This paper looks at the role of translation in metadata. Metadata will help potential users locate, understand and repurpose materials. For digital language materials there are typically a wide range of potential user communities with a wide range of metadata requirements. If there is an uneven distribution of languages among the different user communities, which is likely, then there will also be a wide range of corresponding languages in which the metadata is ideally expressed.
Typically, metadata standards define all metadata names in English and also restrict controlled vocabularies to English. Some standards allow metadata to have values in languages other than English, but do not allow this to be represented as a translation. The newly opened Endangered Languages Archive (ELAR) allows users to add metadata in the language of their choice, translate existing metadata into a language of their choice, and to select language preference(s) for viewing existing metadata. ELAR is the first language archive to implement such flexibility in the translation and use of metadata, raising some important questions: how closely should translations of metadata relate to the original when they are intended for different user communities? Should the content of translations of metadata be moderated, and to what extent? If metadata has been entered in an endangered language, should it always be translated into a single common language?