This paper explores the idea that some Discourse Particles are so culturally embedded that they defy translation. Natural Semantic Metalanguage and Conversation Analysis are used to examine the meanings and functions of (a) the polysemous Arabic Discourse Particle yacni (derived from yacni 'it means'), translateable with 'well,' 'I mean,' 'that it,' 'you see,' 'like,' and 'so,' and ( b) "sorta" and "I mean," the main English translation candidates for yacni. The findings show that yacni's focus on marking relevance is useful in elaborating, correcting, creating narrative suspense, holding a turn, or, as sole constituent of a turn, hedging a response. Similar English functions are achieved using Discourse Particles that focus approximation. The analysis also shows that semantic or pragmatic similarity in Discourse Particles from different languages can predict translation potential. In the same way, similarity in the meaning of a Discourse Particle and of a speech act predicts translateability. However, despite semantic and pragmatic equivalence (of a sort), culture-specific indices may mean that a word is not a suitable translation candidate. The results augment our understanding of cultural semantics and ethnopragmatics, and have applications to the study of translation and intercultural communication.