The popularity of formative assessment has increased since the publication of work by Black and Wiliam in 1998. Even though it is a useful teaching tool, in most cases it has only been possible to use it for students with high levels of cognitive and communicative ability. The aim of this article is to propose a modified, personalisable model of formative assessment for non-verbal students with autism and severe learning difficulties. Five students with autism and severe learning difficulties participated in systematic video observations over a period of eight weeks, during which student behaviour and attainment were recorded. The behaviour checklist gave an accurate representation of the students' level of engagement and predicted attainment, but differences in passive learning objectives (those requiring passive co-operation) and active ones (those requiring active contribution) were noted. The introduction of engaging resources improved engagement, but tangible rewards had a negative effect on attainment. Praise had a positive effect on engagement and attainment.