What the Dog Knows (31 page)

Just a year and a half after
Time
magazine informed dogs that they were working on borrowed time, I started Solo's serious training with Nancy Hook. I was able to read and understand his behavior changes. I knew what happened to scent in heat, in wind, in rain, and water. I was Dr. Dog Science.

That same summer, Glen Rains at the University of Georgia and his colleagues filed their patent for “The Wasp Hound,” an eight-inch portable tube filled with starved parasitic wasps, a fan, a video camera, and a computer. When the target scent wafted into the chamber—whether it was the scent of human remains or of a bomb—the trained wasps would congregate at a pinhole, hoping for their food reward.
The video camera watched the wasps and signaled their behavior change to a computer. The patent application called dogs “subjective and costly.” Wasps were “more sensitive, programmable, portable, and cryptic.” That last adjective was critical in a post-9/11 world. On that point alone, the small, silent wasps in the tube triumphed. From the day he was born, Solo was never, ever cryptic.

At least I could understand the concept of the Wasp Hound as I looked at the application's illustrations. I couldn't say the same about the dog-on-a-chip, although its inventors called it “an elegant fusion of biotechnology and microelectronics.” However, Solo wasn't any more elegant than he was cryptic. He did have more in common with some components of the Wasp Hound than with many of the bioengineered noses. He and the wasps might not be same species, but they both belonged to the kingdom of Animalia.

•  •  •

Mechanical noses, mixtures of mechanisms and organisms, or their newest iteration, genetically engineered cells, have been a Holy Grail for applied scientists. Some researchers continue debating the physics of scent while others labor to understand what dogs are alerting on. A third group has dismissed the dog nose altogether and is working on its replacement. Though the dog's nose might be a black box and a real challenge to reverse-engineer, no scientific law says you can't try to build a better black box before you understand what's inside the original.

The first artificial nose became available in 1982, the same year
USA Today
was launched. The “Warwick nose” had a sensor made of tin oxide and is still produced for commercial uses. In 1988,
Computer Business Review
noted that drug kingpins could use masking odors, “leaving poor pooches puzzled,” but the Warwick nose could do better. The only thing needed, the
Review
urged, was more development money; otherwise, Japan might win the electronic-nose race.

We've lived with the term “electronic nose” for nearly three decades, and despite some very different mash-ups using gas chromatography, grids of nanoparticles, or polymer sensors, any artificial nose needs to do three things: take in a vapor, show it on an array, and identify it to the machine's operator. The dog's nose does something similar: It takes in a smell and sends a signal to the dog's brain. The dog sits: I smell dope. The handler sees the dog sitting and knows she's just smelled dope.

One cannot patent the dog. So “a truly biomimetic olfactory microsystem”—that is, a fake nose—became a goal for bioengineers. Despite its lack of a mammalian reproductive system, the fake nose is a gift that keeps on giving. In early theoretical stages, it helps get papers published. In experimental stages, it helps buy lab space and equipment and pay for postdoctoral lab workers. If the nose ever goes into production, even more people get involved. Most of the current fake noses are expensive but not too expensive, especially in these days of downsizing public infrastructure. Law enforcement, the military, health care, and other industries, such as food production facilities, will buy it. Developers admit that the fake nose will require some maintenance and training. That admission just brings in additional funding.

As each new nose is trotted out to the media and public, the same claims are made: This nose is the nose to end all noses. It will keep working in heat and cold, never false alert, never get tired, never require dog chow. It won't have to retire at the age of eight or nine with hip dysplasia.

Mechanical noses have flourished under scientific competition. But like a bad reality show,
So You Think You Can Make a Better Nose
, the competition can be a polyphony, or even a cacophony, of competing tunes and dueling harmonies: Anything a dog can do, a machine or a mash-up can do better. Your fake nose is good, my fake nose is better. Anything a Georgia Tech chemical engineer can do, an MIT researcher can do better. “There's no further improvement in the sensor part you
can get,” an MIT chemical engineer told
Wired
magazine about his e-nose creation in 2010. “It's the last word in sensors.”

It may be the latest word in sensors. I doubt it's the last. Call me unsentimental, but I can't decide which I prefer: having an airport security employee thrust an electronic nose into my crotch, or having a TSA handler do the same thing with a bomb dog. If my civil rights are going to be violated every time I fly, I do prefer that the search be effective and low on the false positives. And that I not get cancer or get bitten.

•  •  •

A couple of decades before the term “electronic nose” came into being, army researchers, much like those at Southwest Research Institute, started thinking about replacing the dog, a sometimes unwieldy biological system, with a better system.

The experience of using sentry dogs in Vietnam in the early 1960s made for some imaginative leaps of faith, and while some army researchers, like Nick Montanarelli, stuck with the tried and true and played around the edges by trying different breeds of dogs for scent detection, other researchers at the U.S. Army Limited War Laboratory went further. Dogs' noses were great for detecting the enemy, and dogs' barks were great for arousing sleeping soldiers. What about partly mechanizing and miniaturizing that system?

So in 1965, army researchers started playing with “insect ambush detectors,” a reasonable riff on the reality that blood-loving insects were ideal candidates for the “bio” part of the machine: ticks, mosquitoes, bed bugs, giant conenose bugs. All those insects use the warm breath of vertebrates to find their next meal. Researchers rigged up a plastic tube with a bellows, a microphone, and a sanded piano wire that would twang when the bugs' feet started a frenetic dance because they sensed a nearby meal. Ta. Da. An “ambush detector,” ready for a wild rumpus when it smelled the breath of enemy soldiers.

Early work narrowed down the best bugs for the job. When fleas sensed humans, they started jumping violently. They wouldn't settle down. They had to be fed too often. Bed bugs, like fleas, got overly excited about the prospect of a hearty meal. Ticks were an early possibility. They loved the smell of human breath and could move quickly, though they didn't jump. I've watched them crawling up my arm, so I know. They have soft feet. That's why you think you can feel something crawling on you, but it's just a vague sense, and you forget about it. The next time you think about the tick is when someone tells you one is attached to the back of your neck. This is advantageous for ticks but not for army researchers, even with a microphone inside the detector. The researchers tried tying weights to their tiny tick feet, but they still were too silent, even wearing clodhoppers.

Conenose bugs—commonly known as kissing bugs because they most love the fleshy part of your face or lips for their bloodsucking—were just right. Until the researchers tested them. The bug machine performed miserably in the 1966 Panama Canal field test, the army report noted. The false positives were through the roof. Once the conenose bugs got amped up, they were as bad as the bed bugs: They refused to settle down. Their feet just kept noisily sending out signals. Food! Food! Researchers realized it wasn't just food. Motion excited the bugs. Wind excited them. Pretty much everything excited them. It was like a kids' sleepover.

In what would become a template excuse for not finding something as good as the dog's nose, the final report said that using blood-loving bugs as ambush detectors remained “technically feasible.” That term is a big ol' hairy clue to recognize a failed experiment.

In any case, bugs are passé. In the past five years, the entities that have truly turned on tech reporters aren't the hybridized machines with animal names, like the Wasp Hound, but plants. Bomb-detecting ferns, for example, created a minor media stampede. “They turn completely white when they sense something nefarious around them,” a FOX News reporter said. Since we are used to seeing big leafy plants
in malls and in fern bars, I worry that we'll hardly notice them turning from a robust green to a sickly white if someone plants a bomb.

“We actually modify the seed,” University of Colorado biologist June Medford told FOX News, “and then it's a trait that is stable and stays with it forever. It's very empowering because it will tell you that there's an explosive around: ‘Get the security guys here!' ”

Sounds great. So how long does it take that fern to turn white and tell the security guys? While Medford says that her studies show the plants have “detection abilities similar to or better than those of dogs,” it takes the ferns hours to change color. “Work is under way to reduce that to a few minutes,” the report promised.

We don't yet have bioengineered green dogs who turn white when they smell a bomb. The old-fashioned ones are pretty fast at telling us something is amiss. Faster than hours. Or a few minutes. More like a second. When there's a bomb, speed can matter.

That's the problem with the vast majority of the dog replacements, from fern to machine to all the things in between: They don't have the skills that dogs have in one furry package. Researchers realized that in the 1970s when they tried a variety of species to replace dogs; they keep learning that lesson over and over. Dogs are adaptable, mobile, sensitive. They use complex cognitive judgments to avoid lots of false positives. They can do several things at once: sniff, raise an alarm, bite if necessary, act as a deterrent. They interact with people more than a fern does. They're more fun to have around than angry bees. I know this, having been stung a few times tending to our beehive.

Best of all, dogs are comparatively inexpensive. The argument that it takes money and time to train a dog is countered by the argument that training technicians on machines can be just as expensive. Machines don't run themselves any more than dogs do. Machines break down, they need calibrating constantly, and they can be more temperamental about weather than dogs.

“When I started doing work in this area twenty years ago, I originally
thought we would be able to make a machine that could replicate a dog,” said analytical chemist Ken Furton. “But it's not going to happen in my lifetime. We are not going to replicate what a dog can do.”

The Pentagon came to the same conclusion in late 2010 in Afghanistan and Iraq. It shut down a huge program that had spent $17 billion in employees and technologies and failed to make a dent in the problem of IEDs. After five years, hundreds of projects, and a “blizzard of cash” paid to the country's biggest defense contractors, reported the Center for Public Integrity, one system rose above the others: dogs and their handlers, along with observant people: “The most effective IED detectors today . . . don't hum, whir, shoot, scan, or fly. They talk. And they bark,” Peter Cary and Nancy Youssef wrote.

The rate of finding IEDs with other technologies stood “stubbornly” at just 50 percent, Lieutenant General Michael Oates told
National Defense
magazine. Handlers and dogs found IEDs at the rate of 80 percent. The best bomb detectors, Oates said, are dogs working with handlers, local informants, and the trained soldier's eye.

“That combo presents the best detection system we currently have,” Oates said.

A couple of years ago, I happened to be standing next to one of the best detection systems we currently have. I was at a K9 training near a military base in North Carolina. A special forces handler and his military working dog were taking a break from classified work in the Middle East. The Belgian Malinois was lying quietly beside his handler with his back legs tucked under his haunches and his front claws dug slightly into the ground. His eyes weren't fixed on his handler; instead, he was looking out. The dog's eyes never stopped scanning the big field of mowed grass in front of them. The handler looked down at his dog. It was, he said, like having a formidable extension, two and sometimes three hundred feet in front of you, keeping track, watching out for you, making sure you weren't attacked. Or blown up. The handler said he couldn't count the number of times his dog had saved his life.

•  •  •

Not only can turkey vultures detect a dead mouse from more than 1,000 meters (3,300 feet) away, but they have a major advantage over sniffer dogs—they can fly, removing the challenge of difficult terrain.

—
Der Spiegel
, 2011

Bomb detection isn't the only arena where machines and their mash-ups don't appear to be making much progress. The body-location business isn't seeing a great deal of electronic nose success, either. One group of German forensic scientists said in their 2010 conclusion about using machines to locate human bodies (though they used dead rabbits in their experiment): “In principle, an electronic nose based on the sensors applied in this study can be used to find decomposing human bodies in terrain. However, for design and development of a practically applicable device, sampling and measurement procedures have to be optimized.”

In principle, with optimized hooves, pigs can fly.

Between the Germans with their rabbits and the Georgians with their wasps, I was starting to feel better about the dog's place in the body-location universe. Good dogs seem to move through a kind of complex decision tree on difficult searches: “This, not that,” “Up, not down,” and “That thing doesn't belong here, but it's not the thing I'm looking for.” Although dogs aren't perfect, they adapt to a variety of search conditions.