Salience is regarded as one of the key concepts for cognitive studies of language and communication, however there is limited research on how prominence plays out in multimodal discourse. The present study is aimed at investigating how salience comes through in gestures used by simultaneous interpreters.  To distinguish between salient and non-salient gestures by the same participant, two groups of observable parameters were chosen – basic and auxiliary. An empirical study was carried out, based on simultaneous interpreting of the audio of a TED talk (English ®  Russian). The video recordings were integrated into ELAN files and annotated for salient gestures, the functions that were realized by them, and the elementary discourse units (EDUs) that the gestures co-occurred with. It was assumed that, first, salient gestures will be observed less frequently than non-salient gestures; second, prominence in gestures will serve the function of representing various aspects of a situation more often than other functions; third, salient gestures will co-occur more often with elementary discourse units (EDU) containing verb phrases, rather than noun phrases. The hypotheses were partially confirmed via quantitative and qualitative analyses which demonstrated that  every third gesture was salient; the representative function came second after the pragmatic functions; there was no significant difference between the number of gestures used with verbal and nominal EDUs, though it was observed that salient gestures tend to co-occur with the verbs of physical actions and negation, as well as with the nouns accompanied by attributes denoting high degree of a quality.