. They provided false alerts a whopping 36 percent of time.
In short, in the scenarios involving traffic stop-like situations, trained police dogs frequently alerted their handlers to the presence of "drugs" that were never there.
Nonetheless, in 2005, the United States Supreme Court ruled in Illinois vs. Caballes that an alert from a police dog during a traffic stop provides a constitutional basis for law enforcement to search the interior of the vehicle.
The new study, which appears online ahead of print in the journal Forensic Drug International, is far from the first to question the reliability of drug detection dogs. In 2011, researchers at the University of California at Davis reported that the performance of drug-sniffing dogs is significantly influenced by whether or not their handlers believe illicit substances are present. In other words, if their trainers thought that contraband was on the premises, so did their dogs. "Handlers' beliefs that scent was present potentiated handler identification of detection dog alerts," investigators concluded in the journal Animal Cognition. "This confirms that handler beliefs affect outcomes of scent detection dog deployments."

That same year, a review of Australian government statistics, published in the Sydney Morning Herald, found that some 80 percent of drug dog "alerts" in New South Wales yielded no illicit substances. In the first nine months of 2011, "14,102 searches were conducted after a dog sat next to a person, indicating they might be carrying drugs,” the paper reported. “But, in 11,248 cases, no drugs were found."
An abstract of the latest study, “Efficacy of drug detection by fully-trained police dogs varies by breed, training level, type of drug and search environment,” appears online here.