Meta-analytic techniques reveal that corvid causal reasoning in the Aesop’s Fable paradigm is driven by trial-and-error learning

Laura Hennefield, Hyesung G. Hwang, Sara J. Weston, Daniel J. Povinelli

The classic Aesop's fable, Crow and the Pitcher, has inspired a major line of research in comparative cognition. Over the past several years, five articles (over 32 experiments) have examined the ability of corvids (e.g., rooks, crows, and jays) to complete lab-based analogs of this fable, by requiring them to drop stones and other objects into tubes of water to retrieve a floating worm (Bird and Emery in Curr Biol 19:1-5, 2009b; Cheke et al. in Anim Cogn 14:441-455, 2011; Jelbert et al. in PLoS One 3:e92895, 2014; Logan et al. in PLoS One 7:e103049, 2014; Taylor et al. in Gray R D 12:e26887, 2011). These researchers have stressed the unique potential of this paradigm for understanding causal reasoning in corvids. Ghirlanda and Lind (Anim Behav 123:239-247, 2017) re-evaluated trial-level data from these studies and concluded that initial preferences for functional objects, combined with trial-and-error learning, may account for subjects' performance on key variants of the paradigm. In the present paper, we use meta-analytic techniques to provide more precise information about the rate and mode of learning that occurs within and across tasks. Within tasks, subjects learned from successful (but not unsuccessful) actions, indicating that higher-order reasoning about phenomena such as mass, volume, and displacement is unlikely to be involved. Furthermore, subjects did not transfer information learned in one task to subsequent tasks, suggesting that corvids do not engage with these tasks as variants of the same problem (i.e., how to generate water displacement to retrieve a floating worm). Our methodological analysis and empirical findings raise the question: Can Aesop's fable studies distinguish between trial-and-error learning and/or higher-order causal reasoning? We conclude they cannot.