Defined most generally, a translation memory is a computerised archive of existing translations, structured in such a way as to promote translation re-use. In this paper, two types of translation memory are contrasted -an interactive bilingual concordancer, like the RALI's TransSearch system, and a full-sentence repetitions processor, like Trados' Translator's Workbench- bringing out the strengths and weaknesses of each. The article conclude by examining some of the challenges that will have to be met in order to produce more powerful translation memory technology, systems capable of exploiting a larger portion of the knowledge lying dormant in translators' past production.