The adoption of streaming in the primary school (where children are placed in a class on the basis of measures of attainment and remain in that class all of the time) was commonplace when the 11 plus examination was used to select children for grammar school places. During the 1950s and 1960s the practice died out with most children being taught in mixed ability classes with some in-class grouping. During the 1990s successive governments indicated that some form of ability grouping should be introduced in primary schools, setting (children placed in ability groups for some subjects and taught in mixed groups for the remainder of the time) being preferred, however, streaming was introduced in some schools despite evidence that movement between structured ability groups is infrequent and that children tend to remain in the same groups throughout their school careers limiting their educational opportunities. Recent research based on 8875 children, in the Millennium Cohort Study showed that 16.4% of children in Year 2 were in streamed classes. Logistic regression analysis showed that the best predictors of being in the top stream were whether the child was born in autumn or winter 2000, parents owning their own home, and the child's cognitive ability score. The measures predicting being in the bottom stream were being a boy, being born in the spring/summer of 2001, having a behaviour problem, being born into a lone parent family, and cognitive ability score. (Contains 7 tables, 10 figures, and 7 notes.)