Sniffer dogs are a key tool in the arsenal of law enforcement to do their jobs- trained to sniff out everything from drugs to fruit and boasting a reasonably good success rate (with one caveat that we'll get too shortly). As it turns out, though, how these crime-fighting pooches are typically trained is a very simple process anyone can do with a bit of persistence and patience.
Obviously the first thing you need to train a sniffer dog is, well, a dog. While any dog can be trained to sniff for drugs or illegal material, some have a better sense of smell than others. Popular breeds used by law enforcement include members of the retriever family, German shepherds and beagles, all of whom have an impressive sense of smell even by dog standards thanks to generations of selective breeding.
Once a suitable dog is selected, training it to sniff out the desired substance is a simple matter of getting the dog to associate a favourite toy with its smell. One popular toy chosen by sniffer dog trainers is a small white towel, which is easily recognised by the dog, can be cleaned to ensure the towel's scent is minimized, and in the general case dogs love a good game of tug of war with said towels.
After the dog learns to associate the toy with getting to have fun, the trainers begin to introduce the smell of the desired illicit substance they�d like the dog to eventually seek out, be it drugs, chemicals used to make bombs, money, or really whatever. In fact, there are even so-called "porn sniffing" dogs that are in reality trained to sniff out hidden memory cards, hard drives, and other such devices. These dogs are used when investigating hackers or those suspected of being in possession of underage pornography or the like. One such dog, named Bear, was used to help convict Subway's former spokesman Jared Fogle, when the dog found just such a hidden flash drive at Fogle's residence that human searchers had missed.
In a nutshell, if the item has a scent, the dogs can be trained to sniff it out.
As you might have gathered from this, over time, the dog will learn to seek out whatever smell, not for the smell's sake (as has been often erroneously suggested by those insinuating the dogs are addicted to the drugs), but rather because they associate finding the source of the smell with their favorite toy and play time.
And if you're wondering here, yes, the drug sniffing dogs are technically trained via using actual drugs, though the amount needed here for the dogs to pick up the scent is incredibly low thanks to their exceptional sense of smell. This ensures no risk to the dog themselves, at least in training. The same cannot always be said to be the case when the dog is on duty.
While rare owing to the fact that the dogs are being used to find hidden drugs (thus likely barriers between the drug and the dog), it does sometimes happen where the dog may be exposed to something so toxic that even trace amounts pose a risk to them, such as relatively recently happened in Broward County, Florida where sniffer dogs were exposed to the extremely potent opioid fentanyl.
To get around this problem, the dog's handler will keep a vigilant watch and particularly not deploy sniffing dogs at all if there are any loose drugs present. For the unseen drugs that nonetheless may be particularly potent, the handler also might carry things like naloxone with them, which rapidly reverses/blocks the effects of opioids, just in case.
Back to training- once the dog has learned to successfully seek out the smell when commanded, the trainer then moves on to only giving the play reward when the dog does the appropriate action when it finds the source. For example, some drug sniffer dogs are trained to paw at the spot where the illegal substance is located, but this, obviously, isn�t ideal if the dog is searching for something like a bomb. As a result, sniffer dogs are trained to have �passive� and �aggressive� responses based on the situation.
The passive response can, again, be whatever you like, though most sniffer dogs are trained to simply sit down and/or point with their nose and freeze in these cases.� As one security company who trains sniffer dogs summarizes:
"During the early stages of training, the sniffer dog will receive a reward when it displays any form of recognition of the target scent. As the dog�s skills progress, the reward will only be given when the dog responds with the correct reaction (e.g. sit, stand, stare, down, bark, etc�)."
With these quite simple training methods, the resulting sniffer dogs are very accurate in finding concealed illegal items and generally are trained well enough to be immune from being fooled by strategic placement of treats or other masking scents like meats.
This all said, in real world use, there does seem to be a major issue given dogs have also long been bred to please their masters and are exceptionally good at picking up body language from said masters.� This is essentially another example of the Clever Hans Effect. If you're unfamiliar, this was a phenomenon famously observed in a horse named Hans who genuinely seemed to be able to do math, including fractions, multiplication, and even square roots, as well as eventually learned to spell out names, among other remarkable skills.
In the beginning this was widely assumed to be a hoax, but after an independent commission confirmed the horse really could do these things, it seemed as if mankind had been massively underestimating the intelligence of horses. That is, until psychologist Oskar Pfungst ran a series of experiments that uncovered the fact that if the horse couldn't see the questioner or if the questioner didn't know the answer, the horse universally got the answer wrong. When the horse could see the questioner and the questioner knew the answer, however, the horse had remarkable accuracy in the 80%-90% range in many tests.
Thus, what was actually happening was the horse was simply reading very subtle body language of the person conducting the experiments (none of whom were aware they were exhibiting any such cues) and so was able to pick the correct answer to whatever question based on this. (You can read more on this here: The Horse That Could Do Math)
As you might imagine, sniffer dogs suffer from this same phenomenon. In support of this notion is a double-blind 2011 study published in the journal Animal Cognition in which the dogs were tested in a scenario where there were no drugs, but the dog handlers and those observing the handlers were told there were drugs. To further attempt to trick the dogs themselves, certain food/play items were also placed about to entice them.
The results? The dogs performed well with regards to ignoring the treats, but were nonetheless uncharacteristically wildly inaccurate anyway, particularly around areas where the humans were led to believe drugs were located. As there were no drugs, the result was simply a whopping 225 false positives by the dogs and the 18 handlers. In fact, in only 15% of the total 144 runs did a dog and its handler successfully go through the course without indicating they'd found drugs.
Given the number of studies showing both that drug sniffing dogs are very difficult to fool in controlled tests and other studies like this one showing they nonetheless become quite inaccurate when the handlers are fooled, it's generally thought it's really the humans that are the problem here, not the dogs.
On this note, there have been lawsuits in the United States alleging that some handlers are actually training the dogs to respond to intentional subtle cues they give so that the police can then conduct an otherwise illegal search any time they please.
In support of this notion, despite in testing the dogs being extremely accurate, one report from the Chicago Tribune noted that over a three year period the dogs used during traffic stops in the suburbs around Chicago reported false positives a whopping 56% of the time. More disturbingly, that false positive rate rose to 73% when the police had pulled over someone of Hispanic descent. (Although it should also be noted that despite a three year span, the sample-set the Tribune was working with of actual traffic stops where sniffer dogs were deployed was quite small and the data they used was also incomplete.)
That said, during the previously mentioned study published in Animal Cognition, the experimenter "was informed that three handlers admitted to overtly cuing their dogs to alert" where the handler thought there was drugs, implying that this is something at least some handlers have trained their dogs to do when the handler thinks drugs are present.
Nevertheless, while some handlers seem to be doing this, it's generally thought that in many such false positive cases what's actually happening is either trace amounts of drugs from past drug presence (potentially even from years previous) or, perhaps even more likely, simply another instance of the aforementioned Clever Hans Effect, in that the handlers are unintentionally signaling the dogs that they think there may be drugs or the like in, say, a vehicle. The dogs then pick up on this, and so even though the dogs may not yet be smelling drugs, they react as if they do, resulting in a disproportionate number of false positives in these cases, and potentially a bit of racial profiling thanks to the humans' biases, rather than the dogs not being good at their job.
It should be noted here that it wouldn't be that difficult to train the dogs to ignore handler body language when searching. The dogs can also be mostly trained to ignore ultra-low thresholds of detection from drugs present in the distant past. Despite this, to date, this doesn't appear to be something implemented in any sniffer dog training program we could find.
But in the end, the important thing to note in all of this is that sniffer dogs generally see their work as a game and usually work hand in hand with their best friend, sometimes for years with the same handler. And regardless of how accurate they are in any given scenario, each and every one of them is a very, very good dog.
Cognoscente of the profiling problem and potential for rights infringing searches where no such search is warranted by what the dogs are actually smelling, a new tack has been investigated by some agencies in the form of switching to using mice. While less fun to have around than man's best friend, mice have the potential to be more accurate than drug or bomb sniffing dogs for sniffing things out. Another advantage of the mice, according to one security company that trains mice for such, is that they don't need to continually be motivated when not finding anything. In contrast, sniffing dogs must be periodically sent on dummy missions so they don't become bored with the game. For example, in airports, sniffer dog handlers sometimes have to ask employees to carry dummy packages loaded with drugs or some other illegal substance through security so that the dog can �find� them and be rewarded.The dogs must also periodically re-trained to keep their skills sharp. Mice, on the other hand, are good for about four hour shifts at a time with a useful lifespan of about 18 months, all supposedly without need to re-train or re-motivate. Further, the mice are significantly cheaper and easier to maintain. They also can be trained much faster and cheaper than dogs, and in extremely high numbers for mass and quick deployment.
As an example of how the mice might be deployed, the aforementioned security company notes small containers with the hidden mice can be strategically placed at airports in locations where passengers walk by or are made to pause for a moment. When the hidden mice (possibly large groups of them) smell whatever they've been trained to alert for, they press a button alerting the authorities, all in a much less visible fashion than dogs, and not prone to reacting to handler bias or, in this case, even have a handler visible to the mice at all.