This study examined the use of acoustic-phonetic cues to syllable boundaries in on-line lexical segmentation. Two word-spotting experiments manipulated the alignment between word and surface syllable boundaries. A misalignment cost was observed for obstruent-liquid clusters but not for /s/ + obstruent clusters. Durational analyses of the materials as well as of a large corpus of utterances showed systematic variations in segment durations as a function of boundary location in obstruent-liquid clusters only. We conclude that the availability of acoustic-phonetic cues to word boundary location in consonant clusters depends upon the cluster class. When available, these cues are exploited in on-line lexical segmentation of speech.