Authentic experiences of research are seen as valuable elements of undergraduate science, providing motivation for students and linking the research and teaching activities of academics. But as such experiences are made available to increasing numbers of students as formal, graded parts of the curriculum, important questions are raised about their pedagogical function and the ways in which they are assessed. This article draws on interviews with academics involved in the provision of such experiences to ask: what do academics "intend" that their students learn, and do conventional approaches to project assessment relate clearly and effectively to these intended outcomes? We describe four categories of intended learning and suggest that conventional approaches to assessment are fundamentally misaligned with most of these outcomes. We argue that this is due to the focus of these approaches on the products, rather than the processes and experiences, of research, a focus that partly arises from a sense of discomfort with assessment based on context-dependent judgements informed by subconscious expertise. We further suggest that alternative approaches to assessment could build on academics' own descriptions of the experiences and behaviours they value in students.