Chaser (15 page)

Despite the assertions of trainers and farmers that Border collies, as well as other herding and working dogs, could learn a sizable number of words, prior studies had found that dogs knew only around a dozen words, most of them verbs like “sit and “stay.” In contrast, the article said, “Rico's ‘vocabulary size' is comparable to that of language-trained apes, dolphins, sea lions, and parrots.”

Moreover, the article said that Rico was capable of learning by exclusion, or “fast mapping” as it has been dubbed in studies of young children. This meant Rico could fetch an unfamiliar object from a group of familiar ones after hearing its name for the first time.

Summing up their findings, the paper's three authors observed that they could not say if Rico's abilities were the result of “an exceptional mind” or his training. Noting that “dogs appear to have been evolutionarily selected for attending to the communicative intentions of humans,” they wondered if “some of [Rico's] talent may be accounted for by the fact that Border collies are working dogs” who follow verbal commands in herding livestock. They asserted, however, “Our results strongly support the view that a seemingly complex linguistic skill previously described only in human children may be mediated by simpler cognitive building blocks that are also present in another species.”

When an article counters the prevailing scientific consensus or addresses a long-standing debate on a major topic,
Science
sometimes publishes an accompanying “Perspective” on it written by a leading scholar. That was the case here, with a “Perspective” titled “Can a Dog Learn a Word?” by the childhood language learning expert Paul Bloom of Yale University. Bloom accepted that Rico could correctly identify more than two hundred objects by their proper noun names by fetching them on command. He even said, “For psychologists [studying language learning], dogs may be the new chimpanzees.” But he expressed strong reservations about whether Rico really understood that each name had an independent meaning apart from fetching the object associated with it.

Rico's owners said that on command he could also perform other actions with the objects whose names he knew. But the Max Planck Institute researchers had only tested Rico in fetching, and Bloom rightly insisted that that was the only evidence science could evaluate. Among other things, Rico's owners might sincerely believe their dog knew how to do these things but not realize that they were unconsciously cuing him as to what to do.

Bloom noted that human toddlers begin to learn words by associating them on a one-to-one basis with specific things, individuals, and actions. But toddlers quickly develop and display the understanding that words can be used in different ways in different contexts. In addition to learning that proper nouns refer to specific, unique things, toddlers also quickly learn that common nouns refer to whole categories of things. And they can learn words simply by overhearing them, whereas “Rico . . . learn[ed] only through a specific fetching game.” He concluded, “Rico's abilities are fascinating, but . . . it is too early to give up on the idea that babies learn words and dogs do not.”

The Rico paper and Paul Bloom's critique of it provided several goals for my research with Chaser. One striking fact was that the most successful language-trained animals, no matter what their species, all had roughly comparable vocabulary sizes of around one hundred to five hundred words. When the results of a number of experiments cluster, you have to ask if that apparent pattern is the whole story. I wondered if one hundred to five hundred words was the ballpark all animals were in, or if more was possible, and I wanted to test the limit.

I also wanted to see if Chaser could make the leap that toddlers make and understand a variety of words—proper nouns, common nouns, verbs, and modifiers—according to how they are used in different contexts. Exactly what makes this leap possible in children is unknown. But there is no doubt that it happens as every child acquires language. There is also no doubt that one-to-one associative word learning is a necessary first stage along the way. For Chaser to understand that some words represent categories and to learn concepts, our best chance was building a cycle of positive reinforcement in associative word learning.

The Rico paper did not explain how Rico's owners trained him to learn an object's name. But given Chaser's fast-developing memory for the names of her toys, the appearance of the article at just this time felt like a sign that we were on the right path.

As soon as the Jeep incident was behind us, I began to do more intensive training on the names of objects. The procedure began with Chaser sitting in front of me in the living room. So that Chaser could learn without errors and build on a series of small successes, the only toy in sight was a little stuffed Santa Claus doll that we'd played with many times, and whose name—Santa Claus—I had said dozens of times over the previous two months.

Holding up the doll and pointing to it, I said, “Chaser! This is Santa Claus.”

Saying “Pop-Pop hide Santa Claus,” I lowered the doll to the floor in plain view. Chaser's gaze never left it, and I trusted that she was giving me her ear and a glance of her eye.

I dropped the doll on the floor and said, “Chaser, find Santa Claus,” as she darted forward to pick it up in her mouth.

“Good dog!” I told her. She looked up at me triumphantly, tail wagging, ears pricked up, and eyes bright and wide, eager to find out what was next.

Stepping back a few feet I said, “To Pop-Pop, to Pop-Pop. Here,” beckoning her to me with outstretched arms. She came to me and let me take the doll from her mouth, as I again said, “Good dog!”

Tossing the doll to her in a high arc, I said, “Chaser! Catch Santa Claus. Catch Santa Claus.” She reared up on her hind legs to catch the doll, and as her front feet hit the ground again her tail went up and wagged back and forth with excitement and pleasure. “Good dog, good girl!” I said, positively reinforcing her.

“To Pop-Pop. Here,” I said. She brought me Santa Claus, and her tail wagged faster at hearing another enthusiastic “Good dog!”

Finally I tossed Santa Claus into the middle of the living room and put Chaser through the herding behaviors: “Chaser, come by. Come by Santa Claus. Whoa. Way to me, way to me. There, there. Drop. Drop behind Santa Claus. Chaser, crawl to Santa Claus. There. Chaser, one, two, three, take Santa Claus! Good dog!”

After this herding play with as many repetitions of the name Santa Claus and as much positive reinforcement as possible, we began all over. To keep the game interesting I progressively hid Santa Claus so that the doll was harder and harder to find. In each word training session, consisting of a brief trial at finding Santa Claus and a few minutes of play with it, I said the name twenty to forty times. Over the course of the day's training sessions, we went through this exercise with Santa Claus at least twenty times.

Over the next few days we continued to play this game with Santa Claus. We also played the game with other objects that were already familiar to Chaser. The excitement of finding the object, whether it was hidden in plain view or in the other room, put a grin on both our faces and got Chaser's tail wagging with pleasure. After a brief moment to celebrate that victory, I rewarded Chaser with play with the object for three to five minutes. Depending on the object, the play might involve a little tug-of-war, chasing her as she scampered away with it, throwing it for her to catch or fetch, and most of all herding games. All the while, I repeatedly said the name of the object in very simple statements that described what we were doing and positively reinforced her with praise and pets: “Chaser, catch ___. Good dog! Chaser, shake ___. Good dog! Chaser, fetch ___. Good girl, Chaser!”

After a few trials with different objects, Chaser knew what to expect, and her behavior briefly made me fear she had lost interest in what I was doing. When I said, “Chaser, this is ___,” she lay down on the floor and tilted her head as if she were completely ignoring the object and what I did with it. She remained lying on the floor, seeming utterly complacent, even bored. But as soon as I said, “Chaser, find ___,” she sprang to her feet and raced to find the object.

As she did so I repeated, “Chaser, find ___, Chaser, find ___,” in a soft, encouraging tone until she found it. This I greeted with a triumphant “Good girl, Chaser!” followed by a round of play with the object.

We did this many times for each toy, and I was curious to see whether as time went on Chaser needed fewer repetitions to learn the name of a toy. My criterion for saying she had indeed learned a word had two stages. First she had to select the correct object out of a group of objects that included seven other objects whose names she had learned. I called this the 1-of-8 test. And then she had to do the same thing seven more times in a row over the course of a day or so, each time with a different set of seven other objects whose names she knew as distracters. I called this the 8-of-8 test. After asking for the newly learned object in a group of eight toys, I also asked for the seven previously learned objects by name as a way to rehearse and test prior learning.

If Chaser missed fetching any object by its name in the 1-of-8 or 8-of-8 tests, I gave her more training and play with that object alone, with no other objects in view. She missed here and there, but her performances on the 1-of-8 and 8-of-8 tests were usually 90 to 100 percent correct.

As we did this with more objects I also needed to assess Chaser's long-term memory of the objects' names. I had to be sure that Chaser wasn't freeing up memory space for new object names by forgetting older ones. So every month I informally tested Chaser's retention of her complete vocabulary, to that point in time, with random sets of twenty objects on the floor in front of her. In all these tests over a period of three years, Chaser always got at least eighteen of twenty right. At the end of that time it took fifty-one sets of twenty to test her long-term retention of the names of 1,022 toys, a process that I spread over about two weeks every month.

I became an expert at evaluating toys for Chaser's play and training. I could take a brief look at a stuffed animal and gauge if it would last a week or a year without needing to be sewn up. I was always looking for the sturdiest ones possible. The growth in Chaser's flock meant that there were multiple Frisbees, balls, and stuffed lions, tigers, and bears. Despite these similarities, every object had to have unique features that Chaser could distinguish visually. Over time it became harder and harder to find toys that weren't highly similar to ones we already had.

Occasionally Sally and I splurged ten dollars on a dog toy from the pet store, especially during after-Christmas sales. And I was always looking for the perfect Frisbee. I tried to stick to cloth Frisbees as opposed to plastic ones with hard edges. But I bought the majority of the toys mainly at the Salvation Army store, which was closer, and the Miracle Hill thrift shop, which was much cheaper. The Salvation Army store wanted seventy-five cents or a dollar per toy, whereas Miracle Hill would sell a big plastic garbage bag full for a couple of bucks.

Sally was extremely patient as the house filled up with toys. Although we bought many large Rubbermaid containers to store them, there were always at least fifty toys strewn around for training, testing, and pure play. But sometimes the clutter of toys got to be a little too much for Sally and we had to do a rapid cleanup. This got a little easier and more pleasant once Sally decided to teach Chaser to clean up her own toys the same way she'd taught Robin and Debbie when they were small and later helped Debbie teach Aidan.

“Clean up, clean up. Everybody clean up,” Sally sang, as she picked up one of Chaser's toys and dropped it into a plastic storage tub in the middle of the living room. Attracted by her singing, Chaser looked at Sally, head cocked to one side with curiosity. “Clean up, clean up. Everybody clean up,” Sally continued singing. She picked up two more toys and gave one to Chaser, who took it eagerly in her mouth. Continuing to sing, Sally pointed at the plastic tub and dropped the toy she was holding into it. Chaser brought her toy over in response to Sally's song and pointing, and dropped it into the plastic tub.

“Good girl, Chaser!” Sally said, before resuming her song and picking up more toys for Chaser to take and drop in the plastic tub. Midway through the cleanup, Sally stopped picking up toys for Chaser and simply pointed at the ones that were still scattered around on the floor and the furniture. It didn't take too many repetitions of this song and coaching for Chaser to learn the cleanup game. She enjoys it, too, coming eagerly when she hears someone sing, “Clean up, clean up. Everybody clean up.”

Later I learned that telling Chaser “Clean up, and we will play Frisbee in the yard” makes her even happier to play the cleanup game. She darts around picking up her toys, and as soon she has dropped the last one into a Rubbermaid container she heads for the door with her tail wagging.

Sally also sewed up dozens of toys over the years, especially when she became exasperated by my duct tape repairs. That did not include beanbag-type toys, however. Early on we learned to avoid these after Chaser shook one vigorously, a seam ripped, and the beans flew everywhere.

As Chaser's flock expanded to include more named objects, the likelihood that her correct answers were the result of chance became smaller and smaller. That was because the possible answers were not only the named objects in the groups of eight or twenty, but all the objects whose names she had learned at the time of the test. The more object names she learned, the more impressive her consistent, very high level of accuracy became.

The probability of her correct answers resulting from chance was small in her first tests, and her correct answers were well above the threshold that made them statistically significant. In the later tests the probability that her correct answers resulted from chance was vanishingly small. Chaser's proper noun learning reached the point where the probability that her correct answers resulted from chance was less than 1 in approximately 20 million.