AbstractThe ability to formulate testable hypotheses' is now one of the assessment criteria for the recently established National Science Curriculum in England and Wales. This paper looks at this scientific process skill from two perspectives. Firstly, from a broad theoretical viewpoint, the nature and status of hypotheses in scientific reasoning are explored. Secondly, the responses of 11 year‐old schoolchildren to an instrument designed to assess this ability are discussed, with particular reference to the children's use of language. The final section of the paper draws out the most important implications for science education in this and other ‘process’ areas. The difficulties in judging, on the strength of written evidence alone, whether or not a child has demonstrated this skill are highlighted. In terms of pedagogy, the findings described in the paper strongly support the view that the formative aspects of assessment which can be fed back into teaching are most important, and that such evidence from teachers about the performance of individuals should exert a powerful moderating influence on the performance data generated by any external summative assessment s