Cognitive Science 39 (2015) 804–820 Copyright © 2014 Cognitive Science Society, Inc. All rights reserved. ISSN: 0364-0213 print / 1551-6709 online DOI: 10.1111/cogs.12172

How Do Groups Work? Age Differences in Performance and the Social Outcomes of Peer Collaboration Patrick J. Leman Department of Psychology, Royal Holloway, University of London Received 13 December 2012; received in revised form 26 January 2014; accepted 4 February 2014

Abstract Do children derive different benefits from group collaboration at different ages? In the present study, 183 children from two age groups (8.8 and 13.4 years) took part in a class quiz as members of a group, or individually. In some groups, cohesiveness was made salient by awarding prizes to the top performing groups. In other groups, prizes were awarded to the best performing individuals. Findings, both in terms of social outcomes and performance in the quiz, indicated that the 8-year olds viewed the benefits of group membership in terms of the opportunities to receive information from other members. The 13-year olds, in contrast, viewed group collaboration as a constructive process where success was connected with group cohesiveness. Keywords: Collaboration; Group; Interaction; Peer; Sociometric status

1. Introduction Peer collaboration and group work are common features of learning at school, and the potential benefits of peer collaboration are well-established (e.g., Doise & Mugny, 1984; Leman & Bj€ ornberg, 2010). Researchers have also identified benefits in the efficiency of short-term collaborative problem-solving compared with independent problem-solving (e.g., Azmitia, 1988). Cognitive or learning advances through collaboration run alongside social benefits of collaboration for children in terms of social relationships (Tolmie et al., 2010). These positive social experiences appear to occur even when collaboration involves conflict or disagreement (Howe & McWilliam, 2006). However, how children think and how they engage in interactions with peers changes markedly throughout childhood and adolescence. And although many studies highlight the benefits of collaboration, few studies have explored age differences in the processes that underpin collaboration. Correspondence should be sent to Patrick Leman, Department of Psychology, Royal Holloway, University of London, Egham, Surrey, UK TW20 0EX. E-mail: [email protected]

P. J. Leman / Cognitive Science 39 (2015)

805

Thus, the possibility remains that although collaboration works, it works for different reasons at different ages. Most theorists agree that the effectiveness of collaboration is due to a combination of social and cognitive processes (e.g., Howe, 2009). However, collaboration is not always an effective means of learning. For instance, educators may place peers in groups where members lack the relevant social skills and competences to successfully interact or the dynamics of interaction may generate conflicts that prohibit constructive discussion (Damon & Killen, 1982; Psaltis, 2011). Slavin (1996) suggested that a child may benefit from group interaction and collaboration because more group members provide more sources of information and resources to help solve a problem. However, alongside the benefits of the group as a potential source of information, the group may also function as a means of evaluating information by discussion, debate, and incorporating multiple perspectives. Additionally, the group may facilitate the construction of new knowledge (Psaltis, Duveen, & Perret-Clermont, 2009). This second aspect of group work requires that members of a group work together, understand social roles, devise a means for deploying resources effectively, and share goals. If effective collaboration involves both social and cognitive processes, one might expect that as children develop in these domains then their ability to collaborate likewise develops, or their orientation to collaboration changes. However, although teachers often differentiate in terms of group size, task demands and group composition for children at different ages (e.g., Baines, Blatchford, & Kutnick, 2003), surprisingly little research has sought to explore how age relates to the social and cognitive aspects of children’s communication in a collaborative context. Thus, teachers may place children in groups without fully considering whether the tasks or group dynamics provide conditions for productive learning. Understanding the relation between group dynamics and collaboration, particularly in terms of how the relation may differ by age, is the central aim of the present study. Many studies find that even from early childhood, children benefit in some way from collaboration (e.g., Tudge, 1992). However, research that looks more closely at how age relates to collaborative processes suggests that processes of collaboration are affected by age. For instance, Duran and Gauvain (1993) explored collaboration in a task that involved deciding on the best sequence for delivering various items to a farm. Five-yearold children were placed in a pair either with another 5-year-old expert at the task, or with a 7-year-old expert. Duran and Gauvain found that novice children in a pair with an expert child showed significant gains in a same-day posttest. Surprisingly, 5-year-olds performed better when paired with another 5-year-old than when paired with a 7-yearold. The researchers speculated that social comparison processes may have added to a sense of distance between children of different ages. This suggestion implies, on one hand, that same-aged peers may have been more sensitive to a partner’s learning needs and style. On the other hand, more conflict in same-age peer discussion suggested that older children tended to dominate younger children (or encountered less resistance from them) in mixed-age interaction. The resulting asymmetry in interaction may therefore have limited the extent to which younger children participated in the task and hence

806

P. J. Leman / Cognitive Science 39 (2015)

derived fewer benefits from it. Adult domination of “air time” in interaction and decision-making has also been associated with poorer developmental outcomes (vis-a-vis peer interaction) in moral reasoning tasks (e.g., Kruger & Tomasello, 1986). Age may affect collaboration because of expectations (or developmental level) of same-age versus older (expert) partners. While it is worth remembering that some research has shown that under the right conditions, peer tutoring also has clear benefits in terms of learning and problem solving (e.g., Topping, Peter, Stephen, & Whale, 2004), other research suggests that even though both young and older children show benefits from peer collaboration, the processes through which these benefits accrue are not the same (e.g., Gummerum, Leman, & Hollins, 2013; Leman & Oldham, 2005). Leman and Oldham (2005) used a collaborative recall paradigm to explore interaction processes in children age 7 and 9 years. Unlike many forms of collaboration which entail benefits to joint work, adults show a net negative effect of joint, verbal recall of shared information. That is, when asked to recall items from a list in a pair or group with others, adults perform worse compared with the equivalent “pooled” independent recall. The negative effects are a consequence of mutual inhibition of retrieval strategies (Basden & Basden, 1995), which leads to less efficient recall than comparable independent, uninhibited recall. Leman and Oldham found that pairs of 9-year-olds showed adult-type net negative effects of collaboration, whereas pairs of 7-year-olds showed no negative effects. Mixed-age pairs (i.e., 7- and 9-year-olds together) showed only mild effects. The researchers concluded that the older children understood recall as a joint and coordinated activity, whereas the younger children conceived of recall as an individualized activity that required little coordination. In other words, younger children did not grasp the sense in which collaboration involved understanding and coordinating different perspectives. Age-related differences in orientation to collaboration may also be related to development in epistemic understanding. In his work on moral judgment, Piaget (1932) proposed that development proceeded from realist to subjectivist representations of knowledge with the latter becoming more prevalent from around 8–10 years. Realism in thinking corresponds to a sense of knowledge as something that the child acquires from (or is transmitted to the child by) adults and older children, whereas subjectivism corresponds to a view of knowledge as the product of socialized thinking that is constructed through peer interaction (or from within relations of mutual respect). More recently Kuhn, Cheney, and Weinstock (2000) suggested development in epistemological understanding from objective to subjective to mixed (intersubjective) orientations to knowing. These developments in epistemic understanding may feed in to affect collaborative processes. For instance, Leman and Duveen (1996) noted that when 6-year olds’ opinions differ, their conversations with peers often focus on “winning out” with little scope for compromise or negotiation. In contrast, by 11 years of age, children can engage in conversations to find ways to resolve or understand differences and to get a correct answer that all can accept. Development in orientation to knowledge throughout childhood is accompanied by changes in children’s understanding of social groups and social group dynamics (Abrams, Marques, Randsley de Moura, Hutchison, & Bown, 2004) and their own social identity (Serbin, Powlishta, & Gulko, 1993). Early understanding of groups is built around basic

P. J. Leman / Cognitive Science 39 (2015)

807

intergroup (ingroup-outgroup) distinctions, but from 8 years, children begin to understand more about intragroup dynamics and in particular that groups work better when they are more internally cohesive (Abrams, Rutland, Cameron, & Ferrell, 2007). However, it remains unclear how far children’s knowledge about groups influences the strategic aspects of group collaboration. Whether and how children’s understanding of group processes influences collaboration is an important question from a theoretical perspective because knowing more about the intersection of group dynamics and collaboration at different ages can inform work on collaboration and socio-cognitive development. It is also an important question from an applied perspective, because information about what makes groups collaborate effectively can inform classroom practices. The present study explores how children and adolescents, at different ages (8 and 13 years), respond to different forms of group in collaborative decision-making. Of interest is how different forms of group affect the outcomes of interaction in terms of task performance and in terms of social relations between children. Participants were placed in groups and were given a general knowledge quiz. Many previous studies of peer collaboration involve the construction of new knowledge or understanding by children (e.g., Doise & Mugny, 1984). The general knowledge quiz used in the present study does not involve the construction of new knowledge, although the task does involve associated skills such as collective recall, decision-making, discussion and evaluation. The quiz format was chosen because it is familiar to children at all ages and includes a broad range of areas that allows children to contribute to discussions across diverse areas of knowledge. Moreover, a key focus in present study was on how group cohesiveness would affect performance and collaboration at different ages. In addition, differences in cognitive capacity and understanding make it difficult to compare age groups reliably on developmental tasks involving the construction of knowledge. Hence, while the construction or acquisition of new knowledge was not a feature of the present task, better understanding of how group cohesiveness influences judgments and interactions can provide important information about how children at different ages approach peer collaboration. The reward structure and cohesiveness of groups was varied. In a Group Cohesion condition (henceforth, GC), cohesiveness between group members was promoted by rewarding all members of groups with the highest scores. In a Group Information condition (henceforth, GI), participants worked as a team but prizes were awarded to the topscoring individuals. In a control condition participants answered questions individually and the top-scoring individuals were awarded prizes. Changes in social relations were explored in terms of pre- to postinteraction changes in sociometric ratings (whether children would like to “work with” and “play with” other children in their group). For comparison purposes, nominal groups were created for children in the control condition by placing sociometric scores from control condition participants into “virtual” groups after the study had been completed. Also of interest were the features of intersubjectivity that children deployed in conversations, and how these might vary by age and by the different conditions under which children completed the quiz. In light of previous research several hypotheses were proposed. First, and consistent with research showing benefits of collaboration, it was hypothesized that at all ages

808

P. J. Leman / Cognitive Science 39 (2015)

participants in groups (GC and GI) would perform better (get higher scores) than the participants in the control condition. Second, it was expected that younger children (8 years) would be more oriented to the group as a source of information, whereas the older participants (13 years) would benefit from the cohesive group. Specifically, the second hypothesis was that there would be an age 9 condition interaction with 8-yearolds performing better in GI than in GC, and 13-year-olds performing better in GC than in GI. The third and fourth hypotheses related to the social changes (changes in sociometric ratings) after group interaction. For the third hypothesis, it was predicted that all participants would give higher “play with” ratings for peers in their group than a control condition after interaction. This was anticipated because the social benefits of contact will be evident in ratings about social activities such as play but will not feed into strategic decisions about future work. The fourth hypothesis predicted another age x condition interaction: Specifically, after interaction 8-year-olds would give more favorable ratings to peers in GI than in GC, whereas 13-year-olds would give more favorable ratings in GC compared with GI. Finally, a related fifth hypothesis concerned a question about whether children would like to change their group after interaction. Again, an age x condition interaction was predicted that at 8 years, children would be more likely to want to change groups if they had been in GC than in GI, but at 13 years, participants in GI would wish to change more than those in GC.

2. Method 2.1. Participants One hundred and eighty-three children participated in the study. Participants came from schools in suburban areas of South East England, United Kingdom. The social and ethnic mix of participants reflected that of the schools from which children came, which were in middle-class areas with majority White European pupils. A small minority (< 10%) of participants came from ethnic minority groups, including South Asian and African-Caribbean. Participants were in two age groups: an older group aged 12–13 years (Mage 13.4 years, N = 116) and a younger group aged 7–8 years (Mage 8.1 years, N = 67). 2.2. Materials Materials were sociometric rating scales (“work with” and “play with” ratings) for all of a child’s classmates, and the multiple-choice quiz and answer sheet that children completed individually or in groups, depending on the experimental condition. After the quiz children were also asked whether they would like to change groups if they took part in a similar quiz in the future.

P. J. Leman / Cognitive Science 39 (2015)

809

The sociometric rating forms measured the participants’ evaluations of their peers 1 week prior to the quiz, and immediately after hearing the results of the quiz and the award of prizes to successful participants. The sociometric scales included instructions asking participants to rate all members of the class on 5-point Likert-type scales (Merrell, 2003) ranging from “Not at all (I would not like to work/play with this person)” to “Very much (I would like to work/play with this person).” Children indicated preferences by checking one of five faces, which varied from unhappy to happy expressions according to the scale. The general knowledge quiz consisted of 20 multiple-choice, general knowledge questions that were appropriate to each age group. An example question is: “How many bones are there in the human body? (a) 35, (b) 120, (c) 206, (d) 1,013.” Participants completed the quiz by indicating answers on the sheets. The children were given 15 min to complete the quiz. Depending on the condition, half of the participants with the highest scores were awarded a prize. Prizes were chosen that would be desirable to children in the different age groups. For the younger age group the prizes were various sets of stickers from which winners could make a selection. For the older age group, prizes were glow-in-thedark pens. 2.3. Procedure One week prior to the quiz, participants were provided with sociometric rating forms and were asked to rate all members of their class on separate work-with and play-with scales. Subsequently, children were randomly allocated into experimental conditions for the second phase of the study: Group Cohesion (GC, N = 60), Group Information (GI, N = 72), and control condition (N = 51). At this second testing session, participants were told that they were about to take part in a general knowledge quiz, that the quiz would last 15 min, and that there were 20 questions. Children were then given an example question to ensure they understood the multiple-choice quiz format. A different set-up and different further instructions were given for each experimental condition. In both GC and GI conditions, children were placed, at random, in groups of four and seated around a table together. In the control condition, children completed the quiz on their own in their regular, individual classroom seats. For the GC condition, participants were told: “You notice you are sitting in a team with other people. You can work through the quiz in your teams. You can ask people in your team what they think the correct answer is. You can also tell others in your team what you think the correct answer is. At the end, each member of the teams that have the most correct answers will win a prize. Although you will work as a team, it is important that when you think you have the correct answer, tick the correct box on your individual answer sheet.” The GC groups were then given further instructions to stress the possibility for collaborating to produce the best group score; “It is important to note that if you are unsure of the correct answer in your groups, it is still possible for your group to answer the question correctly. For example, if you may be asked: “What are Heathrow

810

P. J. Leman / Cognitive Science 39 (2015)

and Gatwick?: (a) Airports; (b) Football Teams; (c) TV Shows; and (d) Cities.” If your group decides that the answer is definitely not football teams or TV programmes, then some people in you group can answer with (a) Airports, and some people in your group can answer (d) Cities. This means that one of the answers is going to be correct, and therefore your team will definitely get two points rather than none at all. So it is worth making a guess even if you’re not sure of all the answers. Remember, you are working in groups, not individually.” The experimenter checked carefully that all children understood the instructions before proceeding by asking, “Do you understand the instructions for your group? Please ask any further questions if you are unsure.” If participants responded that they did not understand, or appeared uncertain, the instructions were repeated until all participants were clear about the basis of the group and the basis on which prizes were given. In the GI condition, participants were given the following instructions; “You notice you are sitting with other people. You can ask the others on your table what they think the correct answer is. You can also tell others what you think the correct answer is. At the end, there will be a prize for the children with the most correct answers.” Again, the experimenter ensured that participants fully understood the instructions. In the control group, participants were told to work individually on the quiz, and to not ask others for help. The control group was also informed that there would be a prize for the children with the most correct answers. There were 13 groups of four 8-year-olds in the GC condition and 12 groups in the GI condition. There were six groups of 13-year-olds in the GC condition and seven in the GI condition. Following the quiz, participants were supplied the sociometric tests (play-with and work-with ratings for all class members) for the second time. After this, participants responded to a question asking whether they would like to change groups if they were to take a similar quiz again (the response was on a five-point scale from 1 “not at all” to 5 “very much”). Finally, the scores were computed and the top teams or individuals were awarded their prizes and all participants were fully debriefed regarding the aims of the study and were given the opportunity to ask further questions. 2.4. Intersubjectivity in conversations In order to explore the dynamics of interactions more closely, recordings of groups’ interactions were coded in terms of different aspects of intersubjectivity (following G€ onc€ u, 1993; Gummerum et al., 2013). While an appreciation of inter-subjectivity is a skill acquired in childhood, the focus of interest here was how participants at different ages deployed these skills in interactions. The coding scheme used identifies three elements of intersubjectivity: joint focus, meta-communication, and communication. Joint focus involves group members sharing gaze or an affective state (e.g., smiling at one another), or following or replicating the actions of another group member. Meta-communication refers to communications among group members that regulate collaborative activities. For instance, a group member might initiate a conversation topic (e.g., “Let’s come back to that question later”) or propose a strategy for answering (e.g., “Let’s all

P. J. Leman / Cognitive Science 39 (2015)

811

write down our own and then compare?”). Finally, communication is any contribution that is linked to a fellow group member’s previous utterance. This includes repetition or utterances that confirm or support a statement. Individual contributions to group interactions (i.e., GC and GI) were coded by two independent judges. One judge coded 19 of the conversations, with 5 conversations coded by both judges. An event coding technique was used, such that the frequency with which each conversation element was used across the conversation was recorded. Interjudge reliability was calculated for the five conversations that both judges coded, and there was good agreement between judges, j = 0.78.

3. Results 3.1. Design The study employed a 2 9 3 mixed factorial design, with age (8 and 13 years) and experimental condition (GC, GI, control) as independent variables. Sociometric ratings for all members of a class (work-with and play-with ratings, respectively) were collected before and after the quiz. For each child, ratings for all members of the quiz group were used to calculate a change score (by subtracting pre- from posttest ratings on work-with and play-with scales). Thus, a positive value of the change score indicated that participants viewed others more favorably, whereas a negative score indicated that they view others more negatively after the quiz phase. To ensure statistical comparability, a nominal group was created for children in the control condition by placing groups of four children in virtual groups, thus exploring sociometric ratings changes for these children as members of virtual groups. Variations in quiz scores were analyzed first using ANOVA to explore effects of condition and age on performance, and any interaction between independent variables. Changes in sociometric ratings were examined using first MANOVA, with separate follow up ANOVAs for changes in play-with and work-with ratings in terms of effects of age and condition where appropriate. 3.2. Quiz scores In order to establish whether the group structure affected performance at different ages, a 2 9 3 ANOVA explored the effect of age and condition on quiz scores. There was a main effect of condition, F(2, 198) = 12.62, p < .001, gq2 = .113, and of age group, F(1, 198) = 137.08, p < .001, gq2 = .409, on quiz scores. There was also an interaction between condition and age, F(2, 198) =16.37, p < .001, gq2 = .142. Post hoc analysis was conducted to examine condition effects in each age group in turn. The one-way ANOVA exploring effects of condition on quiz scores in the younger age group (8 years) was significant using the Bonferroni corrected value of p < .017 as critical, F(2, 65) = 33.61, p < .001, gq2 = .508. Tukey HSD (p < .05) indicated that all three conditions differed

812

P. J. Leman / Cognitive Science 39 (2015)

significantly from one another (i.e., quiz scores in GI were significantly higher than in GC and control, and scores in GC were significantly higher than in the control condition). In the older age group (13 years), the one-way ANOVA of condition on quiz scores was also significant, F(2, 133) = 15.51, p < .001, gq2 = .189. Tukey HSD (p < .05) tests indicated that scores in GC differed significantly from GI and control. Table 1 gives mean quiz scores by age group and condition. 3.3. Intersubjectivity in conversations An analysis of the use of intersubjectivity in conversations compared different conversation measures at each age in the different, real groups (GC and GI). A MANOVA examined the relations among age group, condition, success, and the three measures of intersubjectivity in conversation: joint focus, meta-communication, and communication. Success in the quiz did not relate to any significant effects or interactions. However, there was a main effect of age group, Wilks’ k = .724, F(3, 139) = 17.67, p < .001, gq2 = .276, and of condition, k = .895, F(3, 139) = 5.45, p < .01, gq2 = .105. There was also an interaction between age and condition, k = .934, F(3, 139) = 3.29, p < .05, gq2 = .066. Table 2 gives means and standard deviations for all measures of intersubjectivity by age and condition. Table 1 Mean (SD) score in the quiz by age group and condition Condition Group Cohesion (GC), N = 60

Group Information (GI), N = 72

Control, N = 51

8 years, N = 116 12.65a (1.53) 15.71b (2.09) 9.71c (3.62) 13 years, N = 67 22.94d (4.10) 18.32e (5.10) 18.33e (4.78) Total, N = 184 19.70f (5.94) 17.35g (4.40) 15.71h (5.97) Note. Means with different subscripts in a row differ significantly from one another (p < .05).

Table 2 Mean (SD) measure of intersubjectivity in conversation—joint focus, meta-communication, and communication—by age group and condition (GC and GI) Conversation Measure

8 years, N = 96 Group cohesion (GC) Group information (GI) Total 13 years, N = 52 Group cohesion (GC) Group information (GI) Total

Joint Focus

Meta-Communication

Communication

4.39 (5.40) 7.48 (5.83) 6.83 (5.61)

1.70 (1.89) 2.10 (2.24) 1.92 (2.09)

13.96 (6.12) 10.79 (5.61) 12.19 (6.00)

11.04 (4.86) 12.47 (6.45) 10.16 (6.23)

7.76 (3.95) 3.70 (3.05) 5.86 (3.84)

15.16 (6.77) 14.68 (6.15) 14.93 (6.45)

P. J. Leman / Cognitive Science 39 (2015)

813

In order to explore the interaction, separate univariate analyses were conducted for each conversation measure. The older participants used more of each conversational measure than the younger participants: joint focus, F(1, 141) = 10.34, p < .001, gq2 = .155; meta-communication, F(1, 141) = 4.22, p < .001, gq2 = .192; communication, F(1, 141) = 6.97, p < .01, gq2 = .047. Additionally, participants engaged in more joint focus in the GI than the GC condition, F(1, 141) = 4.22, p < .01, gq2 = .068. In contrast, participants in GC employed more meta-communication than those in GI, F(1, 141) = 4.21, p < .05, gq2 = .029. However, this latter effect was qualified by an age x condition interaction. Specifically, post hoc independent t-tests indicated that there was no difference in the use of meta-communication comparing 8-year-old children in the GC and GI groups. However, among 13-year olds, those in GC used more meta-communication than those in GI, t(95) = 4.10, p < .001. 3.4. Changes in sociometric ratings The next set of analyses compared participant’s pre- and posttest sociometric ratings of who they would like to work with and play with in a group. Changes in sociometric ratings indicate effects of group structure on participants’ judgments, relationships, and attitudes as a result of engaging in interaction with others. Again, control groups were constructed including ratings from those children who had not interacted yet were randomly placed into a nominal group after data collection had been completed. A MANOVA was carried out to examine the associations between age group, condition, success, and changes in sociometric peer group ratings (“work with” and “play with”). The MANOVA revealed a main effect of condition, Wilks’ k = .925, F(4, 340) = 3.46, p < .01, gq2 = .039, and an interaction between age and condition, Wilks’ k = .937, F(2, 171) = 5.24, p < .05, gq2 = .058. In order to examine variations in sociometric ratings further, separate ANOVAs were conducted on changes in “play with” and “work with” ratings to examine effects of age, condition, and success. 3.5. Play-with ratings A first ANOVA examined age group and condition associations with changes in play with ratings by age group and condition. There were no age effects or interactions. There was a main effect of condition, F(2, 171) = 3.64, p < .05, gq2 = .039, and post hoc tests (Tukey HSD, p < .05) indicated that in the control condition, children’s desire to play with a friend (from their “nominal” group) changed very little from pre- to posttest (M = 0.11), but children indicated a greater desire to play with a friend from the group after the quiz in GI (M = 0.92) and GC (M = 0.99) conditions. 3.6. Work-with ratings Variations by age group and condition in mean pre- to posttest changes in work-with ratings are displayed in Table 3. Across both age group conditions affected changes in

814

P. J. Leman / Cognitive Science 39 (2015)

Table 3 Mean (SD) pre- to posttest change in “work with” ratings of group members by age group and condition Condition Group Cohesion (GC), N = 60

Group Information (GI), N = 72

Control, N = 51

8 years, N = 116 .35a (2.16) 2.01b (3.27) 13 years, N = 67 1.55c (2.56) 1.21c,d (1.98) Total, N = 184 .82e,f (2.57) 1.51e (2.55) Note. Means with different subscripts in a row differ significantly from one another (p < values indicate higher (more favorable) ratings in the posttest.

.48a (3.58) .22d (1.47) .01f (2.38) .05). Positive

the work with ratings, F(2, 171) = 5.24, p < .05, gq2 = .580. However, this main effect was qualified by a condition x age group interaction, F(2, 171) = 5.62, p < .05, gq2 = .060. Post hoc analysis (Tukey HSD, criteria for significance set at p < .05) comparing conditions within age groups indicated that in the younger age group children in GI made more favorable work with ratings for members of their group after the quiz than children in either GC or control conditions. In contrast, in the older age group children in GC were significantly more positive in their work with ratings of peers in their group than the children in the control condition. Comparisons of age differences within conditions further revealed that in the GC condition 13-year-olds made more positive postquiz ratings, whereas the ratings of 8-year-olds changed little, t(58) = 2.96, p < .005. Changes in ratings in GI and control conditions did not differ by age. 3.7. Changing groups A final analysis examined whether there were variations by age group and condition (GC and GI only) in terms of participants’ ratings of whether they wanted to change groups or not. Children’s willingness to remain in or change groups indicates whether group interaction had been a positive or negative experience (see Leman et al., 2011). This analysis revealed an interaction between age and condition, F(1, 143) = 3.98, p < .05, gq2 = .015. The 8-year-olds in the GC condition (M = 2.55) and in the GI condition (M = 2.54) tended to select not to change groups (midpoint of the 5-point scale is 3). Among the 13-year-olds those in the GC condition also, on average, preferred not to change groups (M = 2.42) but were at the midpoint in the GI condition (M = 3.00). Follow-up t tests for each age group indicated that ratings only differed comparing the 13-year-old children in GI versus GC conditions, t(95) = 2.32, p < .05. 4. Discussion Learning and interacting in groups is a common feature of life inside and outside of school. The present study examined, for the first time, how group dynamics influence performance and social outcomes of collaboration for children and adolescents (aged 8 and 13 years). Three sets of hypotheses were proposed concerning children’s reactions to

P. J. Leman / Cognitive Science 39 (2015)

815

different sorts of group: high cohesiveness, groups as a source of information, and a control (independent response) condition. Predictions were made about performance on the quiz task, sociometric outcomes of participating in groups, and group-level decisions. The first two hypotheses predicted variations in children’s performance in the quiz. Specifically, consistent with research from a wealth of sources (e.g., Howe, 2009), it was predicted that at both ages children would perform better in groups (GC and GI) than in the control condition. This hypothesis was partly confirmed. The second, related hypothesis anticipated age differences in the relative effectiveness of the different group conditions. This hypothesis was also confirmed. As expected, at 8 years, children in the GI condition performed significantly better than children in the GC condition and at 13 years the reverse was true (i.e., children in GC performed better than in GI). However, it was noteworthy that among the older age group, performance in GI did not differ significantly from performance in the control condition. That the older age group performed equally (poorly) in GI and control conditions was unexpected. It may be that these older children mistrusted the information given by others or deliberately misinformed others because their responses counted only individually for a prize. The potential benefits for cohesiveness were “structured out” of the GI condition, and for older children the absence of an ability to use the group strategically had clear consequences. In contrast, it is striking that the 8-year-olds performed best in the GI condition where the lack of a strategic understanding might have been advantageous in terms of performance in GI versus GC. A key aim of the present study was to explore the social (sociometric) outcomes of group interaction. Previous research (Tolmie et al., 2010) has suggested that one benefit of the use of collaborative learning in a classroom is increased positive social relations within the class. Another good reason for investigating social outcomes is that understanding more about social changes will inform a further dimension in terms of how effective collaboration works. The third hypothesis predicted that, compared with the control condition, participants at both ages in GC and GI would show a positive change in ratings of whether they would like to play with members of their group after interacting with them. This hypothesis was confirmed. The benefits of group interaction in terms of children’s social relationships have been demonstrated in much previous research (e.g., Gottman, Gonso, & Rasmussen, 1976), and peer contact has been demonstrated to impact positively on social attitudes and friendship choices (Feddes, Noack, & Rutland, 2009). The fourth hypothesis predicted that changes in work with ratings would vary by age and condition. Specifically, it was expected that 8-year-olds would have more positive work with ratings for group members after interaction in GI than GC. The results indicated that at 8 years there was more positive work with nominations for members of a group in GI than GC and control, but nominations in GC and control did not differ from one another (in fact, both showed little movement from pre- to posttest). A corresponding prediction was that in the older age group (13 years), more favorable work with nominations would be made after group interaction in GC than GI. Somewhat consistent with this prediction, adolescents in the GC condition gave more positive ratings than the

816

P. J. Leman / Cognitive Science 39 (2015)

control group, but those in GI did not differ significantly from either GC or control conditions. Findings of sociometric changes underscore the thrust of results from the analysis of performance on the quiz. In this context, 8-year-olds’ responses to groups appear largely to relate to the information they can get from being a member of a group. Hence, 8-yearolds in the GI condition would choose to work with group members again, but those in the GC were indifferent to the prospect. Older (13-year-old) participants appear more sensitive to the social dynamics aspects of groups. The failure to find a significant difference between GC and GI in the older age group may be a consequence of greater understanding of strategic aspects of groups among older children and adolescents. Older children in GI may have worked around the constraints of the GI condition to establish some patterns of working or developed trust and working patterns that could function to everyone’s advantage another time (Gummerum, Keller, Takezawa, & Mata, 2008). Thus, older children in GI may have “made do” with the group structure they were given, and could see worse group functioning, but could also see how a more cohesive group might improve performance and the individual chance of reward. This interpretation of findings of age differences in the analysis of sociometric ratings is consistent with results relating to the fifth hypothesis that also predicted an age x condition interaction in terms of whether participants would want to change groups if they were to take part in a similar quiz. In almost all circumstances, participants showed loyalty to the group and gave lower than midpoint ratings (change vs. no change). However, at 13 years, children in the GI condition gave higher ratings (at the midpoint of the scale) indicating less loyalty to the group. The partial confirmation of this hypothesis again connects with explanations of age differences in the strategic understanding of groups: Older children see practical benefit in continuing with group members in GC but not GI, maybe because in GI they have not built up established patterns of responding and hence do not attain the pragmatic approach to group dynamics children at this age regard as desirable. Analysis of children’s interactions supports the view that the conditions influenced the conversational approaches of children at different ages. It is unsurprising that, while some empirical tests indicate that younger children appreciate aspects of intersubjectivity by 8 years (e.g., Piaget, 1937/1971), they do not deploy this understanding. That is, children do not express this understanding as frequently and perhaps not to the same effect in conversation as older children. What is significant is that the condition did not alter how younger children used different elements of intersubjectivity in interactions. However, older children in GC employed significantly more meta-communication—that is, attempts to regulate collaborative activity, than in GI. The analysis of conversations suggests that older children adapted their conversational behavior to pursue a collaborative (joint) goal. In addition, they appreciated the significance of meta-communicative processes to regulate group activity in pursuit of that goal. Younger children may have a poor grasp of these skills or lack an understanding about when and how to deploy them in collaborative interactions. It is surprising that there was significantly more joint focus coded in interactions among participants in the GI condition compared with those in the GC condition. The

P. J. Leman / Cognitive Science 39 (2015)

817

aspects of interactions that were classified as “joint focus” (e.g., sharing gaze, imitating or replicating others’ actions, or smiling at one another) might be regarded as commonsense indicators of a cohesiveness. This finding may be a consequence of a relative lack of sophistication in the coding scheme or a consequence of interactions in this particular task. However, it might also be the case that these elements of interaction are employed by children at both ages to ensure ongoing group interactions even though a shared objective or goal is absent (or less a feature of the task). In this respect, participants in the GI condition may have employed these conversational indicators of joint focus as a means of compensating for the lack of shared interests or goals. Clearly, more detailed and systematic research into the nuanced nature of children’s collaborative interactions, and perhaps more sophisticated schemes for analyzing features of those interactions, is needed to untangle these different explanations for a seemingly anomalous finding. Taken together, the present findings demonstrate clear age differences in orientation to group collaboration. At 8 years, children benefit from groups acting primarily as sources of information, but at 13 years, group cohesiveness reaps identifiable benefits in terms of performance. In other words, children’s orientation to collaboration shifts from being primarily concerned with information to involving more coordinated forms of interaction and intersubjective exchange (Leman & Duveen, 1996; Leman & Oldham, 2005). The present results also fit well with research that suggests that from middle childhood through to adolescence there are significant changes in children’s group knowledge and understanding, particularly in terms of ingroup dynamics (Abrams et al., 2004). This rapidly emerging knowledge of group dynamics married with developmental advances in discourse processes and intersubjectivity (Baines & Howe, 2010) means that collaboration works for different reasons at these different ages. Between middle childhood and the start of adolescence, increasing knowledge of functioning within groups feeds through to make group cohesiveness more important in peer collaboration. A greater understanding of ingroup dynamics may allow older children and adolescents to make better strategic and planning decisions within groups or to deploy group resources more effectively to solve problems. For instance, older participants may have answered strategically as a group because they made decisions about who possessed expertise in different domains, or “sounded out” answers among one another. Younger children appeared not to respond to or benefit from cohesiveness either because they were not engaged or, more likely, because their strategic understanding of group dynamics was minimal. This, and other research (Gummerum, Hanoch, & Keller, 2008; G€uroglu, van den Bos, & Crone, 2009) points to important age-related changes in children’s strategic understanding in groups and interaction. It is important for educators to remember that group work is not always successful or productive; the right conditions need to be in place, including a sensitivity to age differences in children’s social and cognitive competencies, for collaboration for collaboration to be successful. A limitation to the present study is that these results relate only to children’s and adolescents’ collaborative interaction on a short-term task of recall (a general knowledge quiz). Thus, the present study did not examine learning gains, but rather the influence of group dynamics on processes of communication and collaboration at different ages.

818

P. J. Leman / Cognitive Science 39 (2015)

Implications of the present findings for collaboration in tasks concerning conceptual learning and long-term cognitive change must be inferred. Particular caution is required extending these findings to group-based discovery learning or learning settings where new knowledge is actively tested or constructed. Future work is needed to establish how far and in what ways group dynamics connect with learning. In a similar vein, the measures of sociometric change used here assessed change immediately after the quiz and tell us nothing about longer term changes in social relationships between children. Further research is needed to extend the present findings of age differences to different tasks and age groups. It is also vital to remember that in everyday interactions in educational settings, group members bring a variety of beliefs, expectations, attitudes, and skills to their interactions. While this variation is inevitable, subtle aspects of context may have a significant effect on the effectiveness of collaboration and educators must be careful to select group members who can work in productive ways on tasks that allow for meaningful collaboration and learning. One key source of influence in such interactions is expertise (or the perception of expertise). For instance, if one child is perceived to be smarter, she may unduly influence collaborative interactions or other group members may fail to engage in the task (or “loaf”; see again Gummerum et al., 2013). Data from the present study suggest that what may differ with age is a sense in which such issues (e.g., expertise, being smarter) are used by 13-year-olds strategically to support the group as a unit rather than as a collection of individuals. Thus, compared to the 8-year-olds, the older children in this study were more competent social actors in the sense that they recognized that truly collaborative work could be meaningfully productive, in a way that the younger children did not. Future research needs not only to explore how far these findings extend to children at different ages, and undertaking different tasks, but also how far children’s nuanced knowledge about social groups influences orientation to peer collaboration. It is important to note that children at both ages derive social benefits of peer collaboration, as evidenced by positive changes in sociometric nominations compared with the nominal (virtual) no interaction control group. In the present study, the comparison “no interaction” condition was created virtually. But an alternative may have been to ask children to operate in groups without collaboration, or at least to find a situation that included social presence in a group, rather than a no group setting. It would also be interesting to explore what would happen if group members had been others’ answers without the opportunity for interaction. In this instance, a clear hypothesis is that younger children in cohesive groups would show no benefits (socially or in terms of performance) compared to real interactions in a highly cohesive group, whereas older children would. Nevertheless, an advantage of the method chosen in the present study is that it neatly demonstrates that groups often entail these social benefits for children. However, there may be aspects of being in a group—for instance, sharing a common identity (e.g., Leman & Lam, 2008; Leman et al., 2011)—that affect behavioral dynamics even when children do not collaborate on the same task. The present findings have clear and important implications for educators. Group collaboration works, but it works for different reasons at different ages: At 8 years, the group

P. J. Leman / Cognitive Science 39 (2015)

819

serves as a source of information, but at 13 years, the social opportunities for constructing knowledge together become important. In fact, it may be that many educators have an intuitive grasp of these age differences and apply them in the classroom. For instance, Baines et al. (2003) found that teachers tend to place younger children in larger groups and give them more highly structured activities in the classroom. Conversely, teachers tended to pay more attention to managing intragroup relationships and dynamics for older children and adolescents (11 years and above). Children and adolescents can derive benefits from group collaboration. Yet the reasons for these benefits differ with age. Younger children view the group as a source of information, whereas older children also recognize the group’s potential for collaboration and achieving a better understanding or a situation, or solution to a problem together. Thus, compared with younger children, at 13 years children appear, in some respects, to have a more “socialised” view of group collaboration and a different orientation to it. Researchers and educators need to be sensitive to these underlying social and cognitive changes through development and their impact on peer collaboration.

References Abrams, D., Marques, J. M., Randsley de Moura, G., Hutchison, P., & Bown, N. J. (2004). The maintenance of entitativity: A subjective group dynamics approach. In V. Y. Yzerbyt, C. M. Judd, & O. Corneille (Eds.), The psychology of group perception: Contribution to the study of homogeneity, entitativity, and essentialism. Philadelphia, PA: Psychology Press. Abrams, D., Rutland, A., Cameron, L., & Ferrell, J. (2007). Older but wilier: Ingroup accountability and the development of subjective group dynamics. Child Development, 43, 134–148. Azmitia, M. (1988). Peer interaction and problem solving: When are two heads better than one? Child Development, 59, 87–96. Baines, E., Blatchford, P., & Kutnick, P. (2003). Changes in grouping practices over primary and secondary school. International Journal of Educational Research, 39, 9–34. Baines, E., & Howe, C. (2010). Discourse topic management and discussion skills in middle childhood: The effects of age and task. First Language, 30, 508–535. Basden, B. H., & Basden, D. R. (1995). Some tests of the strategy disruption interpretation of part-list cueing inhibition. Journal of Experimental Psychology: Learning, Memory and Cognition, 21, 1656–1669. Damon, W., & Killen, M. (1982). Peer interaction and the process of change in children’s moral reasoning. Merrill-Palmer Quarterly, 28, 347–367. Doise, W., & Mugny, G. (1984). The social development of the intellect. Oxford, England: Pergammon Press. Duran, R. T., & Gauvain, M. (1993). The role of age versus expertise in peer collaboration during joint planning. Journal of Experimental Child Psychology, 55, 227–242. Feddes, A. R., Noack, P., & Rutland, A. (2009). Direct and extended friendship effects on minority and majority children’s interethnic attitudes: a longitudinal study. Child Development, 80, 377–390. G€ onc€ u, A. (1993). Development of intersubjectivity in social pretend play. Human Development, 36, 185–198. Gottman, J., Gonso, J., & Rasmussen, B. (1976). Social interaction, social competence, and friendship in children. Child Development, 46, 709–718. Gummerum, M., Hanoch, Y., & Keller, M. (2008). When child development meets economic game theory: An interdisciplinary approach to investigating social development. Human Development, 51, 235–261.

820

P. J. Leman / Cognitive Science 39 (2015)

Gummerum, M., Keller, M., Takezawa, M., & Mata, J. (2008). To give or not to give: Children’s and adolescents’ sharing and moral negotiations in economic decision situations. Child Development, 79, 561–576. Gummerum, M., Leman, P. J., & Hollins, T. (2013). Children’s collaborative recall of shared and unshared information. British Journal of Developmental Psychology, 31, 302–317. G€ uro glu, B., van den Bos, W., & Crone, E. A. (2009). Fairness considerations: Increasing understanding of intentionality during adolescence. Journal of Experimental Child Psychology, 104, 398–409. Howe, C. (2009). Collaborative group work in middle childhood: Joint construction, unresolved contradiction and the growth of knowledge. Human Development, 52, 215–239. Howe, C., & McWilliam, D. (2006). Opposition in social interaction among children: Why intellectual benefits do not mean social costs. Social Development, 15, 205–231. Kruger, A. C., & Tomasello, M. (1986). Transactive discussions with peers and adults. Developmental Psychology, 22, 681–685. Kuhn, D., Cheney, R., & Weinstock, M. (2000). The development of epistemological understanding. Cognitive Development, 15, 309–328. Leman, P. J., & Bj€ornberg, M. (2010). Gender, conversation, and development: A study of children’s conceptions of punishment. Child Development, 81, 958–972. Leman, P. J., & Duveen, G. (1996). Developmental differences in children’s understanding of epistemic authority. European Journal of Social Psychology, 26, 683–702. Leman, P. J., & Lam, V. L. (2008). The influence of race and gender on children’s conversations and playmate choices. Child Development, 79, 1329–1343. Leman, P. J., Macedo, A. P., Bluschke, A., Hudson, L., Rawlings, C., & Wright, H. (2011). The influence of gender and ethnicity on children’s peer collaborations. British Journal of Developmental Psychology, 29, 131–137. Leman, P. J., & Oldham, Z. (2005). Do children need to learn to collaborate? The effects of age and age differences on children’s collaborative recall. Cognitive Development, 20, 33–48. Merrell, K. V. (2003). Behavioral, social, and emotional assessment of children and adolescents (2nd ed.). Mahwah, NJ: Lawrence Erlbaum Associates. Piaget, J. (1932). The moral judgment of the child. London: Routledge & K. Paul. Piaget, J. (1937/1971). The construction of reality in the child (trans. M. Cook). New York: Ballantine. Psaltis, C. (2011). The constructive role of gender asymmetry in social interaction: Further evidence. British Journal of Developmental Psychology, 29, 305–312. Psaltis, C., Duveen, G., & Perret-Clermont, A.-N. (2009). The social and the psychological: Structure and context in intellectual development. Human Development, 52, 291–312. Serbin, L. A., Powlishta, K. K., & Gulko, J. (1993). The development of sex typing in middle childhood. Monographs of the Society for Research in Child Development, Serial No.232, 58. Slavin, R. E. (1996). Research on cooperative learning and achievement: What we know, what we need to know. Contemporary Educational Psychology, 21, 43–69. Tolmie, A., Topping, K., Christie, D., Donaldson, C., Howe, C., Jessiman, E., Livingston, K., & Thurston, A. (2010). Social effects of collaborative learning in primary schools. Learning and Instruction, 20, 177–191. Topping, K. J., Peter, C., Stephen, P., & Whale, M. (2004). Cross-age peer tutoring of science in the primary school: Influence on scientific language and thinking. Educational Psychology, 24, 57–75. Tudge, J. R. H. (1992). Processes and consequences of peer collaboration: A Vygotskian analysis. Child Development, 63, 1364–1379.

Copyright of Cognitive Science is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.