Do humans spontaneously take the perspective of others?

Highlights

• We employ a novel method to assess if humans spontaneously compute the visual perspective of others.

• A ‘Perspective taking’ effect occurs even when a person observed has their vision blocked.

• Our findings refute the spontaneous perspective taking theory.

Abstract

A growing number of authors have argued that humans automatically compute the visual perspective of other individuals. Evidence for this has come from the dot perspective task in which observers are faster to judge the number of dots in a display when a human avatar has the same perspective as the observer compared to when their perspectives are different. The present experiment examined the ‘spontaneous perspective taking’ claim using a variant of the dot perspective paradigm in which we manipulated what the avatar could see via physical barriers that either allowed the targets to be seen by the avatar or occluded this view. We found a robust ‘perspective taking’ effect despite the avatar being unable to see the same stimuli as the participant. These findings do not support the notion that humans spontaneously take the perspective of others.

Introduction

As humans, we often make conscious judgments concerning the mental state of other individuals in social situations. This occurs, for instance, when one wonders why a person is gazing at a particular location. Furthermore, the computation of other people's perspective is central to efficient social cognition. A number of authors have argued that certain types of ‘Theory of Mind’ processes can occur automatically such that they are fast and do not require controlled processing. The most notable example was reported by Samson, Apperly, Braithwaite, Andrews, and Bodley Scott (2010), who argued that humans rapidly and spontaneously compute the perspective of other individuals. They employed a paradigm that has become known as the dot perspective task, in which observers are presented with a human avatar (located in the centre of a virtual room) that looks either towards a left or right-hand wall. A number of discs are positioned on the two lateral walls and the participant is asked to judge the number of discs from either their own perspective or the avatar's perspective. The central manipulation concerns the consistency of the avatar's and participant's perspective; on some trials the avatar and participant can see the same number of discs whilst on other trials they see a different number. For example, if the avatar looks to the right-hand wall and one disc is located on each of the two walls, the avatar sees one disc and the participant, by virtue of being able to see the whole room, sees two. By contrast, if two discs appear on the right-hand wall and none on the left, both the participant and the avatar see the same number of discs (i.e., two). Samson et al. found that reaction time (RT) to make the disc number judgement was shorter when the viewpoint of the avatar was consistent with the participant's relative to when their viewpoints were inconsistent. The authors concluded that this consistency effect occurs because the computation of another person's perspective occurs spontaneously. In effect, the observer is said to compute what the avatar can see, and this representation includes the number of discs that can be seen. The knowledge about what the avatar sees then interferes with the observers' knowledge about the total number of discs present, thus increasing RT when the two are inconsistent.

The spontaneous visual perspective taking notion has not however gone unchallenged. For instance, Santiesteban, Catmur, Coughlan Hopkins, Bird, and Heyes (2014) argued that the avatars employed in the Samson et al. experiments act as a cue that shifts attention to one side of the display. That is, the cue was said to activate domain-general processes of spatial cognition. Indeed, the basic spontaneous perspective taking method is similar to the classic central cueing paradigm in which a cue, for instance a human face (a ‘gaze cue’; Langton & Bruce, 1999), is located in the centre of a display and looks towards the left or right hand side. Furthermore, the critical comparison of the Samson et al. method, i.e., ‘consistency-inconsistent’ (of the avatar's and participant's viewpoint), maps directly onto the critical comparison in the central cueing paradigm, i.e., ‘cued–uncued’. Although Samson et al. do include attentional cueing as a process that contributes to spontaneous perspective taking, a cueing effect could solely explain the basic effect. In support of their directional cueing hypothesis, Santiesteban et al. showed that a stimulus known to shift attention laterally (i.e., a centrally located arrow) induced consistency effects of comparable size to that of an avatar (see also Nielsen, Slade, Levy, & Holmes, 2015).

A problem however with the cueing hypothesis is that the perspective and cueing effects may operate independently but still generate a similar pattern of data. Thus, demonstrating that both arrows and avatars generate a consistency effect does not falsify the spontaneous perspective taking theory. As Firestone and Scholl (in press) have recently reminded us, “not only should you observe an effect when your theory calls for it, but you should also not observe an effect when your theory demands its absence”. The principal aim of the present work was to test the claim that perspective taking is indeed spontaneous, as argued by Samson et al. (2010), by setting up a scenario in which visual perspective taking should not occur. As with the original experiment of Samson et al., participants were presented with an avatar located in the middle of a display that looked either towards a left of right hand wall. Importantly, the ability of the avatar to see the stimuli that generate the basic perspective taking effect was manipulated by the positioning of physical barriers either side of the avatar. On ‘non-seeing’ trials these barriers fully occluded stimuli presented to the left or right whilst, on ‘seeing’ trials the barriers included window-like features allowing the stimuli to be seen by the avatar (see Fig. 1¹). The use of physical barriers to manipulate what an agent can see is common when assessing mentalising in non-human animals (e.g., Hare, Call, & Tomasello, 2001). Clearly, if the avatar's perspective is spontaneously taken, no Samson et al.-like effect should occur when the avatar is unable to see the inducing stimuli.