"In 2010, a team of researchers at the University of California, Davis set out to test the reliability of drug- and bomb-sniffing dogs.
The team assembled 18 police dogs and their handlers and gave them a routine task: go through a room and sniff out the drugs and explosives.
But there was a twist. The room was clean. No drugs, no explosives.
In order to pass the test, the handlers and their dogs had to go through the room and detect nothing.
But of 144 runs, that happened only 21 times, for a failure rate of 85 percent.
Although drug-sniffing dogs are supposed to find drugs on their own, the researchers concluded that they were influenced by their handlers, and that's what led to such a high failure rate."

http://www.reviewjournal.com/news/crime-courts/legal-challenge-questions-reliability-police-dogs

What do you guys think of this? Is this a training issue, or a handler issue? I'm assume it's mostly a handler issue. How do we go about better training K9 handlers?

Has anyone seen this actual study?

Thoughts and opinions welcome

Training issue and/or lack of trusting the dog's nose.
#1 rule of detection work is "Trust your dog". That can't be done if your dog's training OR your abilities to read the dog are suspect.

The link doesn't provide info on the study. In fact, the news article quoting the study is extremely biased. Precisely why science doesn't rely on a single trial or study to establish a theory as fact. On top of that, the media wants to present to you a single item that supports their position, and have you accept it as absolute truth.

The biggest problem with the report is that it doesn't detail what the experience or employment of the testing pool is. In the body of the article, the person heading the study admits that they had difficulty obtaining testing subjects. Is the reader supposed to infer that these are dogs and handlers that are actually experienced in the field, with real-time indications to their credit? That's too much of a reach for someone actually schooled in scientific method.

Biggest questions: Would the results have changed if there had been some actual finds in the test runs? Was each test subject isolated from the others, to avoid influencing the rest of the testing pool? Was each dog run with it's own working handler, or were multiple dogs run with a common handler that didn't have real-world experience with each individual dog?

If detection dogs were only 15% accurate, that would be reflected in the results that enforcement authorities see. Real world, dog sniffs are are much more accurate than portrayed in this article. Moreover, I wouldn't gamble my freedom on trying to fool a detection dog.

Good post Duane. First off....any "Study" coming out of a Kalifornia college should be suspect. And yes, it is blatantly biased. Imagine that.

That said...handlers are changing the way sniffs are done so as to eliminate as much influence as they can on the dog during the search.

I did a little snooping and was able to find the actual article.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3078300/

Here is the abstract:

"Our aim was to evaluate how human beliefs affect working dog outcomes in an applied environment. We asked whether beliefs of scent detection dog handlers affect team performance and evaluated relative importance of human versus dog influences on handlers’ beliefs. Eighteen drug and/or explosive detection dog/handler teams each completed two sets of four brief search scenarios (conditions). Handlers were falsely told that two conditions contained a paper marking scent location (human influence). Two conditions contained decoy scents (food/toy) to encourage dog interest in a false location (dog influence). Conditions were (1) control; (2) paper marker; (3) decoy scent; and (4) paper marker at decoy scent. No conditions contained drug or explosive scent; any alerting response was incorrect. A repeated measures analysis of variance was used with search condition as the independent variable and number of alerts as the dependent variable. Additional nonparametric tests compared human and dog influence. There were 225 incorrect responses, with no differences in mean responses across conditions. Response patterns differed by condition. There were more correct (no alert responses) searches in conditions without markers. Within marked conditions, handlers reported that dogs alerted more at marked locations than other locations. Handlers’ beliefs that scent was present potentiated handler identification of detection dog alerts. Human more than dog influences affected alert locations. This confirms that handler beliefs affect outcomes of scent detection dog deployments."

And here is the dog/handler methodology:

"A total of 18 handler/detection dog teams, recruited through word-of-mouth from multiple agencies, participated in this study. These teams were certified by a law enforcement agency for either drug detection (n = 13), explosives detection (n = 3), or both drug and explosives detection (n = 2). Demographic details of teams, including dog age, dog breed, dog years of detection experience and handler years of detection experience are presented in Table 1. Upon detection of target scent, all explosives dogs, both drug/explosives dogs and one drug detection dog were trained to issue a passive alert; that is, the dog would sit at the location of target scent detection. One drug detection dog was trained to issue a passive–active alert (sitting and barking), and all remaining drug dogs were trained to issue an active alert (barking) upon detection of target scent. All drug detection teams and two teams trained to find explosives had successfully identified their target scents in law enforcement deployment situations. Additional demographic information collected included handler years of experience handling detection dogs, dog years of scent detection experience, dog age and handler-reported breed of dog. In order to maintain confidentiality, and so that individual teams could not be identified through demographic information, these data were collected anonymously and cannot be linked to any performance data. Due to subject availability, this study was completed across 2 days, with seven teams completing the experiment on the first day, and the remaining 11 teams completing the experiment on the second day."

Thanks, John. There is still one thing nagging at me. In a real life search, if a dog gave a false indication, it would be known immediately, as no object would be found.

I wonder if any of our LE members could tell us how accurate these dogs are when actually working?

Duane wrote..."There is still one thing nagging at me. In a real life search, if a dog gave a false indication, it would be known immediately, as no object would be found"

Not neccessarily....a dealer who just dropped off a pound of marijuana will still have a dog give an alert. The dog is right but the timing is wrong.




Since the Harris case loomed over our heads for almost 2 years we adjusted the way we kept records in anticipation of a bad outcome for law enforcement K9 and began record keeping in a way that should hold up in court better. We went from recording alerts vs. finds to...

Alert w/find
Alert w/o find
Alert with corroboration
No alert

My new partner currently is at 67% actual finds, and of alerts with no find, the percentage of corroberation is around 90%. Only the most tight lipped, and clean user/dealer won't give corroberation. When I search a vehicle after the dog indicates I look for even the tiniest piece of illegal drug to confirm the alert. Anything more is icing on the cake. Primarily I want to confirm my dogs alert which builds confidence in me and my co-workers that the dog is on task.

My first drug dog was on average about 87% accurate on the street over his working life. In training he was about 92-94%. My bomb dog in training was maintained at 95-97% in training over his working life. Not an easy task when he was dual purpose and had to know 15 odors. Tons of work went into him and most of the training wasn't gimme's becase, well....he was a bomb dog and I enjoy breathing.