You are on page 1of 13

Child Development, September/October 2019, Volume 90, Number 5, Pages 1789–1801

Modeling Prosocial Behavior Increases Helping in 16-Month-Olds


Nils Schuhmacher Moritz K€
oster
University of M€
unster University of Münster and Free University Berlin

Joscha K€artner
University of M€
unster

In two experiments, the imitation of helping behavior in 16-month-olds was investigated. In Study 1 (N = 31),
infants either observed an adult model helping or not helping another individual before they had the opportu-
nity to assist an unfamiliar experimenter. In one of two tasks, more children helped in the prosocial model
condition than in the no model control condition. In Study 2 (N = 60), a second control condition was
included to test whether infants imitated the prosocial intention (no neediness control). Children in the proso-
cial model condition helped more readily than children in the no model condition, with the second control
condition falling in between. These findings propose that modeling provides a critical learning mechanism in
early prosocial development.

Shortly after their first birthday, infants begin to encouragement and praise by primary caregivers
help others, for example by handing them objects during the second year (Dahl, 2015; Dahl et al.,
out of their reach. This early helping behavior fur- 2017; Hammond & Carpendale, 2015; K€ oster, Caval-
ther increases in frequency and complexity during cante, Vera Cruz de Carvalho, D^ ogo Resende, &
children’s second year of life (Dahl, 2015; Warneken, K€artner, 2016), there is no research on another
Chen, & Tomasello, 2006; Warneken & Tomasello, important mechanism, namely observational learn-
2007). Although the development of children’s help- ing based on imitation (see Elsner, 2014).
ing behavior has been described exhaustively in the Influential theoretical frameworks have long
literature (see Eisenberg, Fabes, & Spinrad, 2006), acknowledged the importance of observational
research on the driving forces is scarce. In particular, learning for child development, including children’s
it has been questioned if and how children’s proso- helping behavior. For example, Rogoff (2003) and
cial development is influenced by social learning Grusec and Davidov (2010) emphasize that it is not
from infancy on. One the one hand it was proposed only essential to investigate the degree to which
that infants’ early helping tendencies predominantly adults actively scaffold children’s learning pro-
rest on an evolved biological predisposition (see cesses but also to investigate whether children
Warneken & Tomasello, 2009). On the other hand, have opportunities to observe (and participate in)
current research focuses on different social learning others’ activities such as helping or cooperative
mechanisms such as infant’s sensitization to other’s tasks.
needs by means of parental scaffolding, that is, par- Seminal empirical work on observational learn-
ental encouragement, praise or supportive structur- ing by Bandura (1965) and Bandura, Ross, and Ross
ing in helping situations (see Brownell, 2016). (1963) with older children shows (a) that children
However, although there is first correlational and potentially acquire aggressive behaviors by observ-
experimental evidence on the positive effect of ing corresponding models and (b) that observing
others not only impacts the general acquisition of
new skills but also impacts children’s motivation to
We particularly thank Amelie Helmrich, Micha Hilbert, Karina
L€
udemann, Jennifer Riefer, Benedikt Schuler, Judith Silk- perform previously observed behaviors such as
enb€aumer, Eva Strehlke, and Ulrike Wilde for recruitment, data aggressive acts. Based on these (and additional)
collection, and/or coding. Furthermore, we say thank you to all findings, Bandura (1986) proposed his model of
participating families.
Correspondence concerning this article should be addressed to
Nils Schuhmacher, Unit of Developmental Psychology, Depart-
ment of Psychology, University of M€ unster, Fliednerstr. 21, © 2018 Society for Research in Child Development
48149 M€ unster, Germany. Electronic mail may be sent to All rights reserved. 0009-3920/2019/9005-0023
nils.schuhmacher@uni-muenster.de. DOI: 10.1111/cdev.13054
1790 Schuhmacher, K€
oster, and K€artner

sociocognitive learning that generally builds on infants repeatedly observed a passive model where
humans’ competencies to learn from observing the adult did not provide any help. In the subsequent
others. test phase, infants had the opportunity to instrumen-
Bandura’s Bobo-Doll studies focused on imitation tally assist a third, unfamiliar experimenter. We used
of aggressive behavior in elementary school-aged live models instead of video-mediated models,
children but inspired subsequent research on imita- because live models facilitate observational learning
tive learning of prosocial activities such as sharing in infants (see Barr & Hayne, 1999). We hypothesized
and generosity in elementary schoolers (e.g., Blake, that infants in the prosocial model condition would
Corbit, Callaghan, & Warneken, 2016; Presbie & help the needy experimenter more readily than chil-
Coiteux, 1971; Radke-Yarrow, Zahn-Waxler, & dren in the control condition.
Chapman, 1983; Rushton, 1975; Staub, 1971).
However, observational learning is also a strong
force much earlier in ontogeny. More specifically, Method
infants readily imitate simple activities such as
Sample and Study Design
action sequences performed with toys (Elsner, 2014;
Gergely, Bekkering, & Kir" aly, 2002; Meltzoff, 1995; Our sample consisted of N = 31 16-month-old
Tomasello, 1999; Zmyj & Seehagen, 2013). But, so infants (18 boys; age: M = 16.32; SD = 0.71; range =
far, only a few studies have looked at imitative 14 months; 25 days–17 months; 11 days). Due to
learning of early prosocial activities. In a first study, distress reactions (i.e., intense and/or repeated cry-
Williamson, Donohue, and Tully (2013) found that ing), one child had to be excluded from the final
comforting in 30-month-olds was affected by obser- sample (final N = 30 infants). Participants were
vation of prosocial video models. In particular, chil- recruited from the database of the Developmental
dren who had observed a prosocial model showed Psychology Lab. Parents were fully informed about
more comforting behavior toward their mothers the study’s goals and gave their written consent for
later on, compared to children in two different con- their child’s participation. Participation was volun-
trol conditions (i.e., without a prosocial model or tary, and children received a small gift after the
without any distress of their mothers). testing session. All participants came from Western
On the basis of this literature, we assume that middle-class families living in a mid-sized city in
observational learning is of primary importance in Germany. Data were collected between October
children’s developing prosociality. However, no 2015 and January 2016.
previous study has actually investigated whether The study was based on a between-subjects
observing prosocial models affects infant’s early design with two conditions: (a) prosocial model condi-
helping behavior, namely their instrumental assis- tion (i.e., children observed a prosocial model prior
tance toward others early in their second year to the test phase; n = 15) and (b) no model control
(Warneken & Tomasello, 2007; Warneken et al., condition (i.e., children saw no prosocial model;
2006). To fill this gap, the current experiments (both n = 15). Children were distributed randomly to
Study 1 and Study 2) focus on how observing each condition.
others’ prosocial behavior affects subsequent help-
ing behavior. Furthermore, in Study 2 we specifi-
General Procedure
cally aim to differentiate whether infants just
imitate observed behavior or whether they are The experiment took place in the laboratory of
more selective and take the target’s neediness into the Department of Developmental Psychology at
account. the University of M€ unster and lasted approxi-
mately 40 min. Each testing session began with an
initial warm-up, followed by two different experi-
mental imitation tasks (i.e., container task and
Study 1
slider task), administered in a counterbalanced
In Study 1, we investigated 16-month-olds’ helping order, and a transfer task in between (i.e., plate
behavior after they had observed helping actions task). Each experimental task consisted of a
by others. In particular, we implemented one condi- modeling phase (in which children observed differ-
tion in which infants could repeatedly observe an ent modeling behaviors depending on the condi-
adult model helping an individual in need by mak- tion) followed directly by a test phase (in which
ing desired objects assessable to her (i.e., prosocial children could assist E3). At the end of the
model condition); in a second (control) condition, experiment, children received a small gift for their
Observing a Prosocial Model Increases Helping 1791

participation. Four cameras recorded the sessions Up to this point, the helper (E2) sat at the side and
for the subsequent analysis. observed the scene. He then got up from his seat
Warm-up phase. All three experimenters (E1 = and helped E1 by putting the first container on the
female, E2 = male, and E3 = female) welcomed the table, so that E1 could now reach the cup in this
parent and the child into the lab. The warm-up container. Then E1 said: “Great, now I can go on
phase began with a free play episode, in which the building my tower.” E1 did not thank or socially
child played with E1 using a standard set of toys interact with E2 but focused on the cup and build-
for about 10 min. In the meantime, E3 left the room ing her tower. E2 went back to his seat. The proce-
and E2 informed the mother about the study goals dure was repeated for the second and the third
and instructed her to remain passive during the container (i.e., overall E2 successively put three con-
testing session. tainers on the table).
To test for infants’ level of spontaneous helping In the slider task, the procedure was almost iden-
toward an unfamiliar experimenter (Cortes Barra- tical to the container task. The main difference was
gan & Dweck, 2014), we conducted a classic out-of- in the way in which E2 helped E1 (see also Fig-
reach task at the end of the warm-up session: E2 ures 1 and 2): A larger table was used and three
hung clothes on a clothesline and successively sliders (containing the cups or balls) were posi-
dropped three clothespins. It was then observed tioned on the table, at the opposite side of E1, out
whether infants helped E2, who showed increas- of E1’s reach. In the modeling phase E1 leaned over
ingly ostensive cues within 45 s by reaching for the the table and unsuccessfully reached for the cups
clothespins (Warneken et al., 2006). This task (or balls) on the sliders and E2 helped E1 by sliding
helped to assert that random assignment to experi- the cups (or balls) toward her.
mental and control condition created groups that Modeling phase (no model control). In the no
are initially similar regarding their spontaneous model control condition infants observed a passive
helping toward an unfamiliar adult. model (E2) not helping the target (E1). Like in the
Experimental tasks. After the warm-up phase, prosocial model condition, E1 initially built a tower
each child was presented with two experimental and then uttered that she wanted to build a second
tasks (i.e., container task, slider task). The order of pre- tower and therefore she would need more cups. In
sentation of these tasks was counterbalanced. Each contrast to the prosocial model condition, the three
task consisted of two different phases: (a) a modeling additional cups (or balls) for building the second
phase and (b) a test phase. During the modeling tower were positioned on the table (i.e., from E1’s
phase children either observed a prosocial model perspective: on the right side of the table) and not
(i.e., prosocial model condition) or a passive model in the containers on the ground (which were
(i.e., no model control). Infants were seated in a high empty). E1 looked around, spotted the cups on the
chair so that they could optimally observe the scene. table and, similarly as in the prosocial mode condi-
In the subsequent test phase children had the oppor- tion, leaned over the table, reaching for the object.
tunity to assist an unfamiliar experimenter. After 10 s, E1 finally picked up the first cup while
Modeling phase (prosocial model condition). In commenting “Mhh, I can reach the cup! Although
the prosocial model condition, children observed a it is quite far away!” E1 repeated this sequence for
prosocial model (E2 = helper) helping a target the second and third cup. The model (E2) remained
(E1 = helpee). In the container task, E1 sat at a table passively on a seat during the whole scene.
with three cups (or balls) in front of her (materials Again, the slider task was structurally equivalent
were counter balanced across tasks; see also Fig- but varied in the position of the cups (or balls);
ures 1 and 2). E1 subsequently took these cups and namely, the sliders were positioned in the middle
stacked a tower with them (balls were placed into a of the table, such that E1 could reach them.
transparent tube). After completing the first tower, After the modeling phase, E1 and E2 set up the
the experimenter uttered “Now I want to build a room for the subsequent test phase. The parent
second tower. Therefore, I need more cups. But, interacted with the infant in the meantime, to dis-
where are some more cups?” She then looked tract the child’s attention from the rebuilding of the
around the room and spotted three additional cups, setup. E1 finally instructed mothers to take their
each placed in a container on the floor in front of children out of the high chair at a prearranged sig-
her table. E1 leaned over the table and reached for nal during the test phase and otherwise remain pas-
one of the cups (i.e., from E1’s perspective: the cup sive.
in the leftmost container), commenting “Mhh, I can’t Test phase (prosocial model condition, no model
reach it! It is too far away!” (approximately 10 s). control). Test phase was identical in both
1792 Schuhmacher, K€
oster, and K€artner

Observa!on Phase Test Phase


Prosocial model condi!on E2 (model: ac!ve)

E1

Child

No model control E2 (model: passive) All condi!ons

E1 E3

Child
Child

No neediness control E2 (model: ac!ve)


(only Study 2)

E1

Child

Figure 1. Sketch of initial room setup in each condition and phase for Study 1 and Study 2.
Note. Big rectangle = table; small squares = chairs; big circles = experimenters/child; small circles = containers for target objects (in sli-
der task, a bigger table was used and sliders were positioned on the table); small dots = target objects (i.e., cups, balls/bricks). In Study
1, only the prosocial model condition and no model control were conducted. Furthermore, only three objects were initially on the table
and tables were smaller. [Color figure can be viewed at wileyonlinelibrary.com]

conditions. At the beginning of the test phase in Stage 3 (10 s) E3 furthermore commented: “Look,
each task, E1 and E2 left the room and an unfamil- [name of the child]: I can’t reach the cup (ball)!”.
iar experimenter (E3), blinded to the condition, This sequence was repeated for all three objects.
entered the room. The initial setup was identical to Transfer task. The transfer task was administered
the initial setup of the modeling phase (see Fig- between the slider task and the container task. This
ure 2). E3 demonstrated an instrumental need that task was included to test whether observing a proso-
was very similar to E1’s demonstration during the cial model in the first experimental task (i.e., out-of-
modeling phase. However, in contrast to E1’s reach helping task; Warneken et al., 2006) would also
demonstration, E3 modeled her need with increas- generalize to a different helping context, namely a
ing ostensive cues, giving the infant the opportu- collaborative helping task that requires infants’ spon-
nity to help. In Stage 1 (10 s) E3 stretched out, taneous assistance in jointly tiding up or collecting
focusing the object (“Mhh, I can’t reach it”), in items (Rheingold, 1982). In this task, E3 took a tray
Stage 2 (10 s) E3 additionally alternated her gaze with ten plastic plates and then tripped over a carpet
between the target object and the child, and in so that all the plates fell on the floor. She then sat
Observing a Prosocial Model Increases Helping 1793

Prosocial Model Condition No Model Control


& No Neediness Control
Container Task (With Cups)

Slider Task (With Bricks)

Figure 2. Initial setups of modeling phase in Study 2.


Note. Materials (i.e., cups, bricks) were counterbalanced between children. Initial setup for the test phase was identical to initial setup of
the modeling phase in the prosocial model condition (see left pictures). However, in comparison to the modeling phase, the three left-
most cups/bricks were missing. Furthermore, the initial setup was similar in Study 1. However, in comparison to Study 2, in Study 1
(a) there was no “no neediness control” condition, (b) only one chair, one table (container task) or two tables (slider task) were used, (c)
there were balls instead of bricks and two transparent tubes to throw balls in, (d) handles of the sliders were longer (approximately
50 cm), (e) children observed the scene from a different angle, and (f) three leftmost cups/bricks were not present. [Color figure can be
viewed at wileyonlinelibrary.com]

down and slowly began to collect the plates, com- the target object (i.e., cup or ball) or moved the
menting on her actions (E3: “Oh no, now I have to slider or container with the target objects closer
collect all plates . . . So, I take this plate and put it to E3. The score given was based on the number
back on the tray. Now I take that plate . . .” etc.). of objects the child helped with: 0 to 3 for each
Children now had the opportunity to spontaneously task.
assist E3 in collecting plates and/or putting them Transfer task. Children could assist the experime-
back on the tray. After 60 s, E3 finished the task by nter by handing E3 the plates or by putting the plates
putting the tray on a table. on the tray. The score was the sum of all the plates
the child helped clean up, ranging from 0 to 10 (ten
total plates fell on the ground).
Data coding
Interrater reliability for 30% of the data was per-
Initial clothespin task. Children could pick up fect for helping scores, all Cohen’s js = 1.0.
and hand over up to three dropped clothespins to
E1 during the warm-up phase. Thus, the resulting
initial helping score ranged from 0 to 3. Results
Experimental tasks. Children could help E3 by
Preanalyses
offering her up to three objects in each experimental
task (i.e., slider task and container task). Helping Age was not significantly associated with chil-
was coded if children either directly handed E3 dren’s helping rates, all |r|s < .19, all ps > .10.
1794 Schuhmacher, K€
oster, and K€artner

Furthermore, there was no significant gender set as random effect factor. The full model had a
effect in the experimental tasks, both |t|s < 1.71, all significantly better model fit than the unrestricted
ps > .05. However, there was a significant gender model (i.e., null model with no fixed effects),
effect in the transfer task; that is, boys helped at v2(3) = 26.887, p < .001. Further model fit indices
higher rates than girls in this task, Mgirls = 0.25, also indicated better fit for the full model (Akaike’s
SD = 0.62 and Mboys = 1.94, SD = 2.12, t(28) = information criterion [AIC] = 53.94, Bayesian infor-
!2.67, p = .012. In addition, there were no signifi- mation criterion [BIC] = 64.49; unrestricted model:
cant effects for target objects across tasks, MBalls = AIC = 74.82, BIC = 79.05). There is a significant
0.43, SD = 0.73, and MCups = 0.70, SD = 1.12, t(28) = main effect by task and condition, and there was a
!1.49, p = .147. There was also no order effect; that significant interaction effect Task 9 Condition (see
is, children’s helping rates did not significantly dif- Table 2). Post hoc tests revealed that, during the sli-
fer whether slide task was administered first and der task, significantly more children helped in the
container task second or vice versa, t(29) = 0.47, experimental condition (i.e., 47% helpers) than in
p = .641. Because these control variables did not the control condition (i.e., 13% helpers), v2 = 3.97,
affect infants’ helping rates in the experimental p = .046, two-sided (see Figure 3), x = .36, 95% CI
tasks, they were not considered in further analyses. for the differences between the two proportions
Finally, infants’ initial helping rates during the [2.78, 63.88], whereas there was no significant dif-
clothespin task did not significantly differ between ference between conditions for the container task,
conditions. Because initial helping rates were very v2 = 0.60, p = .439, x = .14, 95% CI [!46.73, 20.07].
low (Mtest = 0.06, SD = 0.26, and Mcontrol = 0.20, Finally, there were no significant differences
SD = 0.77), we analyzed the percentage of infants between conditions in the transfer task, that is, 27%
that helped at least once initially (7% helpers in helpers in the experimental condition versus 47%
both conditions, v2 = 0.00, p > .99). helpers in the control condition, v2 = 1.29, p = .27,
two-sided, x = .21, 95% CI [!55.7, 15.7].
Infants’ Imitation Behavior
During the test phase, infants showed very low
helping rates as well (see Table 1)—on average, Table 2
infants provided only 1.2 of 6 objects during the Results From GLMM on Infant’s Helping Behavior in Study 1
test phase. Thus, we used dichotomous scores for Fixed effects b SE [CI 95%] z Value
further analyses; namely, we categorized helpers
(i.e., children who helped with at least one object) Condition 14.285 6.091 [2.35, 26.22] 2.345*
and nonhelpers (see Supporting Information for Trial type 14.492 4.871 [2.94, 24.04] 2.975**
analyses based on frequencies). We computed a Condition 9 !30.036 8.926 [!21.47, !48.15] !3.365***
generalized linear mixed model using lme4 package Trial Type
in R Version 3.2 (Bates, Maechler, Bolker, & Walker,
*p < .05. **p < .01. ***p < .001.
2015) with three fixed effect factors, namely condi-
tion (i.e., experimental condition vs. no model con-
trol) as the between-subjects factor, task (i.e., slider
100%
vs. container task) as the within-subjects factor and
Condition 9 Task as interaction term. Subjects were 80%

*
60%
Table 1 47% 47%
40% 40%
Children’s Helping Rates in Each Condition by Task in Study 1 40%
27%
Prosocial No model 20% 13%
model condition control
Task type M (SD) M (SD) 0%
Slider Task Container Task Overall
Experimental tasks Prosocial Model Condition No Model Control
Container task (0–3 objects) 0.40 (0.83) 0.80 (1.08)
Slider task (0–3 objects) 0.80 (1.01) 0.27 (0.80) Figure 3. Percentage of helpers during test phase in Study 1.
Overall (0–6 objects) 1.20 (1.70) 1.07 (1.58) Note. Overall describes the percentage of children that helped in
Transfer task (0–10 objects) 0.87 (1.68) 1.67 (2.02) at least one task. *p < .05. [Color figure can be viewed at
wileyonlinelibrary.com]
Observing a Prosocial Model Increases Helping 1795

Discussion children in the prosocial model condition might not


have imitated the observed actions in a narrow
Study 1 provides the first, although mixed, evi- sense (i.e., copying both correct means and goals;
dence that observing a prosocial model helping a Tomasello, 1999; Want & Harris, 2002) but rather
needy person significantly influences infants’ early imitated observed actions based on a different
instrumental helping. In the slider task, significantly understanding of goals. To address these issues and
more children helped in the experimental condition to further elaborate on our research question we
than in the control condition. This finding extends conducted a second study.
previous studies on the positive effect of prosocial
models to, first, younger ages, and, second, another
domain of prosocial behavior, namely instrumental
Study 2
helping (e.g., Radke-Yarrow et al., 1983; Williamson
et al., 2013). The main goal of Study 2 was to replicate and
Yet, the effect in the present study was task thereby substantiate the findings of Study 1 with
dependent. Namely, we found a clear effect for the improved materials and an optimized procedure.
slider task but not for the container task or the Among others, we included an initial warm-up
transfer task. In addition, children’s overall helping phase with E3 to prevent a floor effect of infant’s
rate was relatively low. Because helping during the helping rate (see Discussion above). Furthermore,
test phase occurred under zero acquaintance, that we wanted to explore to which extent infants’ pro-
is, the infant and E3 did not socially interact up to clivity to imitate helping actions is based on a gen-
that point, this finding supports the assumption by uine prosocial understanding of the modeled
Cortes Barragan and Dweck (2014), namely that helping action. Toward this end, we added a sec-
mutual engagement provides the necessary grounds ond control condition (i.e., no neediness control) in
for prosocial interaction. In particular, they com- which children observed exactly the same target
pared 25-month-olds’ helping behavior between a behavior, but in this condition, the “helpee” did not
joint play condition (i.e., toddlers initially played need the model’s help (see description of no needi-
jointly with an experimenter, e.g., rolling a ball ness control below).
back and forth) and a parallel play condition (i.e.,
toddlers did not play jointly with the experimenter,
but both played in parallel, e.g., each one played Method
solitary with his or her ball) and found that tod-
Sample and Design
dlers less frequently assisted a needy experimenter
in a subsequent test phase in the parallel play con- The sample in Study 2 consisted of N = 60 16-
dition (on average, they helped in only 1.2 of 4 out- month-old infants (M = 16.11 months; Range: 15;0–
of-reach trials in this condition), which proposes 18;0). Nine children were excluded, due to distress
that mutual engagement in terms of joint play with during testing (i.e., intense and/or repeated crying,
an unfamiliar adult likely elicits instrumental help- n = 8) or due to technical problems (n = 1), result-
ing in children. Based on these considerations, ing in a final sample of N = 51 children. Children
Study 2 includes a standardized familiarization were recruited in a similar way to Study 1 and thus
phase to increase infants’ helping rates. had comparable socioeconomic backgrounds. Data
Moreover, the design of Study 1 does not resolve were collected between March 2016 and October
if infants considered helpers’ prosocial intention 2016. Children were randomly assigned to one of
during the modeling phase, which was to assist the three conditions (prosocial model condition, n = 16; no
helpee in fulfilling her need. Although previous model control, n = 16; no neediness control, n = 19).
research has established that 9- to 18-month-old
infants have a prosocial understanding of helping
General Procedure
actions (i.e., they understand others’ needs and
actors’ prosocial intentions; K€
oster, Ohmer, Nguyen, The general procedure was almost identical to
& K€ artner, 2016), infants might alternatively have the procedure in Study 1 (see above); that is, we
understood the helper’s behavior during the model- had an initial warm-up phase followed by two
ing phase as driven by different goals, for example, experimental tasks (i.e., slider task and container
following a behavioral rule that “objects must be task) and a transfer task in between. Each experi-
put on tables” or a motive to socially interact with mental task consisted of a modeling phase (in which
the other person (Paulus, 2015). In consequence, children observed different modeling behaviors
1796 Schuhmacher, K€
oster, and K€artner

depending on condition) followed directly by a test Data Coding


phase (in which children could assist E3). Critically,
conditions only differed in the modeling phase. Helping scores. Helping scores were identical to
In contrast to Study 1 we included a standard- those in Study 1, that is, 0–3 cups or bricks could
ized warm-up play between the child and E3. After be passed over to E3 by the child in each
the warm-up with E1, children interacted with E3 experimental task, and children could assist in tid-
(i.e., before test phases) based on a standardized ing up 0–10 plates during the transfer task. 20% of
script. Namely, E3 and the child first played videos were coded by two observers. Interrater reli-
together with a spinning top. E3 activated the spin- ability was perfect for all helping scores, Cohen’s
ning top for approximately 10 s, then handed it js = 1.0.
over to the child for 20 s, after which E3 requested
the toy back from the child. This sequence of recip-
rocal play was repeated twice. The procedure was Results
repeated with a second toy (i.e., mechanical snail,
Preanalyses
mechanical caterpillar, or jingle bell), overall result-
ing in approximately 5 min of standardized recipro- Preanalyses indicated that age was not signifi-
cal play between the child and E3. cantly associated with children’s helping rates, all |r|
Second, we adapted materials as follows: We s < .17, all ps > .05. In addition, there were no signif-
used bricks (which E1 and E3 needed for stacking icant gender, order, or material effects, all |t|s < 1.97,
towers) instead of balls, as Study 1 indicated that all ps > .05. Finally, children did not differ in initial
balls might have been too attractive for children helping (i.e., clothespin task) between conditions,
(i.e., experimenters noticed that children often Mprosocial model = 0.31, SD = 0.87, Mno model control =
started playing with balls during the test phase; 0.06, SD = 0.25, and Mno neediness control = 0.20,
preanalyses in Study 1 indicated that there was a SD = 0.66, F(2, 48) = 0.56, p = .572. Because initial
tendency for children to help less with balls than helping rates were very low, we analyzed the per-
cups). In addition, we cut the handle of the slider centage of infants that helped at least once initially
to further improve usability for infants and opti- and found no significant differences between condi-
mized the setup of the tables. tions (11% helpers in prosocial model, 6% helpers in
Finally, we added a control condition (i.e., no no model control, and 11% helpers in no neediness
neediness control, see Figures 1 and 2). During the control condition; Freeman-Halton extension of Fish-
modeling phase in the no neediness control condi- er’s exact test: p > .99).
tion, infants saw exactly the same modeling actions
by E2 as in the prosocial model condition. How-
Infants’ Imitation Behavior
ever, in contrast to the prosocial model condition,
there was an alternative set of three objects (i.e., As expected, average helping rates were higher
cups or bricks) on the leftmost side of the table (i.e., than in Study 1 (across tasks: Mstudy1 = 1.1, SD =
in addition to the set on the floor; see Figure 2). 1.6 versus Mstudy2 = 2.7, SD = 2.5, t(90) = !3.25,
Furthermore, E1 did not communicate any need for p < .001, d = .72, CI 95% [0.26, 1.19]). A 2 9 2
the objects on the floor but used the three addi- mixed-analysis of variance (ANOVA) with condi-
tional objects on the table instead: In particular, as tion as the between-subjects factor (i.e., experimen-
E1 started building the second tower, she leaned tal condition vs. no model control vs. no neediness
over the table and, similarly as in the other two control) and task (i.e., slider vs. container task) as
conditions, stretched out for the first alternative the within-subject factor yielded a significant
object. After 10 s, she finally took the object and main effect of condition, F(2, 48) = 3.98, p = .025,
said: “Mhh, I can reach the cup! Although it is g2p = .14, 90% CI [0.01, 0.27] (see Figure 4). The
quite far away!”. Simultaneously, E2 (i.e., model) main effect of task was nonsignificant, F(1,
got up, went over to the table, took the (from his 48) = 0.54, p = .466, g2p = .01, 90% CI [0.00, 0.10].
perspective) rightmost container off the ground and Post hoc comparisons indicated that infants in the
put it on the table. E1 and E2 did not interact experimental condition helped significantly more
as both experimenters basically focused on their often than in the no model control, Tukey honestly
actions. This sequence was repeated for the second significant different (HSD) = 1.13, p = .022. Further-
and third target object. more, the no neediness control condition was in
Test phase in all conditions was identical to test between and did not differ from the experimental
phase in Study 1. or the no model control condition, Tukey-
Observing a Prosocial Model Increases Helping 1797

Helping Score (# of Objects Given to E3)


5

4
* *
3

0
Helping Slider Task Helping Container Task Helping Overall

Prosocial Model Condition No Model Control No Neediness Control

Figure 4. Number of objects given to the experimenter during test phase in Study 2. *p < .05. [Color figure can be viewed at wileyonli-
nelibrary.com]

6 control condition (56.25%), v2 = 6.0, p = .014, two-


sided, 95% CI [7.22, 61.14], whereas both conditions
Helping Score (= # of Plates Given to E3)

5
did not differ significantly from the no neediness
4
control condition (73.68%), v2 = 2.46, p = .12, x = .27,
95% CI [!6.35, 43.12], and v2 = 1.17, p = .28, x = .18,
3 95% CI [!13.09, 44.69].

2
Discussion
1
In summary, the adaptations in Study 2 were
0 successful: (a) the overall helping rate was higher
Prosocial Model Condition No Model Control No Neediness Control than in Study 1, (b) the pattern of results was con-
Figure 5. Number of objects given to the experimenter in transfer
sistent across tasks, and (c) the significant main
task in Study 2. [Color figure can be viewed at wileyonlinelibrar- effect for condition could be replicated, that is,
y.com] infants in the prosocial model condition helped sig-
nificantly more often than infants in the no model
control condition (i.e., they reached on average 3.9
HSD = !0.36, p = .625, and Tukey-HSD = 0.76, p = compared to 1.6 of 6 objects). Overall, these find-
.14, respectively. ings further strengthen our conclusion that observ-
Finally, the pattern of results in the transfer task ing a prosocial model (i.e., observing a needy
was similar to the pattern for the main helping person and a prosocial model assisting the person
tasks (see Figure 5). An ANOVA with condition as fulfilling her need) substantially affects early help-
the between-subjects factor and helping (i.e., ing behavior in 16-month-old infants.
infants’ assistance with dropped plates) as the Moreover, findings from the plate task indicated
dependent variable yielded no significant main that infants also generalized imitation to different
effect of condition, F(2, 48) = 1.79, p = .18, g2p = .07, helping contexts, as they helped more sponta-
90% CI [0.00, 0.18]. When dichotomizing the data neously in collecting plates after having observed a
from the transfer task (i.e., helpers vs. nonhelpers), prosocial model in a different context (i.e., model
there were significantly more helpers in the proso- assisting helpee in typical out-of-reach situa-
cial model condition (93.75%) than in the no model tion), zalthough this effect was only significant for
1798 Schuhmacher, K€
oster, and K€artner

dichotomous scores—possibly due to a lack of suffi- These findings raise the intriguing question of
cient power. what infants actually learn from observing prosocial
In Study 2, we used a no neediness control con- models. According to the sociocognitive learning
dition in order to shed some light on the question model proposed by Bandura (1986), the discovered
of whether infants selectively imitate prosocial effects may indicate that the observation of proso-
actions by considering the helpee’s neediness and cial models affects (a) infants’ acquisition of new
the actors’ prosocial intention. Our results showed skills (i.e., observation extends behavioral repertoire
that in the no neediness control condition, infants’ to help others), as well as (b) the performance of
helping rates (i.e., on average, 3.2 of 6 objects) were acts that are already a part of infants’ behavioral
somewhere in between the helping rates in the repertoire (i.e., observation increases motivation to
experimental condition and the no model control perform the observed helping actions). According
condition. If infants’ helping had been similar as in to the first explanation, some children certainly
the no model control condition (and significantly acquired the new skill of handling an arbitrary and
different from the prosocial model condition), this novel slider to assist the needy experimenter, indi-
would have supported the idea that infants particu- cating that infants had learned new means to help
larly imitate genuinely prosocial behavior that is others by observing others. According to the second
oriented at the neediness of the other. However, the explanation, children may have imitated the help-
present findings do not clearly answer this question ing action because of an increased motivation to
and can be interpreted in at least two different perform this behavior after repeated observation;
ways: (a) Infants may attribute alternative goals to that is, infants were more willing to help others if
the modeled behavior that they then imitate, for they had observed prosocial models earlier. This
instance, that things should be put on a table. How- motivational interpretation is further supported by
ever, recent findings indicate that infants do not the finding that infants in the prosocial model con-
simply help to restore order (Hepach, Vaish, Gross- dition also tended to help more in the transfer task.
man, & Tomasello, 2016). (b) In the no neediness Alternatively, one could also argue that infants in
control condition, some infants might (erroneously) the no model control condition had a decreased
have ascribed an unfulfilled need to E1 and, in con- motivation to intervene, as they had observed a
sequence, understood E2’s action as prosocial. passive experimenter in this condition. This
In Study 2, we included a standardized warm- observed passivity might have led children to
up phase including reciprocal play between the assume that the “right” behavior is to remain pas-
infant and the individual requiring help later, sive and not help the needy person in the test
which led to increased helping rates in Study 2 as phase (i.e., infants imitated E2’s passive behaviors).
compared to Study 1, possibly by establishing an Note that these alternatives would still be the result
atmosphere of mutual care and commitment (see of social learning and would thus also support the
Cortes Barragan & Dweck, 2014). Although this conclusion that modeling affects infant’s helping.
interpretation is speculative, our experiments However, the very low initial helping rates during
clearly substantiate the importance of former social the warm-up phase in both studies make the proso-
interaction for infant’s subsequent willingness to cial learning interpretation more plausible. Regard-
assist others. Thereby, we hope to sensitize experi- ing future studies, it would be interesting to
menters to the relevance of warm-up phases and differentiate whether infants imitate others passivity
initial play, as 3–5 min of reciprocal play might be or others helping behavior or whether both effects
sufficient to boost infant’s helping behavior. add up.
Furthermore, the observation of the needy per-
son might also play a crucial role in affecting
infant’s helping behavior in at least two ways.
General Discussion
First, it might be the case that observing the needy
The presented experiments provide the first evidence person alters infant’s empathic arousal, which
that 1-year-olds’ helping behavior is influenced by likely affects their motivation to help the experi-
observing a prosocial model that helps a needy per- menter in the subsequent test phase (see Hepach,
son. The current findings extend the seminal findings Vaish, Grossman, & Tomasello, 2016). However,
by Bandura and others on the importance of observa- the actor in the no model condition similarly
tional learning in children (in the context of antisocial expressed her need for the object, before she
behavior) to the domain of helping behavior at a helped herself. In consequence, in all conditions,
younger age than studied previously. infants have seen a needy other. Second, it could
Observing a Prosocial Model Increases Helping 1799

be that having observed a person who, in the end, of the uncovered effect. Regarding limitations, fur-
helps herself (as in the no model control condition) ther adaptations of the procedure in the no needi-
lowers children’s subsequent helping tendencies, ness control condition might be useful to prevent
because the agent could help herself previously. potential misinterpretation in infants (i.e., minimize
We think that this explanation is unlikely, because probability that children erroneously ascribe unful-
the infants see different experimenter in the mod- filled instrumental need to E2 in this condition;
eling and the test phase. Having said this, future also, see above). Finally, one could implement
research could explicitly address the question further control conditions, for example, a control
whether observing others’ neediness alone (i.e., condition in which E1 repeatedly expresses an
their need for external help) might increase infant’s unfulfilled need (e.g., E1 stretches out for an object,
helping. but is finally unable to reach it autonomously) and
The present findings complement recent research E2 (i.e., the model) remains passive. Such a condi-
on the role of social learning processes in infants’ tion could be useful to strengthen the interpreta-
early helping behavior, as they suggest that chil- tion, that observing prosocial models is an effective
dren’s prosocial development is potentially affected social learning mechanism in infants’ early helping
not only by direct and active structuring of helping behavior. More specifically, with such a condition
situations by others (e.g., Dahl, 2015; Dahl et al., one could test whether observing prosocial models
2017; Hammond & Carpendale, 2015; K€ artner, 2018; is more effective than simply observing needy but
K€oster, Cavalcante, et al., 2016; K€ oster, Schuh- helpless persons, which generally induces high
macher, & K€ artner, 2015; Pettygrove, Hammond, levels of empathic arousal in infants (Hepach et al.,
Karahuta, Waugh, & Brownell, 2013; Waugh, Brow- 2016).
nell, & Pollock, 2015) but also through learning by The presented findings give rise to intriguing
the observation of prosocial models. In particular, future research questions on the role of observa-
our experimental studies show that observing tional learning in infants’ helping behavior. Future
prosocial models is an effective social learning studies could particularly address the following
mechanism by which infant’s early helping behav- questions: (a) How effective are potential “interven-
iors, and possibly, their prosocial motivation can be tions” in the short and the long term? Is there
influenced. Moreover, observational learning something like deferred-imitation of prosocial acts?
might be particularly effective in older infants (b) How effective is repeated observation of paren-
(i.e., 16-month-olds) whereas scaffolding more tal prosocial activities at home? Finally, (c) How do
strongly affects helping in younger infants (i.e., 12– models’ characteristics (e.g., age, competence level,
15 months; see Dahl, 2015; Dahl et al., 2017). reputation, etc.) affect infants’ imitation of helping
Thereby, our findings also refer to the importance behavior?
of prosocial role models (i.e., caretakers and other
persons that children can observe regularly), as Conclusion
they critically affect children’s early prosocial devel-
opment. Thus, the current data provide a proof of To conclude, our findings show that observing
principle for the role of prosocial models for prosocial models helping a needy other shapes 16-
infants’ early helping behavior. Future studies may month-olds’ helping activities. Thereby, our find-
further look at the role of imitation in more natural- ings point to the relevance of observational learning
istic settings, for example, in the scope of home in the socialization of infants’ earliest prosocial
observations, interviewing caretakers and longitudi- behavior, beyond much studied effects of direct
nal correlational studies on caretakers’ prosocial socialization, such as parental scaffolding or parent-
behavior during infant’s everyday activities (see ing style. Thereby the present findings set an
Dahl, 2017). important direction for future research on the role
Of course, our study comes with both specific of social learning in the early ontogeny of human
strengths and limitations. Besides complementing prosocial behavior.
the existing literature on a critical social learning
mechanism in infants’ early helping behavior, the
strengths of the presented studies are the live References
model method, which is adequate for infants (Barr Bandura, A. (1965). Influence of models’ reinforcement
& Hayne, 1999) and the replication of evidence contingencies on the acquisition of imitative responses.
using the refined methods in Study 2 (Open Science Journal of Personality and Social Psychology, 1, 589–595.
Collaboration, 2015), substantiating the robustness https://doi.org/10.1037/h0022070
1800 Schuhmacher, K€
oster, and K€artner

Bandura, A. (1986). Social foundations of thought and action: Hammond, S. I., & Carpendale, J. I. (2015). Helping chil-
A social cognitive theory. Englewood Cliffs, NJ: Prentice- dren help: The relation between maternal scaffolding
Hall Inc. and children’s early help. Social Development, 24,
Bandura, A., Ross, D., & Ross, S. A. (1963). Imitation of 367–383. https://doi.org/10.1111/sode.12104
film-mediated aggressive models. Journal of Abnormal Hepach, R., Vaish, A., Grossmann, T., & Tomasello, M.
and Social Psychology, 66, 3–11. https://doi.org/10. (2016). Young children want to see others get the help
1037/h0048687 they need. Child Development, 87, 1703–1714. https://d
Barr, R., & Hayne, H. (1999). Developmental changes in oi.org/10.1111/cdev.12633
imitation from television during infancy. Child Develop- K€
artner, J. (2018). Beyond dichotomies—(M)others’ struc-
ment, 70, 1067–1081. https://doi.org/10.1111/1467- turing and the development of toddlers’ prosocial
8624.00079 behavior across cultures. Current Opinion in Psychology,
Bates, D., Maechler, M., Bolker, B., & Walker, S.(2015). 20, 6–10. https://doi.org/10.1016/j.copsyc.2017.07.040
Fitting linear mixed-effects models using lme4. Journal K€
oster, M., Cavalcante, L., Vera Cruz de Carvalho, R., D^ ogo
of Statistical Software, 67, 1–48. https//doi.org/10. Resende, B., & K€ artner, J. (2016). Cultural influences on
18637/jss.v067.i01. toddlers’ prosocial behavior: How maternal task assign-
Blake, P. R., Corbit, J., Callaghan, T. C., & Warneken, F. ment relates to helping others. Child Development, 87,
(2016). Give as I give: Adult influence on children’s giv- 1727–1738. https://doi.org/10.1111/cdev.12636
ing in two cultures. Journal of Experimental Child Psy- K€
oster, M., Ohmer, X., Nguyen, T. D., & K€ artner, J.
chology, 152, 149–160. https://doi.org/10.1016/j.jecp. (2016). Infants understand others’ needs. Psychological
2016.07.010 Science, 27, 542–548. https://doi.org/10.1177/0956797
Brownell, C. A. (2016). Prosocial behavior in infancy: The 615627426
role of socialization. Child Development Perspectives, 10, K€
oster, M., Schuhmacher, N., & K€ artner, J. (2015). A cul-
222–227. https://doi.org/10.1111/cdep.12189 tural perspective on prosocial development. Human Ethol-
Cortes Barragan, R., & Dweck, C. S. (2014). Rethinking ogy Bulletin, 30, 71–82. https://doi.org/10.22330/heb.
natural altruism: Simple reciprocal interactions trigger Meltzoff, A. N. (1995). Understanding the intentions of
children’s benevolence. Proceedings of the National Acad- others: Re-enactment of intended acts by 18-month-old
emy of Sciences of the United States of America, 111, children. Developmental Psychology, 31, 838–850. https:
17071–17074. https://doi.org/10.1073/pnas.1419408111 //doi.org/10.1037/0012-1649.31.5.838
Dahl, A. (2015). The developing social context of infant Open Science Collaboration (2015). Estimating the repro-
helping in two US samples. Child Development, 86, ducibility of psychological science. Science, 349, 943.
1080–1093. https://doi.org/10.1111/cdev.12361 https://doi.org/10.1126/science.aac4716
Dahl, A. (2017). Ecological commitments: Why develop- Paulus, M. (2015). Children’s inequity aversion depends
mental science needs naturalistic methods. Child Devel- on culture: a cross-cultural comparison. Journal of Exper-
opment Perspectives, 11, 79–84. https://doi.org/10.1111/ imental Child Psychology, 132, 240–246. https://doi.org/
cdep.12217 10.1016/j.jecp.2014.12.007
Dahl, A., Satlof-Bedrick, E. S., Hammond, S. I., Drum- Pettygrove, D. M., Hammond, S. I., Karahuta, E. L.,
mond, J. K., Waugh, W. E., & Brownell, C. A. (2017). Waugh, W. E., & Brownell, C. A. (2013). From cleaning
Explicit scaffolding increases simple helping in younger up to helping out: Parental socialization and children’s
infants. Developmental Psychology, 53, 407–416. early prosocial behavior. Infant Behavior and Develop-
https://doi.org/10.1037/dev0000244 ment, 36, 843–846. https://doi.org/10.1016/j.infbeh.
Eisenberg, N., Fabes, R. A., & Spinrad, T. L. (2006). Proso- 2013.09.005
cial development. In N. Eisenberg, W. Damon, & R. M. Presbie, R. J., & Coiteux, P. F. (1971). Learning to be gen-
Lerner (Eds.), Handbook of child psychology. Vol. 3: erous or stingy: Imitation of sharing behavior as a func-
Social, emotional, and personality development (6th tion of model generosity and vicarious reinforcement.
ed., pp. 646–718). Hoboken, NJ: Wiley. Child Development, 42, 1033–1038. https://doi.org/10.
Elsner, B. (2014). Theorien zu Imitation und Hand- 2307/1127789
lungsverst€andnis. In L. Ahnert (Ed.), Theorien in der Radke-Yarrow, M., Zahn-Waxler, C., & Chapman, M.
Entwicklungspsychologie [Theories of imitation and (1983). Children’s prosocial dispositions and behavior.
action-understanding] (pp. 310–329). Berlin, Germany: In E. M. Hetherington (Ed.), Handbook of child psychology
Springer. https://doi.org/10.1007/978-3-642-34805-1_12 4: Socialization, personality and social development (4th ed.,
Gergely, G., Bekkering, H., & Kir"aly, I. (2002). Develop- pp. 469–545). New York, NY: Wiley.
mental psychology: Rational imitation in preverbal Rheingold, H. L. (1982). Little children’s participation in the
infants. Nature, 415, 755. https://doi.org/10.1038/ work of adults, a nascent prosocial behavior. Child Devel-
415755a opment, 53, 114–125. https://doi.org/10.2307/1129643
Grusec, J. E., & Davidov, M. (2010). Integrating different Rogoff, B. (2003). The cultural nature of human development.
perspectives on socialization theory and research: A Oxford, UK: Oxford University Press.
domain-specific approach. Child Development, 81, 687–709. Rushton, J. P. (1975). Generosity in children: Immediate
https://doi.org/10.1111/j.1467-8624.2010.01426.x and long-term effects of modeling, preaching, and
Observing a Prosocial Model Increases Helping 1801

moral judgment. Journal of Personality and Social Psychol- Waugh, W., Brownell, C., & Pollock, B. (2015). Early
ogy, 31, 459–466. https://doi.org/10.1037/h0076466 socialization of prosocial behavior: Patterns in parents’
Staub, E. (1971). A child in distress: The influence of nur- encouragement of toddlers’ helping in an everyday
turance and modeling on children’s attempts to help. household task. Infant Behavior and Development, 39,
Developmental Psychology, 5, 124–132. https://doi.org/ 1–10. https://doi.org/10.1016/j.infbeh.2014.12.010
10.1037/h0031084 Williamson, R. A., Donohue, M. R., & Tully, E. C. (2013).
Tomasello, M. (1999). The human adaptation for culture. Learning how to help others: Two-year-olds’ social
Annual Review of Anthropology, 28, 509–529. https://doi. learning of a prosocial act. Journal of Experimental Child
org/10.1146/annurev.anthro.28.1.509 Psychology, 114, 543–550. https://doi.org/10.1016/j.jec
Want, S. C., & Harris, P. L. (2002). How do children ape? p.2012.11.004
applying concepts from the study of non-human pri- Zmyj, N., & Seehagen, S. (2013). The role of a model’s
mates to the developmental study of “imitation” in age for young children’s imitation: A research review.
children. Developmental Science, 5, 1–14. https://doi. Infant and Child Development, 22, 622–641. https://doi.
org/10.1111/1467-7687.00194 org/10.1002/icd.1811
Warneken, F., Chen, F., & Tomasello, M. (2006). Coopera-
tive activities in young children and chimpanzees. Child
Development, 77, 640–663. https://doi.org/10.1111/ Supporting Information
j.1467-8624.2006.00895.x
Warneken, F., & Tomasello, M. (2007). Helping and Additional supporting information may be found in
cooperation at 14 months of age. Infancy, 11, the online version of this article at the publisher’s
271–294. https://doi.org/10.1111/j.1532-7078.2007.tb00 website:
227.x Appendix S1. Study 1—Is the Significant Differ-
Warneken, F., & Tomasello, M. (2009). The roots of ence (Between Conditions) Dependent on Dichoto-
human altruism. British Journal of Psychology, 100, mizing Helping Rates?
455–471. https://doi.org/10.1348/000712608X379061

You might also like