“Andrew, do you want to do this?”
“Hell, yeah!”
“Not to rain on the puppy parade,” Lisa said, “but what is the scientific question?”
There are two types of experiments in science: fishing expeditions, where you start collecting data without a clear idea of what the right questions are, and hypothesis-driven experiments, where you start with a specific question to answer. Every middle school student would recognize the latter type as the foundation of the scientific method. Most people think that hypothesis-driven experiments are the only way scientific progress occurs. And science journals strongly prefer hypothesis-driven experiments.
The recipe for the typical hypothesis-driven experiment is simple: Take a well-accepted scientific theory. Find some minuscule aspect of that theory that nobody has ever verified before. Do an experiment that proves that aspect and supports the theory as a whole. Publish.
These experiments make for easy reading and are a surefire way to get results published, building up a résumé that will ensure promotion and tenure at a university. These types of experiments are also popular with funding agencies because the risk of failure is minimal. By my estimate, nearly all published research falls into this arena.
The thing is, hypothesis-driven experiments are incredibly dull. Most of the time you don’t even need to read the experiment to know that the scientists have proven what they basically knew in the first place. If you already have a well-accepted hypothesis, then you already know the most interesting aspects of the scientific question, and the experimental results will, at best, advance knowledge incrementally. Of course, if the hypothesis turns out to be wrong, that would be really interesting. But those results are almost impossible to publish because nobody believes them.
In answer to Lisa’s question, I said, “This is a fishing expedition. It is an idea in search of a question.”
Andrew frowned, clearly troubled by the conflict this would cause with his dissertation research. The standard curriculum of any graduate program in science drills into students the importance of having a clear hypothesis for their research. But I had no hypothesis for the Dog Project. I had no idea how we were going to do this or how long it would take. Frankly, it probably wouldn’t even work.
“Andrew,” I said, “the Dog Project will be high risk. But it’s going to be a blast, and I guarantee you that if it works, we’ll be the first to have pulled it off.”
“I’m in,” he said. “But are we going to have to sedate the dogs?”
“Why would we do that? If they’re sedated, then we won’t know what they’re thinking.”
“So they’ll be completely awake?” Lisa asked.
“They’ll have to be,” I replied. “Just like humans.”
At the time, none of us realized just how much work lay ahead. We didn’t know what the technical difficulties might be, considering dog brains are much smaller than human ones. We hadn’t even begun to think about the actual experiments we might attempt.
At that point, it was all academic. Before we could go any further, we would have to figure out how to train a dog to go inside an MRI.
4
Puppy Steps
EVEN THOUGH CALLIE HAD BEEN in the house for a year, I had not completely warmed up to her.
I wasn’t even sure that I liked her.
Kat knew how much I had loved Newton. When she and the girls went to the animal shelter, they had deliberately picked a dog that was about as different from a pug as you could get. Callie was the anti-pug. Pugs are short, stocky, and slow. Callie was a lean, mean fighting machine. Her muscles rippled beneath her thin coat.
Where Newton’s face had been fixed in a permanent clownlike expression, Callie’s was always on high alert. Her head was like a periscope, constantly swiveling back and forth in search of prey. Though she was quite friendly, her posture was off-putting to many of the dogs in the neighborhood.
Callie’s strong drive caused endless distress to Helen and Maddy. Whenever Callie killed a chipmunk, the girls would berate her for her cruelty. To make matters worse, Callie wasn’t cuddly. She didn’t like to sit in laps. Sure, she would readily hop on the sofa, but then she would curl up like a cat at the other end—nearby, but not quite touching.
I missed my bedtime ritual with Newton. He would burrow under the covers, seeking refuge in my armpit, and I would pretend to protest. Although Callie wanted to sleep in the bed, her state of alertness never switched off. She would assume a position at the foot of the bed, facing the door, on watch for potential intruders or edible critters. Any attempt to move her unleashed a snarling, snapping bundle of fur. She wanted nothing to do with my armpit.
There was a dog-training facility in a strip mall within walking distance from our house. It was called Comprehensive Pet Therapy—CPT for short. Shortly after Kat adopted her, we signed Callie up for a basic obedience class.
CPT was the brainchild of Mark Spivak, who founded it in 1992. I first met Mark when we signed Lyra up for obedience training in 2005. Mark was not your typical dog trainer. He graduated from the University of Pennsylvania with a degree in economics and then received his MBA from the University of California, Berkeley. Mark bounced around the semiconductor industry in the Bay Area for a while but never meshed well with management. After he moved to Atlanta, he and his German shepherd, Topper, started competing in agility competitions to relieve some of his work stress. They did well, and Mark began helping friends with dog-training problems on the side. Within a few years, he decided to take the plunge and go into the dog-training business full-time.
Mark was a no-nonsense kind of guy. He employed several schools of thought about dog training, choosing the methods most appropriate for each dog and owner. And while he favored positive training methods, he acknowledged that punishment was also necessary from time to time.
Even though I hadn’t yet bonded emotionally with Callie, I did enjoy working with her in Mark’s obedience class. Lyra had taken this class too, but she had never had the level of intensity that Callie brought to the table. Callie wasn’t warm and cuddly, but I had to respect her work ethic. She couldn’t get enough training. She would do anything for a bit of hot dog. I was amazed that she learned basic commands like “sit,” “stay,” and “come” in just a few tries. The CPT teachers loved to use Callie as an example, because she watched them intently and worked tirelessly for a treat.
As Mark was the only dog trainer I knew, it made sense to approach him about the idea of training dogs to go into an MRI. He took an almost academic approach to dog training, so I hoped he would find the idea of scanning dogs’ brains interesting enough to do for fun.
Much to my delight, Mark agreed to meet.
The modern study of dog behavior began with every biologist’s hero, Charles Darwin. In The Expression of the Emotions in Man and Animal, Darwin devoted a great deal of attention to the dog—as an owner himself, his study of dog behavior didn’t require a trip to the Galapagos Islands. What Darwin understood, and what every dog owner knows—but many research scientists seem to have forgotten—is that dogs have a rich set of expressions and body language. Darwin had no problem discerning joy, fear, and rage in dogs. He was primarily concerned with observing the expression of these emotions, not with the intent of training these intelligent animals, but rather to understand how human emotions evolved.
It was the famous Russian physiologist Ivan Pavlov who launched the modern era of dog training. Unlike Darwin, Pavlov had no love for dogs himself. He was just using them to study the digestive system. The problem was that his dogs started salivating before he fed them, and this messed up his data. Regardless of what you think about Pavlov, his “failed” experiment led to the most important discovery in psychology of the twentieth century, for which he was awarded the Nobel Prize in 1904. His discovery has completely dominated theories of dog training ever since.
Pavlov’s discovery is called classical conditioning (although some people honor him by call
ing it Pavlovian conditioning). During the period in which Pavlov was doing his experiments, physiologists thought of the entire nervous system as a collection of reflexes, like the involuntary leg jerk when a doctor raps on your knee. They believed that all behaviors, even complex ones, were basically a series of reflexive actions. A reflex could be broken down into two parts: the unconditioned stimulus (US) and the unconditioned response (UR). For the knee reflex, the US is the hammer hitting the patellar tendon and the UR is the quadriceps contraction that results in the leg jerking upward. Pretty simple.
Pavlov realized that his dogs were having reflexive responses, but they weren’t natural. Hungry dogs will always salivate when presented with food. This is a natural, and thus unconditioned, response. But, as Pavlov discovered, if something neutral, like the ringing of a bell, regularly precedes the presentation of the food, the dog will start salivating at the sound of the bell. The bell, a neutral stimulus, becomes a conditioned stimulus (CS), and the salivation it evokes is now a conditioned response (CR). The terminology of unconditioned and conditioned refers to stimuli and responses that are either natural or created by the experimenter.
By itself, classical conditioning doesn’t say much about dog training. The responses are so simple that they don’t constitute anything remotely resembling a behavior, and it is hard to imagine cobbling together a string of these conditioned responses into something as simple as “sit.” This is where instrumental learning comes in.
In instrumental learning the animal must do a purposeful behavior. While classical conditioning trains an involuntary response like salivation, instrumental learning aims to train a voluntary action. Instrumental learning forms the basis of every dog-training method ever published. Teaching the “sit” command is based on instrumental learning. Here, the stimulus is either a hand signal or a spoken word, and the desired behavior is the act of sitting. When the dog sits and he is immediately rewarded, he makes an association between the act and the reward. In instrumental learning, the link between stimulus (“sit”) and act (sitting) is called the stimulus-response (S-R) relationship. Instrumental learning is also called operant conditioning because the animal learns to operate on, or affect, the environment.
Psychologists have classified four different types of instrumental learning based on whether a behavior is rewarded or punished. A reward is something that the animal likes, such as food or praise. Punishment is something he doesn’t like, such as a loud noise. Rewards and punishments can be either given or withheld, which leads to the four types of learning. For example, the removal of something unpleasant reinforces behavior, so we call it negative reinforcement, negative meaning “removal.” Positive reinforcement comes from the delivery of a reward, while positive punishment comes from the delivery of something unpleasant. The final combination, negative punishment, occurs when you take something desirable away from the animal. Negative punishment is a popular tactic among parents trying to curb undesirable behavior in their children. The suspension of computer privileges is a classic negative punishment and should, according to theory, decrease the frequency of the offending behavior.
The use of instrumental learning to change behavior is broadly referred to as behaviorism. Psychologist Edward Thorndike described many of its basic laws. The law of effect states that S-R relationships are determined by how much the animal likes the reward. The more he likes it, the stronger the S-R link. Thorndike’s law of exercise states that an S-R relationship is strengthened through use and weakened through disuse. Thorndike’s laws were further elaborated by the legendary psychologist B. F. Skinner, who thought that all behavior could be reduced to a set of S-R relationships. He is most famously associated with the Skinner box, a device that automatically trains rats or pigeons to learn behaviors.
After Pavlov’s basic discovery and Thorndike’s and Skinner’s elaborations on it, behaviorism flourished. It reached its peak in popularity in the 1960s, when psychologists and psychiatrists began applying these theories of animal learning to human behavior. Techniques that targeted everything from smoking cessation to learning to make friends were all rooted in the behaviorist tradition. While some of its prominence has waned in recent years, behaviorist techniques remain the most commonly used “talk therapies” for depression and anxiety in humans, which is called cognitive-behavioral therapy (CBT).
When it comes to dogs, much has been said and written about positive and negative training methods. While they are all based in the behaviorist tradition, different schools of thought place different emphases on rewards like food and praise and punishments like noises, scolding, or pain. There is no doubt that the administration of a punishment can cause an immediate effect on a dog’s behavior. What is unclear is whether the dog actually learns anything from it. The child who has lost her TV privileges may have learned not to repeat her offense, or she may simply have learned not to get caught.
This is the limitation of behaviorism: one can never truly know why a person or animal does something. You can only observe the effect of a reward or punishment and whether it increases or decreases a particular behavior. In fact, hard-core behaviorists completely dismiss what goes on in an animal’s head. Since behavior is the only thing that matters to a behaviorist, subjective thoughts and emotions become irrelevant. But if you have tried to curb a dog from a particular bad behavior—chewing furniture or shoes, for example—you know the frustration of trying to understand why none of the punishments are working. How many dog owners have cried out in vain, “Why are you doing that?”
I hoped the Dog Project would someday be able to answer that question.
Until that day, Mark and I would need to figure out a training protocol based on conventional behaviorist methods that would get a dog to willingly climb in an MRI machine.
I met Mark at CPT. The training facilities are basically a large room. The linoleum flooring makes for easy cleanup of the inevitable “accidents.” Apart from a teeter-totter and some ramps and hoops for agility training, the room is devoid of furniture. The spartan decor minimizes dog-induced damage expenses.
Mark was wearing his standard attire for dog training: a polo emblazoned with the CPT logo, athletic shorts, and running shoes. I had seen him only in dog-training mode, so I was surprised when he greeted me with such enthusiasm for the Dog Project.
From the beginning, we agreed training should be done strictly with positive reinforcement. It wouldn’t be right to use punishment to teach a behavior this strange that would not directly benefit either the dogs or their owners. Everything in the Dog Project should be fun. Fun for the dogs, and fun for the owners. Mark suggested that this would be much easier if we could utilize the dogs’ natural behaviors.
Natural behaviors are ones that dogs do on their own. Walking, sitting, and lying down are natural behaviors. If the dog has a drive to hunt small animals, then tracking might be considered a natural behavior too. Retrievers were originally bred to retrieve ducks, so they have a natural drive to carry objects in their mouths and, at least in theory, return them to their handlers. For some dogs, swimming is a natural behavior. For others, water is to be avoided at all costs.
It is safe to say that going into an MRI is not a natural dog behavior. Most humans don’t like it either. But Mark explained how we could train a sequence of behaviors that were mostly natural for the dog.
“Most of what the dog has to do is a ‘down-stay’ position, correct?”
In a “down-stay,” the dog lies down and stays in that position while the handler remains some distance away.
“Yes,” I replied.
“Lying down is a natural behavior, so that is easy to teach with positive reinforcement. What else does the dog need to do?”
“He needs to hold his head perfectly still,” I said.
“How still?”
“Less than two millimeters of movement for periods up to twenty seconds.”
Everything depended on the head being still. Any movement would render the MRI
data useless. When we perform scans on humans, the subject lies on her back with her head surrounded by foam pads. Most people are able to remain still, and the foam makes it easier. But a dog might not like his head being encased in foam. Maybe something less intrusive would suffice.
“We could make a chin rest for the dog,” I suggested.
Mark liked this idea. “When we train dogs for tracking, we will often teach them a ‘touch’ command where they touch their nose to a target. We could do the same thing to teach a dog to ‘touch’ a chin rest.”
Dogs use their nose to touch and sniff everything. This was a brilliant example of taking a natural behavior and turning it into a trained one. That left only the noise. MRIs are as loud as a jackhammer.
Mark stressed the importance of subject selection. He said, “We will need to carefully select the first subjects for the right temperament characteristics.” With the right subjects, the training would be easy. We certainly didn’t want a situation in which the dog didn’t want to be there. Even if we could train the dog to stay in the MRI, if he didn’t want to be there all we would capture would be an anxious dog brain.
Because the patient table of the MRI is elevated, the ideal dog would be unafraid of heights, let alone enclosed spaces. Because we would most likely be studying several dogs, the ideal subjects would need to be social. And because there would be different people at the scanner—including MR techs, vet techs, and people from the lab—the dogs would also have to be unafraid of strangers.
In Georgia, thunderstorms occur with regularity during the spring and summer. I don’t know if there is a higher proportion of dogs with thunder-phobias in the Southeast, but it is very common in Atlanta. Even though the MRI doesn’t sound like thunder, an existing negative association to loud noises might make training difficult. As long as the dog didn’t have a noise phobia, we could gradually acclimate him to the specific type and volume of noise the MRI makes.
How Dogs Love Us: A Neuroscientist and His Adopted Dog Decode the Canine Brain Page 4
