<p>Three experiments investigated efficient belief tracking as described by the two-systems theory of human mindreading (Apperly & Butterfill, 2009) whereupon mindreading implies the operation of a flexible system that is slow to develop and cognitively effortful, and an efficient system which develops early but subject to signature limits. Signature limits have been evidenced by children’s and adults’ difficulty anticipating how someone with a false belief (FB) about an object’s identity, will act. In a recent investigation of signature limits, erroneous pre-activation of the motor system was detected when adults predicted the actions of an agent with an identity FB, suggesting that efficient mindreading and motor processes are linked (Edwards & Low, 2017). Moreover, young children differentiated between true and FBs about an object’s location, but not identity, as revealed by the object children retrieved in an active helping task (Fizke et al., 2017). The aim of the present thesis was to provide new evidence of signature limits in adults, and of the recent conjecture that efficient mindreading and motor processes interact. In helping tasks, participants’ interpretation of another’s actions is crucial to how they coordinate their helping response. Therefore, an ecologically valid helping task was adapted to investigate the proposed interface between efficient mindreading and motor processes. The present work measured adults’ eye movements made prior to helping, and their helping actions across a set of distinct directional full-body movements (around which side of a desk they swerved, which compartment they approached, toward which compartment they reached, and which object they retrieved). In this way, it was possible to investigate whether gaze direction correlated with full-body movements and whether adults’ gaze differed when the agent’s FB was about an object’s location or identity. Results from Experiment 1 indicated that efficient belief tracking is equipped to process location but not identity FBs, and that - in the location scenario - gaze direction correlated with the immediate stage of participants’ helping action (the direction they swerved). To investigate this correlation further, Experiment 2 drew upon research suggesting that temporarily tying an observer’s hands behind their backs impaired their ability to predict the outcome of hand actions (Ambrosini et al., 2012). Results showed that tying adults’ hands behind their back had a negative effect on their gaze behavior and severed the correlation between gaze and swerving, suggesting that the link between efficient mindreading and motor processes is fragile. Experiment 3 tested an alternative interpretation for Experiment 2’s findings (that restraining participants’ hands applied a domain-general distraction, rather than a specific detriment to belief tracking) by tying up participants’ feet. Results were ambiguous: the gaze behavior of participants whose feet were tied did not differ from those who were unrestrained, nor from those whose hands were bound. These findings support the two-systems theory and provide suggestive evidence of a connection between efficient mindreading and motor processes. However, the investigation highlights new methodological challenges for designing naturalistic helping tasks for adult participants.</p>
Prediction of Human Full-Body Movements with Motion Optimization and Recurrent Neural Networks
