Space and rank: infants expect agents in higher position to be socially dominant

Abstract

Social hierarchies exist throughout the animal kingdom, including among humans. Our daily interactions inevitably reflect social dominance relationships between individuals. How do we mentally represent such concepts? Studies show that social dominance is represented as vertical space (i.e. high = dominant) by adults and preschool children, suggesting a space-dominance representational link in social cognition. However, little is known about its early development. Here, we present experimental evidence that 12- to 16-month-old infants expect agents presented in a higher spatial position to be more socially dominant than agents in a lower spatial position. After infants repeatedly watched the higher and lower agents being presented simultaneously, they looked longer at the screen when the lower agent subsequently outcompeted the higher agent in securing a reward object, suggesting that this outcome violated their higher-is-dominant expectation. We first manipulated agents' positions by presenting them on a podium (experiment 1). Then we presented the agents on a double-decker stand to make their spatial positions directly above or below each other (experiment 2), and we replicated the results (experiment 3). This research demonstrates that infants expect spatially higher-positioned agents to be socially dominant, suggesting deep roots of the space-dominance link in ontogeny.

1. Introduction

Social status hierarchy—individuals being ranked above or below other group members—is essential to the stability of a social system [1–8]. Evolved hierarchical organizations in human society are supported by underlying cognitive processes in individuals; studies have shown that people conventionally use various cues (e.g. facial configuration and expressions [9–11], brief social utterances [12]) to recognize whether a person has control and influence over the surroundings and others—social dominance. However, little is known about how we mentally represent abstract concepts of dominance.

Anthropological studies and experimental evidence have shown that power is mentally represented as vertical space: high-ranked individuals are in higher spatial positions than low-ranked individuals [13,14]. For instance, social dominance hierarchies are ubiquitously represented as pyramid-like structures [15,16], and people daily use linguistic expressions such as ‘high’ and ‘low-status’; ‘control is up’, ‘lack of control is down’ [17–19]. As young as 5 years old, children develop adult-like space-dominance associative representation: they judge a person with tilted head up as more powerful (e.g. ‘in charge’) than the one with head down [20]; they respond faster to powerless words with the down key than with the up key and to powerful words with the up key than with the down key [14,19]. These suggest that mental representations of social dominance concepts are associated with their perceptual recognition of vertically spatial orientation [14,21–24].

The current study aimed to explore the developmental origins of the space-dominance link and to further our understanding of the roles that social learning and linguistic experience [19,23,25] play for the acquisition of the associations. These may also have implications for the extent of an evolutionary preparedness [16,26]. We conducted three experiments to test whether 1-year-old infants, who are prelinguistic and have little social experience, associate spatial orientation with social dominance, that is, whether infants attribute social dominance to agents who were presented in a higher spatial position, compared to agents presented in a lower position. Although infants at this age have already developed cognitive capacities for both spatial orientation [27–29] and social dominance [30–32], to our knowledge no previous research has investigated their relationship.

In experiments, infants watched three-dimensional animated events in which two agents first were presented simultaneously in different spatial positions, one higher and one lower, then the same agents contended for one reward object in a zero-sum situation [33–35]. Past studies have shown that infants may regard animated geometric figures as social agents. Indeed, infants not only attribute mental states and processes such as goal-directed intentions [36] and preference [33] to them but also believe that they are capable of demonstrating specific types of social behaviours such as cooperation [34] and interference [35] during interactions. More importantly, infants draw inferences about hierarchical relationships from cues such as relative body and numerical group sizes [30–32]. In the present study, we specifically defined dominance as ‘the tendency to prevail when one's goals conflict with those of another agent’ [32] and considered an agent that successfully secured limited resources to be socially dominant; such a conceptualization of social dominance overlaps with the notion of power [37–39]. If infants attribute social dominance to the higher agent, they should expect the higher agent to outcompete the lower agent in securing the reward. Based on the Violation-of-Expectation paradigm [40–42] which relies on the duration of gaze as an index of surprise (i.e. infants look longer at the scene with unexpected events), the prediction is made that infants should express surprise by looking at the screen longer when they see that the lower agent obtained the reward, not the higher agent, compared to the inverse outcome.

We manipulated the agents' spatial positions by presenting them on a platform with two different side-by-side levels (rather than against a background with no floor) to increase the real-world familiarity of the stimuli [33,43,44]. For experiment 1, we used a podium, as it is known that adults infer social dominance from elevation (i.e. a taller stand indicates a more socially dominant position than a lower one; [23,24]). We conducted experiment 2 with the agents presented on a double-decker stand so that their spatial positions were directly above or below each other, to remove the possible noise of the asymmetrical podium-like set-up and to test the space-dominance association more directly. We also replicated experiment 2 in an altered experimental circumstance (experiment 3). The video stimuli were identical across all experiments.

2. Experiment 1

(a). Method

(i). Ethics statement

In all experiments, participants were recruited from Infant Scientist of Kyushu University. In accordance with the Declaration of Helsinki, written informed consent was obtained from the children's carers before the experiment was conducted. The procedure was approved by the ethics committee of the Faculty of Human-Environment Studies at Kyushu University (2017-012).

(ii). Participants

Eighteen 12- to 16-month-old infants participated in experiment 1 (eight girls; mean age = 446 days, s.d. = 38.72, range = 374–508 days). The sample size was determined before the experiment, based on the previous studies on early recognition of social dominance [31,32]. Twelve additional infants were tested but were excluded from the final sample (the reasons that underlaid the exclusion of these participants from the final sample are presented in the electronic supplementary material).

(iii). Set-up

The experiment was completed in an open booth with three draped walls in a quiet room, located at the Hospital Campus of Kyushu University. The infants were seated on their carers' laps, approximately 60 cm from a monitor (23-inch TFT, 300 Hz, 1280 × 720 pixels) on which the experimental stimuli were presented. The monitor had a built-in remote Tobii TX300 eye-tracking system (Tobii Technology, Danderyd, Sweden). Outside the booth, one experimenter (experimenter A) controlled the calibration, presentation of stimuli and recording of the infants' eye movements, using Tobii Studio 3.2.2. Two desktop speakers, which were connected to the computer and hidden from view behind the monitor, transmitted electronic sounds to maintain the infants' attention towards the stimuli, as mentioned in the stimuli section.

(iv). Stimuli and procedure

All animated events were created in Blender (2.79b, Blender Foundation) and synchronized with custom audio tracks in Final Cut Pro X (figure 1). The animated world had a black background with the stages coloured by a checkerboard with white and purple. Four agents in different shapes but the same volume, with eyes and nose and an approximate height of 5 cm on the screen, were used. Two agent pairs in different colour and shape were prepared—red and blue agents as a pair, and green and orange agents as another pair (electronic supplementary material, Stimuli pattern). A grey podium was used to manipulate the spatial position of the agents.

Figure 1.

Sample of the final frames of spatial position presentation and test events used in experiments. (Online version in colour.)

After a diamond attention getter (3 s), each infant watched a sequence of videos as follows:

• (i)introduction: a pair of agents appeared on the two sides of the stage and moved out simultaneously saying ‘Hmm!’ (4.6 s; electronic supplementary material, movie S1);
• (ii)spatial position presentation (SPP): a podium appeared on the middle of the stage. After 1.6 s, two agents appeared on the podium, with one on the high and another on the low platform. Drum sounds accompanied the visual presentation to get the infants' attention (6 s; electronic supplementary material, movies S2 and S3);
• (iii)single agent collection (SAC): a goal object fell down on the middle of the stage. Then, one agent appeared from the left/right side of the screen, displaced the object from the centre and moved out of the screen from the same side. Sounds for getting infants' attention were used (5.3 s; electronic supplementary material, movies S4 and S5);
• (iv)SPP: identical to (ii);
• (v)SAC: identical to (iii), except that another agent collected the goal object alone, from another side;
• (vi)SPP: identical to (ii);
• (vii)first test: a goal object fell down on the middle of the stage. Two agents appeared simultaneously, went towards the goal object and stopped. Then, one agent took the object back to the side it came from, and the other agent simply moved back to its side without the object (9 s). From this point, the movie froze [30–32] (electronic supplementary material, movies S6 and S7);
• (viii)SPP: identical to (ii); and
• (ix)Second test: identical to (vii), except that the other agent took the goal object.

We did not have any theoretical or empirical evidence to predict the effect of secondary factors (e.g. colour, appearance order, win first in the tests) of the stimuli on social dominance inferences. However, to minimize such possible noise, we created 16 stimuli patterns and randomly assigned one to each participant, ensuring that the patterns are experienced fairly by the participants (electronic supplementary material, Stimuli pattern).

Before the experiment, the infants, carers and experimenters played with toys that were unrelated to the experimental stimuli, in a corner of the same room as the experimental booth. Next, the experimenters briefly introduced the experiment, and the parents signed the informed consent. Then, each parent was seated in front of the screen with the infant on her/his lap. After a 5-point calibration, the infants watched one video sequence. Parents were instructed not to initiate interactions with the infant and that keeping their eyes closed was desirable for reducing the possible experimental noise. Tests (videos of (vii) and (ix) above) were ended if infants looked away from the screen for greater than 2 s, or after 60 s elapsed, after the animation froze [42].

(v). Coding and analysis

Experimenters B and A undertook online and offline coding, respectively (for more details, see the electronic supplementary material, Coding and analysis). They were required to code whether infants had looked towards the screen after the animation had frozen, but they did not possess prior knowledge about which agent was presented in a high or low position [42]. Offline-coded data were analysed. Gaze duration in the SPP and SAC phases were calculated by Tobii Studio [45–47].

Our main question was whether infants looked longer at the frozen screen after the lower agent retrieved the goal object, compared to the scene in which the higher agent took the object. A paired sample t-test on infants' gaze duration was conducted.¹ In addition, to eliminate the possible effects of secondary factors on infants' gaze durations, we used generalized linear mixed models (GLMMs) with a Gaussian distribution and identity link functions and included the following in the model: type of tests (high-position agent dominant: the agent that was presented in a higher position procured the goal object; low-position agent dominant: the agent that was presented in a lower position procured the goal object), test order (i.e. the order of presentation of the two types of tests that infants viewed), stimuli pattern, and the age (in days) and gender of the infants as fixed effects and the identities of the subjects as random intercepts (i.e. subject identity (ID)). The primary dependent measure was log-transformed (log-normal distribution) gaze duration [52].

To rule out the possibility that infants' gaze duration in the tests was influenced by that in the SPP or SAC phases, we also tested whether there was significant difference between (i) gaze duration towards areas of interest of the two agents in SPP phases, or between (ii) gaze duration at the screen when the two agents collected the goal object in SAC phases, and whether these differences influenced the gaze duration.

(b). Results and discussion

In the test phases, infants looked at the screen for mean (M) = 10.33 s (s.d. = 9.648) after the video froze at the point when the higher agent had taken the goal object, and for M = 16.80 s (s.d. = 18.207) when the lower agent had taken the object (figure 2). The duration of gaze significantly differed between the groups (t₁₇ = 3.08, p = 0.007, Cohen's d = 0.725). Planned GLMM confirmed this difference (β = 0.593, t₁₇ = 3.08, p = 0.007) after controlling other factors (e.g. test order; electronic supplementary material, table S1).

Figure 2.

Boxplots of infants' gaze duration(s) towards the screen after the high or low agent won the competition in securing the goal object during tests in experiments 1–3. High and low indicate tests in which high and low agents won, respectively. White diamonds indicate means. Horizontal lines indicate medians, boxes indicate middle quartiles and whiskers indicate points within 1.5 times the interquartile range from the upper and lower edges of the middle quartiles. Light grey circles connected across boxes indicate gaze durations from individual participants.

In SPP phases, during the 4.4 s presentations of the agents, infants did not show different gaze duration towards the high and low agents (M_high = 1.79 s, s.d. = 0.561, M_low = 1.90 s, s.d. = 0.648; t₁₇ = 0.53, p = 0.601, Cohen's d = 0.125). Also, infants did not show different gaze durations in two types of SAC phases (M_high = 6.12 s, s.d. = 0.335, M_low = 5.71 s, s.d. = 1.246; t₁₇ = 1.22, p = 0.238, Cohen's d = 0.288). Further, we used GLMM to examine the effect of the type of tests by controlling differentiated gaze durations between the two agents (which were calculated by subtracting the durations of gazes that were directed towards the low-position agent from the durations of gazes that were directed towards the high-position agent) in the SPP and SAC phases and their interaction, including them as fixed factors. We controlled the differentiated gaze duration to eliminate the possible effects of infants' gazing behaviours in the SPP or SAC phase on gaze duration in the test phase (e.g. it is possible that infants expected the agent they had viewed for a longer duration to be socially dominant). Subjects' identities were included as random intercepts (i.e. subject ID). The results showed that the type of the tests (infants looked longer after they saw the lower agent had secured the reward object than when the higher agent did; β = 0.594, p = 0.007) but not gaze durations in the SPP (β = −0.445, p = 0.392) or SAC phase (β = −0.455, p = 0.511) or their interaction (β = 2.631, p = 0.761) had an effect on gaze durations in the test phase (electronic supplementary material, table S2).

In this experiment, infants looked at the screen longer when they saw that the lower agent obtained the reward over the higher agent, compared to the inverse outcome. This suggests that infants not only distinguished the test outcomes based on the properties of the agents—whether they appeared in a higher or lower position—but also expected the higher one to be more socially dominant in securing the reward. Other factors (order of stimuli presentations, stimuli pattern, infants' age and gender) did not influence infants' gaze behaviour. Further, the longer gaze duration was not owing to previous experience in looking behaviour during the SPP and SAC phases.

However, as young children may experience situations in which socially dominant/submissive individuals are ranged in a podium-like formation (i.e. parents are sitting on higher chairs and young children are sitting on lower chairs), the observed space-dominance inference is possibly formed by such daily experiences, and therefore it is limited to a podium-like set-up rather than vertically spatial presentations in general. Thus, we conducted experiment 2 to generalize the observed space-dominance inference with stimuli presentations including more purely spatial position manipulation (i.e. agents' spatial positions were directly above or below each other), and to conceptually replicate the findings of experiment 1.

3. Experiment 2

Experiment 2 examined infants' dominance evaluation of the higher and lower agents with a more direct set-up. We manipulated the spatial positions so that the agents were presented on a double-decker stand, with their spatial positions directly above or below each other. Based on the results of experiment 1, we hypothesized that infants show longer gaze duration after they saw the lower agent obtaining the reward instead of the higher agent, than when they saw the inverse outcome.

(a). Method

(i). Participants

Eighteen 12- to 16-month-old infants participated in experiment 2 (nine girls; mean age = 445 days, s.d. = 46.52, range = 366–516 days). Seven additional infants were tested but were excluded from the final sample (electronic supplementary material, Participants).

(ii). Set-up

Experiment 2 was conducted in the same physical set-up and with the same experimenter roles as for experiment 1.

(iii). Stimuli and procedure

Stimuli and procedures in experiment 2 were identical to those in experiment 1, except that the animation featured a double-decker stand, instead of the podium (figure 1).

(iv). Coding and analysis

Coding and analysis in experiment 2 were identical to that in experiment 1 (electronic supplementary material, Coding and analysis). In experiments 2 and 3, we used one-tailed p-values to compare the gaze duration in the two tests (high agent dominant, low agent dominant). Directionality was already determined by the previous experiment.

(b). Results and discussion

As predicted, infants looked at the screen for M = 10.68 s (s.d. = 8.300) after the video was frozen at the point when the higher agent took away the goal object, and for M = 22.55 s (s.d. = 17.213) after the lower agent took away the goal object (figure 2). Gaze durations in the two tests differed significantly (t₁₇ = 3.25, p = 0.002, one-tailed, Cohen's d = 0.765). Also, these were consistent with the GLMM results (β = 0.554, t₁₇ = 3.25, p = 0.002, one-tailed) after controlling secondary factors (electronic supplementary material, table S3).

Infants in SPP phases looked at the higher agent for M = 2.58 s (s.d. = 0.524) and the lower agent for M = 1.38 s (s.d. = 0.464), with a significant difference between groups (t₁₇ = 5.56, p < 0.001, Cohen's d = 1.311). Infants did not show different gaze durations in two types of SAC phases (M_high = 6.10 s, s.d. = 0.346, M_low = 6.15 s, s.d. = 0.422; t₁₇ = 0.50, p = 0.625, Cohen's d = 0.117). Further, we applied GLMM to confirm the effect of the type of the tests, as was done for the results of experiment 1. Results showed that the type of the tests (β = 0.554, p = 0.002, one-tailed), but not the gaze duration in the SPP phase (β = 0.312, p = 0.405), in the SAC phase (β = −0.054, p = 0.994), or their interaction (β = 3.600, p = 0.698), had effect on the gaze duration (electronic supplementary material, table S4).

In this experiment, we presented two agents using a symmetrical stand that aligns them vertically above or below each other. Again, infants looked at the screen longer when they saw the lower agent obtain the reward instead of the higher agent, compared to the inverse outcome, suggesting their expectation regarding the higher agent to be more socially dominant. Although infants in experiment 2 showed longer gaze durations for the higher agent than the lower one when they were presented on the stand, this did not influence the differentiated gaze durations in tests, as has been demonstrated in experiment 1.

To increase the reliability of the results so far and confirm that the previous findings were not limited to the experimental circumstance (e.g. the built-in camera of the eye tracking system for capturing infant's looking behaviours, size of the monitor), we replicated the results of experiment 2 with different materials such as a different booth, screen size, manner of handling infants and lighting (electronic supplementary material, experiment 3). We found that, consistent with previous experiments, infants looked at the screen longer when they saw the Violation-of-Expectation scenario in which the lower agent had outcompeted the higher agent in securing a reward object. Further, meta-analysis of the results of the three experiments confirmed the overall effect of type of tests. Differences in gaze durations were not influenced by the experiment (i.e. experiments 1, 2 or 3), experimental circumstances or type of stand that was used to elevate agents (i.e. podium versus double-decker stand; electronic supplementary material, figure S1 and tables S6–S11).

4. General discussion

The study investigated whether 12- to 16-month-old infants attribute greater social dominance to agents in a higher spatial position than to agents in a lower position. Infants watched animated videos in which two agents of the same body size competed for a reward object in a zero-sum situation (test phase) after the agents had been simultaneously presented on different spatial positions (higher/lower). Infants tended to look longer after they saw the lower agent had secured the reward object than when the higher agent did. Neither the gazing behaviour prior to the test (experiments 1, 2), nor the experimental circumstance (experiment 3), the form of the stand used to present agents' position (experiment 2, 3) could explain the findings. Based on the Violation-of-Expectation paradigm [40–42], we consider our results as evidence suggesting that infants expect a spatially higher-positioned agent to be socially dominant.

Past studies have established that even very young infants understand the hierarchical relations and make inferences from cues. Infants as young as six months expected an agent from a numerically larger group to win right-of-way against an agent from a smaller group [30]. From about 10 months old, infants use the agents' relative size to predict the outcome of such conflicts [31]. Fifteen-month-old infants expect hierarchical relations to be stable over time and across situations [32]. More recently, Enright et al. [53] found that 17-month-old infants expect the dominant individual to receive more resources than the submissive one [53]. However, this study neither found that infants infer social dominance from mere spatial orientations (i.e. whether sitting on a higher or lower chair), nor found that the agent sitting on the higher chair/spatial orientation was expected to be given more resources than the agent sitting on the lower chair/spatial orientation. Rather, infants inferred social dominance from a ‘dominance structure’ (i.e. two puppets simultaneously approach and try to take a seat in a more attractive and higher chair by bumping into the chair three times, and the submissive puppet backs away, allowing the dominant puppet to sit on the chair), and expected the resources distribution based on such inferences. By contrast, the current study demonstrates that infants infer social dominance from mere spatial orientation relationships of the agents—they expect spatially higher-positioned agents to be socially dominant.

Young children draw inferences about hierarchical relationships based on several different factors. Although most of the related studies that have been conducted among infants have not focused on such evaluative processes [30–32], a recent study showed that infants can distinguish between leaders (i.e. those who demonstrate respect-based power) and bullies (i.e. those who demonstrate fear-based power) [39,54]. Further, children may expect these two types of high-rank individuals to respond differently to subordinates [55]. The present study regards the socially dominant agent as the one who prevails when the agent's goals conflict with those of another agent, specifically, by securing one reward object in a zero-sum situation. We think that such notion of hierarchical relationships shares a common nature with previous studies (e.g. who prevail to sit on an attractive chair; [53]). However, the present experimental contexts did not include direct ‘dominance structures’ [30,31,53]. Instead, agents were statically presented at different positions during the familiarization process, and they did not come into physical contact with each other during the test phase (the subordinates yielded the resource to the dominant agents). Thus, we expect infants to be more likely to perceive high-position agents as leaders rather than as bullies; however, further research is needed the delineate the evaluative processes that underlie these inferences.

Even with limited linguistic abilities, infants already infer social dominance from a vertical position in space. Researchers in the field of developmental psychology have suggested that the association of space and power in infancy may be based on evolved core cognitive modules [26]. Dahl & Adachi [16] showed that chimpanzees, our closest evolutionary relatives, map higher physical position onto high-ranked individuals and low position to low-ranked individuals. In the study, chimpanzee participants discriminated photos of familiar individuals in a vertical arrangement and showed faster identity discrimination to the position-rank-consistent targets (i.e. high-ranked individual presented in the higher and a low-ranked individual in the lower position). Some have argued that the observed conceptual metaphorical mappings in chimpanzees may be confounded by body size, as the participants and individuals in photos have daily interactions, and high-ranked individuals were possibly taller than the low-ranked individuals [16,56]. By contrast, because infant participants in the current study did not have experience with the presented agents, the results suggest a more direct link between spatial recognition and social dominance evaluation. Despite the methodological difference in the chimpanzee study and ours, these findings strongly support the claims that space-dominance link is language-independent and could have emerged before the development of language.

The eye tracking data in experiment 2 found that infants' gaze duration was longer for the higher agent than the lower one, when they were presented on the stand. Although this difference did not influence the social dominance inferences, it must be recognized when considering the relationship of dominance and attention in social contexts. Past studies have suggested that primate models of group social organization stress that dominant individuals receive more positive social behaviour than other group members [57,58]—socially competent and dominant children receive more attention and attract more affiliative activities from peers [59,60], human infants and chimpanzees acquire social information from dominant individuals [61,62]. In line with this view, the current findings may reflect infants' attentional preference towards socially dominant individuals. Further, although the advantage on mere visual attention does not seem to elicit the inference of social dominance (experiment 2), it will be interesting to test whether the advantage on social attention elicit such inference; for instance, whether individuals who receive more attention from others are evaluated as socially competent and dominant.

There are several potential limitations here which may lead to further research. First, although the current findings showed that infants attribute social dominance to agents in higher spatial positions, thereby indicating that dominance inference and spatial recognition are indeed linked in early development, we do not know if this attributional bias is unidirectional or bidirectional. Do infants also expect dominant agents to be presented in, or to actively secure, a higher spatial position? It is known that preschool children not only attribute dominance to people exhibiting higher body gesture (e.g. head up) as more powerful—a similar unidirectional attribution with the current findings—but also use representations of spatial sensory-motor information while thinking about social hierarchy [19,20], suggesting bidirectional attributions of the dominance-space representations. One may explore this by testing whether infants show surprise responses after they saw an agent who won the competition but finally moved to a lower location.

Second, considering that infants have intuitive understanding of the basic physical rules (e.g. gravity; [33,43,44,63]), we presented the agents on the stands to increase the reality of the current set-up. However, as infants are sensitive to the agents' costs of actions (i.e. climbing a ramp [33,36,64]), they may use information about how high agents can climb to draw inferences about their physical abilities, and thus about their capacity to prevail when agents are engaged in a conflict. Note that this process may be achieved without positing a shared representational system for encoding space and social dominance. In the current study, both the podium-like stand, which seems to be more likely as a stair for climbing, and the double-decker stand, revealed the same effect of the spatial position presentation on infants' social dominance inference. Considering these results, we believe that infants are not inferring social dominance based on the cost-competence account. However, it is worthy to explore whether an even simpler presentation including the vertical location difference of the agents may evoke infants' expectations of socially dominant relationships. This may be tested by having the agents presented with a single coloured background, or even by conducting the experiments in real interactions. This kind of manipulation may be more direct and effective than the current ones.

Third, it is also interesting to explore when and how the space-dominance link emerges in ontogeny, although it is hard to predict its presence before infants understand the concept of social dominance [30]. Possibly, the space-dominance link comes from the habituation of daily experience to the associations of those concepts (i.e. social learning). Infants regularly experience that taller individuals have power over shorter individuals (e.g. parents have power over children). Winners in competitions (e.g. Olympic games) always stand on the tops of asymmetrical podiums [19,23,24]. Based on this, infants may form an association between spatial representations and dominance representations (e.g. associating a high position in space with a high social dominance status). It is also possible that the space-dominance link is rooted in an evolved cognitive base and are at least somewhat independent from the environment.

The current findings demonstrate that infants expect spatially higher-positioned agents to be socially dominant. This seems to be consistent with the cognitive framework which posits that mental representations of concepts are tied to their perceptual basis [19,21,22]. The existence of the space-dominance link during the early developmental stages also suggests that evolved cognitive mechanisms may form the basis for the social status hierarchies that undergird human and non-human societies [13–16].

Ethics

In all experiments, participants were recruited from Infant Scientist of Kyushu University, a database of infants whose parents had volunteered to participate in the studies. In accordance with the Declaration of Helsinki, written informed consent was obtained from the children's carers before the experiment was conducted. The procedure was approved by the ethics committee of the Faculty of Human-Environment Studies at Kyushu University (2017-012).

Data accessibility

Data is available in the Dryad Digital Repository: https://doi.org/10.5061/dryad.5hp6pm8 [65].

Authors' contributions

X.M. conceived of the study, designed the study, collected data, carried out the statistical analyses and drafted the manuscript. Y.M. designed the study, helped draft the manuscript. Y.N. helped design the study and collected data. H.N. participated in data analysis. K.H. helped design the study, coordinated the study. All authors gave final approval for publication and agree to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by JSPS KAKENHI grants (nos 18F18999, 18H01083, 18H04200, 17KT0057, 18K02461, 18H04200, 25118003, 26280049, 19H04431 and 17KT0139).