It is widely accepted that pauses and correspondence with intonation units are among the phonological cues that identify lexical chunks (conventionalised multiword sequences) in spoken first and second language, although conclusive empirical evidence is scant. This article reports on an exploratory study which seeks to identify the main phonological markers of lexical chunks. A corpus of five minutes of speech from four English learners of Spanish was collected and transcribed orthographically. The data-set contained 97 chunks, between 11 and 30 chunks per speaker. The lexical chunks were extracted using speech analysis software and analysed phonologically in respect of pauses, intonational boundaries and sentence position. The analyses revealed that only about half of the chunks in the data were delimited by intonational boundaries. Certain types of lexical chunks, such as discourse markers, consistently coincided with intonational boundaries, especially in sentence initial position. The most proficient speakers used more discourse markers, separated by intonational boundaries, which enhanced the perceived fluency of their utterances.