considering a current dispute about the results from a paradigm purportedly tapping an automatic process for calculating other agents’ visual perspectives. In the basic paradigm, participants view an image of a room in which an avatar is standing in the middle and facing either to the left or to the right (this is varied from one trial to the next). On each trial, anywhere from zero to three red discs are displayed on the walls of the room, such that on some trials the avatar can see all of them (if she, for example, is facing to the left, and three discs appear on the wall to the left but none appear on the wall to the right), but on other trials the avatar can only see some but not all of the discs (e.g. if she is facing to the left, and one disc appears on the wall to the right). Participants have the task of either calculating how many discs the avatar can see or calculating how many discs they themselves see (this is varied from one trial to the next). The basic finding on the latter type of trial is that participants’ performance is impaired when they themselves can see a different number of discs than the avatar can see. In other words, participants appear to have been calculating how many discs the avatar could see even though it was irrelevant to, and indeed interfering with, their task. In a follow-up study using the same paradigm,presented participants with an additional cognitive load during the experiment, and found that the interference from the avatar’s perspective increased. The authors interpret this as evidence that participants calculated what the avatar could see (level-1 perspective taking) automatically, in parallel to the calculation of what the subject herself could see. As the authors put it: ‘This is the first direct evidence of a cognitively efficient process for “theory of mind” in adults that operates independently of executive function. In contrast, the selection of which perspective to draw upon in judging the number of discs is a controlled process requiring executive resources, and is therefore impaired by the cognitive load manipulation.