This page shares the executive summary of our Nuffield Foundation-funded project, developed in partnership with teachers to create a classroom-friendly programme for teaching social conversation skills. You can also download the whole report here.
How to cite this report:
Abbot-Smith, K., Smith, E., Sturrock, A., Alcock, C., Matthews, D., Dockrell, J., Bannard, C. & Wilson, C. (2026) Conversation for Social Interaction: Developing a classroom intervention to improve conversation skills in primary schools. Final Project Report. Available online: https://www.cosi-conversation.org.uk/reports/
Executive Summary of the Final Project Report
Conversation skills allow children and adults to form and sustain friendships and to foster positive relationships with peers. In the general population, there is variability in these skills, with consequences for academic and vocational success as well as social and mental wellbeing. To keep a conversation going, children need to be able to listen attentively, provide relevant responses, engage in appropriate turn-taking, consider the perspectives of their conversation partner and repair misunderstandings. These conversation skills in turn support children to engage in classroom learning and curriculum-based discussion with peers. They also predict and precede better mental health and behaviour as children grow older.
Teaching conversation skills is currently a statutory requirement in the English primary curriculum. The recent UK government-commissioned Curriculum and Assessment Review has highlighted the need for additional oracy support and has recommended the establishment of an oracy framework for schools across all key stages. Supporting conversational ability should be central to this oracy framework, to allow children the opportunity to think out loud, debate, collaborate, share perspectives and question. However, currently teachers receive limited training about oral language development during their initial teaching training and conversation skills are notably absent.
Crucially, there are no evidence-based programmes which aim to support conversation skills in primary school pupils. This Nuffield Foundation-funded project has aimed to co-develop with stakeholders a free, school-based conversation skills programme and to test whether teachers would find it easy to use and view it positively. Below are the four objectives of this project and how these were met.
Objective 1. To develop an effective, teacher-friendly programme and materials in collaboration with upper-primary teachers: Creation of the CoSI programme.
Building on prior preparatory work with teachers, we developed the Conversation for Social Interaction (CoSI) programme in iterative steps with input from a Teacher Advisory Group and four Year 4 teachers who trialled the initial programme. The revised programme (see under the ‘Programme’ tab) is intended to run for 12 weeks, for approximately 35 minutes per week. Informed by evidence-based small group social skills and speech and language therapy programmes, the CoSI teaches explicit reflection on the process of social conversation through direct teaching and structured, active learning, followed by unstructured practice with peers to promote generalisation with feedback from the teacher. The core skills, which the CoSI programme teaches are:
- active attentive listening;
- understanding that one’s conversational behaviour impacts the feelings of others;
- responding in a way that builds on the conversation topic as it gradually shifts;
- taking turns and allowing everyone to join in;
- understanding that everyone makes ‘chatting slip-ups’ and that these can be ‘fixed’;
- accepting and accommodating individual differences in social communication.
Objective 2: To test programme feasibility in classroom settings.
We tested whether it was feasible to deliver the programme in six schools (in Year 4) in Kent and Greater Manchester. We selected four schools in areas of high socio-economic deprivation (two of which had predominately multilingual children) to evaluate feasibility in these contexts. All schools were able to run the six units of the programme, albeit with some taking longer than planned. Most opted to run the programme during lessons for Personal Social Health and Economic / Relationships and Health Education.
Having delivered the CoSI, teachers reported that the direct teaching elements (interactive lessons with PowerPoints) and structured active learning (usually games) were easy to implement and enjoyable. All teachers said they would run these elements of the programme again, albeit with minor amendments. There was more variability in teacher confidence in providing feedback on peer-to-peer unstructured conversations – one of six teachers did not feel this was viable because her class had a high level of need and no teaching assistant support. The remaining five teachers all provided feedback to individual children / groups during peer-to-peer social conversation practice, although many expressed a preference for feedback in the form of whole-class plenaries at the end of these sessions. While some teachers expressed some concerns (e.g. resulting noise levels) about the peer-to-peer unstructured practice, it appears that in most schools it is feasible to deliver the direct teaching elements, structured active learning and (usually) the peer-to-peer unstructured practice.
Objective 3: To determine teacher views of the programme.
We first examined teacher views of the programme via interviews and a focus group with the group of six teachers who had run the programme in their classrooms. All stated that they would recommend the programme to colleagues (allowing for amendments). Five of the six said running the programme had increased their understanding both of social communication and of the social skills of their pupils. All said the programme should run for considerably longer than 12 weeks (to provide time to consolidate the concepts and practise skills), with most recommending it run over multiple primary school years. Several reported broader social impacts for their pupils including reduced peer conflict. Some suggested further linking of CoSI to the curriculum for Relationships and Health Education (RHE).
We then carried out three focus groups with different groups of teachers who had not run the programme but who accessed the programme materials and lesson plans via our website. These teachers found the CoSI materials well-designed and easy to use. They all felt that there is a need for a programme of this type due to poor oral language (which some highlighted as a significant hurdle for literacy) and social communication skills amongst many Key Stage 2 children. However, their focus group discussions indicated that further co-development work with teachers is necessary for certain elements of the current CoSI programme as well as for the corresponding teacher training materials.
All teachers who used or accessed the materials highlighted the importance of the skills taught in the programme for group work in curriculum subjects and for managing disagreements between peers. They also highlighted an increasing need over the past few years for a programme supporting social communication.
Objective 4: To determine the most appropriate means of assessing child conversation skills with a view to a future study comparing the programme to a control group.
To test whether a programme is effective in supporting conversation skills, it is necessary to have valid, statistically reliable and age-appropriate measures of conversational skills, particularly turn-taking and relevant responding. Test reliability was examined in two ways: inter-rater reliability and test-retest reliability. Inter-rater reliability examines whether two independent raters would obtain approximately the same scores for a given child. Test-retest reliability examines whether child rankings on test scores remain the same if tested at two timepoints.
To identify appropriate measures, we first ran a study with 72 Year 4 children exploring different measures, including a) assessing on-topic responding during a semi-structured conversation paradigm with an adult and b) assessing aspects of turn-taking, topic maintenance and question asking in an unstructured conversation with an adult. This identified that the semi-structured conversation measure was in the excellent range for both types of reliability. Six of the eight scales for scoring unstructured conversation were in the excellent range for inter-rater reliability, and all of these had good statistical test-retest reliability.
To explore the feasibility of using these measures in a future evaluation of the CoSI programme, we then used the semi-structured conversation measure and selected four of scales for measuring unstructured conversation. These complemented the structured measure on-topic responding by assessing turn-taking and question asking. We assessed 118 children before and after their teachers ran the CoSI programme in their classrooms. The structured conversation measure and two of the four scales for unstructured conversation (assessing turn-taking) showed statistically significant improvements following the CoSI programme. The change for the structured conversation measure showed a medium effect size[1]; relevant responding rose from 44% to an average of 64%. Separately, for two of the four unstructured conversation measures children showed statistically significant improvements following the CoSI. One of these scales assessed conversational ‘balance’ (use of speaking time relative to the partner), where a higher score indicates better performance. Here, the odds ratio for the effect was 1.97, which means that a child was twice as likely to score higher (for each boundary of the 5-point scale) at post-intervention than at baseline. The other scale assessed conversational reticence scale (where a lower score indicated better performance) and for this the odds ratio was -1.58 indicating that children were 50% more likely to have a lower score (at each boundary) post-intervention. While this is promising, because there was no comparison (control) group, causal inferences cannot be drawn. In addition, the semi-structured conversation measure currently requires researcher time to correct auto-transcription whereas the unstructured conversation measures require approximately a day’s reliability training prior to scoring. This means that while these measures could feasibility be used in a large-scale randomised controlled trial, it would be beneficial to explore ways to reduce researcher time as well as potential additional measures. Nonetheless, our summary is that these findings provide clear motivation for running a future study with a control group to robustly test the effect of the programme.
In addition, 87% of the children reported that the programme had helped them with conversation in real life. Many gave anecdotal examples to support their views, citing increased confidence in chatting, awareness of when and how to change conversation topics, the importance of listening to the conversation partner and allowing them to contribute, all skills which were taught in CoSI. A few children spontaneously gave examples of how the programme impacted their home lives, as in the following quote:
Child 23: “I used to not chat at all like all I did was chill in my room and then my mum said ‘Dinner’s ready’ and I didn’t even barely said anything – I just said ‘Ok’ [oh really?] but now I just talk to my mum” 09:41
Of the six teachers who ran the programme in their classrooms, five gave examples in support of their views that the programme had a positive impact on the children, for example:
“And if you say – if you gave them a stimulus and said talk about this, the number of them who would literally just sit still and stare straight ahead, not even look at the person they were meant to be talking to at the beginning of the year. It was quite staggering and we’ve come so far from that now. But because I think the “in” was making it a social programme and you can talk about your favourite restaurant or you – whether you want the cat or all the things that they wanted to talk about, that’s – now what I’m seeing is if I give them something to talk about in writing or talk about in reading, they’ve got all the skills to be able to look at each other, make eye contact, and be engaged. So it’s had a huge impact” School 5 teacher 5:18
Next steps
There are three areas where further development is needed:
- Further co-development with teachers
First, further co-development of a suite of options for running peer-to-peer unstructured social conversation practice is needed to ensure that all teachers can access materials and activities as appropriate to their classroom needs and that noise is minimalised. In addition, as with all aspects of the curriculum, it is essential that each child is aware of which aspects of conversation they might improve. Therefore, further co-development around providing feedback to children is also required. Connected to this, additional training films should be developed to illustrate running peer-to-peer conversation practice and feedback provision.
- Further development of quantitative child assessment measures
While the measures of child-to-researcher conversation were statistically reliable and age-appropriate, it would be useful to explore child conversation measures, which can be administered and scored with minimal training and without transcription-correction time. In addition, there are further measures which could provide greater insight into the impact of CoSI. For example, given child and teacher anecdotes regarding how the CoSI improved child confidence in oracy with their peers, quantitative measures of impact on peer interactions and / or peer relationships could provide a strong evidence base to justify use of CoSI. Finally, assessing children’s ability to listen attentively and engage in discussion would provide further evidence of the generalisability of the CoSI programme.
- An investigation of causality
Finally, we need a robust test of whether the CoSI programme is effective in promoting conversation skills by comparing an intervention to a comparison group.
Potential implications for policy and our longer-term goals
Both policy guidance and indicative evidence from the current study speak to the need to develop children’s conversations skills. Further, teachers believed there was a real need for a programme like the CoSI. One reason they gave was that strong oral language skills, and social conversation skills in particular, are essential for classroom learning. They argued that strong oral language is essential for literacy and conversation skills are a pre-requisite for engagement in group work as part of all curriculum subjects. This links to findings from the recent Curriculum and Assessment Review which presents evidence that currently “attention to oracy is insufficient” (p42) and highlights “the need for the English curriculum to make speaking and listening requirements more prominent” (p43).
The second key benefit concerned managing disagreements in the classroom and playground. Teachers’ perceptions across the board were that social communication skills are weaker in current cohorts than in the past. Half the teachers in our sample reported that the programme helped to ameliorate peer conflict. This echoes links between teaching ‘genuine listening as perspective-taking’ and conflict reduction found in work on communication for peacekeeping (discursive civility). Indeed, most teachers argued that social communication support is needed for all primary school years groups and that a good programme would go even further in helping children to understand the nature of friendships, the importance of turn-taking and the oral language skills required for negotiation and managing disagreements. This would fit well with the Curriculum and Assessment Review’s proposal that Citizenship should become statutory at primary level (p59) and should foster “oracy skills, such as expressing opinions, listening to others’ points of view, and agreeing and disagreeing respectfully” (p58).
Taken together, these findings make a compelling case for the further development of the CoSI programme as a timely response to both classroom realities and a future-ready national curriculum.
[1] Effect sizes are usually classed as either ‘small’, ‘medium’ or ‘large’. A small effect means the difference (e.g. between pre and post intervention) is fairly subtle in terms of how noticeable it is in daily life. A medium effect means that the average child showed a meaningful improvement, that is, enough that it would likely be felt in their day-to-day life. In studies with children, it is quite rare to find medium effect sizes due the great variability amongst children.