Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Articles

Cooperative Problem-Solving and Teaching in


Preschoolers
Jennifer Ashley and Michael Tomasello, Emory University

Abstract
The current study investigated the ontogenetic origins of children’s skills of cooperative
problem-solving in a task involving two complementary roles. Participants were peer
dyads of 24, 30, 36, and 42 months of age. Primary dyads were initially presented with
an instrumental problem whose solution required them to cooperate by coordinating
two complementary actions. To further investigate their understanding of the task,
these same dyads were then presented with the same problem but with roles reversed.
Finally, after each of these primary participants had demonstrated proficiency in both
roles, each was separately paired with a naive peer and given the opportunity to teach
the naive partner the task. A clear ontogenetic trend emerged. Even with adult assis-
tance, 24-month-old children never became independently proficient at the task. Thirty-
and 36-month-old children became proficient mostly independently, but only relatively
slowly and without demonstrating extensive amounts of behavioral coordination or the
use of explicitly directive language to facilitate coordination. Although they did show
evidence of recognizing when a peer was new to the task, children of this age engaged in
little explicit teaching of naive peers. In contrast, 42-month-old children mastered the
task much more quickly than the other children, responded much more quickly and
accurately when their roles were reversed, coordinated both their actions and language
in the task to a much greater extent, and engaged in more explicit teaching of naive
peers. Results are discussed in terms of the developing social cognitive skills that enable
children from 2 to 4 years of age to understand other persons as mental agents with
whom they may share mental perspectives.

Keywords: Cooperation; peers; peer tutoring; problem-solving; cultural learning

Much of children’s cognitive development takes place as they participate in cultural


activities structured by adults. In children’s peer interactions, on the other hand, it is
up to the children themselves to structure their activities. Piaget (1932) suggested
that such peer interactions may play a special role in cognitive development, and this
hypothesis has been further elaborated and explored by Cultural Psychologists such
as Damon (1977, 1983) and Rogoff (1990). The main point is that peer interactions

This research was supported by a grant from the Spencer Foundation to the second author. Thanks to
Ann Kruger for helpful comments on the manuscript.
Address correspondence to Jennifer Ashley, Department of Psychology, Emory University, Atlanta,
GA 30322, USA.

© Blackwell Publishers Ltd. 1998. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street,
Malden, MA 02148, USA.
144 Jennifer Ashley and Michael Tomasello
are characterized by a symmetry of power, and quite often a symmetry of knowl-
edge, that forces children to play a more active role than when they are interacting
with adults. In general, whereas many adult-child interactions may be described as
adults ‘scaffolding’ children’s activities, peer interactions are best described as chil-
dren ‘co-operating’ and ‘co-constructing’ behavioural outcomes and cognitive prod-
ucts that neither could have created on their own.
Cooperation prototypically involves two or more children coordinating their
behavior in some mutually satisfying way. An important subset of cooperative activ-
ities, however, are those in which two children must cooperate in order to solve an
external problem, typically called cooperative problem-solving. Cooperative prob-
lem-solving has special qualities because it involves the coordinated use of both
cooperative skills and problem-solving skills. Sometimes in a problem-solving con-
text one child is more knowledgeable or skilful than another and so does something
to assist her; this is typically called teaching or instruction.
Almost all of the work on children’s cooperative problem-solving skills has been
conducted with schoolage children, or with children just making the transition to
school (e.g., Gauvain & Rogoff, 1989; Radiszewska & Rogoff, 1988; see Tudge &
Rogoff, 1989, for a review). Studies in a number of different cognitive domains have
shown that children of this age advance more in their understanding of a problem
when they attempt to solve it with a peer than when they attempt to solve it indepen-
dently. These effects have been observed in such tasks as conservation (Doise &
Mugny, 1979), spatial coordination (Emler & Valiant, 1982), mathematical problem
solving (Phelps & Damon, 1989), block building (Azmitia, 1988), computer pro-
gramming (Tudge & Winterhoff, 1993), and moral reasoning (Kruger, 1992). An
interesting variation on this theme occurs when one child knows more about a task
than another, and so attempts to engage in instruction. Again almost all of the stud-
ies have focused on schoolage children, with the basic finding being that children
become more skillful instructors of other children during the schoolage years (e.g.,
Azmitia & Hesser, 1993; Barron & Foot, 1991; Cooper, Ayers-Lopez, & Marquis,
1982; Duran & Gauvain, 1993; Ellis & Rogoff, 1982; and Pratt, Scribner, & Cole,
1977).
There is very little research on the ontogenetic origins of cooperative problem-
solving and teaching skills in the developmental period prior to schoolage. A num-
ber of studies of toddlers and preschoolers have documented their ability to
coordinate behavior in various play activities with peers (e.g., Eckerman, Davis, &
Didow, 1989), and many studies have documented the ability of preschoolers to
solve problems when interacting with adults (see Rogoff, 1990, for a review).
However, to our knowledge there are only two systematic studies of the cooperative
problem solving abilities of preschoolers interacting with peers. First, Cooper (1980)
investigated the abilities of 3.5- and 4.5-year-olds to cooperatively solve a balance
scale problem in which each member of a dyad had to place blocks of equal weight
onto his or her side of the scale so as to make it balance. Not only were 4-year-olds
better at coordinating behavior in this task than 3-year-olds, they were also better at
coordinating their language with the task as well, using more directive language to
assist their partners.
Second, Brownell and Carriger (1990, 1991) investigated the ability of toddlers
and preschoolers to coordinate their behavior in a series of problem-solving tasks. In
all instances one child was required to maneuver a lever so that a reward was made
accessible to another child standing some distance away. The lever was spring-
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 145
loaded, so that if the first child released it to go to retrieve the newly accessible
reward herself, the reward would revert to an inaccessible location. Participants were
children of 12-, 18-, 24-, and 30-months of age. Twelve-month-olds proved to be
very unskillful at the task. Although 18-month-olds managed to be successful on
some occasions, generally they were very unskillful as well; for example, those who
were successful on one occasion were almost never able to repeat their success. Most
of the 24- and 30-month-old children were more successful, repeating their successes
on multiple occasions. However, when solution styles were examined, only 30% of
the 24-month-olds actually showed clear signs of intentionally coordinating their
behavior with that of the partner (e.g., the manipulator might wait for the retriever
to approach the access location before actually manipulating the lever), whereas 55%
of the 30-month-old children intentionally coordinated their behavior with their
partner’s, clearly anticipating what the other would do next. Thirty-month-olds were
also the only age group to make extensive use of directive verbalizations to influence
their partner’s behavior.
In terms of peer teaching, we are aware of only one study in which young
preschoolers were subjects (see Cooper, Ayers-Lopez, & Marquis, 1982, for a study
with children making the transition to school). Koester and Bueche (1980) observed
4-year-olds, who had been given pre-training in a block design task, attempting to
instruct 3-year-olds in that same task. The only clear finding was that the instructors
preferred nonverbal to verbal instruction, using mostly nonverbal demonstrations
and physical assistance to their partners.
The current study was designed to explore in greater detail the cooperative prob-
lem-solving and teaching abilities of toddlers and young preschoolers. Dyads at 24,
30, 36, and 42 months of age were given a moderately difficult task that required
cooperative efforts for a solution. The study is unique for preschoolers in that both
partners had equally difficult and important complementary roles to play. (In the
task of Cooper, 1980, the children had identical roles, and in the task of Brownell &
Carriger, 1990, the role of one child was simply to manually retrieve the reward
made accessible by the other.) In addition, after each child had become proficient in
one role, children’s understanding of the problem was assessed via a role reversal
task in which each had to play the role they had not played previously. Finally, after
these primary participants had become skillful at both roles (some children had to be
trained by an adult), each was paired with a novice peer to see how they might
attempt to instruct a child who clearly did not know how the task operated.

Methods
Participants
Fifty-six preschoolers who attended local daycare centers participated in the study.
There were eight 24-month-olds and 16 in each of the other three age groups: 2-year-
olds (m = 2 years, range = 23–25 months; 3 females, 5 males); 2.5-year-olds (M = 31
months, range = 27–32 months; 12 females, 4 males); 3-year-olds (M = 36 months,
range = 35–37 months; 8 females, 8 males); and 3.5-year-olds (M = 43 months, range
= 41–45 months; 9 females, 7 males). Children were always tested in same-age dyads.
Individuals were chosen to participate based on their age, parental permission, and
teacher recommendation. Teachers were asked to select four children, out of all
available children within a narrow age range, whom they thought would work well
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
146 Jennifer Ashley and Michael Tomasello
together (i.e., who were comfortable with one another and who would not be upset
at being paired with any of the other children participating). From that selection,
teachers were then asked to choose two children whom they thought would work
particularly well together. These two children were designated as ‘primary subjects’,
while the other two were referred to as ‘secondary subjects’. Due to difficulties in
finding four children in one class very close in age, gender was not systematically
considered in the pairings. The breakdown for the primary subjects was: 2-year-olds
= three mixed dyads and one all-male dyad; 2.5-year-olds = one mixed dyad and
three all female dyads; 3-year-olds = three mixed dyads and one all male dyad; and
3.5-year-olds = two mixed dyads and two all-female dyads.

Testing Apparatus
The main part of the testing apparatus consisted of a clear plexiglas tube, approxi-
mately four feet long, mounted on a plywood board (see Figure 1). Protruding out-
ward from the left portion of the tube was a spring-loaded lever that normally rested
in its leftmost position. Moving the lever to the right (toward the middle of the tube)
pulled a string attached to a tray clearly visible within the tube, and caused it to
move through the tube until it rested in front of a plexiglas window. On the tray for
each trial were two clearly visible ‘stickers’ (paper appliqués created for children).
The lever was covered in red duct tape and there was a red mat on the floor directly
beneath it. Protruding outward from the right portion of the tube was a spring-
loaded rotating handle that normally rested in its ‘down’ position. Rotating the han-
dle 180 degrees to its ‘up’ position pulled a string attached to the plexiglas window,
opening it, so that the stickers could be obtained. The handle was covered in blue
duct tape and there was a blue mat on the floor directly beneath it. Note that the
spring loading of the two manipulanda meant that as soon as either was released the
stickers became inaccessible; either the tray would move away from the window or
the window would shut. Thus, in order to be successful in the task (i.e., to obtain the
stickers) one child had to operate the lever, and hold it in place, while the other child
had to operate the handle and hold it in place.
A second apparatus (called the reversed apparatus) was identical in function to
the first. On this apparatus, however, the handle was on the left side of the tube and
the lever was on the right. The colors associated with each end of the apparatus
stayed the same (red on left, blue on right), meaning that their association with the
lever and handle were reversed. This arrangement was designed so that in the role
reversal task (which followed the main cooperation task) children could be
instructed to go to their same colors and ends of the apparatus, but upon arriving
there they would find the manipulandum that their partner had previously used.

Observational Procedure
Children were tested either early in the morning or in mid-afternoon, at the teachers’
discretion, in a separate room located within the daycare centre. Testing lasted from
3–6 days per dyad, with daily sessions lasting as long as children were willing to par-
ticipate, but no longer than one hour. The average session lasted about 30–45 min-
utes and was videotaped via a camera aimed at the apparatus. The single
experimenter (E) was an adult female (first author).
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 147

clear plexiglas tube and door

Door

Reward

lever handle

Red Blue
Mat Mat

Figure 1: Cooperative problem solving apparatus. Reversed apparatus had lever and handle on opposite
sides, with colored mats in their same locations.

Phase 1: cooperation task. Prior to entering the testing room the primary subjects
were told that they were going to play a game that would allow them to get stickers
that they could keep. One child was told that she must stay on the red mat at all
times, while the other child was told that she must stay on the blue mat at all times—
thus specifying the easily manipulable part of the apparatus for each child. If chil-
dren attempted to leave their mats during testing, E made sure they quickly
returned. Children were encouraged to attend to the task, as E said things like ‘Can
you get the stickers? See if you can get the stickers.’ Upon obtaining the stickers,
children had an opportunity to examine them and to attach them to a sheet of
paper that had been identified as belonging to them.
Children were given a series of two-minute trials until they became proficient at
the task, or else were declared non-proficient after 5 unsuccessful trials. A trial began
with E encouraging the children to get the stickers and ended when either (i) the chil-
dren obtained the stickers, or (ii) two minutes elapsed, at which time E removed the
stickers and gave them to the children. She then put new stickers in the apparatus
and again encouraged the children to get the stickers—thus beginning a new trial.
Proficiency was defined as success on five trials in a row, followed by success on
another five trials in a row to begin the following day’s session (to test for retention).
Children who had not learned to be successful by themselves after 5 trials on the first
day were taught by E how the apparatus worked until they became successful (or
showed clearly that they were not going to be able to become successful, in which
case testing ceased).1 To be considered proficient these instructed children had to
succeed for five trials in a row independently after the instruction was completed, as
well as during the following day’s session. Only proficient dyads proceeded to the
second phase of the experiment.
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
148 Jennifer Ashley and Michael Tomasello
Phase 2: reversal task. After a dyad had become proficient, E exchanged the main
apparatus for the reversed apparatus when the children were out of the room. When
the children returned they were reminded of the colors of their respective mats—
which led them to the role in the task they had not played previously. The children
were then given two, two-minute trials to attempt to obtain the stickers from the
reversed apparatus. If, after two trials they were unable to retrieve the stickers, they
were taught by E. Both dyads who were successful on their own as well as those who
required instruction were then required to demonstrate proficiency by successfully
retrieving the stickers five times in a row, and then again five times in a row at the
start of the next day’s session. Only children from proficient dyads proceeded to the
third phase of testing.
Phase 3: teaching. After demonstrating proficiency in both roles of the task in the
previous two phases of the experiment, each of these primary subjects was then
paired with a secondary subject who had never before experienced either version of
the task. Secondary subjects were invited to ‘come play’ by E and were told that they
would have an opportunity to get some stickers. This new dyad (consisting of one
proficient primary subject and one naive secondary subject) received no further
encouragement or instruction. Therefore, it remained up to the primary subject to
attempt to direct her new partner’s attention towards the apparatus and/or to teach
her what needed to be done. The children were given two-minute trials until they
either became proficient (success on five trials in a row) or until they demonstrated
non-proficiency on 5 trials in a row—in which case E helped them to learn the task
until they became proficient (were successful independently on five trials in a row).
Secondary subjects from proficient dyads proceeded to the fourth and final phase of
testing.
Phase 2: reversal for secondary subjects. The final phase of testing was simply a
repeat of the reversal phase for the secondary subjects—to see what they had learned
in the Teaching phase from their peer partners. Secondary subjects were paired
together and first given the opportunity to demonstrate proficiency with their new
partners playing the same role they had played in the previous phase of the experi-
ment. The dyad was then presented with the reversed apparatus for two, two-minute
trials.

Coding Procedure
Videotapes of all phases of the study (excluding any portions in which E was
instructing the children) were coded in two ways. The first coding scheme provided a
second-by-second accounting of the subjects’ problem-solving behavior towards the
apparatus and towards one another. The second coding scheme concerned subjects’
active attempts at communication with one another, and so required a preliminary
transcribing of all the verbalizations and intentional gestures that occurred. These
transcriptions were then coded on an utterance-by-utterance basis (including inten-
tional gestures as instances of, or a part of, utterances). The subjects’ utterances that
were directed to one another were separated from those directed to E. In both cases,
only utterances that were in some way related to the task were coded (i.e., requests
to go to the bathroom or comments about classrooms were ignored for purposes of
coding).
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 149
Problem-Solving behavior. The first coding scheme identified how individual sub-
jects spent their time interacting with the apparatus and their partner in an attempt
to solve the problem. Five mutually exclusive and exhaustive categories were used
to characterize each subject at each second of interaction. For purposes of analysis,
any stretch of time for which a subject was continuously engaged in one of these
behaviors was considered as one instance of that behavior (e.g., 10 continuous sec-
onds of coordinated attempting was considered one instance of a coordinated
attempt).2
Coordination Attempt (Coord Att)—Child is actively attempting to coordinate her
activity with her partner through physical maneuvering or some form of com-
munication—regardless of whether or not these attempts are successful.
Independent Manipulation (Ind Manip)—Child is independently manipulating
either the handle or the lever on the apparatus with apparent disregard for her
partner’s activity.
Investigate (I)—Child is actively investing some part of the apparatus other
than the handle or the lever (typically the stickers through one end of the
tube).
Experimenter Focus (E Focus)—Child’s attention and actions are directly focused
on the experimenter in an attempt to obtain assistance from her.
Nothing (N)—Child is not focused on the task.
Communication (attempt level). The second coding scheme was based on Cooper
(1980) and was used to categorize the different ways in which subjects communi-
cated with one another during their problem-solving activities. Seven mutually
exclusive and exhaustive categories were used to characterize each of a subject’s
attempts at task-relevant communication.
Demonstration (Demo)—Child actively shows partner how to do something,
e.g., leaning over to work her lever.
Specific Directive (Sp Dir)—Child indicates specifically what a partner should
or should not do with the apparatus, e.g., ‘Move it this way’, ‘Pull it up’,
‘No, do it slow’.
Generic Directive (Gen Dir)—Child demands action at a very general level,
without specifying the action to be performed, e.g., ‘Do it’, ‘Go on’, ‘No’,
‘Get it’.
Explanation (Expl)—Child explains to partner what she is doing, what she is
going to do, or what needs to be done. usually related to how the apparatus
works, e.g., ‘See, I’m doing this one’, ‘You’re making it go forwards’, ‘I’ll go
and do that’.
Attention Directing (Att Dir)—Child attempts to obtain or direct the attention
of the partner (either verbally or gesturally), e.g., ‘Alex!’, ‘There!’, pointing,
pushing, tapping.
Responsiveness (Resp)—Child responds to the comments or actions of partner,
e.g., doing what a partner says or asks, saying thank you, praising, imitating
(gestural or verbal).
Experimenter (Exp)—Child directs language or gestures to E. (These were fur-
ther classified as Requests for Help, Requests for Information, or
Comments.)
Communication (trial level). To provide a more general description on larger units of
time for the dyad as a unit, a derivative coding scheme for communication was also
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
150 Jennifer Ashley and Michael Tomasello
constructed on the level of the trial as a whole (the primary examination of commu-
nicative exchanges, just described, was on the level of individual subject’s individual
attempts). For purposes of this coding scheme, each trial was coded via an algorith-
mic combination of all the attempt-level codes for that trial. The five resulting mutu-
ally exclusive and exhaustive categories were:
Teaching (T)—One child uses at least one of each of the following at some time
during the trial: demonstration, specific directive, explanation, and attention
directive.
Helping (H)—One child uses at least one specific directive and attempts to direct
partner’s attention at some point in the trial.
Directing (D)—One partner uses generic directives and attempts to direct partner’s
attention, or else uses a single specific directive with or without a generic direc-
tive or incorporates an explanation into the episode.
Nonsocial (N)—Children either do not communicate or use only generic direc-
tives.
Immediate Success (Imm Succ)—Children immediately solve problem without any
communicative exchanges.
The primary coder for both coding schemes was the first author. For purpose of
obtaining reliability estimates, a research assistant, unaware of the purposes of the
study, coded 20% of the dyads at each phase. Inter-rater agreement was assessed by
computing Cohen’s Kappas. For the Problem-Solving Behavior coding scheme an
overall agreement of 90% yielded a Kappa of .78 (percent agreement ranged from
77% to 87% for the individual categories excluding Experimenter Focus, which
occurred too infrequently in the reliability observations for assessment). For the
attempt-level Communication coding scheme an overall agreement of 91% yielded a
Kappa of .78 on an utterance-by-utterance analysis (percent agreement ranged from
63% to 94% for the individual categories excluding Demonstrations, which occurred
too infrequently in the reliability observations for assessment). The Experimenter-
Directed talk sub-coding scheme yielded an overall agreement of 93% and a Kappa
of .80 (percent agreement ranged from 63% to 100% for the individual categories).
All of these Kappa values are considered to be indicative of very good to excellent
reliability (Bakeman & Gottman, 1986). Because of the algorithmic nature of the
trial-level Communication coding scheme, reliability estimation was not necessary.

Results
Results will be presented separately for subjects’ Problem-Solving Behaviors and
their attempts at Communication. For each of these dimensions of performance,
results will be presented first for the Cooperation phase, then for the Reversal phase,
and then for the Teaching phase. Following these results will be an explicit compari-
son of subjects’ performance in the Cooperation and Teaching phases, when they
had different kinds of communicative partners. Because of their overall lack of suc-
cess, a general summary of the results for 2-year-olds will be presented separately
from that of the other age groups.

Two Year Olds


No 2-year-old dyad succeeded on its own in the main cooperation task for even a
single trial. Moreover, only two of the four 2-year-old dyads were able to be taught
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 151
by E how to jointly manipulate the apparatus successfully, and neither of these sub-
sequently became proficient on their own after this teaching (i.e., neither managed 5
successes in a row independently). Therefore no 2-year-old dyad proceeded to partic-
ipate in the reversal and teaching phases of the experiment.

Problem-Solving Behavior
Table 1 presents the mean values (and statistics) of the three remaining groups
of children on the major problem-solving measures across all three phases of the
task.
Cooperation phase. As an initial assessment of dyads’ skill in the task, we compared
across age groups the average number of trials it took for dyads to have their first
success. There was a significant effect of age group: F(2,9) = 7.82, p , .01. Pairwise
post hoc comparisons revealed that 3.5-year-olds became successful in about half as
many trials (3.5) as the other two age groups, who did not differ significantly from
one another (about 6.4 trials each). The same pattern was found when the average

Table 1. Means and Test Statistics for the Main Problem-solving Measures for the
Three Age Groups Across all Three Phases of the Task (Pairwise Comparisons used
the Tukey Procedure)

3.5 3.0 2.5 Pairwise

Cooperation Phase
Latency to Success (secs) 235 511 645 3.5 , others
F(2,9) = 7.82, p , .01
% Coordinated Attempts 26.5 0.3 1.4 3.5 . others
F(2,21) = 23.45, p , .001
% Independent Manipulation 19.0 21.8 18.0 ns
% Investigate 26.3 36.4 32.5 ns
% Nothing 29.0 41.5 46.5 3.5 , 2.5
F(2,21) = 4.75, p , .05
Reversal Phase
Latency to Success 20 137 133 3.5 , others
F(2,21) = 3.19, p , .06
Teaching Phase
Latency to Success 64 112 310 others , 2.5
F(2,20) = 6.67, p , .01
% Coordinated Attempts 77.8 41.5 20.7 3.5 . others
F(2,20) = 12.10, p , .01
% Independent Manipulation 12.5 32.8 42.7 3.5 , others
F(2,20) = 8.62, p , .01
% Investigate 0.7 7.6 7.9 ns
% Nothing 8.8 18.0 28.6 3.5 , 2.5
F(2,20) = 4.59, p , .05

© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998


152 Jennifer Ashley and Michael Tomasello
latency to first success (in seconds) was compared across age groups, with 3.5-year-
olds being successful in less than half the time of either of the other two age groups
(see Table 1 for means).
To assess more specifically the types of behavior in which dyads engaged, an
overall MANOVA on mean proportions was used, with age as the between subjects
factor and Problem-Solving Behavior category as the within subjects repeated fac-
tor.3 There was an overall effect of age on the type of behavioral activity seen
among age groups, F(10,34) = 6.63, p , .01. At the level of individual categories,
univariate analyses and post hoc comparisons revealed that during this phase 3.5-
year-olds attempted to coordinate with each other proportionally more frequently
than 3-year-olds and 2.5-year-olds, who did not differ from one another. There was
also a significant difference among age groups in the proportion of behaviors
coded as Nothing, with 3.5-year-olds doing Nothing proportionally less often than
2.5-year-olds (3-year-olds did not differ significantly from either of these two age
groups).4
Overall, these results of the subjects’ initial encounters with the task during the
Cooperation phase present a very clear picture: 3.5-year-olds manipulated the appa-
ratus together in a much more coordinated way than the other two age groups, and
they were successful much more quickly and proportionally more often as a result.
One interesting additional piece of information is that when Latency to Success was
correlated with the frequency of Coordinated Attempts across all dyads (using a
Pearson Product-moment correlation), a value of –.88 was obtained (p , .001)—
indicating that the more frequently individuals in a dyad made coordinated attempts
with one another the less time it took them to solve the task.

Reversal phases. After demonstrating proficiency with the initial apparatus, both pri-
mary and secondary subjects experienced the reversal phase (i.e., for primary sub-
jects after the cooperation phase and for secondary subjects after the teaching
phase). The reversed apparatus was used to test for subjects’ general understanding
of how their efforts had succeeded in the initial cooperation phase. Because we were
concerned with subjects’ understanding prior to having extended experience with the
reversed apparatus, only two trials were given (before any adult instruction that
might be needed was administered). The most sensitive measure of the understand-
ing of the new apparatus is the time it took a dyad to reach success within these two
trials, that is, latency (in seconds) to first success. The overall ANOVA using age as
the independent measure approached significance, F(2,21) = 3.19, p = .06. Post hoc
comparisons indicated that 3.5-year-olds’ average latency to first reversal success
was many times shorter than that of both 3-year-olds and 2.5-year-olds, who did not
differ from one another (see Table 1).
As another way of assessing reversal performance as a function of age, dyads were
classified as being either successful or unsuccessful (at any time during the two tri-
als). A Fisher Exact Probability that combined the results of both primary and sec-
ondary reversal (there were thus 8 dyads for each age) showed that more
3.5-year-olds (7) were successful than either 3-year-olds (p , 0.5) or 2.5-year-olds (p
, .05) (2 each). Three-year-olds and 2.5-year-olds did not differ from one another. It
is especially interesting to note that only 3.5-year-olds were successful during sec-
ondary reversal, that is, 3 of 4 3.5-year-old dyads were successful whereas no dyads
from the younger two age groups were successful. Recall that primary subjects par-
ticipated in the reversal task with the same partners with whom they had interacted
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 153
in the initial Cooperation phase, whereas secondary subjects were paired with other
secondary subjects with whom they had never interacted in the context of this task
(each had participated initially with separate primary subjects in the Teaching
phase). This result presumably indicates that the learning of 3.5-year-old secondary
subjects in the Teaching phase was more robust than that of the younger children,
whose learning was apparently more fragile in this phase.
Again in this phase of the experiment, therefore, the clear result was that 3.5-
year-olds understood the task much better than either 3-year-olds or 2.5-
year-olds, neither of whom seemed to understand the reversed apparatus particu-
larly well.

Teaching phase. One 2.5-year-old dyad had to be excluded from all analyses of the
Teaching phase because the secondary subject was unable to learn the task at all.
Again the most sensitive measure of skill is the average latency to first teaching suc-
cess by a dyad. The pattern in this case was a bit different than in the Cooperation
Phase. In this case the 2.5-year-olds again took much more time to succeed than the
3.5-year-olds, but in this case the 3-year-olds patterned with the older children. That
is, the 3-year-olds were also quicker to learn than the 2.5-year-olds, and although
there was a tendency for 3.5-year-olds to reach success faster than 3-year-olds, these
two groups did not differ significantly (see Table 1). Analysis of the average first suc-
cessful trial among age groups found an identical pattern of results, F(2,20) = 9.63, p
, .01, with post hoc analyses revealing again that 2.5-year-olds were slower than the
other two age groups (p , .05 in both cases). These comparisons indicate that while
most secondary 3.5-year-olds and 3-year-olds were learning from their primary part-
ners within two trials, most secondary 2.5-year-olds were not.
As in the Cooperation phase, to assess more specifically the types of behavior in
which dyads engaged during this Teaching phase, an overall MANOVA on mean
proportions was performed, with age as the between subjects factor and Problem-
Solving Behavior category as the within subjects repeated factor. There was an over-
all effect of age on the type of behavioral activity seen among age groups, F(8.34) =
4.50, p , .01. At the level of individual categories, univariate analyses revealed that
3.5-year-olds attempted to Coordinate their behavior with one another proportion-
ally more often than both of the other two age groups, who did not differ from one
another. Conversely, both 2.5-year-olds and 3-year-olds Independently manipulated
the apparatus proportionally more often than 3.5-year-olds (and these two younger
ages did not differ). Finally, 2.5-year-olds did Nothing toward the apparatus pro-
portionally more often than 3.5-year-olds, but 3-year-olds did not differ from either
of these.
During this Teaching phase of the experiment, in which peers of differing levels of
skill combined their efforts in the cooperative task, 3.5-year-olds again coordinated
their behavior proportionally more often and were generally much more skillful than
2.5-year-olds. They also coordinated their behavior much more often than 3-year-
olds, but the 3-year-olds were nevertheless able to approximate their latency to suc-
cess. This pattern of results suggests that when one member of a 3-year-old dyad is
knowledgeable about the cooperative task, they manage to look a bit more like 3.5-
year-olds in terms of outcome, but their style of approach to the task is still more
individualistic. Interestingly, as in the Cooperation phase when Latency to Success is
correlated with the frequency of Coordinated Attempts across all dyads in the
Teaching phase, a value of –.52 is obtained (p , .05)—indicating again that the
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
154 Jennifer Ashley and Michael Tomasello

more frequently individuals in a dyad made coordinated attempts with one another
the less time it took them to solve the task.

Communication
During all phases of the study children communicated with their partners and
with E, both linguistically and nonlinguistically. Obviously, these communicative
attempts are another important source of information about what children under-
stood about the task. In terms of overall amount of communicative attempts,
there were no statistically significant differences among the different age groups:
F(2,45) = 1.31, p > .05. However, numerically the 2.5-year-olds (M = 45.25) pro-
duced more communicative attempts than 3-year-olds (M = 37.37), who produced
more than 3.5-year-olds (M = 24.37). These differences necessitated a method to
correct for differences in the amount of talk for individual children and dyads.
Therefore, all analyses of communicative attempts were conducted using mean
proportions. (Again, all proportions were arcsin transformed prior to statistical
analysis, and analyses of untransformed data produced an almost identical pat-
tern of results.)
Children sometimes chose to talk to E instead of to their partner, for example, to
ask for her help. There were no interesting age differences in this behavior, either as
a whole or in terms of the subclassification into Requests for Help, Requests for
Information, and Comments. All of the subsequent analyses, therefore, focus on the
communicative attempts of subjects with their partners, excluding also any talk that
was clearly not about the task itself (e.g., excluding comments about other children
back in the classroom). These partner-directed, task-relevant communicative
attempts made up 81% of all communicative attempts. Table 2 presents the mean
values (and statistics) of the three groups of children on the major communication
measures across all three phases of the task.

Cooperation phase. Table 2 depicts the mean proportion of primary subjects’ part-
ner-directed, task-relevant communicative attempts during the Cooperation phase.
An overall MANOVA found a clear effect of age: F(12.34) = 2.19, p , .05.
Univariate analyses identified the differences to fall within two categories. Three-
and-a-half-olds used proportionally more Specific Directives than both 3-year-olds
and 2.5-year-olds (who did not differ from each other). Interestingly, 3-year-olds
used proportionally more Generic Directives than both 3.5-year-olds and 2.5-year-
olds (who did not differ from one another). Although the overall numbers were low,
it is important to note that 3.5-year-olds were the only age group to use
Demonstrations during this phase (an average of about 2 per subject across all trials
of the Cooperation phase).
Similar results were found when these individual codes were combined to charac-
terize each trial as a whole in terms of the predominant mode of communication,
that is, using the Trial-level Communication coding scheme. First, there was an
overall difference in Nonsocials, F(2,21) = 3.91, p , .05. Post hoc analyses revealed
that 2.5-year-olds were Nonsocial more than 3.5-year-olds, p , .05. There was no
significant different between 3-year-olds and the other two age groups. Although the
overall numbers were low, it is important to note that 3.5-year-olds were again the
only age group to engage in trials classified as Teaching (an average of about .5 trials
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 155
Table 2. Means and Test Statistics for the Main Communication Measures for the
Three Age Groups Across all Three Phases of the Task (Pairwise Comparisons used
the Tukey Procedure)

3.5 3.0 2.5 Pairwise

Cooperation Phase
% Demonstrations 2.0 0 0 ns
% Specific Directives 17.3 2.5 2.8 3.5 . others
F(2,21) = 3.74, p , .05
% Generic Directives 16.5 40.3 15.5 3.0 . others
F(2,21) = 5.06, p , .05
% Explanations 4.8 9.8 8.1 ns
% Attention Directives 31.3 17.5 32.1 ns
% Responding 27.8 29.8 38.1 ns
Teaching Phase
% Demonstrations 3.4 0 0 3.5 . others
F(2,21) = 6.43, p , .01
% Specific Directives 34.9 15.5 5.25 3.5 . others
F(2,21) = 4.47, p , .05
% Generic Directives 13.8 30.6 19.1 ns
% Explanations 7.9 6.3 16.3 ns
% Attention Directives 29.3 36.8 35.1 ns
% Responding 10.9 11.9 23.1 ns

per subject during the Cooperation phase). Interestingly, however, when Latency to
Success was correlated with the frequency of primary subjects’ clearest Teaching
attempts (proportion of Specific Directives 1 Demonstrations), no correlation was
found.
Teaching phase. Similar to the results found during the Cooperation phase, an over-
all MANOVA found an effect of age on utterance type during the Teaching phase,
F(12,30) = 2.07, p , .05. Univariate analyses revealed that 3.5-year-olds again used
proportionally more Specific Directives than either 3-year-olds or 2.5-year-olds, who
did not differ from one another. Similar to the Cooperation phase, 3.5-year-olds
were also the only age group to use Demonstrations during the Teaching phase, and
in this case they were proportionally higher than the other two age groups, who did
not differ from one another.
Similar results were found for the Trial-level measure of Communication as well.
There were significant differences among age groups in the number of trials charac-
terized as Teaching, F(2,21) = 7.00, p , .05 and Nonsocial, F(2,21) = 4.08, p , .05.
Again 3.5-year-olds (M = .5) were the only age group to actively teaching their
peers, so they differed from the other two groups (p , .05 in both cases), who did
not differ from one another. In terms of Nonsocial, post hoc analyses revealed that
2.5-year-olds (M = 2.0) were Nonsocial more than 3.5-year-olds (M = .5), p , .05,
but there was no difference between 3.5- and 3-year-olds (M = 1.5), or 3-year-olds
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
156 Jennifer Ashley and Michael Tomasello
and 2.5-year-olds. Interestingly, when Latency to Success of the dyad was correlated
with the frequency of primary subjects’ clearest Teaching attempts (proportion of
Specific Directives 1 Demonstrations), a correlation of –.43 (p , .05) was found.
This would suggest that in the Teaching phase the primary subjects’ use of language
is more important to the dyad’s success than in the Cooperation phase.
As a way of characterizing individuals, each child was also classified in terms
of her most sophisticated communicative attempt shown for any trial during the
Teaching phase. A clear development trend was evident in which older children
who were teachers during this phase used more advanced communicative styles,
with almost half of the 3.5-year-olds incorporating some teaching into their commu-
nicative attempts with their naive partners, whereas none of the younger children
did so.

Comparisons Between the Cooperation and Teaching Phases


A final question of interest was differences in the performance of subjects in the
Cooperation and Teaching phases as an indication of the degree to which the differ-
ent age groups were sensitive to the changing skill levels of their partners. During the
initial Cooperation phase primary subjects were paired with equally naive individu-
als. During the Teaching phase, the primary subject could potentially recognize the
disparity of knowledge between herself and her naive partner and adjust her behav-
ior accordingly. The comparisons here are thus between primary subjects in the two
phases, since they are the only subjects to participate both as both a cooperative
partner and as a teacher. Table 3 presents the mean values (and statistics) of the
three groups of children on the major problem-solving and communication measures
across the Cooperation and Teaching Phases.

2.5-Year-Olds. Primary subject 2.5-year-olds produced proportionally more


Coordinated Attempts during the Teaching phase than during the Cooperation
phase. They also increased their amount of Independent Manipulations during the
Teaching as opposed to the Cooperation phase. They also Investigated the appara-
tus less during the Teaching as opposed to the Cooperation phase, and did Nothing
less during the Teaching as opposed to the Cooperation phase. This pattern of
results suggests that once 2.5-year-olds learn the task (they know very little about it
during the Cooperation phase, but by the Teaching phase they are proficient), they
manipulate the apparatus in the correct way more often and monitor the partner
while doing so.
This leaves open the question of how much 2.5-year-olds have increased in their
understanding across the phases, and how much they recognize the needs of their
new, naive partners in the Teaching phase. One clue is how they used their lan-
guage. Comparison of primary subjects across the Cooperation and Teaching
phases found no significant differences in the proportions of the different types of
communicative attempts produced. This suggests that 2.5-year-olds may not be
aware of the lack of knowledge their new partner in the teaching phase. The one
difference was that on the Trial-level, 2.5-year-old primary subjects were Nonsocial
significantly more during the Cooperation phase than during the Teaching phase,
t(13) = 3.34, p , .01, indicating again an understanding of a need for action as the
study progressed.
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 157
Table 3. Means and Test Statistics for Measures Showing Significant Differences
Between Cooperation (C) and Teaching (T) Phases (Utterance Level) for Each Age
Group (Primary Subjects Only)

Cooperation Teaching

2.5 Year Olds


% Coordinated Attempts 1.4 20.7 t(13) = 1.97, p , .05
% Independent Manipulation 18.0 42.7 t(13) = 2.43, p , .0
% Investigation 32.5 7.9 t(13) = 4.23, p , .01
% Nothing 46.5 28.6 t(13) = 3.57, p , .01
3.0 Year Olds
% Coordinated Attempts .3 41.5 t(14) = 12.93, p , .01
% Independent Manipulation 21.8 32.8 t(14) = 2.53, p , .05
% Investigation 36.4 7.6 t(14) = 5.97, p , .01
% Nothing 41.5 18.0 t(14) = 3.03, p , .05
% Communication: Sp. Directives 2.5 15.5 t(14) = 1.80, p , .05
% Communication: Att. Directives 17.5 36.8 t(14) = 2.33, p , .05
% Communication: Responding 29.8 11.9 t(14) = 2.37, p , .05
3.5 Year Olds
% Coordinated Attempts 26.5 77.8 t(14) = 4.26, p , .01
% Investigation 26.3 0.7 t(14) = 9.61, p , .01
% Nothing 29.0 8.8 t(14) = 3.41, p , .01

3-Year-Olds. Similar, 3-year-old primary subjects produced more Coordinated


Attempts with their partners during the Teaching phase than during the
Cooperation phase, and they also increased their Independent Manipulations as
they moved from the Cooperation to the Teaching phase. They also decreased their
use of Investigation as they went from the Cooperation to the Teaching phase, as
well as the amount of time they spent doing Nothing. Similar to 2.5-year-olds these
results suggest that 3-year-old have learned something about the task during the
Cooperation phase, but it is not clear that it means that they recognized the
increased needs of their new partners in the Teaching phase.
Relevant to this question is communication during the two phases. Three-year-
old primary subjects used several times more Specific Directives during the
Teaching as opposed to the Cooperation phase of the experiment. They also used
twice as many Attention Directing utterances during the Teaching than during the
Cooperation phase of the experiment. These two differences, in conjunction with
differences in problem-solving behavior, strongly suggest that 3-year-olds are at
least somewhat sensitive to the changing nature of their new partners in the
Teaching phase of the experiment; they were attempting to elicit their participation
by directing their attention towards the apparatus and providing them with specific
instructions. In addition, it is noteworthy that these children also Responded less
to their partner during the Teaching than during the Cooperation phase, indicating
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
158 Jennifer Ashley and Michael Tomasello
that they were ‘leading’ the interaction and did not feel obligated to respond to the
communicative attempts of an unknowledgeable partner. Corroborating these
results, on the trial level 3-year-old primary subjects were Nonsocial significantly
more during the Cooperation phase than during the Teaching phase of the experi-
ment, and they also encountered significantly more situations where explicit com-
munication was unnecessary (i.e. had Immediate Success) during the Teaching
phase as opposed to the Cooperative phase.
3.5-Year-Olds. As with the younger age groups, 3.5-year-olds also produced signifi-
cantly more Coordinated Attempts with their partners during the Teaching than
during the Cooperation phase. They also Investigated less during the Teaching than
during the Cooperation phase, and did Nothing less during the Teaching than dur-
ing the Cooperation phase. Thus 3.5-year-olds were also changing their behavior as
they moved from the Cooperation to the Teaching phase in a manner that demon-
strated their overall understanding of how the apparatus worked and their need for
partner assistance.
In terms of communication, there were only a few significant differences between
phases in the frequency of the different communicative attempts used by 3.5-year-
olds. Although not significantly different, the 3.5-year-olds used twice as many
Specific Directives during the Teaching phase as during the Cooperation phase, and
they Responded to their partners less than half as often in the Teaching as opposed
to the Cooperation phase. Similar to 3-year-olds, on the trial level 3.5-year-old
primary subjects were Nonsocial significantly more during the Cooperation phase
than during the Teaching phase of the experiment, t(14) = 2.66, p , .05. They also
encountered significantly more situations where explicit communication was unnec-
essary (i.e. had Immediate Success) during the Teaching phase as opposed to the
Cooperation phase, t(14) = –2.25, p , .05. Presumably the reason the 3.5-year-old
children did not show as many differences of communication between the phases as
3-year-olds is that they did not need to; their naive partners were more skillful than
those of the younger children, so that, as reported above, they were successful in
about half the time.

Discussion
The current study examined the skills of toddlers and young preschoolers in a coop-
erative problem-solving task that required the coordination of complementary roles.
Dyads of same-aged children were observed when both members were naive in the
task (cooperation), when they were each proficient in one role of the task but were
then presented with the opposite role (reversal), and when one individual was profi-
cient in both roles but then was paired with a naive partner (teaching). The study
found a clear developmental trend. (1) Two-year-old children did not learn to suc-
ceed in the task, either on their own or after adult instruction. (2) Two-and-a-half-
year-old and 3-year-old children learned to be successful in the task, but only slowly,
with most dyads requiring some adult assistance. Children of this age only infre-
quently made deliberate attempts to coordinate their behavior with their partner,
were relatively slow to learn the reversal task, and only infrequently used specific
directives in their communication with their partners. Children of both of these ages
did show some evidence of recognizing when a peer was new to the task in the teach-
ing phase, the main difference being that the 3-year-old children made more adjust-
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 159
ments for their naive partners during this phase, and so were more successful, than
the 2.5-year-olds. (3) Three-and-a-half-year-old children succeeded in the coopera-
tive task in approximately half the time of the other children, and succeeded in the
reversal task in approximately one-sixth of the time of the other children.
Throughout the experiment their problem-solving behaviors were characterized by a
relatively high number of coordinated attempts (monitoring the behavior of the
partner), and in their communication they used many specific directives and were the
only children to use explicit demonstrations.
The other main study of the cooperative problem-solving skills of children in this
age range found them able to cooperate successfully at a somewhat earlier age than
in the current study. Brownell and Carriger (1990) found that some 2.5-year-olds
cooperated in some fairly complex ways in their task. The reason for this difference
is that in Brownell and Carriger’s task, only one child actively manipulated the
apparatus while the other child simply watched for the reward to become accessible
so she could retrieve it. Thus one child learned to manipulate the apparatus, and the
other learned to anticipate the arrival of the reward. In the current study, both chil-
dren had to learn to operate one role of the apparatus, and they had to do so in a
temporally coordinated manner—arguably a more difficult task. Brownell and
Carriger also had no assessment of children’s comprehension of the ‘other’ role, and
so it is unclear the extent to which the children at this age understood how the two
roles related to one another. In our more demanding task, children had to master
complementary cooperation in all its aspects, and this proved to be a more demand-
ing task.
No previous studies have examined the ability of children this young to teach
other children. In this regard, both 2.5- and 3-year-olds seemed to make at least
some adjustments in their behavior for naive partners (after they had interacted
with knowledgeable partners), for example, engaging in more coordinated attempts
with the naive partners. But this increase in behavioral coordination is most likely
due to the fact that by the time of the teaching phase the children had already mas-
tered the task in both its roles (i.e., the teaching phase followed both the coopera-
tion and reversal phases), and so the children by this time knew for certain that the
role of the other was crucial to success and so they monitored the other’s behav-
ior—whereas they knew this less well in the cooperation phase. In addition, how-
ever, the 3-year-olds (but not the 2.5-year-olds) made significant adjustments in the
language they used with naive partners. They gave more specific directives and
attention directives and responded to their partner’s utterances less often (since
they were clearly leading the interactions). This use of more specific language, in
concert with the behavioral changes, would seem to provide strong evidence that
the 3-year-olds understood something about their naive partner’s need for assis-
tance in the task. In the case of the 2.5-year-olds, the behavioral changes alone,
without linguistic changes, makes it unclear what they understood of their new
partner’s knowledge and skills in the task. Interestingly, the 3.5-year-old children
changed their behavior from the cooperation to the teaching phases (more coordi-
nated attempts), but they did not show marked changes in the nature of their com-
municative attempts (although they doubled their proportion of specific directives,
this was not a statistically significant difference). But one very likely reason is that
their naive partners needed less help than the naive 3-year-olds (recall that 3.5-
year-old primary subjects, who were initially naive, were successful in less than half
the time of 3-year-old primary subjects). The plausibility of this interpretation is
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
160 Jennifer Ashley and Michael Tomasello
evidenced by the fact that in the Teaching phase the 3.5-year-old dyads were again
successful twice as quickly as 3-year-olds.
We believe that the most plausible explanation for these findings is in terms of
the social cognitive skills possessed by children during this age range. The 2.5 and
3-year-old children are beginning to show some signs of appreciating that their
partner has a different role or perspective in the task. This would seem consistent
with the general age range found by Flavell, Everett, Croft, and Flavell (1981) for
Level 1 perspective taking, that is, knowing that another person has a perspective
that differs from one’s own. In addition, the linguistic adjustments that the 3-year-
olds made for their naive partners also suggests the possibility that they may have
demonstrated some Level 2 perspective taking skills, that is, knowing something of
the content of the other person’s perspective. Obviously, the 3.5-year-old children
demonstrated Level 2 perspective taking skills even more clearly, as they were
much better on the reversal task and had overall higher levels of specific directives
for their partners during all phases of the task. They clearly understood that their
partner had a different perspective on the situation than they did, and they had
some ideas about specifically what that perspective was and how it might be
changed.
These findings for the 3.5-year-olds are also consistent with other lines of research
on young children’s ‘theories of mind’, or the understanding that other people are
mental agents with thoughts and beliefs of their own. While children typically do not
pass explicit false belief tasks until four years of age (or more), many children show
clear attempts to deceive others—to create a false belief in another person, in one
interpretation—in the year preceding their fourth birthdays (Chandler, Fritz, &
Hala, 1989). Moreover, observing children in their natural family interactions, Dunn
(1988) found many examples in which 3-year-old children seemed to understand
their sibling’s mental perspective on significant events. Similarly, in a more system-
atic study of children’s talk about the mental states of others, Bartsch and Wellman
(1995) found that it was at around 3.5 years of age children that began to use dia-
logue in which the thoughts and beliefs of others were explicitly contrasted with their
own. The behavioral and linguistic findings of the current study—and in particular
the clearly more skillful performance of the 3.5-year-olds—are thus consistent with
these more naturalistic findings in the social cognition literature that somewhere
between the ages of 3 and 4 young children begin to understand that other people
have their own unique mental perspectives on events and activities to which, in some
situations, they must relate their own mental perspectives (Tomasello, Kruger, &
Ratner, 1993).
The alternative hypothesis to explain the current results is that children are devel-
oping during this age both their cognitive skills to understand the task (e.g., keeping
several things in mind at once) and their skills with conventional language—not
skills of social cognition specifically. But with regard to problem-solving behaviors,
it is notable that even after the younger children were taught the task, they did not
do well in reversal, suggesting the possibility that even while coming to understand
the task in some ways, children of this age still do not understand the role of the
other—a basically social understanding. In terms of language, it should be noted
that the two most sophisticated types of communicative attempts in the current
study (demonstrations and specific directives) were not beyond the capabilities of the
younger children and so language development was not the key factor in their late
emergence. Demonstrations are not linguistic at all, and specific directives do not
© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998
Cooperative Problem-Solving and Teaching 161
require linguistic skills beyond those typically available to 2.5- and 3-year-olds (e.g.,
a specific directive such as ‘Push it up’ would be quite effective and would not tax
the language skills of any normally-developing 2.5-year-old). The significant correla-
tion between children’s most complex teaching activities with their latency to success
in the teaching phase is also compatible with the hypothesis that there is a common
factor—increasing social understanding—that underlies children’s increasing abili-
ties to cooperate, communicate, and teach in this problem-solving task.
It is unfortunate that it was not possible to control the gender composition of the
dyads in the different age groups in the current study. Gender differences in interper-
sonal styles have been observed in children as young as 3 years of age. For example,
Sheldon’s (1990) analysis of conflict talk identified boys as employing more control-
ling discourse strategies, while girls’ strategies tended to be more collaborative (see
also Jacklin & Maccoby, 1978). Nevertheless, in the current study examination of
the dyads of different types (mixed, female only, male only) showed no systematic
trends. Thus, the three most successful dyads of the 3.5-year-olds were two mixed
dyads and one female dyad; the two most successful 3-year-old dyads were one
mixed dyad and one male dyad; and the two most successful 2.5-year-olds were one
mixed dyad and one female dyad. The question of gender is an important issue for
future research.
In conclusion, the current study was able to establish a clear developmental trajec-
tory for the collaborative problem-solving skills that toddlers and preschoolers
employ when faced with a relatively demanding cooperation task, and a similar tra-
jectory for the skills they employ to teach less knowledgeable peers about this task.
The fact that there was a general developmental synchrony between the ontogeny of
these abilities and the ontogeny of some important social cognitive skills is sugges-
tive of a link between these two abilities, and clearly a question worthy of future
research. In general, we believe that further research into the cooperative problem-
solving and peer teaching skills of toddlers and young preschoolers—which would
seem to be among the most cognitively demanding types of social interaction in
which they engage—will yield great insights into young children’s skills of social
cognition.

References
Azmitia, M. (1988). When are two heads better than one? Child Development, 59, 87–96.
Azmitia, M., & Hesser, J. (1993). Why siblings are important agents of cognitive develop-
ment: A comparison of siblings and peers. Child Development, 64, 430–444.
Bakeman, R., & Gottman, J. M. (1986). Observing Interaction: An Introduction to Sequential
Analysis. New York: Cambridge University Press.
Barron, A-M., & Foot, H. (1991). Peer tutoring and tutor training. Educational Research, 33
(3), 174–185.
Bartsch, K., & Welman, H. M. (1995). Children Talk About the Mind. New York: Oxford
University Press.
Brownell, C. A., & Carriger, M. S. (1990). Changes in cooperation and self-other differentia-
tion during the second year. Child Development, 61, 1164–1174.
Brownell, C. A., & Carriger, M. S. (1991). Collaborations among toddler peers: Individual
contributions to social contexts. In L.B. Resnick, J.M. Levine, & S.D. Teasley (Eds.),
Perspectives on Socially Shared Cognition (pp. 365–383). Washington DC: American
Psychological Association.
Chandler, M., Fritz, A. S., & Hala, S. (1989). Small-scale deceit: Deception as a marker of 2-,
3-, and 4-year-olds’ theories of mind. Child Development, 60, 1263–1277.

© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998


162 Jennifer Ashley and Michael Tomasello

Cooper, C. R. (1980). Development of collaborative problem solving among preschool chil-


dren. Developmental Psychology, 16 (5), 433–440.
Cooper, C. R., Ayers-Lopez, S., & Marquis, A. (1982). Children’s discourse during peer
learning in experimental and naturalistic situations. Discourse Processes, 5, 177–191.
Damon, W. (1977). The Social World of the Child. San Francisco: Josey-Bass.
Damon, W. (1983). Social and Personality Development. New York: W.W. Norton &
Company.
Doise, W., & Mugny, G. (1979). Individual and collective conflicts of centrations in cognitive
development. European Journal of Social Psychology, 9, 105–108.
Dunn, J. (1988). The Beginnings of Social Understanding. Cambridge, MA: Harvard
University Press.
Duran, R. T., & Gauvain, M. (1993). The role of age versus experience in peer collaboration
during joint planning. Journal of Experimental Child Psychology, 55, 227–242.
Eckerman, C. O., Davis, C. C., & Didow, S. M. (1989). Toddlers’ emerging ways of achieving
social coordinations with a peer. Child Development, 60, 440–453.
Ellis, S., & Rogoff, B. (1982). The strategies and efficacy of child vs. adult teachers. Child
Development, 53, 730–735.
Emler, N., & Valiant, G. (1982). Social interaction and cognitive conflict in the development
of spatial coordination skills. British Journal of Psychology, 73, 295–303.
Flavell, J., Everett, B., Croft, K., & Flavell, E. (1981). Young children’s knowledge about
visual perception: Further evidence for the Level 1–Level 2 distinction. Developmental
Psychology, 17, 99–103.
Gauvain, M., & Rogoff, B. (1989). Collaborative problem solving and children’s planning
skills. Developmental Psychology, 29, 139–151.
Jacklin, C., & Maccoby, E. (1978). Social behavior at 33 months in same-sex and mixed-sex
dyads. Child Development, 49, 557–569.
Koester, L. S., & Bueche, N. A. (1980). Preschoolers as teachers: Where children are seen but
not heard. Child Study Journal, 10 (2), 107–118.
Kruger, A. (1992). The effect of peer and adult–child transactive discussions on moral reason-
ing. Merrill-Palmer Quarterly, 38, 191–211.
Phelps, E., & Damon, W. (1989). Problem solving with equals: Peer collaboration as a context
for learning mathematics and spatial concepts. Journal of Educational Psychology, 81,
639–646.
Piaget, J. (1932). The Moral Judgement of the Child. London: Routledge & Keegan Paul.
Pratt, M. W., Scribner, S., & Cole, M. (1977). Children as teachers: developmental studies of
communication. Child Development, 48, 1475–1481.
Radziszewska, B., & Rogoff, B. (1988). Influence of adult and peer collaborators on children’s
planning skills. Developmental Psychology, 24, 840–848.
Rogoff, B. (1990). Apprenticeship in Thinking. New York: Oxford University Press.
Sheldon, A. (1990). Pickle fights: Gendered talk in preschool disputes. Discourse Processes,
13, 5–31.
Tomasello, M., Kruger, A. C., & Ratner, H. H. (1993). Cultural learning. The Behavioral and
Brain Sciences, 16, 495–552.
Tudge, J. R. H., & Rogoff, B. (1989). Peer influences on cognitive development: Piagetian and
Vygotskian perspectives. In M.H. Bornstein & J.S. Bruner (Eds.), Interaction in Human
Development (pp. 17–40). New Jersey: Erlbaum.
Tudge, J., & Winterhoff, P. (1993). Can young children benefit from collaborative problem
solving? Tracing the effects of partner competence and feedback. Social Development, 3 (3),
242–259.

Notes
1. This procedure was deemed necessary as children who were unsuccessful for many trials
in a row began to become either uncooperative or upset.
2. This procedure was chosen because the amount of time spent in one of these categories
was often misleading. Thus, spending 20 seconds for one coordinated attempt was not twice
as skillful as spending 10 seconds in a coordinated attempt.

© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998


Cooperative Problem-Solving and Teaching 163
3. For analysis of the problem-solving behavior data, mean proportions were arcsin trans-
formed prior to statistical analysis, as is conventional for the statistical treatment of propor-
tional data (Cohen & Cohen, 1983). All inferential statistics reported used these arcsin
transformations. However, the raw proportion data were also analyzed using MANOVAs
and Univariate analyses prior to the arcsin transformations. The results of these analyses did
not differ from those found after the data were transformed.
4. Parallel analyses using the rate at which children produced these behaviors (i.e., instead
of dividing by the subject’s total number of behaviors, we divided by time in seconds) pro-
duced an identical pattern of results.

© Blackwell Publishers Ltd. 1998 Social Development, 7, 2, 1998

You might also like