Special-language term formation is characterized, inter alia, by the frequent reuse of certain lexical items in the formation of new syntagmatic units and by conceptually motivated restrictions on the position which certain elements can occupy within a compound term. This paper describes how the positional and combinational features of the terminology of a given domain can be identified from relevant existing term lists and used as part of a corpus-based, automatic term-identification strategy within a natural language-processing (e.g. machine translation) system. The methodology described is exemplified and supported with data from the field of satellite communications.