Cases as used in case-based reasoning (CBR) typically record experts' steps of reasoning and action, but not the arguments an expert may consider during the problem-solving. Knowledge that can improve the quality of performance of CBR is therefore lost. The paper describes an approach that tackles this problem by representing arguments in a simple form and treating them, along with the traditional information contained in cases, as case properties. These properties are processed according to methods of numerical taxonomy when similarities between cases are being computed. The cases themselves are structured according to a model (CommonKADS) familiar in knowledge engineering but seldom applied to CBR and argumentation.