In this article, we describe a software system for assessment development in online learning environments in contexts where there are robust links to cognitive modelling including domain and student modelling. BEAR Assessment System Software (BASS) establishes both a theoretical basis for the domain modelling logic, and offers tools for delivery, scoring, reporting, and use of learning evidence. BASS employs the UC Berkeley BEAR Assessment System, which is a four-part approach to modelling that involves a combined domain and student model, in this case for formative learning settings. In the article, we first briefly present an expanded view of learning environments useful for intelligent learning environments, then present background to the development of BASS, elaborate on its approach to construct mapping of the domain and student models, and report findings from studies based on research questions about its use in example projects.