There is currently great concern about levels of obesity in the UK, and its negative physical, psychological and social impacts. Current research suggests that there may be a relationship between obesity and poor educational attainment. It is likely that obesity and poor school performance are elements of a broader picture of inequalities in health and education, whereby disadvantaged socio-economic groups tend to have poorer health and lower levels of education. However, it is possible that other factors influence obesity and attainment, such as gender, discrimination and poor mental and emotional well-being. This systematic review was therefore commissioned to address the question: What do we know about the relationship between childhood obesity and educational attainment, from the research literature? We located studies through sensitive searches of a large number of databases, as well as specialist websites and contact with experts. Included studies were coded and quality-assessed by two reviewers independently. In order to ensure that our analysis was informed by the perspectives of teachers and young people, we held consultations in April and May 2009. While often conflicting, an overall pattern emerges from the research evidence suggesting that there is a weak negative association between obesity and educational attainment in children and young people; i.e. that higher weight is associated with lower educational attainment. Obesity is also associated with other variables, such as socio-economic status, and when these other variables are taken into consideration, the association between obesity and attainment becomes still weaker, and often loses statistical significance. Place of residence, ethnicity, occupation, gender, religion, education, socio-economic status (SES) and social capital were all explored as potential moderating variables in the included research. Twenty-three of 29 studies used a measure of socio-economic status as a moderating variable. Various factors appear to contribute to low educational attainment to some extent, although given the variation in definitions, analyses and quality of data, it is impossible to point to any causative or definite risk factors. Most studies explored the influence of obesity upon attainment. Only two studies examined the influence of attainment upon obesity. Many authors suggested multiple causal pathways, many of which remained untested in their studies. The moderating variables used in statistical analyses of the relationship between obesity and attainment were not consistent with the causal pathways proposed, which is probably a reflection of the constraints imposed upon authors conducting secondary analyses of pre-existing datasets (i.e. they made use of existing variables, rather than collecting their own, tailored data). The most frequently cited factors resulting from obesity and impacting upon educational attainment were poor mental health, stigmatisation and discrimination, disordered sleep, decreased time spent in physical activity and socialising, and absenteeism. First, that obesity should not be understood solely as a health issue. This review, and other research, suggests that one of the most noticeable ways in which obesity affects the lives of children and young people is in their social relationships. Given the paucity of evidence suggesting a causative physiological link between obesity and attainment, any association is likely to be mediated by social factors. We find that stigmatisation, bullying, low self-esteem and young people’s exclusion from opportunities for social interaction are suggested as underlying any relationship between obesity and lower educational attainment. Second, that the variables used in statistical analyses failed to capture many of the potential causative factors identified by the teachers and young people (and, often, also of the researchers undertaking those analyses). If large-scale longitudinal datasets are to deliver on their potential to help us understand people’s lives, they need to engage with the social lives of their participants and amass not simply data that are straightforward to collect, but information that reflects determining characteristics of people’s social experiences, because these are often the key to understanding health and other behaviours. Finally, we find this body of literature to be one of the least cumulative that we have reviewed. Data from the same datasets are analysed in different ways, using different variables, coming to different conclusions with minimal attempts to explain differences in findings. Different statistical models are sometimes employed with little justification for their selection and little acknowledgement that a different method might yield an entirely different result. These are important failings, and it is hoped that work associated with the EQUATOR network will in time lead to improvements (Simera et al. 2010). We should note that these criticisms do not apply to all studies in this review, but it would be true to say that they do apply to many of them.