In this paper we present an investigation into the run-time behaviour of objects in Java programs, using specially adapted coupling metrics. We identify objects from the same class that exhibit non-uniform coupling behaviour when measured dynamically. We define a number of object level run-time metrics, based on the static Chidamber and Kemerer coupling between objects (CBO) measure. These new metrics seek to quantify coupling at different layers of granularity, that is at classclass and object-class level. We outline our method of collecting such metrics and present a study of the programs from the JOlden benchmark suite as an example of their use. A number of statistical techniques, principally agglomerative hierarchical clustering analysis, are used to facilitate the identification of such objects.