This paper focuses on formative assessment in the field of higher education. It examines Bernstein's work on vertical discourses and knowledge structures with the view to deepening understanding of the concept of assessment "for" learning. The first part of the paper draws on Vygotsky's work on concept development and Bernstein's work on knowledge structures to explain why "generalisation" and "hierarchy" are central in knowledge acquisition. It then explores Bernstein's claim that, within the vertical discourse, different knowledge structures (hierarchical and horizontal) afford greater or lesser visibility of their epistemic structure, and thus of their evaluation criteria of what counts as a legitimate text. The second part of the paper investigates the ways epistemic expectations are signalled through the practice of evaluation to first-year university students in a professional education course and proposes that markers do not offer students stuffiest access to recognition rules necessary for producing legitimate texts in the future. Drawing on Maton's distinction between semantic gravity and semantic density, the paper offers an example of how markers could recast what is present in students' work to offer students access to key ordering principles in vertical discourses. (Contains 2 tables and 18 notes.)