Abstract
Human description of the surrounding world may spontaneously rely on others’ perspective, which is a crucial component of social cognition. In five studies, participants were asked to describe the spatial relations between objects in visual scenes including, or not, other agents. In Experiment 1, a substantial proportion of participants used an other-centered perspective in the presence of another agent, replicating classical findings. To our own surprise, we also observed that an even greater number of participants used an other-centered perspective when the human agent was replaced by an armchair. In order to explore this phenomenon, Experiments 2 to 5 compared the respective strength of chair-centered and agent-centered perspectives and/or set them into conflict. A significant proportion of participants spontaneously took the seat’s perspective even when it may not be sat on (Experiments 3 and 4) and even when the seat was not referred to (Experiments 4 and 5). Altogether, these findings suggest that perspective taking may spontaneously apply to inanimate objects. These results question whether such tendencies originate from social cognitive skills—as classically assumed—or reveal a nonsocial phenomenon. Future works should specifically test the widely assumed social nature of spontaneous perspective-taking.
The structure of the surrounding physical world is experienced through the configuration of our body and typical interactions with space (Bryant et al., 1995). In addition to the role of our action capabilities, others’ actions have also been found to influence spatial representations (Coello et al., 2018). The impact of self and others’ physical dispositions seem however dependent of the frames of reference that is used while describing a visual scene (Bryant et al., 1992).
The use of different frame of reference, and particularly the distinction between egocentric and allocentric frames of reference, has a long tradition in neuroscience (e.g., Jeannerod & Rossetti, 1993; Paillard, 1991). Ego-centered perspectives are specific to each individual (e.g., X lies to the left of Y) whereas allo-centered perspectives are common to everyone (e.g., X lies to the north). Arguments for the coexistence of these two reference frames at the psychological level come from various behavioral (e.g., Blouin et al., 1993; Paillard et al., 1983; Rossetti et al., 1995) and neuronal approaches (e.g., Tamura et al., 1990). The development in the field of social cognition emphasized the important role of another type of reference. This third frame of reference (referred here as other-centered perspectives), anchored on another being, can be used to take others’ point of view into account (e.g., X lies to the right of Y from their point of view) and may serve as an important basis for social cognition. Examples of both ego-centered and other-centered intrusions in various tasks have been flourishing over the last decade. On the one hand, it has been shown that humans are easily affected by their own point of view when having to consider another one’s (e.g., Epley et al., 2004; Samson et al., 2010) and that taking another perspective has a cost (e.g., Ferguson et al., 2017; Michelon & Zacks, 2006). On the other hand, egocentric tasks may be affected by the mere presence of a partner with a different perspective (e.g., Freundlieb et al., 2016; Samson et al., 2010).
In an elegantly simple study, Tversky and Hard (2009) showed that humans spontaneously, although unnecessarily, take the perspective of another person. In their experiment, a picture was presented to participants, with a man sitting at a table with a bottle and a book, and participants were required to describe the relationship between the two objects on the table. It was found that about 30% of the participants spontaneously endorsed the perspective of the human in the picture instead of their own. In the absence of the agent in the picture, less than 5% of the participants produced an other-centered answer (or confounded left and right). This spontaneous tendency to adopt another agent’s perspective has been largely replicated (Conson et al., 2017; Furlanetto et al., 2013; Todd et al., 2011). At the psychological level, it is suggested that the occurrence of such a social phenomenon involves representing one’s own body as in the physical location of the other person (e.g., Gardner et al., 2017; Kessler & Rutherford, 2010; Kessler & Thomson, 2010; Michelon & Zacks, 2006; Surtees et al., 2013). By doing so, humans would be able to anticipate others’ behaviors, which would facilitate upcoming social interactions (Freundlieb et al., 2016). According to this view, spontaneous perspective-taking would reflect efficient social cognition skills (we refer to this view as the “social interpretation” in this article).
In a recent experiment inspired by Tversky and Hard (2009), we observed that as many as 30% of our participants used an other-centered reference frame to describe a visual scene with no agent (Quesque et al., 2018). One possible explanation of this difference is that the chair used by the agent remained visible in our “no-agent” condition, whereas it was not visible in the original study (Tversky & Hard, 2009). This interpretation supposes that not only human beings, but also inanimate objects may trigger visuo-spatial perspective-taking. If it was actually the case, this would seriously question the supposedly social aspect of spontaneous perspective-taking (e.g., Bertamini & Soranzo, 2018). Representing the surrounding world from an other-centered perspective could simply reflect specific environmental cues—which would afford oriented behaviors—and not rely on dispositional social cognition abilities (we refer to this view as the “action oriented interpretation” in this article). In this article, we provide, through five experiments, evidence supporting the hypothesis that visuo-spatial perspective taking can be triggered by inanimate objects alone. Further, this tendency can even interfere with the supposedly automatic tendency (Tversky & Hard, 2009) to represent the world from other agents’ perspective (Experiments 2, 3, and 5). Finally, evidence supports that such phenomenon manifests itself even in the absence of any reference to the inanimate object (Experiments 4 and 5), and when its behavioral affordance is clearly impeded (Experiments 3 and 4). Overall, the evidence reported here suggest that humans may spontaneously use other-centered perspectives in the absence of any social agent and instructions to do so, which calls for a reappraisal of the classic conception of perspective-taking as a marker of social cognition.
Experiment 1
Method
Participants
Two hundred and forty-seven French first-year medical students from the University of Lyon, France (mean age = 18.1, SD = 2.2, 169 females) volunteered in Experiment 1, at the end of a lesson. The sample size from this first experiment was based on participant availability. A posteriori sensitivity power analysis (1 – β = .80, α = .05, two-tailed), computed for this sample of 247 participants, allows us to meaningfully identify a Cramer’s V effect size of 0.19 (Faul et al., 2007). All studies in this article were then powered to detect small-to-medium sized effects (using common conventions for Cramer’s V) on the basis of this first experiment. Participants had no prior knowledge about the scientific aim of the study and gave their informed consent prior to their inclusion in the study. The protocols were in accordance with the principles of the Declaration of Helsinki (World Medical Organization, 1996).
Procedure and Materials
This study was initially presented as aiming to collect spontaneous descriptions of diverse visual scenes in order to develop a standardized basic cognition battery for patients. Participants responded to a questionnaire from which only four questions were of interest for the present experiment, the others being fillers (materials are available at: https://osf.io/zdsj7/). In these questions, participants were asked to either describe (e.g., “What is the color of the seat that lies to the left of the fire?” Q1) or localize an item relative to others while presented with visual scenes containing different objects (e.g., “In relation to the TV, where is the key?” Q2 & “In relation to the armchair, where is the ball?” Q3, see Figure 1, left) and potentially another agent (e.g., “In relation to the bottle, where is the book?” Q4, see Figure 1, left). The first scene (Q1) was a control for participants’ right/left discrimination (as no oriented object and no social agent were present). The last scene (Q4) was a replication of Tversky and Hard (2009), relying on a new stimulus for external validity purpose. Participants were free to answer with as many words as they wanted (data are available at: https://osf.io/zdsj7/). At the end of the test, they were informed about the actual aim of the study.

(Left) Visual scenes used in Experiment 1. (1) Three seats of different colors distributed around a wooden fire, (2) three different objects distributed around a TV monitor, (3) three different objects distributed around an armchair, (4) a human seating at a table on which are displayed a book and a bottle (as in Tversky & Hard, 2009). (Right) Proportions of perspective chosen as a function of the scenes used for Experiment 1. Ego-centered perspective referred to the participant’s own perspective, object- or other-centered perspective respectively to the perspective of the object or agent in the picture, and allo-centered perspective to the use of a neutral reference (e.g., table edge) (whiskers represent theoretical errors of binomial distributions).
Results and Discussion
Responses were coded as “Ego-centered” when unambiguously relying on the participants’ perspective to identify or localize the target object (e.g., “to the left” for Q4), as “Other-centered” when they were given from the object/agent’s orientation (e.g., “to the right” for Q4), and as “Neutral” when they did not refer to either perspectives (e.g., “at the same level” for Q4). The response of a single participant referred to both the participant’s and the agent’s perspectives and was recoded depending on the first of the two perspectives mentioned, following Tversky and Hard’s study (2009). Seven other participants had missing data or data which were not interpretable (e.g., making reference to a color not represented in the scene for Q1). Their data were removed from the subsequent analysis.
The distributions of proportions for each localization tasks (Q2, Q3, and Q4) were compared through homogeneity chi-square tests. The control task, relying on another question, was analyzed independently. To draw multiple comparisons, the significance criterion was corrected (α < .05/3). As it is shown in Figure 1 (right), participants predominantly relied on their own ego-centered perspective when describing the visual scenes. The relative contribution of other-centered frames of reference depended on the scene. Participants’ distributions among perspectives significantly differed between the armchair and the TV scenes—χ2(2) = 28.6, p < .001, V = 0.24—as well as between the armchair and the human agent tasks—χ2(2) = 8.43, p < . 05, V = 0.13—and between the TV and the human agent scenes—χ2(2) = 11.4, p < .01, V = 0.15. Specifically, participants tended to endorse an other-centered perspective 2.1 times more often when the scene contained an armchair (46.3%) than when the spatial judgment is made in reference to a TV (22.3%). They used an other-centered perspective more frequently when describing the scene containing the armchair than the one containing a human agent (31.3%). They also endorsed more often an other-centered perspective in the human agent task than in the TV task. Finally, in the control task in which spatial judgments are relative to a not oriented wood fire, participants massively relied on an ego-centered perspective to produce their responses (93.4%, similar to Tversky and Hard’s control task).
Altogether, the present results reveal a human disposition to consider a non-ego-centered perspective rather than their own perspective when describing a visual scene, even in the absence of any other living agent (for congruent results on reaction time tasks, see Cavallo et al., 2017). The proportion of other-centered responses obtained in the presence of another agent (31.3%, Q4) as well as in the absence of any oriented object (6.6%, Q1) are highly congruent with the results reported by Tversky and Hard (2009) respectively in the presence and absence of the agent, suggesting that the original findings (photographs) can be generalized to less ecological conditions such as our simple line-drawings (see also Quesque et al., 2020). In accordance with the “action-oriented interpretation,” the present results show that the mere presence of an object can trigger the use of other-centered perspective that relies on its functional orientation (see Levelt, 1996). It seems that some objects—which may elicit strong bodily oriented affordances (e.g., the armchair)—may lead even more systematically to the use of object-centered perspectives than others (Osiurak et al., 2017). Surprisingly, the human tendency to adopt the perspective of objects (Q3), reported here, seems more important than the tendency to adopt the one of other agents (Q4, but see also Todd et al., 2011; Tversky & Hard, 2009). In Experiment 1, oriented objects and human agents were however not presented simultaneously. In Experiment 2, we investigated frames of reference chosen in the presence of both a human agent and an oriented object. Experiment 2 addressed this question by introducing a conflict between incongruently oriented human agent’s and object’s perspectives.
Experiment 2
Method
Participants
Two hundred and thirty-seven French first-year medical students from the University of Lyon, France (mean age = 18.2, SD = 2.2, 159 females), who did not participate in Experiment 1, voluntarily took part in Experiment 2. A posteriori sensitivity power analysis (1–β = .80, α = .05, two-tailed) allows us to meaningfully identify a Cramer’s V effect size of 0.18. Participants had no prior knowledge about the scientific aim of the study and gave their informed consent prior to their inclusion in the study.
Procedure and Materials
The procedure was the same as for Experiment 1. At the end of a lesson, participants responded to a short questionnaire from which only two questions were of interest for the present experiment. In these questions, participants were asked to localize an item (e.g., the ball) relative to others while presented with visual scenes containing simultaneously different objects and a human agent (see Figure 2, left). In order to investigate the impact of the objects’ orientation, the spatial judgment task used a reference object which was an armchair (oriented) in one question, while it was a stool (not oriented, but same behavior afforded) in the other question.

(Left) Visual scenes used in Experiment 2. (1) A human agent facing three different objects distributed around a stool. (2) A human agent facing three different objects distributed around an armchair. (Right) Proportion of perspective chosen as a function of the scenes used for Experiment 2. Ego-centered perspective referred to the participant’s own perspective, Other-centered perspective to the perspective of the agent in the picture, and Seat-centered perspective to the perspective of the seat in the picture (whiskers represent theoretical errors of binomial distributions).
Results and Discussion
Responses were recoded as “Ego-centered” when participants used their own perspective (e.g., “on top” for Scene 2), as “Other-centered” when they used the agent’s perspective (e.g., “to the right” for Scene 2), or as “Object-centered” when they used the perspective of the object (e.g., “to the left” for Scene 2). Twenty participants produced responses which were not interpretable (e.g., making reference to a location that is not congruent with any of the three perspectives) at least for one question and their data were then discarded from the subsequent analysis. As shown in Figure 2 (right), participants’ distributions among perspectives significantly differed between the armchair and the stool scenes—χ2(2) = 37, p < .001, V = 0.28. Specifically, when facing the stool scene, most of the participants (60.8%) used a non-ego-centered perspective. Around one half of these participants relied on the agent’s perspective (30.1%) while the other half described the scene as if they were seated on the stool (29.7%), that is, facing the agent. The remaining 40.2% of participants used an ego-centered perspective. Concerning the armchair scene, the vast majority of participants (82.8%) used a non-ego-centered perspective. In this case, however, participants took the perspective of the armchair nearly 3 times more often (61.2%) than they took the agent’s perspective (21.6%). Only the remaining 17.2% of participants used an ego-centered perspective. Confirming Experiment 1, the present results showed that contextual factors may lead humans to endorse a non-ego-centered perspective when describing visual scenes. Experiment 2 further extends this result by showing that the tendency to use object-centered perspectives seems to be triggered by at least two characteristics of the objects: both by the type of behavior that they would afford (as shown in Experiment 1) and by their intrinsic orientation (as shown in Experiment 2). Experiments 1 and 2 reported several measures for the same participants. In order to determine how much participants were consistent in their responses (i.e., display a general tendency to consider an ego- or other-centered perspective), we used the intraclass correlation coefficient (ICC) on a linear model with mixed-effect. An high ICC (near 1) means there is a high similitude between responses of the same participants whereas a low ICC (between [0, 0.5]) means the responses of the same participants are different (Koo & Li, 2016). For Experiments 1 and 2, respectively, we obtained an ICC of 0.25 with a 95% confidence of [0.05, 0.45] and 0.40 with a 95% confidence interval of [0.16, 0.55]. According to Koo and Li (2016), these results sustain the idea that respondent are not so consistent in their responses. Finally, it appears that the tendency to consider a non-ego-centered perspective may be even stronger for the armchair than for human agents. One potential explanation for this result is the use of a top view in Experiments 1 and 2, which may have interfered with the natural use of ego-centered perspectives. Accordingly, we designed more natural stimuli in the following experiments. Furthermore, in Experiment 2, the distances between the reference object (i.e., the seats) and the target object (i.e., the ball) vary across condition might have influence the results as a potential confound. In Experiment 3, we controlled this aspect and specifically aimed to test the influence of the affordances (i.e., seating behavior) evoked by the objects.
Experiment 3
Method
Participants
Three hundred and eighty-eight French first-year medical students from the University of Lyon, France (mean age = 18.65, SD = 1.97, 271 females), who did not participate in previous experiments, voluntarily took part in Experiment 3. The total number of participants was almost doubled compared with previous experiments as the two experimental conditions were operationalized with a between-subject design. A posteriori sensitivity power analysis (1 – β = .80, α = .05, two-tailed) allows us to meaningfully identify a Cramer’s V effect size of 0.14. Participants had no prior knowledge about the scientific aim of the study and gave their informed consent prior to their inclusion in the study.
Procedure and Materials
The procedure was the same as for Experiment 1, but this time we reduced the number of questions of interest to prevent any contamination between targeted questions. Participants responded to a questionnaire from which only one question was of interest for the present experiment. They were asked to localize an item relative to others while presented with visual scenes containing simultaneously different objects and a human agent (e.g., “In relation to the seat, where is the ball?”, see Figure 3 left). Contrasting with previous experiments, the reference object was a chair and the scene was no longer displayed from a top viewpoint. In order to test for an affordance effect, the chair displayed was either broken or not. Participants were randomly assigned to one of the two conditions. Once participants responded to the critical question, we controlled for a potential influence of saliency and affordance by asking them to judge the extent the seat attracted their attention and the extent they would like to use that chair, on a 7-point Likert scale (going from 1= not at all to 7 = absolutely) while looking at the picture.

(Left) Visual scenes used in Experiment 3. (Top) A human agent facing three different objects distributed around a chair. (Bottom) A human agent facing three different objects distributed around a broken chair. (Right) Proportion of perspective chosen by participants to describe the scene used for Experiment 3. Ego-centered perspective referred to the participant’s own perspective, Other-centered perspective to the perspective of the agent in the picture, and Seat-centered perspective to the perspective of the seat in the picture (whiskers represent theoretical errors of binomial distributions).
Results and Discussion
As in Experiment 2, responses were recoded as “Ego-centered” when participants used their own perspective (e.g., “in front”), as “Other-centered” when they used the agent’s perspective (e.g., “to the right”), as “Object-centered” when they used the perspective of the object (e.g., “to the left”), and as “Allo-centered” when they did not refer to a particular perspective in their responses (e.g., “near to”). As it is shown in Figure 3 (right), participants’ distributions among perspectives significantly differed between the chair and the broken-chair scenes—χ2(2) = 16.3, p < .001, V = 0.21. Specifically, when facing the “chair scene,” most of the participants (90.4%) used a non-ego-centered perspective. The majority of these participants relied on the seat’s perspective (63.8%), while the others took either the agent’s perspective (only 6.9%) or an allo-centered perspective (19.7%). Only the remaining 9.6% used an ego-centered perspective. Concerning the “broken chair scene,” again, most of the participants (90.0%) used a non-ego-centered perspective. The majority of these participants relied on the seat’s perspective (62.5%), while the other took either the agent’s perspective (18%) or an allo-centered perspective (9.5%). Finally, 10% used an ego-centered perspective. Overall, despite the use of a more natural setting in Experiment 3, it appeared that most of the participants still used another perspective that their own when describing these visual scenes in both conditions. Two-by-two comparisons revealed that participants significantly (Z = 3.29, p < .001) tended to endorse the agent’s perspective more often in the broken-chair than in the normal chair condition. No differences were however observed between conditions concerning the tendency to consider the seat’s perspective (Z = 0.27, ns). This absence of differences suggests that the seat-centered perspectives spontaneously used by participants in Experiments 1 and 2 were not triggered by any seating affordances. For validity purposes, we checked that the broken chair (m = 2.2, SD = 1.83) was indeed preserved as less affording than the control chair (m = 3.97, SD = 2.1, t(341)=8.2, p < .001). In order to further explore the influence of the chair saliency or of its perceived affordance on participants' tendency to use the seat’s perspective, we conducted a cross-conditions multiple logistic regression of the probability of taking the chair's perspective with saliency and perceived affordance scores as predictors. Congruently with the results of the between condition comparison, the perceived affordance did not significantly contribute to the model (Coef = –0.005, Z= –0.04, ns). This was also not the case concerning the saliency of the chair (Coef = 0.02, Z= 0.22, ns), or interaction between the two variables (Coef = –0.01, Z= –0.39, ns). Through study 3, we replicated the findings observed in Experiments 1 and 2 (i.e., a tendency to adopt an object’s perspective when describing a visual scene). The present results also suggest that such tendency to endorse object-centered perspective is not simply linked to the object’s affordance or saliency. This claim is supported by both a regression on participants explicit judgments and by between condition comparisons in which we observed that varying the seating affordance of the seat did not affect the tendency to adopt its perspective. An alternative factor may contribute to the tendency reported here. Because we explicitly referred to the seat in the instructions, it may be the case that our instruction set incited participants to adopt its perspective. This remark would however generalize to previous works in which references to different oriented objects were systematic (e.g., Conson et al., 2017; Mazzarella et al. 2012; Todd et al., 2011; Tversky & Hard, 2009). In order to control for this explanation, reference to the seat was fully removed from the instruction set of the following experiments.
Experiment 4
Method
Participants
One hundred ninety-one French first-year medical students from the University of Lyon, France (mean age= 21.2, SD = 2.4, 119 females), who did not participate in previous experiments of this article, voluntarily took part in Experiment 4. A posteriori sensitivity power analysis (1–β = .80, α = .05, two-tailed) allows us to meaningfully identify a Cramer’s V effect size of 0.20. Participants had no prior knowledge about the scientific aim of the study and gave their informed consent prior to their inclusion in the study.
Procedure and Materials
As in the last two experiments, participants responded to a questionnaire from which only one question was of interest for the present experiment. Despite its presence on the picture (see Figure 4, left), the seat was no longer mentioned as participants were asked to respond to the following question “In relation to the remote-controller, where is the bottle?”. In addition, as an additional attempt to degrade the seat affordance for sitting, a big flower pot (also not relevant for the question) was placed on it.

(Left) Visual scene used in Experiment 4. The scene displays several objects (including an armchair) on a carpet. (Right) Proportion of perspective chosen by participants to describe the scene used for Experiment 5. Ego-centered perspective referred to the participant’s own perspective, Seat-centered perspective to the perspective of the seat in the picture, and Allo-centered to the use of a neutral reference (e.g., “On the same carpet”) (whiskers represent theoretical errors of binomial distributions).
Results and Discussion
Responses were encoded using the same procedure as in previous experiments of this report. Three participants had missing data and were consequently removed from the subsequent analysis. As it can be seen in Figure 4 (right), participants’ distribution among perspectives significantly differs from the flat theoretical distribution—χ2(1)=18.4, p < .001, V = 0.31. Strikingly, most participants (47.9%) used the seat perspective to describe the spatial relation between the objects in the scene; 28.2% of participants used an allo-centered perspective. Finally, 23.9% of participants used an ego-centered perspective. Overall, the present findings show that even when removing any reference to the seat, and when it cannot be sat on, participants still primarily take its perspective, neglecting their own visuo-spatial perspective. This result is particularly important as all previous work relying on a similar paradigm explicitly mentioned the targeted objects in the instructions (e.g., Conson et al., 2017; Mazzarella et al. 2012; Todd et al., 2011; Tversky & Hard, 2009). Quesque et al. (2018) suggested that humans may spontaneously take the perspective of another person even when it is not relevant. In Experiments 2 and 3, we observed that this tendency was largely reduced when the scene also contained an empty seat. In these experiments, the seat was mentioned in the instruction set. In Experiment 5, we aimed to test if participants would still take the perspective of the seat in presence of another person and when the seat is not mentioned in the instructions.
Experiment 5
Method
Participants
One hundred sixty-eight French first-year medical students from the University of Lyon, France (mean age = 19.9, SD = 1.9, 110 females), who did not participate in previous experiments of this article, voluntarily took part in Experiment 5. A posteriori sensitivity power analysis (1–β = .80, α = .05, two-tailed) allows us to meaningfully identify a Cramer’s V effect size of 0.22. Participants had no prior knowledge about the scientific aim of the study and gave their informed consent prior to their inclusion in the study.
Procedure and Materials
As in the last two experiments, participants responded to a questionnaire from which only one question was of interest for the present experiment. Despite its presence on the picture (see Figure 5, left), the seat was no longer mentioned as participants were asked to respond to the following question “In relation to the key, where is the ball?” Contrasting with Experiment 4, the scene now included a human agent.

(Left) Visual scene used in Experiment 5. A human agent facing three different objects distributed between her and a chair. (Right) Proportion of perspective chosen by participants to describe the scene used for Experiment 5. Ego-centered perspective referred to the participant’s own perspective, Other-centered perspective to the perspective of the agent in the picture, Seat-centered perspective to the perspective of the seat in the picture, and Allo-centered to the use of a neutral reference (e.g., “At 30 cm”) (whiskers represent theoretical errors of binomial distributions).
Results and Discussion
Responses were recoded using the same procedure as in previous experiments. Two participants had missing data and were consequently removed from the subsequent analysis. As it is shown in Figure 5 (right), when asked to characterize the spatial relationship between the objects, participants’ distribution among perspectives significantly differs from the flat theoretical distribution—χ2(1)=81.5, p > .001, V = 0.70. Once again, the majority of participants (56.6%) of participants used a non-ego-centered perspective when describing the scene. Contrasting with previous results, most of these participants took the perspective of the human agent (41.6%) rather than the seat’s perspective (9.0%). The other participants used an allo-centered perspective (6.0%). Finally, 43.4% of participants used an ego-centered perspective. In the present experiment, we for the first time observed that the majority of participants took either the perspective of the human agent or rely on their own point of view, which directly reduced the number of participants who used the empty seat’s perspective. It remains notable that even when not mentioning the seat, and in spite of the presence of another human agent, a significant proportion of participants still spontaneously used the seat’s perspective (9%). A supplementary, post review, analysis based on the five experiments has been conducted to test whether gender influenced the tendency to adopt an other-centered perspective. (The authors wish to thank an anonymous reviewer for this suggestion.) We found no significant difference between male and female participants—χ2(1)=0.02, p = .87.
General Discussion
Humans sometimes use other-centered reference frames. We previously found that such spontaneous perspective-taking tendencies may occur in the absence of any social agent (Quesque et al., 2018). In the present study, therefore, we investigated whether the supposedly automatic tendency to adopt other-centered perspectives might simply reflect other explanatory factors (e.g., the “action-oriented interpretation”) than dispositional social cognition abilities, as it is classically conceived (e.g., the “social interpretation”). Participants were asked to describe the spatial relationships between objects presented in an image, in the presence or absence of other human agents. Previous findings (Quesque et al., 2018; Todd et al., 2011; Tversky & Hard, 2009) showed that humans may spontaneously use a non-ego-centered perspective when having to describe the surrounding world in the presence of other agents (e.g., by asking to a person—that is facing us—to take the book in the right drawer although it is presently located to our left). Our series of experiments shows that seats are endowed with the capacity to invite humans to take their perspective, that this does not seem to be modulated by varying the seating affordance of the seat, that removing any reference to the seat does not reduce this spontaneous tendency to adopt the seat perspective, and that even in the presence of a triple choice of perspectives, about 10% of the participants chose the perspective of the seat. The following discussion will address these points.
First, seats are endowed with the capacity to invite humans to take their perspective. As compared to other inanimate objects such as a TV set, the seat implicitly invited participants to take its perspective instead of their own (respectively 22.3 vs. 46.3%, in Experiment 1). This proportion is similar to the proportion described when participants faced another human (Quesque et al., 2018; Tversky & Hard, 2009). In Experiments 1 and 2, comparisons of the seat-centered and agent-centered perspective taking were both favorable to the seat over the human character. One may wonder why this may happen when the question to be answered simply asks for locating a target object. It was already remarkable that participants may take the perspective of others even when this was unnecessary (Quesque et al., 2018; Tversky & Hard, 2009), and the present results appears even more puzzling. The obvious explanation that perspective-taking spontaneously emerge in the presence of a congener in order to facilitate social interactions holds for the previously published observations, but obviously cannot be responsible for taking the perspective of an empty seat. We shall come back to this issue later.
Second, varying the seating affordance of the seat does not seem to affect this implicit tendency to adopt the seat perspective. Whether the chair was broken (62.5%, in Experiment 3) or a large flower pot was placed on it (47.9%, in Experiment 4), the seat-centered perspective remained the most frequently adopted. The present results may seem contradictory with the classical notion that action affordances increase the probability of engagement in perspective-taking (Conson et al., 2017; Mazzarella et al. 2012; Tversky & Hard, 2009). The tendency to use other-centered perspectives when interpreting a visual scene seems to be influenced by the presence of other agents, as classically reported (Quesque et al., 2018; Todd et al., 2011; Tversky & Hard, 2009), but also to be under the influence of the objects present in the scene (Experiments 1 and 2). Interestingly, results from Experiment 3 suggest that neither the perceived affordance nor the objects saliency are linked to this tendency to use an other-centered perspective.
Third, Experiment 4 suggests that removing any reference to the seat does not cancel this spontaneous tendency to adopt the seat perspective.
Fourth, in Experiment 5 about 10% of the participants chose the perspective of the seat in the presence of a triple choice of perspectives (ego-centered, agent-centered, and seat-centered). Interestingly, when not mentioning the seat, and in the presence of both another human agent and a seat, participants were more inclined to adopt the agent’s than the seat’s perspective. A significant proportion of participant still spontaneously took the perspective of the seat, neglecting the agent perspective but also their own.
Altogether, the present findings confirm this human disposition to spontaneously adopt non-ego-centered perspective and additionally demonstrate that this kind of perspective-taking can also be triggered by inanimate objects. This findings are congruent with previous results showing that humans project their own perceptual dispositions (e.g., having a large visual field) to inanimate objects such as mirrors (which actually only reflect items that are presented directly in front of them, see Bertamini et al., 2003). Strikingly, the present results suggest that humans may be even more inclined to use an other’s perspective to describe a visual scene in the presence of an oriented object than in the presence of another human agent. This result calls for more caution in the interpretation of results showing spontaneous perspective-taking as arguments for automatic social cognition. While previous authors proposed that visuo-spatial perspective-taking may represent one of the most basic forms of mental states inference (e.g., Erle & Topolinski, 2017), the present results question the very social nature of such process. It is possible that both the social and the action-oriented interpretation support human’s spontaneous visuo-spatial perspective-taking tendencies. However, we believe that the social interpretation, by itself, would fail to account for the present results as in some conditions participants privileged object’s over human’s perspectives. This new view remains compatible with the idea that the natural use of non-ego-centered perspectives may still find its roots in social interactions (Surtees et al., 2012), in which the use of a spatial reference shared by all may facilitate communication among agents (Levelt, 1996).
Interindividual variability has not been investigated directly in the present study and why some participants spontaneously represent other-centered perspective, while others do not, remains an open question. According to the social interpretation, spontaneous visuo-spatial perspective-taking would constitutes a basic aspect of more general social cognition skills. Concerning other aspects of social cognition, female participants are known to have better socio-emotional decoding skills than male participants, even if meta-analyses show that this advantage is rather limited (Hall et al., 2000; Kirkland et al., 2013; Kret & de Gelder, 2012). Through post-hoc comparisons, we tested whether a potential gender effect exists concerning spontaneous visuo-spatial perspective-taking and failed to find a significant difference between male and female participants. This result argues against the social interpretation of spontaneous visuo-spatial perspective-taking. Also supporting this conclusion, in Experiments 1 and 2 we found that participants displayed a relatively low consistency in the perspectives spontaneously adopted across conditions, which goes against the idea of a personal disposition to endorse other-centered visuo-spatial perspective.
At the methodological level, our results suggest that contextual elements, such as the background features of a visual scene, can deeply influence the way participants will spontaneously represent and process visual information (see also Bryant & Tversky, 1999, who demonstrated that the type of stimuli influenced the perspective adopted by participants). This finding underlines an unsuspected aspect of the necessity to no longer conduct experiments relying on a single stimulus in order to foster the external validity of psychological science (see Judd et al., 2012, 2014; Westfall et al., 2015). Daily life interactions take place in extremely rich contexts (as opposed to experimental settings), and the present study demonstrates that the mere presence of objects can, as it was known for the presence of other humans (e.g., Coello et al., 2018; Freundlieb et al., 2016), influence participants perceptions.
Further lines of research will need to investigate the nature of this tendency to spontaneously adopt the perspective of inanimate objects. It remains however intriguing and significant that people may spontaneously favor the perspective of an object over both the one of another human agents and their own. The present work suggests that a great caution should be exercised when interpreting spontaneous perspective-taking when describing visual scenes as a marker of social cognition.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the CNRS, the INSERM, the University of Lyon and the Hospices Civils de Lyon.
