How should problem-solving representations for complex knowledge domains be designed? Traditional approaches typically address the problem of semantic complexity by designing systems that offer multiple and often heterogonous forms of representation. The REEP approach advocates structure preserving integration of the different classes and perspectives of a domain within a single representation. This paper reports on a novel representational system for nurse rostering that was designed under the REEP approach. An empirical evaluation suggests the kinds of knowledge support provided by the representation and demonstrates that participants prefer fully integrated over selective views of information even though the former increases visual complexity. This knowledge support is explained in terms of more abstract domain independent cognitive benefits that we present as reasons for adopting the REEP approach.