This paper explores the scientific reasoning of 14 children across their first two years of primary school. Children's view of experimentation, their approach to exploration, and their negotiation of competing knowledge claims, are interpreted in terms of categories of epistemological reasoning. Children's epistemological reasoning is distinguished from their ability to control variables. While individual children differ substantially, they show a relatively steady growth in their reasoning, with some contextual variation. A number of these children are reasoning at a level well in advance of curriculum expectations, and it is argued that current recommended practice in primary science needs to be rethought. The data is used to explore the relationship between reasoning and knowledge, and to argue that the generation and exploration of ideas must be the key driver of scientific activity in the primary school.