One of the most important research questions in the economics of education is whether increasing the level of resourcing in schools will lead to improved student outcomes. Research on the education production function has been on-going since the 1960s but there is still considerable controversy in this largely US-dominated literature as to whether additional resources have a positive impact on students' attainment. Perhaps the most crucial methodological difficulty is that the level of resources per pupil often depends on the pupil's attainment. The dependency of resources on pupil attainment can arise for a number of reasons, most commonly funding mechanisms which allocate additional resources to schools with higher concentrations of socially disadvantaged pupils and pupils with learning difficulties. This results in an inverse correlation between resources per pupil and attainment. When this exists, a straightforward multiple regression of attainment on a set of variables, which include resources per pupil, is likely to find either that more resources are associated with lower attainment or that there is no statistically significant relationship at all. Outside of the US the research literature is quite sparse and has also been hampered by the lack of up-to-date, pupil-level data. This study examines the relationship between resources and pupil attainment in England at Key Stage 3 (national tests taken at ages 14-15). Our data are from the Pupil Level Annual Schools Census (PLASC). The dataset contains around 3000 secondary schools and over 430,000 pupils. It includes pupils' attainment at Key Stage 3 in 2003 and prior attainment at Key Stage 2 in 2000, pupil characteristics, such as gender, special educational needs (SEN), ethnicity, English as a first language, age and eligibility for free school meals. Data on pupils' post-codes were used to obtain additional census data on socio-economic indicators of the areas in which the pupils live. PLASC also contains variables describing the context of the school (size, proportion of pupils eligible for free school meals, with additional educational needs, boy or girl only school) and on school type; i.e. age range, selective, denominational, and particular school categories in receipt of additional funding. Data were available on three school- level resource variables: expenditure per pupil, the pupil teacher ratio and the ratio of pupils to non-teaching staff. The study attempts to overcome the endogeneity problem using instrumental variables. - lagged school size and the political control of the local authority. Both are related to resourcing levels, but should not impact on attainment. We also consider the impact of extra resources on pupils who differ by gender, ethnicity, SEN status, poverty and ability and in schools with different mixes of students. Our main findings are that there are indeed positive marginal resource effects on attainment but they are rather small and subject-specific. Resources appear to have a small but significant impact on pupil attainment in mathematics and science but little impact on attainment in English. The gain in attainment from additional resources appears to be greater for pupils from poorer home backgrounds. Also, there is some weak evidence that middle ability pupils benefit from additional spending more than pupils in the top or bottom ability quintiles. High ability pupils from low-income families also benefit more from higher resourcing, particularly in science. The findings suggest that rather than spreading additional resources evenly across all students, policy-makers should concentrate additional spending on projects targeted at specific student groups or curriculum areas. Extra resources would appear to be most effective if targeted at maths and science and at students who are of average ability or from poorer backgrounds, particularly if they have high ability.