Overview
- Researchers pretended to pour juice into two transparent cups and to empty one, and Kanzi chose the cup that would still contain pretend juice 34 of 50 times (68%).
- In control trials offering a real-juice cup versus a pretend one, he selected the real juice 14 of 18 times (about 78%), indicating he distinguished real from imaginary rewards.
- A similar task with imaginary grapes showed comparable performance at roughly 69% correct, suggesting consistent tracking of pretend objects across displacements.
- The study, published Feb. 5 in Science by Amalia Bastos and Christopher Krupenye, adapted child-development “tea party” paradigms for a nonhuman subject and included controls to rule out simple cue-following.
- Experts praised the rigor but highlighted limits and next steps, calling for replication with less-enculturated apes and tests of self-generated pretense; Kanzi, a uniquely language-trained bonobo, died in 2025.