Ratcheting up tool innovation in Goffin's cockatoos (Cacatua goffiniana): The effect of contextually diverse prior experience

Abstract

The ability to gain information from one situation, acquire new skills and/or perfect existing ones, and subsequently apply them to a new situation is a key element in behavioural flexibility and a hallmark of innovation. A flexible agent is expected to store these skills and apply them to contexts different from that in which learning occurred. Goffin's cockatoos (Cacatua goffiniana) are highly innovative parrots renowned for their problem-solving and tool-using skills and are thus excellent candidates to study this phenomenon. We hypothesized that birds allowed to use a tool in a larger variety of contingencies would acquire a broader expertise in handling it, facilitating its transfer to new tasks. In our study, we compared the performance of two groups of captive Goffin's cockatoos (N = 13): A test group received more diverse learning and motor experiences on multiple applications of a hook-type tool, while a control group received intensive, total trial-matched, experience with a single application of the same tool. Then, both groups were tested on two novel tasks to determine whether experience with the tool in multiple contexts would facilitate performance during transfer. While both groups transferred to both novel tasks, group differences in performance were apparent, particularly in the second transfer task, where test birds achieved a higher success rate and reached criteria within fewer trials than control birds. These results provide support for the prediction that experiencing a diverse range of contingencies with a tool appears to allow birds to acquire generalizable knowledge and transferrable skills to tackle an untrained problem more efficiently. In contrast, intensive experience with the tool in a single context might have made control birds less flexible and more fixated on previously learned tool-dependent instances.

Keywords: behavioural flexibility; parrot cognition; problem‐solving; tool use transfer.

FIGURE 1

Apparatuses used for phase 1. (a) Basket: The reward is obtained by probing the hook through the transparent tube and inserting it through the basket's handle, then lifting it up. (b) Skate: The reward is obtained by passing the hook through the front window (left side of the drawing) and grabbing the protruding stick on the skate with the hook, then pulling it out. (c) Trap: The reward is obtained by inclining the hook upwards and grasping the ring with it, then pulling the trap down to open. Elements in yellow indicate target in the apparatus. Dark‐coloured arrows denote the movement displayed by the bird and light‐coloured arrows reflect the directionality of the reward following bird's movement.

FIGURE 2

Apparatuses used for phase 2. (a) Seesaw task: The reward is obtained by actively inserting the hook through a hole located on the top of the box, grasping a lever (target element, coloured in yellow) at the end of the seesaw and pulling it upwards. (b) Cane task: The reward is obtained by pulling the pre‐inserted functional hook towards the bird. Dark‐coloured arrows denote the movement displayed by the bird and light‐coloured arrows reflect the directionality of the reward following bird's movement.

FIGURE 3

Probability of a trial being solved in the cane task as a function of session number. Performance of control group (ctrl) is shown in light grey, and test group (exp) is depicted in dark grey. Point area is proportional to the number of subjects that participated on each session (i.e. areas have different sizes because of some individuals reaching solving criterion earlier than others). Shaded line corresponds to 95% confidence intervals. Horizontal dotted line reflects performance at chance level. The plot shows the model with sex centred and 1000 bootstraps

FIGURE 4

Time to solve a trial (in seconds) in the cane task as a function of session number. Performance of control group (ctrl) is shown in light grey with a solid line, and test group (exp) is depicted in dark grey with a dashed line. Point area corresponds to the number of solved trials per session across all subjects of the same group; lines represent the prediction made by the fitted model. Values on the y‐axis are in seconds.