While adults can readily report another agent’s false belief, theories of belief processing typically assume that this process requires the inhibition of one’s own salient current knowledge: belief processing involves overcoming an initial “egocentric bias” towards one’s own knowledge. However, evidence for the presence of egocentric bias during tasks in which adults explicitly report another agent’s false belief is surprisingly limited, with some studies providing conflicting results (e.g., Wang & Leslie, 2016; Rubio-Fernandez, 2017), failures to replicate (e.g., Ryskin & Brown-Schmidt, 2014; Samuel et al., 2018) or data that do not clearly support the presence of an egocentric bias (e.g., Back & Apperly, 2010). In three lab-based psychology experiments we used computer mouse tracking to attempt to measure, in adults, egocentric bias during an unexpected transfer false belief task. Mouse tracking allows researchers to measure the online competition between different response options when one makes a decision, and thus has the potential to reveal attraction to response options that reflect participants’ own knowledge during a false belief task. In all three experiments, participants viewed video scenarios in which an agent had either a true belief (“TB-scenarios”) or a false belief (“FB-scenarios”) as to the location of a set of keys. In each video, the agent first watched the keys hidden in one of two cups. The keys were then moved to the other cup either in the agent's presence (TB-scenarios) or absence (FB-scenarios). At the end of each video participants used a mouse to answer questions presented on the screen by moving the mouse from the bottom centre of the screen to click on one of two response boxes located in the top left and right of the screen. Key experimental questions required participants to either answer "where are the keys currently hidden?" (“reality” questions) or "where does she think the keys are?" (“belief” questions). Participants also received filler questions, asking them about the colour of the shirts of the actors in the videos or which cup was physically closest to a given actor. Answers to all questions were always either "Red" or "Blue", and the location of the response box corresponding to these two possible answers remained fixed for participants across all trials. Overall, participants received 128 trials: 64 experimental trials (16 of each Scenario x Question combination) and 64 filler trials. On each trial participant response accuracy, response time, time taken to first move the mouse, and time-stamped mouse coordinates as they made their response were recorded. Critically, on belief questions on FB scenarios the alternative incorrect answer reflects participants’ own knowledge of where the keys actually are (if the agent thinks the keys are in the "red" cup, then the participant will know that the keys are in fact in the "blue" cup). In contrast, on all reality questions, and belief questions on TB-scenarios, it is the correct response that is consistent with the participants’ own knowledge of the location of the keys. If there is an egocentric bias towards one’s own knowledge when answering these questions, then the alternative incorrect answer will be most salient during belief questions specifically on FB-scenarios: one would predict more errors, longer response times and greater mouse path deviation towards the incorrect answer on belief compared to reality questions, specifically on FB scenarios (i.e., a statistical interaction between question-type and scenario). Experiment 1 and Experiment 2 both used the same overall design and instructions, but varied on differences in event timings in the video scenarios. In Experiment 1 FB and TB scenario videos had the same overall length, but varied as the to length of time that elapsed between the final hiding event and the appearance of the onscreen question -- TB scenarios had a shorter delay compared to FB scenarios. In Experiment 2, this delay was kept constant between both scenarios, but FB and TB scenario videos varied in overall length, with TB videos lasting longer than FB videos. In neither experiment did we find evidence supporting the presence of egocentric bias. In Experiment 3, participants received the same videos as in Experiment 1, but were now instructed to attend to the agent's belief while watching the videos. This instruction manipulation allowed us to check whether the patterns of performance seen in Experiment 1 and 2 were the result of participants engaging in some belief processing when responding to the questions when they appeared onscreen. If participants in Experiment 1 were actually engaging in belief processing while watching the videos (i.e.

Topics: