Terminology variation in texts is now a well-known phenomenon, whose amount is estimated to range from 15% to 35%, depending on the languages and on the characteristics of the textual documents such as the domain, the genre, the discourse. The kinds of variants that are handled automatically differ depending on the application. In the field of terminology acquisition from texts in a multilingual environment, we will adopt a definition of a term variant that relies on the existence of an original term on which an elementary string operation applies (substitution, expansion, reduction). With regards to the linguistic nature of the source term, i.e. single word, morphological or syntagmatic compound, and the kind of elementary operation, we will describe the linguistic procedures that generate such variants. For some variants, we will propose methods that could be used to detect them as well as the necessary lexical resources. We will provide examples of term variants in the languages studied in the TTC European project. We also will discuss the complex nature of variants by examining a few problems : to distinguish between a term and a variant, to state about the synonymy or non-synonymy nature of the variant with regards to the original term, the language specific issue, as well as a few ambiguities linked to the typology of variant generation procedures.