Does the Dog Know What Its Nose Knows?

Detection dogs are useful but fallible investigative assistants

Posted Nov 04, 2012

 On October 31, the U.S. Supreme Court heard oral arguments in two Florida cases involving the reliability and proper use of drug- sniffing dogs.   Florida v. Jardines, No. 11-569, looks at whether using a trained detector dog to sample odors emanating from inside a home violates the 4th Amendment prohibition against unreasonable searches.

Florida v. HarrisNo. 11-817, directly challenges the accuracy of Aldo, a police dog who seems to have imagined the presence of chemicals for manufacturing methamphetamines in a pick-up truck. Absent proof of training and certification, by what measure is Aldo qualified for detection work, and if he is not qualified, how can any evidence he produces be acceptable in court, the defense asks? 

The New York Times ran a good article outlining the 4th Amendment questions and raising the issue that is central to the cases—the accuracy of the detector dogs.  

If the Court follows its precedents and accepts the demonstrably false claim that dogs are biological machines virtually infallible in their ability to detect the presence of an odor they have been trained to recognize, it will have done no more than deepen the gulf between reality and the law.

But a ruling recognizing that, despite their great promise, many detection dogs are poorly trained using flawed and inconsistent measures of proficiency and improperly handled could conceivably force creation of national standards for training, certifying, and deploying detection dogs.  It also might encourage some jurisdictions with inadequate programs to retool or quit the detection- dog business.

 Dogs have wonderful noses.  Rats, ferrets, and pigs—at least where truffles are concerned—do too.  But dogs are more socially acceptable and so they get most of the detection work.  There is a lot of it.  Detection dogs—they are also called “scent dogs,” “sniffer dogs,” and “detector dogs”—are in demand everywhere and employed to detect not only the old standbys (drugs, guns, explosives, and bodies), but also underground gas and water leaks, termites, agricultural pests, counterfeit money, and residue from chemicals that arsonists use to start fires.  Some studies have indicated that dogs can also detect several types of cancer.

But the training and certification of detection dogs is balkanized and inconsistent. Problems with their accuracy and consistency in performing, especially for dogs involved in the corrosive war on drugs, are well known but ignored or rationalized away.

I wrote about some of these problems In Dog’s Best Friend (1997) and in the December 24, 2002 New York Times (a subscription might be required)..  Newer reports have suggested that drug detection might not be proper work for a dog without major reform.

This dog is being trained on a conveyor belt like the ones in use at airports.

A dog is trained to search luggage at the Customs and Border Patrol Canine Enforcement Training Center.

In the estimation of many people who work with them, a well trained dog and handler team can find its target 100 percent of the time in training with no misses and no false alerts.  But a poorly trained dog and less than competent handler may score worse than chance, especially when false alerts are treated for scoring use to score false alerts as misses argue that the dog has detected something that could be very faint residue of the contraband being sought.  More likely, the dog is making a substitution or taking a short cut or focusing on the wrong substance or responding to unconscious cues from its handler.  

It is well known that the handler’s emotional state and expectations are transmitted to the dog through the leash that joins them. Everything from the dog’s health to the conditions in which it is working also affect its performance..  Some years ago, a police drug-sniffing dog in Alabama was found to be alerting to plastic baggies of the sort used in its training—no matter whether they held marijuana or turkey.  In another case, a bomb dog took a liking for duct tape and began to alert every time it encountered a piece.  The dog was taken from service and retrained.  In another case, medical students from the Middle East were wrongly detained after a police bomb-detection dog falsely alerted to the presence of explosive-making material in their car.

 Judging from the few studies that have been done, when false alerts are counted as errors, drug-detection dogs’ accuracy can fall below 20 percent in real world settings, not the training facility. The problem is that errors are not value neutral:  they affect people’s lives and livelihoods.  No matter what the Supreme Court rules this time, these questions of accuracy and reliability must be addressed if detection  dogs are not to see their credibility steadily eroded.  National standards for training and certification should be implemented to avoid that.

In addition, considerably more research is needed into what a dog’s nose knows and the best ways to put that knowledge to work helping people.