This paper discusses some difficulties Korean speakers experience when they pronounce medial consonant clusters of English. We argue that such pronunciation problems stem from systematic differences between the two languages: especially, realization difficulties are induced when phonological requirements of the Korean language influence Korean speakers pronouncing English medial consonant clusters. In Korean, sonority between consonants over a syllable boundary is either falling or equal. If this requirement is violated, a certain repair strategy is adopted to modify such unallowable sonority relationships between consonants. On the contrary, English does not observe such sonority requirement for consonants over a syllable boundary, which allows to license various types of medial consonant clusters. Among the various combinations of medial clusters, clusters with rising sonority tend to spark problems for Korean speakers because such clusters are not familiar to Korean speakers, as well as are not well-formed from the perspective of Korean phonology. Another problem is that Korean does not allow medial clusters such as [l.n] and [n.l] in the output while they are well-constituted medial clusters in English. The /ln/ and /nl/ sequences of English are also very difficult for Korean speakers since Korean speakers are prone to pronounce them [l.l] and [l.l/n.n], respectively. For these pronunciation problems viewed from OT framework, we propose that highest-ranking constraints such as *nl/*ln, SyllCon, and a medially ranked Id-Lat(pv) in Korean are critical in the analysis. The pronunciation problems for Korean speakers are attributed to the ranking difference of those constraints in English. That is, the undominated constraints in Korean are ranked very low in English. Thus, understanding such constraint ranking difference between the two languages is important to promote the learning of the correct pronunciation of English medial consonant clusters for Korean speakers.