I investigate the parsing strategies of relative clause (RC) attachment preferences by Korean speakers and English learners of Korean as a second language. RC can modify either NP1 or NP2 as its host when the antecedent consists of a complex NP as in `kyosil-ey iss-nun haksayng-uy chayk`, where either haksayng or chayk can be interpreted as a host. Two test methodology (a picture identification test and a written questionnaire test) were used to examine the parsing preferences by the two language groups. The results show that Korean speakers and English speakers converged or diverged depending on the test methodology. While the two language groups behaved similarly in the former test, they did not in the latter test. A strict working memory capacity and lack of knowledge about the verb semantics by L2 learners were assumed as possible reasons for the different parsing preferences. Based on the findings, I propose that native speakers and L2 learners do not have the same source of computational system and that different types of tests are necessary to get a clearer picture of L2 processing system in psycholinguistic research.