Teaching and Learning Research Programme Annual Conference Papers 5th Annual Conference, 22-24 November 2004 Cardiff Marriott Hotel The impact of collaborative group work in primary classrooms and the effects of class composition in urban and rural schools Donald Christie*, Andrew Tolmie*, Christine Howe*, Keith Topping**, Allen Thurston**, Emma Jessiman*, Kay Livingston* and Caroline Donaldson***. *University of Strathclyde **University of Dundee ***Formerly at Dundee, now at Queen Margaret University College NB: This paper was presented at an internal TLRP conference; if you wish to quote from it please contact the authors directly for permission. Contact details for each project and thematic initiative can be found on our website (www.tlrp.org). The impact of collaborative group work in primary classrooms and the effects of class composition in urban and rural schools Donald Christie*, Andrew Tolmie*, Christine Howe*, Keith Topping**, Allen Thurston**, Emma Jessiman*, Kay Livingston* and Caroline Donaldson***. *University of Strathclyde **University of Dundee ***Formerly at Dundee, now at Queen Margaret University College ABSTRACT The paper outlines key findings from the ScotSPRinG project conducted by a team from the Universities of Strathclyde and Dundee funded as a TLRP Scottish Extension project linked to the Phase II Project: “Improving the effectiveness of pupil groups in classrooms” led by Professors Blatchford, Galton and Kutnick. The project involved the participation of pupils and teachers in 24 primary schools in eight local authorities across Scotland. Aim The aim of this intervention study was to evaluate the impact of collaborative group work in selected Scottish Primary classrooms and explore the effects of class composition in urban and rural schools. Method Approximately 600 upper primary stage pupils were involved in the study. A two-phase intervention designed to foster collaborative group work, using similar approaches to those devised by the TLRP Phase II SPRinG Project team, was carried out in four different categories of schools derived from the combination of two factors: urban/rural school; single age/mixed age classes. The first phase of the intervention involved social and communication skills training activities. The second phase focused on collaborative group work in two primary science topic areas: evaporation and forces. A battery of pre- and post-intervention assessments was implemented using a range of cognitive, affective and social measures, including a newly devised sociometric instrument designed to tap into the different patterns of social relationships characterising the four different categories of schools. In addition, systematic classroom observations were conducted during the different phases of the study and detailed evaluation feedback was gathered from participating teachers. Results A number of interesting initial findings have emerged showing significant gains across a number of measures, which were attributable to the group work intervention. In particular there were significant gains in micro measures of attainment in the two specific science topics. Regression analysis showed the cognitive gains to be related to indices of the quality of collaborative dialogue during group work. While there were no consistent overall differences between composite and single age classes nor between urban and rural schools, group work yielded significant gains in social relations in the intervention classes and again collaborative engagement within tasks was found to be associated with the social benefits. The evidence suggests that the socio-emotional gains are independent of the cognitive gains. Evaluative feedback obtained from teachers was generally very positive in terms of the quality of the training and support package and in terms of the impact of the intervention on both pupil learning and their own professional practice . INTRODUCTION Research over the past 20 years has established beyond dispute that interaction between learners is a powerful and natural mechanism for promoting enhanced clarity of conception and articulation, but group work within schools has typically failed to make best use of the potential of interactive learning. The present study (ScotSPRinG) was designed as a Scottish Extension project linked to the TLRP Phase II Project, “Improving the Effectiveness of Pupil Groups in Classrooms,”also described as Social Pedagogic Research into Grouping (SPRinG) based in the South-East of England at the Universities of London (Institute of Education), Brighton and Cambridge. The principal aim of both projects is to enhance the learning potential of pupils working in classroom groups by actively involving teachers in programmes designed to raise the incidence and quality of group work during typical activities. The rationale for the extension of the Phase II project to the Scottish context was based partly upon demographic differences and partly upon the distinct forms of educational provision in Scotland which provided an opportunity to assess the generalisability of the SPRinG project’s resources, and to extend the social pedagogy being developed through the research. The use of group work in schools has much the same tradition in Scotland as in England i.e. an increased incidence during the 1960s, reflecting policy shifts towards child-centred teaching and the view that group work allowed appropriate differentiation of ability levels in classroom activities; followed by its espousal as essentially an organisational device rather than as a teaching strategy. Thus in Scotland, too, the principal issue motivating the SPRinG research holds true: group work exists, but it is frequently not effectively planned (Darling, 1999). Where Scotland does differ markedly from England (particularly the South-East) is in its inherently greater variation in primary school size, with all that follows from this in terms of class and group characteristics. This variation reflects the much wider range of urban and rural communities to be found in Scotland, with small village primary schools still being commonplace in many areas. One consequence of the retention of schools linked to small communities is that pupils typically possess much greater familiarity both with each other and with each other’s families. Initial evidence from the SPRinG Project (Kutnick, personal communication, 2003) pointed to the importance of a relational basis (e.g. via trust and social communication exercises) for competence in group work to operate outside of friendship pairings. The evidence also tended to confirm that it is possible to foster through dyadic and other group activities productive forms of communication, such as the exploratory talk identified by Mercer (1995), or the disagreement-explication cycle defined by Howe & Tolmie (1998). Thus in terms of the framework of influences on effective group work identified by the SPRinG project, children in schools linked to small communities might be expected to have stronger preexisting relational bases for joint activity, and to fall more naturally and competently into group work as a result (cf. early research by Shapira & Madsen, 1969, on the greater cooperative tendencies within school of children from more collective communities). At the same time, though, community schools of this kind by necessity also make considerable use of composite classes because small school numbers make age banding unfeasible, and they have more individualised learning programmes to redress the resulting within-class diversity. Joint activity in such classes may therefore be less common. Where it does occur, it will often mean interaction between children of different ages and differing levels of expertise, altering the dynamic from one of collaboration between peers to one of tutoring of the junior partner by the senior (cf. Piaget, 1932; Rogoff, 1990; Howe, Tolmie, Duchak-Tanner & Rattray, 2000). Given the positive outcomes reported in the literature on peer tutoring (Topping & Ehly, 1998), this does not entail that group work is necessarily less effective under these conditions. However, it may well alter the operation of factors such as relational ease. For instance, while better understanding on the part of the tutor of the knowledge held by the tutee contributes to effective tutoring (Foot, Shute, Morgan & Barron, 1990), greater relational distance might plausibly be needed for the tutee to be prepared to be guided by the tutor. Moreover, a substantial literature attests to the fact that peer collaboration and peer tutoring have different natural ambits of applicability, with the former lending itself more to conceptual explication, and the latter to procedural understanding (see Howe et al., 2000). The implication is, then, that the ways in which teachers make use of group work might need to be different in small school settings, undermining the generalisability of the SPRinG resources, and indicating the need for more differentiated, context-specific models of effectiveness. Investigation of this issue simply via a comparison between small rural schools and larger urban ones is problematic, however, because of the confound between cross-age group work and greater personal familiarity. If results suggested there were problems in extending the SPRinG resources to rural schools, it would be unclear whether this was attributable to the lack of relational distance, or to a mismatch between the resources and the use of cross-age combinations per se. The distinction is an important one, since the former would imply that the situation might be rectified to some extent simply by choosing pupil groups more carefully, whereas the latter would indicate the need to develop separate support regimes for teachers working in these contexts. Fortunately, however, the Scottish primary system provides a means of disentangling the possible influences, since compositing is not restricted to rural schools, but is also present to some degree in many urban primary schools, where it is used to cope with variation in the size of intake. Thus some cross-age group work occurs in contexts where out of school familiarity is lower. Application of the SPRinG resources to selected types of school in Scotland, therefore, allows a form of natural experiment to be conducted, comparing the effects of supporting group work in this way under four conditions: 1) classes in rural schools, where group work is typically cross-age; 2) classes in larger rural schools where compositing is not used and hence group work is between same-age children; 3) classes in urban schools with compositing, where at least some group work is cross-age and 4) classes in larger urban schools where compositing is not used, and group work is between same-age children. Comparison of the these conditions in terms of the relative effectiveness of the existing support programmes would make it possible to pinpoint exactly how far it is necessary to devise different support systems and different pedagogical frameworks for promoting productive group work in rural schools, and also in other contexts where composite classes exist. Previous research has emphasised the value of peer activity in promoting social competence (Warden & Christie, 1997). By adopting a fine-grained approach to promoting effective group work at primary school level, especially in the later years, the present study was designed to provide evidence across a range of typically Scottish contexts on the direct effects of group work on educational attainment, on pupils’ social relationships and willingness to work independently of teachers and on how these effects interact with social circumstance. The study was also designed to foster a network of interested and experienced teachers and a set of resources adapted as necessary to suit the Scottish context, which could be used by schools as part of the self-evaluation of their provision. At a more theoretical level, the present study was aimed at providing systematic insight into the influence of an important set of contextual factors on productive group work, a previously under-researched area, both in Scotland and more generally. This has a direct bearing on the wider SPRinG objective of developing a social pedagogy for group work, since the role of contextual variation (cf. Brofenbrenner, 1979; Pellegrini & Blatchford, 2000) has already been identified as a central concern within such a model. A number of theoretical perspectives cast light on the functioning of pupil groups in classrooms. Piagetian and neoPiagetian theories address the effects on collaborative group work of differences in perceived status, relative facility with the subject matter and the cognitive consequences of friendship (Piaget, 1932; Doise & Mugny, 1984; Howe & Tolmie, 1998; Williams & Tolmie, 2000; Azmitia & Montgomery, 1993). Vygotskian and neo-Vygotskian theories articulate the role of relative differences in procedural expertise, and the tactical use of such differences by the more expert to scaffold the activity of the less expert until they achieve autonomy (Vygotsky, 1978, Wood, 1986; Rogoff, 1990). Activity theory focuses on activities as consensual practices or ways of achieving particular objectives which are shared between members of a community, and which, therefore, depend on various forms of mutual knowledge and conjoint past experience (Leontiev, 1981; Engeström, 1987; Cole & Engeström, 1993). Taken together, these different theories provide a framework for interpreting the observed impacts of urban vs rural context and same vs cross-age grouping on methods of supporting the implementation of effective group work in different types of school; and for generating a unified social pedagogy which explicitly takes into account the processes underlying these influences, and ways in which these might be managed or capitalised upon. Key differences between the SPRinG and ScotSPRinG projects are that while SPRinG was designed to investigate work in pupil groups across the curriculum and across Key Stages 1,2 and 3, ScotSPRinG focused mainly on the curriculum area of Primary Science and confined its investigation to classes of pupils aged 10-12 at the Primary 6 and Primary 7 stage in Scottish primary schools, a stage equivalent to the boundary of KS2 and KS3. The choice of science as the main curricular focus had a number of advantages. Firstly, it was the focus of the Key Stage 2 and 3 interventions in the SPRinG work, so it provided a crucial linkage between the projects. Secondly, it is an area in which the use of group work is more prevalent (34% of activity, as against 16% for language and literacy, and only 6% for mathematics; see Blatchford, Kutnick & Baines, 1999), so that there is likely to be a track record to assist school selection and provide a point of departure. Thirdly, despite the greater prevalence of group work in science, this has commonly been dictated by the limited availability of equipment rather than any pedagogic strategy. Thus there is considerable scope for improvement via the planning of effective activities, rendering outcomes potentially more detectable and more discriminable from each other. Fourthly, science work is not typically associated with ability setting, thus avoiding potential confounds between the variables under investigation and constraints on relative levels of expertise. At the same time, however, whilst the intervention was focused on science, it was important to establish something of its wider effects on other areas of the curriculum. Thus, bearing in mind the fact that 5-14 Guidelines are explicitly concerned with integration across the curriculum, it was seen as important in the present study to gather observational and descriptive data on group work in other areas of the curriculum and to assess attainment in Mathematics and Language as a means of gauging spill-over impacts. Original aims of the project: To examine how far the Phase II SPRinG project’s programmes for supporting effective group work need to be adapted for use in rural and urban schools with composite classes, where interactional styles are different from those in same-age urban classes To identify a representative sample of teachers in rural and urban schools with and without composite classes, and recruit these teachers to in-service programmes which will support them in the planning and implementation of group work activities in science To collect data on learning outcomes for participating pupils, both in science and more widely, and also on changes in self-esteem and quality of collaborative behaviour in class To establish whether there are differences in outcomes associated with cross-age vs sameage pupil groups and rural vs urban settings, and to compare patterns of outcome to those obtained by SPRinG To interpret the findings in the light of current models of factors affecting collaboration between pupils, and extend the social pedagogy being developed by SPRinG To make recommendations about modifications to group work support programmes for teachers of composite classes in rural and urban schools. METHODOLOGY Initial Survey A questionnaire was issued to a randomly selected, one-in-three sample of primary schools in eight Scottish local authorities. The survey instrument comprised four sections dealing with: 1. Background information about the school (roll, number of classes, number of composite classes, numbers of children at P6 and P7 stages) 2. Current approaches to group work (form, frequency) 3. Science topics planned for P6 and P7 in the current school year. 4. Willingness to participate in the intervention study. Altogether 221 questionnaires were issued and 85 were returned (38%). Sample for Intervention Study The final set of schools for the intervention study was selected in a way which aimed to capitalise on systematic patterns of natural variation. Geographical and demographic spread was achieved by using 2 clusters of local authority areas, yielding 12 schools in West-Central Scotland (the Strathclyde sample) and 12 schools in the North-East (the Dundee sample). Within each group of 12 schools there were 6 urban and 6 rural schools. Within each of the urban and rural sub-samples of schools 3 “single-age” P6 or P7 classes and 3 composite or “mixed age” P6/P7 classes were identified as the sites for the intervention, each school involved providing one class in which the study would take place, as shown in Table 1. Table 1. Composition of the sample of schools/classes taking part in the intervention study Strathclyde Sample Dundee Sample Urban Rural Urban Rural 6 schools 6 schools 6 schools 6 schools Single Mixed Single Mixed Single Mixed Single Mixed age age age age age age age age 3 classes 3 classes 3 classes 3 classes 3 classes 3 classes 3 classes 3 classes Where school roll and composition allowed, schools were also asked to provide ‘control’ classes who would take part in the assessment but would not be involved in the group work training during the present school session in which the planned intervention would take place. Only three such control classes were recruited comprising approximately 70 pupils. This was not seen as a major concern for the study, since the research design primarily rests on comparisons between outcomes for different categories of school, and teachers and classes will in a sense provide their own controls in as much as the study will track change over time attributable to the intervention. Nevertheless it was felt that a small control sample would provide useful data to enable evaluation of the impact of the intervention. The Intervention – Phase 1, Group Work Skills Training The teachers of each of the 24 classes involved in the intervention study attended three days of in-service training, 12 attending the sessions arranged at the University of Dundee and a total of 14 at the University of Strathclyde (including one job-share pairing and one additional teacher who took over responsibility for one of the intervention classes during the study). The focus of the first day of training, which took place before the classroom activities were introduced to the children, was on potential benefits of group work and effective strategies for introducing classroom activities designed to enhance children’s social skills and communication skills. A comprehensive package of materials to support group work training was issued to each teacher, based on the materials produced by the TLRP Phase II SPRinG team. Care was taken to ensure that the in-service sessions included the opportunity for discursive input by teachers themselves, in order to mirror the SPRinG approach. Teachers were encouraged to complete a 12-week programme, during which approximately one hour per week was devoted to group work training activities using the package of resources developed by the SPRinG team. These were designed to foster social and relational skills and communication skills. In addition, teachers were expected to devote approximately a further one hour per week of class time to group work within ‘normal’ curricular contexts. The Intervention – Phase 2, Group Work in Science Training The second training day, which took place towards the end of the first phase of the intervention, focused on curricular applications of group work and in particular on the use of group work in two units of Primary Science, namely, Forces and Evaporation, which are specified at Levels D and E of the 5-14 Environmental Studies Guidelines. During this phase of the intervention, teachers were expected to devote at least an hour per week over a 6- to 8week period to classroom activity including structured group work connected with the two Science topics, supported by further comprehensive classroom resources. In addition, teachers were expected to continue with their use of group work across the rest of the curriculum Evaluation of the Intervention On each of the training days, time was spent explaining the teachers’ role with respect to the pre- and post-intervention assessment procedures to be implemented. In addition, the pattern of researcher visits and the observation procedures to be adopted throughout the study were explained to participating teachers. The third in-service day, which took place after all classroom data collection was complete, was devoted to debriefing, feedback and evaluation. An evaluation questionnaire was issued to each of the participating teachers and focus group discussions were recorded dealing with the implementation and outcomes of the group work schemes and general reflections on experiences of the intervention. Assessment and Observation Instruments A comprehensive battery of assessment was adopted for pre- and post-testing of both cognitive and affective outcomes. In the cognitive domain, the Performance Indicators in Primary School (PIPS) tests (CEM, Durham) were used to assess children in intervention and control classes on measures of academic attainment in Mathematics, Language and Science as well as on their non-verbal reasoning. Micro measures of learning outcomes relating the two specific Science topics, Forces and Evaporation, were obtained using parallel forms of tests similar to those used by the SPRinG team, before and after the group work in science intervention phase of the study. In the affective domain, the PIPS tests included measures of attitudes to school and to school subjects. A questionnaire, “My Feelings”, devised by the SPRinG project team to measure children’s attitudes to group work, was administered along with the “general worth” 6-item subscale of the Harter Self-Esteem questionnaire. A new sociometric instrument, “People in your Class”, was developed to provide data on children’s social relationships and patterns of interaction both in school and outside of school. Presented in the form of a matrix, this asked children to consider six key context questions (columns) regarding their relationships with all other members of their class (rows). The context questions included, for example, “Who do you like to play with in school?” and “Who do you like to play with out of school?” as well as questions to do with familiarity with children’s families and whether peers would be likely to be met locally, or at ‘clubs’ outside school. The observation schedule used by the researcher during observation visits focused on the collaborative activity occurring in participating classrooms at three time points over the course of the study. At each time point two observation sessions were undertaken, one during a conventional class lesson and one during a planned group work lesson. During each session, six target children per class were observed for eight 40-second time windows. The SPRinG schedule was adapted to focus on “interactants” and quality of dialogue. Within “dialogue” there were two main categories of codes: i. Collaborative codes: proposition; disagreement; explanation; reference back; and resolution/compromise ii. Tutoring codes: instruction; and question In addition to this fine-grained observation schedule, a general class-level observation instrument devised by the SPRinG team, namely, “S-TOP” was used at each of the three time points to capture, on a 3-point scale, global ratings of the quality of the learning context (4 items); suitability of tasks and activities (7 items); degree of adult (teacher) involvement (10 items); group work skills (11 items) evident in the intervention classrooms. RESULTS Initial survey of schools Schools were asked to indicate the frequency with which group work was used as a general classroom strategy. The percentage frequencies were as follows (N=81): “Frequently” 90%; “Sometimes” 9%; “Rarely” 1%; “Never” 0%. Schools were also asked to indicate their use of different forms of group work across the curriculum. The pattern of responses regarding use of collaborative group work in curriculum areas is shown in Table 2. The range of percentage frequencies was striking, with the highest percentage of schools indicating that they used collaborative group work in Mathematics and in Environmental Studies and the lowest percentages in Religious and Moral Education and in English Language. Table 2: Percentages of schools indicating use of collaborative group work in each area of the curriculum CURRICULUM AREA Mathematics Environmental Studies Expressive Arts Personal and Social Development Health Education Religious and Moral Education English Language %YES %NO 88 87 76 72 48 39 32 12 13 24 28 52 61 68 Pre- and post-testing The data from the pre- and post-tests was analysed with respect to the nature and extent of any differences between a) rural vs urban school context, and b) cross-age vs same-age grouping. Of particular concern in the analysis were the following three questions: i. whether the range of measures employed indicated that the programme of support for group work had been more productive in terms of incidence or effectiveness under some conditions than others; ii. whether observed differences were general or restricted to particular aspects of performance; and iii. crucially, whether cross-age urban groups were more similar in pattern of effects to rural cross-age groups (indicating a general cross-age effect) or to urban same-age groups (indicating that more specific effects are operating in the rural context). Cognitive measures The PIPS performance measures (Science, Maths Reading, Vocabulary and Non-Verbal Reasoning) all showed significant pre- to post-test shifts. However, there was no obvious effect of the intervention in that effect sizes were similar across intervention and control children and there were no pre- vs post-test x condition interactions in overall 3-way Anovas. There were, however, significant area x class interactions on every measure except reading, indicating that urban, straight-age classes perform at the lowest level whilst rural straight age classes perform at the highest. The pre- and post-test comparisons for the micro measures associated with the two specific science topics were consistent were more promising in terms of showing an effect potentially attributable to the intervention. The mean scores (and s.d.) for the “Evaporation” pre- and post-tests in composite and straight age classes are shown in Table 3. Table 3 Mean pre- and post-test scores (and s.d) for Evaporation topic in the intervention and control conditions in composite and straight age classes in rural and urban areas (min. score = 0; max. score =19). Condition/Time Rural Comp Rural st age Urban comp Urban st age of test (n=117) (n=121) (n=135) (n=141) Intervention Pre9.26 9.45 9.85 8.34 test (2.73) (3.02) (5.07) (2.70) Intervention 12.48 12.37 12.67 11.32 Post-test (3.97) Control Pre-test Control Post-test (3.41) Rural st age (n=23) 12.39 (2.55) 11.00 (1.88) (3.17) (3.60) Urban st age (n=48) 9.29 (3.10) 9.73 (3.34) Three separate analyses of variance were carried out on this data. Three-way Anova of the scores of classes in the intervention condition showed the pre- vs post-test difference was highly significant: F(1,510) = 270.27, P < .001. The effect size (partial eta-squared) for this comparison was 0.35. The effect of type of class (straight age vs. composite) was also significant: F (1,510) = 7.27, P = .007 with children in composite classes scoring higher overall than those in straight age classes. The Type of Class x Area (Urban vs. Rural) interaction was also significant: F(1,510) = 8.27, P = .004. In the 2-way Anova of scores of classes in the control condition the pre- vs post- test difference was not significant. In the overall 3-way Anova of pre- and post-test scores in the intervention and control conditions in urban and rural classes the critical Pre- vs Post x Condition interaction was significant, F(1,329) 29.12, P < .001 (effect size = .14). A very similar pattern of scores was obtained from the “Forces” pre- and post-tests (See Table 4). Again the analyses of variance showed significant pre- to post-test gains in the intervention classes: F (1,461) = 158.02, P < .001 (partial eta-squared effect size = 0.25), but no significant gains in control classes. Table 4 Mean pre- and post-test scores (and s.d) for Forces topic in the intervention and control conditions in composite and straight age classes in rural and urban areas (min. score = 0; max. score =37). Condition/Time Rural Comp Rural st age Urban comp Urban st age of test (n=92) (n=116) (n=143) (n=114) Intervention Pre19.86 22.46 19.87 19.85 test (5.12) (4.69) (4.47) (4.97) Intervention 23.59 24.86 22.78 23.04 Post-test (4.93) (5.04) (5.19) (5.28) Control Pre-test Control Post-test Rural st age (n=20) 21.35 (3.25) 21.60 (4.42) Urban st age (n=40) 24.05 (5.62) 25.02 (4.98) Group influences on cognitive gains Regression analysis was used to establish significance of the observational data and other measures of the quality of group work in predicting the pattern of attainment in the two science topic post-tests. The pattern of results was fairly consistent across the two topics. Firstly, observed frequency of children giving explanations during group work positively predicted post-test scores for Evaporation (Beta = 0.28, t = 2.20, p<.05) and similarly for Forces, though failing to reach significance, (Beta = 0.25, t = 1.81, p=.07). Secondly, the STOP ‘Adult’ score proved to be a negative predictor for both topics: Evaporation (Beta = 0.36, t = 3.27, p<.001); Forces (Beta = -0.39, t = 3.38, p<.001). This indicated that the less adults intervened, and the more they monitored, guided without directing and encouraged pupil interaction, the better. Thirdly the S-TOP ‘Learning Context’ scores proved to be a positive predictor of gains in Evaporation (Beta = 0.26, t = 3.67, p<.001) and in Forces (Beta = 0.19, t = 2.48, p<.05). Affective and Social measures Sociometric measures The sociometric instrument yielded the predicted pattern of overall differences between the rural and urban schools. Compared with children in urban schools, children in rural schools nominated a higher percentage of classmates that they: play with in school: F(1,477) = 30.40, P < .001; work with in school: F (1,477) = 6.12, P = .014; play with out of school: F(1,477) = 34.37, P < .001; see at clubs: F(1,477) = 22.43, P < .001; know also within their family context: F(1,477) = 94.07, P < .001; and see locally: F(1,477) = 164.46, P < .001. The anticipated improvements in classroom relationships were evident in positive pre- to post- test shifts in the first two of these measures in intervention classes (“play with” F(1,477) = 8.13, P = .005; “work with” F(1,477) = 22.03, P < .001). However, a degree of caution is required, since a significant shift was also recorded in the “play with” measure in control classes, though the pattern of scores here was very irregular with unusually high percentages of children being nominated overall in the single rural control class. Self-Esteem measure The aggregate scores on the Harter “General Worth” subscales yielded a significant difference between urban and rural schools, with children in urban schools scoring more positively on this measure of self-esteem than children in rural schools: F(1,493) = 7.74, p = .006. There was no evidence of an overall pre- to post-test improvement in self-esteem. However, there was a significant three-way Pre/post x Area x Class interaction, which indicated that while self esteem stayed the same or slightly declined in both types of rural classrooms and in urban composite classes, there were gains in the urban single-age classes: F(1,493) = 9.77, P = .002. Group work influences on socio-emotional measures Regression analysis was used to establish the significance of the observational data and other measures of the quality of group work in predicting the pattern of gains in the two key sociometric measures (percentage of the members of their class children liked to ‘work with’ and ‘play with’). The S-TOP measure of ‘task suitability’ positively predicted sociometric gains in percentage of the class children indicated they liked to ‘play with’ (Beta = 0.313, t = 6.49, p<.001) and ‘work with’ (Beta = 0.395, t = 5.78, p<.001). The teachers’ ratings of children’s group work skills at the end of the intervention also predicted the social gains both in terms of ‘work with’(Beta = 0.451, t = 8.84, p<.001) and ‘play with’ (Beta = 0.423, t = 6.61, p<.001). Teacher Evaluations Teachers (N=26) were issued with an evaluation questionnaire which asked them to rate how useful they had found the various aspects of the intervention package (in-service staff development, support and materials provided) and to evaluate its impact on pupils’ learning and their own practice. Teachers across all four conditions gave equally positive ratings of all of these aspects of the intervention. Individual comments were also gathered from teachers during the final in-service day. A small selection of these indicates the tone of much of what was said. For example, one very experienced teacher commented: “As a teacher, I am obviously aware of the effectiveness of group work and have always grouped children socially to make their behaviour and learning as effective as possible. The SPRinG study took this a good few steps further and made me realise that there is a variety of structures of grouping that can be even more effective. The materials given by the group were excellent and made me aware of how my class learned, how they reacted to one another.... I was inspired to take the use of group learning into many areas of the curriculum and use it as an effective learning tool. I will continue this into next year, with a new class, and be able to monitor their progress as well as develop a whole host of new skills – not only for the pupils, but for the teacher as well!” Commenting particularly on the quality of the Science materials and activities, another said: “I thought the project presented an interesting way to teach science. The group work skills package has proved to be very beneficial. There has been a positive change in their attitudes towards working in a group and getting along with each other.” Finally, teachers had also given thought to the implications for pedagogy of their experiences on the project. One illuminating piece of advice offered by one of the participating perhaps shows an awareness the value of teachers supporting but not to closely intervening in group work: “Learn to let go – this approach gives children more responsibility for their own learning” DISCUSSION AND CONLUSIONS The results of the initial survey together with the observational data gathered during the study tend to suggest that while the primary schools in the present sample report that group work is a common feature of classroom activity, there was considerable scope for improvement in the quality of group work and in children’s mastery of the skills of cooperation and collaboration. Supporting group work yielded a number of important gains. In particular the interventions related to the science topics proved highly successful. The Science activities involved a number of components which may have contributed to the gains made, including whole class lessons, practical activities and well structured materials as well as collaborative group work activity. However, the regression analyses of the observational data clearly showed that the group work component played a significant role. It was also important to note that the properties of effective group work were similar across the two different topic areas covered and that these properties observed in a range of real classroom contexts corresponded closely to those observed in experimental settings (e.g. Mercer, 1995; Howe & Tolmie, 1998). Classroom observations showed that the quality and effectiveness of collaborative group work improved significantly in participating classrooms over the course of the study. The predicted differences between urban and rural schools in terms of pre-existing patterns of social relations were confirmed with children in the rural schools being more likely to have social contacts with other children both in school and outside in the community. However, the socio-emotional gains attributable to the intervention were broadly similar across the rural and urban schools as well as across composite and non composite classes. In many ways this is an encouraging finding. It was in the area of social relationships, as indicated by the sociometric data, where the clearest gains were made. Regression analysis confirmed that it was the quality of collaborative engagement within tasks that was central to the benefits obtained. It is also worth noting that the socio-emotional gains were broadly independent of cognitive gains. While our large structural variables of rural vs urban area and class type did not account for the variation in affective outcomes, the potential impact of group work needs to be understood in terms of a much finer grained analysis set against the background of demographic factors that may also be having an influence In terms of the evaluative feedback received from participating teachers, the intervention was highly valued by teachers across the different types of class/school who were equally likely to report benefits both in terms of their professional practice and in terms of pupil learning. The SPRinG approach was seen by teachers as effective in bringing about this improvement. The intervention was designed in such a way as to be sustainable and it was encouraging that many of the participating teachers indicated they planned to continue to develop this aspect of their practice in the future. REFERENCES Blatchford, P., Kutnick, P. & Baines, E. (1999). The nature and use of classroom groups. End of Award Report to ESRC. Cole, M. & Engestrom, Y. (1993). A cultural-historical approach to distributed cognition. In G. Salomon (Ed.), Distributed Cognitions: Psychological and Educational Considerations. Cambridge: Cambridge University Press. Darling, J. (1999). Scottish Primary Education: Philosophy and Practice. In T.G.K.Bryce and W.M.Humes (Eds.), Scottish Education. Edinburgh: Edinburgh University Press. Doise, W. and Mugny, G. (1984). The Social Development of the Intellect. Oxford: Pergamon. Engestrom, Y. (1987). Learning by Expanding: An Activity-Theoretical Approach to Developmental Research. Helsinki: Orienta-Konsultit. Foot, H.C., Shute, R.J., Morgan, M.J. & Barron, A. (1990). Theoretical issues in peer tutoring. In H.C. Foot, M.J. Morgan & R.J. Shute (Eds.), Children Helping Children. Chichester: John Wiley. Howe, C.J. & Tolmie, A. (1998). Productive interaction in the context of computer-supported collaborative learning in science. In K. Littleton & P. Light (Eds.), Learning with Computers: Analysing Productive Interaction. London: Routledge. Howe, C.J., Tolmie, A., Duchak-Tanner, V. & Rattray, C. (2000). Hypothesis testing in science: group consensus and the acquisition of conceptual and procedural knowledge. Learning & Instruction, 10, 361-391. Leontiev, A.N. (1981). Problems in the Development of Mind. Moscow: Progress Publishers. Mercer, N. (1995). The Guided Construction of Knowledge: Talk Amongst Teachers and Learners. Clevedon: Multilingual Matters. Pellegrini, A. & Blatchford, P. (2000). Children’s Interactions at School: Peers and Teachers. London: Edward Arnold. Piaget, J. (1932). The Moral Development of the Child. London: Routledge & Kegan Paul. Rogoff, B. (1990). Apprenticeship in Thinking: Cognitive Development in Social Context. Oxford: Oxford University Press. Shapira, A. & Madsen, M.C. (1969). Cooperative and competitive behaviour of kibbutz and urban children in Israel. Child Development, 4, 609-617. Topping, K. J. & Ehly, S. (Eds.) (1998). Peer-assisted learning. Mahwah NJ & London UK: Lawrence Erlbaum. Vygotsky, L.S. (1978). Mind in Society: The Development of Higher Psychological Processes. Cambridge, MA: Harvard University Press. Warden, D. and Christie, D. (1997) Teaching Social Behaviour: Classroom activities to foster interpersonal awareness. London: David Fulton. Williams, J.M. & Tolmie, A. (2000). Conceptual change in biology: group interaction and the understanding of inheritance. British Journal of Developmental Psychology, 18, 625-649. Wood, D. (1986). Aspects of teaching and learning. In M. Richards & P. Light (Eds.), Children of Social Worlds. Cambridge: Polity Press.