Since 2010, the whole national cohort Key Stage 2 (KS2) National Curriculum test in science in England has been replaced with a sampling test taken by pupils at the age of 11 from a nationally representative sample of schools annually. The study reported in this paper compares the performance of different subgroups of the samples (classified by gender, ethnicity, English as an additional language, eligibility for free school meals and special education needs) in the test over the 2010 and 2011 test series. To study measurement invariance of the test across the different subgroups, factor analysis and Rasch modelling have been used to analyse item level-data for differential item functioning and differential test functioning in order to investigate the functioning of the items in the tests and the overall tests across the subgroups. A range of measurement properties related to the science construct being measured by the tests, including dimensionality, item difficulty, expected scores on individual items and expected total test scores, have been examined. It is concluded that the KS2 science sampling tests function similarly across the different subgroups at both item level and the overall test level and provide a measure of pupils' attainment in science on the same measurement scale for all subgroups studied over the past two years.