This study investigates the role of universal markedness based on sonority distance and L2 phonotactic knowledge in the perception of onset consonant clusters. Ten native English listeners (NE), ten native Korean listeners with higher English proficiency (NKH), and ten native Korean listeners with lower English proficiency (NKL) participated in a syllable count task. The participants listened to four types of onset clusters based on sonority distance in monosyllables (i.e., large sonority rises, small rises, plateaus and falls) and matched disyllabic counterparts with vowel epenthesis. The results showed that for response accuracy, the NE and NKH groups showed more sensitivity to sonority distance than the NKL group. In monosyllables, both the NE and NKH groups were more accurate as the sonority distance increased, but they did not distinguish between sonority plateaus and falls. For disyllables, all the listeners differentiated consonant clusters with a large sonority rise from the other types of clusters. For response time, the NE group was faster than the two Korean groups. Sonority-based markedness was not able to explain all the results of the present study. Different misperception patterns shown in the three listener groups are partly due to linguistic experience in English. Universal markedness could be restricted by L2 phonotactic knowledge in Korean listeners’ interlanguage. (Cyber Hankuk University of Foreign Studies)