This study examined two possible sources of asymmetrical lexical access: phonetic proximity to the nearest L1 category and orthographic information. Three groups of native Korean speakers learned Arabic non-words with sound pairs with/without an L1-dominant category (/l-r/ vs. /?-h/), and then their phonetic categorization and lexical encoding abilities were evaluated. One group was presented with the same letters for the target pair (e.g., <l> for both /l/ and /r/), the second group, different letters (e.g., <l> for /l/, <r> for /r/), and the third group, auditory input only. The results of discrimination did not show any effect of these two sources, whereas in lexical encoding, (1) a pair with an L1-dominant category was more accurately encoded; and (2) orthographic information hindered the lexical encoding. In the following spelling recall task, the scores from the learners with different letters for the target pair were similar to a ceiling. Thus, orthographic information might help them to have target-like representation, despite difficulties in online processing.