This paper challenges the familiar hierarchical partitioning of data fusion problems into 'levels'. The JDL data fusion model and its variants are seen as a method to partition a problem space in a way that tends to support different types of solutions. The layered view of fusion presented in these models is a rough engineering-based representation of a domain that has been addressed in analytically- and empirically-based models developed over centuries by philosophers and cognitive scientists. These ontological and cognitive models involve distinctions that are not all necessarily hierarchical or sequential. A hierarchical partitioning - while often convenient in characterizing fusion problems - should not be an impediment to fusion solutions that span the levels. A more flexible and comprehensive partitioning scheme is suggested
