Different file formats, access protocols, query languages etc.
Often called syntactic heterogeneity from the point of view of data.
Table decompositions may vary, column names (data labels) may be different (but have the same semantics), data encoding schemes may vary (i.e., should a measurement scale be explicitly included in a field or should it be implied elsewhere).
Perhaps a database system must be able to integrate genomic and proteomic data.
Relationships may be inferred between data according to rules specified in domain ontologies.