Dominant discourses of quality assurance in UK higher education posit assessment as a transparent and rigorous process through the ascription of the two key and inter-related concepts of validity and reliability. Drawing on ethnographic research into the assessment procedures and practices of one teacher-training course in England (a part-time course for teachers in the learning and skills sector which is delivered on a franchise basis across a network of further education colleges), this thesis demonstrates that claims to assessment validity and reliability are contestable. The thesis draws on three complementary social practice theories (communities of practice, new literacy studies and actor-network theory) in order to reveal assessment as being a complex, localised practice characterised by contingency and improvisatory behaviours on the part of both tutors and students, mediated by a variety of genres of textual artefacts. These divergent and complex practices are shown to disrupt dominant managerialist discourses of assessment practice in higher education. They are also shown to disrupt dominant definitions of learning, teaching and assessment in higher education, which predominantly rest on models of individual cognition and transferable skills, and which this thesis critiques through the use of social practice accounts of learning within communities of practice. The thesis demonstrates that assessment is more contingent and complex than dominant discourses of assessment practice in HE allow, thereby problematising claims to reliability and validity. The thesis makes contributions to current literature and research in two ways. Firstly, it concludes by offering a series of suggestions as to how assessment validity and reliability might be enhanced or reframed. Secondly, it demonstrates how communities of practice theory can be used critically to explore pedagogic activity, including assessment, within formal educational settings.