Everything was going faster than I had anticipated. Melissa was working with McKenzie in parallel; unsurprisingly, they made even quicker progress. In fact, the training was going so well, I hadn’t even stopped to ponder what we were doing to their brains, which, after all, was the subject of the Dog Project.
Even though we were using basic behaviorist principles to shape a complex behavior in the dogs, it couldn’t explain what the dogs thought of all this. If we cared about just behavior, the reason a dog did something or what she thought wouldn’t matter. But if we got to the point of actually scanning their brains, the dogs’ motivations could have a big impact on what we found. Doing something for food would look very different from doing something for social praise or, dare I say, love.
I had the nagging feeling that the ease with which dogs slip into human lives could not be fully explained by behaviorist theories. For dogs to do what they do, they must have a rich inner life that goes beyond a chain of actions resulting in food. Dogs must have a rich mental model of their environment. As highly social animals, these mental models are likely to be weighted heavily toward social relationships. Not just dominance and subordination, but more fluid models of how they should behave with members of their household, either dog or human, and how these interactions will affect their current state of well-being.
It makes you wonder who is training whom. Skinner and Pavlov were partly right. Their principles are highly effective in training behavior. But they studied animals in a laboratory—a place where the human controls pleasure and pain. Dogs in their native environment—the human household—interact with us in a much more natural way. There is give-and-take, and testing, on both sides.
10
The Stand-In
WITH CALLIE AND MCKENZIE making rapid progress in their training, we would soon be ready to make the jump to the real scanner. Although the mock head coil and tube were good facsimiles of the MRI, they weren’t the real things. There was no way to simulate the smells and sounds of the hospital, for instance. We wouldn’t know how the dogs would react until we actually brought them there.
The initial introduction to the real MRI would be critical. Dogs can form negative imprints of environments based on one event: a loud noise such as a slamming door, an encounter with someone who doesn’t like dogs. Any of these could permanently affect a dog’s impression of the MRI. If that happened, and the dog didn’t want to go near the scanner as a result, all the training we’d done would have been wasted.
Mark and I were particularly concerned about noise. The MRI scanner makes a wide range of sounds. The magnetic field is always on, and this requires constant tending by an array of devices, like the pumps that circulate chilled water around the magnet. When you enter the room, the first thing you hear is the heartbeat of the circulation pumps. If you listen carefully, you will also hear the machine breathing, the sound of the “cold-head”—a compressor that keeps the helium under pressure.
How would the dogs react to this living machine?
When scanning a subject, MRIs are loud. Depending on the particular settings, an MRI can reach nearly 100 decibels. Every 6 decibels means a doubling in the sound pressure. Normal conversation is about 60 decibels. Busy traffic, 80 or 90. A jackhammer is 100. Hearing damage for humans begins at around 120 decibels, equivalent to a jet engine at a hundred meters. Nobody knows at what level hearing damage occurs in dogs, but one point of reference is hunting. The report of a hunting-caliber bullet is 170 decibels, and hearing loss after repeated gunshot exposure is a well-known phenomenon in both hunting and military working dogs.
Earmuffs can cut sound levels by 20 or even 30 decibels, so assuming that a dog’s hearing is more sensitive than that of a human, Callie and McKenzie would be fine as long as they wore ear protection.
The loudness of the MRI wasn’t the whole story, though. The type of sound made could also have a big effect on the dogs. The MRI makes different types of sounds depending on the type of brain scan being performed. Some scans sound like a swarm of bees, while other scans are like the klaxon of a submarine preparing to dive. The specific sound depends on dozens of parameters that are programmed for each scan. These parameters indicate how many slices through the brain will be made, how thick they will be, and whether they should focus on gray matter (the neurons) or white matter (the connections between neurons) or on changes in blood flow like we do in fMRI.
The only way to get an accurate recording of the sounds the dogs would experience during their scans would be to program the actual scan sequence with the exact parameters necessary to scan the dogs’ brains. But since nobody had scanned a dog’s brain before, at least not with fMRI, we had no idea what the correct settings might be. With a dog in the scanner, we could figure out the correct settings, but we needed the right settings in order to record the sounds to train him to get him into the scanner in the first place.
I felt like a dog chasing his tail.
At lab meeting, I brought up the conundrum.
“What if you use the standard human settings to record the scanner noise?” Lisa suggested.
“It might be good enough,” I said. “But what if it isn’t?”
“I bet the dogs could tell the difference,” Andrew said. “If we train them with the wrong sounds, they might freak out when they hear the real thing.”
“We need a stand-in,” I said. “Something that can take the place of the dog while we fiddle with the scanner settings.”
Lisa’s forehead knitted up in thought. Everyone else looked at the floor. Before anyone suggested it, I headed off the obvious.
“We’re not using a dead dog.”
“Why don’t you just go to the supermarket and buy a steak and scan that?” Gavin joked.
“You mean like the famous dead-salmon study?” Andrew asked.
A few years before, neuroscientists had used fMRI to scan a salmon purchased at a local fish market. As they wrote in their findings, the fish “was not alive at the time of scanning.” They presented their results at a conference, but most scientists dismissed it as a joke. It wasn’t. The point was to measure the accuracy of fMRI and how the technique could sometimes lead to the appearance of brain activations that weren’t actually there. Obviously a dead salmon couldn’t have brain activation, but the scientists showed that with poor statistical technique, it might appear that way.
Gavin’s joke wasn’t half bad. But a steak (or a salmon) would be a lousy stand-in for a dog.
“We need something more doglike,” I said.
“A pig?” Gavin said.
“Too big.”
“How about a lamb?” Andrew suggested.
“Can you buy a whole lamb at the market?” I wondered.
After a few phone calls to some local butchers, Andrew found a lead. It wasn’t a whole lamb—you needed to get that directly from a farm—but there was a market that might sell us a lamb’s head.
“I think he said they get their delivery of lamb heads on Wednesdays,” Andrew explained. “I’m not completely sure because I couldn’t understand some of what he was saying. But he definitely said they go fast.”
“Today is Wednesday,” I pointed out.
“Giddyup!”
The halal meat market had no sign. The “market” consisted of a counter at the rear of a convenience store, itself sandwiched in a rundown strip mall and sharing a wall with a video store specializing in bootlegged Middle Eastern movies.
Andrew and I walked in to find a trio of bearded men hanging out at the cash register, smoking cigarettes and watching soccer on TV. They said nothing as we made our way to the rear of the store. I noticed some elaborate water pipes on display.
At the butcher’s counter, a spread of organ meats glistened beneath the glass case. Kidneys I recognized. The rest—not so much. The animals of origin were a mystery to me too.
A squat guy wearing a tight soccer jersey peered over the counter.
“You guy call about lamb head?” he asked in
a Middle Eastern accent.
“Yes.”
“How many you want?”
Andrew and I looked at each other.
“How many do you have?” I replied.
“Lots.”
We conferred briefly and decided that we should have a backup in case something went wrong.
“Two,” I said.
The butcher disappeared through a doorway covered with vinyl slats. A moment later he returned and deposited two heads on the counter with an authoritative clank.
“They’re frozen,” I said.
“Yes,” said the butcher, “fresh frozen.”
They bore a resemblance to a lamb, but as all the wool had been removed, it was hard to tell what they were. The lips had retracted a bit, and the faces were fixed in permanent grimaces.
The size was right, I had to admit. In fact, they were about the same size as Lyra’s head. I shivered and pushed that unpleasant image out of my mind.
“Where is the rest of the lamb?” I asked.
“Just head,” he replied.
“Do they still have their brains?”
The butcher brought his fingers to his mouth in the sign known to foodies around the world and said, “Yes. Delicacy.”
Ideally, we would have gotten a whole lamb to stand in for a dog. Anything you put inside an MRI disturbs the magnetic field. The bigger the object, the greater the disturbance, and as the scanner compensates for these disturbances, it makes different kinds of sounds. The lamb’s head was not going to have enough mass to replicate the disturbance created by a dog. We needed something else.
Andrew pointed to a pair of hooves in the butcher’s case. They appeared to be the front legs of a calf starting just above the ankle joint.
In the actual MRI, the dogs would be scanned in a sphinx position. Their heads would be upright, supported by a chin rest, and their front paws would be sticking straight forward. Andrew realized we could use the calf hooves to simulate the front paws of the dog. The combination would give a close approximation to the shape and mass of the part of the dog that would be at the center of the scanner. We paid for our meats and headed back to the lab with two lamb heads and a pair of calf hooves.
The vegetarians in the lab weren’t going to be happy.
We let the heads thaw overnight and reserved time on the MRI scanner for the following evening. Scanning dead animal parts in the MRI is the kind of thing best done discreetly. Once thawed, the heads, now swimming in their own juices, looked even worse. Their eyes had taken on an opaque haze. Andrew and I double-bagged everything and headed to the scanner.
We were greeted by Lei Zhou, a Chinese postdoc on duty that evening. Lei had received his PhD in physics and was intimately familiar with the technical wizardry behind MRI. His English, however, had a ways to go. I could only hope that we understood each other during this unusual procedure.
Lei and Andrew preparing to scan the lamb’s head.
(Gregory Berns)
Andrew unloaded our cargo, and we proceeded to arrange it in the head coil of the scanner. With foam pads propping up the body parts, Lei snapped on the top of the coil and sent the whole mess into the center of the scanner.
When you place something in the MRI, the magnetic field tugs on the atoms inside the object. In living tissue or, as in the case of the lamb’s head, formerly living tissue, hydrogen is the most common atom. There are two hydrogen atoms in every water molecule, and water accounts for 60 percent of body weight in humans. Hydrogen is also abundant in the brain. The outer membranes of neurons and their supporting cells, called glia, are rich in fat and cholesterol, which have large numbers of hydrogen atoms.
A hydrogen atom has one proton and one electron. The proton is like a spinning top. Normally, the protons spin in random directions, but inside the MRI they line up with the magnetic field. Like spinning tops, the protons also wobble a little bit. The stronger the magnetic field, the faster they wobble. If you hit the protons with radio waves exactly in sync with their wobbling, the protons jump into a higher energy state. This is called magnetic resonance. Different types of atoms resonate at different frequencies. For the strength of scanner we use, hydrogen resonates at 127 MHz, which falls in the range of radio waves—just beyond the FM dial. Carbon, another common element in the body, resonates at 32 MHz. MRI works by sending in a blast of radio waves that excite the atom of interest—in most cases hydrogen because of its abundance and superior sensitivity to magnetic fields. When the radio waves are turned off, the protons relax back to their original state and, in the process, cause an oscillating magnetic field that can be picked up by an antenna. The head coil is nothing more than a fancy FM radio antenna that picks up these signals from the protons in the brain.
Not all protons behave the same way. The protons in a water molecule wobble slightly different from the protons in a fat molecule. These slight differences can be detected by the MRI and, with the help of a computer, be used to construct a visual image representing the types and locations of these different molecules.
We would need to do three types of scans on each subject. The localizer, which lasts only a few seconds, gives a snapshot of the location and orientation of the head in the magnet. The localizer scan of the lamb’s head came out well. We could clearly make out the brain. The human settings for the localizer seemed to work. Next up was the structural image. For humans, we like as much anatomical detail as possible, but this has to be weighed against the time it takes to get high-resolution images. Images clear enough to resolve features as small as one millimeter take six minutes to complete. Humans have no problem holding still for that long, but there was no way our dogs would. I told Lei that we needed to come up with a structural sequence that would take no more than thirty seconds. I figured that would be the limit for most dogs.
This turned out to be somewhat difficult. The normal structural scans couldn’t be completed that quickly, so we had to switch to a different type of scan. This one didn’t show as much detail, but we were able to find a combination of parameters that produced a usable image in under thirty seconds.
We spent an awful lot of time figuring out the best orientation of the brain. If you think of the MRI as being a digital bread slicer, we had to decide which way to cut the slices: left to right, top to bottom, or front to back. Since the human head is pretty close to a sphere, it doesn’t make a whole lot of difference which way you slice it. But a dog’s head, like the lamb’s, is elongated front to back and generally pretty flat from top to bottom.
As the images of the lamb’s head came up on the screen, we saw how little of the head was actually occupied by brain. Most of it was nose and muscle. Those air pockets in the nose can wreak havoc with the MRI images too. Abrupt transitions in tissue density, such as going from air to skull, cause distortions in the magnetic field, which result in warped images. By carefully selecting the orientation of the slices, you can minimize this effect. Slicing from front to back seemed to give us the best results.
Anatomical images of the lamb’s head. The slices go from front to back. The eyeballs are visible in the top row, while the brain appears prominently in the middle and lower rows. The large black cavities are nasal sinuses.
(Gregory Berns)
Finally, it was time to attempt some functional scans, which are two-second glimpses of the brain in action. By continuously acquiring these functional scans while the subject does something, we can measure changes in brain activity. Think of the functional scans as the individual frames of a movie. Even though each one takes only two seconds, the subject might be in the scanner for half an hour during functional scanning. During such a session, we would acquire nine hundred functional images, at a rate of thirty scans a minute for thirty minutes.
Of course, the lamb was dead, so we didn’t expect to see much “activation.” But we only needed to figure out how many slices it took to cover the brain and how to orient the brain for the most efficient coverage. Once we worked that out, Andrew and I recor
ded the sounds of the scanner running this sequence.
We could now introduce Callie and McKenzie to the actual noise they would experience in the scanner and gradually let them get used to it.
11
The Carrot or the Stick?
THE CHALLENGE OF ENTERING the head coil and placing her chin on the boogie board chin rest had long been overcome. As soon as Callie heard the rustling of the plastic baggie containing bits of chopped-up hot dog, she knew. She would bound into the kitchen, wagging her whole rear end, and look at me with excitement and anticipation.
“Wanna do some training?” I would ask in a high-pitched voice.
Our training regimen had outgrown the basement. The only room in the house big enough to contain what was now a full-blown MRI simulator was the living room. Kat eyed the monstrosity in her living room, a space formerly occupied by an elegant sofa set and coffee table now pushed off to the side.
“There isn’t any other place for this?” she asked.
“It’s too heavy to move down in the basement,” I replied. “And I don’t think it will fit through the door.”
“You mean you constructed this in the living room without a way to get it out?”
“No, no,” I reassured her. “It comes apart.”
I had dusted off a PA system left over from my guitar-swinging days in a garage band. As I set the speakers on a stand facing the tube, Helen came into the living room.
“What’s that for?”
“To simulate the noise of the scanner,” I explained. “It’s the only thing we have that’s loud enough.”
She nodded, and together we snaked cables from the speakers to the amplifier. We aimed one speaker at the side of the tube to simulate the vibrations that course through the MRI. The other speaker went at the end of the tube to achieve the full decibel level inside.
“Daddy?”
“Yes?”
“Can I come with you when you scan Callie?”
This question took me by surprise. I wondered what had motivated it.
How Dogs Love Us: A Neuroscientist and His Adopted Dog Decode the Canine Brain Page 9
