The use of metadata as description of statistical datasets has been with us for some time, and the value of this approach is demonstrated by the success of initiatives for metadata standards, such as triple-s, SPSS Dimensions and DDI. The need to support discovery of resources and services over the internet has widened interest in and use of the term metadata, focussed by initiatives such as the Dublin Core, the (derived) UK Government Metadata Standard (e-GMS) and ISO-11179. These standards and structures are now being influenced by more formal ideas from IT systems design. The recent proposals from SDMX and (version 3 of) DDI are strongly influenced by object-oriented design principles, using UML as the specification medium. This extends the specification beyond the structure of the information represented to include the formal representation of the semantics and functionality that are needed in order to use the metadata. These proposals cover information about the data capture process (including sample and instrument design), about the structure and location of datasets (including coding), and the exchange of data (with metadata) in micro or macro form. The Opus Project has attempted to go further than this and represent the statistical models used in the analysis of data, and thus to provide a link back from conclusions and results based on statistical information (evidence) to the assumptions and data used.