The principal aim of this project was to develop an assessment of metapragmatic ability to use with school-age children who have language impairments, thereby providing a means of sampling and coding a child’s ability to identify and reflect upon communication acts. This would be of clinical relevance to children who have difficulty in understanding and implementing the rules of language use (pragmatics) and to a broader group of children who have communication impairments including those on the autism spectrum. All of these children at some stage require support from specialist speech and language therapy services, yet practitioners have no assessments available to profile children’s metapragmatic ability. A clinically valid and reliable task which provides normative data on the development of metapragmatic ability in typically developing children and children who have language impairments is therefore likely to be welcomed by user groups such as speech and language practitioners and specialist teachers. The study of metapragmatics also has potential to add to the theoretical debate around the nature of developmental language impairments, especially those involving pragmatics. Children who have pragmatic language impairment or CwPLI have long-term difficulties with discourse, social understanding, narrative and conversation and therefore might be expected to perform poorly on a task of metapragmatics. It is accepted that CwPLI share some difficulties in social cognition or ‘theory of mind’ with children who have autism, but they also have some of the problems of grammar and vocabulary organisation seen in children who have ‘pure’ language disorders or specific language impairment (CwSLI). In this study we hypothesised that CwPLI would perform more poorly than CwSLI on the AMPs task because of the former’s social cognitive limitations and that both groups would be out-performed by the children with typical language development. Factors which could predict metapragmatic awareness, such as language ability, non-verbal performance, parent ratings of autism and social cognition were measured in children with PLI and SLI. It was hypothesised that language would be a predictor across groups but that autism ratings and social cognition measures would also predict poor AMPs performance in the PLI group.In the first stage of this research an experimental task, the Assessment of Metapragmatics or AMPs, was developed as a child-friendly procedure in which the participant watches scenarios performed by child actors and scripted by the researchers. Each scenario contained a deliberate pragmatic ‘error’, such as ‘talking too much’, ‘over-exaggerating’, which the participant was asked to identify and reflect on using a list of questions asked by the researcher. After a substantial revision and piloting process with both child participants and potential end users, a final version of AMPs, incorporating 13 scenarios and 8 questions was drawn up. A scoring guide was developed by detailed examination of the pilot data.42 children with typical language development aged between 6;0 years and 11;11 years then participated in the development of AMPs normative data. Participants were grouped into three age bands for the purpose of analysis: 6;0-7;11, 8;0-9;11, 10;0-11;11. Statistical analysis showed that there was a difference in scores amongst the first two age bands and the first and third age bands but not between the 8-9 and 10-11 year old groups. Further statistical analysis revealed that for 9 videos there was a clear developmental progression with age, but that 3 videos had floor or ceiling effects.Normative data for each age group were then drawn up. A reliability analysis on 20% of the video coding established inter-rater reliability of 81% agreement. Furthermore, analysis of the 13 video items in AMPs revealed good split-half reliability; all items correlated well with total AMPs scores.40 CwPLI and 15 CwSLI participated in the experimental phase. They all performed more poorly than CwTLD on the AMPs task by some considerable margin. CwPLI tended to be better than CwSLI on the AMPs tasks but this was partly explained by the fact that CwPLI tended to come into the study with better overall language skills. By controlling for language levels of children (and their ages) in statistical analysis, it was clear that CwPLI and CwSLI were generally performing at similar levels on AMPs, except for the part of AMPs which taps into social understanding where PLI children perform more poorly than expected given their language skills. This is a finding which could be pursued in future research.Factors which best predict AMPs scores in CwPLI and CwSLI were identified as language test scores and performance on a verbal test of theory of mind/inference (the Happe Strange Stories). Performance on AMPs was not predicted by autism ratings or by non-verbal IQ in either the PLI or SLI group. Future analysis of narrative data and other social cognition tasks may help to clarify the role of moderators of AMPs performance.In the final phase of this research, 18 of the 40 CwPLI, who were already participating in an intervention study aimed at improving social communication, were tested on AMPs before and after intervention. Outcome measures confirming intervention change will not be released from the companion study until end July 2009. At that stage an exploratory analysis of AMPs as a potential mediator of intervention will be carried out. ConclusionsThere is a clear developmental shift in metapragmatic performance between 6/7 and 8/9 age groups, with some plateauing after this point. Children who have typical language development show much stronger ability than children with language impairments on the AMPs task, suggesting that this assessment may be an appropriate high-level procedure for more able school age children with communication impairments. There is no difference between CwPLI and CwSLI on a task of metapragmatic ability when language status is controlled for, suggesting that there is no special characteristic of PLI which makes these children vulnerable to metapragmatic difficulties. Language performance was the strongest predictor of AMPs scores in PLI and SLI groups, but parental autism ratings did not predict AMPs performance, neither did non-verbal performance. These findings tend to confirm recent reports of overlap in communication abilities between children who have pragmatic or specific language impairments.