The signature-testing approach to exploring the evolution of dogs' social cognition

Abstract

This thesis explores whether the signature-testing approach is effective for investigating dogs’ social cognition. I begin by examining whether dogs show a similar ‘watching-eye’ effect as humans and take longer to steal food when in the presence of pictures of eyes. I then demonstrate whether scaffolding behaviour improves dogs’ performance on an A-not-B detour task as it does for infants. Next, I test whether dogs engage in contagious yawning by reanalysing data from six previous studies. I test also whether this contagious yawning is reliable indicator of empathy by examining whether dogs show any of the familiarity, gender, or sociality biases in their propensity to contagiously yawn. Finally, I examine whether dogs show a ‘justification of effort’ effect and whether this effect is better explained in terms of cognitive dissonance or a within-trial contrast effect by seeing how hard dogs pull to reach toy vary depending on the effort required to obtain the toy or how they had to wait for the toy. My results show that dogs do not show the ‘watching-eye’ effect as humans, nor do they respond to scaffolding behaviour in a similar way to infants. Dogs do contagiously yawn but it is not a reliable signal of empathy as they do not exhibit any of the three predicted biases. Finally, dogs do show cognitive dissonance, pulling harder for high effort toys but not long delay toys. Together, my results provide support for the Evolved-Potential hypothesis for the evolution of dog cognition: dogs have a specific suite of cognitive mechanisms adapted for living humans but that these cognitive mechanisms did not necessarily result from the same selection pressures shaping human social cognition. I conclude by assessing how useful the signature-testing approach was for exploring dogs’ social cognition and whether there are any conclusions that can be drawn about the usefulness or the limitations of the signature-testing more generally. The resulting general principals can be applied to a wide range of cognitive domains.