Philip was already a trained assistant to his disabled owner, and he readily followed Topal’s commands. First, Topal told him to stay, and then commanded, “Do as I do.” The researcher then performed a simple action, such as jumping in place, barking, putting an object in a box or carrying it to Philip’s owner. Next, Topal ordered, “Do it!” and Philip responded by matching the scientist’s actions.
Despite Philip’s abilities, “nobody really cared, or saw that it could be useful for investigating how dogs learn or see their world,” says Adam Miklosi, a behavioral ethologist who was part of Topal’s team. In 2009, another team concluded that dogs were able to imitate only if there was no more than a five-second delay between watching the action and repeating it. With such a short retention span, the imitation skills seemed useless.
But Miklosi didn’t fully accept this; he thought that the experiment needed to be adjusted by teaching the dogs more explicitly.
In his system, there are two basic commands the dogs must learn: “Do as I do,” which means to pay attention to what is being demonstrated. The second command, “Do it,” requires the dog to imitate what it has seen. A third command instructed the dog to wait before performing the imitation. Deferred imitation is considered a sophisticated cognitive skill because it requires recalling an action after a delay of a minute or more.
Miklosi and a graduate student who is also a dog trainer worked with eight dogs — all females of various breeds — and their owners. For instance, an owner would tell her dog, “Stay,” and then command, “Do as I do,” whereupon the owner might walk around a traffic cone or put her head in a bucket placed on the ground or ring a bell suspended from a bar. After returning to her dog, the owner would wait five seconds, then order, “Do it!” The dog was expected to copy her owner’s behavior.
To see how long the dogs retained the memory, the owners were asked to add another step to the test. After saying, “Do it!”they walked their pets behind a screen so that the animals couldn’t continue to look at the objects used in the training exercises. Then they waited for up to 30 seconds before returning to the starting position.
Once the dogs could imitate the behavior twice in a row after waiting for 30 seconds, they were ready for the testing phase. Each dog was given 19 tests, including copying a familiar action, a novel action and a distracting action.
All the dogs were shown the same novel action: Each watched her owner enter a wooden box. In those cases, they were expected to wait behind the screen for a full minute before returning to the starting position and being told, “Do it!” For the distracting-action tests, the dogs watched the owner do something they had seen before. Again, they were led behind the screen, but this time commanded to lie down or fetch a ball. The waiting periods during these distraction sessions lasted from 30 seconds to four minutes.
The dogs endured their longest breaks after watching a familiar action — with times varying from 24 seconds to 10 minutes. The dogs also showed their smarts by repeating the action that they’d witnessed, even when a person who did not know which action the dog was expected to copy gave the “Do it!” command.
All the dogs completed 18 trials, scoring almost perfect marks: Six dogs made one error each, one dog made two, and another made six, the team reports in the journal Animal Cognition. This suggests, says Claudia Fugazza, a graduate student who worked on the study, that dogs have declarative memory — long-term memory about facts and events that can be consciously recalled. Until now, only humans have been shown to have this type of memory.
“It is a very nice demonstration of deferred imitation in dogs,” says Frans de Waal, a primatologist at Emory University who suggests that now that this ability has been found in dogs, it’s likely to be found in many other animals. Still, the discovery will probably be a surprise to even the most experienced dog trainers, says Brian Hare, a comparative psychologist at Duke University.
“I doubt that they would have predicted that dogs can learn new actions by observing what a human does, remembering the actions and then repeating those actions, after translating them to their own doggy body plan,” he said. And while de Waal agrees with the researchers that the dogs must be using declarative memory to do this type of imitation, Hare and others are less certain. “That’s the weakest part of the study,” says Jonathon Crystal, a comparative psychologist at Indiana University at Bloomington. “But the evidence for delayed imitation is solid and impressive.”
The researchers hope that trainers, especially those teaching guide dogs and other working dogs, capitalize on the animals’ willingness to learn by watching. “They do it so naturally, because dogs are predisposed to learn socially from us,” Miklosi says.
This article was produced by ScienceNOW,the daily online news service of the journal Science.