<jats:title>Abstract</jats:title>
<jats:p>This paper contributes to the question of how discourse relations are realised in <jats:sc>TED</jats:sc> talks. Drawing on an annotated, multilingual discourse corpus of <jats:sc>TED</jats:sc> talk transcripts, we examine discourse relations in English and Lithuanian, Portuguese and Turkish translations by concentrating on three aspects: the degree of explicitness in discourse relations, the extent to which explicit and implicit relations are encoded inter- or intra-sententially, and whether top-level discourse relation senses employed in English differ in the target languages. The study shows that while the target languages differ from English in the first two dimensions, they do not display considerable differences in the third dimension. The paper thus reveals variations in the realisation of discourse relations in translated transcripts of a spoken genre in three languages and offers some methodological insights for dealing with the issues surrounding discourse relations.</jats:p>