How to Teach Pigeons to Play Pacman: Virtual 2D spatial navigation using a touchscreen video game

Video-game experimental procedures have been commonly used to study problem-solving behavior in human psychology studies, as they allow for rigorous experimental control and heightened behavior quantification. Here, we present the methodology and empirical findings from a redesigned touchscreen navigation procedure resembling the classic video arcade game Pac-Man® (hereafter “pacman”) played by pigeons. Our objective was to develop and test a procedure to study insight behavior, similar to Epstein et al. (1984). The results of our cumulative insight experiment have been published in Rodrigues et al., (in press). In this manuscript, we detail the training methodology and procedure used to teach pigeons to play our pacman task. In Experiment 1, pigeons were trained to guide the virtual pacman across an arena on a touchscreen display using four directional guides pointing in the cardinal directions. When the pacman reached the goal – a visual icon of a banana – the pacman “ate” it, and the pigeon received food reward. In Experiment 2, we trained pigeons to navigate the pacman around visual barriers. Sometimes, the barriers intervened between the pacman and the banana, and, thus, pigeons were rewarded for successfully navigating the pacman around these barriers. In Experiment 3, pigeons were trained to pass through portals that could transport the pacman from one part of the digital arena to another. Finally, we presented novel configurations of puzzle tasks incorporating all the previous tools in Experiment 4. Overall, results across experiments showed that some subjects (3) were able to learn how to navigate the pacman with a high degree of control, while other subjects (5) struggled with different points of training. We discuss the methodological contributions of this virtual procedure and possible refinements, concluding with its potential extension to other problem-solving tasks and its applicability to non-human species, thereby enabling comparative investigations of problem-solving across animals. This work was supported by National Science Foundation BCS-1844144 (Blaisdell), and FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo) grants 19/27401-1 and 21/09666-8. notReviewed other