We investigated the possibility of whether there is a post-vocalic ‘r’ in phonological representations of Korean L2 English speakers and the extent to which they exploit their knowledge of mapping graphemes onto phonemes within each correspondence between orthography and phonemes. First, the results obtained in the pseudohomophone task showed that R-items were responded to with higher accuracy and shorter RT than Non-R items. It suggests that there is no post-vocalic ‘r’ in phonological representations of Korean L2 speakers unlike Australian native English speakers and that Korean L2 learners are truly non-rhotic speakers. Another striking finding is that accuracy and RT for visual lexical access varied depending on the transparency between orthography and its corresponding phonemes. This indicates that Korean L2 speakers’ knowledge about the association of graphemes and phonemes varies depending on each type of correspondence between spellings and phonemes. Finally, it was found that the frequency of the base words also affected the retrieval of words along with the orthographic depth in grapheme-to-phoneme correspondences.