Motor imitation in kea parrots

Humans learn many of their skills by imitating others in their environment. While social learning has been shown time and time again in a wide range of animal species, evidence of true imitation is largely lacking. Motor imitation, the copying of physical movements or interactions with objects, has been shown in a few species, although often under very specific conditions. The task either relied on very distinct actions, meaning that the accuracy with which the observer imitated was impossible to assess. Or it relied on having highly encultured subjects, e.g. domestic dogs or human-raised apes, to be able to use a human as the demonstrator, and thus be able to control how precise the demonstration was. Here we propose a project where we would test the kea parrot of New Zealand on their ability to imitate the actions of another kea. First, we would use the standard two-action task, where demonstrators are trained to open a box in one of two specific ways. The observers ability to imitate can then be seen in choosing the action of their demonstrator over the other possibilities. Then we would train a kea to peck a specific sequence of dots on a touch screen computer which was first shown by a demonstrator. While the basic task is the same, imitate the actions seen, the higher number of possible choices means that we can look for degrees of accuracy in the imitations. Lastly, we would use the so called Do-as-I-do paradigm, where a demonstrator performs different known actions, and the observer is rewarded if they imitate these accurately. Subsequently, novel actions are demonstrated to evaluate the spontaneous imitation of these. The difference to the previous tests lies in the inclusion of actions that are not needed to solve a task, for example raising a wing. Such intransitive actions are readily imitated by human children, but so far the evidence from non-human animals is limited; only a recent dog study was able to show that their subjects would imitate an irrelevant action (touching of a colour spot) when their caregiver demonstrated it in a sequence to get to a reward. Thus this project would break new ground in many ways, and advance our knowledge of imitation in non-human animals.

In the first study investigating kea parrots ability to imitate, we matched an experiment previously done with budgies, which had been shown to match a demonstrator's action when opening a box, as well as the colour and side of the lid demonstrated. When the kea were tested, the test group, which got a demonstration, solved the task quicker than the control group, which did not get a demonstration, a clear effect of social learning. However, none matched the action, lid colour or side of the demonstration. It was thus not imitation, where the goal is reached by copying all observed details, but rather of emulation, where the goal itself is learned, as well as some aspects of the solution, for example that removing the lids gave access to the reward. As the task was solved by every kea in every trial, independent of group, may have also simply been too easy for the kea. The second task was a two-step sequence, where the kea had to pull out a peg which then unlocked a sliding door, allowing them to pull it open and access a reward. There were two pegs and two doors, and the kea needed to walk around the box between actions, as the left peg unlocked the right door, and vice-versa.; both doors would also lock again, if both pegs were pulled out. More birds from the test group, solved the task compared with those from the control group. Yet, almost half of the successful birds did not copy the solution they observed but solved it in the opposite direction. They had learned about the action and the goal, but again did not copy the colours of the pegs and doors or the position/direction of the demonstration. This further suggested that emulation rather than imitation, was the normal social learning mechanism for kea, which can be explained through their ecology. In the wild, the kea feed from a very large number of food resources, which are scattered throughout their mountain habitat. If a kea watches another bird dig up a root and eat it, copying the same actions on the same plant in the same place would not be useful, as that root is already gone. Rather, as in emulation, learning about some aspects of the task (this kind of plant can be dug up) and the goal (there is an edible root underneath), would be more useful when searching for food. Through the third and fourth tasks, both designed to require the kea to match colour/shape, we found that the kea could learn to pay attention to these aspects when necessary.