The success of an integrated system is highly dependent on the successful development of a global data model. Any integrated system, a data warehouse, a mediator, or a composite system, must be able to handle multiple interpretations of a concept's symbols, extensions, and intensions. This paper discusses the under-pinning subjective nature of data integration and illustrates the resulting semantic problems that occur. Annotations are introduced to extend the current models' ability to represent the semantic nuances of the underlying databases