Preverbal infants' evaluation of sharing behaviors within a “taking turns” context is modulated by the perspectives they adopt

Socio-moral evaluations may differ depending on the perspectives adopted. In the field of early socio-moral development, it remains unclear whether preverbal infants evaluate interactions differently when they adopt the perspectives of different social agents. Using a “taking turns” scenario involving one agent initially occupying a swing and a newcomer requesting a turn, this study investigates 11-month-old infants' evaluations of agents who shared the swing by taking turns and the agents who did not take turns. By manipulating both the roles of the agents (i.e., an agent requesting a turn versus an agent asked to yield a turn) and the test phase design, we conducted four eye-tracking experiments to examine whether infants' preferences for agents involved in sharing and non-sharing events vary depending on perspectives from which the interaction is evaluated. In Experiments 1a and 1b, we examined infants' predictions regarding an involved agent's preferences between the agents. In Experiments 2a and 2b, we examined infants' own preferences as a third-party observer. The total sample comprised 103 infants (M = 10 months 29 days; 55 girls), distributed across Experiment 1a (N = 30), Experiment 1b (N = 30), Experiment 2a (N = 21), and Experiment 2b (N = 22). Anticipatory gaze metrics revealed that infants looked longer at an agent involved in the sharing event only when predicting the choices of a newcomer requesting a turn (Experiment 1a). Conversely, this preference was not observed when infants inferred the choices of the initial possessor of the swing who was asked to take a turn (Experiment 1b), nor in the third-party preference conditions (Experiments 2a and 2b). These findings suggest that infants' socio-moral evaluations may be context-dependent and modulated by perspectives. This variability further indicates that, rather than examining whether infants show prosocial preferences, research should focus on elaborating when and under what criteria these evaluations emerge.

Studies on socio-moral evaluation in infancy have provided substantial evidence that preverbal infants prefer prosocial agents such as those who help others achieve their goals, distribute resources fairly, or intervene to protect victims from aggressors (Hamlin et al., 2007; Hamlin and Wynn, 2011; Geraci and Surian, 2011; Kanakogi et al., 2017; Woo et al., 2022). However, research on early social reasoning also indicates that infants' preferences for prosocial behavior are not unconditional. In particular, infants use cues such as spatial relationships, physical size, and numerical group size to infer social dominance, and they often expect dominant individuals to prevail over others rather than to kindly yield to others. Studies show that 12- to 16-month-olds use spatial relationships as cues to infer dominance relationships and show violation-of-expectation reactions when the agent in a lower position obtains the resource in a competition (Meng et al., 2019). Similar violation-of-expectation responses were observed when 10- to 13-month-old infants saw a larger agent bow and yield to a smaller agent (Thomsen et al., 2011); and when 6- to 12-month-olds saw the member of a larger group yield to the member of a smaller group (Pun et al., 2016). By 21 months of age, toddlers further distinguish between legitimate authority and coercive dominance, expecting the commands of a leader—but not those of a bully—to continue being obeyed even after the individual had left the scene (Margoni et al., 2018). Further, infants' expectations of prosocial behavior are shaped by information regarding an agent's moral characteristics. For example, 14-month-olds expect fair distribution from agents who previously engaged in helping behavior, but not from agents who hindered others' goal achievement (Surian et al., 2018). Together, these findings suggest that infants' socio-moral evaluations are flexible and context-sensitive, rather than uniformly prosocial. To reconcile evidence concerning both the prosocial and prudential aspects of early socio-moral development, research should move beyond asking whether infants prefer prosocial behavior and instead examine the conditions under which such preferences emerge.

One potentially important factor affecting socio-moral evaluations is the evaluator's perspective. Socio-moral evaluation is shaped not only by the content of an action but also by the perspective or social role from which the action is evaluated. The ability to recognize others' goals, desires, feelings, knowledge, and beliefs has been regarded as a foundation for prosocial behaviors, such as helping and cooperation (Tomasello et al., 2005; Brownell et al., 2013). However, perspective-taking and empathy do not always lead to impartial moral judgment. Empathy can encourage the over-evaluation of some people, the neglect of others, and the privileging of one over many; thus, it is not a reliable source of information in moral decision-making (Decety, 2021). Developmental psychology has long emphasized the importance of perspective taking in social understanding and moral reasoning. For example, Selman (1971) proposed that reciprocal role-taking abilities are associated with more advanced moral judgments in children aged 8–10 years. Sobel and Blankenship (2021) showed that when asked to take the perspective of the recipient with fewer resources, 3- to 6-year-olds were more likely to make distributions leading to equal results. Regarding the emergence of third-party intervention, Rossano et al. (2011) found that both 2- and 3-year-old children protested violations of possession when their own property was involved, whereas only 3-year-olds frequently protested violations involving a third party's property. Regarding group affiliation and socio-moral evaluation, research has shown that children over 3 years of age judge the wrongness of harmful behaviors differently depending on whether the action occurs within or across group boundaries (Chalik and Rhodes, 2015, 2020). By 5–6 years of age, children's group identity is associated with their assessment of fair and unfair advantages (D'Esterre et al., 2022), and children between 4 and 12 years of age may preferentially allocate limited resources to themselves or their ingroup members (Sudo and Ishikawa, 2025). Preverbal infants are also capable of adopting the perspective of a specific agent in certain contexts. For instance, 7-month-olds can infer another individual's belief in a visual object detection task (Kovács et al., 2010); 10-month-olds can reason about how an agent evaluates and selects among different goal objects (Liu et al., 2017); 11-month-olds are able to predict agents' positive or negative responses toward other group members (Köster and Hepach, 2024). Although developmental research has extensively examined perspective taking and its influence on moral reasoning in childhood, it remains unclear whether socio-moral evaluations in infancy vary according to the perspective adopted.

Infants' socio-moral evaluation of sharing in “taking turns” contexts

To examine whether preverbal infants' socio-moral evaluations vary depending on the roles and perspectives they adopt, we proposed a “taking turns” situation which involves a single swing with an initial possessor (i.e., the agent who arrived first and controls the swing) and a newcomer requesting a turn. In such situations, sharing the swing may be positively evaluated because it helps the other agent achieve their goal. Conversely, refusing to yield and remaining on the swing may also be justified, as the initial possessor has control over the swing and thus is in a dominant position.

Regarding the prosocial aspect, taking turns and sharing the swing can also be explained as being fair (i.e., the newcomer should also have a turn to play on the swing). Infants are expected to reflect an emerging sense of fairness in resource distribution (Sloane et al., 2012). Sharing the swing also involves interactions similar to turn-taking. Children before 2 years of age are reported to be able to conduct joint activities with cooperative turn-taking to achieve a shared goal (Warneken et al., 2006) and 3-year-olds show sensitivity to turn-taking violations in cooperative contexts (Gräfenhain et al., 2009). Sensitivity to signals indicating the intention to use objects emerges in infancy. 15- to 16-month-olds reliably recognize begging gestures (i.e., the individual turning their palm upwards while facing an individual who possesses food or a desired object) as requests and respond to them (Rossano and Liebal, 2014). Infants also show sensitivity to turn-taking interactions involving only signal exchange but no object transfer; for example, 10.5-month-olds are reported to show sensitivity to turn-taking behaviors and the variability of signal sequences influences their inference about agency attribution (Tauzin and Gergely, 2019).

However, the interaction of “taking turns” discussed in the current study is not synonymous with object exchange, since the agent offering a turn in the present study will not receive immediate reward. Taking turns in a situation with limited resources entails an asymmetrical relationship between the giver and taker, and there is a possible conflict of intentions between the initial possessor (who wants to keep the turn) and the newcomer (who requests a turn). Regarding infants' representation of the “giving” and “taking” relationship, it is reported that infants younger than 6 months prefer a giver who returns a dropped ball to the initial possessor over a recipient who takes it away (Hamlin and Wynn, 2011). There is also evidence that 10.5- and 12-month-olds, but not 9-month-olds, represent three-argument interactions (giver, taker, and a flower) and recognize changes in action roles (Schöppner et al., 2006). Tatone et al. (2015) further suggest that 12-month-olds represent giving behaviors in the structure of a three-argument interaction and recognize it as a transfer-based social interaction but often filter out the agent who initially possessed the object in taking events. This further indicates the asymmetry between giving and taking relationships.

Based on the above research, the present study focused on 11-month-old infants, who are capable of representing a three-argument structure (i.e., the initial possessor of the swing, the newcomer asking for a turn, and the swing).

Moreover, infants' socio-moral evaluation of agents refers to their positive or negative responses—or expectations of such responses—toward specific agents involved in a social interaction, based on the assessment of the agents' actions or intentions (Hamlin et al., 2007; Van de Vondervoort and Hamlin, 2018; Pun et al., 2018). In the present study, we specifically examine infants' expectations regarding positive responses to agents. Within a “taking turns” context, we investigated infants' preferences for different agents within Sharing vs. Non-sharing events by comparing (a) the preference for an initial possessor who yielded the swing to a newcomer with one who remained on the swing (Condition A, Figure 1), and (b) the preference for a newcomer who got on the swing with one who did not get on the swing (Condition B, Figure 1). We hypothesized that preverbal infants' evaluation of agents involved in “taking turns” interactions would vary depending on the perspective and role they adopt. Given the asymmetrical nature of giving and taking, we predicted that when reasoning about these social relationships from the perspective of a stakeholder, adopting the newcomer's vs. the initial possessor's perspective would lead to different socio-moral evaluations. From the perspective of a newcomer requesting a turn, we predicted that infants would prefer an agent who shared the swing over an agent who did not (Experiment 1a). Conversely, when infants adopted the initial possessor's perspective, we predicted that a preference for the agent who took turns might not be observed (Experiment 1b). One possible reason is that, for the initial possessor, yielding a turn results in reduced playtime on the swing. Finally, we predicted that when evaluating agents from a third-party perspective, infants would not exhibit a clear preference between agents involved in Sharing event by taking turns and agents who did not share the swing (Experiments 2a and 2b). This prediction was based on indication from previous research on infants' socio-moral evaluation. For example, agents who share a resource by taking turns are likely to be evaluated as prosocial and kind (e.g., Hamlin and Wynn, 2011), whereas agents who continue to control access to the resource may be perceived as dominant (e.g., Pun et al., 2016).

By combining two Familiarization Phase conditions with two Test Phase designs, we conducted four eye-tracking experiments that manipulated both the roles of focal agents and the evaluative perspectives (Figure 1). Across Familiarization Conditions A and B, we manipulated which agent remained constant across the Sharing and Non-sharing events. In Condition A, the newcomer requesting a turn was identical across events. In Condition B, we reversed the roles of the two characters so that the initial agent occupying the swing remained identical. Test Phase Pattern 1 (Experiments 1a and 1b) examined infants' reasoning regarding social relationships between the agents from an involved party's perspective. Conversely, Test Phase Pattern 2 (Experiments 2a and 2b) assessed infants' preferences as third-party observers. All four experiments followed a between-participants design. For instance, participants in Experiment 1a were familiarized with Condition A, in which the green agent acted as the newcomer, and were subsequently presented with Test Phase Pattern 1. In contrast, participants in Experiment 1b were familiarized with Condition B, in which the green agent served as the initial possessor of the swing, before viewing Test Phase Pattern 1.

Experiment 1a and experiment 1b

Participants

Recordings of 60 preverbal infants were included in the data analysis. Thirty infants were assigned to participate in Experiment 1a and the remaining 30 participated in Experiment 1b. The age of the participants who provided valid data ranged from 10 months 13 days to 11 months 17 days (Experiment 1a: M = 10 months 26 days; Experiment 1b: M = 11 months 0 days), and there were 27 girls among the 60 participants. Across both experiments, recordings from another seven participants were excluded because of calibration failure or other technical problems, and recordings from another nine infants were excluded because they did not clear the looking rate standard in the Familiarization Phase or failed to complete the experiment because of fussiness (see Results section 2.1.2 for details about data exclusion criteria). Sample size determination was based on prior studies utilizing similar test phase designs (e.g., Kuhlmeier et al., 2003, Exp. 1; Liu et al., 2017; Powell and Spelke, 2013, Exp. 1; Tauzin and Gergely, 2019), in which a sample size of at least 20 to 24 participants per condition is typically established. In the present study, to test our specific hypotheses, we set a higher target of 30 participants per condition for Experiments 1a and 1b. For Experiments 2a and 2b, which were positioned to validate third-party perspective evaluations, we targeted a sample of approximately 20 participants per condition, resulting in a final sample of 21 and 22 participants, respectively (see Section 2.2.1). Data validation was assessed on an ongoing basis after each session concluded, in accordance with our established exclusion criteria. All infants were recruited from a pool of participants at the University of Tokyo Baby Lab (Komaba Campus) and were born and raised in Japan.

Experimental design and gaze movement measures

Experiments 1a and 1b consisted of a Familiarization Phase and a Test Phase. The scheme of overall structure of the experiments is shown in Figure 1; for a sample of the stimulus animations used in Experiments 1a and 1b, please refer to Supplementary Video 1. Note that for ease of comparison in Video 1, the Non-sharing condition is displayed as a picture-in-picture insert in the upper-right corner of the Sharing condition; however, during the actual experiment, these two conditions were presented to participants in an alternating sequence.

In the Familiarization Phase, animation clips of the Sharing and Non-sharing conditions were alternately presented twice. The Familiarization Phase involved a swing on the left side of the screen. We employed a swing-like object to represent the goal because it allowed only one agent to use it at a time, while providing a salient positive outcome of elevating the agent above the ground. Compared with other designs, the swing depicted the repeated exclusive use of a desirable resource more naturally. For example, hill-climbing paradigms (e.g., Hamlin et al., 2007) or slide-like objects (e.g., Lewkowicz, 2013) make it difficult to represent the repeated monopolization of an activity, whereas trampoline-like designs introduce ambiguity as to whether the upward movement results from the object itself or from the agent's own ability. Therefore, we adopted an elevated reciprocating object that can be occupied by only one agent at a time.

In the experimental scenario, an agent (the initial possessor of the swing) arrives from the right side of the screen and jumps onto the swing. The swing then begins to move, helping the agent move upwards at the highest point, accompanied by sparkling positive sound effects. Subsequently, another agent (a newcomer requesting a turn) enters the screen and jumps several times while facing the swing. The design was informed by prior research on intention and goal representation in preverbal infants. For instance, Hamlin et al. (2007) utilized a scenario where an agent repeatedly attempts to climb a hill but fails and slides down, and Köster et al. (2016) presented two agents attempting to reach for and play with a ball, with one being blocked by an obstacle. In the present experiment, the newcomer arrives and jumps upward to reach the swing, which is currently occupied and unavailable. The agent occupying the swing notices this action and turns toward the newcomer. The newcomer then continues jumping while maintaining eye contact with the occupying agent, signaling a desire to access the swing and a change of turn. Two kinds of results follow. In the Sharing condition, the initial possessor gets off the swing and changes position with the agent who arrived later. In Non-sharing condition, the agent on the swing turns back and continues to use the swing. In the Sharing condition, the newcomer agent ends up playing on the swing at the end, but in the Non-sharing condition, the initial possessor is on the swing from the beginning to the end.

To manipulate the role of the evaluating agent in the subsequent Test Phase, character assignments differed across Experiments 1a and 1b (see Figure 2). In Experiment 1a, the newcomer requesting a turn was consistently represented by a green sphere-shaped character, whereas the initial-possessor agent who did or did not yield the swing was represented by one of two inverted cup-shaped characters, with the purple and orange inverted cup-shaped characters counterbalanced between participants. In Experiment 1b, these role assignments were reversed, such that the initial possessor of the swing was represented by a green sphere-shaped character, while the newcomer requesting the swing was represented by one of the two inverted cup-shaped characters. The colors of the three characters were selected using a triadic color scheme to ensure equal distances between the colors of any two characters on the color wheel. Across all four distinct animation scenarios, the timing, duration, and types of sound effects were identical. This design ensured that the four animations were entirely indistinguishable based on the auditory cues alone.

The Test phase design in Experiment 1a and Experiment 1b illustrated one agent (the green ball) in the middle choosing between the two agents. Although the overall structure of the Test Phase was identical in Experiments 1a and 1b, the specific behaviors and roles assigned to the agents differed between the two experiments. In Experiment 1a, the central agent was the newcomer agent (receiver) who chose between the character that had yielded the swing and the character that had not by moving closer to either of the agents. In Experiment 1b, the central agent was the initial occupant of the swing (giver), who chose between the character to whom they had yielded the swing and the character to whom they had not.

At the beginning of each trial, to indicate that the central agent was choosing between the other two characters, it sequentially glanced at the left and right characters once before moving to the center position (following the experimental paradigm of Liu et al., 2017). The screen then remained static for 3 s (for examples of this still frame, see, e.g., Köster et al., 2016). Infants' anticipatory gaze during this pre-choice 3-s time window served as our primary measure. Following this interval, the central agent approached either the agent involved in the Sharing or Non-sharing event and jumped together, accompanied by a positive sound effect. After the central agent demonstrated its choice, the screen froze for 5 s (for a similar 5-s post-choice still frame, see, e.g., Köster and Hepach, 2024).

Each participant underwent six test trials. The choice conditions (selecting either the agents involved in the Sharing or Non-sharing event) were presented alternately and repeated across the three blocks (resulting in a 6-trial sequence with an ABABAB structure). Across participants, the spatial location of the choice (left vs. right) and colors of the characters (orange vs. purple) in the first trial were counterbalanced. However, to minimize the cognitive load associated with spatial shifting, the physical positions of the characters were held constant across all six trials for all participants.

The entire experiment was conducted in an experimental booth with the infants seated on their parents' laps. Stimuli were presented on a 24-inch display connected to a Tobii Pro Spectrum eye tracker, which captured the infants' eye movements at 300 Hz. The animated stimuli were created and rendered at a frame rate of 24 frames per second using Blender (Version 3.4.1). All video clips were presented, and the overall progression of the experiment was controlled using Tobii Pro Lab (Version 1.207). Eye tracking was established using a 9-point calibration procedure. Participants' behaviors were monitored by the experimenter via a separate computer. Because the progression of the experiment, the number of trials per phase, and the durations were pre-programmed, no experimenter intervention occurred after the calibration phase, except to abort the session if a participant became too fussy to proceed. Data inclusion was restricted to participants who successfully completed both the calibration protocol and the entire experimental procedure.

The study focused on the following measures to evaluate infants' looking behaviors and their socio-moral evaluations. First, we assessed infants' fixation proportion in the Familiarization Phase. Since the Sharing and Non-sharing clips shared an identical scenario for the first 27.4 s of the 40.6-s clip, the divergence (sharing vs. not sharing) occurred exclusively during the final 13.2 s. Thus, verifying attention during the Familiarization Phase was vital to ensure that infants recognized the differences between the Sharing and Non-sharing events. Based on the characteristic of our animation stimuli and experimental design, we applied a 60% minimum fixation-time cutoff to the entire screen to ensure sufficient exposure during the Familiarization Phase. Recordings of infants who fell below the 60% fixation threshold during trials of either the Sharing or Non-sharing condition in the Familiarization Phase were excluded from the analysis.

Then, to examine infants' anticipation for which agents would be chosen by the agent in the middle, we measured the infants' looking time at each of the agent during the Test phase before the agent in the middle made a choice. Referring to Tauzin and Gergely (2019), we created two Regions of Interest (ROIs) covering either side of the screen where the agent that differed between the Sharing and Non-sharing events were placed (Figure 3). We compared the differences in the total fixation duration for each ROI. We used these differences to reflect infants' anticipation of which agent would be chosen by the agent in the middle (i.e., which direction the agent in the middle would move). As a supplementary measure, we also examined the infants' fixation duration during the 5-s frozen scene after the agent in the middle made its choice. The six test trials presented each agent being chosen three times each. Because repeated test trials may lead infants to adapt their looking behavior to previously presented test conditions, we restricted our analysis to the first two test trials to assess infants' initial expectations based on the interactions in the Familiarization Phase (for an example of a design with only two test trials, see Tatone et al., 2015). However, to evaluate the effect of cumulative exposure to stimuli in the Test Phase, we also tracked the infants' anticipatory gazes across all six test trials. All raw data were exported from Tobii Pro Lab (Version 1.207) and subsequently analyzed in RStudio (Version 2023.06.0+421) (RStudio Team, 2024).

Results

Anticipatory gaze before the choice was made

As a measure indicating infants' anticipation of which agent would be preferred, we first compared participants' fixation duration at the two agents (i.e., in Experiment 1a, the initial possessor who shared the swing vs. the initial possessor who did not; in Experiment 1b, the newcomer who was given a turn vs. the newcomer who was not) during the 3-second time window before the choice was made. Based on the mean total fixation times for the ROIs corresponding to each agent, we calculated the differences in fixation time between the two ROIs (the agent involved in the Sharing minus the agent involved in the Non-sharing event) for each participant on every trial. The distribution of the difference scores for each test trial in Experiments 1a and 1b is shown in Figure 3.

To examine how these differences in fixation at each agent varied across experiments and test trials, we constructed a linear mixed-effects model (LMM) with the Experiment and Trial as fixed effects and a random intercept for participants. Although visual inspection of the residual Q-Q plot revealed slightly heavy tails, the residuals were generally normally distributed. Relying on the robustness of linear mixed-effects models to minor deviations (Schielzeth et al., 2020), we fit the model using raw fixation duration data, without applying transformations. Due to the alternating block design (ABABAB), Trials 3–4 and 5–6 were identical to Trials 1–2. To capture this test-trial repetition effect, the six chronological trials were collapsed into a three-level repetition factor: Repetition 1 (pooled Trials 1 and 2), Repetition 2 (pooled Trials 3 and 4), and Repetition 3 (pooled Trials 5 and 6). Our analysis focused particularly on infants' initial responses during Repetition 1, where each condition, in which either the agent involved in the Sharing event or the agent involved in the Non-sharing event was chosen, was presented exactly once. A Type III analysis of variance (ANOVA) using Satterthwaite's method yielded a significant main effect of Repetition [ F (2, 296) = 4.92, p = 0.008, η 2 = 0.03]. The main effect of Experiment [ F (1, 58) = 0.04, p = 0.847, η 2 < 0.01] and the interaction between Experiment and Repetition [ F (2, 296) = 2.67, p = 0.071, η 2 = 0.02] were not significant. Post hoc pairwise comparisons were conducted using the estimated marginal means with Tukey's adjustment for multiple comparisons. For Experiment 1a, the post hoc analysis revealed that the difference scores significantly decreased from the first to the second repetition [ t (296) = 3.54, p = 0.001, d = 0.21] and subsequently increased from the second to the third repetition [ t (296) = −2.91, p = 0.011, d = −0.17]. Conversely, the difference in scores in Experiment 1b did not vary significantly across repetitions. Given our experimental design, the first two trials constituted a block in which the two agents were each chosen once across separate trials. The subsequent repetition of this two-trial set over the remaining four trials appeared to have driven a trial-repetition effect in Experiment 1a.

The main purpose of the study was to compare these difference scores against zero, where a positive difference score would indicate the anticipation that the central agent would approach the initial possessor who shared the swing. Therefore, follow-up post hoc tests were performed on the estimated marginal means with Bonferroni adjustment for multiple comparisons. For Experiment 1a, the difference scores were significantly greater than zero only during the first repetition (Trials 1 and 2), as confirmed by two-tailed t (264) = 3.04, p = 0.016, d = 0.28, and directional one-tailed tests examining positive deviations (i.e., whether fixation time was significant longer for the agent in Sharing event), p = 0.008. Conversely, difference scores did not significantly deviate from zero during the second (two-tailed: p = 0.447, d = −0.17) or third test repetitions (two-tailed: p = 0.179, d = 0.21). In Experiment 1b, the difference scores did not significantly exceed zero in any of the repetitions (two-tailed: Repetition 1: p = 0.518, d = 0.16; Repetition 2 and Repetition 3: p > 0.1, d < 0.01). In summary, these anticipatory gaze results indicated that infants predicted the central agent would approach the character involved in the taking turns and sharing interaction when the central agent was a newcomer requesting a turn (Experiment 1a). However, this predictive preference was not observed when the central agent was the initial possessor who had been asked to yield the swing (Experiment 1b).

Looking time after the choice was presented

The study also examined infants' looking time after the agent in the middle made a choice by comparing the total duration of fixation on the whole screen of each condition (the agent in the Sharing event being chosen vs. the agent in the Non-sharing event being chosen) during the 5-second still screen after the central agent made a choice. To normalize the bounded eye-tracking data for linear modeling, the total fixation duration during this 5-second time window was converted into empirical logits. We then calculated the empirical logit values by taking the natural log of the ratio of fixation time to the remaining trial time, incorporating a standard 0.5 smoothing offset, to account for absolute zero or ceiling fixation scores. We then constructed a linear mixed-effects model (LMM) with empirical logit fixation duration as the dependent variable. The condition (Sharing vs. Non-sharing), repetition (First, Second, and Third blocks), and experiment (Experiment 1a vs. Experiment 1b) were entered as fixed main effects alongside participant-specific random intercepts. A Type III ANOVA test with Satterthwaite's method on this model revealed a significant main effect only for Repetition [ F (2, 902) = 5.15, p = 0.006, η 2 = 0.03]; no other main effects or interactions reached statistical significance. Post-hoc pairwise comparisons with Tukey's adjustment indicated that this main effect of repetition was driven by a decline in infants' looking time at the final scene from the first repetition (Trials 1 and 2) to the third repetition (Trials 5 and 6). [ t (290) = 3.13, p = 0.006, d = 0.26]. Furthermore, to explore the differences between the conditions, follow-up pairwise comparisons were conducted with False Discovery Rate correction. The only marginal trend observed emerged during the third repetition of Experiment 1a (Trials 5 and 6), where fixation durations were slightly shorter when the agent in Sharing event was chosen compared to when the agent in Non-sharing event was chosen [ t (290) = −1.76, p = 0.080, d = −0.45]. All other pairwise comparisons were non-significant. Collectively, these findings provide insufficient evidence to suggest that infants' visual attention systematically differentiated between trials where the agent involved in the Sharing or Non-sharing event was chosen in either Experiment 1a or 1b. Therefore, no significant differences in looking times were observed between conditions where the focal agent in the center chose the agent involved in Sharing event vs. the agent involved in Non-sharing event.

Experiment 2a and experiment 2b

In Experiments 2a and 2b, we modified the test phase to further investigate whether infants would show the same preference for the agent involved in the Sharing event from a third-party observer's perspective.

Participants

Recordings from 21 and 22 infants participating in Experiments 2a and 2b, respectively, were included in the data analysis. For both experiments, the average age was 11 months 0 days. The age range was 10 months 20 days to 11 months 15 days. Data from another four participants were excluded because of technical problems regarding eye-tracking recording, and data from nine participants were excluded because the looking time during the Familiarization phase did not reach the standard or they became fussy and failed to complete the entire recording. Participants were recruited from the same pool as in Experiment 1a and Experiment 1b.

Experimental design and gaze movement measures

The stimuli for the Sharing and Non-sharing events in the Familiarization phase were identical to those used in Experiments 1a and 1b (with the Familiarization phases of Experiments 2a and 2b corresponding directly to Experiments 1a and 1b, respectively). However, to examine infants' preferences for characters as third-party observers, we modified the test phase design and timeline of the experiment. The overall experimental design is illustrated in Figure 4.

In the test phase of Experiments 1a and 1b, before the 3-s duration of the freezing scene, there was an agent in the middle choosing between the two agents, indicating that the agent would make a choice later and trigger infants' anticipation gaze about which agent would be chosen. In the Test Phase of Experiments 2a and 2b, only the two agents positioned on the left and right sides of the screen were presented on the screen. Therefore, an additional modification was introduced to indicate whether one of the two agents would be selected. Following the presentation of two agents positioned side by side on the screen, a ring-shaped object resembling the seat of a swing appeared beneath one of the agents and lifted the agent upward once while a positive sound effect was played simultaneously. This procedure was designed to assess the infants' selective preferences by measuring which agent they expected to experience a positive event. For similar reasons, the test timeline was modified; rather than presenting six consecutive post-familiarization trials, we administered two baseline trials prior to familiarization (Pre-tests 1 and 2), followed by two post-familiarization test pairs that replicated the exact configurations of the pre-tests. This design allowed us to examine shifts in infants' looking preferences toward each character by comparing their preference for looking at each character before the Familiarization Phase with that after observing the “taking turns” interactions. A sample of the test-phase design in Experiments 2a and 2b is shown in Supplementary Video 1.

Results

Following the data analysis procedures for Experiments 1a and 1b (see Section 2.1.3), we first conducted an analysis by grouping the six test trials into three blocks of repetitions. The two baseline trials (Pre-tests 1 and 2) formed the first block, during which each of the two agents was positively evaluated once. The four subsequent post-test trials (Post-tests 1 through 4) replicated this identical pair of test-phase configurations twice. We then analyzed the contrasts between infants' preferences for the agents in the pre-test and post-test trials.

Anticipatory gaze before the choice was made

We maintained the same ROIs from Experiments 1a and 1b to represent the agents that differed in the Sharing and Non-sharing event. We then calculated the differences in fixation duration during the 3-s still frame across each test trial. The infants' fixation duration difference values (Sharing minus Non-sharing ROIs) are shown in Figure 5. To analyze potential changes in infants' visual preferences across the trial timeline, we constructed an LMM with fixation duration difference scores (Sharing minus Non-sharing) as the dependent variable. Repetition, experiment (Experiment 2a vs. Experiment 2b), and their interactions were entered as fixed effects alongside participant-specific random intercepts. A Type III ANOVA test using Satterthwaite's approximation revealed no significant main effects or interactions.

To assess whether infants exhibited a looking preference toward either character, we extracted the estimated marginal means from the model and conducted two-tailed tests against zero using Bonferroni adjustment. These post hoc comparisons indicated that the difference scores did not significantly deviate from zero in any repetition block in either experiment. Next, we substituted the trial repetition factor in the LMM with a binary factor representing Pre/Post-test phases to predict looking differences. Similarly, a Type III ANOVA using Satterthwaite's approximation of this revised model revealed no statistically significant main effects for the Pre/Post-test phase, experiment, or their corresponding interactions (all p s > 0.10). These results indicated that infants' anticipatory gaze toward the agents involved in Sharing vs. Non-sharing event did not differ significantly, either during the Post-test phase or in terms of the change between Pre- and Post-test.

Looking time after the choice was presented

Next, we evaluated infants' total fixation duration during the final 5-second static scene in Experiments 2a and 2b across different conditions (i.e., the agent involved in the Sharing event being praised vs. the agent involved in the Non-sharing event being praised). As in Section 2.1.3, an LMM was executed on empirical logit fixation times, specifying conditions, test trial repetition, experiment, and their interactions as fixed effects and participants as a random intercept. A Type III ANOVA revealed a significant main effect of test trial repetition [ F (2, 205) = 20.57, p < 0.0001, η 2 = 0.17). No other main effects or interactions, including those involving conditions, were statistically significant. Post-hoc pairwise comparisons demonstrated that infants' looking time significantly declined in the third repetition block (Post-test 3 and 4) compared to both the first two trials (Pre-tests 1 and 2; t (205) = 6.20, p < 0.0001, d = 0.95) and the second test trial block (Post-tests 1 and 2; t (205) = 4.51, p < 0.0001, d = 0.69). Similarly, when the model was modified to collapse the four post-test trials into a single phase, in contrast to the pre-test trials, the main effect of the condition and its interaction with the experiment remained non-significant (all p s > 0.10). This revised model yielded only a significant main effect of Phase [ F (1, 209) = 19.12, p < 0.0001, η 2 = 0.08]. Taken together, while infants' looking time at the last scene decreased across repeated test trial blocks, looking durations did not differ between trials in which either agent was praised.

Across four experiments, we manipulated roles (the initial possessor of the swing vs. a newcomer asking for a turn) and perspectives (involved party of the characters vs. third-party observer) and examined whether infants preferred an agent who took part in a “taking turns” interaction over one who refused or failed to take a turn. Anticipatory gaze measures showed that 11-month-olds exhibited a preference for the agent who took turn and shared the swing only when they adopted the perspective of a newcomer requesting a turn (Experiment 1a). This indicates that infants anticipated that a newcomer requesting a turn would approach the agent who helped it achieve its goal (i.e., by allowing it to use the swing), reflecting a preference for the agent who behaved beneficially toward it. In contrast, this preference for the agent involved in the sharing interaction did not emerge when they evaluated the interaction from the initial possessor's perspective (Experiment 1b), nor did it appear during third-party evaluations. In addition, infants' fixation durations during the final scene of each trial did not differ significantly according to whether the agent involved in the Sharing or Non-sharing event was chosen.

The present study aimed to examine the conditions under which infants favor prosocial agents engaging in resource-sharing behaviors. Based on our findings, 11-month-old infants' preference for agents involved in sharing interactions appears to be relatively constrained in the “taking turns” context, where resources are limited, and the relationship between the initial possessor and the newcomer is asymmetrical. These findings suggest that early socio-moral preferences may be context dependent and potentially modulated by the specific social perspective adopted by the infants.

We manipulated the perspective adopted by the infants primarily by varying the roles and agents presented during the test phases (e.g., whether the agent making a choice acted as a receiver or a giver). This design assumes that preverbal infants can process and adopt the perspectives of animated characters. Kuhlmeier et al. (2003) altered the test-phase configurations and found that 12-month-olds looked significantly longer when a protagonist (a ball) approached the helper. However, when the helper or hinderer approached the ball, the infants showed no difference in looking time. Previous studies have also shown that preverbal infants initially utilize an other-reference framework to encode objects in their memories (Grosse Wiesmann et al., 2025). By 12 months of age, they do not merely track an agent's perspective but also balance it against their own representation of the event layout (Manea et al., 2023).

Regarding gaze measures, expectations of the agent's selection were captured solely via anticipatory gaze rather than post-choice looking times (Experiment 1a). One plausible explanation is that the fixed 5-second window imposed on the final scene may have been too brief to capture classical violation-of-expectation (VoE) effects. Indeed, in previous looking-time paradigms utilizing a similar test phase architecture, infants' baseline attention to expected or congruent outcomes frequently exceeded an average of 5 seconds when evaluated under standard, infant-controlled look-away criteria (e.g., Kuhlmeier et al., 2003). Moreover, using a fixed predetermined duration, rather than a gaze-contingent design, might prevent look-away behavior from meaningfully driving the trial progression. It is also possible that the 3-second static window prior to the choice presentation interfered with looking behaviors during the final 5-second scene, as this setup introduced two separate still-frame intervals within a single trial.

When interpreting our experimental results, it is also necessary to evaluate several perceptual and attentional factors beyond the act of taking turns that could potentially influence infants' fixation times toward each agent during the Test Phase. First, the familiarity and visual richness of the conditions varied during the Familiarization Phase. In the Sharing event, the physical positions of the characters shifted as they traded places, thus providing a dynamic visual layout. By contrast, the spatial configuration in the Non-sharing event remained static throughout the trial. Consequently, infants may have lost interest during the Non-sharing events in the Familiarization Phase, leading to a decline in looking rates. However, our data exclusion criteria (see Section 2.1.2) strictly required infants to maintain a fixation time of at least 60% for the entire video duration in at least one trial for each condition. This rigorous threshold ensured that infants adequately attended to the agents in Non-sharing event and were capable of detecting structural differences between the sequences where agents engaged in taking turns vs. those where they did not. Second, as infants spent a substantial amount of time observing the agent occupying the swing, cumulative exposure could exert a familiarity or novelty effect during the Test Phase (Houston-Price and Nakai, 2004; Kidd et al., 2012). For instance, in Experiment 1a, the sharing initial possessor voluntarily dismounted the swing, resulting in shorter cumulative duration of swing occupancy. This relative novelty might explain why infants preferentially looked at the agent shared the swing during the early trials (Trials 1 and 2). However, if this perceptual novelty hypothesis were correct, infants in Experiment 1b should have shown an even stronger preference for the newcomer who never got on the swing and was therefore the most novel character. Crucially, the Test Phase data for Experiment 1b revealed that the differences in scores did not deviate significantly from zero. Similarly, the structural differences in the Familiarization Phases between Experiments 1a/2a and 1b/2b rule out the possibility that the infants simply preferred the character who either initially occupied or remained on the swing because of recency or primacy effects. Neither of these low-level perceptual accounts fully explains the pattern of results observed in our experiments. Therefore, the findings across our four experiments demonstrate that 11-month-old infants' visual preferences during the Test Phase cannot be attributed to superficial features such as the duration of swing occupancy or the temporal order of character appearances. Instead, their evaluations are best explained by the socio-moral framework of the “taking turns” interaction and its direction (giver or receiver).

Regarding the relationship between social interactions and socio-moral evaluation, previous studies also suggest that infants' and children's inferences about social ties between agents are heavily influenced by observed behaviors such as imitation (Over and Carpenter, 2015; Pepe and Powell, 2025) or pedagogical demonstrations (Ling et al., 2024). Sharing behaviors in the “taking turns” situation shares structural similarities with these interactions, primarily because the newcomer replicates the actions of the character who came out and got on the swing first. Crucially, however, taking turns to use a swing is characterized by a unique constraint: the swing can only be occupied by a single character at a time, and the newcomer's access to the swing is decided by the choice of the initial possessor (for reasons concerning the use of a swing-like object to represent the goal, see Section 2.1.2). This inherent resource limitation and asymmetrical dominance relationship, coupled with role alternation between the two agents, distinguishes this situation from standard collaborative or imitative contexts. Furthermore, by focusing on the roles and perspectives in the context of “taking turns”, this study advances the understanding of early socio-moral evaluations within asymmetric relationships.

Moreover, evaluating interactions that involve taking turns or sharing behaviors inherently entails a sense of possession. The agent who initially occupied and played on the swing possessed it not only through physical control over the resource, but also by virtue of being the first to arrive. Previous developmental research has indicated that preschoolers infer ownership from who controls permission or accessibility (Neary et al., 2009; Pesowski and Powell, 2023) and has traced how young children utilize temporal information, possession history, and a first-possessor bias in their ownership judgments (for a review, see (Pesowski et al., 2022); see also Friedman and Neary, 2008; Rossano et al., 2011; Kanngiesser et al., 2015; Noles and Keil, 2019; Verkuyten et al., 2015). Although our experiment proceeded under the assumption that preverbal infants recognize the swing as a rewarding resource and perceive the occupying agent as having control over it, the current findings do not provide direct evidence on whether preverbal infants possess an abstract sense of possession. Consequently, it remains unclear whether their socio-moral evaluations and behavioral expectations were predicated on ownership concepts, or what specific cues they utilized to infer possession. Future studies could clarify this by manipulating the nature of agent interactions; for instance, by comparing a scenario in which the swing is voluntarily yielded vs. one where a newcomer forcibly takes over the resource (cf. Blake and Harris, 2009, for a similar paradigm exploring gift-giving vs. stealing in preschoolers). Further research is required to map the early emergence of possession judgments and systematically examine their relationships with socio-moral evaluations during early development.

In the present study, we found no evidence that preverbal infants, as third-party observers, preferred either the agent who yielded the swing or the agent who received the turn after observing a “taking turns” interaction (Experiment 2a and Experiment 2b). However, one limitation of the present study is that we examined only 11-month-old infants raised in Japan in a laboratory setting. Previous research has investigated infants' preferences for prosocial behaviors, such as helping, across cultures (e.g., Lucca et al., 2025) and has examined the developmental trajectory of prosocial behaviors using observations of infants' everyday interactions (e.g., Hammond et al., 2017; Silletti et al., 2026). Future research could examine how infants' own taking-turn behaviors, such as sharing limited resources or yielding a place to others, emerge in everyday contexts across cultures. Another notable limitation of the present study is that our experimental stimuli isolated a single, non-reciprocal episode of taking turns. Specifically, the agent who initially occupied the swing received no gratitude or social reinforcement upon relinquishing the resource, was subsequently forced into a waiting role and never regained access. However, in natural settings, complying with another person's request to share resources is typically accompanied by positive emotional rewards. For instance, Tan et al. (2026) reported that toddlers aged 16–23 months demonstrated greater happiness when giving resources to a puppet than when keeping the resources for themselves. Furthermore, older infants and toddlers are capable of prioritizing prosocial giving over possessive behaviors. Research has shown that 19-month-old infants can override their possessive tendencies to share highly valued personal items with others (Barragan and Meltzoff, 2021). Similarly, resource scarcity does not necessarily increase conflict in early childhood; 2-year-olds are more likely to resolve disputes through sharing when toys are scarce, although individual factors such as age and gender composition can further modulate these prosocial resolutions (Caplan et al., 1991). Nevertheless, empirical evidence also indicates that an understanding of indirect reciprocity emerges during toddlerhood (Nava et al., 2019). To bridge the gap between infancy and later development, future research should incorporate early social experiences, such as children's practice of giving or requesting toys from others, and adults' reactions to social conflicts (e.g., Hay and Ross, 1982; Ross et al., 2011). Furthermore, the lack of significant differences in looking time toward the agents under some experimental conditions indicates the potential role of individual differences. Future longitudinal studies should systematically examine early developmental trajectories of altruistic behavior from infancy to toddlerhood by accounting for individual variability, such as infant temperament, as well as nursery school and home environments.