Generic relationships are abstraction patterns used for structuring information across application domains. They play a central role in information modeling. However, the state of the art of handling generic relationships leaves open a number of problems, like differences in the definition of some generic relationships in various data models and differences in the importance given to some generic relationships, considered as first-class constructs in some models and as special cases of other relationships in other models. To address those problems, we define a list of dimensions to characterize the semantics of generic relationships in a clear and systematic way. The list aims to offer a uniform and comprehensive analysis grid for generic relationships, drawn from a careful analysis of commonalities and differences among the generic relationships discussed in the literature. The usefulness of those dimensions is illustrated by reviewing significant generic relationships, namely, materialization, role, aggregation, grouping, and ownership. Based on those dimensions, a new metamodel for relationships is proposed.