ArticlePDF Available

Infants’ Social Evaluation of Helpers and Hinderers: A Large-Scale, Multi-Lab, Coordinated Replication Study

November 2024
Developmental Science

DOI:10.1111/desc.13581

Authors:

Yiyi Wang

University of Chicago

Nicolás Alessandroni

Concordia University

Show all 85 authorsHide

Evaluating whether someone's behavior is praiseworthy or blameworthy is a fundamental human trait. A seminal study by Hamlin and colleagues in 2007 suggested that the ability to form social evaluations based on third-party interactions emerges within the first year of life: infants preferred a character who helped, over hindered, another who tried but failed to climb a hill. This sparked a new line of inquiry into the origins of social evaluations; however, replication attempts have yielded mixed results. We present a preregistered, multi-laboratory, standardized study aimed at replicating infants’ preference for Helpers over Hinderers. We intended to (1) provide a precise estimate of the effect size of infants’ preference for Helpers over Hinderers, and (2) determine the degree to which preferences are based on social information. Using the ManyBabies framework for big team-based science, we tested 1018 infants (567 included, 5.5–10.5 months) from 37 labs across five continents. Overall, 49.34% of infants preferred Helpers over Hinderers in the social condition, and 55.85% preferred characters who pushed up, versus down, an inanimate object in the nonsocial condition; neither proportion differed from chance or from each other. This study provides evidence against infants’ prosocial preferences in the hill paradigm, suggesting the effect size is weaker, absent, and/or develops later than previously estimated. As the first of its kind, this study serves as a proof-of-concept for using active behavioral measures (e.g., manual choice) in large-scale, multi-lab projects studying infants.

Sample image of choice phase. Note: Choice phase foam 3-D stimuli presented on a white board. Board dimensions: 45 cm × 60 cm; Shape dimensions: 9.9 cm × 9.9 cm (blue square), 14.8 cm (base) × 13 cm (height; yellow triangle).

…

Age × Choice plot. Note: Probability of choosing the Helper/Push-up character (over Hinderer/Push-down character) across ages. The smoothing line shows the predicted marginal effects from our noninformative Bayesian regression model along with their credible interval. The dashed line on the y-axis represents chance performance. Data are jittered slightly on the vertical axis to avoid overplotting.

…

Plot of median testing date × proportion of infants choosing Helper/Push-up character in each laboratory. Note: Proportion of infants choosing the Helper/Push-up character (over Hinderer/Push-down character) in each laboratory plotted by the laboratory's median testing date. The dashed line on the y-axis represents chance performance. The smoothing line shows the estimated effects with credible intervals. Data are jittered slightly on the vertical axis to avoid overplotting. Each point represents data from one lab.

…

Figures - uploaded by Hernando Taborda

Content may be subject to copyright.

Content uploaded by Hernando Taborda

Content may be subject to copyright.

Developmental Science

REGISTERED REPORT

Infants’ Social Evaluation of Helpers and Hinderers: A

Large-Scale, Multi-Lab, Coordinated Replication Study

Kelsey Lucca1Francis Yuen2Yiyi Wang3Nicolás Alessandroni4Olivia Allison5

Mario Alvarez1Emma L. Axelsson6Janina Baumer7Heidi A. Baumgartner8Julie Bertels9

Mitali Bhavsar10 Krista Byers-Heinlein4Arthur Capelier-Mourguy11 Hitomi Chijiiwa12

Chantelle S.-S. Chin2Natalie Christner13 Laura K. Cirelli14 John Corbit15 Moritz M. Daum16,17

Tiffany Doan14 Michaela Dresel18 Anna Exner19 Wenxi Fei20 Samuel H. Forbes21

Laura Franchin22 Michael C. Frank23 Alessandra Geraci24 Michelle Giraud25 Megan E. Gornik26

Charlotte Grosse Wiesmann27 Tobias Grossmann5Isabelle M. Hadley26 Naomi Havron28,29

AnnetteM.E.Henderson

30 Emmy Higgs Matzner31 Bailey A. Immel32 Grzegorz Jankiewicz33

Wiktoria Jędryczka33 Yasuhiro Kanakogi12 Jonathan F. Kominsky34 Casey Lew-Williams35

Zoe Liberman32 Liquan Liu36,37,38 Yilin Liu39 Miriam T. Loeffler16,17 Alia Martin18

Julien Mayor40 Xianwei Meng41 Michal Misiak42,43 David Moreau30,44 Mira L. Nencheva23,35

Linda S. Oña45,46 Yenny Otálora47 Markus Paulus13 Bill Pepe48 Charisse B. Pickron31

Lindsey J. Powell48 Marina Proft49 Alyssa A. Quinn6Hannes Rakoczy49 Peter J. Reschke50

Ronit Roth-Hanania51 Katrin Rothmaler27,52 Karola Schlegelmilch45,46 Laura Schlingloff-Nemecz34,53

Mark A. Schmuckler14 Tobias Schuwerk13 Sabine Seehagen19 Hilal H. Şen54 Munna R. Shainy23,55

Valentina Silvestri25 Melanie Soderstrom26 Jessica Sommerville56 Hyun-joo Song57

Piotr Sorokowski33 Sandro E. Stutz16,17 Yanjie Su58 Hernando Taborda-Osorio59 Alvin W. M. Tan23

Denis Tatone34 Teresa Taylor-Partridge60 Chiu Kin Adrian Tsang2,61 Arkadiusz Urbanek62

Florina Uzefovsky63 Ingmar Visser7Annie E. Wertz32,46 Madison Williams4Kristina Wolsey30

Terry Tin-Yau Wong61 Amanda M. Woodward64 Yan g Wu14 Zhen Zeng38,65 Lucie Zimmer13

J. Kiley Hamlin2

Correspondence: Kelsey Lucca (klucca@asu.edu)

Received: 12 November 2019 Revised: 26 September 2024 Accepted: 29 September 2024

Funding: Piotr Sorokowski was supported by Being Human Incubator funds; Yenny Otálora was supported by CI: 5348, funded by Universidad del Valle;

Hernando Taborda-Osorio was supported by code 00010364 Universidad Javeriana; Nicolás Alessandroni was supported by Concordia Horizon Postdoctoral

Fellowship; Hannes Rakoczy was supported by DFG RA 2155/7-2; Tobias Schuwerk was supported by DFG SCHU3060/2-1; Denis Tatone and Laura Schlingloff-

Nemecz were supported by ERC Horizon 2020 742231 awarded to Gergely Csibra; Lindsey J. Powell was supported by Hellman Fund Fellowship; Kelsey Lucca

was supported by funds from Arizona State University Department of Psychology; Mario Alvarez was supported by an APA SUPER fellowship; Florina Uzefovsky

was supported by ISF 561/18; Liquan Liu was supported by MSCA-IF-798658; Foundation of Graduates in Early Childhood Studies under the Forest Hill section of

the Trust; Yanjie Su was supported by National Natural Science Foundation of China, 32071075; Mark A. Schmuckler was supported by NSERC Discovery Grant;

Melanie Soderstrom was supported by NSERC Discovery RGPIN-2023-04285 and RGPIN-05367-2019; Tobias Grossmann was supported by NSF 2017229; Annette

M. E. Henderson was supported by PBRF grants, School of Psychology, University of Auckland; John Corbit was supported by NSERC 2023–05954; Moritz

M. Daum was supported by SNF 10001G_20768; J. Kiley Hamlin was supported by SSHRC Partnership Development Grant 890-2020-0059 and a Social Sciences

and Humanities Research Council of Canada (SSHRC) grant (12R20573); Hilal H. Şen was supported by University of Akureyri Internal Grant R2308; Jessica

Sommerville was supported by a grant from NICHD (1R01HD076949-01); Terry Tin-Yau Wong was supported by University of Hong Kong, Seed Fund for Basic

Research, 104006653; Hyun-joo Song was supported by grant NRF-2020S1A5A2A01042840; Annie E. Wertz was supported by funding from the Max Planck Society.

Kelsey Lucca and Francis Yuenshare the first authorship.

A full breakdown of CRediT contributions can be viewed here https://osf.io/qntyd/.

The authors from ManyBabies4 Consortium are listed in the byline.

Foraffiliations,refertopage24.

Developmental Science, 2025; 28:e13581

https://doi.org/10.1111/desc.13581

1of28

ABSTRACT

Evaluating whether someone’s behavior is praiseworthy or blameworthy is a fundamental human trait. A seminal study by Hamlin

and colleagues in 2007 suggested that the ability to form social evaluations based on third-party interactions emerges within

the first year of life: infants preferred a character who helped, over hindered, another who tried but failed to climb a hill. This

sparked a new line of inquiry into the origins of social evaluations; however, replication attempts have yielded mixed results. We

present a preregistered, multi-laboratory, standardized study aimed at replicating infants’ preference for Helpers over Hinderers.

We intended to (1) provide a precise estimate of the effect size of infants’ preference for Helpers over Hinderers, and (2) determine

the degree to which preferences are based on social information. Using the ManyBabies framework for big team-based science,

we tested 1018 infants (567 included, 5.5–10.5 months) from 37 labs across five continents. Overall, 49.34% of infants preferred

Helpers over Hinderers in the social condition, and 55.85% preferred characters who pushed up, versus down, an inanimate object

in the nonsocial condition; neither proportion differed from chance or from each other. This study provides evidence against

infants’ prosocial preferences in the hill paradigm, suggesting the effect size is weaker, absent, and/or develops later than previously

estimated. As the first of its kind, this study serves as a proof-of-concept for using active behavioral measures (e.g., manual choice)

in large-scale, multi-lab projects studying infants.

1 Introduction

As adults, we are quick to judge other individuals’ actions as

praiseworthy or blameworthy. These judgments have a pervasive

impact on our social interactions—we gravitate toward and

befriend individuals with a history of behaving prosocially and

avoid those with a history of behaving antisocially. Notably, our

judgments and selective social preferences are not limited to those

with whom we directly interact. Humans readily judge individu-

als on the basis of the prosocial and antisocial actions they direct

toward unrelated third parties, and even incur personal costs

to punish those who behave antisocially (Fehr and Fischbacher

2003). From where does this propensity to morally evaluate others

originate?

Historically, some scholars have contended that infants are born

either amoral, with no moral sense, or immoral, motivated solely

by selfish impulses (Freud et al. 1961; Kohlberg 1969; Piaget

1932). According to these perspectives, it is only through cognitive

development and extensive socialization that humans come to

develop an adult-like moral sense (reviewed in Brownell 2013;

Carpendale and Lewis 2004). This proposal has been recently

challenged by a series of studies on the social evaluative abilities

of preverbal infants (reviewed in Margoni and Surian 2018).

This work suggests that key precursors of full-fledged moral

competencies, such as the ability to evaluate others on the basis of

their prosocial or antisocial acts, may already be in place within

the first year after birth.

The first study to suggest that preverbal infants engage in social

evaluations presented 6- and 10-month-old infants with scenarios

in which novel characters directed prosocial and antisocial

acts toward a third party (Hamlin, Wynn, and Bloom 2007).

Specifically, infants watched a puppet show featuring a “Climber”

(a red wooden circle with googly eyes) who repeatedly tried but

failed to climb to the top of a steep hill. Infants were shown two

distinct events in alternation: a hindering event and a helping

event. During hindering events, a Hinderer character (e.g., a blue

wooden square with googly eyes) prevented the Climber from

reaching its goal by pushing it down to the bottom of the hill.

During helping events, a Helper character (e.g., a yellow wooden

triangle with googly eyes; character identity was counterbalanced

across infants) facilitated the Climber’s goal by pushing it to the

top of the hill. Helping and Hindering events were repeated until

infants reached a preset habituation criterion. Finally, infants

were presented with the Helper and the Hinderer and were

prompted to choose between the two. Infants at both 6 and 10

months of age robustly reached for the Helper over the Hinderer

(12 of 12 6-month-olds and 14 of 16 10-month-olds), suggestive of

early social evaluation.

A key control condition tested whether preferences for the Helper

were truly social in nature or whether they were based on

low-level features of the display. In this control condition, the

procedure was similar to the experimental condition, except that

the animate Climber was replaced with an inanimate (eyeless)

red circular ball that produced no self-propelled motion. On

alternating events, this ball was pushed up or down the hill by

the same square and triangle characters that played the roles of

Helper and Hinderer in the original condition. Critically, in this

condition, pushing up and down was present, but should not have

been interpreted as helping and hindering, given that the ball

did not exhibit cues of agency or goal-directedness. Here, infants

did not demonstrate a preference for either character, suggesting

that infants’ evaluations of the Helper and Hinderer in the

experimental condition were guided by the social consequences

of their actions (Hamlin, Wynn, and Bloom 2007).

In recent years, research has extended these findings across a

variety of paradigms and social scenarios. This work demon-

strated that infants, as early as 5 months, prefer characters who

helped, rather than hindered, an agent achieve different types

of goals, including opening a box containing an appealing toy,

retrieving a toy from a high shelf, and obtaining a preferred

object (Hamlin and Wynn 2011; Hamlin et al. 2013;Wooetal.

2017). Infants’ preference for prosocial individuals has also been

shown to extend to morally relevant actions beyond helping and

hindering; for example, 6–10-month-olds disprefer those who

physically batter others but prefer those who prevent others from

being battered (Kanakogi et al. 2017; see also Buon et al. 2014),

and by 12–16 months, infants demonstrate a preference for agents

who act fairly versus unfairly (e.g., distributed resources equally

2of28 Developmental Science,2025

14677687, 2025, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/desc.13581 by Kelsey Lucca - Arizona State University Acq , Wiley Online Library on [27/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

Summary

∙We examined infants’ preferences for prosocial (helping)

over antisocial (hindering) individuals through a large-

scale, multi-lab, coordinated replication study.

∙Using the ManyBabies framework for team-based science,

we tested 1018 infants (567 usable; ages 5.5–10.5 months)

from 37 labs across five continents.

∙Infants did not prefer a Helper over a Hinderer, nor did

they prefer a character who pushed up, versus pushed

down, an inanimate object.

∙This study provides evidence against infants’ prosocial

preferences in the hill paradigm, suggesting the effect

is weaker, absent, and/or develops later, than previously

estimated.

vs. unequally; Burns and Sommerville 2014; Geraci and Surian

2011; Lucca, Pospisil, and Sommerville 2018). As in the original

Helper/Hinderer study, control conditions have often helped to

rule out various low-level explanations for infants’ preferences,

suggesting that they were based on the social nature of the

interactions.

Findings suggesting that infants possess precocious social evalu-

ation capacities have led researchers to probe the replicability of

these results, as well as the underlying explanation for infants’

success in these tasks. Direct replication attempts have been met

with varying levels of success. Though some studies have found

that infants prefer prosocial individuals in manual choice and

preferential-looking paradigms (Buon et al. 2014, with 10- and

29-month-olds; Chae and Song 2018, with 6- and 10-month-olds;

Hamlin, Wynn, and Bloom 2010, with 3-month-olds; Loheide-

Niesmann, Lijster, and Hall 2020, with 24-month-olds; Scola et al.

2015; with 12–24 and 24–36-month-olds; Shimizu et al. 2018,with

15–18-month-olds), others have not (i.e., Abramson et al. 2016,

unpublished, with 9- and 18-month-olds; Cowell and Decety 2015,

with 12- to 24-month-olds; Nighbor et al. 2017,with5-to16-

month-olds; Salvadori et al. 2015, with 9-month-olds; Schlingloff,

Csibra, and Tatone 2020, with 14- to 16-month-olds; Shimizu et al.

2018, with 6-, 9-, and 12-month-olds; Vaporova and Zmyj 2020,

with 9- and 14-month-olds). These disparate findings have raised

critical questions about the robustness of the effect, as well as the

conditions under which the effect can be elicited.

Other research has focused on identifying the reason for infants’

choices in social evaluation tasks. Most notably, Scarf and col-

leagues (2012a) raised the possibility that infants’ preference for

helping characters could be explained not by a preference for

helpful over unhelpful agents, but by perceptual features that

co-occurred with the helping action in Hamlin and colleagues’

(Hamlin, Wynn, and Bloom 2007) experiment. Specifically, they

noted that the climbing character bounced after being helped,

but not after being hindered. Thus, infants may have selected

the Helper solely due to its association with a positively valenced

stimulus (i.e., the bouncing event). To examine this possibility,

Scarf and colleagues (2012a) conducted a series of experiments

with 10-month-old infants that closely matched the original study,

but with additional conditions in which bouncing actions also

occurred during hindering events. Results revealed that infants

preferred characters associated with bouncing, regardless of the

type of event (helping or hindering) in which the action occurred.

These results suggested that infants’ evaluations may be based

on physical rather than social aspects of the displays. Subsequent

research challenged this interpretation (cf. Hamlin 2015), noting

that stimuli differences may have hindered infants’ ability to

represent the Climber’s behavior as goal-directed. Nevertheless,

the study by Scarf and colleagues (2012a) highlights that infants’

evaluations in these tasks could be based on features of the

events devoid of sociomoral significance, thereby confirming the

necessity of proper control conditions in studies claiming to

examine the roots of social evaluation in infancy.

A recent meta-analysis aimed to provide an estimate of the

effect size of infants’ preference (or lack thereof) for prosocial

over antisocial characters, as well as to provide insights into

potential moderators of the effect (Margoni and Surian 2018;

see also Holvoet et al. 2016). The meta-analysis included data

from 26 studies (reporting a total of 61 effect sizes) in which a

prosocial agent was defined in various ways, including helping

(vs. hindering), giving (vs. taking), or distributing goods fairly (vs.

unfairly), and in which preference was defined either by selective

reaching or selective helping. Overall, the estimated average

proportion of infants who preferred the prosocial character was

0.68, 95% CI [0.64, 0.72].

The overall effect size did not vary as a function of age (4

to 32 months), the type of dependent variable (reaching vs.

helping), target type (foam shapes, hand puppets, or human

experimenters), or type of stimulus presentation (real events

or video displays). In contrast, infants’ preference for prosocial

characters did depend on the type of action presented. Studies

depicting giving versus taking yielded larger effect sizes than did

studies depicting helping versus hindering (although this finding

should be interpreted with caution due to the small number of

datapoints for giving vs. taking). The authors also noted higher

effect sizes in studies with small sample sizes (N=16 or fewer),

suggestive of a file-drawer problem, as well as in the laboratories

that produced the original studies on the topic (e.g., Hamlin’s),

indicative of a lab effect. Finally, they explored for evidence of

a publication bias. Although most of the examined effects were

published (N=44), several were unpublished (N=17). A common

“trim-and-fill” procedure (Duval and Tweedie 2000) provided

evidence of a publication bias (but see Carter et al. 2019). When

adjusted for publication bias, the meta-analytic estimate was

slightly smaller, revealing an average proportion of choice for the

Helper character of 0.64, 95% CI [0.60, 0.69].

The results of the meta-analysis raised several important direc-

tions for future work. Since nearly half of the studies included

in the meta-analysis were conducted by a single lab using small

sample sizes (N=16 infants per study), replication studies that

incorporate a higher diversity of labs as well as larger samples

are clearly needed. Moreover, some of the studies incorporated in

the meta-analysis were failed replications in which no preference

for prosocial individuals was observed. These failures suggest

either that the true effect size might be smaller than originally

thought (or nonexistent), and/or that slight variations in method

or population were responsible for the failures (Makel, Plucker,

and Hegarty 2012). Moreover, variability between studies can

also be partially explained by sampling error or chance alone

3of28

(see Margoni and Shepperd 2020; Stanley and Spence 2014).

Finally, the meta-analysis did not examine infants’ behavior in

control conditions, leaving open questions about whether infants’

preferences are truly based on the social aspects of prosocial and

antisocial actions.

The current study set out to address these outstanding questions

as well as provide new insights into infants’ early social evalua-

tions. Specifically, we first aimed to establish a precise estimate

of the true effect size of infants’ preference for helping over hin-

dering agents. Second, we sought to determine whether infants’

preferences are social in nature; namely, that they require that

positive and negative acts be directed toward agentic third parties

capable of goal-directed behavior. To achieve these objectives, we

conducted a large-scale, multi-site replication study of infants’

preferences for helping characters, with a preregistered method-

ological and analytical plan (https://osf.io/qntyd/). Our multi-site

replication approach provides crucial insights beyond those of

the existing systematic reviews (Holvoet et al. 2016; Margoni and

Surian 2018). First, by using a consistent methodology across

laboratories around the world, we are able to more precisely iden-

tify sources of variation in infants’ social evaluations (e.g., age,

geographic location) that go beyond variation in the stimuli and

experimental procedure, since each lab in our study will follow

the same protocol and use the same video stimuli. Second, this

approach allows us to compare infants’ preferences for characters

across social and nonsocial contexts, enabling us to measure

whether infants’ preferences are driven by social features of the

helping/hindering events versus nonsocial or perceptual aspects

of the events. Finally, given that meta-analyses in psychology

have been shown to report effect sizes approximately three times

as large as preregistered multi-lab replication projects (e.g., due

to publication bias or selective reporting), our approach will

allow us to obtain a more accurate estimate of the true effect

size of infants’ preference for Helpers (Kvarven, Strømland, and

Johannesson 2020).

We utilized the ManyBabies model of coordinated replication

efforts (Byers-Heinlein et al. 2020;Franketal.2017;The

ManyBabies Consortium 2020; see also ManyLabs: Klein et al.

2014), wherein a hypothesis of interest is chosen and explored by a

large group of interested laboratories, all following a standardized

protocol in order to create a data set larger than any one

laboratory could produce on its own. This method allows for the

exploration of both participant- and laboratory-level variables.

The ManyBabies model strives to adhere to Open Science

principles and practices. For instance, all stimuli, protocols, code,

and data are shared on open-access repositories. As opposed to

exactly reproducing previously published methodologies, group-

level decision-making is used in order to converge on a method

that provides the best possible test of a hypothesis of interest.

Laboratories from across the globe, especially from countries that

are traditionally underrepresented in developmental research,

are invited to contribute at all research stages. Whereas these

collective efforts focus on carefully deciding and standardizing

a study’s critical manipulations, individual labs still utilize their

own general research practices (e.g., recruitment strategies,

research assistant training).

To examine whether infants do indeed engage in social evaluation

of prosocial and antisocial characters, we chose to replicate the

hill study by Hamlin and colleagues (Hamlin, Wynn, and Bloom

2007). This study was selected because (1) it is the most widely

cited demonstration of infant social evaluation in the literature;

(2) it has been successfully replicated in subsequent research by

the original lab (Hamlin, Wynn, and Bloom 2010) and at least

one independent laboratory (Chae and Song 2018), but not by

others (Cowell and Decety 2015; Scarf et al. 2012a; Schlingloff,

Csibra, and Tatone 2020); (3) the effect has been reported in

studies employing multiple response measures (including prefer-

ential looking, anticipatory looking, and selective reaching) and

different presentation formats (video stimuli presented on screen

rather than live displays: Hamlin 2015). We reasoned that video

stimuli would be easier to utilize on a global scale.

Consistent with the ManyBabies goal of conducting the best

possible test of a given hypothesis as opposed to exact replications,

we made several modifications to the original Hamlin and

colleagues (Hamlin, Wynn, and Bloom 2007) paradigm. First,

as just noted, we utilized a filmed puppet show as opposed to

a live puppet show as in the original study. Using prerecorded

stimuli standardizes the stimuli presented to infants across

laboratories, thereby ensuring that any differences in results

cannot be attributed to variations in habituation events. Further,

videotaping these events rather than producing them in real-time

allowed us to, within condition, match the speed and timing of

the pushing-up and down actions along with the overall exposure

to the push-up/push-down characters down to the millisecond.

This method of presentation also increases the number of labs

eligible for participation because it substantially reduces barriers

to participation, such as financial/time constraints involved in

purchasing puppet stage materials, constructing a puppet stage,

and training researchers to execute a live puppet show. The video

stimuli used here were recorded in Hamlin’s lab and closely

matched videos used in a successful replication (Hamlin 2015)of

the live puppet show paradigm (Hamlin, Wynn, and Bloom 2007).

Because these two studies (Hamlin 2015; Hamlin, Wynn, and

Bloom 2007) differed only in the stimulus presentation modality

and found results with comparable effect sizes, we expected

that the effect size would not be moderated by this decision.

Importantly, meta-analytic results also found no moderating

effect of the modality of stimulus presentation (i.e., animations,

videotaped, and real events; Margoni and Surian 2018).

Second, our design implemented controls for perceptual dif-

ferences between helping and hindering events that were not

present in the initial Hamlin and colleagues (Hamlin, Wynn, and

Bloom 2007) study. As previously discussed, Scarf and colleagues

(2012b) argued that infants’ preferences in the hill paradigm were

due to the Climber character bouncing after being helped but not

after being hindered (but see Hamlin 2015, for evidence against

these criticisms). To avoid this issue, our study utilized videos

in which the Climber remained motionless, instead of bouncing,

upon reaching its final position.

Third, rather than including two separate age groups (6- and

10-month-olds), as in the original study, we included a single

group of infants ranging from 5.5 to 10.5 months. This age

range was selected for several reasons. First, a manual reaching

choice task can be used across this age window, allowing us to

fully standardize the task across all infants. Second, infants in

this age range demonstrate sensitivity to the causal power of

4of28 Developmental Science,2025

agents (Liu, Brooks, and Spelke 2019), and to both successful

(e.g., Woodward 1998) and failed goal-directed actions (e.g.,

Brandone and Wellman 2009; Hamlin, Hallinan, and Woodward

2008). Third, although Margoni and Surian (2018) did not find a

significant influence of age on infants’ preference for prosocial

individuals, several successful and failed replications fall within

this age range (Hamlin 2015; Hamlin et al. 2011; Salvadori et al.

2015; Scarf et al. 2012a). Thus, including this broad age range

allowed us to assess whether there are developmental changes

in infants’ preferences for prosocial others. Finally, as recruiting

participants across a broad age range is presumably easier

than recruiting within a narrow age range, we selected a wide

age window to maximize the number of laboratories able to

participate.

As in the original Hamlin and colleagues (Hamlin, Wynn, and

Bloom 2007) study, we included a nonsocial control condition

to examine whether infants’ preferences are driven by the

social aspects of helping versus hindering actions as opposed

to nonsocial perceptual features of the displays. In the control

condition, infants viewed events similar to the helping and

hindering events of the social condition, but with several notable

differences. Most critically, the Climber was replaced by an

inanimate, eyeless object that did not engage in self-propelled

motion. Specifically, infants viewed an inert red ball being pushed

up or down the hill by triangle and square agentic characters

with eyes. Based on the estimate from Margoni and Surian’s

(2018) meta-analysis, we predicted that approximately 64% of

infants would choose the Helper in the social condition where

the animate Climber, a red ball with eyes, demonstrated an

unfulfilled goal to climb the hill. We predicted that infants would

not demonstrate a preference for the character who pushed an

inanimate red ball (that had no eyes and demonstrated no goal-

directed behavior) in the nonsocial control condition. Relatedly,

we also predicted significantly greater preference for the Helper

in the social compared to the pusher-upper in the nonsocial

condition.

The social and nonsocial videos were designed to convey funda-

mentally different events—helping/hindering an animate char-

acter versus pushing an inanimate character up/down; therefore,

it was necessary that they differed in several ways. First, we had to

ensure that the ball was perceived as animate in the social videos,

and as inanimate in the nonsocial videos. To do so, social videos

included a hill-climbing action at the start, which demonstrated

the ball’s goal to go uphill. The nonsocial videos do not have this

portion of the video, since the ball is inanimate and not capable of

self-propelled motion. This difference led to the nonsocial videos

being 4.4 s shorter than the social videos. Although the timing

of the nonsocial videos could have been matched to the social

videos by introducing additional still frames and/or adding in

novel actions, we reasoned that these modifications might lead to

inattentiveness and fussiness in the nonsocial displays, insofar as

they do not add anything directly relevant to, or may even hamper

the interpretability of the push-up or push-down character’s

goals. Despite the overall length across social and nonsocial

videos, the length of the videos was nevertheless equated within

condition (i.e., both social videos are 13.3 s and both nonsocial

videos are 8.9 s), and the amount of time the Helper/Push-up and

Hinderer/Push-down characters are on stage is exactly matched

within and closely matched across conditions (social =4.7 s,

nonsocial =5.9 s). Although we do not expect these differences

in timing to impact our main results of interest, we will test for

the possible influence of these timing differences by analyzing

infants’ attention (e.g., as measured by the number of habituation

trials and the overall looking to the still frame events presented

after each video), and explore whether (a) attention differs across

conditions, and (b) differences in attention following each event

relate to differences in infants’ choices.

A second difference between the social and nonsocial conditions

involves the location of the Climber at the start of each event. In

the social condition, the Climber always has the goal to climb up a

steep hill and needs assistance in doing so. Therefore, the Climber

starts at the bottom of the hill during both helping and hindering

events (before being pushed up or down). In the nonsocial

condition, infants also see the ball being pushed up or down the

hill. But, because the ball is inanimate and not-goal directed, the

ball must begin at the bottom of the hill during pushing-up events,

but at the top of the hill during pushing-down events. Although

we could have created events where the character was trying to

climb down a hill, we thought this goal might be perceived as

relatively trivial in terms of costs (compared to climbing up a steep

hill), rendering it more difficult to determine that the Climber

needs assistance. We have no reason to hypothesize that these

differences in starting location across conditions would influence

infants’ choices.

Because the critical event across conditions is the pushing action

itself, it was important that the mechanics of the pushing actions

were as similar as possible across conditions so the characters

that interacted with the ball would appear equally agentive and

goal-directed. Both conditions featured two pushing actions, each

of which began with a forceful “smack.” In the social condition,

these smacks occur during the climbing action (i.e., on the steep

part of the hill). In the nonsocial condition, on the other hand,

the smacks occur on the flat portions of the hill. This difference

across conditions is due to the physical properties of pushing

objects: smacking an inanimate ball on the steep part of the

hill should cause the ball to roll down on its own, rendering

the second smack unnecessary. This difference in the location

of the smacks required the characters in the nonsocial condition

to stay in contact with the red circle for longer (∼3 s) than the

characters in the social condition (∼1 s). Importantly, however,

the contact time between characters within conditions is the

same. Nevertheless, as with the differences in overall timing, we

will conduct exploratory analyses to examine possible relations

between infants’ attention (e.g., as measured by looking-time

to the freeze frames presented at the end of each video and

the number of trials it took for the infants to habituate) and

subsequent choices.

2 Methods

2.1 Participation Details

2.1.0.1 Time-Frame. We circulated an open call for lab

participation after our Registered Report received an in-principle

acceptance, in April 2021. Data collection took place during a

one and a half-year window, initiating in the summer of 2022

and ending in the fall of 20231. Prior to receiving an in-principle

5of28

acceptance, we made an announcement on relevant listservs (i.e.,

Cogdevsoc and Infancy) to gauge general interest in the project,

as having a sense of the number of labs that might be interested

in participating helped us plan methodological decision-making

(i.e., whether we would have sufficient power to propose two

experimental conditions and whether labs would be using one vs.

two experimenters). At the time we submitted the Stage 1 version

of this Registered Report, 61 labs in 17 countries expressed interest

in collecting data, pledging to test a total of 1414 infants. The

obtained sample is described in the below Participants section.

A data collection log that includes the range of testing for each

participating lab can be found on OSF https://osf.io/qntyd/.

2.1.0.2 Age Distribution. Participating labs were asked to

recruit participants between 5.5 months (167 days) and 10.5

months (319 days), covering a 5-month age window.

2.1.0.3 Power. Findings from the meta-analysis by Margoni

and Surian (2018) revealed that the mean proportion of infants’

preference for the Helper was 0.64, 95% CI [0.60, 0.69]. Using

the Bayesian analysis tools described below, we computed the

number of participants necessary to have a 0.8 probability

(corresponding to 80% power) to find evidence against the null

hypothesis of no preference between Helpers and Hinderers,

indicated by a Bayes factor greater than 3 (Kruschke and Liddell

2018). In our simulations, this probability was achieved by

including 140 participants in a study with a single condition (i.e.,

the social condition). To achieve the same probability in a study

designed to detect a difference between an experimental group

that shows an effect and a control group that performs at chance

(i.e., the nonsocial control condition), our simulations revealed

that 500 participants were needed.

2.1.0.4 Lab Participation Criteria. Since the precision of

our analysis is affected by the number of labs contributing data

more than by the number of participants within each lab (Judd,

Westfall, and Kenny 2017), and since including a large number of

labs is important to the long-term objective of building a diverse

community of researchers to engage in replication projects, we

adopted a liberal inclusion criterion. We specified that labs should

recruit typically-developing infants within the specified age range

(see below for a full list of demographics collected). Sample sizes

were asked to be calculated on the basis of the total number of

infants who saw the entire video presentation, made a choice, and

did not fit any of the exclusion criteria (described in detail below),

meaning that most labs would likely need to recruit more than 16

infants. Labs were required to make sample size decisions prior to

testing, and signed a contract confirming that they would not stop

data collection based on results (e.g., whether or not an expected

effect was observed). Thus, if labs were unable to achieve their

initial goal of testing at least 16 infants, or if they tested more

participants than they had initially registered for, we included

their data.

2.1.0.5 Ethics. Prior to collecting data, all labs were

required to agree to a “Code of Conduct” where they agree to

maintain a high level of integrity and ethics in their involvement

in the project, including the explicit, detailed information about

data collection integrity in the project instruction manual (e.g.,

not making decisions about stopping/extending data collection

based on the data itself). Labs acquired their own Ethics

Review Board protocols for data collection. All central data

analyses used de-identified data. Individual video recordings

of participants were coded and stored at each individual lab.

In addition, if permitted by individual laboratories’ Ethics

protocols, participant videos were uploaded to a centralized

video library accessible by ManyBabies investigators on the

University of British Columbia’s Chinook server, and by other

co-investigators on Databrary (https://nyu.databrary.org/), a

video data library used by behavioral scientists world-wide. Since

not all laboratories had permission to share their videos, all

primary data coding and exclusion criteria were determined by

individual laboratories.

2.1.0.6 Lab Research Practices. Labs were instructed to

follow their individual protocols for training of research assistants

and to maintain the same quality standards for this study as

for other studies they conduct. Each lab completed a general

questionnaire to report on their training practices, academic

standing and experience of the experimenters involved in the

present study (e.g., volunteer, undergraduate, graduate, postdoc-

toral, professor), and protocol for greeting families. Additionally,

laboratories were required to create “lab tour videos” in which

they walked through their lab setup and experimental procedure.

These videos were uploaded to a central database.

2.1.0.7 Lab Training and Reliability. All labs were pro-

vided with a “ManyBabies4 Big Manual” that provided a detailed

overview of the experimental procedures (see https://osf.io/

qntyd/). Prior to initiating data collection, labs were required

to send three videos of the interactive part of the procedure

(i.e., the “choice phase”) to a centralized team of researchers

for approval. To be approved, the pilot videos were required

to follow the procedures outlined in an Experimenter’s Manual

located within the Big Manual linked above (e.g., timing of verbal

prompts, presentation distance, and angle of choice characters).

This review procedure was implemented to ensure labs strictly

adhered to the experimental protocol. The decision to switch

from piloting to data collection within each lab was made prior

to data collection, and no pilot data was included in the final

analyses.

To ensure standardization of coding for the primary variable of

interest (i.e., infants’ choice of puppet during the test phase), we

created a centralized training process for choice coding using a

centralized bank of videos with varying levels of coding difficulty.

Prior to collecting data, all labs were required to complete this

training with a reliability of 90%. The full protocol can be found

at https://osf.io/qntyd/. During data collection, the experimenter

recorded and coded the infants’ choice behavior. Offline, a

second researcher who was unaware of the conditions, coded

infants’ choices from video as a reliability check for the first

experimenter’s online coding. If the offline coder disagreed with

the experimenter, the offline coder’s choice was used for the

analyses. Labs were required to report the percent agreement

between the two coders in their final data reporting. The average

agreement across labs was 98% (range =80%–100%, 26/33 labs

reported 100% agreement).

Labs were not required to participate in a standardized train-

ing process for coding looking-time data because some labs

used automated techniques for coding (e.g., eye-tracking), and

6of28 Developmental Science,2025

looking-time was not the primary variable of interest in this

experiment. Infants’ looking time was coded either manually

(82.5% of infants) or with an automated eye-tracker (17.5% of

infants). Regardless, we asked all labs to conduct their own

reliability coding for looking-time data within their lab on 25% of

their sample and upload this data to the project’s data repository

(n=35 labs reported, sample size =1489). We calculated the

intercoder reliability using the intraclass coder coefficient (ICC).

The reliability was high, ICC =0.96, p<0.001, 95% CI [0.95, 0.96].

2.2 Participants

A total of 567 typically developing, full-term infants participated.

A total of 37 labs collected data, of which 34 collected data that

met all inclusion criteria and were included in data analysis

(mean laboratory sample size of babies included in data analysis

=16.68, SD =10.16, range: 1–47). Labs participated from 18

regions, including: Australia (n=25 included in analyses/44

tested), Austria (n=16/34), Belgium (n=17/23), Canada (n

=85/151), Colombia (n=18/24), Germany (n=71/134), Hong

Kong (n=29/63), Iceland (n=15/30), Israel (n=7/25), Italy (n

=19/48), Japan (n=21/27), New Zealand (n=44/73), Poland

(n=34/43), South Korea (n=3/12), Switzerland (n=19/28),

United States (n=80/131), mainland China (n=47/93), and

the Netherlands (n=17/35). The average age of infants included

in the study was 253 days (SD =42.95, range: 167–319), and

46.91% were female. An additional 451 infants were tested but

were not included in the final sample due to not having met the

inclusion or eligibility criteria (see below). This exclusion rate

(44.30%) is much higher than the original Hamlin et al. work

(13%; 4/32), but more comparable to past replication attempts

(e.g., 37.25% in Schlingloff, Csibra, and Tatone 2020). Further,

the majority of exclusions resulted from experimenter errors and

equipment errors, which highlights the complexity and difficulty

of standardizing and implementing methodology with behavioral

components at a large scale. Information on all participating

and included labs, including information on all babies tested, is

provided in Table 1.

2.3 Procedure

The approved Stage 1 protocol can be found on the Open Science

Framework (https://osf.io/qntyd/). Procedural instructions pro-

vided to each participating lab are viewable on OSF.io (https://osf.

io/qntyd/). Infants sat in front of the screen, either in an infant

seat or on their parent’s lap. Infants were randomly assigned to

one of two conditions: a social condition (N=302) or a nonsocial

control condition (N=265; described below). The conditions

differed only in the content of the video stimuli—the overall

procedure was otherwise identical across conditions. While the

video stimuli were displayed, parents were asked to close their

eyes or wear occluding glasses. Parents also received standardized

instructions on how to maintain proper positioning throughout

the experiment so that infants could fully view the display and

to avoid inadvertently biasing their infant: they were told to sit

still (e.g., act like infants’ “chairs,” try to keep them in their lap,

and hold them around the rib cage), not talk or gesture during the

experiment, and not redirect infants’ attention to the events if the

TABLE 1 Details about participating labs.

University Region N

Peking University Mainland

China

The University of Hong Kong Hong Kong 63

University of British Columbia Canada 59

University of Auckland New Zealand 50

University of Wroclaw Poland 43

University of California, Santa Barbara United States 42

University of Amsterdam The

Netherlands

Central European University (Vienna) Austria 34

University of Toronto Scarborough Canada 34

University of Manitoba Canada 33

University of Akureyri Iceland 30

Università degli Studi di

Milano-Bicocca

Italy 29

Ludwig-Maximilians-Universität

München: babylabLMU

Germany 28

University of Minnesota United States 28

University of Zurich Switzerland 28

Osaka University Japan 27

Ludwig-Maximilians-Universität

München: lmuMunich

Germany 25

Western Sydney University Australia 25

Ruhr University Bochum Germany 24

Universidad del Valle Colombia 24

University of Göttingen Germany 23

Université Libre de Bruxelles Belgium 23

Victoria University of Wellington New Zealand 23

Max Planck Institute for Human

Cognitive and Brain Sciences

Germany 22

Concordia University Canada 19

University of Newcastle Australia 19

University of Trento Italy 19

Princeton University United States 18

University of Virginia United States 16

The Academic College of Tel Aviv-Yaffo Israel 14

University of California, San Diego United States 14

Max Planck Institute for Human

Development

Germany 12

Yonsei University South Korea 12

University of Haifa Israel 11

Arizona State University United States 10

St. Francis Xavier University Canada 6

University of the Incarnate Word United States 3

Note: The Nrepresents the number of total tested participants, including those

who needed to be excluded from the analyses.

7of28

FIGURE 1 Screenshots of study stimuli. Note: Screenshots of helping (top left) and hindering (top right) events in the social condition; pushing

up actions (bottom left) and pushing down actions (bottom right) in the nonsocial condition.

infants grew uninterested. Infants were video recorded during the

entire experiment.

2.3.0.1 Apparatus. Laboratories were instructed to display

the video stimuli using the setup with which they were most

familiar (e.g., TV screen, projection screen, computer monitor).

The mean screen size was 91.94 cm diagonal (SD:57.48,range:

53.34–294.64) and screens stood at a mean distance of 94.67 cm

(SD: 45.70, range: 23.60–196.00) from infants. Laboratories were

instructed to position infants’ faces at the center of the screen, so

they were not required to look up or down to view the stimuli. To

facilitate standardization across labs and experimenter blinding,

laboratories were encouraged to use PyHab (Kominsky 2019)

to display the video stimuli. PyHab is a free software program

based in a PsychoPy environment (Kanbe 2019;Peirceetal.2019)

that randomly selects a counterbalanced order, presents stimuli,

and records looking time, all while allowing the experimenter

to remain blind to the onscreen stimuli display. Through this

procedure, the experimenter is able to watch the infant’s face,

initiate a trial by pressing a key, and register when the infant

is looking at the screen by pressing and releasing a key. Pyhab

expects the experimenter and the infant to be viewing separate

monitors, and rather than viewing what the infant is watching,

the experimenter sees a display that simply indicates whether a

trial is active or not. In this setup, it is possible for the entire

experiment to be run by a single experimenter who is unaware

of the puppets’ identities (i.e., which puppet is the Helper vs.

Hinderer). Labs that did not use PyHab were instructed to

reproduce these testing conditions as closely as possible, and

were strongly encouraged to use a second experimenter whenever

possible in order to ensure the experimenter remained naive to

the puppets’ identities. Regardless of setup, we required that the

experimenter administering the choice phase be naive to puppet

identity. Thus, if labs did not use PyHab or a second experimenter,

they were required to report how they ensured the experimenter

remained naive to the puppets’ identities.

2.3.0.2 Personnel. Labs were given the option of running

the experiment with one or two experimenters, as long as the

experimenter administering the choice phase was naive to the

puppets’ helping/hindering identities. Labs were required to

report to the central planning team any minimum academic

achievements required to participate in data collection (e.g., a

completed undergraduate degree), as well as a summary of the

training procedures that the experimenters underwent prior to

data collection.

2.3.0.3 Calibration. Prior to displaying the habituation

trials, an attention-getting stimulus (a spinning multi-colored

circle) appeared in turn on each of the four corners of the screen

and once in the center, allowing the online coder to calibrate

infants’ looking at the screen. The attention-getter moved to a

new location after the experimenter pressed a key, indicating that

the infant looked at it.

2.3.0.4 Onset of Each Trial. Each trial began with an

attention-getting stimulus at the center of the screen (e.g., a

twirling/squeaking multi-colored circle). When the experimenter

determined that the infant was fixating on the screen, the

experiment proceeded to the next trial.

2.3.0.5 Habituation Phase. The stimuli (Figure 1)usedfor

habituation were modeled off those in Hamlin and colleagues

(Hamlin, Wynn, and Bloom 2007), and are viewable at the

following link in the folder “Video Files March 2021” (https://osf.

io/qntyd/). The habituation phase of the experiment was the only

aspect of the study that differed between the social and nonsocial

conditions.

2.3.0.6 Habituation Phase: Social Condition. Stimuli

consisted of four 13.2-s, prerecorded, live-action puppet shows

displayed via video depicting either a “helping event” or a “hin-

dering event.” Each event was filmed against a white background

and featured a green hill rising from the bottom right to the top

left of the scene. The hill had two inclines, the second steeper

than the first, with a small flat plateau between them. Characters

consisted of three colored wooden shapes (i.e., a red circle, a blue

square, and a yellow triangle), each featuring googly eyes. The

red circle was always the “Climber” character who tried but failed

to climb the hill; the Climber’s eyes were fixed, pointing upward

8of28 Developmental Science,2025

so that it continuously gazed toward her goal. The blue square

and yellow triangle were the “Helper” and “Hinderer” characters;

their eyes were not fixed, and the characters’ identities were

counterbalanced across participants. Soft instrumental music

played in the background of the events to maintain infants’

interest.

For each event, the Climber was observed trying but failing to

reach the top of the hill. The Climber was first shown resting

at the bottom of the hill, at the right edge of the display. The

Climber then moved up the first mild incline to a plateau, where

it “danced” briefly. Next, the Climber attempted but failed twice

to reach the top of the second, steeper incline, moving slightly

higher but sliding back down to the plateau after each attempt.

During the Climber’s third attempt, a second agent (the Helper

or the Hinderer) entered the scene. In order to draw infants’

attention to the screen just before the critical “hitting” aspects of

the events occurred, a “ding” sound was played 300 ms prior to

the second agent’s appearance on stage. During helping events,

the Helper entered the scene from the bottom of the hill, moved

up the hill, and eventually contacted the Climber. The Helper hit

the Climber twice from below, pushing it upward to the top of

the hill. During hindering events, the Hinderer entered the scene

from the top of the hill and moved down, eventually contacting

the Climber. The Hinderer hit the Climber twice from above,

pushing it downward to the bottom of the hill. Once the Climber

reached its final destination (the top or bottom of the hill), the

Helper or Hinderer exited the stage from where they entered, and

the event paused. This last frame remained on the screen until

either the infant looked away for 2 consecutive seconds or 30 s

elapsed (Hamlin 2015). The helping and hindering events were

matched in timing, speed, and sound down to the millisecond

(see above OSF link for a comparison of videos). For example, the

timing of the initial attempts of the Climber, the first frame where

the Helper/Hinderer enters the stage, the first contact between

the Helper/Hinderer and the Climber, and the exit speed of the

Helper/Hinderer are identical across videos. All variations of the

videos used the same audio track.

To ensure that infants viewed the helping/hindering events in

their entirety, we identified a Critical Period from the first

frame the Helper or Hinderer appeared on stage to the last

frame where the Helper or Hinderer was moving (2500 ms).

If infants looked away during the Critical Period for more

than 1000 cumulative ms (i.e., 40% of the window), the trial

was terminated and repeated up to three times2. Failure to

attend the events within this Critical Period after three showings

resulted in the exclusion of the infant due to inattentiveness.

If such occurred within the first six trials, the experiment was

nevertheless continued until the 6th trial and ended immediately

after. This was done to avoid any frustration for caregivers that

could arise from ending the experiment abruptly, and to ensure

that each family had the opportunity to fully participate in the

experiment.

Helping and hindering trials were shown in alternation until

infants fulfilled the preset habituation criteria, adopted from

Hamlin and colleagues (Hamlin, Wynn, and Bloom 2007). To

fulfill these criteria, infants’ summed looking times to any

three consecutive trials had to decrease to less than half the

summed looking time to the first three trials for which looking

totaled 12 s or more. Once this criterion was met, or after a

maximum of 14 trials, infants moved on to the choice phase. The

presentation order of the videos (helping first vs. hindering first)

was counterbalanced across participants.

2.3.0.7 Habituation Phase: Nonsocial Control Condi-

tion. Infants viewed events highly similar to the helping and

hindering events of the social condition but with one critical

difference: the Climber was replaced by an inanimate, eyeless

object that did not exhibit self-propelled motion. Because there

were no initial climbing actions, videos in the nonsocial con-

dition were 4.4 s shorter in duration than those in the social

condition. In these videos, infants viewed an inert red ball get

pushed up or down the hill by the same animated triangle and

square characters from the social condition. Aside from these

differences, the videos across conditions were closely matched

on visual and sound cues. The screen time of the square/triangle

characters was closely matched across conditions (i.e., screen

time of Helper/Hinderer: 4.7 s per video; screen time of up-

pusher/down-pusher: 5.9 s per video). The critical window was

defined using the same boundaries as the social condition:

it started from the first frame when the Helper or Hinderer

appeared on stage to the last frame when the Helper or Hinderer

stopped moving (3.5 s). As in the social condition, if infants

looked away for more than 40% cumulative ms during this critical

window (i.e., 1400 ms), the trial was terminated and repeated

up to three times. The habituation criteria were identical across

conditions. Within the nonsocial condition, pushing up versus

down videos were closely matched on speed, timing, and sound

cues. The presentation order of the videos (pushing up vs. down

first) was counterbalanced across participants.

2.3.0.8 Choice Phase. Complete instructions for admin-

istering the choice phase are viewable in Appendix Aof the

Supplemental Materials. The choice phase was identical across

conditions.

Immediately following the end of the habituation phase, an exper-

imenter (naive to the identity of the characters) presented infants

with foam versions of the yellow triangle and blue square charac-

ters, attached with Velcro to a 45 cm ×60 cm board with a white

background (Figure 2). The characters were standardized in size:

blue square, 9.9 cm ×9.9 cm, and yellow triangle, 14.8 cm (base) ×

13 cm (height). Labs were provided with hyperlinks to the mate-

rial used and templates for creating the characters (see Appendix

Bfor excerpts from the links). A subset of characters was centrally

created by the project team and distributed to some participating

labs via mail or at the International Congress for Infancy Studies

conference in July 2018. Seventeen labs received characters made

by the project team. Each character was located 7.5 cm from the

base to the bottom of the board. Velcro pieces were placed at the

center of each character. Measuring from the center of the Velcro,

the pieces were 30 cm apart from each other, 15 cm from each side

of the board, with each character placed ∼9cmfromitsoutermost

point to the side edges of the board. The location of characters

(left/right) was counterbalanced across participants.

At the start of the choice phase, parents were instructed to grasp

their infants securely around the waist and position them close to

their knees to facilitate infants’ reaching. Parents were then asked

9of28

FIGURE 2 Sample image of choice phase. Note: Choice phase foam 3-D stimuli presented on a white board. Board dimensions: 45 cm ×60 cm;

Shape dimensions: 9.9 cm ×9.9 cm (blue square), 14.8 cm (base) ×13 cm (height; yellow triangle).

to close their eyes or wear occluding glasses to prevent them from

biasing the child’s attention toward a particular character.

Next, the experimenter leaned over in front of the infant and said,

“Hi! Look!” while lowering the board directly at an approximately

30-degree angle. The board was placed just out of the infant’s

reach until the infant looked at both characters. When the infant

had done so, the experimenter said, “Hi!” to direct the infant’s

attention away from the characters and back to the experimenter.

Upon making eye contact with the infant, the experimenter said,

“Who do you like?” while moving the board within the infant’s

reach. The experimenter kept the board extended toward the

child for 60 s, while keeping track of time either in their head,

using a stopwatch, or by referencing a wall clock. Any choices

that were made after 60 s were excluded from the analyses (e.g.,

in the case the experimenter inadvertently extended the choice

phase past 60 s).

Infants’ choice of character was coded online by the experimenter

conducting the choice phase. A visually guided reach (touching

one character while looking at it) was indicative of infants’

choices. Occasionally, infants touched both puppets at once;

some of these instances were counted as valid choices according

to our predefined criteria. In usable both touches, the infant

clearly directed her gaze and reached toward one character but

touched the other character as well. On the other hand, unusable

both touches involved the infant touching both characters with

unclear or inconsistent visual attention. Experimenters were

instructed to continue the choice procedure until a usable touch

was recorded by encouraging and reprompting the baby after

30 s had passed. If infants did not make a choice during the 60

s choice phase, their data were excluded from final analyses.

Experimenters were required to administer the choice phase

within 2 min of ending the habituation videos. If more than 2

min passed, the session was considered an “experimenter error”

and excluded from data analysis.

2.3.0.9 Order of Testing. Laboratories occasionally tested

participants in a separate (unrelated) experiment during their

visit. We encouraged, but did not require, labs to run the

Helper/Hinderer experiment as their first experiment. All

labs recorded whether another experiment was run with the

same participants and whether it preceded or followed the

Helper/Hinderer experiment.

2.3.0.10 Demographics. Each lab collected a set of partic-

ipant demographic information for each infant: gender, date of

birth, estimated proportion of language exposure for language(s)

heard daily, preterm/full-term status (i.e., more than 36 weeks

gestation), hearing or visual impairments, developmental con-

cerns (e.g., developmental disorders), infant handedness (right,

left, not sure), parent handedness (right, left), and color blindness

in the immediate family. Labs were given a standard participant

questionnaire that they were encouraged to use (see Appendix C

of the Supplemental Materials).

2.4 Exclusion Criteria

All data collected for the study (i.e., every infant for whom a data

file was generated, regardless of how many trials were completed)

were uploaded to the central database for analysis. We instructed

10 of 28 Developmental Science,2025

labs to make note of any instances in which a procedural error

or anomaly occurred during testing. Participants were excluded

from the final analysis if they met any of the criteria below.

2.4.1 Eligibility Criteria

1. Full-Term. Full term was defined as 36 weeks or more

gestation. Caregivers were asked to report their child’s due

date, and a centralized research team calculated the child’s

gestational age. Ten (1.0%) of tested infants were excluded for

not meeting this criterion.

2. No Diagnosed Developmental Disorders. If parents reported

any known developmental disorders, their infants were

excluded. One (0.1%) of the tested infants were excluded for

not meeting this criterion.

3. Within Age Range. Infants were excluded if they were older or

younger than our target age range. Fifty-four (5.3%) of tested

infants were excluded for not meeting this criterion. Note that

this category was added after the Stage 1 submission when we

conducted an initial check for the data uploaded by labs.

2.4.2 Experimental Exclusion Criteria

4. Failure to Set a Habituation Criterion. Infants set the habitu-

ation criterion if their looking time toward the paused frame

at the end of the habituation video on any 3 consecutive

trials during the first 6 trials summed to at least 12 s. If an

infant did not set a habituation criterion in the first 6 trials,

the study terminated, and their data was excluded. A total of

142 (14.0%) of tested infants were excluded for not setting a

habituation criterion.

5. Failure to View the Critical Period. If infants looked away for

more than 750 ms cumulatively during the Critical Period

of a trial, the experimenter repeated the trial up to two

times. If infants failed to attend a Critical Period after three

consecutive trials, they were excluded. A total of 126 (12.4%)

of tested infants were excluded for this criterion.

6. Unclear Choice. Infants must have produced visually

guided reaches (as defined in the Supplementary Materials,

Appendix A) toward one of the two characters in the choice

phase. A total of 151 (14.8%) of the tested infants were

excluded for not meeting this criterion.

7. Parental/Outside Interference. Infants were excluded if dis-

tracting events that were not part of the study protocol

occurred during the Habituation or choice phase (e.g., a

noise outside the testing room, the parent gesturing to the

child). Thirty-nine (3.8%) of tested infants were excluded for

parental interference.

8. Experimenter Error. Experimenter errors included any

actions the experimenter inadvertently made that may have

influenced infants’ behavior (e.g., the experimenter failed to

record an infant’s looking time, failed to repeat the Critical

Period after infants missed viewing it, or became aware of

puppets’ identity prior to the choice phase). A total of 133

(13.1%) of tested infants were excluded for this criterion.

9. Equipment Error. Any technical deviations that may have

influenced infants’ behavior were considered equipment

errors (e.g., stimulus froze during presentation, software did

not accurately set habituation criteria). A total of 232 (22.8%)

of tested infants were excluded for equipment error.

10. Fussiness. Infants became too fussy to finish the experiment

(e.g., by showing visible signs of distress, fussiness, or

crying). Twenty-nine (2.9%) of tested infants were excluded

for fussiness. Note that this category was added after the

Stage 1 submission, but prior to the onset of data collection.

The training manual defined fussiness as follows: “Infant

becomes visibly distressed to the point where the experiment

cannot continue. Note that this decision must be made by

the experimenter and not the parent.”

11. Data Entry Error. Labs entered the data incorrectly while

reporting the experiment condition or infants’ choices.

Thirteen (1.3%) of tested infants were excluded for this

criterion. Note that this category was added after the Stage

1 submission when we conducted a data entry check for the

data uploaded by labs.

Exclusion based on the reasons above led to 451 total excluded

and ineligible participants (44.30%). The categories above are not

mutually exclusive (e.g., parental influence may have co-occurred

with experimenter error), so the numbers in the individual cate-

gories listed above will total to more than 451. If we do not con-

sider ineligible infants (i.e., infants who were tested but should

not have been due to their age, preterm status, or diagnosed devel-

opmental disorder), the exclusion rate is 40.63%. Labs reported

their exclusion rates and reasons, the central project team

also manually checked the data for any missed exclusions and

removed them from the final data analysis. Of the eligible infants

tested by labs, the average (unweighted) exclusion rate across labs

was 43.70% (SD =23.03%), with a range between 10% and 100%.

3Results

Note: Pilot data are reported in Supplementary Materials only

(see Appendix D for details on pilot methods and Appendix E

for pilot ananalyses). Pilot data only include infants in the social

condition, as the decision to run a nonsocial control condition

was made during the peer review process, after initial piloting

was complete. We made a few minor modifications to the social

condition videos used in the final data collection so they could

be more closely matched to the nonsocial videos and more closely

match the original Hamlin, Wynn, and Bloom (2007) videos. Data

simulations with the expected effect size were used to confirm the

validity of our analysis structure to test the effects of interest. All

analysis scripts, data, including simulations for Bayesian power

analyses and choices for the statistical plan, can be found online

(https://github.com/manybabies/mb4-analysis)and(https://osf.

io/qntyd/). Bayesian power analysis was conducted in a similar

fashion to frequentist power analysis, with the difference that power

was estimated as the number of samples in which Bayes factors

(BFs) were greater than the threshold value 3.

3.1 Analysis Structure

We used a Bayesian analysis framework to calculate BFsthro-

ugh Bayesian model comparisons. This framework provides

11 of 28

a principled method for testing evidence in favor of and against

the null hypothesis (i.e., no difference in preference for Helpers

over Hinderers), and allows us to fit hierarchical models to

account for sparse data and variation across labs.

We fit generalized linear mixed-effects Bayesian models using

Stan (Gelman, Lee, and Guo 2015)andthebrmspackage(version

2.20.4; Bürkner 2017); 95% credible intervals (CI) were computed

using the HPDInterval function from the coda package (version

0.19.4; Plummer et al. 2006). While standard frequentist 95%

Confidence Intervals are defined such that if repeated samples

are drawn from a population, the Confidence Interval will

include the true population mean in 95% of the observed

samples, Bayesian credible intervals represent the interval in

which the true population mean is likely to be, given the data at

hand and the model assumptions (Morey et al. 2016).

We were also interested in assessing the evidence for specific

hypotheses (e.g., a nonzero preference for Helpers over Hinder-

ers). While Bayesian CIs allow us to assess the precision of our

measurement of effects, they do not allow for the assessment

of evidence in this way. For this purpose, BFs are the appro-

priate Bayesian tool. Crucially, while pvalues only allow us

to reject the null hypothesis H0,BFs obtained through model

comparison allow us to either reject H0in favor of H1, accept

H0to the detriment of H1, or conclude that the data do not

provide sufficient evidence to support either. To obtain BFs, we

computed two models for each specific research question, one

representing the null hypothesis (H0) and the other an alternative

hypothesis (H1). We then estimated the posterior distributions

of each model and compared them to obtain a BF, using the

bridgesampling package (version 1.1.2; Gronau, Singmann, and

Wagen makers 2017).

3.2 Confirmatory Analyses

Our primary research questions were (1) whether there was

an above-chance choice of the Helper in the social condition

compared to the choice of the Push-Up Character in the nonsocial

control condition, and (2) whether there were developmental

differences in choice. Based on the original study (Hamlin, Wynn,

and Bloom 2007), we predicted that the proportion of children

choosing the Helper would be above chance.

In our Stage 1 Registered Report, we proposed two analysis plans

with power analyses computed for each plan based on the final

sample size (reported below). In the event that we did not have

enough participants to reach 80% power to compare the social

to the nonsocial condition (nmin =500), we planned to revert

to testing the social condition only against chance (nmin =140).

Below, we describe the analysis plan with a nonsocial control

condition. If our sample size had only been sufficient for compar-

ing the choices of infants in the social condition against chance

performance, the same procedure would have been used,

removing the main effect of the condition from the model, and

instead testing the intercept. Our final sample size was 567,

which is sufficient for achieving 80% power to compare the social

to the nonsocial condition, so we used the below analysis and

included the main effect of the condition and its interaction with

age in the model.

We tested our prediction via a Bayesian generalized linear mixed

effects model with a Bernoulli response model. In such models,

the probability for the dependent variable (here, participants’

choice) is transformed through the logit function such that an

estimate of zero corresponds to a 50% chance for either choice,

with values greater than zero representing a preference for the

Helper. The specification of our model was:

choice ∼ 1 +condition +age +condition ∶ age +

(1+condition +age +condition ∶ age |lab)

Choice (Helper vs. Hinderer, or Push-Up vs. Push-Down charac-

ter) and condition (social vs. nonsocial) were entered as binary

variables. Age (in days) was scaled and centered to allow for better

convergence of the statistical models and easier interpretation of

the results. This type of variable scaling is standard (e.g., Mar-

quardt 1980), and we adopt it because (a) keeping raw age in days

would result in a meaningless intercept and main effect of con-

dition (representing a hypothetical preference for either group at

age =0), and (b) estimating an intercept and a difference between

conditions at age =0 as well as a slope for age would lead to

less precise estimates (as we found in simulations). To control for

possible variation in preference across labs as well as in develop-

mental trends across labs, the model also included random inter-

cepts for each lab and random slopes for condition, age, and their

interaction, by lab. For the effect of condition, we used treatment

contrasts with the non-social condition as the reference level.

In Bayesian analyses, the most appropriate method for pooling

data from a new experiment with previous knowledge is to set

priors based on that previous knowledge. For our key effect of

interest, namely the effect of condition (a higher or lower pro-

portion of Helper choices in the social vs. nonsocial condition),

we used an informative normal prior based on the effect size

and Confidence Interval from the meta-analysis of Margoni and

Surian (2018). We used the following formula to compute the

standard deviation from the reported Confidence Interval (𝜎=

𝐶𝐼𝑜𝑛𝑒−𝑠𝑖𝑑𝑒 𝑤𝑖𝑑𝑡ℎ ×√𝑁𝑚𝑒𝑎𝑛∕𝑧(95); here, M=logit(0.64) =0.58,

SD =0.1). We believe that this choice was warranted given

that the meta-analysis included both studies that showed and

did not show the hypothesized effect, as well as published and

unpublished data, making them unlikely to reflect the outcome

of a biased literature selection toward the effect of interest.

Further, the estimate of 0.64 represents the proportion of infants

choosing prosocial characters after having corrected for possible

publication bias. Sensitivity analyses using noninformative priors

are presented alongside our results (hereafter referred to as the

noninformative model, as compared to our main model, the

informative model). For these analyses, we only specify a prior

on the random effects to improve model convergence.

We further used weakly informative normal priors for the

intercept (M=0, SD =0.1), relationship of age to choice and

its interaction by condition (M=0, SD =0.5), and a restricted

Student prior on random effects (with df =3, M=0, and SD

=2, as compared to the default in the brms package). Priors

of this type provide very little information in cases like ours

where we have large amounts of data; their primary purpose is

to improve model convergence and subsequent bridge sampling

for the computation of BFs.

12 of 28 Developmental Science,2025

FIGURE 3 Age ×Choice plot. Note: Probability of choosing the Helper/Push-up character (over Hinderer/Push-down character) across ages. The

smoothing line shows the predicted marginal effects from our noninformative Bayesian regression model along with their credible interval. The dashed

line on the y-axis represents chance performance. Data are jittered slightly on the vertical axis to avoid overplotting.

Relative to the first research question, the main finding we aimed

to replicate was infants’ preference for the Helper at levels greater

than chance, as reflected by a greater than zero estimate for the

effect of the condition. To assess the evidence for this hypothesis,

we computed the BF in favor of the full model described above

(H1) compared to a model that did not include a main effect of

the condition or any higher-level terms including this effect (H0).

Given that random effects are only interpretable with respect

to the corresponding fixed effects for nested model comparisons

(Stroup 2012), random effects corresponding to the dropped fixed

effects were also dropped3.Moreprecisely,themodelforH0was

specified as:

choice ∼ 1 +age+(1+age |lab)

Following Kass and Raftery (1995), we interpreted a BF >3infavor

of either H0or H1as substantial evidence and a BF >10 as strong

evidence4.

For all preregistered main models, models were checked for

divergent transitions, ˆ

𝑅, and effective sample sizes; all models

converged properly, indicated by no divergent transitions, an

effective sample size greater than 10% of the total sample size,

and 

𝑅<1.1. Since dispersion issues can arise with these models,

they were also checked for dispersions and posterior predictions

visualized against the data.

We found a BF01 >100 in the informative model and a BF01 =

10.23 in the noninformative model, providing further evidence

that infants’ choices were not influenced by condition or the

condition-by-age interaction. In other words, infants did not

preferentially choose the Helper over the Hinderer more often

in the social condition than in the nonsocial condition, which

did not replicate previous findings (e.g., Hamlin 2015, and Bloom

2007). In the social condition, 49.34% of infants chose the Helper.

In the non-social condition, 55.85% of infants chose the Helper

(See Figures 3and 4). Parameter estimates for this model with and

without informative priors are reported in Table 2and Table 3,

along with estimated error and 95% credible intervals (CI) for

those parameters. It is important to note that the 95% CI for the

condition parameter did not overlap with zero in the informative

model (b=0.42, SE =0.09, 95% CI [0.24, 0.61]). However, given

aBF01 >100 for the informative model, this parameter estimate

alone cannot support the effect of condition on infants’ choices

(i.e., Type II conflict, Lovric 2019). It is likely that this effect

was driven by our strong priors with a relatively small standard

deviation of 0.1.

Relative to the second research question concerning developmen-

tal differences in infants’ choice, we were interested in whether

the addition of the condition-by-age interaction to the model

contributed to model fit. We assessed evidence for this question

via the same procedure as above, fitting a null model without the

condition-by-age interaction:

choice ∼ 1 +condition +age +(1+condition +age |lab)

The BFs for this comparison was BF01 =8.40 in the informative

model and BF01 =7.35 in the noninformative model, suggesting

that the condition-by-age interaction did not have an effect on

infants’ choices (informative: b=0.11, SE =0.20, 95% CI [−0.28,

0.49]; noninformative: b=0.20, SE =0.20, 95% CI [−0.20, 0.59]).

In other words, infants’ choices were not influenced by condition

or age.

As a follow-up analysis, we refitted the same model, including

only infants who successfully habituated to the events (n=

441, 77.78% of babies successfully habituated). We again fitted a

Bayesian Bernoulli linear mixed effects model. Choice (Helper

vs. Hinderer) and condition (social vs. nonsocial) were entered

as binary variables. Age (in days) was scaled and centered. To

control for possible variation in preference across labs as well

as in developmental trends across labs, the model also included

random intercepts for each lab and random slopes for condition

and age by lab. Priors were as above. We used the same model

comparisons and the same method to obtain BFs. We found

aBF01 >100 in the informative model and a BF10 =1.45 in

the noninformative model, suggesting that condition and the

13 of 28

FIGURE 4 Forest plot of character choice for all contributing labs. Note: “Forest plot” of estimates for proportion of participants selecting the

Helper/Push-up character (over Hinderer/Push-down character) for each contributing lab. The dashed line at 0.50 on the x-axis represents chance

performance. Error bars represent 95% Bayesian credible intervals.

TABLE 2 Bayesian model parameter estimates—Informative pri-

ors.

Parameter Estimate Est. error 95% CI

Intercept −0.11 0.1 [−0.3, 0.08]

Condition 0.42 0.1 [0.23, 0.61]

z_age_days −0.12 0.16 [−0.44, 0.19]

condition:z_age_days 0.17 0.22 [−0.28, 0.60]

Note: Parameter estimates along with estimated error and 95% credible

intervals (CI) for those parameters for the full Bayesian model with informative

priors.

TABLE 3 Parameter estimates for Bayesian

models—Noninformative priors.

Parameter Estimate Est. Error 95% CI

Intercept 0.45 0.17 [0.12, 0.77]

Condition −0.49 0.23 [−0.95, −0.02]

z_age_days −0.23 0.18 [−0.58, 0.13]

condition:z_age_days 0.29 0.24 [−0.19, 0.74]

Note: Parameter estimates along with estimated error and 95% credible inter-

vals (CI) for those parameters for the full Bayesian model with noninformative

priors.

condition-by-age interaction did not have an effect on infants’

choices. In other words, habituated infants did not preferentially

choose the Helper over the Hinderer more often in the social

condition than in the nonsocial condition, failing to replicate

previous findings. In the social condition, 47.88% of habituated

infants chose the Helper. In the non-social condition, 59.51% of

habituated infants chose the Helper.

Additionally, we were interested in variation in infants’ choice

behavior across labs as, descriptively, effect sizes from individual

labs greatly varied (see Figure 4). Following Klein and colleagues

(Klein et al. 2014), we calculated the binary intraclass correlation

coefficient (ICC) using the ICCbin package in R (version 1.1.1;

Chakraborty and Hossain 2018). This measure captures the

degree to which the proportion of Helper choices was correlated

across labs by comparing participant-level deviations from the

within-lab mean to the lab-level deviations from the between-

lab mean. There was a low intra-class correlation of effects

across labs (ANOVA Estimate ICC =0.02, Smith’s Large Sample

Confidence Interval [0, 0.06], excluding one lab with only one

data point), suggesting that the within-lab variance is much

greater than the across-lab variance. Thus, it is unclear whether

the observed variations in effect size reflect genuine cross-lab

differences, or are a product of random variability inherent to

complex behavioral paradigms.

Finally, we completed all confirmatory analyses using a fre-

quentist approach to ensure their consistency with the Bayesian

approach. Overall, the findings aligned with those reported above.

Because the frequentist analyses were not included in our Stage 1

Registered Report, outputs from the frequentist analyses (includ-

ing a summary table of the parameter estimates), additional

information regarding how the frequentist analyses were carried

out, and any discrepancies from the Bayesian analyses5,canbe

14 of 28 Developmental Science,2025

FIGURE 5 Plot of habituation status ×Probability of choosing Helper/Push-up character. Note: Probability of choosing the Helper/Push-up

character (over Hinderer/Push-down character) based on infants’ habituation status (i.e., whether they successfully habituated or not). The dashed

line on the y-axis represents chance performance. The smoothing line shows the estimated effects with credible intervals. Data are jittered slightly on

the vertical axis to avoid overplotting.

found in an annotated code script, available on GitHub (https://

github.com/manybabies/mb4-analysis).

3.3 Exploratory Analyses on Infants’ Choices

An additional set of exploratory analyses was conducted. In our

Stage 1 Registered Report we included the following exploratory

analyses to test potential moderators of infants’ preference for

Helpers and Hinderers: (1) attention to the video events (i.e., as

measured by the number of trials to habituation, mean looking

time following the pushing events), (2) clear versus ambiguous

choice actions (i.e., whether infants touched both characters

during the choice phase), and (3) whether the experimenter

had knowledge of the participant’s condition (i.e., whether the

experimenter administering the choice phase knew whether the

infant participated in the social vs. nonsocial condition).

We also conducted a set of exploratory analyses that were not

included in our Registered Report, using the following modera-

tors: (1) experiment-level factors, including: the order of helping vs.

hindering videos, Helper identity (yellow vs. blue shape), Helper

side during the choice phase (right vs. left), infants’ visual angle

(calculated from infants’ distance from the screen and screen

size), whether the experiment was conducted as the first session

of the infant’s visit or not, whether the experimenter wore a mask

or not, (2) child-level characteristics, including: habituation status

(i.e., whether infants reached the habituation criterion prior to

the choice phase), handedness, color blindness, participant sex,

the percentage of primary language exposure, and (3) lab-level

factors, including: the lab’s overall exclusion rate (not including

participants who were tested but were ineligible, e.g., due to age),

the exclusion rate due to failure to make a choice for each lab,

and the median dates of data collection of each lab (transformed

into the number of days between the median date and the earliest

median date). Most factors were selected based on theoretical rel-

evance and interest. Others, especially experiment and lab-level

factors, were selected to shed light on potential sources of cross-

lab variation. The median testing date was analyzed to explore

potential cohort effects as a result of the COVID-19 pandemic6.

For each moderator, we fitted a noninformative maximal model

by including both the main effect of the moderator and its

interaction with the condition:

choice ∼ 1 +condition +moderator +condition ∶ moderator +

(1 +condition +moderator +condition ∶ moderator |lab)

We no longer used informative models here because previous

studies provided little evidence on setting priors. We used

treatment contrasts for all categorical variables.7

There was a main effect of habituation status, such that infants

who habituated were more likely to choose the Helper/Push-Up

character (53.29% chose the Helper/Push-Up character) than

infants who did not habituate (49.21%; b=0.70, SE =0.32, 95%

CI [0.09, 1.33]). However, this main effect was qualified by an

interaction between condition and habituation status, BF10 =

4.25 (b=−1.00, SE =0.45, 95% CI [−1.87, −0.13]). To unpack

the interaction, we examined the effect of habituation status on

infants’ choices in the social and nonsocial conditions separately.

We found no evidence in favor of the effect of habituation status

in either condition, BF01 =2.56 (social condition) and BF10 =1.77

(nonsocial condition). However, when we examined the effect

of condition on infants’ choice of the Helper in nonhabituated

and habituated infants separately, we found moderate evidence

in favor of the effect of condition for habituated infants, BF10 =

4.34 (b=−0.48, SE =0.22, 95% CI [−0.92, −0.06], See Figure 5)8.

Habituated infants in the nonsocial condition were more likely to

choose the Helper/Push-Up Character (59.51%) than habituated

infants in the social condition (47.88%). For nonhabituated

infants, we found no evidence in favor of the effect of condition,

BF10 =1.14. Nonhabituated infants’ choices of the Helper/Push-

Up Character in the nonsocial condition (43.33%) and the social

condition (54.55%) were not different.

15 of 28

FIGURE 6 Plot of median testing date ×proportion of infants choosing Helper/Push-up character in each laboratory. Note: Proportion of infants

choosing the Helper/Push-up character (over Hinderer/Push-down character) in each laboratory plotted by the laboratory’s median testing date. The

dashed line on the y-axis represents chance performance. The smoothing line shows the estimated effects with credible intervals. Data are jittered slightly

on the vertical axis to avoid overplotting. Each point represents data from one lab.

There was also a main effect of laboratory median testing date,

such that infants from labs with later median testing dates were

more likely to choose the Helper/Push-Up character, BF10>100

(b=0.39, SE =0.10, 95% CI [0.01, 0.77]). However, the critical

interaction between median testing date and condition did not

have an effect on infants’ choices, BF01 =3.70 (b=−0.29, SE =

0.27, 95% CI [−0.82, 0.25], See Figure 6). All the other moderators

and interaction effects had CIs that overlapped with zero (for

details of the model parameter estimates, see Table 4).

3.4 Exploratory Analyses on Infants’ Looking

Time

The above exploratory analyses focused on the effect of condition

and possible moderators on infants’ choices of the Helper. We

also explored possible attentional differences across the social

and nonsocial conditions. We conducted exploratory analyses on

infants’ gaze behavior during habituation, in particular how long

they looked at the display following the pushing events (i.e., at

the still image of the final frame presented after each video). To

compare infants’ looking time following the pushing events in the

two conditions, we used the below model comparisons and the

same method as above to obtain BFs. Noninformative priors were

used in the models.

The model for H1was specified as:

l𝑜𝑜𝑘𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑓𝑜𝑙𝑙𝑜𝑤𝑖𝑛𝑔 𝑝𝑢𝑠ℎ𝑖𝑛𝑔 𝑒𝑣𝑒𝑛𝑡𝑠 ∼ 1+𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 +𝑎𝑔𝑒

+𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛 ∶𝑎𝑔𝑒 +(1+condition +age +condition ∶ age |lab)

The model for H0was specified as:

looking time following pushing ev ents ∼ 1 +age +

(1+age |lab)

We found a BF10 =111.20 in favor of H1, providing strong evidence

for the effects of condition and the interaction between age and

condition on infants’ looking time. The 95% CI for the condition

parameter did not overlap with zero, b=0.97, SE =0.34, 95%

CI [0.29, 1.62]. In other words, infants looked longer following

the pushing events in the social condition (M=8.99, SD =

4.20) than in the nonsocial condition (M=8.02, SD =3.58),

suggesting infants were more attentive following the social videos

relative to the nonsocial videos. Furthermore, we examined

whether the addition of the condition-by-age interaction to the

model contributed to model fit. The H1model was the same

as above.

The model for H0was specified as:

looking time following pushing ev ents ∼ 1 +age +

condition +(1+age +condition |lab)

We found a BF10 =3.33 in favor of H1, providing weak evidence

for the effect of the interaction between age and condition on

infants’ looking time (b=−0.69, SE =0.33, 95% CI [−1.35,

−0.05]). Furthermore, we unpacked the interaction and analyzed

the effect of condition for younger infants (i.e., 1 SD below the

mean age) and older infants (i.e., 1 SD above the mean age). For

younger infants, we found strong evidence in favor of the effect

of the condition, BF10 =40.03. Younger infants looked longer

following the pushing events in the social condition (M=10.10,

SD =4.83) than in the nonsocial condition (M=8.19, SD =

3.69), b=1.80, SE =0.83, 95% CI [0.17, 3.43]. For older infants,

we found no evidence in favor of the effect of the condition,

BF01 =1.75. We also examined the effect of age on infants’

looking time in the social and nonsocial conditions separately.

In the social condition, we found strong evidence in favor of

the effect of age, BF10 =358.87. Infants’ looking time following

the social events decreased with age, b=−0.88, SE =0.27, 95%

CI [−1.43, −0.34] (see Figure 7). In the non-social condition,

16 of 28 Developmental Science,2025

TABLE 4 Parameter estimates for Bayesian models—Exploratory moderators.

Moderator Parameter Estimate Est. error 95% CI

Number of trials to

habituation

Intercept −0.01 0.55 [−1.08, 1.08]

Condition −0.91 0.76 [−2.46, 0.52]

hab_trial 0.05 0.06 [−0.07, 0.16]

condition:hab_trial 0.05 0.08 [−0.11, 0.21]

Looking time following

pushing events

Intercept −0.12 0.34 [−0.79, 0.53]

Condition 0.15 0.46 [−0.75, 1.06]

Looking 0.05 0.04 [−0.04, 0.12]

condition:looking −0.05 0.05 [−0.15, 0.06]

Clear versus ambiguous

choice

Intercept 0.23 0.15 [−0.07, 0.51]

Condition −0.23 0.21 [−0.63, 0.19]

touch_both 0.17 0.41 [−0.62, 0.98]

condition:touch_both −0.25 0.69 [−1.60, 1.12]

Experimenter blindness Intercept 0.27 0.56 [−0.80, 1.37]

Condition 0.38 0.77 [−1.12, 1.91]

Blindness −0.02 0.58 [−1.18, 1.09]

condition:blindness −0.68 0.80 [−2.24, 0.89]

Order of helping versus

hindering events

Intercept 0.30 0.18 [−0.07, 0.65]

Condition −0.10 0.26 [−0.61, 0.40]

Order −0.11 0.27 [−0.65, 0.41]

condition:order −0.32 0.36 [−1.02, 0.40]

Helper color Intercept 0.34 0.18 [0.00, 0.69]

Condition −0.32 0.26 [−0.84, 0.17]

helper_color −0.22 0.27 [−0.75, 0.30]

condition:helper_color 0.17 0.37 [−0.55, 0.90]

Helper side during choice Intercept 0.26 0.19 [−0.11, 0.63]

Condition −0.24 0.29 [−0.82, 0.33]

choice_side 0.01 0.31 [−0.60, 0.61]

condition:choice_side −0.09 0.39 [−0.81, 0.71]

Infants’ visual angle to the

display

Intercept 0.36 0.48 [−0.53, 1.33]

Condition −0.67 0.64 [−1.93, 0.57]

v_angle 0.00 0.01 [−0.02, 0.01]

condition:v_angle 0.01 0.01 [−0.02, 0.03]

Whether the experiment

was the first session of

infant’s visit

Intercept 0.21 0.14 [−0.06, 0.48]

Condition −0.22 0.20 [−0.61, 0.17]

first_sess 0.89 1.17 [−1.39, 3.31]

condition:first_sess −1.16 1.82 [−4.77, 2.41]

Whetherornotthe

experimenter wore a mask

Intercept 0.31 0.16 [−0.01, 0.62]

Condition −0.26 0.23 [−0.71, 0.18]

Mask −0.23 0.38 [−1.00, 0.51]

condition:mask 0.18 0.59 [−0.98, 1.32]

Infants’ handedness Intercept −0.05 0.24 [−0.53, 0.42]

Condition −0.01 0.35 [−0.71, 0.66]

(Continues)

17 of 28

TABLE 4 (Continued)

Moderator Parameter Estimate Est. error 95% CI

HandednessL 0.38 1.86 [−2.76, 4.29]

HandednessR 0.36 0.43 [−0.47, 1.23]

HandednessU 0.49 0.31 [−0.11, 1.09]

condition:handednessL −1.92 2.80 [−7.26, 2.41]

condition:handednessR −0.21 0.59 [−1.40, 0.9]

condition:handednessU −0.50 0.45 [−1.39, 0.37]

History of colorblindness in

infant’s family

Intercept 0.23 0.15 [−0.06, 0.52]

Condition −0.34 0.21 [−0.74, 0.09]

Colorblind 0.12 0.74 [−1.23, 1.70]

condition:colorblind −0.06 1.09 [−2.23, 1.94]

Infant’s sex Intercept 0.24 0.20 [−0.16, 0.63]

Condition −0.19 0.27 [−0.72, 0.36]

Sex 0.02 0.27 [−0.53, 0.52]

condition:sex −0.14 0.39 [−0.90, 0.62]

Percentage of primary

language exposure

Intercept −0.19 0.93 [−2.00, 1.65]

Condition −1.25 1.20 [−3.58, 1.13]

lang_exp 0.00 0.01 [−0.01, 0.02]

condition:lang_exp 0.01 0.01 [−0.01, 0.04]

Laboratory’s overall

exclusion rate

Intercept 0.88 0.41 [0.10, 1.70]

Condition −0.18 0.58 [−1.25, 1.01]

ex_rate −1.80 1.11 [−3.94, 0.44]

condition:ex_rate −0.19 1.56 [−3.31, 2.84]

The exclusion rate due to

failure to make a choice for

each lab

Intercept 0.59 0.24 [0.10, 1.06]

Condition −0.06 0.34 [−0.71, 0.61]

no_choice −2.70 1.58 [−5.90, 0.40]

condition:no_choice −1.53 2.21 [−5.92, 2.78]

Median data collection date

of laboratories

Intercept 0.34 0.17 [−0.01, 0.67]

Condition −0.35 0.25 [−0.83, 0.15]

median_date 0.39 0.20 [0.00, 0.78]

condition:median_date −0.29 0.27 [−0.84, 0.23]

Note: Parameter estimates along with estimated error and 95% credible intervals (CI) for those parameters for Bayesian models testing exploratory moderators.

we found no evidence in favor of the effect of age, BF01 =6.67

(Figure 7).

Following the approach taken for confirmatory analyses, we

completed all exploratory analyses using a frequentist approach.

The results are highly consistent with those obtained through

the Bayesian approach. Any discrepancies are documented

in the code script available on GitHub (https://github.com/

manybabies/mb4-analysis).

4 Discussion

The extent to which infants make early social evaluations and

prefer prosocial over antisocial agents continues to be a highly

studied and frequently debated topic. However, methodological

and procedural differences between studies render it difficult to

directly compare findings. The present international, large-scale,

and multi-site replication—using the ManyBabies framework

for collaborative team-based science—allowed us to recruit 1018

infants from 37 labs from around the world, with a final sample

of 567 infants from 34 labs included in the analyses. We used

this model to address critical limitations of prior work and to

establish a precise estimate of the effect size of infants’ preference

for helping over hindering agents in the Hill paradigm (Hamlin,

Wynn, and Bloom 2007). Furthermore, we sought to test the

extent to which infants’ preferences for Helpers are social in

nature by including a nonsocial control condition that was

physically matched to the experimental condition. As the first

of its kind, this project served as a proof-of-concept for the use

of behavioral components (e.g., manual choice) in a large-scale,

multi-lab replication and as a model for openly sharing video data

18 of 28 Developmental Science,2025

FIGURE 7 Plot of age ×Estimated freeze frame looking time. Note: Estimated looking time following the pushing events across ages. The

smoothing line shows the predicted marginal effects from our noninformative Bayesian regression model along with their credible interval. Data are

jittered slightly on the vertical axis to avoid overplotting.

across international borders. These videos can be reused in future

“spin-off” projects to answer new research questions.

4.1 Summary of Planned Analyses

The current study included infants across five continents, who

participated in a standardized, preregistered paradigm that

closely replicated the original hill study by Hamlin and colleagues

(Hamlin, Wynn, and Bloom 2007) with a few key differences.

Following our preregistered data analysis plan, we employed a

Bayesian analysis framework and predicted that (1) infants in

the social condition would show an above-chance tendency to

choose the Helper over the Hinderer compared to infants in the

nonsocial condition, and (2) infants’ preference for the Helper

over the Hinderer would not increase with age. In addition, we

tested whether infants who habituated (vs. those who did not)

would show a stronger preference for the Helper.

Inconsistent with our predictions, infants did not demonstrate

a preference for the Helper over the Hinderer in the social

condition, nor did they show a greater tendency to prefer

Helpers in the social than in the nonsocial condition. Thus,

the current work did not replicate the effect (preference for the

Helper) observed in the original paradigm (Hamlin, Wynn, and

Bloom 2007) and is inconsistent with a host of prior research

demonstrating that infants show social preferences for Helpers

and other prosocial agents over antisocial agents (e.g., Buon

et al. 2014; Geraci, Simion, and Surian 2022; Geraci and Surian

2011, 2023; Hamlin, Wynn, and Bloom 2010; Kanakogi et al.

2017). Instead, these findings more closely align with past work

that found no preference between Helpers and Hinderers (e.g.,

Schlingloff, Csibra, and Tatone 2020). These null results do not

appear to be driven by age-related differences: Our analyses,

consistent with Margoni and Surian (2018) and our prediction,

provide no evidence that infants’ social preferences for Helpers

over Hinderers change across the second half of the first year of

life. Similarly, our analyses showed that infants who habituated

did not choose the Helper more often than infants who never

habituated. Finally, the preregistered analysis revealed a low

intra-class correlation, suggesting that while there was consider-

able variability in infants’ preference for the Helper across labs,

this variability was even higher across individuals. Together, this

study provides evidence against infants’ prosocial preferences in

the hill paradigm, which suggests the effect is weaker, absent,

and/or develops later, than previously estimated (e.g., Margoni

and Surian 2018), and highlights that inconsistencies in past

work may be partially due to overall high levels of variability in

behavioral paradigms. We discuss these sources of variability in

the sections below.

Our collaboration yielded a large sample size, which allowed

a unique opportunity to explore potential moderating variables

and interactions related to preferences for the Helper or Hin-

derer, usually not evaluated due to lack of power. The planned

exploratory analyses tested whether infants’ preference for the

Helper was moderated by visual attention to video events, clarity

of infants’ choice actions, and experimenter awareness of the

condition before choice presentation. We found no evidence

supporting any moderating or interaction effects.

4.2 Summary of Exploratory Analyses

We additionally conducted exploratory analyses to examine

the relation between infants’ attention to the stimuli and their

subsequent social preferences. Infants’ visual attention to the still

frame following the videos varied by condition: Overall, infants

in the social condition looked longer than those in the nonsocial

condition. This conditional difference also interacted with age,

such that younger infants showed a larger difference in visual

attention between the social and nonsocial conditions than older

infants. Together, these patterns suggest that infants, especially

younger ones, found the social videos more engaging. Although

this result is consistent with a host of other research demonstrat-

ing young infants’ preferences for social over nonsocial stimuli

19 of 28

(e.g., preference for biological motion in Bardi, Regolin, and

Simion 2011; face preference in Valenza et al. 1996), it remains

unclear whether differences in infants’ looking to videos across

conditions were due to differences in the “plot” depicted in

social versus nonsocial events (i.e., helping/hindering vs. push-

up/push-down) or to various other lower-level differences (e.g.,

more eyes in the social videos, more motions, slight timing

differences between social and nonsocial videos). Critically,

infants’ attention did not differ between helping/push-up trials

and hindering/push-down trials, and the conditional differences

did not predict infants’ preference for the Helper or Hinderer.

This finding is consistent with past work, which also found

that infants’ visual attention to the still frames following hill

Helper/Hinderer events did not predict their preference, albeit

with 15-month-olds and using a fixed number of familiarization

trials rather than habituation (Schlingloff, Csibra, and Tatone

2020).

We also explored whether infants’ preference was moderated

by various lab-level and experiment-level factors, as well as

participant-level characteristics, and probed potential sources of

cross-lab variation in infants’ preference for the Helper. We found

a main effect of median laboratory testing date, such that infants

tested by labs with later median testing dates more often chose the

Helper/Push-up character. However, the interaction of interest

between median testing date and condition was not present.

Other analyses revealed that none of the lab- or experiment-level

variables of interest, such as exclusion rates or visual angle of

the stimuli (i.e., the ratio between the screen size and infant

distance from the screen), were associated with individual labs’

results. Similarly, there were no moderating or interaction effects

of counterbalanced experiment-level factors (e.g., order of video

presentation). Finally, participant-level characteristics such as

sex, handedness, and primary language exposure did not predict

infants’ preference for the Helper. This lack of sex difference,

in particular, is consistent with a recent meta-analysis (Margoni

et al. 2023). We found weak evidence for habituation status as

a moderator of infants’ choices, such that habituated infants

tended to choose the Helper/Push-Up character more often in the

nonsocial condition, compared to the social condition. However,

this effect was not present when age was included in the model

(in the confirmatory analyses), suggesting it is not robust and

therefore difficult to interpret. Altogether, results from these

exploratory analyses suggest that infants’ preference for Helpers

was not moderated by any variables collected in our study.

4.3 Interpretations and Implications

There are several possible ways to interpret the null results

observed in the current study. First, it may be that infants between

5 and 10 months of age do not prefer prosocial characters over

antisocial characters after all. That is, past research on infants’

social evaluations that obtained positive results (e.g., Buon et al.

2014, with 10- and 29-month-olds; Dunfield and Kuhlmeier 2010,

with 21-month-olds; Geraci and Surian 2023 with 4-month-olds;

Geraci, Simion, and Surian 2022, with 9-month-olds; Hamlin

and Wynn 2011, with 3-, 5-, and 9-month-olds; Hamlin 2013,with

5- and 8-month-olds; Kanakogi et al. 2017, with 6-month-olds;

Kanakogi et al. 2022, with 8-month-olds; Scola et al. 2015,with

12- to 14-month-olds) have all significantly overestimated the

true effect size of infants’ preference for prosocial others, perhaps

due to factors such as publication bias, other researcher degrees

of freedom, and small sample size (Margoni and Surian 2018).

Therefore, with a large and different sample size and a tightly

controlled methodology, the present work provided the most

robust and accurate estimate of this (null) effect. Indeed, the

present findings are consistent with a large body of work on

infants’ social evaluations which failed to detect a preference

for helpful agents (e.g., Abramson et al. 2016, unpublished,

with 9- and 18-month-olds; Cowell and Decety 2015,with12-

to 24-month-olds; Nighbor et al. 2017, with 16-month-olds;

Salvadori et al. 2015, with 9-month-olds; Schlingloff, Csibra, and

Taton e 2020, with 14- to 16-month-olds; Shimizu et al. 2018,with

6-, 9-, and 12-month-olds; Vaporova and Zmyj 2020, with 9- and

14-month-olds). Under this view, these findings serve as strong

evidence against infants’ social evaluative capacities at this

early age. Although demonstrating a lack of sensitivity to moral

actions during infancy would not in itself provide irrefutable

evidence that morality is not innate (nor would positive evidence

“prove” innateness), these findings highlight the importance of

considering nonnativist possibilities for the origins of morality.

For instance, future work should consider the role of children’s

early experiences in shaping socio-moral cognition and prosocial

behaviors (Dahl 2013;2015; Rogoff et al. 2018).

A second, narrower interpretation is that infants do not prefer

Helpers in the Hill paradigm specifically, but may nonetheless

prefer Helpers (and prosocial agents more generally) in other

contexts. This view is plausible, and perhaps more likely than the

first, for at least two reasons. First, although the Hill paradigm

is the first scenario in which evidence was found for infants’

social evaluation, it is far from the only scenario to show infants’

preference for prosocial agents in a helping/hindering context,

nor is helping/hindering the only context in which infants have

demonstrated positive evaluation of prosocial agents. Recall that

the Hill paradigm was chosen due to its high impact on and

pervasiveness in the literature, rather than its capacity to elicit the

strongest effects. Its simplicity,which allowed for the most precise

test of the construct of interest, comes with lower ecological

and face validity compared to other paradigms that boast richer

cues to agency or more familiar actions (Kominsky et al. 2022),

and these tradeoffs may have ultimately resulted in weaker

average effects. In fact, in the time since the decision to replicate

the Hill paradigm was made, meta-analyses have shown that

helping/hindering scenarios in general have yielded the smallest

average effect size relative to other socially-relevant scenarios,

such as fair versus unfair resource distributions and giving versus

taking (note that giving vs. taking is sometimes, but sometimes

not, referred to as helping vs. hindering in the literature; Margoni

and Surian 2018;A.Tan2024). Therefore, even if we were to

assume the evidence presented herein to be the “nail in the coffin”

for the Hill paradigm, it need not directly speak to other help-

ing/hindering scenarios (e.g., the Box opening/closing and Ball

giving/taking show in Hamlin and Wynn 2011), let alone other

significantly different but nonetheless socially-relevant scenarios

such as fair versus unfair resource distributions (e.g., Burns and

Sommerville 2014; Lucca, Pospisil, and Sommerville 2018)and

physical battery versus protection (e.g., Kanakogi et al. 2017).

Second, since the conception of the present project in 2018,

several works have been published by various research groups

20 of 28 Developmental Science,2025

using a diverse set of stimuli (e.g., Geraci and Franchin 2021;

Loheide-Niesmann et al. 2021; Schlingloff, Csibra, and Tatone

2020; Singh 2020; Vaporova and Zmyj 2020; Woo and Spelke

2020;2023;seeWooetal.2022 for recent review; see A. Tan,

2024 for recent unpublished meta-analysis), most of which have

found a preference for agents performing different kinds of

prosocial actions. For these two reasons, it seems plausible that

even if infants do not prefer Helpers over Hinderers in the Hill

paradigm by 10.5 months, they might nonetheless prefer Helpers

(or prosocial agents more generally) in other scenarios, perhaps

when the intentions of prosocial/antisocial agents and/or the

protagonist are overt (Geraci, Simion, and Surian 2022; Hamlin

2013; Strid and Meristo 2020; for pro-environmental behaviors,

see also Geraci, Franchin, and Benavides-Varela 2023). Of course,

it is unreasonable to require every single possible paradigm to

succeed or fail in large-scale collaborations before one makes

broad generalizations about social evaluation. However, it may be

equally unreasonable, or at least premature, to discount all other

paradigms based on results from one study alone. Indeed, these

findings highlight the need for alternative methods of conducting

large-scale replications that might be better suited for assessing

the truth-value of a given theoretical hypothesis. For instance,

one could randomly assign infants and/or laboratories to distinct

instances of prosocial and antisocial behaviors from across the

literature, which (given sufficient power) would have allowed

paradigm-level similarities/differences to be explored while also

assessing evidence for/against the broader question of whether

infants prefer prosocial to antisocial others.

It is also possible that the overall weak/null effect observed

herein is a product of cross-lab variations in methodology, an

interpretation supported by the cross-lab variation in infants’

preference for Helpers. Indeed, although we standardized critical

components of the study (e.g., stimuli, methods of presentation)

as much as possible, certain parts of the procedure inevitably

deviated between labs due to practical considerations. As but one

example, we allowed labs to “warm up” with infants according

to their own practices; variation in warm-up practices may result

in some infants being more/less comfortable in the testing envi-

ronment, and subsequently lead to different patterns in selective

reaching. Overall, given the large number of ways that labs vary

in their routine practices, we were not able to quantify the full

range of ways labs differed in their procedures, and so we were

unable to test whether and how they may have impacted infants’

behaviors.

More specific to the methodology is the potential variance in

the manual choice procedure. Indeed, whereas other multi-lab

replication attempts in the past used a “plug-and-play” method-

ology that required virtually no experimenter interaction with the

participants and relied entirely on looking-based measures (e.g.,

ManyBabies1; The ManyBabies Consortium 2020), the current

project was the first to use a behavioral component for the

primary dependent measure. The difficulty in standardizing our

procedure is evident in our relatively high exclusion rate (40.63%)

if we do not consider ineligible infants, and this exclusion rate

dropped to 14.86% if we do not consider experimenter or equip-

ment errors. Although all participating experimenters underwent

a standardized training process, and required approval from

the central team to be eligible to both administer and code

the choice phase, due to ethics constraints (e.g., laboratories

unable to share participant videos), we were unable to centrally

code all infant reaches. As such, variations both in how each

individual laboratory administered the choice phase, as well as

how infants’ manual reaches were subsequently coded could

have led to differences in infants’ likelihood of choosing the

Helper. That said, it should be noted that our attrition rate is

comparable to past replication attempts that also involved the

original author (e.g., 37.25% in Schlingloff, Csibra, and Tatone

2020). Thus, despite the null findings, we nonetheless consider

the current project a successful proof-of-concept for adapting and

scaling up paradigms with behavioral components.

On a different methodological level, our null findings could have

been driven by differences between our stimuli and the original

Helper/Hinderer “show” in Hamlin, Wynn, and Bloom (2007).

First, we used videos versus live events as had been used in the

original studies, both in order to maximize consistency across labs

and to make it possible for labs that lacked puppet stage setups

to run the study. Although we are aware of one published study

that successfully utilized video recordings of the Hill show with

a habituation design (Hamlin 2015), another lab failed to elicit

a preference for Helpers using an animated cartoon version of

the Hill show with a familiarization design (Schlingloff, Csibra,

and Tatone 2020). Notably, the videos used in the current study

were designed to match both each other and example videos of

the original live show used in Hamlin, Wynn, and Bloom (2007)

as closely as possible. However, because it is neither feasible nor

practical for puppeteers performing the Hill events to perfectly

reproduce the same show, this matching was achieved via heavy

editing in postproduction (e.g., by speeding up and slowing down

sections of the video).

It should also be noted that all of the videos used in the current

study were created following the initial peer review process due

to the need to include matching non-social videos, and therefore,

due to COVID-19, none of these videos were piloted prior to

testing (though these videos were closely matched to the ones

that were piloted during Stage 1 submission, albeit slightly faster

in order to minimize the timing difference between the social

and nonsocial condition videos). Although there is no reason

to suspect that the artificially sped up/slowed down sections

impacted the perception of helping and hindering, it is notable

that the live shows used in the original studies (12 s long on

average), as well as the videos used in the social condition in

the current study (13.2 s), were quite fast, whereas the cartoon

videos used in subsequent hill displays shown on screens were

on average significantly slower (e.g., E. Tan and Hamlin 2022;

16.8 s). These timing differences may be crucial, given that a

recent video-based eye-tracking study found a positive association

between anticipatory looking to the top of the hill during the

Climber’s failed attempts and subsequent visual preference for

the Helper (E. Tan and Hamlin 2022). Specifically, infants who

showed gaze shifts between the Climber and the hilltop (which

has been argued to signal goal understanding) visually preferred

the Helper at the test, whereas infants whose gaze fixated solely

on the Climber (which has been argued as signaling a lack of

goal understanding) did not. Thus, it is possible that increasing

the overall pace of the show in turn decreased the amount of

time infants had to process the Climber’s goal, which may have

influenced their subsequent social evaluations. This possibility

should be explored in future work.

21 of 28

Second, as evidenced by a recent unpublished meta-analysis

(A. Tan 2024), infant social evaluation studies in which stimuli

are presented on video elicit smaller effect sizes than those

with live puppet shows (note that a previous meta-analysis,

with less data, did not observe this effect; Margoni and Surian

2018); this effect size reduction may be due to relatively lower

attention and affective responses to video stimuli compared to

real-life events (e.g., Barr 2010; Diener et al. 2008). Although,

as noted above, there has been a successful replication using a

similar video-based procedure (Hamlin 2015), those shows were

slower than the ones in the current study. Thus, it is possible

that the combination of faster-paced videos with potentially

lower attentiveness/interest in video stimuli may have negatively

impacted infants’ understanding of the stimuli’s overall message.

An additional notable methodological difference between our

current work and the original 2007 Hill study was the (inten-

tional) omission of a bouncing event that occurred when the

Climber reached the top of the hill in the helping event of the

social condition. This omission may have been problematic, as

a previous study had failed to replicate a preference for Helpers

over Hinderers when the bouncing event was also included in

the hindering event (Scarf et al. 2012). Although this alternative

explanation was subsequently ruled out in a successful replica-

tion that did not include bouncing during helping events and

uncovered a preference for Helpers (Hamlin 2015), the absence

of a significant preference for the Helper over the Hinderer in the

current study with a broader sample could nonetheless suggest

that bouncing was an important component of infants’ social

evaluations in the original Hill studies. Importantly, as we did not

compare across conditions where the bouncing event was present

or absent in the current study, we are not able to determine the

impact of this factor relative to other differences from the original

stimuli. We encourage future work to systematically investigate

what lower-level factors, such as the bouncing action as well as

the overall pacing discussed above, impact infants’ understanding

and interpretation of helping/hindering actions, which would in

turn shed light on reasons behind failures to replicate the effects

of seminal work.

Finally, the present findings may simply reflect fundamental

differences between meta-analyses and multi-laboratory replica-

tions. Indeed, prior work has revealed that the effect sizes in meta-

analyses tend to be three times larger than in multi-laboratory

replication studies (Kvarven, Strømland, and Johannesson 2020),

though the exact reasons underlying these discrepancies remain

unclear (Lewis et al. 2022). It is noteworthy that even when

effect size estimates are similar between meta-analyses and large-

scale replication projects, the two sources may diverge in the

effects of key moderators such as infant age and experimental

method. This is the case of infant preference for infant-directed

speech (IDS), where preference of IDS over adult-directed speech

(ADS) reportedly increased with infant age according to a

large-scale replication (The ManyBabies Consortium 2020), but

remained stable according to meta-analytic estimates (Bergmann

et al., 2018; Dunst, Gorman, and Hamby 2012, and community-

augmented meta-analyses https://metalab.stanford.edu; Zetter-

sten et al. 2024), despite having the same overall effect size

(Zettersten et al. 2024). This may be because applying the very

same procedure to infants across age, as happens in large-scale

collaborations like ours but not in meta-analyses, can impact

the size of an observed effect given age-related differences in

attention, processing speed, and so forth. This suggests that meta-

analyses and large-scale replications should likely be used in tan-

dem to assess the robustness of effects of interest in the literature.

4.4 Challenges and Limitations

Though this project was successful on many fronts, there are

limitations that might have influenced the observed pattern of

results. First, the initial planned window for participating labs

to collect data spanned from summer 2021 to summer 2022.

However, the COVID-19 pandemic posed significant challenges

for participating laboratories and necessitated an extension, from

summer 2022 to fall 2023. In addition, the pandemic introduced

large cross-lab, -region, and -country variabilities in research

protocols over which our researchers had no control (e.g., mask-

ing, time researchers spent warming up with infants prior to

testing, infant familiarity with new environments and people).

For instance, many labs had to shut down for substantial periods

of time, whereas others remained open or had brief closures.

These closures could have led to recruiting challenges when

testing became possible (e.g., parents not yet comfortable with

in-person sessions), and researchers who were out of practice for

testing infants in person. In addition, we were unable to conduct

planned in-person training at international conferences, which

may have further contributed to across-lab procedural deviations.

Perhaps a more important factor is the potential impact on partic-

ipating infants. Though many of the infants who participated in

the current study were born after the height of pandemic-related

restrictions (e.g., social distancing, mask wearing), the lingering

effects of the pandemic may have led to a reduction in opportuni-

ties for social interaction and observation, which may have lim-

ited exposure to the types of social experience that are important

for sociomoral development in infancy. Reduced social interac-

tions with strangers may have also resulted in infants feelingmore

anxious about the testing environment, and this increased anxiety

may have subsequently affected their performance. Although

the exact impact of the COVID-19 pandemic on the study is

difficult to quantify, and a growing body of research suggests that

infants are surprisingly resilient toward direct adverse effects of

the pandemic (see Lobue et al. 2023,forareview),noworkto

our knowledge has directly tested whether infants who grew up

during the pandemic show heightened stranger anxiety, and how

this potential heightened anxiety affects subsequent performance

in social evaluative tasks. The present work crudely explored

potential cohort effects by testing the effect of median testing

dates; however, it should be noted that the date of testing is only

meaningful if considered alongside related environmental factors

(e.g., regional lockdown policies, laboratory testing policies). For

instance, two laboratories with similar median testing dates may

in fact have wildly different masking policies, or be located in

regions with varying degrees of COVID restrictions. In addition,

other factors that have been shown to impact infants’ sociocog-

nitive development (e.g. maternal stress; Nazzari et al. 2024)may

be correlated with some or all of these environmental variables.

As such, we note the theoretical and practical relevance of testing

the effects of the pandemic, but are cautious in interpreting these

exploratory results. We look forward to other future work, such as

longitudinal work that includes cohorts tested before, during, and

22 of 28 Developmental Science,2025

after the pandemic, as well as spin-off work that capitalizes on the

large dataset collected by this project and can more thoroughly

shed light on these questions.

Finally, despite the participation of 37 labs from 18 different

world regions, it is important to note that the participants

included in this study mostly came from industrialized and edu-

cated societies. This concern applies broadly to previous studies

investigating social evaluations in infancy (see the populations

included in the Margoni and Surian 2018 meta-analysis), and

within the field of developmental psychology broadly (Nielsen

et al. 2017). Intriguingly, a recent study documented variation

in infants’ (12- to 20-months of age) expectations about third-

party resource distribution between a Western (Sweden) and

two non-Western populations (Samburu and Kikuyu infants in

Kenya; Meristo and Zeidler 2022); these differing expectations

for fairness in infants in the second year might reflect that

infants are sensitive to the high levels of variation in fairness

norms and behaviors that exist across diverse societies (Blake

et al. 2015; Henrich, Heine, and Norenzayan 2010). In contrast,

because helping behaviors have been shown to emerge early and

consistently across diverse sociocultural contexts (Callaghan et al.

2011; Callaghan and Corbit 2018; Kärtner, Keller, and Chaudhary

2010), it may be that an early preference for helpful agents, if and

when it appears, will emerge similarly consistently across diverse

societies. It is crucial that future work continues to include partic-

ipants from diverse sociocultural environments, especially from

small-scale traditional societies, which are underrepresented in

many developmental psychology samples, including the current

study, in order to address these possibilities and otherwise gain

a globally representative view of social evaluations in early

childhood.

5 Conclusion

Rigorous replication efforts are cornerstones to scientific

progress; they contribute to bolstering the accuracy of findings,

examining conditions under which findings are observed,

and approximating findings’ true effect sizes (Open Science

Collaboration 2015; Zwaan et al. 2018). The current study,

carried out through the ManyBabies collaboration, found no

preference for Helpers over Hinderers in the Hill paradigm

(Hamlin, Wynn, and Bloom 2007) in a large and diverse sample

of 6- to 10-month-olds; these findings are inconsistent with the

original study and with past meta-analyses. We additionally

tested a number of moderators that could potentially explain

the variability in results among the labs (e.g., attention to the

stimuli, age, and sex), but did not find support for any moderating

effects. Our study also found substantial variation in effect size

across labs spanning diverse regions, though it remains unclear

what exact factors contribute to these differences, and whether

these differences merely reflect random variability associated

with complex behavioral paradigms. Finally, this work took the

first steps toward building best-practices for utilizing research

paradigms with behavioral components in large-scale, multi-

site collaborative research. We hope that the work presented

herein lays the groundwork upon which future work on the

development of social evaluations can build, and that novel

works will elucidate what factors underlie replication failures

and cross-lab differences.

Author Contributions

Conceptualization: Annette M. E. Henderson, Denis Tatone, Florina

Uzefovsky, Francis Yuen, J. Kiley Hamlin, Jessica Sommerville, Jonathan

F. Kominsky, Kelsey Lucca, Laura Schlingloff-Nemecz, Michael C. Frank,

Yiyi Wang, and Zoe Liberman. Data curation:AnnetteM.E.Henderson,

Bill Pepe, Denis Tatone, Francis Yuen, Kelsey Lucca, Samuel H. Forbes,

Wenxi Fei, and Yiyi Wang. Formal analysis: Alvin W. M. Tan, Amanda

M. Woodward, Annette M. E. Henderson, Arthur Capelier-Mourguy,

David Moreau, Denis Tatone, Francis Yuen, J. Kiley Hamlin, Julien

Mayor, Kelsey Lucca, Michael C. Frank, Nicolás Alessandroni, Samuel H.

Forbes, Wenxi Fei, Arthur Capelier-Mourguy, and Yiyi Wang. Funding

acquisition: Annette M. E. Henderson, Arkadiusz Urbanek, Casey Lew-

Williams, Francis Yuen, Hannes Rakoczy, Hernando Taborda-Osorio,

Hilal H. Şen, Hyun-joo Song, J. Kiley Hamlin, Kelsey Lucca, Laura K.

Cirelli, Lindsey J. Powell, Liquan Liu, Mark A. Schmuckler, Melanie

Soderstrom, Moritz M. Daum, Piotr Sorokowski, Tobias Grossmann,

Tobias Schuwerk, Yanjie Su, and Yenny Otálora. Investigation:Alia

Martin, Alyssa A. Quinn, Anna Exner, Annette M. E. Henderson,

Annie E. Wertz, Arkadiusz Urbanek, Arthur Capelier-Mourguy, Bailey

A. Immel, Bill Pepe, Casey Lew-Williams, Chantelle S. -S. Chin, Charisse

B. Pickron, Charlotte Grosse Wiesmann, Chiu Kin Adrian Tsang, Denis

Tatone, Emma L. Axelsson, Emmy Higgs Matzner, Florina Uzefovsky,

Francis Yuen,Grzegorz Jankiewicz, Hannes Rakoczy, Hernando Taborda-

Osorio, Hilal H. Şen, Hitomi Chijiiwa, Hyun-joo Song, Ingmar Visser,

Isabelle M. Hadley, J. Kiley Hamlin, Janina Baumer, John Corbit, Julie

Bertels, Karola Schlegelmilch, Katrin Rothmaler, Kelsey Lucca, Krista

Byers-Heinlein, Kristina Wolsey, Laura K. Cirelli, Laura Franchin, Laura

Schlingloff-Nemecz, Linda S. Oña, Lindsey J. Powell, Liquan Liu, Lucie

Zimmer, Madison Williams, Marina Proft, Mario Alvarez, Mark A.

Schmuckler, Markus Paulus, Megan E. Gornik, Melanie Soderstrom,

Michaela Dresel, Michal Misiak, Michelle Giraud, Mira L. Nencheva,

Miriam T. Loeffler, Moritz M. Daum, Naomi Havron, Natalie Christner,

Nicolás Alessandroni, Olivia Allison, Peter J. Reschke, Piotr Sorokowski,

Ronit Roth-Hanania, Sabine Seehagen, Sandro E. Stutz, Teresa Taylor-

Partridge, Terry Tin-Yau Wong, Tiffany Doan, Tobias Grossmann, Tobias

Schuwerk, Valentina Silvestri, Wiktoria Jędryczka, Xianwei Meng, Yang

Wu, Yanjie Su, Yasuhiro Kanakogi, Yenny Otálora, Yiyi Wang, Zhen

Zeng, and Zoe Liberman. Methodology: AnnetteM.E.Henderson,

Chantelle S.-S. Chin, Denis Tatone, Florina Uzefovsky, Francis Yuen,

J. Kiley Hamlin, Jessica Sommerville, Jonathan F. Kominsky, Kelsey

Lucca, Laura Schlingloff-Nemecz, Megan E. Gornik, Michael C. Frank,

Sandro E. Stutz, and Zoe Liberman. Project administration: Francis

Yuen, Heidi A. Baumgartner, J. Kiley Hamlin, Kelsey Lucca, Michael

C. Frank, Michal Misiak, Munna R. Shainy, Wiktoria Jędryczka, Yilin

Liu, and Yiyi Wang. Resources: Arkadiusz Urbanek, Emma L. Axelsson,

Francis Yuen, Grzegorz Jankiewicz, Hannes Rakoczy, J. Kiley Hamlin,

Katrin Rothmaler, Laura Franchin, Mario Alvarez, Moritz M. Daum,

Sandro E. Stutz, and Yilin Liu. Software: Francis Yuen, J. Kiley Hamlin,

Jonathan F. Kominsky, Kelsey Lucca, Michael C. Frank, Samuel H.

Forbes, and Sandro E. Stutz. Supervision: Alia Martin, Annette M. E.

Henderson, Annie E. Wertz, Arkadiusz Urbanek, Bill Pepe, Casey Lew-

Williams, Chantelle S. -S. Chin, Charisse B. Pickron, Charlotte Grosse

Wiesmann, Chiu Kin Adrian Tsang, Denis Tatone, Emma L. Axels-

son, Emmy Higgs Matzner, Florina Uzefovsky, Francis Yuen, Hannes

Rakoczy, Hernando Taborda-Osorio, Hyun-joo Song, Isabelle M. Hadley,

J. Kiley Hamlin, Janina Baumer, Jessica Sommerville, Julie Bertels,

Karola Schlegelmilch, Kelsey Lucca, Krista Byers-Heinlein, Laura K.

Cirelli, Laura Franchin, Laura Schlingloff-Nemecz, Lindsey J. Powell,

Liquan Liu, Lucie Zimmer, Mario Alvarez, Mark A. Schmuckler, Markus

Paulus, Melanie Soderstrom, Michael C. Frank, Michal Misiak, Michelle

Giraud, Miriam T. Loeffler, Moritz M. Daum, Natalie Christner, Piotr

Sorokowski, Sandro E. Stutz, Teresa Taylor-Partridge, Terry Tin-Yau

Wong, Tiffany Doan, Tobias Schuwerk, Valentina Silvestri, Yang Wu,

Yanjie Su, Yenny Otálora, Yiyi Wang, Zhen Zeng, and Zoe Liberman.

Validation: Alvin W. M. Tan, Amanda M. Woodward, Francis Yuen,

Julien Mayor, Kelsey Lucca, Michael C. Frank, Samuel H. Forbes, and

Yiyi Wang. Visualization: Alvin W. M. Tan, Francis Yuen, Julien Mayor,

Kelsey Lucca, and Yiyi Wang. Writing–original draft: Alessandra

23 of 28

Geraci, Annette M. E. Henderson, Arthur Capelier-Mourguy, David

Moreau, Denis Tatone, Florina Uzefovsky, Francis Yuen, J. Kiley Hamlin,

John Corbit, Jonathan F. Kominsky, Kelsey Lucca, Laura Schlingloff-

Nemecz, Mitali Bhavsar, Munna R. Shainy, Peter J. Reschke, Samuel H.

Forbes, Teresa Taylor-Partridge, Yilin Liu, Yiyi Wang, and Zoe Liberman.

Writing–review and editing: Alessandra Geraci, Alvin W. M. Tan,

Amanda M. Woodward, Annette M. E. Henderson, Charisse B. Pick-

ron, Charlotte Grosse Wiesmann, David Moreau, Denis Tatone, Emma

L. Axelsson, Florina Uzefovsky, Francis Yuen, Heidi A. Baumgartner,

Hilal H. Şen, J. Kiley Hamlin, Janina Baumer, John Corbit, Jonathan

F. Kominsky, Julien Mayor, Kelsey Lucca, Laura Schlingloff-Nemecz,

Liquan Liu, Michael C. Frank, Mitali Bhavsar, Moritz M. Daum, Munna R.

Shainy, Nicolás Alessandroni, Peter J. Reschke, Samuel H. Forbes, Teresa

Taylor-Partridge, Wenxi Fei, Yanjie Su, Yilin Liu, Yiyi Wang, and Zoe

Liberman.

Affiliations

1Department of Psychology, Arizona State University, Tempe, Arizona, USA

2Department of Psychology, University of British Columbia, Vancouver,

British Columbia, Canada 3Department of Psychology, University of

Chicago, Chicago, Illinois, USA 4Department of Psychology, Concordia

University, Montreal, Quebec, Canada 5Department of Psychology,

University of Virginia, Charlottesville, Virginia, USA 6School of

Psychological Sciences, University of Newcastle, Callaghan, New South

Wales, Australia 7Department of Psychology, University of Amsterdam,

Amsterdam, Noord-Holland, Netherlands 8Center for the Study of

Language and Information, Stanford University, Stanford, California, USA

9Center for Research in Cognition and Neurosciences, Université Libre de

Bruxelles, Brussels, Belgium 10ARISA Foundation, Baner, Maharashtra,

India 11Department of Psychology, Lancaster University, Lancaster,

Lancashire, UK 12Graduate School of Human Sciences, Osaka University,

Suita, Osaka, Japan 13Department of Psychology,

Ludwig-Maximilians-Universität München, Munich, Germany

14Department of Psychology, University of Toronto Scarborough,

Scarborough, Ontario, Canada 15Department of Psychology, St. Francis

Xavier University, Antigonish, Nova Scotia, Canada 16Jacobs Center for

Productive Youth Development, University of Zurich, Zurich, Switzerland

17Department of Psychology, University of Zurich, Zurich, Switzerland

18School of Psychology, Victoria University of Wellington, Wellington,

Wellington, New Zealand 19Faculty of Psychology, Ruhr University

Bochum, Bochum, Germany 20Department of Chinese and Bilingual

Studies, The Hong Kong Polytechnic University, Hung Hom, Hong Kong

21Department of Psychology, Durham University, Durham, UK

22Department of Psychology and Cognitive Science, University of Trento,

Trento, Italy 23Department of Psychology, Stanford University, Stanford,

California, USA 24Department of Educational Sciences, University of

Catania, Catania, Sicily, Italy 25Department of Psychology, University of

Milano-Bicocca, Milano, Italy 26Department of Psychology, University of

Manitoba, Winnipeg, Manitoba, Canada 27 Minerva Fast Track Group

Milestones of Early Cognitive Development, Max Planck Institute for Human

Cognitive and Brain Sciences, Leipzig, Sachsen, Germany 28School of

Psychological Sciences, University of Haifa, Haifa, Israel 29Center for Child

Development, University of Haifa, Haifa, Israel 30School of Psychology, The

University of Auckland | Waipapa Taumata Rau, Auckland, New Zealand

31Institute of Child Development, University of Minnesota Twin Cities,

Minneapolis, Minnesota, USA 32Department of Psychological & Brain

Sciences, University of California Santa Barbara, Santa Barbara, California,

USA 33Institute of Psychology, University of Wrocław, Wrocław, Poland

34Department of Cognitive Science, Central European University, Vienna,

Austria 35Department of Psychology, Princeton University, Princeton, New

Jersey, USA 36 Graduate School of Health, University of Technology Sydney,

Sydney, New South Wales, Australia 37School of Psychology, Western

Sydney University, Pentrith, New South Wales, Australia 38The MARCS

Institute for Brain, Behaviour and Development, Western Sydney University,

Westmead, New South Wales, Australia 39School of Behavioral and Brain

Sciences, University of Texas at Dallas, Richardson, Texas, USA

40Department of Psychology, University of Oslo, Oslo, Norway 41Graduate

School of Informatics, Nagoya University, Nagoya, Aichi, Japan 42IDN

Being Human, Institute of Psychology, University of Wroclaw, Wroclaw,

Poland 43School of Anthropology & Museum Ethnography, University of

Oxford, Oxford, UK 44Centre for Brain Research, University of Auckland,

Auckland, New Zealand 45Institute of Biology, Department of Human

Biology and Primate Cognition, University of Leipzig, Leipzig, Germany

46Max Planck Research Group Naturalistic Social Cognition, Max Planck

Institute for Human Development, Berlin, Germany 47Facultad de

Psicología, Universidad del Valle, Cali, Colombia 48Department of

Psychology, University of California San Diego, La Jolla, California, USA

49Institute of Psychology, University of Göttingen, Gottingen, Niedersachsen,

Germany 50School of Family Life, Brigham Young University, Provo, Utah,

USA 51Department of Psychology, The Academic College of Tel-Aviv Yaffo,

Tel-Aviv, Israel 52Humboldt Research Group for Child Development,

Faculty of Education, Leipzig University, Leipzig, Sachsen, Germany

53TUM School of Social Sciences and Technology, Technische Universität

München, Munich, Bavaria, Germany 54Faculty of Psychology, University

of Akureyri, Akureyri, Iceland 55Axxonet Brain Research Laboratory,

Bengaluru, Karnataka, India 56Department of Psychology, University of

Toronto, Toronto, Ontario, Canada 57Department of Psychology, Yonsei

University, Seoul, Republic of Korea 58School of Psychology and Cognitive

Sciences, Peking University, Beijing, China 59Departamento de Psicología,

Pontificia Universidad Javeriana, Bogota, Colombia 60Department of

Psychology, University of the Incarnate Word, San Antonio, Texas, USA

61Department of Psychology, The University of Hong Kong, Hong Kong,

Hong Kong 62Institute of Pedagogy, Faculty of Pedagogical and Historical

Sciences, University of Wrocław, Wroclaw, Poland 63Department of

Psychology, Ben-Gurion University in the Negev, Beer-Sheva, Israel

64Department of Psychology, University of Minnesota Twin Cities,

Minneapolis, Minnesota, USA 65Department of Linguistics and Modern

Languages, The Chinese University of Hong Kong, Sha Tin, New Territories,

Hong Kong SAR

Acknowledgments

We would like to thank the many families who participated in this

research. We would also like to thank all of the researchers who

participated in various aspects of this study, including planning, pilot

testing, participant recruitment, data collection, data management, and

providing feedback on the manuscript, including: Siobhan Kennedy-

Costantini, Florian Bednarski, Francesco Margoni, Yuju Shin, Alexander

Withers, Georgina Mariyadas, Ninthujah Suthaharan, Angela Le, Beverly

Lundeen, Cheyenne Engeser, Dulce Erdt, Duygu Hilal Bayir, Janek

Stahlberg, Daniel Tatlow-Devally, Julia Wolf, Dorottya Mészégető, Chiara

Vollmerhausen, Hafdís Kristný Haraldsdóttir, Stefanía Guðrún Eyjólfs-

dóttir, Léa Simon, Solenn Gouadon, Madeleine Gale, Mateo Belalcazar,

Laura Mejía, Mayte Alvarado, Madalyn Brown, Meghan Pierce, Moxuan

Liu, Xingyue Han, Sara Trnovska, Ariana Martinez, Silvia Saletti, Vanessa

Y. Dawidowicz, and Laura I. Blank.

Conflicts of Interest

The opinions expressed in this publication are those of the authors and do

not necessarily reflect the views of the funding agencies listed here. The

authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are openly available on

OSF at https://osf.io/qntyd/ and the code is available on GitHub at https://

github.com/manybabies/mb4-analysis.

24 of 28 Developmental Science,2025

Endnotes

1Note that a brief extension was given to one lab.

2Pilot testing revealed that brief look-aways outside this Critical Period

were common, especially during trials near the end of the habituation

phase.

3In our Registered Report, we planned to run simulations to determine

whether it is more appropriate to keep or drop random effects corre-

sponding to the dropped fixed effects from our model comparisons. We

planned to choose the model comparison method that yielded the best

Type I and II error rates, so long as these models successfully converge.

In analyzing the data, we came to the conclusion that because there are

no fixed effects included in the null model, the random effect should

be dropped because it indicates random variation of the fixed effect,

which is not included in the null model (Stroup 2012). Therefore, in our

analysis, we dropped the random effects of the predictors whose fixed

effects were dropped for nested model comparisons.

4The notation BF10 refers to the Bayes factor for the alternative hypothe-

sis. Conversely, the notation BF01 refers to the Bayes Factor for the null

hypothesis.

5The frequentist analyses revealed a main effect of condition, where

habituated infants were more likely to choose the helper in the non-

social condition relative to the social condition. However, the addition

of condition did not improve overall model fit, so this main effect was

not interpreted further.

6Median testing date was added as a moderator after the helpful

suggestion from one of the reviewers.

7The categorical variables and their reference levels are as follows: clear

versus ambiguous choice actions (clear), whether the experimenter had

knowledge of the participant’s condition (naive), the order of helping

versus hindering videos (helping first), Helper identity (blue shape),

Helper side during the choice phase (left), whether the experiment was

conducted as the first session of the infant’s visit or not (first session),

whether the experimenter wore a mask or not (no mask), habituation

status (nonhabituated), handedness (use both equally), color blindness

(no), participant sex (female).

8The model here is “choice ∼1+condition +(1 +condition | lab)” for

habituated and nonhabituated infants run separately. Note that in our

confirmatory analysis, described earlier, we included both age and the

interaction between age and condition in the model, and did not detect

an effect of condition.

References

Abramson, L., M. Dar, A. Te’eni, and A. Knafo-Noam. 2016. Preferences

for Helpers and Hinders in 9- and 18-Month-Old Infants. Unpublished

raw data.

Bardi, L., L. Regolin, and F. Simion. 2011. “Biological Motion Preference

in Humans at Birth: Role of Dynamic and Configural Properties.”

Developmental Science 14, no. 2: 353–359. https://doi.org/10.1111/j.1467-

7687.2010.00985.x.

Barr, R. 2010. “Transfer of Learning Between 2D and 3D Sources During

Infancy: Informing Theory and Practice.” Developmental Review 30, no. 2:

128–154. https://doi.org/10.1016/j.dr.2010.03.001.

Bergmann, C., S. Tsuji, P. E. Piccinini, et al. 2018. “Promoting Replicability

in Developmental Research Through Meta-Analyses: Insights From Lan-

guage Acquisition Research.” Child Development 89: 1996–2009. https://

doi.org/10.1111/cdev.13079.

Blake, P. R., K. McAuliffe, J. Corbit, et al. 2015. “The Ontogeny of

Fairness in Seven Societies.” Nature 528, no. 7581: 258–261. https://

link.gale.com/apps/doc/A437223659/HRCA?u=anon~1cb3f10e&sid=

googleScholar&xid=915c05ec.

Brandone, A. C., and H. M. Wellman. 2009. “You Can’t Always Get

What You Want: Infants Understand Failed Goal-Directed Actions.”

Psychological Science 20, no. 1: 85–91. https://doi.org/10.1111/j.1467-9280.

2008.02246.x.

Brownell, C. A. 2013. “Early Development of Prosocial Behavior: Current

Perspectives.” Infancy 18, no. 1: 1–9. https://doi.org/10.1111/infa.12004.

Buon, M., P. Jacob, S. Margules, et al. 2014. “Friend or Foe? Early Social

Evaluation of Human Interactions.” PLoS ONE 9, no. 2: e88612. https://

doi.org/10.1371/journal.pone.0088612.

Bürkner, P.-C. 2017. “brms: An R Package for Bayesian Multilevel Models

Using Stan.” Journal of Statistical Software 80, no. 1: 1–28. https://doi.org/

10.18637/jss.v080.i01.

Burns, M. P., and J. A. Sommerville. 2014. “‘I Pick You’: The Impact of

Fairness and Race on Infants’ Selection of Social Partners.” Frontiers in

Psychology 5: 93. https://doi.org/10.3389/fpsyg.2014.00093.

Byers-Heinlein, K., C. Bergmann, C. Davies, et al. 2020. “Building a

Collaborative Psychological Science: Lessons Learned From ManyBabies

1.” Canadian Psychology/Psychologie Canadienne 61, no. 4: 349–363.

https://doi.org/10.1037/cap0000216.

Callaghan, T., and J. Corbit. 2018. “Early Prosocial Development Across

Cultures.” Current Opinion in Psychology 20: 102–106. https://doi.org/10.

1016/j.copsyc.2017.07.039.

Callaghan, T., H. Moll, H. Rakoczy, et al. 2011. “Early Social Cognition in

Three Cultural Contexts.” Monographs of the Society for Research in Child

Development 76, no. 2: vii–viii, 1–142. https://doi.org/10.1111/j.1540- 5834.

2011.00603.x.

Carpendale, J. I., and C. Lewis. 2004. “Constructing an Understanding

of Mind: The Development of Children’s Social Understanding Within

Social Interaction.” Behavioral and Brain Sciences 27, no. 1: 79–96. https://

doi.org/10.1017/s0140525X04000032.

Carter, E. C., F. D. Schönbrodt, W. M. Gervais, and J. Hilgard. 2019.

“Correcting for Bias in Psychology: A Comparison of Meta-Analytic

Methods.” Advances in Methods and Practices in Psychological Science 2,

no. 2: 115–144. https://doi.org/10.1177/2515245919847196.

Chae, J. J., and H. Song. 2018. “Negativity Bias in Infants’ Expectations

About Agents’ Dispositions.” British Journal of Developmental Psychology

36: 620–633. https://doi.org/10.1111/bjdp.12246.

Chakraborty, H., and A. Hossain. 2018. “R Package to Estimate Intraclus-

ter Correlation Coefficient With Confidence Interval for Binary Data.”

Computer Methods and Programs in Biomedicine 155: 85–92. https://doi.

org/10.1016/j.cmpb.2017.10.023.

Cowell, J. M., and J. Decety. 2015. “Precursors to Morality in Develop-

ment as a Complex Interplay Between Neural, Socioenvironmental, and

Behavioral Facets.” Proceedings of the National Academy of Sciences of the

United States of America 112, no. 41: 12657–12662. https://doi.org/10.1073/

pnas.1508832112.

Dahl, A. 2015. “The Developing Social Context of Infant Helping in Two

U.S. samples.” Child Development 86, no. 4: 1080–1093. https://doi.org/10.

1111/cdev.12361.

Dahl, A., R. K. Schuck, and J. J. Campos. 2013. “Do Young Toddlers Act on

Their Social Preferences?” Developmental Psychology 49, no. 10: 1964–1970.

https://doi.org/10.1037/a0031460.

Diener, M. L., S. L. Pierroutsakos, G. L. Troseth, and A. Roberts.

2008. “Video Versus Reality: Infants’ Attention and Affective Responses

to Video and Live Presentations.” Media Psychology 11, no. 3: 418–

441.

Dunfield, K. A., and V. A. Kuhlmeier. 2010. “Intention-Mediated Selective

Helping in Infancy.” Psychological Science 21, no. 4: 523–527. https://doi.

org/10.1177/0956797610364119.

Dunst, C. J., E. Gorman, and D. W. Hamby. 2012. “Preference for Infant-

Directed Speech in Preverbal Young Children.” Center for Early Literacy

Learning 5, no. 1: 1–13.

Duval, S., and R. Tweedie. 2000. “A Nonparametric “Trim and Fill”

Method of Accounting for Publication Bias in Meta-Analysis.” Journal of

25 of 28

the American Statistical Association 95, no. 449: 89–98. https://doi.org/10.

1080/01621459.2000.10473905.

Fehr, E., and U. Fischbacher. 2003. “The Nature of Human Altruism.”

Nature 425, no. 6960: 785–791. https://doi.org/10.1038/nature02043.

Frank, M. C., E. Bergelson, C. Bergmann, et al. 2017. “A Collaborative

Approach to Infant Research: Promoting Reproducibility, Best Practices,

and Theory-building.” Infancy 22, no. 4: 421–435. https://doi.org/10.1111/

infa.12182.

Freud, S., J. Strachey, A. Freud, A. Strachey, and A. Tyson. 1961. The Ego

and the Id: The Standard Edition of the Complete Psychological Works of

Sigmund Freud, Volume XIX (1923-1925): The Ego and the ID and Other

Wor ks . London: The Hogarth Press: The Institute of Psycho-Analysis.

Gelman, A., D. Lee, and J. Guo. 2015. “Stan: A Probabilistic Program-

ming Language for Bayesian Inference and Optimization.” Journal of

Educational and Behavioral Statistics 40, no. 5: 530–543.

Geraci, A., and L. Franchin. 2021. “Do Toddlers Prefer That Agents Help

Similar or Dissimilar Needy Agents?” Infant and Child Development 30,

no. 5: e2247. https://doi.org/10.1002/icd.2247.

Geraci, A., L. Franchin, and S. Benavides-Varela. 2023. “Evaluations of

Pro-Environmental Behaviors by 7-Month-Old Infants.” Infant Behavior

and Development 72: 101865. https://doi.org/10.1016/j.infbeh.2023.101865.

Geraci, A., F. Simion, and L. Surian. 2022. “Infants’ Intention-Based

Evaluations of Distributive Actions.” Journal of Experimental Child

Psychology 220: 105429. https://doi.org/10.1016/j.jecp.2022.105429.

Geraci, A., and L. Surian. 2011. “The Developmental Roots of Fairness:

Infants’ Reactions to Equal and Unequal Distributions of Resources.”

Developmental Science 14, no. 5: 1012–1020. https://doi.org/10.1111/j.1467-

7687.2011.01048.x.

Geraci, A., and L. Surian. 2023. “Intention-Based Evaluations of Dis-

tributive Actions by 4-Month-Olds.” Infant Behavior and Development 70:

101797. https://doi.org/10.1016/j.infbeh.2022.101797.

Gronau, Q. F., H. Singmann, and E.-J. Wagenmakers. 2017. “Bridgesam-

pling: An R Package for Estimating Normalizing Constants.” Journal of

Statistical Software. Published ahead of print, October 23, 2017. https://

doi.org/10.31222/osf.io/v94h6.

Hamlin, J. K. 2013. “Failed Attempts to Help and Harm: Intention Versus

Outcome in Preverbal Infants’ Social Evaluations.” Cognition 128, no. 3:

451–474. https://doi.org/10.1016/j.cognition.2013.04.004.

Hamlin, J. K. 2015. “The Case for Social Evaluation in Preverbal Infants:

Gazing Toward One’s Goal Drives Infants’ Preferences for Helpers Over

Hinderers in the Hill Paradigm.” Frontiers 5: 1563. https://doi.org/10.3389/

fpsyg.2014.01563.

Hamlin, J. K., E. V. Hallinan, and A. L. Woodward. 2008. “Do As I

Do: 7-Month-Old Infants Selectively Reproduce Others’ Goals.” Develop-

mental Science 11, no. 4: 487–494. https://doi.org/10.1111/j.1467-7687.2008.

00694.x.

Hamlin, J. K., T. Ullman, J. Tenenbaum, N. Goodman, and C. Baker. 2013.

“The Mentalistic Basis of Core Social Cognition: Experiments in Preverbal

Infants and a Computational Model.” Developmental Science 16, no. 2:

209–226. https://doi.org/10.1111/desc.12017.

Hamlin, J. K., and K. Wynn. 2011. “Young Infants Prefer Prosocial to

Antisocial Others.” Cognitive Development 26, no. 1: 30–39. https://doi.

org/10.1016/j.cogdev.2010.09.001.

Hamlin, J. K., K. Wynn, and P. Bloom. 2007. “Social Evaluation by

Preverbal Infants.” Nature 450, no. 7169: 557–559. https://doi.org/10.1038/

nature06288.

Hamlin, J. K., K. Wynn, and P. Bloom. 2010. “Three-Month-Olds Show a

Negativity Bias in Their Social Evaluations.” Developmental Science 13, no.

6: 923–929. https://doi.org/10.1111/j.1467-7687.2010.00951.x.

Hamlin, J. K., K. Wynn,P. Bloom, and N. Mahajan. 2011. “How Infants and

Toddlers Reactto Antisocial Others.” Proceedings of the National Academy

of Sciences of the United States of America 108, no. 50: 19931–19936. https://

doi.org/10.1073/pnas.1110306108.

Henrich, J., S. J. Heine, and A. Norenzayan. 2010. “The Weirdest People

in the World?” Behavioral and Brain Sciences 33, no. 2–3: 61–83. https://

doi.org/10.1017/S0140525X0999152X.

Holvoet, C., C. Scola, T. Arciszewski, and D. Picard. 2016. “Infants’

Preference for Prosocial Behaviors: A Literature Review.” Infant Behavior

& Development 45: 125–139. https://doi.org/10.1016/j.infbeh.2016.10.008.

Ioannidis, John P. A. 2018. “Why Replication Has More Scientific Value

Than Original Discovery.” Behavioral and Brain Sciences 41: e137. https://

doi.org/10.1017/s0140525X18000729.

Judd, C. M., J. Westfall, and D. A. Kenny. 2017. “Experiments With More

Than One Random Factor: Designs, Analytic Models, and Statistical

Power.” Annual Review of Psychology 68: 601–625. https://doi.org/10.1146/

annurev-psych-122414-033702.

Kanakogi, Y., Y. Inoue, G. Matsuda, D. Butler, K. Hiraki, and M. Myowa-

Yamakoshi. 2017. “Preverbal Infants Affirm Third-Party Interventions

That Protect Victims From Aggressors.” Nature Human Behaviour 1, no.

2: 1–7. https://doi.org/10.1038/s41562-016-0037.

Kanakogi, Y., M. Miyazaki, H. Takahashi, H. Yamamoto, T. Kobayashi,

and K. Hiraki. 2022. “Third-Party Punishment by Preverbal Infants.”

Nature Human Behaviour 6, no. 9: 1234–1242. https://doi.org/10.1038/

s41562-022-01354-2.

Kanbe, F. 2019. “Generating Strictly Controlled Stimuli for Figure

Recognition Experiments.” Journal of Visualized Experiments.

https://app.jove.com/t/59149/generating-strictly-controlled-stimuli-

for-figure-recognition.

Kärtner, J., H. Keller, and N. Chaudhary. 2010. “Cognitive and Social

Influences on Early Prosocial Behavior in Two Sociocultural Contexts.”

Developmental Psychology 46, no. 4: 905–914. https://doi.org/10.1037/

a0019718.

Kass, R. E., and A. E. Raftery. 1995. “Bayes Factors.” Journal of the

American Statistical Association 90, no. 430: 773–795. https://doi.org/10.

2307/2291091.

Klein, R. A., K. A. Ratliff, M. Vianello, et al. 2014. “Data From Inves-

tigating Variation in Replicability: A “Many Labs” Replication Project.”

Journal of Open Psychology Data 2, no. 1: e4. http://doi.org/10.5334/jopd.

ad.

Kohlberg, L. 1969. Stages in the Development of Moral Thought and Action.

New York, NY: Holt, Rinehart & Winston.

Kominsky, J.F. 2019. “PyHab: Open-Source Real Time Infant Gaze Coding

and Stimulus Presentation Software.” Infant Behavior & Development 54:

114–119. https://doi.org/10.1016/j.infbeh.2018.11.006.

Kominsky, J. F., K. Lucca, A. J. Thomas, M. C. Frank, and J. K. Hamlin.

2022. “Simplicity and Validity in Infant Research.” Cognitive Development

63: 101213. https://doi.org/10.1016/j.cogdev.2022.101213.

Kruschke, J. K., and T. M. Liddell. 2018. “The Bayesian New Statistics:

Hypothesis Testing, Estimation, Meta-Analysis, and Planning From a

Bayesian Perspective.” Psychonomic Bulletin and Review 25, no. 1: 178–206.

https://doi.org/10.3758/s13423-016-1221-4.

Kvarven, A., E. Strømland, and M. Johannesson. 2020. “Comparing Meta-

Analyses and Preregistered Multiple-Laboratory Replication Projects.”

Nature Human Behaviour 4, no. 4: 423–434. https://doi.org/10.1038/

s41562-019-0787-z.

Lewis, M., M. B. Mathur, T. J. VanderWeele, and M. C. Frank. 2022. “The

Puzzling Relationship Between Multi-Laboratory Replications and Meta-

Analyses of the Published Literature.” Royal Society Open Science 9, no. 2:

211499. https://doi.org/10.1098/rsos.211499.

Liu, S., N. B. Brooks, and E. S. Spelke. 2019. “Origins of the Concepts

Cause, Cost, and Goal in Prereaching Infants.” Proceedings of the National

Academy of Sciences 116, no. 36: 17747–17752. https://doi.org/10.1073/pnas.

1904410116.

Lobue, V., K. Pérez-Edgar, N. Kirkham, and J. Herbert. 2023. “The Impact

of COVID-19 on Infant Development: A Special Issue of Infancy.” Infancy

28, no. 1:. https://doi.org/10.1111/infa.12528.

26 of 28 Developmental Science,2025

Loheide-Niesmann, L., J. de Lijster, R. Hall, H. van Bakel, and M. Cima.

2020. “Toddlers’ Preference for Prosocial Versus Antisocial Agents: No

Associations With Empathy or Attachment Security.” Social Development

30, no. 2: 410–427. https://doi.org/10.1111/sode.12487.

Loheide-Niesmann, L., J. de Lijster, R. Hall, H. van Bakel, and M. Cima.

2021. “Toddlers’ Preference for Prosocial Versus Antisocial Agents: No

Associations With Empathy or Attachment Security.” Social Development

30, no. 2: 410–427.

Lovric, M. M. 2019. “Conflicts in Bayesian Statistics Between Inference

Based on Credible Intervals and Bayes Factors.” Journal of Modern

Applied Statistical Methods 18.

Lucca, K., J. Pospisil,and J. A. Sommerville. 2018. “Fairness Informs Social

Decision Making in Infancy.” PLoS ONE 13, no. 2: e0192848. https://doi.

org/10.1371/journal.pone.0192848.

Makel, M. C., J. A. Plucker, and B. Hegarty. 2012. “Replications in

Psychology Research: How Often do They Really Occur?” Perspectives on

Psychological Science: A Journal of the Association for Psychological Science

7, no. 6: 537–542. https://doi.org/10.1177/1745691612460688.

Margoni, F., K. Block, K. Hamlin, N. Zmyj, and T. Schmader. 2023.

“Meta-Analytic Evidence Against sex Differencesin Infants’ and Toddlers’

Preference for Prosocial Agents.” Developmental Psychology 59, no. 2:

229–235. https://doi.org/10.1037/dev0001421.

The ManyBabies Consortium. 2020. “Quantifying Sources of Variabil-

ity in Infancy Research Using the Infant-Directed-Speech Preference.”

Advances in Methods and Practices in Psychological Science 3, no. 1: 24–52.

Margoni, F., and M. Shepperd. 2020. “Changing the Logic of Replication:

a Case From Infant Studies.” Infant Behavior and Development 61: 101483.

https://doi.org/10.1016/j.infbeh.2020.101483.

Margoni, F., and L. Surian. 2018. “Infants’ Evaluation of Prosocial and

Antisocial Agents: A Meta-Analysis.” Developmental Psychology 54, no. 8:

1445–1455. https://doi.org/10.1037/dev0000538.

Marquardt, D. W. 1980. “Comment: You Should Standardize the Predictor

Variables in Your Regression Models.” Journal of the American Statisti-

cal Association 75, no. 369: 87–91. https://doi.org/10.1080/01621459.1980.

10477430.

Meristo, M., and H. Zeidler. 2022. “Cross-Cultural Differences in Early

Expectations About Third Party Resource Distribution.” Scientific Reports

12, no. 1: 11627. https://doi.org/10.1038/s41598-022-15766-7.

Morey, R. D., R. Hoekstra, J. N. Rouder, M. D. Lee, and E. J. Wagenmakers.

2016. “The Fallacy of Placing Confidence in Confidence Intervals.”

Psychonomic Bulletin & Review 23, no. 1: 103–123. https://doi.org/10.3758/

s13423-015-0947-8.

Nazzari, S., M. P. Pili, Y. Günay, and L. Provenzi. 2024. “Pandemic Babies:

A Systematic Review of the Association Between Maternal Pandemic-

Related Stress During Pregnancy and Infant Development.” Neuroscience

& Biobehavioral Reviews 162: 105723. https://doi.org/10.1016/j.neubiorev.

2024.105723.

Nielsen, M., D. Haun, J. Kärtner, and C. H. Legare. 2017. “The Persistent

Sampling Bias in Developmental Psychology: A Call to Action.” Journal

of Experimental Child Psychology 162: 31–38. https://doi.org/10.1016/j.jecp.

2017.04.017.

Nighbor, T., C. Kohn, M. Normand, and H. Schlinger. 2017. “Stability of

Infants’ preference for Prosocial Others: Implications for Research Based

on Single-Choice Paradigms.” PLoS ONE 12, no. 6: e0178818. https://doi.

org/10.1371/journal.pone.0178818.

Open Science Collaboration. 2015. “Estimating the Reproducibility of

Psychological Science.” Science 349, no. 6251: aac4716. https://doi.org/10.

1126/science.aac4716.

Peirce, J., J. R. Gray, S. Simpson, et al. 2019. “PsychoPy2: Experiments

in Behavior Made Easy.” Behavior Research Methods 51, no. 1: 195–203.

https://doi.org/10.3758/s13428-018-01193-y.

Piaget, J. 1932. The Moral Judgment of the Child. London, England: Kegan

Paul, Trench, Trubner.

Plummer, M., N. Best, K. Cowles, and K. Vines. 2006. “CODA: Con-

vergence Diagnosis and Output Analysis for MCMC.” RNews6, no. 1:

7–11.

Rogoff, B., A. Dahl, and M. Callanan. 2018. “The Importance of Under-

standing Children’s Lived Experience.” Developmental Review 50: 5–15.

https://doi.org/10.1016/j.dr.2018.05.006.

Salvadori, E., T. Blazsekova, A. Volein, et al. 2015. “Probing the Strength of

Infants’ Preference for Helpers Over Hinderers: TwoReplication Attempts

of Hamlin and Wynn (2011).” PLoS ONE 10, no. 11: e0140570. https://doi.

org/10.1371/journal.pone.0140570.

Scarf, D., K. Imuta, M. Colombo, and H. Hayne. 2012a. “Social Evaluation

or Simple Association? Simple Associations May Explain Moral Reason-

ing in Infants.” PLoS ONE 7, no. 8: e42698. https://doi.org/10.1371/journal.

pone.0042698.

Scarf, D., K. Imuta, M. Colombo, and H. Hayne. 2012b. “Golden Rule or

Valence Matching? Methodological Problems in Hamlin et al.” Proceed-

ings of the National Academy of Sciences 109, no. 22: E1426. https://doi.

org/10.1073/pnas.1204123109.

Schlingloff, L., G. Csibra, and D. Tatone. 2020. “Do 15-Month-Old Infants

Prefer Helpers? A Replication of Hamlin et al. (2007).” Royal Society Open

Science 7, no. 4: 191795. https://doi.org/10.1098/rsos.191795.

Scola, C., C. Holvoet, T. Arciszewski, and D. Picard. 2015. “Further

Evidence for Infants’ Preference for Prosocial Over Antisocial Behaviors.”

Infancy 20, no. 6: 684–692. https://doi.org/10.1111/infa.12095.

Shimizu, Y.,S. Senzaki, and J. S. Uleman. 2018. “The Influence of Maternal

Socialization on Infants’ Social Evaluation in Two Cultures.” Infancy 23,

no. 5: 748–766. https://doi.org/10.1111/infa.12240.

Singh, L. 2020. “Bilingual Infants are More Sensitive to Morally Relevant

Social Behavior Than Monolingual Infants.” Journal of Cognition and

Development 21, no. 5: 631–650.

Stanley, D., and J. Spence. 2014. “Expectations for Replications: Are Yours

Realistic?” Perspectives on Psychological Science 9, no. 3: 305–318. https://

doi.org/10.1177/1745691614528518.

Strid, K., and M. Meristo. 2020. “Infants Consider the Distributor’s

Intentions in Resource Allocation.” Frontiers in Psychology 11: 596213.

https://doi.org/10.3389/fpsyg.2020.596213.

Stroup, W. W. 2012. Generalized Linear Mixed Models: Modern Concepts,

Methods and Applications. Boca Raton, Florida: CRC Press.

Tan, A. W. M. 2024. Prosocial Evaluation: A New Meta-Analysis. RPubs.

osf.io/preprints/psyarxiv/qnh92.https://doi.org/10.31234/osf.io/qnh92.

Tan, E., and J. K. Hamlin. 2022. “Mechanisms of Social Evaluation

in Infancy: A Preregistered Exploration of Infants’ Eye-Movement and

Pupillary Responses to Prosocial and Antisocial Events.” Infancy 27, no.

2: 255–276. https://doi.org/10.1111/infa.12447.

Valenza, E., F. Simion, V. M. Cassia, and C. Umiltà. 1996. “Face Preference

at Birth.” Journal of Experimental Psychology: Human Perception and

Performance 22, no. 4: 892–903. https://doi.org/10.1037/0096-1523.22.4.

892.

Vaporova, E., and N. Zmyj. 2020. “Social Evaluation and Imitation of

Prosocial and Antisocial Agents in Infants, Children, and Adults.” PLoS

ONE 15, no. 9: e0235595. https://doi.org/10.1371/journal.pone.0235595.

Woo, B. M., and E. S. Spelke. 2020. “How to Help Best: Infants’

changing Understanding of Multistep Actions Informs Their Evaluations

of Helping.” In Proceedings of the 42nd Annual Conference of the Cognitive

Science Society. COGSCI, 384–390.

Woo, B. M., and E. S. Spelke.2023. “Toddlers’ Social Evaluations of Agents

Who Act on False Beliefs.” Developmental Science 26, no. 2: e13314. https://

doi.org/10.1111/desc.13314.

Woo, B. M., C. M. Steckler, D. T. Le, and J. K. Hamlin. 2017. “Social

Evaluation of Intentional, Truly Accidental, and Negligently Accidental

Helpers and Harmers by 10-month-old Infants.” Cognition 168: 154–163.

https://doi.org/10.1016/j.cognition.2017.06.029.

27 of 28

Woo, B. M., E. Tan, and J. K. Hamlin. 2022. “Human Morality Is Based

on an Early-emerging Moral Core.” Annual Review of Developmental

Psychology 4: 41–61. https://doi.org/10.1146/annurev-devpsych-121020-

023312.

Woodward, A. L. 1998. “Infants Selectively Encode the Goal Object of an

Actor’s Reach.” Cognition 69, no. 1: 1–34.

Zettersten, M., C. Cox, C. Bergmann, et al. 2024. “Evidence for Infant-

Directed Speech Preference Is Consistent Across Large-Scale, Multi-Site

Replication and Meta-Analysis.” Open Mind: Discoveries in Cognitive

Science 8: 439–461. Advance publication. https://doi.org/10.1162/opmi_a_

00134.

Zwaan, R. A., A. Etz, R. E. Lucas, and M. B. Donnellan. 2018. “Making

Replication Mainstream.” Behavioral and Brain Sciences 41: E120. https://

doi.org/10.1017/S0140525X17001972.

Supporting Information

Additional supporting information can be found online in the Supporting

Information section.

28 of 28 Developmental Science,2025

ResearchGate has not been able to resolve any citations for this publication.

Pandemic babies: A systematic review of the association between maternal pandemic-related stress during pregnancy and infant development

Article

Full-text available

May 2024
NEUROSCI BIOBEHAV R

Cross-cultural differences in early expectations about third party resource distribution

Article

Full-text available

Jul 2022

Research using non-verbal looking-time methods suggests that pre-verbal infants are able to detect inequality in third party resource allocations. However, nothing is known about the emergence of this capacity outside a very narrow Western context. We compared 12- to 20-month-old infants (N = 54) from one Western and two non-Western societies. Swedish infants confirmed the pattern from previous Western samples by looking longer at the unequal distribution, suggesting that they expected the resources to be distributed equally. Samburu infants looked longer at the equal distribution, suggesting an expectation of unequal distribution. The Kikuyu infants looked equally at both distributions, and did not show any specific exactions. These results suggest that expectations of equal distributions in third party allocations are affected by experience of cultural variations of distributive norms and social interaction early in development.

Prosocial evaluation: A new meta-analysis

Preprint

May 2024

Alvin Wei Ming Tan

In this project, we provide an updated meta-analysis on infants' preference for prosocial agents, expanding an existing meta-analysis (Margoni & Surian, 2018) to include newer studies, studies using looking time measures, and additional conditions.

Evidence for Infant-directed Speech Preference Is Consistent Across Large-scale, Multi-site Replication and Meta-analysis

Article

Apr 2024

There is substantial evidence that infants prefer infant-directed speech (IDS) to adult-directed speech (ADS). The strongest evidence for this claim has come from two large-scale investigations: i) a community-augmented meta-analysis of published behavioral studies and ii) a large-scale multi-lab replication study. In this paper, we aim to improve our understanding of the IDS preference and its boundary conditions by combining and comparing these two data sources across key population and design characteristics of the underlying studies. Our analyses reveal that both the meta-analysis and multi-lab replication show moderate effect sizes (d ≈ 0.35 for each estimate) and that both of these effects persist when relevant study-level moderators are added to the models (i.e., experimental methods, infant ages, and native languages). However, while the overall effect size estimates were similar, the two sources diverged in the effects of key moderators: both infant age and experimental method predicted IDS preference in the multi-lab replication study, but showed no effect in the meta-analysis. These results demonstrate that the IDS preference generalizes across a variety of experimental conditions and sampling characteristics, while simultaneously identifying key differences in the empirical picture offered by each source individually and pinpointing areas where substantial uncertainty remains about the influence of theoretically central moderators on IDS preference. Overall, our results show how meta-analyses and multi-lab replications can be used in tandem to understand the robustness and generalizability of developmental phenomena.

Evaluations of pro-environmental behaviors by 7-month-old infants

Article

Jul 2023
INFANT BEHAV DEV

Environmental morality is the foundation of a sustainable future, yet its ontogenetic origin remains unknown. In the present study, we asked whether 7-month-olds have a sense of ‘environmental morality’. Infants’ evaluations of two pro-environmental actions were assessed in both visual and reaching preferential tasks. In Experiment 1, the overt behavior of protecting (i.e., collecting artificial objects spread on a lawn) was compared with the action of harming the environment (i.e., by disregarding the objects). In Experiment 2, the covert behavior of protecting the environment (i.e., maintaining artificial objects inside a container) was compared with the action of harming the environment (i.e., littering the artificial objects on a lawn). The results showed infants’ reaching preference for the agent who performed overt pro-environmental actions (Experiment 1), and no preference for the agent who performed covert pro-environmental actions (Experiment 2). These findings reveal a rudimentary ecological sense and suggest that infants require different abilities to evaluate overt impact-oriented and covert intend-oriented pro-environmental behaviors.

Generating Strictly Controlled Stimuli for Figure Recognition Experiments

Article

Mar 2019
JoVE

Fumio Kanbe

The impact of COVID‐19 on infant development: A special issue of infancy

Article

Jan 2023

Intention-based evaluations of distributive actions by 4-month-olds

Article

Feb 2023
INFANT BEHAV DEV

Four-month-olds' ability to consider the intentions of agents performing distributive actions was investigated in four experiments, using the Violation of Expectation paradigm (VoE) (Experiments 1-3) and the Preferential Looking paradigm (Experiment 4). In Experiment 1, infants were presented with two events showing two types of failed attempts to perform a distribution. In an attempt to distribute fairly, the distributor first tried to reach one of the recipients to deliver an apple, he failed, and then attempted to reach the other recipient to deliver a second apple and also failed. In an attempt to distribute unfairly, a different distributor tried unsuccessfully to bring resources always to the same recipient. Infants looked reliably longer at failed fair distribution events, suggesting that they did not just react to the actions outcomes and they attended to agents' intentions. Experiments 2 and 3 assessed alternative explanations based on perceptual factors or affiliative behaviors. In Experiment 4, during the test trials, infants were shown both distributors simultaneously and they preferred to look at the fair rather than at the unfair distributor. Overall, these findings reveal an early ability to take into account distributors' intentions and a preference for watching agents that tried to distribute resources fairly.

Human Morality Is Based on an Early-Emerging Moral Core

Article

Sep 2022

Scholars from across the social sciences, biological sciences, and humanities have long emphasized the role of human morality in supporting cooperation. How does morality arise in human development? One possibility is that morality is acquired through years of socialization and active learning. Alternatively, morality is instead based on a “moral core”: primitive abilities that emerge in infancy to make sense of morally relevant behaviors. Here, we review evidence that infants and toddlers understand a variety of morally relevant behaviors and readily evaluate agents who engage in them. These abilities appear to be rooted in the goals and intentions driving agents’ morally relevant behaviors and are sensitive to group membership. This evidence is consistent with a moral core, which may support later social and moral development and ultimately be leveraged for human cooperation. Expected final online publication date for the Annual Review of Developmental Psychology, Volume 4 is December 2022. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Toddlers’ Social Evaluations of Agents Who Act on False Beliefs

Article

Aug 2022

Mature social evaluations privilege agents' intentions over the outcomes of their actions, but young children often privilege outcomes over intentions in verbal tasks probing their social evaluations. In three experiments (N = 118), we probed the development of intention-based social evaluation and mental state reasoning using nonverbal methods with 15-month-old toddlers. Toddlers viewed scenarios depicting a protagonist who sought to obtain one of two toys, each inside a different box, as two other agents observed. Then, the boxes' contents were switched in the absence of the protagonist and either in the presence or the absence of the other agents. When the protagonist returned, one agent opened the box containing the protagonist's desired toy (a positive outcome), and the other opened the other box (a neutral outcome). When both agents had observed the toys move to their current locations, the toddlers preferred the agent who opened the box containing the desired toy. In contrast, when the agents had not seen the toys move and therefore should have expected the desired toy's location to be unchanged, the toddlers preferred the agent who opened the box that no longer contained the desired toy. Thus, the toddlers preferred the agent who intended to make the protagonist's desired toy accessible, even when its action, guided by a false belief concerning that toy's location, did not produce a positive outcome. Well before children connect beliefs to social behavior in verbal tasks, toddlers engage in intention-based evaluations of social agents with false beliefs. This article is protected by copyright. All rights reserved.

Infants’ Social Evaluation of Helpers and Hinderers: A Large-Scale, Multi-Lab, Coordinated Replication Study

Abstract and Figures

Recommended publications

Infants’ Social Evaluation of Helpers and Hinderers: A Large‐Scale, Multi‐Lab, Coordinated Replicati...

Infants’ Social Evaluation of Helpers and Hinderers: A Large-Scale, Multi-Lab, Coordinated Replicati...

Do 15-month-old infants prefer helpers? A replication of Hamlin et al . (2007)

Human Morality Is Based on an Early-Emerging Moral Core